Browse the corpus

The mucus barrier is an essential part of the innate immune system which hydrates and protects the underlying epithelia. The gel-like properties of the barrier are mainly due to the polymeric mucins that are the main secretory products of epithelial goblet cells.1–3 The colonic epithelium expresses mainly MUC2/Muc2 in large amounts which is stored in bulky apical granules of the goblet cells and is the most important factor determining the goblet cell morphology.4–6 Muc2 forms a heterogeneous mucus barrier that is proposed to contain 2 distinct layers; a “loose” outer layer that bacteria can penetrate and an adherent inner layer that excludes bacteria from direct contact with the underlying epithelia.7 Alterations or absence of MUC2 production can lead to many common human disorders such as colon carcinoma,8 ulcerative colitis,9 and celiac disease.10 A role for Muc2 in the suppression of colorectal carcinoma has also been suggested because Muc2 knockout (KO) mice spontaneously develop colitis and adenomas that progress to invasive adenocarcinoma,11 suggesting an important function for this mucin in colonic protection.6 Furthermore, missense mutations in the Muc2 gene results in aberrant Muc2 oligomerization, leading to endoplasmic reticulum stress and subsequently increased susceptibility to colitis.12

itis and adenomas that progress to invasive adenocarcinoma,11 suggesting an important function for this mucin in colonic protection.6 Furthermore, missense mutations in the Muc2 gene results in aberrant Muc2 oligomerization, leading to endoplasmic reticulum stress and subsequently increased susceptibility to colitis.12 Hyperplasia of mucin-producing goblet cells has been described in a number of parasitic infections, including Nippostrongylus brasiliensis, Hymenolepis diminuta, Trichinella spiralis, and Trichuris muris.13–17 Putative mechanisms underlying the protective role of mucins against infectious agents include the demonstration of trapping of Hymenolepis diminuta17 and Trichinella spiralis18 in the mucus and inhibition of parasite motility and feeding capacity.18–20 Goblet cell response, in all 4 of these nematode models, is thought to be under the control of a T helper (TH) 2-type immune response and is considered as a potential effector mechanism.21–23 A number of goblet cell bioactive factors such as resistin-like molecule-β (Relm-β), intelectin, and calcium-activated chloride channel-3 have been suggested to play an important role in nematode infection.24,25 However, a definitive and precise role of mucins, the main secreted product of goblet cells, in host defense in intestinal nematode infection remains to be elucidated.

like molecule-β (Relm-β), intelectin, and calcium-activated chloride channel-3 have been suggested to play an important role in nematode infection.24,25 However, a definitive and precise role of mucins, the main secreted product of goblet cells, in host defense in intestinal nematode infection remains to be elucidated. The nematode T muris inhabits the cecum of mice and is closely related at the morphologic, physiologic, and antigenic levels to Trichuris trichuria, the causative agent of chronic trichuriasis in human beings.26 In this parasitic infection, strains resistant to chronic infection (BALB/c, C57BL/6) expel the parasites through the generation of a TH2-type immune response, whereas susceptible strains (AKR), which do not expel the worms, develop a TH1-type immune response.22,27 In this study, we demonstrated that the increase in Muc2, the main determinant of mucus barrier properties, correlates with worm expulsion. In the absence of Muc2 there is a delay in worm expulsion, but interestingly Muc5ac is up-regulated in the Muc2-deficient mice before expulsion. Moreover, Muc5ac is up-regulated in the wild-type (WT) mice that are resistant to infection, but not in those unable to expel. The physical properties of the mucus barrier are also altered during infection, resulting in a less-porous network, with overall changes having a direct effect on the viability of the whipworm. Collectively, these data show for the first time a protective role for mucins in nematode infection.

in those unable to expel. The physical properties of the mucus barrier are also altered during infection, resulting in a less-porous network, with overall changes having a direct effect on the viability of the whipworm. Collectively, these data show for the first time a protective role for mucins in nematode infection. Materials and Methods Animals Breeding pairs of Muc2-KO mice originally produced by gene mutation11 and their WT (C57BL/6) littermates (Albert Einstein Medical College, New York, NY) were kept at the animal facilities of McMaster University (Hamilton, ON, Canada). AKR, BALB/c (Harlan, UK), and severe combined immunodeficient (SCID) mice were maintained in the Biological Services Unit at Manchester University. The protocols used were in accordance with guidelines by the McMaster University Animal Care Committee, Canadian Council on the Use of Laboratory Animals, and the Home Office Scientific Procedures Act (1986). All mice were kept in sterilized, filter-topped cages, and fed autoclaved food in the animal facilities. Only 6- to 10-week-old male mice were used. Parasitologic Techniques The techniques used for T muris maintenance and infection were described previously.28 Mice were orally infected with approximately 100–300 eggs for a high-dose infection and <15 eggs for a low-dose infection. Worm burdens were assessed by counting the number of worms present in the cecum as described previously.28

ues The techniques used for T muris maintenance and infection were described previously.28 Mice were orally infected with approximately 100–300 eggs for a high-dose infection and <15 eggs for a low-dose infection. Worm burdens were assessed by counting the number of worms present in the cecum as described previously.28 Histology, Immunohistochemistry, and Immunofluorescence A 1-cm segment or the whole cecum (rolled) was fixed in 10% neutral buffered formalin or 95% ethanol and processed with the use of standard histologic techniques. Sections were treated with 0.1 mol/L KOH for 30 minutes before staining with periodic acid Schiff (PAS) reaction.29 Slides were counterstained with either H&E or 1% fast-green. Standard immunohistochemical and immunofluorescent staining methods29,30 were used to determine the levels of Muc2, Muc5ac, Relm-β, and trefoil factor 3 (Tff3). Antibodies Immunodetection was carried out with the use of a polyclonal antibody raised against a murine Muc2 (mMuc2).12 Commercially available 45M1 antibody was used for the detection of mouse Muc5ac.31 The mouse Muc5b-specific antibody32 was a kind gift from Dr Camille Ehre (University of North Carolina, Chapel Hill). Commercially available mRelm-β (Abcam, Cambridge, UK) and mITF (Santa Cruz Biotechnology Inc, Santa Cruz, CA) antibodies were used to detect Relm-β and Tff3, respectively. Detection of bromodeoxyuridine (BrdU) incorporated into nuclei was carried out with the use of a monoclonal anti-BrdU antibody (AbD Serotec, Oxford, UK).33

ially available mRelm-β (Abcam, Cambridge, UK) and mITF (Santa Cruz Biotechnology Inc, Santa Cruz, CA) antibodies were used to detect Relm-β and Tff3, respectively. Detection of bromodeoxyuridine (BrdU) incorporated into nuclei was carried out with the use of a monoclonal anti-BrdU antibody (AbD Serotec, Oxford, UK).33 Mucus Extraction and Agarose Gel Electrophoresis The cecum was gently flushed with phosphate-buffered saline and scraped, and mucus was solubilized in 8 mol/L guanidium chloride. Subsequently, extracted mucus samples were reduced with 50 mmol/L dithiothreitol and carboxylmethylated with 0.125 mol/L iodoacetamide before electrophoresis on a 1% (wt/vol) agarose gel. Mucins were detected after Western blotting with mucin-specific antisera.34 Analysis of Mucus Network Properties Cecal tissue isolated from BALB/c and AKR mice was cut longitudinally, washed with phosphate-buffered saline, and kept hydrated in a 6-well plate. Blue fluorescently labeled polymer microspheres (0.1 μm; Dukes Scientific, Dorchester, United Kingdom) were placed on top of the luminal surface of the cecum (set as a reference) and their position was analyzed with the use of the Nikon (Melville, NY) C1 Upright confocal microscope. Three-dimensional optical stacks were taken every 5 μm and combined to obtain a z-stack at the time points stated.

chester, United Kingdom) were placed on top of the luminal surface of the cecum (set as a reference) and their position was analyzed with the use of the Nikon (Melville, NY) C1 Upright confocal microscope. Three-dimensional optical stacks were taken every 5 μm and combined to obtain a z-stack at the time points stated. Energy Status of Worms The CellTiter-Glo luminescent cell viability assay (Promega, Madison, WI) was carried out according to manufacturer's instructions. Relative light units were calculated per worm as follows: relative light unit = (sample light units − blank light units)/number of worms. Substrate only was used as a blank control, whereas worms were boiled before homogenization for negative controls. To determine recovery of energy status, worms recovered were washed extensively in Dulbecco's modified Eagle's medium, added to 6-well plates with LS174T cells (maintained as previously described by Hayes et al35) for 24 hours before measuring adenosine triphosphate (ATP) levels. Statistical Analysis All results are expressed as the mean ± standard error of the mean. Statistical analysis was performed with the use of SPSS Version 16.0 (SPSS Inc, Chicago, IL). Statistical significance of different groups was assessed with parametric tests (one-way analysis of variance with post test after statistical standards or paired Student t test). P < .05 was considered statistically significant.

60–2000 and showed a similar reduction in mortality.27 However, these control groups were from different geographical populations with different study exclusion criteria. Comparisons were therefore susceptible to selection bias. Other studies of trends in variceal hemorrhage mortality contained less than 1000 patients. The other finding of note in our study in relation to variceal hemorrhage is the small proportion of overall hemorrhages that they represent. In the context of the increasing burden of liver disease28 and an apparent increase in variceal hemorrhage in the recent BSG audit,8 a higher proportion might have been expected. Our finding, however, was similar to that from the 1993 BSG audit (4%) and to other studies.9,29 It is possible that some of the variceal hemorrhages in our study may have been incorrectly coded to esophageal hemorrhage, but a sensitivity analysis, assuming the most likely misclassification of all esophageal hemorrhage codes being miscoded variceal bleeds, did not alter the adjusted reduction in mortality.

tistical analysis was performed with the use of SPSS Version 16.0 (SPSS Inc, Chicago, IL). Statistical significance of different groups was assessed with parametric tests (one-way analysis of variance with post test after statistical standards or paired Student t test). P < .05 was considered statistically significant. Results Increased Muc2 Production Correlates With Worm Expulsion It has been well documented that susceptible (AKR) mice harbor the T muris worms until patency (day 35 after infection; Figure 1A), whereas the resistant (BALB/c) mice start expelling worms by day 14 after infection, and expulsion is achieved by day 21 after infection22,36 (Figure 1A). Changes in the production of Muc2, the main gel-forming constituent of intestinal mucus, were explored within the cecum of AKR or BALB/c mice exposed to a high-dose T muris infection. Immunohistochemical staining and reverse transcription–polymerase chain reaction (RT-PCR) analysis for Muc2 (Figure 1B and C) showed that significantly higher amounts of Muc2 were expressed within the cecal crypts of the resistant mice on day 21 after infection than in the naïve and susceptible mice; a similar staining pattern was observed with the PAS reagent (Supplementary Figure 1). This increase in goblet cell number and Muc2 levels was restricted to the niche of the parasite, was not observed in the colon (Supplementary Figure 1), and correlated with worm expulsion. Therefore, to further understand the role of Muc2 in T muris infection, we performed a high-dose infection in Muc2-deficient mice on the resistant C57BL/6 background.

and Muc2 levels was restricted to the niche of the parasite, was not observed in the colon (Supplementary Figure 1), and correlated with worm expulsion. Therefore, to further understand the role of Muc2 in T muris infection, we performed a high-dose infection in Muc2-deficient mice on the resistant C57BL/6 background. Muc2 Deficiency Delayed T muris Worm Expulsion From Infected Mice A high-dose T muris infection established in both WT and Muc2-KO mice showed no marked difference in the number of worms at day 13 after infection (Figure 2A). However, as infection progressed, there was a significant decrease in worm burden in the WT mice, evident by day 15 after infection (46% reduction) and with a 84% decrease over establishment levels by day 20 after infection. In contrast, in the Muc2-deficient mice there was no decrease in worm burdens until after day 20 after infection, although mice did eventually expel their parasites.

worm burden in the WT mice, evident by day 15 after infection (46% reduction) and with a 84% decrease over establishment levels by day 20 after infection. In contrast, in the Muc2-deficient mice there was no decrease in worm burdens until after day 20 after infection, although mice did eventually expel their parasites. Muc2 Deficiency Had No Significant Effect on TH2-Type Immune Response Elicited by T muris Infection We next sought to determine whether the delay in worm expulsion in the KO mice was due to an alteration to the adaptive immune response to T muris infection. Interleukin-4 (IL-4) and interferon-γ levels in intestinal tissue were not detectable in the naïve, WT, and KO mice. Furthermore, there was no significant difference in IL-4 or interferon-γ levels in intestinal tissues between both strains on day 20 after infection (Figure 2B). Consistent with the local immune response, there was no significant difference in IL-4 and IL-13 production from in vitro concanavalin A–stimulated spleen cells (Figure 2B). Thus, despite the delay in worm expulsion, Muc2 deficiency had no significant effect on generation of the TH2-type immune response in T muris infection. The crypt architecture, an indicator of inflammation, changed during infection; there was an increase in crypt length on day 15 after infection in WT and KO mice which was more pronounced in the KO mice (Figure 2C). With the use of the highest position of BrdU-positive cells in the crypts as a measure of rate of epithelial cell turnover,33 it was clear that cell turnover was higher in the naïve KO mice than in the WT mice. However, there was no significant difference in epithelial cell turnover between the KO and WT mice on day 20 after infection (Figure 2D), indicating that the delay in worm expulsion in KO mice was not associated with an alteration of the “epithelial escalator.”37

er was higher in the naïve KO mice than in the WT mice. However, there was no significant difference in epithelial cell turnover between the KO and WT mice on day 20 after infection (Figure 2D), indicating that the delay in worm expulsion in KO mice was not associated with an alteration of the “epithelial escalator.”37 T muris Infection Induced Expression of PAS-Positive Goblet Cells in Muc2-Deficient Mice Despite the similar number of goblet cells (as defined by Relm-β and Tff3; Supplementary Figure 2) in the infected and noninfected KO and WT mice, there was a significant difference between the number of PAS-positive goblet cells (Figure 3A). As with the resistant BALB/c mice, there was a significant increase in the numbers of PAS-positive goblet cells in WT mice after infection. Although there was significant impairment in the development of hyperplastic goblet cells in the KO mice, unexpectedly by day 15 after infection, there was an increase in PAS-positive goblet cells, with significant elevation by day 30 after infection (Figure 3A; Supplementary Figure 3).

ells in WT mice after infection. Although there was significant impairment in the development of hyperplastic goblet cells in the KO mice, unexpectedly by day 15 after infection, there was an increase in PAS-positive goblet cells, with significant elevation by day 30 after infection (Figure 3A; Supplementary Figure 3). T muris Infection Triggers Muc5ac Mucin Production After exposure to T muris the levels of Muc2 were significantly elevated in the WT mice (Supplementary Figure 3). As expected, no Muc2-positive goblet cells were seen in the KO mice. Similarly, higher amounts of Muc2 (assessed by Western blotting after agarose gel electrophoresis) were present in the content of mucus collected in the WT mice after infection (Figure 3B). Although there was little evidence of mature, glycosylated Muc2 in the KO mice, interestingly, on day 21 after infection, there was a faint band consistent with the electrophoretic migration of Muc2 in these mice (red box; Figure 3B). This, along with the PAS-positive goblet cells, suggested the presence of another polymeric mucin after infection. To identify this mucin, the mucus (pooled from 5 KO mice) was analyzed by Western blotting after agarose gel electrophoresis. A mouse Muc5b-antiserum did not show any bands (data not shown). In contrast, a Muc5ac monoclonal antibody31,38 identified bands in the mucus samples from infected mice (Figure 4A). Immunofluorescence microscopy (Figure 4C), RT-PCR (Figure 4D), and tandem mass spectrometry (data not shown) confirmed the de novo expression of Muc5ac after infection in the KO mice. Furthermore, the PAS-stained material after agarose gel electrophoresis showed coincidence with the Muc5ac reactive band (Figure 4B arrow), suggesting that Muc5ac is a significant component of the mucus in the KO animals. No marked changes were observed in the expression of the cell surface mucins, Muc1, Muc4, and Muc17, which are thought to contribute to mucosal protection (Supplementary Figure 3).

ncidence with the Muc5ac reactive band (Figure 4B arrow), suggesting that Muc5ac is a significant component of the mucus in the KO animals. No marked changes were observed in the expression of the cell surface mucins, Muc1, Muc4, and Muc17, which are thought to contribute to mucosal protection (Supplementary Figure 3). Muc5ac Is Up-Regulated As Part of the “Normal” Response to Worm Expulsion Unexpectedly, Muc5ac expression was also significantly up-regulated in the WT mice on days 14 and 21 after infection (Figure 5). In contrast to the KO mice, Western blotting showed that Muc5ac mucin was not the main component in the mucus, because the main PAS bands migrated further than the broad, Muc5ac-reactive band (Figure 5arrow) and was coincident with Muc2 staining bands (data not shown). However, the de novo expression of Muc5ac was only observed in the resistant mouse models (high dose in C57BL/6 and BALB/c mice) and not in the susceptible models (low dose in BALB/c, high dose in AKR and SCID mice) (Figure 5C). Immunofluorescence microscopy and immunohistochemistry confirmed the expression of Muc5ac after infection (days 15 and 21) in the cecal crypts of the resistant models (Figure 5D). No reactivity was observed in the susceptible models (data not shown).

els (low dose in BALB/c, high dose in AKR and SCID mice) (Figure 5C). Immunofluorescence microscopy and immunohistochemistry confirmed the expression of Muc5ac after infection (days 15 and 21) in the cecal crypts of the resistant models (Figure 5D). No reactivity was observed in the susceptible models (data not shown). Susceptibility Is Associated With Altered Mucus Porosity Fluorescently labeled beads were used to investigate mucus permeability after infection (day 19) in BALB/c and AKR mice (Figure 6A). The beads traveled to a depth of approximately 100 μm over a 60-second period in both strains. Thereafter, there was a reduction in diffusion rate of the beads in the resistant (BALB/c) mice. However, the beads traveled significantly further in the mucus of susceptible (AKR) mice over a 20-minute period.

e (Figure 6A). The beads traveled to a depth of approximately 100 μm over a 60-second period in both strains. Thereafter, there was a reduction in diffusion rate of the beads in the resistant (BALB/c) mice. However, the beads traveled significantly further in the mucus of susceptible (AKR) mice over a 20-minute period. Worms in a Resistant Environment Have a Reduced Energy Status ATP measurements were carried out to determine the energy status of worms in resistant and susceptible mice as a measure of worm vitality. As infection progressed (day 21 after infection) in the AKR mice, there was a significant increase in the ATP production by the worms (Figure 6B). In contrast, there was a marked reduction in ATP production in the worms isolated from the BALB/c mice. However, these worms were not irreversibly damaged because they recovered their ATP production when transferred to in vitro culture with the colonic LS174T cell line for 24 hours (Supplementary Figure 4). Importantly, worms taken from Muc2-deficient mice showed a comparable drop in energy status during worm expulsion (Supplementary Figure 4).

e not irreversibly damaged because they recovered their ATP production when transferred to in vitro culture with the colonic LS174T cell line for 24 hours (Supplementary Figure 4). Importantly, worms taken from Muc2-deficient mice showed a comparable drop in energy status during worm expulsion (Supplementary Figure 4). Discussion It is well established that T muris survives by eliciting a TH1 response in mice susceptible to chronic infection in the absence of a TH2 response. In common with all other studies of intestinal helminth immunity, multiple effectors under immunologic (TH2) control are probably operating during worm expulsion. Although we already know that IL-13–mediated regulation of epithelial cell turnover and smooth muscle contractility can contribute to worm expulsion, little detail is known about the protective role of the secreted barrier, ie, mucus.22,24,39 Previously, we have shown that the TH2-type immune response in resistance plays an important role in the development of goblet cell hyperplasia.14,40 Some reports have also suggested that mucus produced from goblet cells has an important role in trapping and removing nematodes from the intestine.17,18,20 The polymeric mucins are responsible for the physical properties of the mucus barrier,41,42 and changes in mucins are associated with pathophysiology of a number of gastrointestinal disorders.6,12 It has also been shown that deficiency in the main component of the intestinal mucus barrier, Muc2, leads to an abnormal morphology of the colon and contributes to the onset and perpetuation of dextran sulfate sodium–induced experimental colitis.6,11

d with pathophysiology of a number of gastrointestinal disorders.6,12 It has also been shown that deficiency in the main component of the intestinal mucus barrier, Muc2, leads to an abnormal morphology of the colon and contributes to the onset and perpetuation of dextran sulfate sodium–induced experimental colitis.6,11 In this study we demonstrated, using the T muris model, that Muc2 increased in resistance (restricted to the cecum, the niche of the parasite) which correlated with worm expulsion. However, this was not the case for the mice susceptible (AKR) to T muris infection, supporting the hypothesis that Muc2 contributed to host protection in nematode infection. A distinct functional role for Muc2 in host protective immunity in T muris infection was shown in the Muc2-deficient mice. These animals exhibited a significant delay in worm expulsion even though the adaptive immune response was unaltered; similar TH2-type immune responses were shown in Muc2-deficient and WT control mice after infection. Unexpectedly, de novo expression of Muc5ac was observed just before worm expulsion in the Muc2 KO mice and resistant mouse models, but not in the susceptible models. Overall, the network properties of the intestinal mucus barrier are different between resistance and susceptibility, and the changes in the parasitic niche can have damaging effects on the vitality of the parasite. To our knowledge this is the first direct demonstration for a functionally protective role of gel-forming mucins in nematode infection.

s of the intestinal mucus barrier are different between resistance and susceptibility, and the changes in the parasitic niche can have damaging effects on the vitality of the parasite. To our knowledge this is the first direct demonstration for a functionally protective role of gel-forming mucins in nematode infection. Analysis of cecal mucus from Muc2-deficient mice showed that Muc5ac was the only polymeric mucin present in the mucus after infection. Moreover, in WT mice, although not the main mucin (which is Muc2), for the first time in a nematode infection we show the up-regulation of Muc5ac after intestinal infection. Several studies have elucidated that TH2-type cytokines such as IL-13 have the ability to up-regulate MUC5AC/Muc5ac expression levels.43,44 Therefore, the up-regulation in Muc5ac expression observed after infection, in both WT and Muc2-deficient mice, may be a result of IL-13 production. Interestingly, this de novo expression of Muc5ac was observed in all the resistant models (TH2-type response) but in none of the susceptible models (TH1-type response) of T muris infection. Although this mucin is predominantly found in airway and stomach mucus,42,45 studies on patients with ulcerative colitis and adenocarcinomas have shown MUC5AC expression in the intestine along with MUC2.31,46 However, this is the first time that Muc5ac expression has been implicated in response to an enteric parasitic infection.

his mucin is predominantly found in airway and stomach mucus,42,45 studies on patients with ulcerative colitis and adenocarcinomas have shown MUC5AC expression in the intestine along with MUC2.31,46 However, this is the first time that Muc5ac expression has been implicated in response to an enteric parasitic infection. We observed no discernible PAS-positive goblet cells throughout the cecum of Muc2-deficient mice without infection. However, this was not due to the absence of goblet cell lineage as the expression of Tff3, and Relm-β was observed in the cecum of both infected and noninfected WT and Muc2-deficient mice. This observation corroborates with the findings of Van der Sluis et al6 whereby the expression of Tff3 was observed, despite the lack of PAS-positive goblet cells. Muc2 seems to be the main phenotypic determinant of goblet cells, and, in the absence of Muc2, goblet cells lose their characteristic goblet-like shape and specific staining, but the goblet cell lineage is still present.6,11 Interestingly, after infection there was an increase in PAS-positive goblet cells in the Muc2-deficient mice. Although the size of the goblet cells in Muc2-deficient mice was smaller than in those in WT mice, their emergence correlated with worm expulsion.

fic staining, but the goblet cell lineage is still present.6,11 Interestingly, after infection there was an increase in PAS-positive goblet cells in the Muc2-deficient mice. Although the size of the goblet cells in Muc2-deficient mice was smaller than in those in WT mice, their emergence correlated with worm expulsion. We have shown a functional role for the mucus barrier in host protective immunity to T muris infection because in the absence of Muc2, worm expulsion is significantly delayed. Moreover, the physical properties of the mucus barrier are changed after infection, although the details of how these changes contribute to protection remains to be fully elucidated. However, one possibility is that, in the susceptible mice, the lower levels of Muc2 result in a network that may compromise defense because of inappropriate presentation or concentration of other host defense proteins in the environment of the worms. Whereas, in the resistant mice, other proteins (such as Relm-β, Tff3, and angiogenin) may be retained and effectively concentrated at the sites of worm infection. This may be by specific interactions with Muc2 or with the infected induced Muc5ac, or by the physical constraints imposed by the mucin network, thus rendering the host interface unsuitable for worm reproduction and/or survival which results in expulsion.19 Indeed, changes in the niche of the parasite do have a detrimental effect on the parasite, because worms extracted from mice during worm expulsion clearly have a reduced energy status than worms extracted from the susceptible mice. This reduction in the worm vitality is reversible if worms are transferred to a “favorable” environment, supporting the notion that expulsion reflects damaged, but not killed parasites.

e worms extracted from mice during worm expulsion clearly have a reduced energy status than worms extracted from the susceptible mice. This reduction in the worm vitality is reversible if worms are transferred to a “favorable” environment, supporting the notion that expulsion reflects damaged, but not killed parasites. Another explanation supported by our finding, which is by no means mutually exclusive, is that the physical nature of the mucus barrier is changed in such a way as to facilitate worm expulsion. We have shown that around the time of worm expulsion the mucus barrier is less porous in the resistant mice than in the susceptible mice, and this alteration in physical properties of the barrier after infection may directly affect the niche of the worms. The intestinal mucus barrier is proposed to comprise “loose” outer layer and a less porous, adherent inner layer.7 The results from the bead penetration assay showed that after 60 seconds the beads travelled to a depth of approximately 100 μm in the mucus from the susceptible and resistant mice, suggesting the properties of the loose layer are similar in both. However, the beads traveled at different rates thereafter, suggesting that the differences in network properties observed between the resistant and susceptible mice are mainly in the inner adherent layer of the barrier. This alteration may physically constrain the worms, thus affecting the niche.

r are similar in both. However, the beads traveled at different rates thereafter, suggesting that the differences in network properties observed between the resistant and susceptible mice are mainly in the inner adherent layer of the barrier. This alteration may physically constrain the worms, thus affecting the niche. What might be the role of the infection-induced mucin Muc5ac in protection against the worms? Muc5ac is assembled in a different manner to Muc2 and does not possess the disulphide-resistant cross-links present in Muc247–50 and may result in a mucus gel with different rheologic properties. Indeed, Muc5ac is a main component of airway mucus, and, unlike the intestinal barrier which is normally an adherent Muc2-rich gel, a specific functional requirement in the airways makes a transportable mucus gel. Thus, Muc5ac may change the rheologic nature of the mucus gel and, in conjunction with the intestinal muscle hypercontractility (controlled by TH2 response14,40), could physically aid worm expulsion. This is consistent with the observations of mucus trapping in N. brasiliensis and T. spiralis infection, in which globules of mucus trap worms, which are then transported out of the intestine.18,20 Another interesting possibility raised by the data is that expulsion occurs in 2 phases: an early phase influenced by Muc2 and a final, clearance phase that occurs independently of Muc2, possibly involving Muc5ac.

is infection, in which globules of mucus trap worms, which are then transported out of the intestine.18,20 Another interesting possibility raised by the data is that expulsion occurs in 2 phases: an early phase influenced by Muc2 and a final, clearance phase that occurs independently of Muc2, possibly involving Muc5ac. In conclusion, this study clearly shows that the mucus barrier is a significant component of a well-coordinated response in the gut to worm expulsion. Even though T muris has an intracellular niche within the gut epithelium, in resistance as the “epithelial escalator” displaces worms, it may be that the overall changes in barrier have a subsequent significant detrimental effect on the worm itself, and the additional changes in the physical properties of mucus contribute to the efficient elimination of the worms from the intestinal lumen. Moreover, it further highlights the functionally dynamic and highly regulated nature of the mucus barrier during immunologically mediated intestinal disease.

on the worm itself, and the additional changes in the physical properties of mucus contribute to the efficient elimination of the worms from the intestinal lumen. Moreover, it further highlights the functionally dynamic and highly regulated nature of the mucus barrier during immunologically mediated intestinal disease. Supplementary Methods RT-PCR Total RNA from epithelial cells was isolated with the use of the previously described method.27 cDNA was generated with the use of an IMPROM-RT kit (Promega) and Absolute QPCR SYBR Green (ABgene Epsom, Surrey, United Kingdom) was used for quantitative PCR. Primer efficiencies were determined with the use of cDNA dilutions, and genes of interest were normalized against housekeeping gene, β-actin, and expressed as a fold difference to uninfected naïve message levels. mRNA expression was investigated with the primers 5′-GTGGGCCGCTCTAGGCACCAA-3′ and 5′-CTCTTTGATGTCACGCACGATTTC-3′ for β-actin, 5′-GTCCAGGGTCTGGATCACA- 3′ and 5′-CAGATGGCAGTGAGCTGAGC-3′ for Muc2; 5′-GTGATGCACCCAT GATCTATTTTG-3′ and 5′-ACTCGGAGCTATAACAGGTCATGTC-3′ for Muc5ac, GGTTGCTTTGGCTATCGTCTATTT and AAAGATGTCCAGCTGCCCATA for Muc1, CCACCTCCTCGACCCTTACT and CTCCGACTTCAGACCCGTAG for Muc4, 5′-GTGGGACGGGCTCAAATG-3′ and 5′-CTC TACGCTCTCCACCAGTTCCT-3′ for Muc17, 5′-TTGCTGGGTCCTCTGGGATA-3′ and 5′-GCCGGCACCATA CATTGG-3′ for Tff3, and 5′-GCTCTTCCCTTTCCTTCTCCAA-3′ and 5′-ACCACAGTGTAGGCTTCATGCTGTA-3′ for Relm-β. RT-PCR products were directly sequenced to verify the identity of the amplified genes. In brief, products were digested with Exonuclease I and calf intestinal phosphatase and subsequently sequenced with the use of the ABIPRISM (Applied Biosystems, Foster City, CA) Big-Dye Terminator cycle sequencing reaction at the Sequencing Facility in the University of Manchester. The data were analyzed with Chromas Pro v1.34 (Technelysium P/L, Tewantin, QLD, Australia), and the sequences obtained were compared against the GenBank database (http://www.ncbi.nlm.nih.gov/BLAST).

Foster City, CA) Big-Dye Terminator cycle sequencing reaction at the Sequencing Facility in the University of Manchester. The data were analyzed with Chromas Pro v1.34 (Technelysium P/L, Tewantin, QLD, Australia), and the sequences obtained were compared against the GenBank database (http://www.ncbi.nlm.nih.gov/BLAST). Rate of Epithelial Cell Turnover The rate of intestinal epithelial cell turnover was assessed by visualizing BrdU incorporated into nuclei after mice were injected with 10 mg of BrdU 16 hours before killing, as described previously.37 Evaluation of In Vitro Cytokines Production From Splenocytes Single-cell suspensions of spleen were prepared in RPMI 1640 containing 10% fetal calf serum, 5 mmol/L l-glutamine, 100 U/mL penicillin, 100 mg/mL streptomycin, 25 μm/L HEPES, 0.05 mmol/L 2-ME (all from Gibco-BRL, Carlsbad, CA). Cells (107) were incubated in the presence of 5 mg/mL concavalin A (Con A). IL-4 and IL-13 levels in the supernatant were measured by enzyme immunoassay with the use of a commercially available kit (R&D Systems, Minneapolis, MN).

00 mg/mL streptomycin, 25 μm/L HEPES, 0.05 mmol/L 2-ME (all from Gibco-BRL, Carlsbad, CA). Cells (107) were incubated in the presence of 5 mg/mL concavalin A (Con A). IL-4 and IL-13 levels in the supernatant were measured by enzyme immunoassay with the use of a commercially available kit (R&D Systems, Minneapolis, MN). Investigation of Intestinal Tissue Cytokine Levels Frozen intestinal tissues were homogenized in lysis buffer containing protease inhibitor cocktail (Sigma, Indianapolis, IN). The homogenates were freeze-thawed 3 times and centrifuged, and then supernatant was collected and stored at −20°C until analyzed. Interferon-γ, IL-4, and IL-13 levels in the supernatant were measured by enzyme immunoassay technique with the use of a commercially available kit purchased from R&D Systems. Concentration of protein in the intestinal tissue was determined by a commercially available DC Protein Assay kit (Bio-Rad, Hercules, CA), and the amount of cytokines in the tissues was expressed per milligram of tissue protein. Quantification of Histologic Staining The numbers of goblet cells expressed per crypt were counted in 50 longitudinally sectioned crypt units. The area stained (pixel/mm2) per 100 crypts was determined by using the ImageJ software Version 1.39a (National Institutes of Health, Bethesda, MD).

Investigation of Intestinal Tissue Cytokine Levels Frozen intestinal tissues were homogenized in lysis buffer containing protease inhibitor cocktail (Sigma, Indianapolis, IN). The homogenates were freeze-thawed 3 times and centrifuged, and then supernatant was collected and stored at −20°C until analyzed. Interferon-γ, IL-4, and IL-13 levels in the supernatant were measured by enzyme immunoassay technique with the use of a commercially available kit purchased from R&D Systems. Concentration of protein in the intestinal tissue was determined by a commercially available DC Protein Assay kit (Bio-Rad, Hercules, CA), and the amount of cytokines in the tissues was expressed per milligram of tissue protein. Quantification of Histologic Staining The numbers of goblet cells expressed per crypt were counted in 50 longitudinally sectioned crypt units. The area stained (pixel/mm2) per 100 crypts was determined by using the ImageJ software Version 1.39a (National Institutes of Health, Bethesda, MD). Worm Isolation for ATP Analysis The cecum was longitudinally cut and segmented before incubation with 0.1 mol/L NaCl for 2 hours at 37°C with frequent shaking. Worms were counted after separation from debris and epithelial cells with the use of a 0.7-μm filter and kept in RPMI 1640 supplemented with 10% fetal calf serum.. Alive worms were subsequently homogenized with the use of the FastPrep homogeniser (MP Biomedicals, Irvine, CA).Supplementary Figure 1 PAS staining in the cecum showed a significant increase in goblet cell numbers only in the resistant (BALB/c) mice with infection (A). Worms are highlighted by arrows visible in the sections from susceptible mice. No main changes in goblet cell numbers in the colon of resistant (BALB/c) and susceptible (AKR) mice on day 14 and day 21 after infection compared with naïve (B). Representative of 3 mice. *P < .05.

esistant (BALB/c) mice with infection (A). Worms are highlighted by arrows visible in the sections from susceptible mice. No main changes in goblet cell numbers in the colon of resistant (BALB/c) and susceptible (AKR) mice on day 14 and day 21 after infection compared with naïve (B). Representative of 3 mice. *P < .05. Supplementary Figure 2 Expression of Tff3 (A) and Relm-β (B) were determined with the use of immunohistochemistry and RT-PCR in cecal tissue of Muc2 KO mice, and their resistant WT littermates on day 15 and day 20 after infection, respectively. RT-PCR showed no main changes in the mRNA expression of cell surface mucins, Muc1 (C), Muc4 (D), or Muc17 (E) in the WT and KO mice on day 20 after infection. Red dashed lines indicate naïve levels. Scale bar, 10 μm. Representative of 5 mice. *P < .05. Supplementary Figure 3 PAS staining with and without fast green counterstaining, and immunofluorescent staining with mMuc2 antibody of cecal tissue of Muc2 KO mice and their resistant WT littermates (A). Arrows highlight the emergence of smaller PAS-positive goblet cells in the Muc2 KO mice. Quantification of mMuc2 antibody staining represented as area stained in pixels per mm2 (B). RT-PCR confirms the increase in Muc2 levels after infection in resistant mice (C; red dashed line indicates naïve levels). Representative of 5 mice. *P < .05.

emergence of smaller PAS-positive goblet cells in the Muc2 KO mice. Quantification of mMuc2 antibody staining represented as area stained in pixels per mm2 (B). RT-PCR confirms the increase in Muc2 levels after infection in resistant mice (C; red dashed line indicates naïve levels). Representative of 5 mice. *P < .05. Supplementary Figure 4 (A) ATP production (data presented as relative light units per worm) was determined in the worms isolated from the resistant (BALB/c) or susceptible (AKR) mice compared with isolated worms transferred onto LS174T cell culture on day 19 after infection. (B) ATP production by worms isolated from Muc2-deficient mice and their WT littermates was determined on days 18 and 23 after infection. Representative of 3 mice. *P < .05. Acknowledgments The authors thank Prof Timothy E. Hardingham (University of Manchester), Associate Prof Michael A. McGuckin (MMRI, Brisbane, Australia), and Dr Stephen Collins (McMaster University, Canada) for their invaluable input, and Trish Blennerhassett for technical support. S.Z.H. and H.W. contributed equally. D.J.T. and W.I.K. share the senior authorship. Conflicts of interest The authors disclose no conflicts. Funding This work is supported by grants from the Canadian Institutes of Health Research (CHIR), the Crohn's and Colitis Foundation of Canada (CCFC), BBSRC, and the Wellcome Trust. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi:10.1053/j.gastro.2010.01.045.

Funding This work is supported by grants from the Canadian Institutes of Health Research (CHIR), the Crohn's and Colitis Foundation of Canada (CCFC), BBSRC, and the Wellcome Trust. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi:10.1053/j.gastro.2010.01.045. Figure 1 Worm burdens were assessed in both resistant (BALB/c) and susceptible (AKR) mice (A). Immunohistochemistry with mMuc2 antibody (B) and RT-PCR (C) were used to determine changes in Muc2 levels during infection. Nematodes are depicted by arrows (B). Red dashed line indicates naïve levels (C). Representative of 3 mice. Scale bar, 50 μm. *P < .05, **P < .01. Figure 2 Muc2-deficient mice and their resistant WT (C57BL/6 background) littermates were infected orally with 300 eggs of T muris, and worm burdens were investigated on days 13, 15, 20, 25, and 30 after infection (A). Cytokine levels were determined in intestinal tissues (in pg/mg) or by concanavalin A stimulation of spleen cells (in pg/mL) (B). Cecal crypt length was measured (C), and crypt position of the highest BrdU+ cell (D) in Muc2-deficient and WT mice was determined. Representative of 5 mice. †P < .05 compared with day 13 after infection; *P < .05 compared with wild types. IFN-γ, interferon-γ.

concanavalin A stimulation of spleen cells (in pg/mL) (B). Cecal crypt length was measured (C), and crypt position of the highest BrdU+ cell (D) in Muc2-deficient and WT mice was determined. Representative of 5 mice. †P < .05 compared with day 13 after infection; *P < .05 compared with wild types. IFN-γ, interferon-γ. Figure 3 Quantification of goblet cell numbers in the cecum of WT and Muc2-deficient mice during infection (A); goblet cells marked by arrows in deficient mice can be visualized on day 30 after infection (PAS staining without fast green counterstain). Total mucus scraped from WT and Muc2-deficient mice were reduced/alkylated, separated by agarose gel electrophoresis, analyzed by Western blot, and probed with the mMuc2 antibody (B). The relative staining intensity of the mMuc2 antibody in the portion of the blot indicated by brackets was measured. A faint band (red box highlighted) was observed on day 21 after infection in the Muc2-deficient mice. The 2 Muc2 bands in the WT animals most likely represent the monomeric (●) and dimeric (▲) forms of Muc2 (B). Representative of 5 mice. *P < .05, **P < .01. ND indicates not detectable. Figure 4 Muc5ac (A) and total glycoprotein (B) levels present in cecal mucus, determined by Western blotting using 45M1 antibody and PAS staining, respectively, in the Muc2-deficient mice. Immunofluorescence microscopy (C) and RT-PCR (D) illustrated Muc5ac was present in the Muc2-deficient mice after infection. D; Red dashed line = naïve levels. Representative of 5 mice. Scale bar; 10 μm. *P < .05, **P < .01.

Western blotting using 45M1 antibody and PAS staining, respectively, in the Muc2-deficient mice. Immunofluorescence microscopy (C) and RT-PCR (D) illustrated Muc5ac was present in the Muc2-deficient mice after infection. D; Red dashed line = naïve levels. Representative of 5 mice. Scale bar; 10 μm. *P < .05, **P < .01. Figure 5 Muc5ac (A) and total glycoprotein (B) levels present in cecal mucus, determined by Western blotting with the use of 45M1 antibody and PAS staining, respectively, in the WT resistant (C57BL/6) mice. RT-PCR showed that Muc5ac levels increase significantly only in the resistant models (high-dose infection in BALB/c and C57BL/6 mice) and not in the susceptible models (low-dose infection in BALB/c and high-dose infection in AKR and SCID mice) (C; red dashed line indicates naïve levels). Immunofluorescence microscopy and immunohistochemistry showed that Muc5ac was present in some of the goblet cells of resistant mice after infection (D). Representative of 5 mice. Scale bar, 10 μm. *P < .05. Figure 6 Fluorescent beads were used to determine the permeability of the mucus barrier of the susceptible (AKR) and resistant (BALB/c) mice on day 19 after infection, represented as the distance traveled from the top of the mucus barrier in the time stated (A). Energy levels (data presented as relative light units per worm) were determined in worms extracted from BALB/c and AKR mice during infection (B).

Hepatitis C virus (HCV) is a common chronic viral infection with only a minority of individuals exposed to HCV infection being able to resolve infection spontaneously. Clearance of HCV is dependent on a successful immune response, which likely involves T cells, B cells, dendritic cells, and also natural killer cells (NK) cells.1 Consistent with a broad immune response being important, polymorphisms of both the innate and adaptive immune system are associated with spontaneous resolution of HCV infection.2 Recent work has highlighted that polymorphisms in the Interleukin-28B (IL28B) gene (interferon [IFN]-λ3) are strongly associated with both spontaneous resolution of HCV infection and also resolution of infection with pegylated interferon and ribavirin.2–7 Similarly, the killer cell immunoglobulin-like receptors (KIR) and their human leukocyte antigen class I ligands have also been implicated in spontaneous and treatment-induced resolution of HCV infection.8–11 In particular, KIR2DL3 and its ligands, the group 1 HLA-C allotypes (HLA-C1), are protective against chronic HCV infection and, hence, are beneficial factors in outcome following exposure to HCV.

cyte antigen class I ligands have also been implicated in spontaneous and treatment-induced resolution of HCV infection.8–11 In particular, KIR2DL3 and its ligands, the group 1 HLA-C allotypes (HLA-C1), are protective against chronic HCV infection and, hence, are beneficial factors in outcome following exposure to HCV. A minority of long-term injection drug users (IDU) demonstrate apparent resistance to HCV infection and remain seronegative and aviremic despite likely repeated exposure to HCV through the sharing of drug injection equipment. These exposed but uninfected (EU) IDU cases have been shown to have detectable HCV-specific T-cell responses, indicating their exposure to HCV infection.12,13 They also have increased NK cell activity.14 Consistent with this, we have recently shown that, similar to conventional spontaneous resolvers (SR), the combination of KIR2DL3 and HLA-C1 is also over-represented in the exposed seronegative aviremic population.10 Additionally, both groups of protected individuals have an increased frequency of a functional interleukin-12 (IL-12) polymorphism as compared with chronically infected individuals.15,16

gdom) machine using 96-well plates and 10–100 ng genomic DNA with 0.5 μmol/L of each primer in a reaction mix of total volume 20 μL. The thermal cycling protocol consisted of an initial denaturation step of 95°C for 10 minutes, followed by 40 two-step amplification cycles of 95°C for 20 seconds and 58°C for 20 seconds. KIR2DL2/3 genotyping was performed on the Hencore cohort and the 32 additional exposed uninfected individuals by polymerase chain reaction using sequence specific primers as previously described.19 HLA typing was performed on the Hencore and EU cohorts as described elsewhere.20 HLA types that were not resolved by sequencing or that gave unusual results were also tested by sequence-specific oligonucleotide probe typing using commercial kits (RELI SSO; Dynal, Wirral, United Kingdom). Other cohorts had previously been typed for KIR2DL2/3 and HLA-C.8,10 Statistical Analysis GraphPad Prism 5 software (GraphPad, Inc, La Jolla, CA) was used to calculate 2-tailed P values and odds ratios (OR) from 2 × 2 contingency tables by Fisher exact test. Logistic regression analysis was performed using SPSS statistical software version 17 (SPSS, Inc, Chicago, IL) with the ENTER function. Synergy between IL28B and KIR:HLA was calculated using the method of Cortina-Borja et al.21

ous resolvers (SR), the combination of KIR2DL3 and HLA-C1 is also over-represented in the exposed seronegative aviremic population.10 Additionally, both groups of protected individuals have an increased frequency of a functional interleukin-12 (IL-12) polymorphism as compared with chronically infected individuals.15,16 To date, the protective effect of IL28B in this subgroup of individuals has not been investigated. Furthermore, it is not well understood whether protective polymorphisms in the immune system work together to increase protection against chronic HCV infection or whether these components of the innate immune system act independently. The aim of this study was therefore to determine whether the EU population have a protective IL28B genotype and to determine how protective IL28B and KIR:HLA-C polymorphisms may interact to influence the outcome of HCV infection in untreated individuals. Patients and Methods Patients Three hundred ninety-seven patients (74 exposed uninfected, 89 SR, and 234 chronically infected patients) were studied for the distribution of the IL28B.rs12979860 single nucleotide polymorphism (SNP), KIR2DL2/3 and HLA-C genotypes. All patients gave informed consent with approval by the relevant ethics committees as previously described.8,10,17 Patients were excluded if they were human immunodeficiency virus positive or hepatitis B virus surface antigen positive. The patients are classified into the following 3 cohorts: (1) exposed uninfected (EU) cohort, (2) spontaneous resolving (SR), and (3) chronically infected individuals.

as previously described.8,10,17 Patients were excluded if they were human immunodeficiency virus positive or hepatitis B virus surface antigen positive. The patients are classified into the following 3 cohorts: (1) exposed uninfected (EU) cohort, (2) spontaneous resolving (SR), and (3) chronically infected individuals. Exposed uninfected cohort Seventy-four individuals were recruited from Dartmoor Prison, needle exchanges, community drug services, and hostels in Plymouth, United Kingdom. All these individuals were of Caucasian ethnicity. They had an extensive history of past or present injection drug use. This group was defined as being both HCV antibody (third generation enzyme linked immunosorbent assay, Abbott IMx, Abbott Diagnostics, Maidenhead, Berkshire, United Kingdom) and HCV RNA (Amplicor, Roche Diagnostics, Pleasanton, CA) negative on at least 2 occasions, 3–6 months apart with subsequent testing on an approximate 6 monthly basis to ensure that this profile remained unchanged. Forty-two of these cases had been genotyped previously for KIR2DL2/3 and HLA-C.10 Detailed information about drug injecting behavior was ascertained by means of a structured questionnaire, and the median duration of intravenous drug use was 8.62 ± 6.05 years (range, 0.3–24) with a median number of injections of 4927 (range, 36–41,620).10 Their median age was 28 years, and 64 (79%) were male.

LA-C.10 Detailed information about drug injecting behavior was ascertained by means of a structured questionnaire, and the median duration of intravenous drug use was 8.62 ± 6.05 years (range, 0.3–24) with a median number of injections of 4927 (range, 36–41,620).10 Their median age was 28 years, and 64 (79%) were male. SRs. Individuals were classified in this group if they had detectable anti-HCV by second-generation enzyme-linked immunosorbent assay (Abbott IMx; Abbott Diagnostics, Maidenhead, Berkshire, United Kingdom) and no detectable HCV viremia by Quantiplex HCV RNA 2.0 assay (Chiron, Emeryville, CA) or HCV COBAS Amplicor system (Roche Diagnostics, Pleasanton, CA) on at least 2 occasions 6 months apart. They were recruited between 1995 and 1998 as part of the Hepatitis C European Network for Cooperative Research (Hencore) collaboration17,18 and between 1999 and 2005 from Addenbrookes Hospital, Cambridge, United Kingdom, and Southampton General Hospital, United Kingdom.8 Eighty-seven (98%) were Caucasian, 59 (66%) were male, and their median age was 36 years. Forty-four had been genotyped previously for KIR2DL2/3 and HLA-C.8

ch (Hencore) collaboration17,18 and between 1999 and 2005 from Addenbrookes Hospital, Cambridge, United Kingdom, and Southampton General Hospital, United Kingdom.8 Eighty-seven (98%) were Caucasian, 59 (66%) were male, and their median age was 36 years. Forty-four had been genotyped previously for KIR2DL2/3 and HLA-C.8 Chronically infected individuals These individual were all persistently anti-HCV and HCV RNA positive, by second-generation enzyme-linked immunosorbent assay (Abbott IMx) and HCV COBAS Amplicor system (Roche Diagnostics, Pleasanton, CA), respectively. They were recruited from the general hepatology clinic at Southampton General Hospital, United Kingdom, between 2003 and 2007. Two hundred seventeen (93%) were of Caucasian origin, with a median age of 45 years, and 138 (59%) were male.10 All had been genotyped previously for KIR2DL2/3 and HLA-C.8

n, CA), respectively. They were recruited from the general hepatology clinic at Southampton General Hospital, United Kingdom, between 2003 and 2007. Two hundred seventeen (93%) were of Caucasian origin, with a median age of 45 years, and 138 (59%) were male.10 All had been genotyped previously for KIR2DL2/3 and HLA-C.8 Genotyping Genomic DNA was extracted from peripheral blood lymphocytes using a salt precipitation method,17 Nucleon DNA extraction kit (Tepnel Lifesciences, Manchester, United Kingdom) or the QIAamp blood kit (Qiagen, Crawley, United Kingdom). All samples were typed for the rs12979860 SNP using a real-time polymerase chain technique incorporating Sybr Green (Qiagen QuantiTect SYBR; Qiagen). The primers used were as follows: 5′-GCTTATCGCATACGGCTAGGC-3′ (forward common), 5′-GCAATTCAACCCTGGTTCG-3′ (C- allele specific reverse) and 5′-GCAATTCAACCCTGGTTCA-3′ (T-allele specific reverse). Reactions were performed on a 5700 Perkin Elmer (Cambridge, United Kingdom) machine using 96-well plates and 10–100 ng genomic DNA with 0.5 μmol/L of each primer in a reaction mix of total volume 20 μL. The thermal cycling protocol consisted of an initial denaturation step of 95°C for 10 minutes, followed by 40 two-step amplification cycles of 95°C for 20 seconds and 58°C for 20 seconds.

e 2-tailed P values and odds ratios (OR) from 2 × 2 contingency tables by Fisher exact test. Logistic regression analysis was performed using SPSS statistical software version 17 (SPSS, Inc, Chicago, IL) with the ENTER function. Synergy between IL28B and KIR:HLA was calculated using the method of Cortina-Borja et al.21 Results IL28B Polymorphism Distinguishes Exposed Uninfected Individuals From Anti-HCV-Positive Spontaneous Resolvers The frequency of the protective CC genotype at the SNP rs12979860-CC in the 74 EU individuals was significantly lower than in the 89 SR (41.9% vs 69.7%, respectively, P = .0005; OR, 0.31; 95% confidence interval [CI]: 0.16–0.60) but was similar to that found in the 234 individuals with chronic HCV infection (41.9% vs 43.6%, respectively) (Table 1). Consistent with previous work, the frequency of the IL28B.rs12979860-CC genotype was significantly higher in the spontaneous resolving population compared with those with chronic infection (69.7% vs 43.6%, respectively, P < .0001; OR, 2.97, 95% CI: 1.76–5.00). We also found that CT heterozygosity was more prevalent in the EU as compared with the SR population (43.2% vs 24.7%, respectively, P = .019; OR, 2.32, 95% CI: 1.19–4.52), and this genotype was lower in the SR population as compared with the chronically infected individuals (24.7% vs 48.7%, respectively, P < .0001; OR, 0.35, 95% CI: 0.20–0.60). Additionally, we found that there was a trend toward an increase in TT homozygosity in the EU population as compared with both SR (14.9% vs 5.6%, respectively, P = .06; OR, 2.93, 95% CI: 0.97–8.87) and also chronically infected individuals (14.9% vs 7.7%, respectively, P = .07; OR, 2.09, 95% CI: 0.94–4.67). This is despite the overall T allele frequency being similar between EU and chronically infected individuals (36.5% vs 32.1%, respectively) (Supplementary Tables 1 and 2). These observations remained similar if only Caucasian individuals were considered (Supplementary Table 3). Thus, the rs12979860 polymorphism distinguishes the EU population from those that spontaneously resolve HCV infection.

hronically infected individuals (36.5% vs 32.1%, respectively) (Supplementary Tables 1 and 2). These observations remained similar if only Caucasian individuals were considered (Supplementary Table 3). Thus, the rs12979860 polymorphism distinguishes the EU population from those that spontaneously resolve HCV infection. IL28B and KIR:HLA Define Distinct Populations of HCV Protected Individuals Although IL28B.rs12979860-CC was not associated with protection in the EU cohort, these individuals are genetically distinct from those with chronic HCV because homozygosity for KIR2DL3:HLA-C1 is over-represented in this population as compared with those with chronic HCV (31.1% vs 13.3%, respectively, P = .0008; OR, 2.95, 95% CI: 1.59–5.49) (Supplementary Table 4). KIR2DL3:HLA-C1 was found at a similar frequency to the anti-HCV-positive SR population (31.1% vs 29.2%, respectively, P = ns), as we have previously shown in a subgroup of these individuals.10 We therefore hypothesized that KIR and IL28B genes might define distinct groups of individuals who are protected against chronic HCV infection using different genetic pathways. To study the interrelationship of these genes on the outcome of hepatitis C, we compared the frequency of IL28B.rs12979860-CC in individuals with and without the protective KIR2DL3:HLA-C1 homozygous genotype from all 3 cohorts (EU, SR, and chronic).

tected against chronic HCV infection using different genetic pathways. To study the interrelationship of these genes on the outcome of hepatitis C, we compared the frequency of IL28B.rs12979860-CC in individuals with and without the protective KIR2DL3:HLA-C1 homozygous genotype from all 3 cohorts (EU, SR, and chronic). In individuals who had spontaneously resolved infection and were not KIR2DL3:HLA-C1 homozygous, the frequency of the rs12979860-CC genotype was significantly higher compared with chronically infected individuals (68.3% [SR] vs 41.9% [chronic], P = .0003; OR, 2.98, 95% CI: 1.64–5.43, Table 2). The effect was similar in individuals who were KIR2DL3:HLA-C1 homozygous, but this did not reach statistical significance (73.1% vs 54.8%, respectively, P = .18; OR, 2.23, 95% CI: 0.73–6.84), most likely because of the small sample size. Likewise, the protective effect of KIR2DL3:HLA-C1 homozygosity was similar in individuals with the rs12979860-CC genotype (30.6% [SR] vs 16.7% [chronic], P = .051; OR, 2.21, 95% CI: 1.04–4.68) and also without the rs12979860-CC genotype (25.9% SR vs 10.6% chronic, P = .055; OR, 2.95, 95% CI: 1.06–8.21). Similarly, we found an under-representation of rs12979860-CC in EU as compared with SR in both the KIR2DL3:HLA-C1 homozygous and nonhomozygous subgroups (P = .046; OR, 0.28, 95% CI: 0.09–0.94 and P = .0046; OR, 0.33, 95% CI: 0.15–0.70, respectively, Table 2).

SR vs 10.6% chronic, P = .055; OR, 2.95, 95% CI: 1.06–8.21). Similarly, we found an under-representation of rs12979860-CC in EU as compared with SR in both the KIR2DL3:HLA-C1 homozygous and nonhomozygous subgroups (P = .046; OR, 0.28, 95% CI: 0.09–0.94 and P = .0046; OR, 0.33, 95% CI: 0.15–0.70, respectively, Table 2). In univariate analysis, the frequency of the combination of rs12979860-CC and KIR2DL3:HLA-C1 homozygosity in the SR group was 21% as compared with only 7.3% in the chronically infected group (P = .0007; OR, 3.47, 95% CI: 1.71–7.03). However, it is not clear whether these 2 protective genetic factors are acting synergistically or independently. To determine this, we performed multivariate logistic regression analysis using 3 variables: rs12979860-CC+KIR2DL3:HLA-C1 homozygosity; rs12979860-CC with or without KIR2DL3:HLA-C1 homozygosity; and KIR2DL3:HLA-C1 homozygosity with or without rs12979860-CC. This analysis tests whether the combination of the 2 factors provides additional benefit above that due to each factor individually. This demonstrated that the rs12979860-CC genotype and KIR2DL3:HLA-C1 homozygosity are protective in isolation (P < .001 and P = .04, respectively), and having both rs12979860-CC and KIR2DL3:HLA-C1 homozygosity together does not confer any additional protection (P > .1) (Table 3). Additionally, we applied a recently described test to evaluate the synergistic effects of genetic factors.21 This method is based on logistic regression analysis and compares the ORs of protection among the different groups. It determines whether the observed OR for 2 factors considered in combination is greater than that of having both protective factors assuming independent effects of each factor. In this case, 4 groupings were tested: (1) rs12979860-CC positive, not KIR2DL3:HLA-C1 homozygous; (2) KIR2DL3:HLA-C1 homozygous, rs12979860-CC negative; (3) rs12979860-CC positive and KIR2DL3:HLA-C1 homozygous; and (4) neither rs12979860-CC positive nor KIR2DL3:HLA-C1 homozygous. Using this test, we confirmed the absence of synergy between the 2 protective factors in the SR population (synergy factor = 1.3 [95% CI: 0.37–4.75], Psynergy = .6). Because the synergy factor can uncover unexpected synergies, we determined this statistic for the EU population in comparison with the chronically infected individuals. However, no synergy was found (synergy factor = 1.53 [95% CI: 0.44–5.37], Psynergy = .5).

on (synergy factor = 1.3 [95% CI: 0.37–4.75], Psynergy = .6). Because the synergy factor can uncover unexpected synergies, we determined this statistic for the EU population in comparison with the chronically infected individuals. However, no synergy was found (synergy factor = 1.53 [95% CI: 0.44–5.37], Psynergy = .5). Thus, these polymorphisms of the innate immune system distinguish EU from both SR and chronically infected individuals and operate independently to protect individuals against chronic HCV infection.

on (synergy factor = 1.3 [95% CI: 0.37–4.75], Psynergy = .6). Because the synergy factor can uncover unexpected synergies, we determined this statistic for the EU population in comparison with the chronically infected individuals. However, no synergy was found (synergy factor = 1.53 [95% CI: 0.44–5.37], Psynergy = .5). Thus, these polymorphisms of the innate immune system distinguish EU from both SR and chronically infected individuals and operate independently to protect individuals against chronic HCV infection. Discussion HCV causes chronic infection in the majority of exposed individuals, thus protection from HCV infection is the exception rather than the norm. Individuals with beneficial immune responses have traditionally been identified as anti-HCV positive, HCV RNA negative. More recently, individuals who remain seronegative and aviremic despite high-risk behavior have also been shown to be relatively protected against chronic infection. These individuals have detectable T-cell responses,12,13,22 a favorable KIR2DL3:HLA-C genotype,10 and also a protective IL12 genotype.15,16 In this respect, they are indistinguishable from conventional SR. However, our data show that protection in this subgroup of individuals is not associated with the IL28B.rs12979860-CC genotype, which marks them as distinct from SR. Indeed, they are the first subgroup of individuals identified who have a favorable outcome following HCV exposure who do not have an over-representation of this genotype. This is unlikely to represent a bias related to the ethnicity of our population because all EU individuals were Caucasian, and the frequency of the IL28B.rs12979860-C allele (63.5%, Supplementary Table 2) is comparable with that of 67.4% reported by Thomas et al in Americans of European extraction and is also similar to the frequency found in other European populations.6 Therefore, it seems that IL28B distinguishes the population of SR from other healthy and HCV exposed populations.

allele (63.5%, Supplementary Table 2) is comparable with that of 67.4% reported by Thomas et al in Americans of European extraction and is also similar to the frequency found in other European populations.6 Therefore, it seems that IL28B distinguishes the population of SR from other healthy and HCV exposed populations. Overall, given our understanding of the protective nature of the rs12979860-CC genotype, it may be that this genotype fails to deliver protection against acute HCV infection. One alternative explanation could be that the rs12979860TT genotype is protective against acute HCV infection. Potentially, this genotype could be associated with a weaker antibody response and a bias toward both innate and adaptive cell mediated immunity. Interestingly, the rs12979860-TT genotype was over-represented in our EU cohort as compared with both SR and chronically infected individuals, consistent with a role in skewing the immune response away from antibody production. This difference is unlikely to be related to a population bias because the trend was present when Caucasian individuals alone were considered, and, also, the overall T allele frequency was similar between EU and chronically infected individuals.

with a role in skewing the immune response away from antibody production. This difference is unlikely to be related to a population bias because the trend was present when Caucasian individuals alone were considered, and, also, the overall T allele frequency was similar between EU and chronically infected individuals. Within the spontaneously resolving group are 2 distinct populations: those resolving HCV via an IL28B-associated mechanism and those with a protective KIR:HLA combination. We found that the combination of KIR2DL3:HLA-C1 and IL28B.rs12979860-CC homozygosity did not provide any additional protection above that due to each genetic factor in isolation as determined both by logistic regression and calculation of a synergy factor. This indicates that they function as independent genetic protective factors and do not have a synergistic interaction. The calculation of a synergy factor allows separation of a true synergistic interaction from an apparent one, that is, one that is due to the expected increase in OR caused by combining 2 protective factors.21 This analysis also complements that performed by logistic regression, which demonstrated that the combination of the 2 protective factors had no advantage over that due to each factor in isolation. Additionally, the synergy factor is designed to be robust for small samples sizes, even when individual cells are zero.21 Thus, overall, the absence of a synergistic interaction between these factors is consistent with the observation that KIR:HLA, but not IL28B, is protective in the EU cohort.

ach factor in isolation. Additionally, the synergy factor is designed to be robust for small samples sizes, even when individual cells are zero.21 Thus, overall, the absence of a synergistic interaction between these factors is consistent with the observation that KIR:HLA, but not IL28B, is protective in the EU cohort. Both KIR2DL3:HLA-C1 and IL28B have predominantly innate immune functions. KIR2DL3-positive NK cells are activated in the acute phase of HCV infection, and we have shown that KIR2DL3-positive NK cells from individuals who resolve HCV have higher levels of degranulation than healthy controls, but those from individuals who become chronically infected do not.23 Thus, KIR2DL3 protection operates at the level of the NK cell. At present, the mechanism of action of IL28B in resolving HCV infection, either spontaneously or with treatment, is not clear. Although, as a type III interferon that shares signaling pathways with type I interferons,24 it most likely protects via a direct mechanism on the hepatocyte, possibly inhibiting HCV replication like the related molecule IFN-λ125 or rendering cells less susceptible to infection.26 Additionally, IFN-λ2 does not appear to directly affect NK cells.27 Therefore, there is a biologic rationale for the separation of these 2 genetic effects.

es.9,29 It is possible that some of the variceal hemorrhages in our study may have been incorrectly coded to esophageal hemorrhage, but a sensitivity analysis, assuming the most likely misclassification of all esophageal hemorrhage codes being miscoded variceal bleeds, did not alter the adjusted reduction in mortality. The previous difficulties in detecting a reduction in mortality might imply that we are reaching the point where mortality becomes unavoidable because of age and comorbidity. However, because the mortality in our study continued to improve right up to the end of the study period, improvements in management would appear to be continuing to have an impact on mortality following gastrointestinal hemorrhage. The reasons for the reduction in mortality we have observed are likely to be complex. There were similar reductions in mortality whether or not an endoscopy was recorded and for all associated diagnoses, implying that endoscopic therapy was not a major contributor to the reduction in mortality. Instead, our data perhaps suggest that improvement in standard nonendoscopic care has led to improved survival, such as the routine administration of intravenous proton pump inhibitor infusions, the routine use of risk scoring, the implementation of standardized clinical guidelines, and the subsequent local auditing of practice.4,5,30

via a direct mechanism on the hepatocyte, possibly inhibiting HCV replication like the related molecule IFN-λ125 or rendering cells less susceptible to infection.26 Additionally, IFN-λ2 does not appear to directly affect NK cells.27 Therefore, there is a biologic rationale for the separation of these 2 genetic effects. One model for resistance to, or resolution of, HCV infection is that possession of multiple independent protective factors may synergize to provide protection against chronic infection so that individuals with more protective factors have a greater chance of resolution of HCV infection. Our data do not support this hypothesis. Instead, we propose that KIR:HLA and IL28B define 2 genetically distinct subpopulations of individuals who are relatively protected against chronic HCV infection. Future genetic studies of resolution of HCV infection should stratify for these genotypes to take this heterogeneity into account. Supplementary material Supplementary Table 1 Frequency and Comparisons of the IL28B.rs12979860-T Alleles in 74 Exposed Uninfected, 89 Spontaneous Resolvers, and 234 Chronically Infected HCV Patients rs12979860-T 2N (%) P value (Pc) OR (95% CI) EU vs SR 54 (36.5) vs 32 (18.0) .0002 (0.0006) 2.62 (1.58−4.36) EU vs chronic 54 (36.5) vs 150 (32.1) >.1 1.22 (0.83−1.79) SR vs chronic 32 (18.0) vs 150 (32.1) .0004 (.0012) 0.46 (0.30−0.71) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc).

.5) vs 32 (18.0) .0002 (0.0006) 2.62 (1.58−4.36) EU vs chronic 54 (36.5) vs 150 (32.1) >.1 1.22 (0.83−1.79) SR vs chronic 32 (18.0) vs 150 (32.1) .0004 (.0012) 0.46 (0.30−0.71) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; chronic, chronically infected HCV patients; 2N, number of alleles; OR, odds ratio. Supplementary Table 2 Frequency and Comparisons of the IL28B.rs12979860-C Alleles in 74 Exposed Uninfected, 89 Spontaneous Resolvers, and 234 Chronically Infected HCV Patients rs12979860-C 2N (%) P value (Pc) OR (95% CI) EU vs SR 94 (63.5) vs 146 (82.0) .0002 (.0006) 0.38 (0.23−0.63) EU vs chronic 94 (63.5) vs 318 (67.9) >.1 0.82 (0.56−1.21) SR vs chronic 146 (82.0) vs 318 (67.9) .0004 (.0012) 2.15 (1.40−3.31) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; chronic, chronically infected HCV patients; OR, odds ratio. Supplementary Table 3 Comparison of the IL28B.rs12979860-CC, -CT, and -TT Genotypes in Caucasian Individuals Genotypes EU, n (%) SR, n (%) Chr, n (%) EU vs SR EU vs Chr SR vs Chr CC 31 (41.9) 60 (69.0) 93 (42.9) P = .0008 (Pc = .002) OR = 0.32 95% CI: 0.17−0.62 P > .1 P < .0001 (Pc = .0002) OR = 2.96 95% CI: 1.75−5.02 CT 32 (43.2) 22 (25.3) 106 (48.8) P = .019 (Pc = .06) OR = 2.25 95% CI: 1.16−4.39 P > .1 P = .0002 (Pc = .0006) OR = 0.35 95% CI = 0.20−0.62 TT 11 (14.9) 5 (5.7) 18 (8.3) P = .07 (Pc = .2) OR = 2.86 95% CI: 0.05−8.67

Genotypes EU, n (%) SR, n (%) Chr, n (%) EU vs SR EU vs Chr SR vs Chr CC 31 (41.9) 60 (69.0) 93 (42.9) P = .0008 (Pc = .002) OR = 0.32 95% CI: 0.17−0.62 P > .1 P < .0001 (Pc = .0002) OR = 2.96 95% CI: 1.75−5.02 CT 32 (43.2) 22 (25.3) 106 (48.8) P = .019 (Pc = .06) OR = 2.25 95% CI: 1.16−4.39 P > .1 P = .0002 (Pc = .0006) OR = 0.35 95% CI = 0.20−0.62 TT 11 (14.9) 5 (5.7) 18 (8.3) P = .07 (Pc = .2) OR = 2.86 95% CI: 0.05−8.67 P > .1 P > .1 NOTE. Includes 74 exposed uninfected, 87 spontaneous resolvers, and 217 chronically infected Caucasian individuals. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; Chr, chronically infected HCV patients; 2N, number of alleles; OR, odds ratio. Supplementary Table 4 Frequency of KIR2DL3:HLA-C1 Homozygosity in the Whole Cohort KIR2DL3:C1 homozygosity, n (%) P value (Pc) OR (95% CI) EU vs SR 23 (31.1) vs 26 (29.2) .86 0.91 (0.47−1.79) EU vs chronic 23 (31.1) vs 31 (13.3) .0008 (.002) 2.95 (1.59−5.49) SR vs chronic 26 (29.2) vs 31 (13.3) .002 (.005) 2.70 (1.49−4.89) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; chronic, chronically infected HCV patients; OR, odds ratio. Acknowledgments The authors thank Dr Bernard North for statistical advice. Conflicts of interest The authors disclose no conflicts. Funding Supported by a Wellcome Trust Senior Clinical fellowship (to S.I.K.).

KIR2DL3:C1 homozygosity, n (%) P value (Pc) OR (95% CI) EU vs SR 23 (31.1) vs 26 (29.2) .86 0.91 (0.47−1.79) EU vs chronic 23 (31.1) vs 31 (13.3) .0008 (.002) 2.95 (1.59−5.49) SR vs chronic 26 (29.2) vs 31 (13.3) .002 (.005) 2.70 (1.49−4.89) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; chronic, chronically infected HCV patients; OR, odds ratio. Acknowledgments The authors thank Dr Bernard North for statistical advice. Conflicts of interest The authors disclose no conflicts. Funding Supported by a Wellcome Trust Senior Clinical fellowship (to S.I.K.). Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi:10.1053/j.gastro.2011.04.005. Table 1 Frequency of the IL28B.rs12979860-CC, CT, and TT Genotype in 74 Exposed Uninfected, 89 Spontaneous Resolvers, and 234 Chronically Infected HCV Patients rs12979860 Genotype EU, n (%) SR, n (%) Chr, n (%) EU vs SR EU vs Chr SR vs Chr CC 31 (41.9) 62 (69.7) 102 (43.6) P = .0005 (Pc = .002) OR = 0.31 95% CI: 0.16−0.60 P > .1 P < .0001 (Pc = .0002) OR = 2.97 95% CI: 1.76−5.00 CT 32 (43.2) 22 (24.7) 114 (48.7) P = .019 (Pc = .057) OR = 2.32 95% CI: 1.19−4.52 P > .1 P < .0001 (Pc = .0002) OR = 0.35 95% CI: 0.20−0.60 TT 11 (14.9) 5 (5.6) 18 (7.7) P = .06 (Pc > .1) OR = 2.93 95% CI: 0.97−8.87 P = .07 (Pc > .1) OR = 2.09 95% CI: 0.94−4.67

rs12979860 Genotype EU, n (%) SR, n (%) Chr, n (%) EU vs SR EU vs Chr SR vs Chr CC 31 (41.9) 62 (69.7) 102 (43.6) P = .0005 (Pc = .002) OR = 0.31 95% CI: 0.16−0.60 P > .1 P < .0001 (Pc = .0002) OR = 2.97 95% CI: 1.76−5.00 CT 32 (43.2) 22 (24.7) 114 (48.7) P = .019 (Pc = .057) OR = 2.32 95% CI: 1.19−4.52 P > .1 P < .0001 (Pc = .0002) OR = 0.35 95% CI: 0.20−0.60 TT 11 (14.9) 5 (5.6) 18 (7.7) P = .06 (Pc > .1) OR = 2.93 95% CI: 0.97−8.87 P = .07 (Pc > .1) OR = 2.09 95% CI: 0.94−4.67 P > .1 NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; Chr, chronically infected HCV patients; OR, odds ratio. Table 2 Effect of rs12979860-CC Homozygosity in 89 Spontaneously Resolving, 234 Chronically Infected, and 74 Exposed Uninfected Individuals With and Without the KIR2DL3:HLA-C1 Genotype Comparison (n) rs12979860-CC positive, n (%) P value OR (95% CI) KIR2DL3:C1 homozygous SR (26) vs Chr (31) 19 (73.1) vs 17 (54.8) >.1 2.23 (0.73−6.84) EU (23) vs SR (26) 10 (43.5) vs 19 (73.1) .046 (Pc > .1) 0.28 (0.09−0.94) Not KIR2DL3:C1 homozygous SR (63) vs Chr (203) 43 (68.3) vs 85 (41.9) .0003 (Pc = .0009) 2.98 (1.64−5.43) EU (51) vs SR (63) 21 (41.2) vs 43 (68.3) .0046 (Pc = .013) 0.33 (0.15−0.70) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; Chr, chronically infected HCV patients; OR, odds ratio.

Comparison (n) rs12979860-CC positive, n (%) P value OR (95% CI) KIR2DL3:C1 homozygous SR (26) vs Chr (31) 19 (73.1) vs 17 (54.8) >.1 2.23 (0.73−6.84) EU (23) vs SR (26) 10 (43.5) vs 19 (73.1) .046 (Pc > .1) 0.28 (0.09−0.94) Not KIR2DL3:C1 homozygous SR (63) vs Chr (203) 43 (68.3) vs 85 (41.9) .0003 (Pc = .0009) 2.98 (1.64−5.43) EU (51) vs SR (63) 21 (41.2) vs 43 (68.3) .0046 (Pc = .013) 0.33 (0.15−0.70) NOTE. Two-tailed P values were calculated for 2 × 2 contingency tables using Fisher exact test, and the Bonferroni correction was applied (Pc). 95% CI, 95% confidence interval; Chr, chronically infected HCV patients; OR, odds ratio. Table 3 Multivariate Logistic Regression Analysis of the Interaction Between Individual Protective Factors in 89 Spontaneous Resolvers and 234 Chronically Infected Individuals SR, n (%) Chronic, n (%) P value OR 95% CI 860-CC 62 (69.7) 102 (43.6) <.001 2.99 1.64−5.43 2DL3:C1 homozygous 26 (29.2) 31 (13.2) .04 2.94 1.06−8.20 860CC+2DL3:C1 homozygous 19 (21.3) 17 (7.3) >.1 0.75 0.21−2.67 NOTE. The 3 variables included in this analysis were IL28B.rs12979860CC (860-CC), KIR2DL3:HLA-C group 1 homozygosity (2DL3-C1), and IL28B.rs12979860CC in combination with KIR2DL3:HLA-C group 1 homozygosity (860-CC+2DL3-C1). 95% CI, 95% confidence interval; Chronic, chronically infected HCV patients; OR, odds ratio.

Gastrointestinal hemorrhage is the commonest cause of acute hospital admission to gastroenterology and therefore has a large impact on the acute medical admission workload. Changes in management have been shown in randomized controlled trials to improve outcome from gastrointestinal hemorrhage, but the largest observational studies of mortality trends following upper gastrointestinal hemorrhage report no improvement in overall mortality over the last 2 decades.1–3 This failure to demonstrate an improvement suggests either that clinical guidelines4,5 derived from the results of randomized controlled trials are not generalizable to the clinical population, that they are not being implemented appropriately, or that the patients have changed at the same time as the treatments. This latter explanation, with increasing age and comorbidity confounding the effects of therapy, has been proposed as the likely explanation.6,7 However, this has not been proven because to reliably measure the effect of changes in age and comorbidity on mortality necessitates larger studies than have been published. Therefore, we aimed to investigate current trends in mortality following admission from upper gastrointestinal hemorrhage in England and investigate whether these can be explained by population changes in age and comorbidity.

nges in age and comorbidity on mortality necessitates larger studies than have been published. Therefore, we aimed to investigate current trends in mortality following admission from upper gastrointestinal hemorrhage in England and investigate whether these can be explained by population changes in age and comorbidity. Patients and Methods Database The Hospital Episodes Statistics database (HES) contains information on all admissions to an NHS hospital in England, with over 12 million new records added each year. It is managed by the NHS information center and is available for research with ethical approval. All NHS hospitals within England are required to contribute to the database. There are currently 168 acute trusts in England; however, each of these trusts can manage more than 1 hospital, and over time trusts can merge and split. Over the course of our study, approximately 150–200 providers were contributing to the database. The available data consist of a number of records for each admission, which are called episodes. Each episode represents the time period of the admission that a patient was under the clinical care of a particular consultant team during their inpatient stay. A unique patient identifier allows all records for each patient to be identified and linked together. Each episode's time span is defined with a start and finish date as well as being assigned an admission and discharge date for the whole period of the inpatient stay. Each episode will have up to 14 diagnoses coded using International Classification of Diseases 10th revision (ICD-10); and up to 12 procedures coded using the United Kingdom Tabular List of the Classification of Surgical Operations and Procedures (OPCS) (version OPCS4). This database has been linked to the Office of National Statistics (ONS) death register since 1998.

ing International Classification of Diseases 10th revision (ICD-10); and up to 12 procedures coded using the United Kingdom Tabular List of the Classification of Surgical Operations and Procedures (OPCS) (version OPCS4). This database has been linked to the Office of National Statistics (ONS) death register since 1998. Study Population Inclusion criteria All admissions older than 15 years (chosen to be consistent with the lower age limit of previous British Society of Gastroenterology (BSG) audits of mortality in gastrointestinal hemorrhage8,9), which had an ICD-10 code for upper gastrointestinal hemorrhage, with a date of hemorrhage between January 1, 1999, and December 31, 2007, were extracted. Data were available for 2008 to allow complete follow-up of mortality for admissions occurring in December 2007. Upper gastrointestinal hemorrhage was defined as an ICD-10 code that specifically implied either variceal gastrointestinal hemorrhage: esophageal varices with hemorrhage (I85.0) or nonvariceal hemorrhage: Mallory–Weiss syndrome (K22.6), esophageal hemorrhage (K22.8) acute, or chronic gastric ulcer with hemorrhage including perforation with hemorrhage (K25.0, K25.2, K25.4, K25.6), acute or chronic duodenal ulcer with hemorrhage including perforation with hemorrhage (K26.0, K26.2, K26.4, K26.6), acute or chronic peptic ulcer with hemorrhage including perforation with hemorrhage (K27.0, K27.2, K27.4, K27.6), acute or chronic gastrojejunal ulcer with hemorrhage including perforation with hemorrhage (K28.0, K28.2, K28.4, K28.6), hematemesis (K92.0), melena (K92.1), or unspecified gastrointestinal hemorrhage (K92.2). This ICD-10 code list has previously been used in hospital data.10 Subsequent readmissions with upper gastrointestinal hemorrhage were included in the study and recorded as a readmission. We performed 2 sensitivity analyses to assess the affect of inaccuracies in coding. First, to assess the effect of under-reporting, we expanded the definition for variceal hemorrhage to include all admissions coded for esophageal hemorrhage (K22.8) and then reassessed the trends in mortality. Second, to assess whether there was over-reporting of cases that might not be a genuine upper gastrointestinal hemorrhage, we analyzed separately those who had and those who did not have an intervention of upper gastrointestinal endoscopy recorded (as defined by an OPCS4 code for an endoscopic procedure of the upper gastrointestinal tract).

her there was over-reporting of cases that might not be a genuine upper gastrointestinal hemorrhage, we analyzed separately those who had and those who did not have an intervention of upper gastrointestinal endoscopy recorded (as defined by an OPCS4 code for an endoscopic procedure of the upper gastrointestinal tract). Exclusion criteria The study population was geographically limited to patients who were residents within England at the time of hospital admission. Admissions were excluded if they were coded with unspecified gastrointestinal hemorrhage (K92.2) and had a lower gastrointestinal endoscopy/diagnosis code but no upper gastrointestinal endoscopy code. Admissions were also excluded with the following: day case admission codes with no overnight stay (a majority of these admissions were for an outpatient endoscopy and would not have represented an acute presentation of hemorrhage but either a complication of endoscopy or a follow-up endoscopy to a previous bleed), invalid date codes as flagged by HES, date codes that were out of chronological order, invalid date of birth codes, invalid sex codes, or duplicate records for 1 episode. Outcome Short-term mortality was defined as a date of death within 28 days of the start of the recorded episode of upper gastrointestinal hemorrhage. This included deaths that occurred after discharge from hospital but within the 28 days. The date and fact of death were obtained from the ONS death register using a probability matching algorithm based on NHS number, date of birth, postcode, and sex.11

tart of the recorded episode of upper gastrointestinal hemorrhage. This included deaths that occurred after discharge from hospital but within the 28 days. The date and fact of death were obtained from the ONS death register using a probability matching algorithm based on NHS number, date of birth, postcode, and sex.11 Exposures The exposure of interest was defined as the year of upper gastrointestinal hemorrhage. Charlson index,12 sex, and age were assessed as potential confounders. The Charlson index was calculated for each upper gastrointestinal hemorrhage admission based on the diagnoses coded for all admissions up to and including the first upper gastrointestinal hemorrhage admission for each patient. The Charlson index is a validated comorbidity score that has been weighted to predict 1-year mortality. For analysis and reporting, it is combined into 3 groups: no comorbidity (0), a single comorbidity (1), and multiple or serious comorbidity (2). For analysis of variceal hemorrhage, the comorbidity of liver disease was excluded from the calculation of Charlson index because most variceal patients will have liver disease. The Charlson index has been adapted and validated for ICD-10 coding in administrative data13,14 and has previously been used in HES.15 As a sensitivity analysis, we also assessed the use of an alternative measure of comorbidity called the Elixhauser index16 that was derived to predict mortality during the inpatient stay. Although it has the potential to be a more appropriate measure for our study than the Charlson index, it has not been previously validated within HES, so it was not used for our primary analysis. The recorded age was grouped into age bands of 15–29 years, 30–59 years, 60–79 years, and older than 80 years. A further analysis assessed whether using a higher minimum age limit of 18 years altered the results. We calculated the length of inpatient stay as the number of days between admission and discharge dates. We defined admissions as either having a higher probability of being an acute bleed on admission (if an upper gastrointestinal hemorrhage was coded on the first episode in a nonelective admission) or as lower probability of being an acute bleed on admission with a higher probability of being an inpatient bleed (if the coding occurred after the first episode within a nonelective admission, or during an elective [nonemergency] admission). Hereafter, these are referred to, respectively, as acute admissions and inpatient bleeds.

robability of being an acute bleed on admission with a higher probability of being an inpatient bleed (if the coding occurred after the first episode within a nonelective admission, or during an elective [nonemergency] admission). Hereafter, these are referred to, respectively, as acute admissions and inpatient bleeds. To assess trends in diagnoses that were associated with a gastrointestinal hemorrhage code, we extracted additional diagnoses for gastritis/duodenitis, Mallory–Weiss syndrome, any peptic ulcer, gastric ulcer, duodenal ulcer, and malignancy.

robability of being an acute bleed on admission with a higher probability of being an inpatient bleed (if the coding occurred after the first episode within a nonelective admission, or during an elective [nonemergency] admission). Hereafter, these are referred to, respectively, as acute admissions and inpatient bleeds. To assess trends in diagnoses that were associated with a gastrointestinal hemorrhage code, we extracted additional diagnoses for gastritis/duodenitis, Mallory–Weiss syndrome, any peptic ulcer, gastric ulcer, duodenal ulcer, and malignancy. Statistical Analysis We analyzed variceal and nonvariceal hemorrhage admissions separately. After the exclusions described above, 28-day case fatalities were calculated by age group, sex, year, grouped Charlson index, and acute or inpatient hemorrhage. A case-control study analysis was carried out with cases defined as patients who had died by 28 days and controls as patients who were alive at 28 days. The primary exposure of interest was defined as year of upper gastrointestinal hemorrhage. A logistic regression model was constructed to adjust for the change in mortality over the study period by sex, age group, and Charlson index. Variables that changed the odds of mortality were judged to be confounders. We assessed whether there was a trend in mortality over time and whether this could be modelled as a linear trend using likelihood ratio tests. We also performed a secondary analysis comparing trends in mortality that occurred before discharge and trends in mortality that occurred after discharge. The calculation of postdischarge mortality excluded patients who had died as inpatients. In addition, to determine whether the changes in mortality varied for different ages, sex, and comorbidities, the model was also tested for interactions between each of the variables and year of bleed with likelihood ratio testing. If there was evidence against the null hypothesis of no interaction, stratified results were presented. The use of the a priori age groups was assessed against alternative groupings of 5-year age bands or age as a linear variable. All analysis was performed using Stata version 10 (StataCorp LP, College Station, TX).

g. If there was evidence against the null hypothesis of no interaction, stratified results were presented. The use of the a priori age groups was assessed against alternative groupings of 5-year age bands or age as a linear variable. All analysis was performed using Stata version 10 (StataCorp LP, College Station, TX). Results Study Population and Exclusions There were 516,153 upper gastrointestinal hemorrhage admissions identified after exclusions (shown in Figure 1) of which 501,471 (97%) were nonvariceal bleeds, and 14,682 (3%) were variceal bleeds. Mortality Ascertainment Seventy-four thousand nine hundred ninety-two deaths occurred within 28 days of the date of upper gastrointestinal hemorrhage, giving an overall case fatality rate of 14.5% (95% confidence interval [95% CI]: 14.4%–14.6%). Of these, 10,977 deaths (15%) occurred after discharge from hospital but within 28 days of hemorrhage. Only 312 (3%) of postdischarge deaths were coded as a subsequent hospital admission within the HES dataset.

al hemorrhage, giving an overall case fatality rate of 14.5% (95% confidence interval [95% CI]: 14.4%–14.6%). Of these, 10,977 deaths (15%) occurred after discharge from hospital but within 28 days of hemorrhage. Only 312 (3%) of postdischarge deaths were coded as a subsequent hospital admission within the HES dataset. Univariable Analysis The population characteristics for nonvariceal and variceal hemorrhage are shown in Table 1. The median age for nonvariceal bleeds was 71 years (interquartile range, 50–81 years) and, for variceal bleeds, was 55 years (interquartile range, 45–66 years). Forty-six percent of those presenting with nonvariceal hemorrhage had no comorbidity recorded, compared with 67% of those presenting with variceal hemorrhage after the exclusion of liver disease from the calculation of comorbidity. The population age structure and comorbidity varied over the study period (Figure 2) with a peak in the proportion of nonvariceal admissions over 80 years old in 2002. This matched the peak in case fatality in the same year (Table 1). There was a reduction over time in the proportion of those presenting with variceal hemorrhage who were less than 60 years old (Figure 2). The comorbidity for both groups increased over the study period. Median length of stay for nonvariceal hemorrhage was 4 days (interquartile range, 1–8 days) and for variceal hemorrhage was 7 days (interquartile range, 4–12 days). The length of stay reduced over the study period for nonvariceal hemorrhage from 4 (interquartile range, 2–8 days) to 3 (interquartile range, 1–6 days) (P < .001 nonparametric test for trend), but there was no reduction for variceal hemorrhage.

s) and for variceal hemorrhage was 7 days (interquartile range, 4–12 days). The length of stay reduced over the study period for nonvariceal hemorrhage from 4 (interquartile range, 2–8 days) to 3 (interquartile range, 1–6 days) (P < .001 nonparametric test for trend), but there was no reduction for variceal hemorrhage. Nonvariceal and Variceal Hemorrhage The overall 28-day case fatality following a nonvariceal hemorrhage admission was 14% and, following a variceal hemorrhage admission, was 23% (Table 1). From 1999 to 2007, the unadjusted 28-day mortality following nonvariceal hemorrhage reduced from 14.7% to 13.1% (unadjusted odds ratio [OR], 0.87; 95% CI: 0.84–0.90). The unadjusted mortality following variceal hemorrhage reduced from 24.6% to 20.9% (unadjusted OR, 0.81; (95% CI: 0.69–0.95). Acute Hemorrhage on Admission Compared With Inpatient Hemorrhage Twenty-eight-day mortality for an acute admission with hemorrhage reduced over the study period for nonvariceal hemorrhage from 11.3% to 9.3% (unadjusted OR, 0.81; 95% CI: 0.77–0.85) and, for variceal hemorrhage, from 21.3% to 17.3% (unadjusted OR, 0.77; 95% CI: 0.62–0.95). Twenty-eight-day mortality for cases with an inpatient hemorrhage also reduced over the study period, for nonvariceal hemorrhage from 20.0% to 18.4% (unadjusted OR, 0.91; 95% CI: 0.86–0.95) and, for variceal hemorrhage, from 32% to 29% (unadjusted OR, 0.88; 95% CI: 0.67–1.14).

(unadjusted OR, 0.77; 95% CI: 0.62–0.95). Twenty-eight-day mortality for cases with an inpatient hemorrhage also reduced over the study period, for nonvariceal hemorrhage from 20.0% to 18.4% (unadjusted OR, 0.91; 95% CI: 0.86–0.95) and, for variceal hemorrhage, from 32% to 29% (unadjusted OR, 0.88; 95% CI: 0.67–1.14). Multivariate Analysis The odds of mortality for each year were altered when adjusted separately for each of the potential confounders of age, sex, and Charlson index. The slight peak in mortality in 2002 was removed when adjusting for the increase in age in 2002. The use of alternative groupings for age did not alter the estimates. An alternative minimum age limit of 18 years did not alter the findings of the analysis for mortality. Adjusting for increases in comorbidity had the largest effect on the reduction in mortality. The multivariate model adjusting for all these variables is shown in Table 2. Age and comorbidity were stronger confounders for nonvariceal than variceal hemorrhage.

ars did not alter the findings of the analysis for mortality. Adjusting for increases in comorbidity had the largest effect on the reduction in mortality. The multivariate model adjusting for all these variables is shown in Table 2. Age and comorbidity were stronger confounders for nonvariceal than variceal hemorrhage. There was evidence of a linear trend in mortality over time, for both nonvariceal hemorrhage and variceal hemorrhage (P < .001), and there was minimal evidence to suggest that a linear model was inappropriate for the data (test for departure from a linear trend; nonvariceal hemorrhage, P = .061; variceal hemorrhage, P = .94). The adjusted average annual reduction in odds of mortality for nonvariceal hemorrhage was 2.5% (average annual OR, 0.97; 95% CI: 0.97–0.98) and, for variceal hemorrhage, was 3.5% (average annual OR, 0.96; 95% CI: 0.95–0.98). Assessing age, sex, and comorbidity adjusted trends following the diagnoses of gastritis/duodenitis, Mallory–Weiss syndrome, any peptic ulcer, gastric ulcer, duodenal ulcer, or malignancy associated with nonvariceal hemorrhage found that there were similar reductions in mortality following all these diagnoses (see Table 3). A sensitivity analysis was conducted including esophageal hemorrhage codes (K22.8) as a variceal hemorrhage admission, and this estimated an annual reduction in odds of mortality of 3.6% (average annual OR, 0.96; 95% CI: 0.95–0.98). The second sensitivity analysis found a similar reduction in nonvariceal hemorrhage admissions who had an endoscopy recorded (average annual OR, 0.97; 95% CI: 0.96–0.97) to those who did not have an endoscopy recorded (average annual OR, 0.96; 95% CI: 0.96–0.97). This was also the case for variceal hemorrhage, although because only a few cases did not have an endoscopy, there was greater uncertainty (with endoscopy: average annual OR, 0.98; 95% CI: 0.96–0.99; without endoscopy: average annual OR, 0.95, 95% CI: 0.92–0.98). The third sensitivity analysis used the Elixhauser index to adjust for comorbidity, and this showed a slightly increased average annual reduction compared with using the Charlson index to adjust for comorbidity (nonvariceal hemorrhage OR, 0.96; 95% CI: 0.96–0.97). However, the overall model with the Elixhauser index did not have as good a fit to the data as when the Charlson index was used to adjust for comorbidity.

a slightly increased average annual reduction compared with using the Charlson index to adjust for comorbidity (nonvariceal hemorrhage OR, 0.96; 95% CI: 0.96–0.97). However, the overall model with the Elixhauser index did not have as good a fit to the data as when the Charlson index was used to adjust for comorbidity. Reanalyzing the age, sex, and comorbidity adjusted trends for mortality only occurring before discharge demonstrated the same reduction in inpatient mortality as in the main analysis (nonvariceal average annual adjusted mortality OR, 0.97; 95% CI: 0.97–0.98). However, the mortality after discharge increased slightly (nonvariceal average annual adjusted mortality OR, 1.02; 95% CI: 1.02–1.03). Further analyses for interactions demonstrated different time trends for different ages and different levels of comorbidity for nonvariceal hemorrhage (likelihood ratio tests for interactions of both age and comorbidity with year, P < .001) but not for variceal hemorrhage (year and age, P = .29; year and comorbidity, P = .67). Consequently, the age-specific stratum average annual changes in odds of mortality for nonvariceal hemorrhage are presented in Table 4. The annual improvement in odds of mortality was minimal for those presenting 80 years and older compared with all the other age groups. Further stratifying the model by age and comorbidity (Table 5) demonstrated that, within each age-specific stratum, the improvement in mortality did not differ by the level of comorbidity. Therefore, the final model of a linear trend in 28-day mortality for nonvariceal hemorrhage is the model shown in Table 4, with confounding by comorbidity adjusted for by logistic regression and effect modification demonstrated by stratifying the results by age. The final model of a linear trend in 28-day mortality for variceal hemorrhage demonstrated only confounding by both comorbidity and age with no effect modification.

own in Table 4, with confounding by comorbidity adjusted for by logistic regression and effect modification demonstrated by stratifying the results by age. The final model of a linear trend in 28-day mortality for variceal hemorrhage demonstrated only confounding by both comorbidity and age with no effect modification. Discussion The failure of previous studies to demonstrate improvements in mortality after upper gastrointestinal hemorrhage at the population level calls into question the value of therapeutic changes that are of proven benefit to individuals. In an increasingly challenging economic environment, clinicians will need to be able to demonstrate that increased therapeutic expenditure really does bring benefits. That 28-day mortality for equivalent patients, following hospital admission for both nonvariceal and variceal upper gastrointestinal hemorrhage, has reduced by 2% and 3%, respectively, year on year in England over the period 1999 to 2007 is therefore of great importance. The demonstration that this can be shown through the analysis of routinely collected data may be of great value in the assessment of other conditions.

ariceal upper gastrointestinal hemorrhage, has reduced by 2% and 3%, respectively, year on year in England over the period 1999 to 2007 is therefore of great importance. The demonstration that this can be shown through the analysis of routinely collected data may be of great value in the assessment of other conditions. Strengths and Limitations When, as in this case, a study's findings differ from the previous literature, we must ask whether this is because the current or previous studies were in error or whether they are in reality observing different things. The data source chosen for our study provides key advantages. The study is the largest to date of mortality after hospital admission for gastrointestinal hemorrhage and therefore has power to demonstrate trends that would be missed in smaller studies. It also has power to demonstrate variations in trends between subgroups of the population such as the smaller reduction in mortality in those over 80 years old with nonvariceal hemorrhage. The provision within the dataset of information on the previously suggested confounders of age and comorbidity is also of great benefit and has allowed us to clearly show and correct for this confounding.

f the population such as the smaller reduction in mortality in those over 80 years old with nonvariceal hemorrhage. The provision within the dataset of information on the previously suggested confounders of age and comorbidity is also of great benefit and has allowed us to clearly show and correct for this confounding. Another key advantage of the current study is the linkage of clinical data with the ONS death register, ensuring that almost all deaths are captured in the study population. Hospital admission data only capture deaths occurring before discharge, which we found to be 86% of the deaths occurring within 28 days. Studies without such linkage will have missed a proportion of these deaths because postdischarge deaths will have been difficult to capture. Furthermore, any change in this capture over time may have biased results. The linkage used in the current study, depending as it does on probability matching, still leaves potential for some underestimation of mortality, but the robustness of the linkage coupled with its uniform methodology throughout the study period mean that bias because of this is unlikely to have occurred. The reduction in length of stay over the course of the study further emphasises the importance of identifying deaths following discharge to accurately calculate trends in mortality. The slight increase in postdischarge mortality might imply that the observed earlier discharge of patients was inappropriate; however, if management in hospital was no longer of benefit to a patient who is dying, then discharge might well be the most appropriate decision. The observed trends might therefore indicate a shift of unavoidable in-hospital mortality into the postdischarge period.

ved earlier discharge of patients was inappropriate; however, if management in hospital was no longer of benefit to a patient who is dying, then discharge might well be the most appropriate decision. The observed trends might therefore indicate a shift of unavoidable in-hospital mortality into the postdischarge period. Patients who died in the emergency department before admission for endoscopy were not included in our study because hospital admissions data contain information only on admitted patients. However, because acute admission to the hospital for all upper gastrointestinal hemorrhages was standard practice within England, the admissions data will have captured almost all other relevant bleed presentations. We excluded patients who had a nonspecific code for gastrointestinal hemorrhage with a colonoscopy but no gastroscopy, and it is possible that these could have had an upper gastrointestinal bleed if they had died before a planned gastroscopy. However, this would be unlikely because usual practice would be to perform a gastroscopy before colonoscopy because of the easier access and greater therapeutic potential of gastroscopy.

troscopy, and it is possible that these could have had an upper gastrointestinal bleed if they had died before a planned gastroscopy. However, this would be unlikely because usual practice would be to perform a gastroscopy before colonoscopy because of the easier access and greater therapeutic potential of gastroscopy. There have been concerns about the accuracy of routine hospital admissions coding, in particular the coding of specific operations and the ascertainment of death for generating mortality rates for specific hospitals. However, a systematic review found a 91% median accuracy in diagnostic coding prior to our study period, and the most recent audit of selected samples of UK hospital data confirmed accuracy approaching 90%.17 Other comparisons of procedure coding have reported similar or higher rates of coding in the HES database compared with specialist clinical databases,18,19 and, with specific regard to upper gastrointestinal hemorrhage, the incidence of peptic ulcer hemorrhage in the HES data from 1992 to 1995 has been shown to be comparable with the 1993 regional BSG audit (32 vs 29 per 100,000 per year, respectively). Furthermore, by choosing our study period, we have ensured no systematic changes in coding because the ICD-10 coding system has been in continuous use in HES from 1995 to present. This, of course, does not exclude variation in rates of coding over the study period affecting our estimates. For example, if the potential error in coding was systematically changing over time with increased coding of patients' comorbidity rather than patients having more comorbidity, then clearly that could bias our results. However, the different trends in comorbidity for variceal and nonvariceal bleed admissions and different trends in mortality in different age and comorbidity strata suggest that there was no systematic change in comorbidity coding over the time period of our study. Under-reporting of the comorbidities in the Charlson index may have resulted in incomplete adjustment for comorbidity. However, although the alternative Elixhauser index assessed almost twice the number of comorbidities, it did not alter the adjustment of comorbidity in the model. Comorbidity adjustment by either index increased the magnitude of the mortality reduction, and, therefore, any residual confounding in this regard would only, we believe, cause an underestimate of the real mortality trend in our study.

umber of comorbidities, it did not alter the adjustment of comorbidity in the model. Comorbidity adjustment by either index increased the magnitude of the mortality reduction, and, therefore, any residual confounding in this regard would only, we believe, cause an underestimate of the real mortality trend in our study. Other Studies A PubMed search, to October 2010, found the largest comparable population-based study for nonvariceal hemorrhage mortality trends used a Canadian hospital discharge database with ICD-10 and ICD-9 codes. However, it identified less than one-third of the number of bleeds used for this study (n = 142,363) and was not able to identify a reduction in case fatality for nonvariceal hemorrhage between 1993 and 2003.3 The researchers adjusted for changes in age but not for changes in comorbidity. They also only identified deaths that occurred before discharge. The low mortality identified in this study (3.5%) is similar to other North American20 and Mediterranean1,21 studies but is much lower than other European studies.2,22,23 However, a study of Medicare patients in the United States found that the proportion being managed as outpatients varied between states from 18.6% to 45.3%.24 These differences in practice would lead to differences in inpatient study populations and confound comparisons with countries such as England where outpatient management is not routine.

are patients in the United States found that the proportion being managed as outpatients varied between states from 18.6% to 45.3%.24 These differences in practice would lead to differences in inpatient study populations and confound comparisons with countries such as England where outpatient management is not routine. Although the most recent report from the US National Inpatient Sample showed a 23% reduction in upper gastrointestinal hemorrhage mortality from 1998 to 2006 (n = unreported because only extrapolated estimates from the 20% sample are provided),20 this was a global figure for the reduction seen at the end of the study rather than year on year, and it did not distinguish variceal and nonvariceal hemorrhage. Another report from the US National Inpatient Sample noted an adjusted reduction in variceal hemorrhage from 18% to 12%.25 However, in the study period of both these reports, the number of states in the sampling frame almost doubled from 22 to 40. The reports therefore compare different populations from each time period, and, although a number of weighting procedures are used, the estimates remain susceptible to selection bias.

% to 12%.25 However, in the study period of both these reports, the number of states in the sampling frame almost doubled from 22 to 40. The reports therefore compare different populations from each time period, and, although a number of weighting procedures are used, the estimates remain susceptible to selection bias. One smaller study from Wales (n = 24,421) used the same ICD-10 definitions as our study and also found an overall reduction in case fatality but did not report variceal and nonvariceal hemorrhage mortality trends separately or trends in different age and comorbidity strata.10 Other nonvariceal hemorrhage studies from Spain (n = 17,663),1 The Netherlands (n = 1720),2 Greece (n = 1304),21 France (n = 1165),23 and Italy (n = 1126)22 did not identify reductions in nonvariceal inpatient mortality. Although these were large studies, they may have been underpowered to detect a change, and none of them adjusted the trends in case fatality for changes in comorbidity. Furthermore, none of these studies identified deaths that occurred after discharge. The remainder of the studies contained less than 1000 patients and therefore could not provide accurate estimates of mortality trends.

to detect a change, and none of them adjusted the trends in case fatality for changes in comorbidity. Furthermore, none of these studies identified deaths that occurred after discharge. The remainder of the studies contained less than 1000 patients and therefore could not provide accurate estimates of mortality trends. For variceal hemorrhage, the largest study on mortality after hospitalization because of varices (n = 12,281; compared with 14,682 for this study) did not differentiate between hemorrhage and nonhemorrhage admissions.26 The next largest study (n = 1475) compared variceal hemorrhage mortality between control groups in randomized trials 1960–2000 and showed a similar reduction in mortality.27 However, these control groups were from different geographical populations with different study exclusion criteria. Comparisons were therefore susceptible to selection bias. Other studies of trends in variceal hemorrhage mortality contained less than 1000 patients.

a perhaps suggest that improvement in standard nonendoscopic care has led to improved survival, such as the routine administration of intravenous proton pump inhibitor infusions, the routine use of risk scoring, the implementation of standardized clinical guidelines, and the subsequent local auditing of practice.4,5,30 In conclusion, contrary to previous smaller studies, we have found an encouraging substantial improvement in mortality following hospital admission for upper gastrointestinal hemorrhage. Our study shows that this is partially obscured by changes in age and comorbidity and that the improvements are less marked in the elderly individuals in a manner not explained by comorbidity. We believe that this improvement reflects the effect of changes in the care of gastrointestinal hemorrhage over the last decade, but it also suggests the need to focus our ongoing attention on the elderly individuals who may not yet have benefited to the maximum possible extent from these changes. The recent demonstration of under-utilization of endoscopic techniques in the United Kingdom, coupled with the fact that other interventions such as use of proton pump inhibitors are more readily available to the admitting physician worldwide, may suggest areas that could be further improved.4,5,31–33 Video Abstract Video Abstract Acknowledgments The funding bodies had no role in the collection, analysis, or interpretation of the data. Conflicts of interest The authors disclose the following: Dr Tim R. Card is married to an employee of AstraZeneca. The remaining authors disclose no conflicts.

In conclusion, contrary to previous smaller studies, we have found an encouraging substantial improvement in mortality following hospital admission for upper gastrointestinal hemorrhage. Our study shows that this is partially obscured by changes in age and comorbidity and that the improvements are less marked in the elderly individuals in a manner not explained by comorbidity. We believe that this improvement reflects the effect of changes in the care of gastrointestinal hemorrhage over the last decade, but it also suggests the need to focus our ongoing attention on the elderly individuals who may not yet have benefited to the maximum possible extent from these changes. The recent demonstration of under-utilization of endoscopic techniques in the United Kingdom, coupled with the fact that other interventions such as use of proton pump inhibitors are more readily available to the admitting physician worldwide, may suggest areas that could be further improved.4,5,31–33 Video Abstract Video Abstract Acknowledgments The funding bodies had no role in the collection, analysis, or interpretation of the data. Conflicts of interest The authors disclose the following: Dr Tim R. Card is married to an employee of AstraZeneca. The remaining authors disclose no conflicts. Funding Supported by an MRC population health scientist fellowship (to C.C.), by a Walport senior lectureship (to T.C.), and by an NIHR clinician scientist fellowship (to J.W.). View this article's video abstract atwww.gastrojournal.org. Figure 1 Flowchart of exclusions from study population.

Conflicts of interest The authors disclose the following: Dr Tim R. Card is married to an employee of AstraZeneca. The remaining authors disclose no conflicts. Funding Supported by an MRC population health scientist fellowship (to C.C.), by a Walport senior lectureship (to T.C.), and by an NIHR clinician scientist fellowship (to J.W.). View this article's video abstract atwww.gastrojournal.org. Figure 1 Flowchart of exclusions from study population. Figure 2 Trends in age and comorbidity measured by grouped Charlson index (percentage of population shown). (A) Percentage of nonvariceal hemorrhage patients in each age band. (B) Percentage of nonvariceal hemorrhage patients in each comorbidity group. (C) Percentage of variceal hemorrhage patients in each age band. (D) Percentage of variceal hemorrhage patients in each comorbidity group. Table 1 Population Characteristics

Figure 2 Trends in age and comorbidity measured by grouped Charlson index (percentage of population shown). (A) Percentage of nonvariceal hemorrhage patients in each age band. (B) Percentage of nonvariceal hemorrhage patients in each comorbidity group. (C) Percentage of variceal hemorrhage patients in each age band. (D) Percentage of variceal hemorrhage patients in each comorbidity group. Table 1 Population Characteristics Nonvariceal bleed admissions Variceal bleed admissions Number of admissions (n) Percentage of all admissions 28-Day case fatality (%) Number of admissions (n) Percentage of all admissions 28-Day case fatality (%) Year 1999 51,843 10.3 14.7 1559 10.6 24.6 2000 53,206 10.6 14.8 1592 10.8 25.1 2001 53,268 10.6 14.9 1496 10.2 25.0 2002 53,735 10.7 14.9 1581 10.8 24.2 2003 55,656 11.1 14.7 1619 11.0 23.6 2004 57,450 11.5 14.1 1768 12.0 22.3 2005 59,362 11.8 13.9 1612 11.0 21.7 2006 58,737 11.7 13.7 1736 11.8 20.7 2007 58,214 11.6 13.1 1719 11.7 20.9 Total 501,471 100.0 14.3 14,682 100.0 23.1 Sex Male 276,304 55.1 13.3 9565 65.1 23.0 Female 225,167 44.9 15.5 5117 34.9 23.2 Age, y <30 39,973 8.0 0.5 375 2.6 10.7 30 to 59 135,507 27.0 5.5 8749 59.6 21.2 60 to 79 174,181 34.7 15.1 4688 31.9 25.9 ≥80 151,810 30.3 24.8 870 5.9 31.3 Charlson index No comorbidity 229,941 45.9 6.8 9825 66.9 21.6 Single comorbidity 150,004 29.9 13.6 3832 26.1 25.2 Multiple or serious comorbidity 121,526 24.2 29.2 1025 7.0 29.5 Acute hemorrhage on admission or inpatient Acute hemorrhage on admission 295,887 59.0 10.5 10,176 69.3 20.1 Inpatient bleed 205,584 41.0 19.7 4506 30.7 29.8 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total.

e currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. Table 2 Logistic Regression Model Predicting 28-Day Mortality

e currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. Table 2 Logistic Regression Model Predicting 28-Day Mortality Nonvariceal hemorrhage Variceal hemorrhage Unadjusted odds ratio Adjusted odds ratioa 95% Confidence interval Unadjusted odds ratio Adjusted odds ratioa 95% Confidence interval Year of presentation 1999 1.00 1.00 1.00 1.00 2000 1.00 0.98 0.94–1.01 1.02 1.02 0.87–1.20 2001 1.01 0.97 0.93–1.00 1.02 1.02 0.86–1.20 2002 1.01 0.95 0.92–0.99 0.98 0.98 0.83–1.15 2003 0.99 0.94 0.90–0.97 0.94 0.95 0.80–1.11 2004 0.95 0.90 0.86–0.93 0.88 0.88 0.75–1.03 2005 0.93 0.89 0.86–0.92 0.85 0.83 0.70–0.98 2006 0.92 0.85 0.82–0.88 0.80 0.79 0.67–0.94 2007 0.87 0.80 0.77–0.83 0.81 0.80 0.67–0.94 Age, y <30 1.00 1.00 1.00 1.00 30–59 10.09 7.22 6.37–8.19 1.93 1.92 1.44–2.55 60–79 30.04 16.80 14.84–19.02 2.51 2.37 1.77–3.17 ≥80 55.62 34.14 30.15–38.65 3.26 3.05 2.22–4.20 Sex Male 1.00 1.00 1.00 1.00 Female 1.20 1.01 0.99–1.03 1.01 0.96 0.88–1.04 Charlson index No comorbidity 1.00 1.00 1.00 1.00 Single comorbidity 2.16 1.70 1.66–1.74 0.99 1.17 1.07–1.27 Multiple or serious comorbidity 5.64 4.37 4.28–4.47 1.31 1.37 1.18–1.58 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total.

e currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for all variables in Table. Table 3 Trends in 28-Day Mortality for Diagnoses Associated With an Upper Gastrointestinal Hemorrhage Diagnosis associated with upper gastrointestinal hemorrhage Adjusted odds ratioa 95% confidence intervals Change in mortality for an increment of 1 yearb No specific diagnosis 0.97 0.97–0.98 Gastritis/duodenitis 0.96 0.94–0.98 Mallory–Weiss syndrome 0.96 0.95–0.97 Any peptic ulcer 0.96 0.93–0.99 Gastric ulcer 0.94 0.93–0.95 Duodenal ulcer 0.96 0.95–0.97 Malignancy 0.95 0.95–0.96 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for age, sex, and comorbidity by Charlson index. b Year as a continuous variable. Table 4 Age Stratified Logistic Regression Model Predicting 28-Day Mortality for Nonvariceal Hemorrhage

Diagnosis associated with upper gastrointestinal hemorrhage Adjusted odds ratioa 95% confidence intervals Change in mortality for an increment of 1 yearb No specific diagnosis 0.97 0.97–0.98 Gastritis/duodenitis 0.96 0.94–0.98 Mallory–Weiss syndrome 0.96 0.95–0.97 Any peptic ulcer 0.96 0.93–0.99 Gastric ulcer 0.94 0.93–0.95 Duodenal ulcer 0.96 0.95–0.97 Malignancy 0.95 0.95–0.96 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for age, sex, and comorbidity by Charlson index. b Year as a continuous variable. Table 4 Age Stratified Logistic Regression Model Predicting 28-Day Mortality for Nonvariceal Hemorrhage Adjusted odds ratioa 95% Confidence interval Change in mortality for an increment of 1 yb <30 y 0.92 0.88–0.97 30–59 y 0.97 0.96–0.97 60–79 y 0.97 0.96–0.97 ≥80 y 0.99 0.98–0.99 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for comorbidity by Charlson index and sex. b Year as a continuous variable.

Adjusted odds ratioa 95% Confidence interval Change in mortality for an increment of 1 yb <30 y 0.92 0.88–0.97 30–59 y 0.97 0.96–0.97 60–79 y 0.97 0.96–0.97 ≥80 y 0.99 0.98–0.99 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for comorbidity by Charlson index and sex. b Year as a continuous variable. Table 5 Age and Comorbidity Stratified Logistic Regression Model Predicting 28-Day Mortality Age, y Charlson index Adjusted odds ratioa 95% Confidence interval Change in mortality for an increment of 1 yb <80 0 0.96 0.95–0.97 1 0.96 0.95–0.97 2 0.95 0.95–0.96 ≥80 0 1.00 0.99–1.01 1 0.99 0.98–0.99 2 0.98 0.97–0.99 NOTE. Linked HES/ONS mortality records are currently provided on a provisional basis. An issue has arisen whereby a small number of mortality records may have been incorrectly rejected. The algorithm that links HES to ONS mortality is currently being amended to rectify this issue, which affects approximately 1000 mortality records or about 0.02% of the total. a Adjusted for sex. b Odds ratio for year as a continuous variable.

See Covering the Cover synopsis on page 1327. Helicobacter pylori infection, nonsteroidal anti-inflammatory medications (NSAIDs), and aspirin are believed to be the main causes of nonvariceal upper gastrointestinal bleeding,1 and with the discovery of proton pump inhibitors (PPIs) and H pylori eradication therapy, the burden of peptic ulcer disease has been decreasing.2 Despite this, upper gastrointestinal hemorrhage remains the most common acute severe medical admission for gastroenterology,3,4 and its incidence in population-based studies remains virtually unchanged.5,6 This suggests that other (previously unidentified) risk factors are contributing to its population burden.

asing.2 Despite this, upper gastrointestinal hemorrhage remains the most common acute severe medical admission for gastroenterology,3,4 and its incidence in population-based studies remains virtually unchanged.5,6 This suggests that other (previously unidentified) risk factors are contributing to its population burden. Historically, nongastrointestinal comorbidity was believed to be associated with stress ulceration7 but, currently, the role of comorbidity in the etiology of gastrointestinal bleeding (GIB) is not recognized apart from in severe illness; for example, sicker cirrhotic patients are known to have an increased risk of variceal bleeding,8 and sicker patients in intensive therapy units (ITUs) have an increased risk of nonvariceal bleeding.9 However, as the proportion of bleed patients with comorbidity has increased during the last decade,5 we wondered if exposure to less severe but chronic comorbidity could itself be responsible for the persisting incidence of bleeding. Outside of ITU though, the effect of comorbidity has only been assessed as a confounder in studies that focused on the effect of medications on gastrointestinal bleeds.10 Although these studies do support a role for comorbidity, they do not allow us to understand whether it is an important independent contributor to the persisting burden of upper GIB.

of comorbidity has only been assessed as a confounder in studies that focused on the effect of medications on gastrointestinal bleeds.10 Although these studies do support a role for comorbidity, they do not allow us to understand whether it is an important independent contributor to the persisting burden of upper GIB. We have therefore conducted a study aimed primarily at assessing whether comorbidity might have an important role in the etiology of upper GIB. To do this we have conducted a case-control study and formed a model fully corrected for known measured risk factors of upper GIB. We have then calculated the additional explanatory effect of adding comorbidity to our model to understand its effect on bleeding incidence in the general population. Methods Study Design We conducted a matched case control study.

We have therefore conducted a study aimed primarily at assessing whether comorbidity might have an important role in the etiology of upper GIB. To do this we have conducted a case-control study and formed a model fully corrected for known measured risk factors of upper GIB. We have then calculated the additional explanatory effect of adding comorbidity to our model to understand its effect on bleeding incidence in the general population. Methods Study Design We conducted a matched case control study. Data To provide the detailed longitudinal data and necessary power for this study, we have used the recently linked English Hospital Episodes Statistics data (secondary care data) and General Practice Research Database (GPRD) (primary care data). Because of the comprehensive English primary care system, the population registered to the GPRD is representative of the general English population.11 The data are subject to quality checks and a practice's data are only used when they are of high enough quality to be used in research, at these times the data are said to be “up to research standard.”12 The GPRD has been extensively validated for a wide range of diagnoses, with a mean positive predictive value of 89%.13 Ethical approval for this study was obtained from the Independent Scientific Advisory Committee for Medicines and Healthcare products Regulatory Agency database research. Fifty-one percent of English practices in GPRD have consented to record level linkage of their population to Hospital Episodes Statistics. This records all hospital admissions from the population registered to one of the linked primary care practices contributing to the GPRD. For this study, the linked dataset was available between April 1, 1997 and August 31, 2010.

ave consented to record level linkage of their population to Hospital Episodes Statistics. This records all hospital admissions from the population registered to one of the linked primary care practices contributing to the GPRD. For this study, the linked dataset was available between April 1, 1997 and August 31, 2010. Case Definition We have previously published the codes and methods used to define upper gastrointestinal bleeds in this study.14 In brief, we selected as exposed all patients with a first nonvariceal upper gastrointestinal bleed. A bleed was defined by a specific code for an upper gastrointestinal nonvariceal bleed in either primary or secondary care who had a supporting code in the linked dataset (defined as a likely symptom, cause, therapy, investigation, or outcome of upper gastrointestinal hemorrhage). Variceal bleeds or nonspecific gastrointestinal bleed codes with either a lower gastrointestinal diagnosis or procedure were excluded. Further exclusions were temporary patients (patients not registered permanently at a GPRD primary care practice, who might just be visiting the area of the practice briefly, and who are therefore not part of the GPRD's underlying population), children younger than 16 years old, cases with invalid date codes, or cases outside the up-to-research-standard observed time periods. Patients were required to be registered with the primary care practice for at least 3 months before an upper gastrointestinal bleed event to avoid including prevalent cases that might have been coded at the initial registration consultation. Only the first event for each patient was included. We have previously demonstrated that this selection strategy minimizes selection bias in studies of upper GIB in these data.14 A secondary analysis was then stratified by whether the defining bleed code or supporting code specifically referred to a peptic ulcer (Read codes J11 to J14 or International Classification of Diseases, 10th Revision codes K25–K28). The Read codes had high positive predictive values (>95%) for peptic ulcers and upper gastrointestinal complications when validated in English primary care routine records.15,16

code specifically referred to a peptic ulcer (Read codes J11 to J14 or International Classification of Diseases, 10th Revision codes K25–K28). The Read codes had high positive predictive values (>95%) for peptic ulcers and upper gastrointestinal complications when validated in English primary care routine records.15,16 Matched Controls Each case was age (±5 years) and sex matched without replacement to 5 controls selected randomly who were alive at the time of the gastrointestinal bleed and registered to the same primary care practice. Controls were required to have been registered with the primary care practice for at least 3 months before the match date to be consistent with the definition for cases.

replacement to 5 controls selected randomly who were alive at the time of the gastrointestinal bleed and registered to the same primary care practice. Controls were required to have been registered with the primary care practice for at least 3 months before the match date to be consistent with the definition for cases. Exposures Potential final common causal pathways of an upper gastrointestinal bleed were defined a priori for erosions/ulceration, varices, angiodysplasia, fistula/trauma and coagulopathy, and code lists derived for diagnoses and medications that might be associated with each pathway based on published literature (Figure 1). Although variceal bleeds were excluded from the cases and controls, cirrhosis itself was included as a risk factor, as cirrhotic patients can have nonvariceal bleeds. Medication risk factors were included if there was a coded prescription within the year before the admission. Exposures coded within 2 months of the admission date were excluded to avoid identifying events and prescriptions related to the actual bleed event. PPIs were included as an indicator of physicians' judgement of the risk of upper gastrointestinal hemorrhage that was not captured by other measured risk factors. Alcohol consumption was classified as either nondrinker, alcohol mentioned, ex–alcohol dependency, alcohol excess, alcohol complications, and missing. Smoking was classified as never smoked, current smoker, ex-smoker, and missing. Cirrhosis was classified as uncomplicated, with varices, with ascites, or with encephalopathy or liver failure coded. All other exposures were binary variables.

ned, ex–alcohol dependency, alcohol excess, alcohol complications, and missing. Smoking was classified as never smoked, current smoker, ex-smoker, and missing. Cirrhosis was classified as uncomplicated, with varices, with ascites, or with encephalopathy or liver failure coded. All other exposures were binary variables. Comorbidity Comorbidity was defined using the Charlson Index.17 This is a well-validated weighted comorbidity score derived from unselected hospital admissions that predicts 1-year mortality after hospital discharge. It has since been used in many contexts and has repeatedly measured the burden of comorbidity reliably. The original article demonstrated a graded increase in the risk in mortality associated with an increase in total score. The different comorbidities were assigned weights of 1, 2, 3, and 6, depending on their association with mortality. Where a graded effect was observed within a disease, for example, in diabetes or malignancy, these diseases were further stratified according to their severity. The conditions included in the original score (in order of weighting) were myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, mild liver disease, diabetes, hemiplegia, moderate or severe renal disease, diabetes with end organ damage, leukemia, lymphoma, moderate or severe liver disease, metastatic solid tumor, and acquired immunodeficiency syndrome. For our study, any codes already used to define risk factors of upper GIB in Figure 1 were excluded when calculating the index, ie, peptic ulcer and cirrhosis codes. For clarity in reporting in the tables, the index was summarized as no comorbidity (Charlson Index = 0), single comorbidity (Charlson Index = 1), and multiple or severe comorbidity (Charlson Index = 2).

risk factors of upper GIB in Figure 1 were excluded when calculating the index, ie, peptic ulcer and cirrhosis codes. For clarity in reporting in the tables, the index was summarized as no comorbidity (Charlson Index = 0), single comorbidity (Charlson Index = 1), and multiple or severe comorbidity (Charlson Index = 2). Analysis Unadjusted analysis Unadjusted odds ratios (ORs) were calculated for each exposure using conditional logistic regression to allow for the matched study design. Multivariable analysis Adjusted ORs for each exposure of interest were calculated with conditional logistic regression adjusting for all exposures in addition to age, PPI use, and previous gastrointestinal procedures. As calendar year, sex, and primary care practice were precisely matched on in the controls, it was not necessary to include them in the model. Comorbidity was added last, and its association with bleeding tested using a likelihood ratio test. The variance inflation factor (a measure of the increase in model variance due to correlation between variables) was calculated for each exposure of interest to assess the effect of correlation between variables. All exposures with a variance inflation factor >5 were excluded from the final conditional logistic regression model.18 The final model was then stratified into cases with a recording of peptic ulcer and those without.

) was calculated for each exposure of interest to assess the effect of correlation between variables. All exposures with a variance inflation factor >5 were excluded from the final conditional logistic regression model.18 The final model was then stratified into cases with a recording of peptic ulcer and those without. Sequential (or Extra) Population Attributable Fractions Sequential (or extra) population attributable fractions (PAFs) were calculated for each exposure, using the prevalence among the cases and the respective coefficients from the conditional logistic regression model.19 Sequential PAFs differ from the standard adjusted PAFs that are usually presented. They are calculated by estimating the additional proportion of cases attributable to each exposure, after removing the proportion of cases already attributed to the combined effect of all other exposures in the model. The final model was then stratified into cases with a recording of peptic ulcer and those without. All analysis was performed using Stata software, version 12 (StataCorp LP, College Station, TX).

osure, after removing the proportion of cases already attributed to the combined effect of all other exposures in the model. The final model was then stratified into cases with a recording of peptic ulcer and those without. All analysis was performed using Stata software, version 12 (StataCorp LP, College Station, TX). Sensitivity Analyses Previous studies of risk factor medications, such as NSAIDs,20 have been conducted in study populations that excluded patients with known risk factors for GIB. To allow comparisons with these, we re-estimated the crude ORs for each of the risk factor medications after excluding any cases and their controls with nonmedication bleed risk factors. To assess the effect of the choice of the exposure exclusion time window before the bleed event on the effect of NSAIDs, we also re-estimated a model that included NSAID use up to 30 days before the index date.

e risk factor medications after excluding any cases and their controls with nonmedication bleed risk factors. To assess the effect of the choice of the exposure exclusion time window before the bleed event on the effect of NSAIDs, we also re-estimated a model that included NSAID use up to 30 days before the index date. Two additional sensitivity analyses were performed to assess the effect of potential under-reporting. First the analysis was restricted to those older than 65 years old and who were eligible for free prescriptions, to assess the effect of potential under-reporting of nonprescribed NSAID use. Secondly, multiple imputation was used to re-estimate the association with comorbidity by imputing missing values for alcohol and smoking status. Alcohol and smoking were categorised as binary exposures of excess alcohol or current smoking to fit the logistic regression imputation model. All previously extracted exposures were used in the imputation model with addition of the socioeconomic status, and 20 sets of imputations were calculated. Socioeconomic status was measured by the Index of Multiple Deprivation quintiles obtained from linked Office of National Statistics data. Finally, to assess the effect of using the aggregated and weighted Charlson Index, the model was re-estimated to assess the effect of the individual component comorbidities from the Charlson Index.

Two additional sensitivity analyses were performed to assess the effect of potential under-reporting. First the analysis was restricted to those older than 65 years old and who were eligible for free prescriptions, to assess the effect of potential under-reporting of nonprescribed NSAID use. Secondly, multiple imputation was used to re-estimate the association with comorbidity by imputing missing values for alcohol and smoking status. Alcohol and smoking were categorised as binary exposures of excess alcohol or current smoking to fit the logistic regression imputation model. All previously extracted exposures were used in the imputation model with addition of the socioeconomic status, and 20 sets of imputations were calculated. Socioeconomic status was measured by the Index of Multiple Deprivation quintiles obtained from linked Office of National Statistics data. Finally, to assess the effect of using the aggregated and weighted Charlson Index, the model was re-estimated to assess the effect of the individual component comorbidities from the Charlson Index. Results Cases and Matching There were 16,355 unique cases identified with a first nonvariceal bleed; 13,372 with specific code in Hospital Episodes Statistics, 10,938 with a specific code in GPRD, and 7955 with a specific code in both datasets. There were 16,304 (99.7%) cases matched to 5 controls each and only 8 cases (0.05%) were not matched to any controls. Median observed time before admission for cases was 7.4 years (interquartile range, 3.4–11.5) compared with 7.5 years (interquartile range, 3.5–11.5) for controls.

with a specific code in both datasets. There were 16,304 (99.7%) cases matched to 5 controls each and only 8 cases (0.05%) were not matched to any controls. Median observed time before admission for cases was 7.4 years (interquartile range, 3.4–11.5) compared with 7.5 years (interquartile range, 3.5–11.5) for controls. Unadjusted Analysis Table 1 shows the proportion of cases and controls with each exposure. As expected, aspirin and NSAIDs were the most frequently prescribed risk factor medications, and peptic ulcer and gastritis/duodenitis/esophagitis were the most frequent risk factor diagnoses. All a priori risk factors were associated with upper GIB. Peptic ulcers were coded as a diagnosis within the linked data in 4,823 patients (29% of cases). The exposures stratified by coding of peptic ulcer are shown in Supplementary Table 1.

gastritis/duodenitis/esophagitis were the most frequent risk factor diagnoses. All a priori risk factors were associated with upper GIB. Peptic ulcers were coded as a diagnosis within the linked data in 4,823 patients (29% of cases). The exposures stratified by coding of peptic ulcer are shown in Supplementary Table 1. Multivariable Analysis and PAFs There was strong evidence for an association between the nongastrointestinal Charlson Index and upper GIB after adjusting for all measured risk factors (single comorbidity adjusted OR = 1.43; 95% CI: 1.35–1.52; multiple or severe comorbidity adjusted OR = 2.26; 95% CI: 2.14–2.38; P < .001 likelihood ratio test). Table 2 shows the adjusted ORs from the final model for each exposure. We found the largest association with a bleed was with a previous Mallory-Weiss syndrome, which reflects the inherent risk of bleeding in recurrent vomitters. The variables for angiodysplasia and dialysis had the highest variance inflation factors, 1.48 and 2.35, respectively. As both of these were less than the a priori threshold of 5, all exposures were included in the final conditional logistic regression model. Stratifying this model demonstrated similar associations with comorbidity, whether or not peptic ulcer coding was present, and slightly higher associations for a peptic ulcer with exposure to previous peptic ulcers, NSAID, or aspirin use (Table 3). Associations with other risk factors were higher in the nonpeptic ulcer cohort.

ing this model demonstrated similar associations with comorbidity, whether or not peptic ulcer coding was present, and slightly higher associations for a peptic ulcer with exposure to previous peptic ulcers, NSAID, or aspirin use (Table 3). Associations with other risk factors were higher in the nonpeptic ulcer cohort. The proportion of cases attributable in the population to the combined effect of all available measured exposures was 48%, not including the effect of nongastrointestinal comorbidity. The additional proportion of cases attributable to nongastrointestinal comorbidity (or the sequential PAF) was 20%, and this was higher in magnitude than for any other measured exposure (Table 4). The next largest PAFs were 3% for aspirin and NSAID use. The PAF for comorbidity associated with peptic ulcer bleeds was slightly lower than that for nonulcer bleeds (18% vs 21%), with a higher contribution from previous peptic ulcer bleeds and aspirin and NSAIDs (Table 5). In contrast, for nonulcer bleeds, the PAF was slightly increased for gastrointestinal cancer, alcohol, anticoagulants, and selective serotonin reuptake inhibitors.

as slightly lower than that for nonulcer bleeds (18% vs 21%), with a higher contribution from previous peptic ulcer bleeds and aspirin and NSAIDs (Table 5). In contrast, for nonulcer bleeds, the PAF was slightly increased for gastrointestinal cancer, alcohol, anticoagulants, and selective serotonin reuptake inhibitors. Sensitivity Analyses The crude ORs were re-estimated for medications after excluding cases with nonmedication risk factors and these are shown in Supplementary Table 2. NSAID use was strongly associated with bleeding, with an OR of 1.67, and this increased to 2.80 with the exclusion of nonmedication risk factors. The corresponding adjusted ORs associated with NSAIDs were 1.59 with nonmedication risk factors included and 1.73 without. Altering the exposure exclusion window for NSAIDs to 30 days rather than 60 days before the bleed slightly increased the effect of NSAIDS, but had only a minimal effect on the other results, including comorbidity (see Supplementary Table 3).

th NSAIDs were 1.59 with nonmedication risk factors included and 1.73 without. Altering the exposure exclusion window for NSAIDs to 30 days rather than 60 days before the bleed slightly increased the effect of NSAIDS, but had only a minimal effect on the other results, including comorbidity (see Supplementary Table 3). Restricting the analysis to those older than 65 years old increased the proportion of cases attributable to the combined effect of all exposures from 48% to 63%, and reduced the additional proportion of cases attributable to nongastrointestinal comorbidity from 19.8% to 16.1%. Re-estimating the model using multiple imputation for missing alcohol and smoking status (modeled as binary exposures) slightly reduced the PAF associated with comorbidity from 22.9% to 22.4%, but when alcohol and smoking status were omitted from the model, the PAF was almost unaltered at 22.2%. Finally, the full model was re-estimated for each component of the Charlson Index (Table 6). The contribution of these individual comorbidities was minimal in comparison with their combined weighted effect in the Charlson Index in the main analysis.

us were omitted from the model, the PAF was almost unaltered at 22.2%. Finally, the full model was re-estimated for each component of the Charlson Index (Table 6). The contribution of these individual comorbidities was minimal in comparison with their combined weighted effect in the Charlson Index in the main analysis. Discussion This study has demonstrated that a combined measure of nongastrointestinal comorbidity is a significant independent predictor of upper GIB, even after accounting for all other recognized and measured risk factors. In addition, it explained a greater proportion of the burden of bleeding than any other risk factor in the population. The effect of this combined measure of nongastrointestinal comorbidity was far in excess of that which would be expected from its constituent diseases. The association of comorbidities with upper GIB has been studied previously, but only in smaller secondary care surveys with comorbidity as a confounder and not as the primary exposure. We searched PubMed using variants of comorbidity, etiology, causality, risk factors, and gastrointestinal hemorrhage; however, no studies were identified that set out to address the question of our article. Studies were most frequently designed to measure the association of a single medication while adjusting for any confounding by comorbidity.21,22

ts of comorbidity, etiology, causality, risk factors, and gastrointestinal hemorrhage; however, no studies were identified that set out to address the question of our article. Studies were most frequently designed to measure the association of a single medication while adjusting for any confounding by comorbidity.21,22 Two studies assessed a larger range of medications in cross-sectional hospital-based surveys.10,23 First, Weil found that only 2 comorbidities, heart failure and diabetes, contributed to upper GIB with adjusted PAFs of 5% and 4%, respectively.23 However, the study was retrospective, and with <1000 cases limiting its power. In contrast to the “extra PAF” we calculated, the adjusted PAFs in their article calculated the effect of each exposure in a pseudo-population with no other risk factors present, potentially overestimating the effect in the general population, in which a case can be caused by many risk factors. The second comparable paper of Gallerani et al found an association with comorbidity and a similar 2-fold increase in risk in those exposed to NSAIDs to what we found in our peptic ulcer cohort.10 However, it was also a retrospective survey–based study potentially subject to recall bias, and had <1000 cases. Furthermore, the authors did not separate out gastrointestinal comorbidity from nongastrointestinal comorbidity and used hospital controls, therefore limiting comparisons with our population-based study.

.10 However, it was also a retrospective survey–based study potentially subject to recall bias, and had <1000 cases. Furthermore, the authors did not separate out gastrointestinal comorbidity from nongastrointestinal comorbidity and used hospital controls, therefore limiting comparisons with our population-based study. Other studies assessed higher alcohol intake,24 H pylori,25 smoking,26 acute renal failure,27 and acute myocardial infarction28 and found associations with upper GIB. But these studies were in small selected hospitalised cohorts (n < 1000 bleeds) with limited assessments of individual comorbidity and no measure of their PAFs.

tudies assessed higher alcohol intake,24 H pylori,25 smoking,26 acute renal failure,27 and acute myocardial infarction28 and found associations with upper GIB. But these studies were in small selected hospitalised cohorts (n < 1000 bleeds) with limited assessments of individual comorbidity and no measure of their PAFs. Our study has a number of important strengths when compared with these previous works because we set out specifically to assess the degree to which nongastrointestinal comorbidity predicts nonvariceal upper GIB after removing the effects of all the available known risk factors in a much larger general population. In addition, we used a method of defining cases and exposures that utilized information from both primary and secondary care, thereby maximizing the evidence supporting each case while not excluding severe events.14 Furthermore, due to the comprehensive coverage of the English primary care system, our study's results are likely to be generalizable to the whole English population and, we believe, further afield. The linked dataset used for our study remained representative of the GPRD overall, as whole practices rather than individual patients declined or consented to the linkage. Consequently, we were able to estimate the additional attributable fraction for comorbidity in the English population that was not already attributable to other risk factors.19

our study remained representative of the GPRD overall, as whole practices rather than individual patients declined or consented to the linkage. Consequently, we were able to estimate the additional attributable fraction for comorbidity in the English population that was not already attributable to other risk factors.19 As our study was one of the first to assess the effect of the burden of comorbidity as a risk factor for upper GIB, no measure of comorbidity had been specifically validated for this purpose. We decided to use the Charlson Index because it is a well-validated score for measuring comorbidity in many different contexts. Other comorbidity scores that could be used, such as the Elixhauser Index or a simple counts of diagnoses, have been used and validated less frequently and in fewer contexts. In addition, some of the other scores also include other outcomes, such as financial cost, which are not necessarily a measure of the severity of disease. The Charlson Index was therefore selected as the most appropriate comorbidity score for our study.

ave been used and validated less frequently and in fewer contexts. In addition, some of the other scores also include other outcomes, such as financial cost, which are not necessarily a measure of the severity of disease. The Charlson Index was therefore selected as the most appropriate comorbidity score for our study. We do need to consider alternative explanations for our observed association of comorbidity with upper GIB. A potential weakness of our study is the inevitably imperfect data on some recognized risk factors that might have caused us to underestimate their importance. The GPRD contains comprehensive recording of all available diagnoses and prescriptions. However, under-reporting is likely to have occurred for H pylori infection, NSAID use, alcohol, and smoking. In the case of H pylori, there was inevitable under-reporting because there was no population screening. However, if the under-reporting of H pylori infection was to explain our study's findings, it would have to be strongly associated with comorbidity, and the evidence for this is conflicting and underpowered.29,30 In studies of ischemic heart disease, for which there is the largest body of evidence, any significant association with H pylori was minimal after adjustments for confounding.31 In our study, the apparent protective effect of H pylori after adjustments for confounding was not surprising because H pylori will have been eradicated when found.

heart disease, for which there is the largest body of evidence, any significant association with H pylori was minimal after adjustments for confounding.31 In our study, the apparent protective effect of H pylori after adjustments for confounding was not surprising because H pylori will have been eradicated when found. NSAID use might also have been under-reported, as NSAIDs can be bought over the counter from a pharmacy without a prescription, potentially explaining the low association between NSAIDs and bleeding in our study compared with a previous meta-analysis.20 However, we had higher recorded NSAID use than was reported in a recent national audit,32 and the studies used in the meta-analysis excluded patients with other known GIB risk factors.20 When we made the same exclusions in our study (Supplementary Table 2), or restricted to peptic ulcers, the association of bleeding with NSAIDs increased and became comparable with figures in the literature. With regard to over-the-counter use, nondifferential under-reporting has been shown to reduce the measured effect of prescribed medications.33 In our study, this would cause an underestimate of the effect of NSAIDs. However, in England, certain groups receive free prescriptions, such as patients older than 65 years or those with certain chronic diseases, and these groups have been shown to purchase far fewer medications over the counter than those who have to pay for prescriptions.34,35 When we restricted our analysis to those older than 65 years, thereby reducing confounding by over-the-counter medications, we found only a small reduction in the estimated PAF for comorbidity, but no change in PAF for NSAIDs. The final area of under-reporting that could affect our study was missing data for alcohol and smoking status, but these variables were not strong confounders of the association between comorbidity and bleeding and there was only a minimal effect on the PAF of comorbidity when missing data were imputed conditional on all available data and socioeconomic status.

d affect our study was missing data for alcohol and smoking status, but these variables were not strong confounders of the association between comorbidity and bleeding and there was only a minimal effect on the PAF of comorbidity when missing data were imputed conditional on all available data and socioeconomic status. We therefore believe that potential under-reporting of exposures does not explain the association we found between upper GIB and a general measure of comorbidity. This suggests that comorbidity itself, or other factors not included in our study that are associated with comorbidity, might be causing the association. It is possible that other medications not included in the study were responsible for some of this association, however, we are not aware of any additional prescribed or nonprescribed medication that would fulfill the requirements of common usage and a strong association with bleeding. Historically, nongastrointestinal comorbidity itself was commonly recognized as a risk factor for upper GIB.7 However, the concept of “stress ulceration” is no longer accepted, aside from patients on ITU who are exposed to severe acute physiological stresses from ventilation, coagulopathy, liver failure, renal failure, septic shock, or nutritional support.9 The physiological effects from chronic comorbidities in our study are unlikely to be as severe as those that occur on ITU and, therefore, what we are describing is likely to have a different mechanism than that seen in the ITU setting. Many potential mechanisms for our observed association can be hypothesized; for example, reduced epithelial microperfusion in cardiac failure,36 decreased oxygen levels in chronic obstructive pulmonary disease,37,38 poor nutritional status in many diseases, or the platelet and clotting dysfunction in end-stage renal failure.27,39 However, it is unlikely that there is a single mechanism that accounts for the association we found, but rather that multiple illnesses and mechanisms have a cumulative effect. This was shown by the graded effect of the Charlson Index and by Table 6, in which no individual disease accounted for the magnitude of the overall association with comorbidity.

is a single mechanism that accounts for the association we found, but rather that multiple illnesses and mechanisms have a cumulative effect. This was shown by the graded effect of the Charlson Index and by Table 6, in which no individual disease accounted for the magnitude of the overall association with comorbidity. Our findings contrast with current beliefs that the main burden of bleeding in the general population comes from known iatrogenic causes, such as NSAIDs prescribed for analgesia or antiplatelet agents prescribed for cardiac and cerebrovascular disease,40 and that this burden would be reduced by increasing PPI use.41 Instead, we have demonstrated that the extra contribution of these medications to bleeding cases was not large after considering the contributions of other risk factors present in the population. Therefore, simply increasing PPI prescriptions in patients on high-risk medications might not have as large an impact as previously thought. In conclusion, the largest measurable burden of upper gastrointestinal hemorrhage in this study was attributed to nongastrointestinal comorbidity. In a proportion of patients, a bleed is an indicator of the burden of their comorbidity, and recognizing this will help guide management, particularly in the absence of modifiable gastrointestinal risk factors. However, our finding also explains why the incidence of nonvariceal bleeding is likely to remain high in an aging population, thereby necessitating continued acute gastroenterology service provision.

nd recognizing this will help guide management, particularly in the absence of modifiable gastrointestinal risk factors. However, our finding also explains why the incidence of nonvariceal bleeding is likely to remain high in an aging population, thereby necessitating continued acute gastroenterology service provision. Supplementary Material Supplementary Table 1 Proportion of Cases Exposed 2 Months Before Bleed Date or Match Date Stratified by Coded Peptic Ulcer

nd recognizing this will help guide management, particularly in the absence of modifiable gastrointestinal risk factors. However, our finding also explains why the incidence of nonvariceal bleeding is likely to remain high in an aging population, thereby necessitating continued acute gastroenterology service provision. Supplementary Material Supplementary Table 1 Proportion of Cases Exposed 2 Months Before Bleed Date or Match Date Stratified by Coded Peptic Ulcer Peptic ulcer coded, n Exposed, % No peptic ulcer coded, n Exposed, % Charlson index+ No comorbidity 883 18.3 2557 22.2 Single comorbidity 916 19.0 2306 20.0 Multiple or severe 3024 62.7 6669 57.8 Gastrointestinal Cirrhosis—none coded 4753 98.5 11,251 97.6 Cirrhosis only 17 0.4 46 0.4 Cirrhosis—varices 8 0.2 57 0.5 Cirrhosis—ascites 32 0.7 140 1.2 Cirrhosis—encephalopathy 13 0.3 38 0.3 Gastritis, duodenitis, or esophagitis 710 14.7 2341 20.3 Peptic ulcer 864 17.9 988 8.6 Helicobacter pylori 162 3.4 447 3.9 Angiodysplasia 1 0.0 5 0.0 Mallory-Weiss syndrome 11 0.2 85 0.7 Crohn's disease 19 0.4 95 0.8 GI cancer 254 5.3 920 8.0 Lifestyle Alcohol—not coded 3299 68.4 7727 67.0 Alcohol—non-drinker 97 2.0 278 2.4 Alcohol—ex-drinker 20 0.4 44 0.4 Alcohol—mentioned 284 5.9 693 6.0 Alcohol—over limits 1105 22.9 2658 23.0 Alcohol—complications 18 0.4 132 1.1 Smoking—not coded 2753 57.1 6434 55.8 Smoking—nonsmoker 646 13.4 1686 14.6 Smoking—ex-smoker 288 6.0 600 5.2 Smoking—passive 405 8.4 1050 9.1 Smoking—current 731 15.2 1762 15.3 Medications Aspirin 1831 38.0 3561 30.9 NSAIDs 1431 29.7 2467 21.4 COX II inhibitors 222 4.6 383 3.3 Clopidogrel 198 4.1 470 4.1 Oral steroids 428 8.9 1150 10.0 Anticoagulants 427 8.9 1190 10.3 SSRIs 460 9.5 1565 13.6 Other diagnoses Aortic stenosis 125 2.6 225 2.0 Repair of AAA 36 0.7 79 0.7 Dialysis 30 0.6 58 0.5 Confounders Previous upper GI procedure 817 16.9 2621 22.7 Previous PPI use 906 18.8 3679 31.9 Age (median and interquartile range) 75 64–83 72 54–82 AAA, Abdominal aortic aneurysm; SSRI, selective serotonin reuptake inhibitors.

iagnoses Aortic stenosis 125 2.6 225 2.0 Repair of AAA 36 0.7 79 0.7 Dialysis 30 0.6 58 0.5 Confounders Previous upper GI procedure 817 16.9 2621 22.7 Previous PPI use 906 18.8 3679 31.9 Age (median and interquartile range) 75 64–83 72 54–82 AAA, Abdominal aortic aneurysm; SSRI, selective serotonin reuptake inhibitors. aNongastrointestinal comorbidity is catogorized as: Charlon Index = 0 is no comorbidity, Charlson Index = 1 single or comorbidity, and Charlson Index = 2 is multiple or severe comorbidity. Supplementary Table 2 The Association of Medications With Upper Gastrointestinal Bleeding After Excluding Patients With Nonmedication Risk Factors (Age, Year, Practice, and Sex Matched) Crude OR Adjusteda OR Lower 95% CI Upper 95% CI Aspirin 2.39 1.73 1.60 1.87 NSAIDs 2.80 1.78 1.64 1.93 COX II inhibitors 2.59 1.50 1.23 1.83 Clopidogrel 7.30 2.15 1.70 2.73 Oral steroids 1.23 1.23 1.08 1.41 Anticoagulants 4.83 2.26 1.99 2.57 SSRIs 2.78 1.52 1.34 1.71 SSRI, selective serotonin reuptake inhibitors. a Adjusted for all other variables in Table and in Figure 1. Supplementary Table 3 Sequential Population Attributable Fractions for Nonvariceal Upper Gastrointestinal Hemorrhage (NSAIDS Prescribed Between 1 Month and 1 Year Before Bleed)

Crude OR Adjusteda OR Lower 95% CI Upper 95% CI Aspirin 2.39 1.73 1.60 1.87 NSAIDs 2.80 1.78 1.64 1.93 COX II inhibitors 2.59 1.50 1.23 1.83 Clopidogrel 7.30 2.15 1.70 2.73 Oral steroids 1.23 1.23 1.08 1.41 Anticoagulants 4.83 2.26 1.99 2.57 SSRIs 2.78 1.52 1.34 1.71 SSRI, selective serotonin reuptake inhibitors. a Adjusted for all other variables in Table and in Figure 1. Supplementary Table 3 Sequential Population Attributable Fractions for Nonvariceal Upper Gastrointestinal Hemorrhage (NSAIDS Prescribed Between 1 Month and 1 Year Before Bleed) Sequential population attributable fractions,a,b% Nongastrointestinal comorbidity 19.92 Gastrointestinal Cirrhosis 0.48 Gastritis, duodenitis, or esophagitis 1.86 Peptic ulcer 2.01 Helicobacter pylori −0.06 Angiodysplasia 0.01 Mallory-Weiss syndrome 0.28 Crohn's disease 0.14 GI cancer 1.09 Lifestyle Alcohol use 2.83 Smoking 0.93 Medications Aspirin 2.91 NSAIDs 3.18 COX II inhibitors 0.36 Clopidogrel 0.32 Oral steroids 0.58 Anticoagulants 1.11 SSRIs 1.60 Other diagnoses Aortic stenosis 0.16 Repair of aorta 0.02 Dialysis 0.06 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures, and age.

Sequential population attributable fractions,a,b% Nongastrointestinal comorbidity 19.92 Gastrointestinal Cirrhosis 0.48 Gastritis, duodenitis, or esophagitis 1.86 Peptic ulcer 2.01 Helicobacter pylori −0.06 Angiodysplasia 0.01 Mallory-Weiss syndrome 0.28 Crohn's disease 0.14 GI cancer 1.09 Lifestyle Alcohol use 2.83 Smoking 0.93 Medications Aspirin 2.91 NSAIDs 3.18 COX II inhibitors 0.36 Clopidogrel 0.32 Oral steroids 0.58 Anticoagulants 1.11 SSRIs 1.60 Other diagnoses Aortic stenosis 0.16 Repair of aorta 0.02 Dialysis 0.06 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures, and age. b The estimate in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed. This article has an accompanying continuing medical education activity on page e18. Learning Objective: Upon completion of this CME exercise, and reading of the associated paper, successful learners will be able to appraise how multiple risk factors can contribute to upper gastrointestinal bleeding and recognize the large contribution of comorbidity.

continuing medical education activity on page e18. Learning Objective: Upon completion of this CME exercise, and reading of the associated paper, successful learners will be able to appraise how multiple risk factors can contribute to upper gastrointestinal bleeding and recognize the large contribution of comorbidity. Conflicts of interest This author discloses the following: Timothy Richard Card is married to an employee of Takaeda Pharmaceuticals (recently changed jobs from AstraZeneca). The remaining authors disclose no conflicts. Funding This work was supported by personal fellowships: Colin Crooks is a research fellow and gastroenterology trainee supported by a Medical Research Council Population Health Scientist fellowship (grant number G0802427), Tim Card is a Clinical Associate Professor and Consultant Gastroenterologist supported by Nottingham University and the National Health Service, and Joe West is a Clinical Associate Professor, Reader in Epidemiology, and Consultant Gastroenterologist supported by a University of Nottingham/Nottingham University Hospitals NHS Trust Senior Clinical Research fellowship. These funding bodies had no role in the study design, or the collection, analysis, or interpretation of data. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at http://dx.doi.org/10.1053/j.gastro.2013.02.040. Figure 1 Risk factors for upper GIB. SSRI, selective serotonin reuptake inhibitor. Table 1 Proportion of Cases and Controls Exposed 2 Months Before Bleed Date or Match Date

Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at http://dx.doi.org/10.1053/j.gastro.2013.02.040. Figure 1 Risk factors for upper GIB. SSRI, selective serotonin reuptake inhibitor. Table 1 Proportion of Cases and Controls Exposed 2 Months Before Bleed Date or Match Date Controls, n Exposed, % Cases, n Exposed, % Charlson Indexa No comorbidity 30,194 37.0 3440 21.0 Single comorbidity 18,714 22.9 3222 19.7 Multiple or severe 32,728 40.1 9693 59.3 Gastrointestinal Cirrhosis—none coded 81,385 99.7 16,004 97.9 Cirrhosis only 65 0.1 63 0.4 Cirrhosis—varices 62 0.1 65 0.4 Cirrhosis—ascites 86 0.1 172 1.1 Cirrhosis—encephalopathy 38 0.0 51 0.3 Gastritis, duodenitis, or esophagitis 7904 9.7 3051 18.7 Peptic ulcer 3830 4.7 1852 11.3 Helicobacter pylori 1964 2.4 609 3.7 Angiodysplasia 14 0.0 6 0.0 Mallory-Weiss syndrome 34 0.0 96 0.6 Crohn's disease 222 0.3 114 0.7 GI cancer 2494 3.1 1174 7.2 Lifestyle Alcohol—not coded 61,536 75.4 11,026 67.4 Alcohol—nondrinker 1485 1.8 375 2.3 Alcohol—ex-drinker 176 0.2 64 0.4 Alcohol—mentioned 4317 5.3 977 6.0 Alcohol—over limits 14,073 17.2 3763 23.0 Alcohol—complications 49 0.1 150 0.9 Smoking—not coded 51,751 63.4 9187 56.2 Smoking—nonsmoker 11,666 14.3 2332 14.3 Smoking—ex-smoker 4075 5.0 888 5.4 Smoking—passive 5574 6.8 1455 8.9 Smoking—current 8570 10.5 2493 15.2 Medications Aspirin 18,079 22.1 5392 33.0 NSAIDs 12,722 15.6 3820 23.4 COX II inhibitors 1687 2.1 605 3.7 Clopidogrel 1297 1.6 668 4.1 Oral steroids 4135 5.1 1578 9.6 Anticoagulants 3799 4.7 1617 9.9 SSRIs 4813 5.9 2025 12.4 Other diagnoses Aortic stenosis 782 1.0 350 2.1 Repair of AAA 307 0.4 115 0.7 Dialysis 70 0.1 88 0.5 Confounders Previous upper GI procedure 10,471 12.8 3438 21.0 PPI 10,909 13.4 4585 28.0 Age (median and interquartile range) 73 57–82 72 57–81 AAA, Abdominal aortic aneursym; SSRI, selective serotonin reuptake inhibitors.

her diagnoses Aortic stenosis 782 1.0 350 2.1 Repair of AAA 307 0.4 115 0.7 Dialysis 70 0.1 88 0.5 Confounders Previous upper GI procedure 10,471 12.8 3438 21.0 PPI 10,909 13.4 4585 28.0 Age (median and interquartile range) 73 57–82 72 57–81 AAA, Abdominal aortic aneursym; SSRI, selective serotonin reuptake inhibitors. a Nongastrointestinal comorbidity is catogorized as: Charslon Index = 0 is no comorbidity, Charlson Index = 1 single or comorbidity, and Charlson Index = 2 is multiple or severe comorbidity. Table 2 Adjusted Model for Nonvariceal Upper Gastrointestinal Bleeding All Cases (Age, Year, Practice, and Sex Matched)

her diagnoses Aortic stenosis 782 1.0 350 2.1 Repair of AAA 307 0.4 115 0.7 Dialysis 70 0.1 88 0.5 Confounders Previous upper GI procedure 10,471 12.8 3438 21.0 PPI 10,909 13.4 4585 28.0 Age (median and interquartile range) 73 57–82 72 57–81 AAA, Abdominal aortic aneursym; SSRI, selective serotonin reuptake inhibitors. a Nongastrointestinal comorbidity is catogorized as: Charslon Index = 0 is no comorbidity, Charlson Index = 1 single or comorbidity, and Charlson Index = 2 is multiple or severe comorbidity. Table 2 Adjusted Model for Nonvariceal Upper Gastrointestinal Bleeding All Cases (Age, Year, Practice, and Sex Matched) Adjusted OR Lower 95% CI Upper 95% CI Charlson Index No comorbidity 1.00 1.00 1.00 Single comorbidity 1.43 1.35 1.52 Multiple or severe 2.26 2.14 2.38 Gastrointestinal Cirrhosis—none 1.00 1.00 1.00 Cirrhosis only 3.89 2.61 5.77 Cirrhosis—varices 3.75 2.51 5.61 Cirrhosis—ascites 5.96 4.46 7.96 Cirrhosis—encephalopathy 5.05 3.14 8.10 Gastritis, duodenitis, or esophagitis 1.46 1.39 1.55 Peptic ulcer 2.11 1.98 2.26 Helicobacter pylori 0.96 0.86 1.07 Angiodysplasia 1.67 0.58 4.80 Mallory-Weiss syndrome 12.39 8.16 18.82 Crohn's disease 2.19 1.71 2.81 GI cancer 2.13 1.97 2.31 Lifestyle Alcohol—not 1.00 1.00 1.00 Alcohol—nondrinker 1.25 1.10 1.42 Alcohol—ex-drinker 1.39 1.01 1.92 Alcohol—mentioned 1.05 0.96 1.14 Alcohol—over limits 1.42 1.35 1.49 Alcohol—complications 9.33 6.48 13.44 Smoking—not 1.00 1.00 1.00 Smoking—non-smoker 0.97 0.92 1.04 Smoking—ex-smoker 0.94 0.86 1.02 Smoking—passive 1.03 0.95 1.11 Smoking—current 1.29 1.22 1.37 Medications Aspirin 1.50 1.43 1.57 NSAIDs 1.59 1.52 1.66 COX II inhibitors 1.52 1.37 1.69 Clopidogrel 1.74 1.57 1.94 Oral steroids 1.38 1.29 1.48 Anticoagulants 1.94 1.81 2.08 SSRIs 1.72 1.62 1.83 Other diagnoses Aortic stenosis 1.58 1.38 1.82 Repair of AAA 1.29 1.02 1.64 Dialysis 3.59 2.55 5.05 Confounders Previous upper GI procedure 1.10 1.04 1.15 PPI 1.59 1.52 1.67 Age 1.09 1.08 1.10 AAA, Abdominal aortic aneurysm; SSRI, selective serotonin reuptake inhibitors.

ticoagulants 1.94 1.81 2.08 SSRIs 1.72 1.62 1.83 Other diagnoses Aortic stenosis 1.58 1.38 1.82 Repair of AAA 1.29 1.02 1.64 Dialysis 3.59 2.55 5.05 Confounders Previous upper GI procedure 1.10 1.04 1.15 PPI 1.59 1.52 1.67 Age 1.09 1.08 1.10 AAA, Abdominal aortic aneurysm; SSRI, selective serotonin reuptake inhibitors. Table 3 Adjusted Model for Nonvariceal Upper Gastrointestinal Bleeding Stratified by Coding of Peptic Ulcer (Age, Year, Practice, and Sex Matched)

ticoagulants 1.94 1.81 2.08 SSRIs 1.72 1.62 1.83 Other diagnoses Aortic stenosis 1.58 1.38 1.82 Repair of AAA 1.29 1.02 1.64 Dialysis 3.59 2.55 5.05 Confounders Previous upper GI procedure 1.10 1.04 1.15 PPI 1.59 1.52 1.67 Age 1.09 1.08 1.10 AAA, Abdominal aortic aneurysm; SSRI, selective serotonin reuptake inhibitors. Table 3 Adjusted Model for Nonvariceal Upper Gastrointestinal Bleeding Stratified by Coding of Peptic Ulcer (Age, Year, Practice, and Sex Matched) Peptic ulcer Nonpeptic ulcer Adjusted OR Lower 95% CI Upper 95% CI Adjusted OR Lower 95% CI Upper 95% CI Charlson Indexa No comorbidity 1.00 1.00 1.00 1.00 1.00 1.00 Single comorbidity 1.45 1.30 1.62 1.42 1.33 1.52 Multiple or severe 2.28 2.06 2.52 2.27 2.13 2.42 Gastrointestinal Cirrhosis—none 1.00 1.00 1.00 1.00 1.00 1.00 Cirrhosis only 3.98 2.03 7.80 3.80 2.30 6.27 Cirrhosis—varices 2.33 0.92 5.94 4.15 2.63 6.54 Cirrhosis—ascites 4.67 2.63 8.29 6.85 4.85 9.65 Cirrhosis—encephalopathy 3.16 1.39 7.20 6.66 3.70 12.01 Gastritis, duodenitis, or esophagitis 1.22 1.10 1.36 1.58 1.48 1.68 Peptic ulcer 4.36 3.92 4.85 1.37 1.25 1.49 Helicobacter pylori 1.04 0.85 1.27 0.94 0.83 1.06 Angiodysplasia 1.71 0.16 18.64 1.49 0.44 5.00 Mallory Weiss syndrome 3.75 1.43 9.84 16.54 10.23 26.77 Crohns disease 1.18 0.68 2.05 2.65 1.99 3.54 GI cancer 1.45 1.23 1.69 2.45 2.23 2.70 Lifestyle Alcohol—not 1.00 1.00 1.00 1.00 1.00 1.00 Alcohol—nondrinker 1.14 0.89 1.47 1.30 1.11 1.51 Alcohol—ex-drinker 1.58 0.89 2.81 1.30 0.88 1.93 Alcohol—mentioned 1.02 0.87 1.20 1.04 0.94 1.16 Alcohol—over limits 1.34 1.22 1.47 1.45 1.36 1.54 Alcohol—complications 3.88 1.70 8.87 11.85 7.76 18.10 Smoking—not 1.00 1.00 1.00 1.00 1.00 1.00 Smoking—nonsmoker 0.99 0.88 1.11 0.96 0.90 1.04 Smoking—ex-smoker 0.96 0.82 1.13 0.92 0.83 1.03 Smoking—passive 0.95 0.82 1.09 1.06 0.97 1.16 Smoking—current 1.35 1.21 1.51 1.28 1.19 1.37 Medications Aspirin 1.69 1.56 1.82 1.42 1.34 1.50 NSAIDs 2.21 2.04 2.39 1.37 1.29 1.45 COX II inhibitors 1.81 1.51 2.17 1.42 1.24 1.62 Clopidogrel 2.04 1.68 2.48 1.70 1.49 1.93 Oral steroids 1.31 1.16 1.49 1.40 1.29 1.51 Anticoagulants 1.67 1.47 1.90 2.10 1.94 2.28 SSRIs 1.47 1.30 1.66 1.84 1.71 1.97 Other diagnoses Aortic stenosis 1.79 1.41 2.26 1.46 1.23 1.75 Repair of AAA 1.33 0.87 2.04 1.27 0.95 1.68 Dialysis 5.56 2.95 10.48 2.92 1.94 4.41 Confounders Previous upper GI procedure 0.88 0.80 0.98 1.20 1.13 1.28 PPI 0.82 0.74 0.91 2.01 1.90 2.13 Age 1.10 1.08 1.11 1.09 1.08 1.10 AAA, Abdominal aortic aneursym; SSRI, selective serotonin reuptake inhibitors.

2.26 1.46 1.23 1.75 Repair of AAA 1.33 0.87 2.04 1.27 0.95 1.68 Dialysis 5.56 2.95 10.48 2.92 1.94 4.41 Confounders Previous upper GI procedure 0.88 0.80 0.98 1.20 1.13 1.28 PPI 0.82 0.74 0.91 2.01 1.90 2.13 Age 1.10 1.08 1.11 1.09 1.08 1.10 AAA, Abdominal aortic aneursym; SSRI, selective serotonin reuptake inhibitors. a Nongastrointestinal comorbidity is catogorized as: Charlon Index = 0 is no comorbidity, Charlson Index = 1 single or comorbidity, and Charlson Index = 2 is multiple or severe comorbidity. Table 4 Sequential Population Attributable Fractions for Nonvariceal Upper Gastrointestinal Hemorrhage (All Cases) Sequential population attributable fractionsa,b % 95% CI Nongastrointestinal comorbidity 19.80 18.43 to 21.18 Gastrointestinal Cirrhosis 0.49 0.41 to 0.57 Gastritis, duodenitis or esophagitis 1.98 1.66 to 2.30 Peptic ulcer 2.05 1.81 to 2.28 Helicobacter pylori −0.04 −0.15 to 0.08 Angiodysplasia 0.01 −0.01 to 0.02 Mallory-Weiss syndrome 0.29 0.22 to 0.37 Crohn's disease 0.14 0.08 to 0.19 GI cancer 1.11 0.96 to 1.27 Lifestyle Alcohol use 2.89 2.39 to 3.39 Smoking 0.83 0.27 to 3.42 Medications Aspirin 2.95 2.54 to 3.36 NSAIDs 3.07 2.72 to 3.42 COX II inhibitors 0.33 0.23 to 0.44 Clopidogrel 0.34 0.26 to 0.43 Oral steroids 0.59 0.44 to 0.74 Anticoagulants 1.19 1.04 to 1.35 SSRIs 1.58 1.36 to 1.80 Other diagnoses Aortic stenosis 0.16 0.10 to 0.22 Repair of aorta 0.03 0.00 to 0.06 Dialysis 0.07 0.04 to 0.09 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures, and age.

Sequential population attributable fractionsa,b % 95% CI Nongastrointestinal comorbidity 19.80 18.43 to 21.18 Gastrointestinal Cirrhosis 0.49 0.41 to 0.57 Gastritis, duodenitis or esophagitis 1.98 1.66 to 2.30 Peptic ulcer 2.05 1.81 to 2.28 Helicobacter pylori −0.04 −0.15 to 0.08 Angiodysplasia 0.01 −0.01 to 0.02 Mallory-Weiss syndrome 0.29 0.22 to 0.37 Crohn's disease 0.14 0.08 to 0.19 GI cancer 1.11 0.96 to 1.27 Lifestyle Alcohol use 2.89 2.39 to 3.39 Smoking 0.83 0.27 to 3.42 Medications Aspirin 2.95 2.54 to 3.36 NSAIDs 3.07 2.72 to 3.42 COX II inhibitors 0.33 0.23 to 0.44 Clopidogrel 0.34 0.26 to 0.43 Oral steroids 0.59 0.44 to 0.74 Anticoagulants 1.19 1.04 to 1.35 SSRIs 1.58 1.36 to 1.80 Other diagnoses Aortic stenosis 0.16 0.10 to 0.22 Repair of aorta 0.03 0.00 to 0.06 Dialysis 0.07 0.04 to 0.09 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures, and age. b The estimate in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed. Table 5 Sequential Population Attributable Fractions for Nonvariceal Upper Gastrointestinal Hemorrhage Strafied by Coding for Peptic Ulcer

b The estimate in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed. Table 5 Sequential Population Attributable Fractions for Nonvariceal Upper Gastrointestinal Hemorrhage Strafied by Coding for Peptic Ulcer Sequential population attributable fractions,a,b% Peptic ulcer Nonpeptic ulcer Nongastrointestinal comorbidity 18.44 20.50 Gastrointestinal Cirrhosis 0.32 0.57 Gastritis, duodenitis, or esophagitis 0.69 2.74 Peptic ulcer 5.31 0.69 Helicobacter pylori 0.05 −0.07 Angiodysplasia 0.01 0.00 Mallory-Weiss syndrome 0.06 0.39 Crohn's disease 0.02 0.19 GI cancer 0.35 1.48 Lifestyle Alcohol use 1.93 3.30 Smoking 0.80 0.81 Medications Aspirin 3.99 2.42 NSAIDs 5.40 2.00 COX II inhibitors 0.47 0.28 Clopidogrel 0.38 0.35 Oral steroids 0.36 0.66 Anticoagulants 0.78 1.41 SSRIs 0.74 2.02 Other diagnoses Aortic stenosis 0.22 0.12 Repair of aorta 0.02 0.03 Dialysis 0.09 0.05 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures and age.

Sequential population attributable fractions,a,b% Peptic ulcer Nonpeptic ulcer Nongastrointestinal comorbidity 18.44 20.50 Gastrointestinal Cirrhosis 0.32 0.57 Gastritis, duodenitis, or esophagitis 0.69 2.74 Peptic ulcer 5.31 0.69 Helicobacter pylori 0.05 −0.07 Angiodysplasia 0.01 0.00 Mallory-Weiss syndrome 0.06 0.39 Crohn's disease 0.02 0.19 GI cancer 0.35 1.48 Lifestyle Alcohol use 1.93 3.30 Smoking 0.80 0.81 Medications Aspirin 3.99 2.42 NSAIDs 5.40 2.00 COX II inhibitors 0.47 0.28 Clopidogrel 0.38 0.35 Oral steroids 0.36 0.66 Anticoagulants 0.78 1.41 SSRIs 0.74 2.02 Other diagnoses Aortic stenosis 0.22 0.12 Repair of aorta 0.02 0.03 Dialysis 0.09 0.05 SSRI, selective serotonin reuptake inhibitors. a Age, year, practice, and sex matched and adjusted for PPI use, previous upper gastrointestinal procedures and age. b The estimate in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed. Table 6 The Adjusteda Association of the Component Comorbidities of Charlson Index With Nonvariceal Bleeding

b The estimate in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed. Table 6 The Adjusteda Association of the Component Comorbidities of Charlson Index With Nonvariceal Bleeding Proportion of cases exposed, % ORa Lower 95% CI Upper 95% CI PAF,b% Myocardial infarction 13.98 1.04 0.97 1.10 0.12 Congestive cardiac disease 19.90 1.49 1.41 1.58 1.95 Peripheral vascular disease 11.17 1.31 1.23 1.41 0.70 Dementia 8.98 1.40 1.30 1.50 1.00 Chronic pulmonary disease 31.80 1.11 1.06 1.16 1.10 Cerebrovascular disease 23.1 1.13 1.08 1.19 0.80 Rheumatological disease 10.13 1.06 0.99 1.13 0.17 Uncomplicated diabetes 17.88 1.01 0.96 1.06 0.04 Hemiplegia 4.73 1.79 1.62 1.97 0.67 Renal disease 14.42 1.71 1.61 1.82 1.74 Diabetes with complications 12.30 1.00 0.94 1.06 −0.01 Any malignancy 13.11 1.21 1.14 1.28 0.78 Lymphoproliferative disorders 2.21 1.95 1.70 2.24 0.43 Metastatic solid tumour 5.89 2.35 2.14 2.57 1.29 HIV/AIDS 0.06 0.69 0.31 1.55 −0.00 HIV/AIDS, human immunodeficiency virus/acquired immunodeficiency syndrome. a Adjusted for all other variables in this Table and Table 2 and matched for year, practice, and sex.

Proportion of cases exposed, % ORa Lower 95% CI Upper 95% CI PAF,b% Myocardial infarction 13.98 1.04 0.97 1.10 0.12 Congestive cardiac disease 19.90 1.49 1.41 1.58 1.95 Peripheral vascular disease 11.17 1.31 1.23 1.41 0.70 Dementia 8.98 1.40 1.30 1.50 1.00 Chronic pulmonary disease 31.80 1.11 1.06 1.16 1.10 Cerebrovascular disease 23.1 1.13 1.08 1.19 0.80 Rheumatological disease 10.13 1.06 0.99 1.13 0.17 Uncomplicated diabetes 17.88 1.01 0.96 1.06 0.04 Hemiplegia 4.73 1.79 1.62 1.97 0.67 Renal disease 14.42 1.71 1.61 1.82 1.74 Diabetes with complications 12.30 1.00 0.94 1.06 −0.01 Any malignancy 13.11 1.21 1.14 1.28 0.78 Lymphoproliferative disorders 2.21 1.95 1.70 2.24 0.43 Metastatic solid tumour 5.89 2.35 2.14 2.57 1.29 HIV/AIDS 0.06 0.69 0.31 1.55 −0.00 HIV/AIDS, human immunodeficiency virus/acquired immunodeficiency syndrome. a Adjusted for all other variables in this Table and Table 2 and matched for year, practice, and sex. b Sequential population attributable fractions: The estimates in each row are calculated separately conditional on all the other variables in the model. They should therefore not be interpreted as summing over the column to 100%. Sequential PAF estimates the additional proportion of nonvariceal bleeding cases attributable to each risk factor after cases attributable to all the other risk factors in the model have been removed.

The gastrointestinal (GI) tract is essential for digestion of foods, absorption of nutrients and water, energy balance, and protection from pathogenic microorganisms. Most aspects of GI physiology are under neural control, which is exerted via extrinsic nerves (which include both primary afferent and autonomic fibers that ultimately connect the gut tissues with the central nervous system [CNS]) and a vast network of intrinsic enteric neurons (1–5 × 108) and glial cells that are organized into the myenteric and submucosal plexi of the enteric nervous system (ENS).1 On the basis of their neurochemical properties, enteric neurons are subdivided into multiple subtypes that share molecular, morphologic, and physiological characteristics.2, 3 Unlike enteric neurons whose cell bodies are restricted to the myenteric and submucosal ganglia, enteric glial cells (EGCs), and neuronal fibers are distributed throughout the gut wall, including the lamina propria of the mucosa.4, 5 Among other functions, the ENS regulates GI peristalsis, epithelial secretion, intestinal blood flow, and transmucosal movement of liquids largely independently of central nervous system input.1

al cells (EGCs), and neuronal fibers are distributed throughout the gut wall, including the lamina propria of the mucosa.4, 5 Among other functions, the ENS regulates GI peristalsis, epithelial secretion, intestinal blood flow, and transmucosal movement of liquids largely independently of central nervous system input.1 In addition to its motor and secretory functions, the GI tract is the largest sensory organ of the body, which incessantly monitors the dynamic microenvironment of the gut wall and its lumen.6 Several cellular systems contribute to the sensory function of the gut, including the enteroendocrine cells of the intestinal epithelium, the mucosal immune system, and the ENS.1 The integrated responses of these cellular networks enable the gut to build highly selective anatomic and functional barriers that allow absorption of useful nutrients and exclusion of harmful chemicals and micro-organisms. Information relating to the chemical composition and caloric value of ingested food, the dynamic equilibrium of the microbial ecosystem of the gut (microbiota), and the physiological state of the gut wall reaches the brain via the neurohumoral pathways of the microbiota–gut–brain (MGB) axis and allows the CNS to generate appropriate homeostatic and behavioral responses.6, 7, 8

c value of ingested food, the dynamic equilibrium of the microbial ecosystem of the gut (microbiota), and the physiological state of the gut wall reaches the brain via the neurohumoral pathways of the microbiota–gut–brain (MGB) axis and allows the CNS to generate appropriate homeostatic and behavioral responses.6, 7, 8 Emerging evidence suggests that gut microflora can have dramatic effects on the development and function of the nervous system, both at the local as well as at the systemic level. Although disruption of the physiological microbiota composition (dysbiosis) is known to influence cognitive activity and behavior, such as stress response, anxiety, and memory,7, 9 the mechanistic understanding of microbe–neural interactions remains obscure. Because the ENS constitutes a key relay station along the MGB axis, understanding the mechanisms of ENS–microbe communication is essential for deciphering how the gut microenvironment influences physiology at the local and organismal level. Here, we provide a brief overview of the impact of microbiota and the mucosal immune system on ENS development and homeostasis.

along the MGB axis, understanding the mechanisms of ENS–microbe communication is essential for deciphering how the gut microenvironment influences physiology at the local and organismal level. Here, we provide a brief overview of the impact of microbiota and the mucosal immune system on ENS development and homeostasis. Genetic Programs That Control ENS Development The ENS is derived from neuroectodermal progenitors that originate mainly from the vagal neural crest, invade the foregut during embryogenesis, and migrate rostrocaudaly to colonize the entire GI tract. In mammals, enteric neurogenesis and gliogenesis occur mostly during embryonic and fetal stages but a considerable fraction of enteric neurons and glia are born within the postnatal gut.10 Furthermore, functional maturation of intestinal neural circuits also occurs within the early postnatal period.11 To date, the development of the ENS has been examined primarily from the point of view of genetic and molecular mechanisms that operate within the neuroectodermal lineages of the gut. These studies have identified several transcription factors, such as SOX10 (an SRY-related HMG-box transcription factor), FOXD3 (a member of the forkhead protein family) and HAND2,12, 13, 14, 15 which control the survival and lineage choices of ENS progenitors. In addition, ENS lineages express several types of cell surface receptors, such as the receptor tyrosine kinase RET and the G-protein–coupled endothelin receptor B (EDNRB), which control multiple aspects of ENS development and neural circuit assembly. RET and EDNRB are activated by members of the GDNF family of ligands and endothelin-3, respectively, which are produced by the intestinal mesenchyme, highlighting the key role of the cellular microenvironment on the development of ENS lineages.16 For a comprehensive recent review of the cellular and molecular mechanisms underlying ENS development, please refer to Lake et al.16

f ligands and endothelin-3, respectively, which are produced by the intestinal mesenchyme, highlighting the key role of the cellular microenvironment on the development of ENS lineages.16 For a comprehensive recent review of the cellular and molecular mechanisms underlying ENS development, please refer to Lake et al.16 The Role of Gut Microbial Factors on the Development and Homeostasis of the ENS Immediately after birth the GI tract is colonized by complex microbial communities (>100 trillion microbes belonging to ∼1000 species), which influence multiple aspects of host physiology, including metabolism, immune responses, behavior, and circadian rhythm.7, 17, 18, 19 The role of microbiota on ENS organization is highlighted by the reduced number of enteric neurons and the associated deficits in gut motility observed in germ-free (GF) mice.20, 21, 22 In addition, GF mice show attenuated excitability of intrinsic primary afferent neurons23 that are part of the hard-wired gut-brain neural pathways.24 Furthermore, the development and continuous homeostatic influx of EGCs into the intestinal mucosa is defective in GF mice.25 These observations argue that microbiota is essential for the assembly of intestinal neural circuits and for signaling along the gut–brain axis. Interestingly, reconstitution of GF mice with conventional microbiota normalized the density of EGC network and gut physiology,25, 26 raising interesting questions relating to the cellular plasticity of the ENS and the mechanisms by which microbiota influence its homeostasis.

and for signaling along the gut–brain axis. Interestingly, reconstitution of GF mice with conventional microbiota normalized the density of EGC network and gut physiology,25, 26 raising interesting questions relating to the cellular plasticity of the ENS and the mechanisms by which microbiota influence its homeostasis. A recent report showed that the reduction of myenteric neurons in GF mice is present by postnatal day 327 when the number and diversity of gut microbiota has not been established,28 raising the possibility that, in addition to factors associated with changes in early postnatal gut physiology or feeding, maternal microbial factors may play a role in ENS development during pregnancy via uteroplacental circulation. Consistent with this idea, microbial colonization of the gut may occur before birth.29, 30 Furthermore, recent evidence indicates that maternal microbe-derived factors and the maternal immune system contribute to the offspring’s immune and neuronal homeostasis.31, 32, 33, 34 Taken together, these observations suggest that dynamic host–microbe interactions during critical developmental periods could increase the risk of neurodevelopmental disorders and have long-term consequences on neuronal function. Here, we summarize the impact of gut microbial factors on ENS development and homeostasis.

Taken together, these observations suggest that dynamic host–microbe interactions during critical developmental periods could increase the risk of neurodevelopmental disorders and have long-term consequences on neuronal function. Here, we summarize the impact of gut microbial factors on ENS development and homeostasis. Toll-Like–Receptor Ligands Gut microbe-derived signals are recognized partly by pattern recognition receptors, such as Toll-like receptors (TLRs). TLR4-/- mice are characterized by abnormal intestinal motility and a reduced number of nitrergic (neuronal nitric oxide synthase [nNOS]+) inhibitory neurons, a phenotype similar to that observed in GF and antibiotic-treated animals.22 This phenotype also was reproduced in mice with ENS-specific deletion of MyD88, an adaptor molecule essential for TLR-mediated signal transduction, suggesting key roles of the TLR4 pathway on the development and functional organization of intestinal neural networks.22 A separate study showed that deletion of TLR2, which is expressed by enteric neurons, EGCs, and smooth muscle cells of the gut wall, also resulted in reduction of nNOS+ neurons and acetylcholine-esterase–stained fibers in the myenteric ganglia.35 The altered neurochemical profiles of enteric neurons in the gut of TLR2-deficient mice was accompanied by gut dysmotility and attenuated chloride production by intestinal explants. Interestingly, expression of glial markers, such as glial fibrillary acidic protein (GFAP) and S100β, also decreased in the myenteric plexus of mutant mice. Considering that probiotic and pathogenic bacteria up-regulate TLR2 expression on human EGCs,36 these studies suggest that the microbiota–TLR2 pathway promotes functional maturation of EGCs. Interestingly, expression of GDNF is reduced significantly in TLR2-deficient and microbiota-depleted mice, while administration of GDNF rescued the ENS deficits of these animals, suggesting that the effects of TLR/microbiota pathways on ENS development and homeostasis are mediated via mesenchyme-derived neurotrophic factors. Outside the gut, TLR4 regulates the expression of Sox10 and Foxd3, raising the possibility that these transcriptional regulators also are targets of TLR4 in the ENS.37, 38 Nevertheless, the molecular mechanisms by which TLR ligands control enteric neurogenesis during neonatal stages require further investigation.

tors. Outside the gut, TLR4 regulates the expression of Sox10 and Foxd3, raising the possibility that these transcriptional regulators also are targets of TLR4 in the ENS.37, 38 Nevertheless, the molecular mechanisms by which TLR ligands control enteric neurogenesis during neonatal stages require further investigation. Short-Chain Fatty Acids Gut microbes metabolize dietary fiber and resistant starch to produce a wide variety of metabolites, including short-chain fatty acids (SCFAs), which are used as nutrients by colonic epithelial cells and influence host physiology.39, 40, 41, 42 Recent studies have uncovered a role of SCFAs in the production of serotonin (5-hydroxytryptamine [5-HT]) by enterochromaffin cells (ECs) of the intestinal epithelium.43, 44 ECs are the largest source of serotonin in the body, which, among other functions, regulate GI motility and platelet function.45, 46 Spore-forming bacteria from healthy human beings and mouse microbiota increase colonic and serum 5-HT levels in GF mice (by increasing expression of the colonic biosynthetic enzyme tryptophan hydroxylase 1-Tph143) and ameliorate GF-associated gut dysmotility. These bacteria produce SCFAs,39, 47 which are capable of increasing 5-HT production by cultured chromaffin cells43 and up-regulating Tph1 expression in a human-derived EC cell line.44 In addition, the extracellular availability of 5-HT within the gut is regulated by the serotonin-selective reuptake transporter (SERT) which is expressed by intestinal epithelial cells.48 Expression of SERT is lower in neonatal gut in comparison with adult tissues, resulting in higher availability of 5-HT, which during these early stages is essential for the maturation of intestinal motor reflexes.48, 49 SERT expression by intestinal epithelial cells also is regulated by microbiota-derived factors such as TLR ligands because treatment of an epithelial cell line (Caco-2) with lipopolysaccharide diminishes the expression and activity of SERT.50 These observations suggest that different kinds of microbial factors (eg, SCFAs and lipopolysaccharide) contribute to functional maturation of ENS by regulating the production and availability of 5-HT in a coordinated way.

lial cell line (Caco-2) with lipopolysaccharide diminishes the expression and activity of SERT.50 These observations suggest that different kinds of microbial factors (eg, SCFAs and lipopolysaccharide) contribute to functional maturation of ENS by regulating the production and availability of 5-HT in a coordinated way. SCFAs also can influence the neurochemical phenotype of the ENS in adult rat.51 Resistant starch diet (RSD), which enhances luminal SCFA concentration,52 specifically increased the proportion of excitatory cholinergic neurons in the colon, but had no effect on nNOS+ neurons, resulting in decreased colonic transit time. Intrarectal administration of butyrate, but not acetate or propionate, mimicked the effect of RSD. Of note, the butyrate-induced increase in excitatory choline acetyltransferase (ChAT)+ neurons depended on the butyrate transporter monocarboxylate transporter (MCT)2, which is expressed by enteric neurons,51 but the factors regulating neuronal MCT2 expression remain unknown. A recent study has shown that the EGC cell line, JUG-2, also expresses MCT1 and MCT2,53 although the physiological role of these enzymes on glial homeostasis in vivo has not been determined.

transporter (MCT)2, which is expressed by enteric neurons,51 but the factors regulating neuronal MCT2 expression remain unknown. A recent study has shown that the EGC cell line, JUG-2, also expresses MCT1 and MCT2,53 although the physiological role of these enzymes on glial homeostasis in vivo has not been determined. SCFAs also activate G-protein–coupled receptors, such as GPR41 and GPR43.54, 55 Analysis of transgenic reporter mice has shown that GPR41 is expressed by ECs and enteric neurons,55 although the role of neuronal GPR41 on ENS function has not been characterized. GPR43 is expressed by intestinal immune cells and sporadically by ECs.55 Activation of GPR43 on enteroendocrine cells by SCFAs promotes secretion of the incretin hormone glucagon-like peptide-1 (GLP-1),56 which also controls gastric emptying and gastrointestinal transit.57 Moreover, RSD is known to increase GLP-1 expression by cecal and colonic epithelial cells in vivo,58 suggesting that gut microbiota may increase GLP-1 levels through the SCFAs/GPR43/GPR41 pathway. An additional pathway inducing GLP-1 by microbiota is mediated by intestinal Escherichia coli–derived protein caseinolytic protease B. Caseinolytic protease B serves as an antigen mimetic of the α-melanocyte–stimulating hormone59 and is capable of stimulating melanocortin receptor 4 on enteroendocrine L cells to produce GLP-1.60 However, a recent study showed that GF mice (which lack colonic SCFAs) show significantly higher levels of GLP-1 in the plasma, while colonization of GF mice with microbiota or treatment with SCFAs reduced GLP-1 expression in the colon.61 Interestingly, blocking the GLP-1 signaling with exendin 9-39 (Ex-9) (GLP-1 antagonist) completely rescued the slow intestinal transit observed in GF mice, and antibiotic-dependent reduction of intestinal transit was rescued by deletion of GLP-1, indicating that GLP-1 signaling is required to slow intestinal transit in GF conditions. Perhaps slower intestinal transit provides extra time for nutrient absorption during insufficient colonic energy availability (lack of SCFAs) under GF conditions. Further studies will be necessary to determine how SCFAs regulate the levels of GLP-1 and how GLP-1 controls the activity of intestinal neural circuits.

ons. Perhaps slower intestinal transit provides extra time for nutrient absorption during insufficient colonic energy availability (lack of SCFAs) under GF conditions. Further studies will be necessary to determine how SCFAs regulate the levels of GLP-1 and how GLP-1 controls the activity of intestinal neural circuits. An alternative mechanism by which SCFAs could affect host gene expression is the inhibition of histone deacetylase activity.62, 63 Histone deacetylase inhibition enhances histone acetylation of gene regulatory elements and increases gene transcription.64 The epigenetic regulation of the immune system by gut microbial butyrate has been shown in colonic T cells39, 65 and macrophages.66 Although little is known about the epigenetic modifications of ENS by SCFAs, butyrate treatment enhances acetylation of the H3K9 in primary cultured enteric neurons and the EGC cell line JUG-2.51, 53 It would be interesting to determine the extent to which microbiota-derived SCFAs modulate the epigenetic status of genes and its role in enteric neurogenesis and gliogenesis.

ions of ENS by SCFAs, butyrate treatment enhances acetylation of the H3K9 in primary cultured enteric neurons and the EGC cell line JUG-2.51, 53 It would be interesting to determine the extent to which microbiota-derived SCFAs modulate the epigenetic status of genes and its role in enteric neurogenesis and gliogenesis. Bile Acid Metabolism and Dietary Factors Gut microbes also participate in the conversion of primary bile acids synthesized de novo in the liver into secondary bile acids.17 Secondary bile acids can activate the G-protein–coupled bile acid receptor 1, which is known as TGR5. TGR5 is highly expressed in enteric neurons and enteroendocrine L cells67 and TGR5-deficient mice showed delayed colonic transit and reduced defecation frequency relative to wild-type mice.68 In addition, stimulation with TGR5 agonists induced colonic peristalsis in wild-type but not TGR5-deficient mice, suggesting the important role of TGR-5 signaling on intestinal propulsive activity. TGR5-dependent enhancement of peristalsis could be mediated partly by production of 5-HT, because stimulation of isolated distal colon with bile acids increased 5-HT production.68 Although a better mechanistic understanding is required, targeting of TGR5 emerges as a potential therapeutic strategy to alleviate symptoms of constipation and diarrhea.

peristalsis could be mediated partly by production of 5-HT, because stimulation of isolated distal colon with bile acids increased 5-HT production.68 Although a better mechanistic understanding is required, targeting of TGR5 emerges as a potential therapeutic strategy to alleviate symptoms of constipation and diarrhea. Diet ingredients also can influence gut motility and ENS function in combination with microbiota. A recent study identified deconjugated bile acids as fecal metabolites associated with intestinal transit time phenotype.69 Consumption of a Bangladeshi diet containing turmeric, a spice that increases luminal bile acids, significantly slowed motility in mice that had been colonized with microbiota isolated from a Bangladeshi donor, compared with a turmeric-free Bangladeshi diet, suggesting that a single food ingredient can influence gut motility. Because deconjugation of bile acids is mediated by activity of bacterial bile salt hydrolases (BSHs), these investigators generated gnotobiotic mice composed of either BSH-positive or BSH-negative bacterial consortia cultured from the microbiota of a Bangladeshi donor. Intestinal transit time significantly decreased in mice colonized with the BSH-positive microbiota than BSH-negative microbiota only when they were fed a turmeric-containing diet, indicating that the motility phenotypes are dependent on the abilities of microbiota to deconjugate the bile acids produced by turmeric consumption. Interestingly, this phenotype was not observed in mice heterozygous for a Ret null mutation, suggesting that the effects of turmeric on motility phenotypes depend on both genetic background and bacterial bile acid metabolism.69 Taken together, gut motility is influenced in a coordinated manner by the interaction between luminal environmental factors (eg, diet, SCFAs, bile acids, TLR ligands), microbial factors (composition and metabolism activity), and host factors (nutritional condition and functional ENS) (Figure 1).

bile acid metabolism.69 Taken together, gut motility is influenced in a coordinated manner by the interaction between luminal environmental factors (eg, diet, SCFAs, bile acids, TLR ligands), microbial factors (composition and metabolism activity), and host factors (nutritional condition and functional ENS) (Figure 1). Bidirectional Communication Between the Gut Immune System and ENS The GI tract harbors the highest concentration of immune cells in the body. In particular, macrophages, which have key roles in innate immune responses and tissue homeostasis, are present in intestinal muscularis (called muscularis macrophages [MMs]) and are in close contact with ENS cells.70 MMs have a unique surface marker profile (CX3CR1hiMHCIIhiCD11cloCD103-CD11b+) and their development is dependent on colony stimulatory factor (CSF)1 receptor, a receptor for macrophage CSF that regulates mononuclear phagocyte development.71 Interestingly, treatment of adult mice with the anti-CSF1R antibody induced depletion of MMs and resulted in gut dysmotility, manifested as colonic hyperactivity and increased colonic transit time. Bone marrow chimeric mice with Csf1-receptor–deficient hematopoietic progenitors also showed increased colonic transit time, indicating the importance of MMs on the physiological control of GI motility. MM-dependent control of gut motility is mediated by bone morphogenetic protein (BMP)2, which is expressed by MMs and activates enteric neurons.72, 73 Administration of the BMP signaling inhibitor dorsomorphin reproduced the phenotype of MM-depleted mice whereas exogenous BMP2 rescued the dysmotility of these animals. Remarkably, enteric neurons selectively express BMP receptor type II, a component of the BMP receptor, and can produce CSF1 that is required for MM development. CSF1-deficient mice harbored an increased number of enteric neurons and showed a less-organized ENS architecture. These results suggest that neuronal CSF1 contributes to the homeostasis of MMs, which then are required for normal ENS activity. Interestingly, production of CSF1 and BMP2 is dependent on gut microflora because antibiotic treatment decreased the production of both signaling mediators.71

less-organized ENS architecture. These results suggest that neuronal CSF1 contributes to the homeostasis of MMs, which then are required for normal ENS activity. Interestingly, production of CSF1 and BMP2 is dependent on gut microflora because antibiotic treatment decreased the production of both signaling mediators.71 More recently, Gabanyi et al74 performed RNA sequencing–based transcriptome analysis of MMs and lamina propria macrophages, and showed that MMs preferentially express tissue-protective and wound-healing genes resembling alternatively activated (M2-type) macrophages, while lamina propria macrophages express proinflammatory genes. Interestingly, MMs express Adrb2 (encoding β2 adrenergic receptors), which is essential for norepinephrine signaling, and reside in close proximity to enteric neurons labeled with the calcium indicator GCaMP3, suggesting that MMs interact with active neurons in gut muscularis. Of note, intestinal infection with a mutant strain of Salmonella typhimurium activated tyrosine hydroxylase–expressing extrinsic neurons in the sympathetic ganglia innervating the gut, leading to production of norepinephrine in the muscular layer, which was accompanied by a significant increase in intestinal transit time. The noradrenaline signaling on MMs through β2 adrenergic receptors also contributed to their polarization into M2-type–related phenotype. These studies indicate that microbiota-driven interactions between innate immune cells and the ENS control gut motility and enhance the tissue-protective phenotype in MMs in response to intestinal infection, even in sites distal from an initial pathogen entry.74

ted to their polarization into M2-type–related phenotype. These studies indicate that microbiota-driven interactions between innate immune cells and the ENS control gut motility and enhance the tissue-protective phenotype in MMs in response to intestinal infection, even in sites distal from an initial pathogen entry.74 In conclusion, these recent studies provide further support for the concept that specialized interactions between the ENS and the gut immune system are essential for GI tract homeostasis (Figure 2). Given the capacity of the nervous system to respond rapidly to diverse stimuli by releasing neurotransmitters and neuropeptides, which include among their targets immune cell functions,75 it is conceivable that enteric neural reflexes are an integral part of early response mechanisms that operate continuously to restore the balance between innocuous and pathogenic micro-organisms and thus maintain the symbiotic host–microbe relationships.

tides, which include among their targets immune cell functions,75 it is conceivable that enteric neural reflexes are an integral part of early response mechanisms that operate continuously to restore the balance between innocuous and pathogenic micro-organisms and thus maintain the symbiotic host–microbe relationships. Parkinson’s Disease: A Disease of the Microbiota–Gut–Brain Axis? The critical role of the ENS in controlling GI tract physiology is highlighted by the high morbidity of congenital enteric neuron deficits, such as Hirschsprung’s disease.76 Acquired dysmotility syndromes, such as irritable bowel syndrome (IBS), also are attributed to ENS deficits, although additional local factors, including luminal microbiota, mucosal immune cells, epithelial barrier functions, and serotonin metabolism have been implicated in the pathogenesis of this condition.77 IBS also is associated with deficits in the bidirectional gut–brain communication,78 and studies on this condition are likely to provide insight into the role on the MGB axis in health and disease. A detailed presentation of the relationship between IBS and the MGB axis is beyond the scope of this review and the reader is directed to excellent recent literature.77, 79 Here, we highlight an emerging hypothesis implicating ENS deficits in the pathogenesis of neurodegenerative diseases, including Parkinson’s disease (PD).80 PD is characterized by selective degeneration of dopaminergic neurons in the midbrain substantia nigra, and the abnormal deposition of α-synuclein (Lewy bodies) in the surviving dopaminergic neurons, resulting in characteristic motor symptoms.81 However, a high percentage of PD patients also are characterized by abnormal GI motility and constipation.82 Interestingly, PD-associated α-synuclein accumulations also are found in enteric neurons, which precede the development of motor symptoms by several years,83 suggesting that the ENS is an initial site of α-synuclein aggregations, which subsequently spread to the brain through vagus nerve fibers.

ity and constipation.82 Interestingly, PD-associated α-synuclein accumulations also are found in enteric neurons, which precede the development of motor symptoms by several years,83 suggesting that the ENS is an initial site of α-synuclein aggregations, which subsequently spread to the brain through vagus nerve fibers. In support of this notion, the risk of PD is lower in vagotomized individuals in comparison with the healthy population.84 Furthermore, a recent study has shown that α-synuclein injected into the gut wall can translocate into the dorsal motor nucleus of the vagus nerve via vagus nerve fibers in a time-dependent manner.85 Nevertheless, it remains unclear whether in PD patients intestinal PD pathology spreads to the brain and initiates motor symptoms, and how luminal factors (eg, microbiota and diet) could influence the potential gut–brain translocation, severity of intestinal symptoms, and loss of midbrain dopaminergic neurons. Interestingly, PD patients show dysbiosis, which is correlated with the clinical phenotype,86 although it has not been determined whether the observed changes of microbiota contribute to the pathogenesis of the disease or instead are the consequence of PD-associated nonmotor symptoms. The other alterations found in the gut of PD patients was an abnormal increase of proinflammatory cytokine genes and the glial cell markers GFAP, SOX10, and S100β,87 suggesting an association of intestinal inflammation and glial dysregulation with PD development. In addition, colonic biopsy specimens from PD patients showed the presence of enteric glial reactivity characterized by the up-regulation of GFAP expression but a reduction in phosphorylation,88 although the pathophysiological significance of these abnormalities remains unknown. Given that PD patients are diagnosed only after the onset of motor symptoms and are not treated until significant loss of dopaminergic neurons already has occurred, intestinal PD pathology could be an early and potentially useful biomarker for this condition.

logical significance of these abnormalities remains unknown. Given that PD patients are diagnosed only after the onset of motor symptoms and are not treated until significant loss of dopaminergic neurons already has occurred, intestinal PD pathology could be an early and potentially useful biomarker for this condition. Conclusions and Future Perspectives Accumulating evidence suggests that the development and function of ENS is controlled by luminal microbial factors and the host immune system. In addition to the importance of ENS on GI homeostasis, ENS also serves as a relay station along the MGB axis that conveys information from the luminal microenvironment to the CNS. The mechanisms underlying MGB axis communication involve the immune and endocrine system, neural pathways via the vagus nerve, and the microbiota-dependent modulation of CNS.8, 89, 90, 91, 92, 93, 94, 95 For instance, SCFAs produced by microbiota ensure the integrity of the blood-brain barrier by up-regulating tight junction proteins,42 and regulate the maturation and activation of microglial cells.41 On the other hand, the mechanism directly controlling the neural pathway connecting the CNS and ENS by microbial factors remains elusive. Considering that defects in ENS cause the development of CNS diseases, understanding the molecular mechanism of microbiota–ENS interactions could help us generate novel therapeutic strategies for multiple types of neurodegenerative diseases. Conflicts of interest The authors disclose no conflicts.

Conclusions and Future Perspectives Accumulating evidence suggests that the development and function of ENS is controlled by luminal microbial factors and the host immune system. In addition to the importance of ENS on GI homeostasis, ENS also serves as a relay station along the MGB axis that conveys information from the luminal microenvironment to the CNS. The mechanisms underlying MGB axis communication involve the immune and endocrine system, neural pathways via the vagus nerve, and the microbiota-dependent modulation of CNS.8, 89, 90, 91, 92, 93, 94, 95 For instance, SCFAs produced by microbiota ensure the integrity of the blood-brain barrier by up-regulating tight junction proteins,42 and regulate the maturation and activation of microglial cells.41 On the other hand, the mechanism directly controlling the neural pathway connecting the CNS and ENS by microbial factors remains elusive. Considering that defects in ENS cause the development of CNS diseases, understanding the molecular mechanism of microbiota–ENS interactions could help us generate novel therapeutic strategies for multiple types of neurodegenerative diseases. Conflicts of interest The authors disclose no conflicts. Funding Work in Vassilis Pachnis's laboratory is funded by the Francis Crick Institute and the BBSRC (Biotechnology and Biological Sciences Research Council). Also supported by a long-term EMBO (European Molecular Biology Organization) fellowship (Y.O.).

Conclusions and Future Perspectives Accumulating evidence suggests that the development and function of ENS is controlled by luminal microbial factors and the host immune system. In addition to the importance of ENS on GI homeostasis, ENS also serves as a relay station along the MGB axis that conveys information from the luminal microenvironment to the CNS. The mechanisms underlying MGB axis communication involve the immune and endocrine system, neural pathways via the vagus nerve, and the microbiota-dependent modulation of CNS.8, 89, 90, 91, 92, 93, 94, 95 For instance, SCFAs produced by microbiota ensure the integrity of the blood-brain barrier by up-regulating tight junction proteins,42 and regulate the maturation and activation of microglial cells.41 On the other hand, the mechanism directly controlling the neural pathway connecting the CNS and ENS by microbial factors remains elusive. Considering that defects in ENS cause the development of CNS diseases, understanding the molecular mechanism of microbiota–ENS interactions could help us generate novel therapeutic strategies for multiple types of neurodegenerative diseases. Conflicts of interest The authors disclose no conflicts. Funding Work in Vassilis Pachnis's laboratory is funded by the Francis Crick Institute and the BBSRC (Biotechnology and Biological Sciences Research Council). Also supported by a long-term EMBO (European Molecular Biology Organization) fellowship (Y.O.). Figure 1 Microbiota and diet control the activity of multiple cell types in the gut wall, including the ENS. For example, the bacterial metabolites SCFAs activate G-protein coupled receptors (eg, GPR41 and GPR43) on enteroendocrine cells of the intestinal epithelium resulting in enhanced production of GLP-1 and 5-HT and changes in gut motility. Gut microbiota also contribute to the conversion of primary bile acids into secondary bile acids, which activate TGR5 expressed by enteroendocrine cells and enteric neurons. TLR signalling (eg, TLR2 and TLR4) maintains subsets of enteric neurons and influences gut motility. In addition, microbiota is essential for the maintenance of mucosal glial cells, which express the neurotrophic factor GDNF and GFAP. 5-HT, Serotonin; α-MSH, α-melanocyte-stimulating hormone; GDNF, glial cell-derived neurotrophic factor; GFAP, glial fibrillary acidic protein; GLP-1, glucagon-like peptide-1; SERT, serotonin-selective reuptake transporter; Tph1, tryptophan hydroxylase 1.

glial cells, which express the neurotrophic factor GDNF and GFAP. 5-HT, Serotonin; α-MSH, α-melanocyte-stimulating hormone; GDNF, glial cell-derived neurotrophic factor; GFAP, glial fibrillary acidic protein; GLP-1, glucagon-like peptide-1; SERT, serotonin-selective reuptake transporter; Tph1, tryptophan hydroxylase 1. Figure 2 BMP2 from muscularis macrophages (MMs) regulates the activity of enteric neurons (by activating BMPRII) while CSF1 from enteric neurons is essential for the development of MMs (which express CSF1R). Production of CSF1 and BMP2 is dependent on gut microbiota. Activation of MMs by norepinephrine (via β2 adrenergic receptors) contributes to their polarization into M2-type phenotype, which is associated with tissue homeostasis and wound healing. BMP2, bone morphogenetic protein 2; β2AR: β2 adrenergic receptors, CSF1, colony stimulating factor 1; NE, norepinephrine.

See Covering the Cover synopsis on page 1459. Editor's Notes Background and Context Although HBV infects hepatocytes, evasion and modulation of the immune response is key to the subsequent physiopathology and the mechanisms of disease progression are unclear. New Findings Chronic HBV immunopathology is modeled in a dually humanized mouse developing a complete viral life cycle and immune responses proportionally to the dose of viral inoculum. Limitations HBV-infected mice do not develop fibrosis. Impact Liver-specific immune responses are investigated during chronic HBV and innovative therapeutic approaches can be tested in this small animal model.

New Findings Chronic HBV immunopathology is modeled in a dually humanized mouse developing a complete viral life cycle and immune responses proportionally to the dose of viral inoculum. Limitations HBV-infected mice do not develop fibrosis. Impact Liver-specific immune responses are investigated during chronic HBV and innovative therapeutic approaches can be tested in this small animal model. With over 350 million people infected by hepatitis B virus (HBV) and despite an efficient prophylactic vaccine, HBV prevalence is on the rise, constituting a major global health burden. The sequelae of HBV infection include acute and chronic hepatitis, which may lead to the development of fibrosis, cirrhosis, and hepatocellular carcinomas (HCC). The clinical course depends in part on the age of the host: most neonatally acquired infections develop into chronicity, whereas in adults they are mostly self-resolving.1 Current treatment strategies lead to viral suppression, but rarely establish a “functional cure” with HBsAg (hepatitis B surface antigen) loss. Furthermore, these life-long therapies do not eradicate the virus, nor do they efficiently train the immune response to control viremia, since viral rebound is frequent following discontinuation of treatment.2, 3 HBV is not directly cytopathic for hepatocytes, disease pathophysiology is conditioned by the ensuing anti-viral immune response.4 The breadth and scope of this response is essential for viral eradication; therefore, understanding the mechanisms that lead to viral clearance or persistence represents an essential goal.

HBV is not directly cytopathic for hepatocytes, disease pathophysiology is conditioned by the ensuing anti-viral immune response.4 The breadth and scope of this response is essential for viral eradication; therefore, understanding the mechanisms that lead to viral clearance or persistence represents an essential goal. Viral clearance is the result of a coordinated immune response initially mediated by Kupffer, natural killer (NK), and antigen-presenting cells, leading to robust and polyclonal CD4+ and CD8+ T-cell responses and the development of neutralizing anti-HBs antibodies that ensures protective immunity to functionally cured patients, despite the maintenance of covalently closed circular DNA (cccDNA).4, 5 In contrast, chronic viremia is associated with increased immunosuppressive cytokines (TGF-beta, IL-10) and regulatory T cells, combined with functional exhaustion of virus-specific T cells expressing programmed death-1 (PD-1), cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) and tumor necrosis factor (TNF)-related apoptosis-inducing ligand death receptor (TRAIL-R2), targeting them for NK cell-mediated deletion.4, 6, 7, 8 In chronic HBV (CHB) patients, an increased risk of progression to cirrhosis and HCC has been correlated with the viral load; however, the mechanisms driving these virally induced pathologies are unclear.9, 10, 11

optosis-inducing ligand death receptor (TRAIL-R2), targeting them for NK cell-mediated deletion.4, 6, 7, 8 In chronic HBV (CHB) patients, an increased risk of progression to cirrhosis and HCC has been correlated with the viral load; however, the mechanisms driving these virally induced pathologies are unclear.9, 10, 11 Deciphering the cross-talk between the immune system and HBV-infected hepatocytes has been hampered by the viruses’ restricted tropism to humans and chimpanzees. Although HBV transgenic or transduced mice, and HBV-infected human liver chimeric mice, are valuable for exploring the viral life cycle and testing antiviral compounds, they have limited immune responses and cannot assay immunotherapeutic approaches because significant differences between mice and humans predominate.12, 13 To establish an immunocompetent animal model for hepatotropic infections, several mouse models harboring both a humanized immune system and human hepatocytes have been recently described.14, 15, 16, 17, 18, 19 Although HBV infection of these models resulted in immune responses, low levels of liver chimerism limited viral output and the analysis of chronic pathophysiologic responses.15, 19

s, several mouse models harboring both a humanized immune system and human hepatocytes have been recently described.14, 15, 16, 17, 18, 19 Although HBV infection of these models resulted in immune responses, low levels of liver chimerism limited viral output and the analysis of chronic pathophysiologic responses.15, 19 Previously, we established a dually humanized mouse model in BALB/c Rag2-/-Il2rg-/-SirpaNOD Alb-uPAtg/tg recipients, stably engrafted with a humanized immune system and human hepatocytes (HIS-HUHEP mice).18 In this study, we demonstrate that HIS-HUHEP mice are susceptible to HBV, developing a full viral life cycle (HBeAg+, HBsAg+, cccDNA+) similar to CHB patients. By varying the viral inoculum in HIS-HUHEP mice, we found that distinct immune responses were obtained that led to different levels of viral control. Subsequent biomarker analysis revealed a cluster of effectors that may be useful to further dissect the complexity of HBV immune control. Nucleos(t)ide analog (NUC) treatment of chronically infected mice efficiently reduced viral loads, resulting in the resolution of liver inflammation.

fferent levels of viral control. Subsequent biomarker analysis revealed a cluster of effectors that may be useful to further dissect the complexity of HBV immune control. Nucleos(t)ide analog (NUC) treatment of chronically infected mice efficiently reduced viral loads, resulting in the resolution of liver inflammation. Materials and Methods Generation of Humanized Mice, HBV Infection, and NUC Treatment HUHEP and HIS-HUHEP mice were established in BALB/c Rag2-/-Il2rg-/-SirpaNODuPAtg/tg male and female mice as previously described.18 Mice with >100 μg/mL human Albumin (hAlbumin) and >10% hCD45+ cells in peripheral blood mononuclear cells (PBMC), including human T and B cells, were HBV infected intraperitoneally at 15 (±3) weeks with either 1x107 or 1x109 HBV genome equivalents (GE) purified from concentrated supernatants of the HepG2.2.15 stably producing cell line (HBV genotype D subtype ayw).20, 21 NUC-treated mice had been inoculated with 107 HBV GE for 14 ±1 weeks prior to receiving Entecavir at 0.3 mg/kg/day either injected intraperitoneally or dispensed in the drinking water (Baraclude; Bristol-Myers Squibb, Princeton, NJ). Animals were housed in isolators under pathogen-free conditions with humane care. Experiments were approved by an institutional ethical committee at the Institut Pasteur (Paris, France) and validated by the French Ministry of Education and Research (MENESR #02162.02).

r (Baraclude; Bristol-Myers Squibb, Princeton, NJ). Animals were housed in isolators under pathogen-free conditions with humane care. Experiments were approved by an institutional ethical committee at the Institut Pasteur (Paris, France) and validated by the French Ministry of Education and Research (MENESR #02162.02). Enzyme-Linked Immunosorbent Assay (ELISA) Analysis Species-specific enzyme-linked immunosorbent assays (ELISA) of plasma samples for human albumin, human IgM, and human IgG were performed as previously described.18 M65 ELISA was performed according to manufacturer’s specifications (M65 EpiDeath ELISA; Peviva, Nacka, Sweden) on plasma samples pre- and post-HBV infection, or the equivalent time points for controls. M65 levels were normalized to hAlbumin levels for each time point. Human cytokines and chemokines were quantified in mouse plasma with the Human Cytokine Magnetic 25-plex Panel (catalog # LHC0009M; Life Technologies, Carlsbad, CA) according to manufacturer's specifications and analyzed with MAGPIX (Luminex, Austin, TX).

65 levels were normalized to hAlbumin levels for each time point. Human cytokines and chemokines were quantified in mouse plasma with the Human Cytokine Magnetic 25-plex Panel (catalog # LHC0009M; Life Technologies, Carlsbad, CA) according to manufacturer's specifications and analyzed with MAGPIX (Luminex, Austin, TX). Flow Cytometry Analysis Mononuclear cells from blood, spleen, and liver were isolated as previously described.18 Cell labeling was performed with directly conjugated monoclonal antibodies against human CD3, CD4, CD8, CD14, CD16, CD33, CD45, CD45RA, CD45RO, CD56, CD69, HLA-DR, Nkp46, PD-1 (BD Biosciences, San Jose, CA; eBiosciences, San Diego, CA; Miltenyi Biotech, Bergisch Gladbach, Germany) according to standard techniques. Dead cells were excluded using Fixable viability dye (eBioscience). Intracellular labeling was performed as previously described18 (see Supplementary Materials and Methods). Acquisitions were performed using BD Fortessa and LSRII flow cytometers (Becton Dickinson, Franklin Lakes, NJ). Analysis was performed with FlowJo Version 8.8 (TreeStar, Ashland, OR) and Prism 6 (GraphPad, La Jolla, CA).

ellular labeling was performed as previously described18 (see Supplementary Materials and Methods). Acquisitions were performed using BD Fortessa and LSRII flow cytometers (Becton Dickinson, Franklin Lakes, NJ). Analysis was performed with FlowJo Version 8.8 (TreeStar, Ashland, OR) and Prism 6 (GraphPad, La Jolla, CA). Immunofluorescence Analysis Liver cryostat sections were prepared as previously described22 and immunostained with antibodies against HBc (B0586 [Dako, Glostrup, Denmark] or C1-5 [Santa Cruz, Dallas, TX]), and human antigens: albumin (A0001 [Dako] or A80-129A [Bethyl]), CD3 (A0452 [Dako] or UCHT1 [eBiosciences]), CD45 (2D1 [BD Biosciences]), CD68 (KP1 [Dako]), CK7 (M7018 [Dako]) EpCAM (OP187 [Calbiochem, San Diego, CA]), PD-L1 (MAB1561 [R&D Systems]). Secondary antibodies were coupled to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (Molecular Probes, Eugene, OR). Photomicrographs were taken with an Axioimager Apotome microscope (Zeiss, Oberkochen, Germany).

[Dako]), CK7 (M7018 [Dako]) EpCAM (OP187 [Calbiochem, San Diego, CA]), PD-L1 (MAB1561 [R&D Systems]). Secondary antibodies were coupled to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (Molecular Probes, Eugene, OR). Photomicrographs were taken with an Axioimager Apotome microscope (Zeiss, Oberkochen, Germany). Virological Measurements HBV DNA was extracted from plasma and homogenized liver tissue using QIAamp DNA blood mini kit (Qiagen, Hilden, Germany). Plasmatic HBV DNA was quantified by quantitative polymerase chain reaction (qPCR) with a LightCycler system (Roche, Basel, Switzerland) as described by Brezillon et al.21 Intrahepatic HBV DNA was quantified by SYBR Green qPCR on an ABI PRISM 7900HT system (Applied Biosystems, Foster City, CA). For HBV cccDNA, DNA was pre-treated with 10 U plasmid-safe DNase (Epicenter, Madison, WI) for 1 hour at 37°C. HBV-specific primers were: 5′-GTTGCCCGTTTGTCCTCTAATTC-3′ and 5′-GGAGGGATACATAGAGGTTCCTTG-3′ for total HBV DNA; and 5′-GTGCACTTCGCTTCACCTCT-3′ and 5′-AGCTTGGAGGCTTGAACAGT-3′ for cccDNA amplification. Standard curves were generated from an HBV genome-containing plasmid (payw1.2), data was normalized to human cell numbers by qPCR amplification of the human IL8 promoter as described by Rivière et al.23 HBe and HBs antigens were quantified with a clinical test (DiaSorin, Saluggia, Italy).

T-3′ for cccDNA amplification. Standard curves were generated from an HBV genome-containing plasmid (payw1.2), data was normalized to human cell numbers by qPCR amplification of the human IL8 promoter as described by Rivière et al.23 HBe and HBs antigens were quantified with a clinical test (DiaSorin, Saluggia, Italy). Antigen-Specific Antibody Responses Clinical assays for HBsAb and HBcAb have detection thresholds well above the IgG antibody titers developed in HIS-HUHEP mice. HIS-HUHEP mouse plasma was assayed for antigen-specific antibodies by immunocytochemistry on transfected BHK-21 cells producing either HBs or HBc protein, or β-galactosidase as a control.24 BHK-21 cells were incubated with humanized mouse sera previously normalized to 80 μg/mL total human IgG followed by 2-fold serial dilutions (details provided in Supplementary Materials and Methods). Gene Expression Analysis by Quantitative Reverse-Transcription Polymerase Chain Reaction cDNA was prepared from 200 ng of total liver RNA with 2 steps using Superscript III RT (Invitrogen) for reverse transcription followed by TaqMan PreAmp Master Mix (Applied Biosystems) for target amplification with all TaqMan gene expression assays pooled at 0.2x. Subsequently, specific gene expression was performed using BioMark 48.48 Dynamic Arrays (Fluidigm, San Francisco, CA) following manufacturer's protocols. For each sample, 3 independent experiments were analyzed with 2 housekeeping genes (HPRT and GAPDH). Human specificity of each gene expression assay was verified on C57BL/6 mouse liver samples.

expression was performed using BioMark 48.48 Dynamic Arrays (Fluidigm, San Francisco, CA) following manufacturer's protocols. For each sample, 3 independent experiments were analyzed with 2 housekeeping genes (HPRT and GAPDH). Human specificity of each gene expression assay was verified on C57BL/6 mouse liver samples. Statistical Analysis Data sets were tested with 2-tailed unpaired Student t tests or Mann Whitney U tests, correlations were analyzed with Pearson's χ2 test using Prism version 6 (GraphPad Software, San Diego, CA). Significant P values are shown as: *P < .05, **P < .005, and ***P <.0005.

expression was performed using BioMark 48.48 Dynamic Arrays (Fluidigm, San Francisco, CA) following manufacturer's protocols. For each sample, 3 independent experiments were analyzed with 2 housekeeping genes (HPRT and GAPDH). Human specificity of each gene expression assay was verified on C57BL/6 mouse liver samples. Statistical Analysis Data sets were tested with 2-tailed unpaired Student t tests or Mann Whitney U tests, correlations were analyzed with Pearson's χ2 test using Prism version 6 (GraphPad Software, San Diego, CA). Significant P values are shown as: *P < .05, **P < .005, and ***P <.0005. Results Control of Viremia in Immunocompetent Humanized Mice Chronically Infected With HBV We analyzed the outcome of HBV infection after ‘high dose’ inoculation (109 HBV DNA copies) in HUHEP and HIS-HUHEP mice. Both models could be productively infected by HBV over the 4–5 month time course of the experiment (Figure 1, individual animals are shown in Supplementary Figure 1). While high levels of viral replication were evident in both models, viral progression was markedly different in the presence of human immune cells. Interestingly, viremia increased unchecked in HUHEP mice, whereas it stagnated in HIS-HUHEP mice, resulting in 10-fold lower viral titers in HIS-HUHEP compared with HUHEP mice (respectively, up to ≈108 vs 109 HBV DNA copies/mL); Figure 1A and 1B). We determined whether viral loads correlated with hAlbumin levels in HUHEP vs HIS-HUHEP models. In HUHEP mice, this correlation was significant (r2=0.51), whereas in immunocompetent humanized HIS-HUHEP mice there was no correlation (r2=0.07) (Figure 1C and 1D). Moreover, virus progression (ratio HBV DNA/hAlbumin) increased proportionally over time in the HUHEP mice (r2=0.72), but not in the HIS-HUHEP model (r2=0.03) (Figure 1E). The difference between HUHEP and HIS-HUHEP models was highly statistically significant (P < .0001) indicating that HBV infection was controlled in HIS-HUHEP mice, presumably through anti-HBV human immune responses. The cross-talk between infected hepatocytes and immune effectors may differ according to the viral load; chimpanzees inoculated with low or high doses of HBV had distinct intrahepatic immune responses, resulting in either a delayed or rapid viral clearance.25 We therefore analyzed HIS-HUHEP mice infected with a lower dose (107 HBV DNA copies) to assess the impact of HBV load on viral progression. Following infection with the lower inoculum, viremia was globally reduced (maximum ≈107 HBV DNA copies/mL), more efficiently controlled, and 1 HIS-HUHEP mouse cleared the infection (Supplementary Figures 2 and 3).Figure 1 Viral progression is controlled in immunocompetent HIS-HUHEP mice.

HBV load on viral progression. Following infection with the lower inoculum, viremia was globally reduced (maximum ≈107 HBV DNA copies/mL), more efficiently controlled, and 1 HIS-HUHEP mouse cleared the infection (Supplementary Figures 2 and 3).Figure 1 Viral progression is controlled in immunocompetent HIS-HUHEP mice. (A and B) HBV viremia (black line) and hAlbumin (grey bars) were measured in the plasma of HBV-infected HUHEP (A) and HIS-HUHEP (B) mice inoculated at 10e9. Means of HUHEP (n=4) and HIS-HUHEP (n=12) mice. (C and D) Analysis of viral load: for each plasma sample viremia was plotted against the hAlbumin concentration from HUHEP (C) (n=4) or HIS-HUHEP mice (D) (n=12). (E) Analysis of viral progression over time by linear regression analysis of the ratio of HBV DNA over hAlbumin concentration, plotted against time post infection, from HUHEP (dotted line) or HIS-HUHEP mice (filled line). The P value determines whether the slopes of each linear regression are significantly different from each other. (F and G) Plasma viral antigen loads (HBeAg and HBsAg) correlated with HBV viremia in HBV-infected HIS-HUHEP mice. (H, I, and J) Quantification of cccDNA and total HBV DNA in the liver of infected HIS-HUHEP mice. The ratio of total liver HBV DNA/cccDNA indicates the virus’ replicative activity (n=3). Bars show mininum to maximum, line at the mean. (K) Human hepatotoxicity (M65) was normalized to the degree of human liver chimerism (hAlbumin) per mouse. The fold change of M65/hAlbumin pre- and post-infection (19 ±4 wpi) are shown for each group (control n=8; HBV 10e7, n=7; HBV 10e9, n=4) Histograms show means and SEM. Statistical significance: Mann Whitney U test. Correlation analysis: r2 and P values calculated using 2-tailed Pearson's χ2 test.

(hAlbumin) per mouse. The fold change of M65/hAlbumin pre- and post-infection (19 ±4 wpi) are shown for each group (control n=8; HBV 10e7, n=7; HBV 10e9, n=4) Histograms show means and SEM. Statistical significance: Mann Whitney U test. Correlation analysis: r2 and P values calculated using 2-tailed Pearson's χ2 test. We characterized the viral life cycle in these HBV-infected humanized mouse models. Both HBeAg and HBsAg were within clinical ranges and correlated to HBV DNA viral titers (Figure 1F and 1G). cccDNA levels were higher in ‘high dose’ compared with ‘low dose’ inoculated mice (respectively, mean 1.9 vs 0.06 copies/cell; Figure 1H and 1I). However, the replicative activity (intrahepatic total HBV DNA/ cccDNA) was similar for both inocula (Figure 1J) and comparable to non-treated CHB patients.26

gure 1F and 1G). cccDNA levels were higher in ‘high dose’ compared with ‘low dose’ inoculated mice (respectively, mean 1.9 vs 0.06 copies/cell; Figure 1H and 1I). However, the replicative activity (intrahepatic total HBV DNA/ cccDNA) was similar for both inocula (Figure 1J) and comparable to non-treated CHB patients.26 In our previous work, we found that human hepatocyte graft function, measured as plasma hAlbumin, generally increased over time in non-infected HUHEP and HIS-HUHEP mice.18 Interestingly, hAlbumin also increased in HBV-infected HUHEP mice, but this failed to occur in HBV-infected HIS-HUHEP mice (Figure 1, Supplementary Figure 1, Supplementary Figure 2, and 3). In uPA-based models, active tissue remodeling of the host, independently of the grafted cells, precludes use of ALT measurements (alanine aminotransferase) due to its lack of species specificity.27 Hepatotoxicity results in soluble and fragmented Cytokeratin-18 measurable in the plasma with a human-specific antibody M65 by ELISA.28 In non-infected HIS-HUHEP mice, M65 was stable over time, yet levels increased following HBV infection, indicating that human hepatocyte cytolysis was occurring (Figure 1K). No significant differences were observed in control or HBV-infected HUHEP mice (Supplementary Figure 2).

a human-specific antibody M65 by ELISA.28 In non-infected HIS-HUHEP mice, M65 was stable over time, yet levels increased following HBV infection, indicating that human hepatocyte cytolysis was occurring (Figure 1K). No significant differences were observed in control or HBV-infected HUHEP mice (Supplementary Figure 2). Cellular Immune Responses in HIS-HUHEP Mice With Chronic HBV-Infection To assess human immune responses to HBV infection in situ, we immunostained liver sections of infected and control HIS-HUHEP mice. Most human hepatocytes were infected (hAlb+ HBcAg+) and strong inflammation (hCD45) composed of T cells (hCD3) and Kupffer cells (hCD68) was observed in the parenchyma of ‘high dose’ infected animals (Figure 2). Inflammatory foci of hCD45+ cells were grouped adjacent to HBcAg+ hepatocytes. Specifically, CD3+ T cells were concentrated within clusters of infected hepatocytes, whereas CD68+ Kupffer cells were present throughout the liver parenchyma and intertwined between infected hepatocytes (Figure 2). In ‘low dose’ infected mice, fewer hepatocytes were HBcAg+, and a diffuse recruitment of CD3+ T cells and CD68+ Kupffer cells to the liver was apparent (Supplementary Figure 4). Histologically, liver fibrosis was not observed and α-smooth muscle actin staining was normal (data not shown). Human CD45+ cellularity was significantly increased in the liver of HIS-HUHEP mice following HBV infection (Figure 3A). HBV-infected mice demonstrated a striking increase in the numbers of hepatic NK cells (hCD45+CD3-NKp46+) (for low and high dose, respectively, in the liver, 12x and 16x; in the spleen, 3x and 7x) (Figure 3B). More NK cells were activated (CD69+) and showed a mature phenotype (CD56+CD16+) in HBV-infected mice compared with controls (Figure 3C and 3D). Liver-derived NK cells from high-dose HBV-infected mice produced significantly more IFN-γ and TNF-α than from control mice after stimulation with IL-12, IL-15, and IL-18, suggesting heightened responsiveness subsequent to viral stimulation in vivo (Figure 3E). A similar trend was observed in the spleen, although the differences were not significant.Figure 2 Human immune cells are recruited to the liver in chronically infected HIS-HUHEP mice. Immunofluorescence analysis of liver sections from control (top panels) and HBV-infected (bottom panels) mice co-stained for hAlbumin (blue) and HBc (green), with either hCD45 (red), or hCD3 (red), or hCD68 (red). DAPI-stained nuclei are shown in grey. Scale bar represents 100 μm.

n chronically infected HIS-HUHEP mice. Immunofluorescence analysis of liver sections from control (top panels) and HBV-infected (bottom panels) mice co-stained for hAlbumin (blue) and HBc (green), with either hCD45 (red), or hCD3 (red), or hCD68 (red). DAPI-stained nuclei are shown in grey. Scale bar represents 100 μm. Figure 3 NK-cell mediated immune response in HBV-infected HIS-HUHEP mice. (A) Absolute numbers of total human leukocytes (hCD45+) and (B) NK cells (CD3-NKp46+) from the liver and spleen of control and HBV-infected mice. Representative FACS plots show the percent positive cells in each gate. (C) Frequency of CD69+ cells among NK cells in the liver and spleen from control (grey filled line) and HBV-infected (black open line) HIS-HUHEP mice. The percentage of positive cells from HBV-infected mice is shown. (D) Expression of CD56 and CD16 in NK cells from the liver of control or HBV-infected HIS-HUHEP mice. (E) Liver leukocytes or splenocytes were restimulated ex vivo overnight and analyzed by FACS for the expression of IFN-γ or TNF-α by NK cells. Histograms show the mean and SEM. In plots each dot represents a mouse, data obtained at 14–20 wpi. Statistical analysis in A–D: Mann Whitney U test, E: 2 way ANOVA.

HUHEP mice. (E) Liver leukocytes or splenocytes were restimulated ex vivo overnight and analyzed by FACS for the expression of IFN-γ or TNF-α by NK cells. Histograms show the mean and SEM. In plots each dot represents a mouse, data obtained at 14–20 wpi. Statistical analysis in A–D: Mann Whitney U test, E: 2 way ANOVA. Persistence of viral antigens during chronic hepatitis induces dysfunctional hyporesponsive T cells with increased expression of inhibitory molecules PD-1 and CTLA-4.8, 29, 30 We found that HBV-infected HIS-HUHEP mice had increased numbers of CD3+ T cells in the liver (low dose, 7x; high dose, 6x) and spleen compared with controls (Figure 4A). Following infection, CD4+ TH cells were expanded in both the liver (5x for both low and high doses) and spleen. Furthermore, CD8+ TC cell numbers increased massively in the liver (low dose, 9x; high dose, 6x), which is consistent with their role as key effectors in HBV clearance (Figure 4B). Strikingly, in the liver of high-dose inoculated mice, the majority of T cells shifted from a naïve (CD45RA+) to a memory phenotype (CD45RO+) (Figure 4C), resulting in the generation of an increased pool of effector memory cells (CD8+CD45RO+HLA-DR+) (Supplementary Figure 5). Liver-derived memory CD4+ and CD8+ cells in high-dose inoculated mice showed an exhausted phenotype with increased expression of PD-1 (Figure 4D). Interestingly, the frequency of PD-1+ memory CD4+CD45RO+ T cells correlated to viral loads in the liver of low-dose inoculated mice (Figure 4E).Figure 4 Characterization of human T-cell subsets during chronic HBV infection. (A) Total numbers of human T lymphocytes (hCD45+CD3+) and (B) CD4+ or CD8+ T cells isolated from the liver and spleen. (C) Representative FACS plots of CD4+ or CD8+ T cells from the liver of control or HBV-infected samples, histograms show normalized frequencies of naïve (CD45RA+ [white bar]) or memory (CD45RO+ [black bar]) cells among CD4+ or CD8+ T cells. Mean and SEM are shown. (D) Frequency of PD-1+ cells among the CD4+CD45RO+ or CD8+ CD45RO+ memory T cells from a representative HIS-HUHEP control (grey filled line) and HBV 10e9 inoculated (black empty line) mouse liver. Each dot represents a mouse, data obtained at 14–20 wpi. A–D: Mann Whiteny U test. (E) Analysis of an exhaustion marker (PD-1+) on intrahepatic CD4+CD45RO+ T cells as a function of viral load with Pearson’s correlation test.

ve HIS-HUHEP control (grey filled line) and HBV 10e9 inoculated (black empty line) mouse liver. Each dot represents a mouse, data obtained at 14–20 wpi. A–D: Mann Whiteny U test. (E) Analysis of an exhaustion marker (PD-1+) on intrahepatic CD4+CD45RO+ T cells as a function of viral load with Pearson’s correlation test. We next analyzed PD-1 ligand (PD-L1) expression in the liver of HIS-HUHEP mice. In non-infected mice, PD-L1 was expressed by few human hepatocytes (hAlb+) as well as non-parenchymal cells (hAlb-) (Supplementary Figure 6). During chronic HBV infection, PD-L1+ cells were far more abundant and distributed among immune cells as well as infected and non-infected hepatocytes (HBcAg+ or HBcAg-) (Supplementary Figure 6). These results suggest that in the context of chronic HBV infection, enhanced expression of PD-L1 by hepatocytes and non-parenchymal cells may engage PD-1 on T cells, thereby contributing to their exhaustion.

among immune cells as well as infected and non-infected hepatocytes (HBcAg+ or HBcAg-) (Supplementary Figure 6). These results suggest that in the context of chronic HBV infection, enhanced expression of PD-L1 by hepatocytes and non-parenchymal cells may engage PD-1 on T cells, thereby contributing to their exhaustion. Antigen-Specific Antibody Responses in HBV-Infected HIS-HUHEP Mice Absolute numbers of splenic B cells (CD19+CD20+) were unchanged after ‘low dose’ or ‘high dose’ HBV infection (data not shown). Total hIgM levels were similar for control and ‘high dose’ infected mice, whereas in the ‘low dose’ inoculated mice antibody titers increased at 7–9 week post-infection (wpi) (Figure 5A). Class-switched hIgG antibodies increased significantly in both the ‘low dose’ and ‘high dose’ infected animals compared with controls, with an accelerated maturation in the ‘low dose’ inoculated mice (Figure 5A). We assessed anti-HBs (HBsAb) and anti-HBc (HBcAb) specific IgG antibody responses using a previously validated, highly sensitive immunohistochemistry assay.24 Both HBsAb and HBcAb responses were detected in the plasma in a majority of HBV-infected mice (low dose, 75%, n=4; high dose, 55%, n=9) (Figure 5B and 5C). HBV-specific antibody responses developed in a slightly higher fraction in ‘low dose’ vs ‘high dose’ infected mice. In line with previous reports,31 polyreactive IgGs could be detected in a few control HIS mice (16%, n=19). These results indicate that HIS-HUHEP mice generate HBV-specific IgG antibodies in response to infection.Figure 5 Humoral immune responses in HBV-infected HIS-HUHEP mice. (A) Total human IgM and IgG concentrations in the plasma of HIS-HUHEP control (white bar; n=11) or HBV-infected mice (10e7 inoculum: grey bar; n=5, 10e9 inoculum: black bar, n=9) plotted against weeks post infection (wpi). Bars show the mean with SEM. Statistics used unpaired two-tailed t test. (B) Analysis of anti-HBV antibodies in HIS-HUHEP mice. The frequency of mice positive for HBcAb IgG or HBsAb IgG from serially diluted plasma at the indicated concentration of total human IgG was analyzed by nonlinear regression. HIS-HUHEP control (n=19, dotted line), HBV-inoculated at 10e7 (n=4, grey line) or 10e9 (n=9, black line). P values indicate the differences between the slopes compared with the control data. (C) Representative images of IHC for anti-HBsAb.

indicated concentration of total human IgG was analyzed by nonlinear regression. HIS-HUHEP control (n=19, dotted line), HBV-inoculated at 10e7 (n=4, grey line) or 10e9 (n=9, black line). P values indicate the differences between the slopes compared with the control data. (C) Representative images of IHC for anti-HBsAb. Human Biomarker Disease Profiles for Chronic HBV in Humanized Mice Although HBV is considered a “stealth” virus, modest elevations in cytokine and chemokine levels are observed in HBV-infected patients during the acute phase or hepatic flares.32, 33, 34, 35 To identify serum biomarkers in HBV-infected mice, plasma was screened by multi-analyte profiling. In non-infected HUHEP mice, very low levels of IL-8, CCL4, IFN-α, CXCL9, and CXCL10 could be detected that did not significantly increase following HBV infection (Supplementary Figure 7). Similarly, in non-infected HIS-HUHEP mice, IL-1b, IL-12, IL-1Ra, IL-2R, CCL5, and IFN-γ, could be detected but did not significantly change post-infection. In contrast, HBV-infected HIS-HUHEP mice had significantly increased levels of inflammatory mediators IL-6, IL-8, CCL2, CCL3, CCL4, TNF-α, as well as IFN-α and the interferon stimulated genes CXCL9 and CXCL10 (Figure 6A). Furthermore, the immunosuppressive cytokine IL-10 was also upregulated (Figure 6A). Biomarker profiles from low-dose inoculated mice had a moderate inflammatory signature, with significant increases observed only in IL-6, IL-8, and CCL2. Interestingly, plasma levels of CXCL10, IL-10, and to a lesser extent IFN-α, correlated with the expression levels of PD-1 on memory CD4+ T cells (Figure 6B).Figure 6 Biomarker analysis in HBV-infected HIS-HUHEP mice. (A) Plasma from HIS-HUHEP control (n=11) and HBV-infected mice (inoculum 10e7, n=5; or 10e9, n=10) at endpoint were analyzed using human cytokine multiplex assay. No cross-reactivity with mouse cytokines was detected. Plasma levels of IL-1β, IL-1Ra, IL-2, IL-2R, IL-4, IL-12, IL-13, IL-15, IFN-γ, GM-CSF, and RANTES were similar between control and HBV-infected mice; while IL-5, IL-17, and Eotaxin were not detected (data not shown). (B) Correlation of PD-1+ memory CD4+ CD45RO+ T cells with either IFN-γ, or IP-10/CXCL10, or IL-10 plasma cytokines quantified in (A). Each dot represents a mouse: grey or black inoculated with, respectively, HBV 10e7 or HBV 10e9. (C) RT-qPCR analysis of liver samples from HIS-HUHEP control (n=14) and HBV-infected mice (inoculum 10e7, n=5; or 10e9, n=9).

+ CD45RO+ T cells with either IFN-γ, or IP-10/CXCL10, or IL-10 plasma cytokines quantified in (A). Each dot represents a mouse: grey or black inoculated with, respectively, HBV 10e7 or HBV 10e9. (C) RT-qPCR analysis of liver samples from HIS-HUHEP control (n=14) and HBV-infected mice (inoculum 10e7, n=5; or 10e9, n=9). Fold changes in gene expression of HBV-infected compared with control mice are shown. Data was normalized to the internal control human GAPDH (hGAPDH) to account for differences in humanization levels on triplicate samples. Dotted line indicates fold change of 1. Histograms show the mean and SEM. Data from 14–20 wpi. Statistical significance: Mann Whitney U tests.

d compared with control mice are shown. Data was normalized to the internal control human GAPDH (hGAPDH) to account for differences in humanization levels on triplicate samples. Dotted line indicates fold change of 1. Histograms show the mean and SEM. Data from 14–20 wpi. Statistical significance: Mann Whitney U tests. We further analyzed gene expression profiles in the liver. Expressions of pro-inflammatory molecules CCL3, CCL4, IL-8, IL-18, CXCL9, CXCL12, and of acute phase response genes IL-6, C-reactive protein, and serum amyloid A2, were up-regulated in HBV-infected HIS-HUHEP mice (Figure 6C and data not shown). Although a weak IFN-α response was detected in the plasma of low-dose infected HIS-HUHEP mice, a clear increase in interferon-stimulated genes ISG56, and RIG-I was observed in the livers (Figure 6C). Previous studies showed that NK and T cells eliminate HBV-infected hepatocytes via Fas (CD95) and TRAIL-mediated apoptosis.32, 36 In the liver of HIS-HUHEP infected mice, we also observed increased expression of Fas/CD95, TRAIL, and TRAIL-R (Figure 6C). HBV antigens act as immunosuppressors by inhibiting the activity of innate sensors Toll-like receptor (TLR) 7 and 9 that recognize viral nucleic acids.37, 38 Similarly, TLR7 and TLR9 expression were downregulated in HIS-HUHEP–infected livers (Figure 6C). Moreover, immunosuppressive mediators (IL-1Ra and PD-L1) were also overexpressed in HBV-infected mouse liver (Figure 6C). In HUHEP livers following HBV infection, IL-18 and CXCL-12 were not detected and C-reactive protein, SAA-2, ISG56, RIG-I, FAS, TRAIL, TRAIL-R, TLR7, TLR9, IL-1Ra, and PD-L1 were not significantly modulated (data not shown).

ve mediators (IL-1Ra and PD-L1) were also overexpressed in HBV-infected mouse liver (Figure 6C). In HUHEP livers following HBV infection, IL-18 and CXCL-12 were not detected and C-reactive protein, SAA-2, ISG56, RIG-I, FAS, TRAIL, TRAIL-R, TLR7, TLR9, IL-1Ra, and PD-L1 were not significantly modulated (data not shown). HBV-Infected Liver Progenitor Cells Inflammatory cytokines (IL-6, TNF-α, IFN-γ) produced by macrophages or activated T and NK cells have been shown to induce proliferation of liver-progenitor cells (LPC) in response to liver damage.22, 39, 40 The more aggressive HCC subtypes express LPC markers (EpCAM, CK7, CK19, CD133)41, 42 and increased incidences of HCC have been correlated to high viral loads in CHB patients.9 EpCAM+Alb+ and CK7+Alb+ LPCs were observed in non-infected HIS-HUHEP mice, as expected in the context of liver regeneration because of continuous expression of the uPA transgene (Supplementary Figure 8A).43 Interestingly, in HBV-infected mice, some LPC were HBc+ and the more dedifferentiated progenitor-like CK7+Albumin-HBc+ cells were only observed in the high-dose infected mice (Supplementary Figure 8B and 8C). These HBc+ LPCs could originate from infected hepatocytes that had dedifferentiated, or may have been infected after their dedifferentiation to LPCs. Some LPCs localized near CD3+ T cell clusters, suggesting inflammation may impact on their development (Supplementary Figure 8D).44

mice (Supplementary Figure 8B and 8C). These HBc+ LPCs could originate from infected hepatocytes that had dedifferentiated, or may have been infected after their dedifferentiation to LPCs. Some LPCs localized near CD3+ T cell clusters, suggesting inflammation may impact on their development (Supplementary Figure 8D).44 Testing Anti-Viral Therapies in HBV-Infected HIS-HUHEP Mice To determine whether suppression of viremia modifies intrahepatic immunophenotypes, we treated chronically HBV-infected HIS-HUHEP mice with the nucleoside analogue entecavir (ETV). ETV efficiently reduced viral loads (> 3 logs) in HBV-infected mice; viremia was undetectable in 2 of 4 treated mice (Figure 7A and Supplementary Figure 9). ETV-treated mice showed decreased liver inflammation with reduced monocytes and NK cells, including pro-inflammatory NK cells (CD56+CD16+), at levels comparable with noninfected mice (Figure 7B, C, and D). CD4+ and CD8+ T cellularity diminished and returned to a more naïve phenotype (CD45RA+), while the frequency of PD-1+ memory CD4+CD45RO+ cells tended to decrease (Figure 7E and Supplementary Figure 9). Among immunoregulatory mechanisms induced by HBV infection, TREG accumulation in the liver of CHB patients dampens the antiviral immune response.8 Interestingly, whereas HBV-infected HIS-HUHEP mice had more TREG CD4+CD25+Foxp3+ cells in the liver than controls, they were reduced after ETV treatment, and TREG cellularity correlated with viral loads (Figure 7F). Thus, ETV treatment resulted in decreased viral loads and restored naïve immune phenotypes in the liver.Figure 7 ETV treatment of HBV chronically infected HIS-HUHEP mice reverses hepatitis. (A) Liver engraftment (hAlbumin: grey bars) and viremia (HBV DNA: black line) in the plasma of mice previously infected for 3 months undergoing ETV treatment (n=4 mice). (B to F) FACS analysis of intrahepatic leukocytes and splenocytes from control, HBV-infected, and HBV-infected post-ETV treated mice. Absolute numbers of (B) total human leukocytes (hCD45+), (C) monocytes (HLA-DR+CD14+CD33+), (D) NK cells (CD3-NKp46+), and the frequency of CD56+CD16+ cells among NK cells, (E) CD4+, CD8+ (F), and Foxp3+ T cells. The number of intrahepatic Foxp3+ TREG cells was plotted against the viral load at end point for each mouse and analyzed with Pearson’s correlation test.

(hCD45+), (C) monocytes (HLA-DR+CD14+CD33+), (D) NK cells (CD3-NKp46+), and the frequency of CD56+CD16+ cells among NK cells, (E) CD4+, CD8+ (F), and Foxp3+ T cells. The number of intrahepatic Foxp3+ TREG cells was plotted against the viral load at end point for each mouse and analyzed with Pearson’s correlation test. Discussion HIS-HUHEP mice constitute a robust small animal model to investigate bi-directional interactions between the immune system and the liver during hepatotropic infections by HBV. The strength of this model relies on the combination of stable hepatic chimerism that supports high levels of HBV replication, coupled with multilineage development of myeloid and lymphoid cell subsets recruited to the liver following infection. Although HIS and HUHEP grafts are not HLA matched, non-infected HIS-HUHEP mice showed no signs of graft vs graft responses. No inflammation or hepatotoxicity was observed, most likely because of the timing of grafts and the tolerant liver environment.18 Nevertheless, HBV-infected HIS-HUHEP mice developed chronic hepatitis that persisted for several months with viral titers, HBe and HBs antigenemia, cccDNA, and anti-viral immune reactions that were similar to those observed in CHB patients. Inoculation with lower or higher doses in HIS-HUHEP mice resulted in viremia similar to patient samples categorized as moderate vs elevated.45 Taken together, HIS-HUHEP mice represent a versatile system in which the molecular and cellular mediators of human antiviral immune responses can be studied. In this report, we investigated how viral loads affected the quality of the anti-HBV immune response and the outcome of viral infections.

moderate vs elevated.45 Taken together, HIS-HUHEP mice represent a versatile system in which the molecular and cellular mediators of human antiviral immune responses can be studied. In this report, we investigated how viral loads affected the quality of the anti-HBV immune response and the outcome of viral infections. The cellular distribution and phenotypes of immune mediators in the liver are distinct from those found in the periphery, thus analysis at the site of infection is key to deciphering anti-viral immunity.4 In HBV-infected HIS-HUHEP mice, Kupffer cells swarmed the liver, while NK- and T-cell recruitment increased proportionally to the dose of viral inoculum. Chemokine attraction has been suggested to play a role in this process via CXCL9 and CXCL10/CXCR3 interactions during hepatic flares in patients.35 We found that hepatic MIP-1α, MIP-1β, CXCL9, and CXCL10 levels were increased in chronically infected mice, which could participate in continuous monocyte, NK-cell, and T-cell recruitment to the liver. Intrahepatic NK cells were activated (CD56+CD16+CD69+) and functionally producing IFN-γ and TNF-α ex vivo following re-stimulation. The role of NK cells in mediating viral clearance vs persistence is still unresolved: they are essential in killing virally infected cells, yet they can eliminate HBV-specific T cells via TRAIL-mediated pathways, thereby contributing to pathogen maintenance.7, 46, 47 Liver-resident CXCR6+ NK cells have elevated TRAIL expression that may play a role in modulating T-cell–dependent HBV immunity.48 Because NK- and T-cell recruitment were positively correlated with viral infection in HIS-HUHEP mice, it seems unlikely that NK cells were eliminating hepatic HBV-reactive T cells in this context. Still, HIS-HUHEP mice should offer a useful tool to further study intrahepatic NK-cell responses during different phases of HBV infection where NK cells may modulate anti-viral responses.

th viral infection in HIS-HUHEP mice, it seems unlikely that NK cells were eliminating hepatic HBV-reactive T cells in this context. Still, HIS-HUHEP mice should offer a useful tool to further study intrahepatic NK-cell responses during different phases of HBV infection where NK cells may modulate anti-viral responses. A broad and robust T-cell response is essential for viral clearance, yet in the context of CHB patients, excess stimulation by viral antigens induces hypo-responsive PD-1+ T cells.8, 11, 30 We found that PD-1+ T cells were similarly enriched in the livers of HBV-infected HIS-HUHEP mice. Infected hepatocytes and antigen-presenting cells constituted a source of PD-L1 expression that may create an immunosuppressive environment and promote PD-1 mediated T-cell exhaustion. Interestingly, in low-dose inoculated and ETV-treated mice, the frequency of PD-1+ memory CD4+ T cells correlated with viral loads. This result suggests that low viral loads can be effectively controlled via anti-viral T cells up to a threshold level. Once this viral load is exceeded, T-cell responses may become ineffective, in part through PD-1 induction. Accordingly, in high-dose infected HIS-HUHEP mice, excessive viral production may overwhelm T-cell responses with uniform conversion to PD-1+ phenotype. While PD-1 blockade can improve anti-HBV T-cell responses in vitro,11, 30 this does not occur in all patients. Other co-inhibitory receptors, including CTLA-4, Tim-3, LAG-3, and 2B4, may have non-redundant roles in regulating HBV-specific T-cell responses.29

T-cell responses with uniform conversion to PD-1+ phenotype. While PD-1 blockade can improve anti-HBV T-cell responses in vitro,11, 30 this does not occur in all patients. Other co-inhibitory receptors, including CTLA-4, Tim-3, LAG-3, and 2B4, may have non-redundant roles in regulating HBV-specific T-cell responses.29 CD3+ T cells were specifically recruited to HBV-infected human hepatocyte foci (but not uninfected human hepatocytes), suggesting that human T cells are activated in a virus-specific fashion. However, because of the HLA-mismatched setting of these HIS-HUHEP mice (and eventually PD-1–dependent exhaustion), we were unable to detect HLA-specific T-cell responses (data not shown), which could have added to the usefulness of the model. Future studies in HLA transgenic and haplotype-matched immune system/hepatocyte HIS-HUHEP mice may help to address this issue. Still, antigens from virally infected hepatocytes may be cross presented by dendritic cells via MHC class I to CD8+ T cells, and in chronic HBV, CD14+ monocyte-derived dendritic cells carry HBV antigens to induce CD8+ T-cell responses.49 These mechanisms may allow HBV proteins to gain access to the MHC class I pathway to generate virus-specific T-cell responses in our model.

cross presented by dendritic cells via MHC class I to CD8+ T cells, and in chronic HBV, CD14+ monocyte-derived dendritic cells carry HBV antigens to induce CD8+ T-cell responses.49 These mechanisms may allow HBV proteins to gain access to the MHC class I pathway to generate virus-specific T-cell responses in our model. The balance between immunosuppressive and stimulatory signals regulates immune cell recruitment and activation in the liver of CHB patients and contributes to the pathophysiology during hepatic flares. This interplay was mirrored in HBV-infected HIS-HUHEP mice that had elevated pro-inflammatory mediators along with immunoregulatory cytokines in the plasma and TREG recruitment to the liver. Plasmatic IL-10, CXCL10, and IFN-α levels correlated with the frequency of intrahepatic PD-1+ memory CD4+ T cells, which also correlated to HBV viremia, suggesting that the viral antigen load is able to fine tune the immune response via multiple pathways. These secreted biomarkers could facilitate the non-invasive evaluation of intrahepatic immune responses.

s correlated with the frequency of intrahepatic PD-1+ memory CD4+ T cells, which also correlated to HBV viremia, suggesting that the viral antigen load is able to fine tune the immune response via multiple pathways. These secreted biomarkers could facilitate the non-invasive evaluation of intrahepatic immune responses. Antiviral treatments with NUC analogs control viral replication yet do not deplete cccDNA pools and rarely lead to functional cures. However, long-term treatments lead to improved clinical outcomes with reversion of hepatitis, fibrosis, and cirrhosis in CHB patients. HBV-specific polyfunctional T-cell responses and quiescent NK-cell phenotypes are partially restored in peripheral blood mononuclear cells from treated patients, yet the mechanisms regulating viral-host responses following NUC treatments are unclear.47, 50 ETV treatment of HBV-infected HIS-HUHEP mice reduced viral loads and restored naïve immune profiles, with diminished liver infiltration of monocytes, inflammatory NK cells, CD4+ and CD8+ T and TREG cells. These results suggest that HBV viral loads are sensitively detected by the immune system and demonstrate the proof-of-concept utility of HIS-HUHEP mice for evaluating therapeutic strategies.

ed naïve immune profiles, with diminished liver infiltration of monocytes, inflammatory NK cells, CD4+ and CD8+ T and TREG cells. These results suggest that HBV viral loads are sensitively detected by the immune system and demonstrate the proof-of-concept utility of HIS-HUHEP mice for evaluating therapeutic strategies. Because of the lack of robust prognostic biomarkers of HCC development and the invasiveness of repeated biopsy sampling in human liver, the evolution of HBV-related pathology to HCC development is still obscure. HCC tumor samples from CHB patients express elevated levels of LPC gene transcripts (EpCAM, CK19, AFP) and these LPCs are tumorigenic in immunodeficient NOD/SCID mice, suggesting that they may constitute tumor-initiating cells.42, 51 Hepatic ectopic lymphoid structures have been associated to LPCs with poor HCC prognosis.44 Following 4 months of high viremia and chronic inflammation in HIS-HUHEP livers, HBV-infected LPCs were observed in inflammatory sites, although HCC did not develop in our model. The human timeframe for HCC development (several decades) would be difficult to match in a mouse model. Nevertheless, chronically infected HIS-HUHEP mice may help to define tumorigenic pathways that are relevant to HCC initiation and progression.

bserved in inflammatory sites, although HCC did not develop in our model. The human timeframe for HCC development (several decades) would be difficult to match in a mouse model. Nevertheless, chronically infected HIS-HUHEP mice may help to define tumorigenic pathways that are relevant to HCC initiation and progression. The lack of effective therapeutic strategies to eliminate HBV, or induce a sustained anti-viral immune response off treatment, remains a major health problem. Adults co-infected with HIV and HBV have a higher incidence of liver-related mortality, although the underlying mechanisms remain obscure. Our HIS-HUHEP model could offer a means to understand disease pathophysiology in this context. Taken together, humanized immune system and liver mouse models provide a potential platform to test innovative therapeutic anti-HBV strategies that combine direct-acting antivirals with immunomodulatory drugs to assess and eventually harness the full potential of the human innate and adaptive immune response to cure HBV infections. Supplementary Materials and Methods Generation of Humanized Mouse Model BALB/c Rag2-/-Il2rg-/-SirpaNODuPAtg/tg male and female newborn mice (<1 week old) were sublethally irradiated (2.5 Gy), injected intrahepatically with ∼2x105 CD34+ human fetal liver cells (Advanced Bioscience Resources, Alameda, CA), and subsequently injected intrasplenically with 7x105 freshly thawed human hepatocytes (BD, Corning, Corning, NY) at 5–8 weeks. HUHEP mice were injected only with human hepatocytes as described above.

Gy), injected intrahepatically with ∼2x105 CD34+ human fetal liver cells (Advanced Bioscience Resources, Alameda, CA), and subsequently injected intrasplenically with 7x105 freshly thawed human hepatocytes (BD, Corning, Corning, NY) at 5–8 weeks. HUHEP mice were injected only with human hepatocytes as described above. Leukocyte Stimulation In Vitro and Intracellular Staining For ex vivo cell stimulations, splenocytes (depleted of mouse CD45 cells with microbeads, Miltenyi Biotech) and intrahepatic lymphocytes were cultured overnight in RPMI 10% fetal calf serum penicillin streptomycin, with or without IL-12 (5 ng/mL), IL-15 (5 ng/mL), and IL-18 (20 ng/mL). GolgiPlug (BD Biosciences) was added 3 hours prior to cell staining with antibodies. IFN-γ, TNF-α (Miltenyi Biotech) and Foxp3 (eBioscience) intracellular labeling was performed after staining for extracellular proteins, fixation, and permeabilization of cells using either BD Cytofix/Cytoperm reagents (BD Bioscience) for IFN-γ and TNF-α, or Foxp3 Transcription factor Fixation/Permeabilization kit (eBioscience), according to manufacturer’s instructions.

nce) intracellular labeling was performed after staining for extracellular proteins, fixation, and permeabilization of cells using either BD Cytofix/Cytoperm reagents (BD Bioscience) for IFN-γ and TNF-α, or Foxp3 Transcription factor Fixation/Permeabilization kit (eBioscience), according to manufacturer’s instructions. Analysis of Anti-HBs and Anti-HBc Antibody Responses Human plasma samples were initially titrated with the Abbott Architect i2000SR analyzer (Chicago, IL). BHK-21 cells were transfected by electroporation with RNA transcribed in vitro from either pSFV1-SHBsadw or pSFV1-HBcadw or pSFV3-β-Gal constructs.24 Sixteen hours after transfection, cells were fixed in 80% acetone for 10 minutes at -20°C, and immediately used for immunohistochemical staining. Cells were incubated for 30 minutes at 37°C with serial 2-fold dilutions of humanized mouse plasma or patient serum first normalized to 80 μg/mL total immunoglobulin G (IgG). After 4 phosphate-buffered saline (PBS) washes, cells were incubated for 30 minutes at 37°C with a peroxidase-conjugated goat anti-human IgG antibody Fc gamma fragment specific (cat #109-035-008; Jackson ImmunoResearch Laboratories, West Grove, PA) diluted 1:500 in PBS. Cells were washed with PBS (4 times) and stained with an AEC (3-Amino-9-ethylcarbazole) substrate-chromogen solution (Sigma-Aldrich, Saint Louis, MI) for 5 minutes. After a wash with distilled water, wells containing red stained BHK-21 cells were scored using an Olympus IX51 microscope. The experimenter was not informed of the plasma's origin so as to ensure a blind interpretation of the results.Supplementary Figure 1 Longitudinal follow-up of HBV viremia (black line) and hAlbumin (grey bars) in the plasma of individual HUHEP and HIS-HUHEP mice inoculated with high-dose HBV 10e9.

mpus IX51 microscope. The experimenter was not informed of the plasma's origin so as to ensure a blind interpretation of the results.Supplementary Figure 1 Longitudinal follow-up of HBV viremia (black line) and hAlbumin (grey bars) in the plasma of individual HUHEP and HIS-HUHEP mice inoculated with high-dose HBV 10e9. Supplementary Figure 2 Follow-up of viral loads and liver chimerism in HIS-HUHEP mice inoculated at low-dose HBV 10e7. (A and B) HBV viremia (black line) and hAlbumin (grey bars) were measured longitudinally in the plasma of HBV-infected HUHEP (A) and HIS-HUHEP (B) mice. Graphs show the mean from 6 HUHEP and 6 HIS-HUHEP mice. (C and D) Correlative analysis of the viral load vs liver humanization: for each plasma sample viremia was plotted against hAlbumin concentration from HUHEP (C) (n=6) or HIS-HUHEP mice (D) (n=6). Correlation was evaluated using a 2-tailed Pearson's χ2 test. (E and F) Analysis of viral progression over time: linear regression analysis of the ratio of HBV viral load over hAlbumin concentration plotted against time post infection, from (E) HUHEP (dotted line) or HIS-HUHEP mice inoculated at 10e7 or (F) from HIS-HUHEP mice inoculated at 10e7 vs 10e9 (respectively, grey or black filled line). The P value determines whether the slopes are significantly different. (G) Fold change of M65/hAlbumin pre- and post-infection (17 ±6 wpi) in HUHEP mice (control n=4 HBV 10e7; and 10e9 n=6).

IS-HUHEP mice inoculated at 10e7 or (F) from HIS-HUHEP mice inoculated at 10e7 vs 10e9 (respectively, grey or black filled line). The P value determines whether the slopes are significantly different. (G) Fold change of M65/hAlbumin pre- and post-infection (17 ±6 wpi) in HUHEP mice (control n=4 HBV 10e7; and 10e9 n=6). Supplementary Figure 3 Longitudinal follow-up of HBV viremia (black line) and hAlbumin (grey bars) in the plasma of individual HUHEP and HIS-HUHEP mice inoculated with low-dose HBV 10e7. Supplementary Figure 4 Immunofluorescence analysis of liver sections from HIS-HUHEP control or low-dose (HBV 10e7) infected mice co-stained for hAlbumin (blue) and HBc (green), with either hCD45 (red), or hCD3 (red), or hCD68 (red). DAPI-stained nuclei shown in grey. Scale bar represents 100 μm. Supplementary Figure 5 Characterization of human T-cell subsets in HIS-HUHEP–infected mice by FACS analysis. Percentage of HLA-DR–positive cells in memory CD4+CD45RO+ or CD8+CD45RO+ T cells from control or HBV-infected mice at low (HBV 10e7) or high (HBV 10e9) doses.

Supplementary Figure 4 Immunofluorescence analysis of liver sections from HIS-HUHEP control or low-dose (HBV 10e7) infected mice co-stained for hAlbumin (blue) and HBc (green), with either hCD45 (red), or hCD3 (red), or hCD68 (red). DAPI-stained nuclei shown in grey. Scale bar represents 100 μm. Supplementary Figure 5 Characterization of human T-cell subsets in HIS-HUHEP–infected mice by FACS analysis. Percentage of HLA-DR–positive cells in memory CD4+CD45RO+ or CD8+CD45RO+ T cells from control or HBV-infected mice at low (HBV 10e7) or high (HBV 10e9) doses. Supplementary Figure 6 Immunofluorescence analysis of liver sections from HIS-HUHEP control or HBV-inoculated mice at either low dose (HBV 10e7) or high dose (HBV 10e9) co-stained for hAlbumin (blue), HBc (green), and PD-L1 (red) with DAPI-stained nuclei shown in grey. Scale bar represents 100 μm. In the second, third, and fourth vertical panels 1 color channel has been removed to visualize the staining patterns by pairs. White arrowheads: double-stained cells for Albumin and PD-L1; white open arrows: triple-stained cells for Albumin, HBc, and PD-L1; white filled arrows: cells mono-stained cells for PD-L1.

0 μm. In the second, third, and fourth vertical panels 1 color channel has been removed to visualize the staining patterns by pairs. White arrowheads: double-stained cells for Albumin and PD-L1; white open arrows: triple-stained cells for Albumin, HBc, and PD-L1; white filled arrows: cells mono-stained cells for PD-L1. Supplementary Figure 7 Biomarker expression patterns in HBV-infected and control HUHEP mice. Plasma from control (n=4) and HBV-infected (n=10) mice at the time of sacrifice (20 ±2 wpi) was analyzed by multiplex assay for human cytokines. IL-1b, IL-2, IL-2R, IL-4, IL-5, IL-6, IL-10, IL-12, IL-13, IL-15, IL-17, CCL2, CCL3, CCL5, GM-CSF, IFN-γ, and TNF-α were not detected. Dotted lines indicate the lower limit of quantification for each assay. Supplementary Figure 8 HBV-infected liver progenitor cells in highly viremic HIS-HUHEP mice. Immunofluorescence analysis of liver sections from (A) control, or HBV-infected mice inoculated with (B) HBV 10e7 or (C) HBV 10e9 for Albumin (blue), HBcAg (green) and either (left column) EpCAM (red) or (right column) CK7 (red). White open arrows show mono-stained cells for either EpCAM or CK7; white arrowheads show double-stained cells co-expressing HBc and CK7; white filled arrows show triple-stained cells co-expressing Albumin, HBc, and either EpCAM or CK7. (D) HBV-infected liver co-stained for Albumin (blue), EpCAM (green) and CD3 (red). Scale bar represents 100 μm.

-stained cells for either EpCAM or CK7; white arrowheads show double-stained cells co-expressing HBc and CK7; white filled arrows show triple-stained cells co-expressing Albumin, HBc, and either EpCAM or CK7. (D) HBV-infected liver co-stained for Albumin (blue), EpCAM (green) and CD3 (red). Scale bar represents 100 μm. Supplementary Figure 9 (A) Longitudinal analysis of viremia and liver chimerism in HBV-infected (10e7) ETV-treated HIS-HUHEP mice. The dotted line indicates start of treatment. (B–D) FACS analysis of T-cell phenotypes in control, HBV-infected (10e7), and HBV-infected (10e7) ETV-treated, HIS-HUHEP mice. (A) The frequency of naïve (CD45RA+) vs memory (CD45RO+) CD4+ and CD8+ T cells was evaluated in the liver and spleen of each cohort. (B) The frequency of PD-1+ cells in memory (CD45RO+) CD4+ or CD8+ T cells from the liver and spleen. (C) Correlation analysis of the viral load (HBV DNA) relative to the frequency of memory CD4+CD45RO+PD-1+ T cells in the liver using Pearson’s correlation test. Acknowledgments The authors thank all the members of the Innate Immunity Unit and Dr Raymond Schinazi for helpful discussions, Eric Giang and Helene Massinet for technical assistance, and Dr Agnès Durand at the CBCV for help with virological measurements. We gratefully acknowledge the Centre de Recherche Translationnelle and the Animalerie Centrale of the Institut Pasteur for their productive collaboration. Conflicts of interest J.P.D. is a founder of AXENIS and a member of its advisory board. All other authors disclose no conflicts.

Acknowledgments The authors thank all the members of the Innate Immunity Unit and Dr Raymond Schinazi for helpful discussions, Eric Giang and Helene Massinet for technical assistance, and Dr Agnès Durand at the CBCV for help with virological measurements. We gratefully acknowledge the Centre de Recherche Translationnelle and the Animalerie Centrale of the Institut Pasteur for their productive collaboration. Conflicts of interest J.P.D. is a founder of AXENIS and a member of its advisory board. All other authors disclose no conflicts. Funding This work was supported in part by grants from the Agence Nationale de la Recherche programme Emergence (grant no. ANR-11-EMMA-026) and Laboratoire d’Excellence Investissement d’Avenir REVIVE (grant no. ANR-10-LBX-73), Agence Nationale de Recherches sur le sida et les hépatites virales (grant nos. 2013-105, 2016-16180, and 2016-16365), European Commission Seventh Framework Programme PATHCo (grant no. HEALTH-F3-2012-305578), Bill & Melinda Gates Foundation (global health grant no. 37869), Institut Pasteur, and Institut National de la Santé et de la Recherche Médicale. M.D. was supported by an ANRS postdoctoral grant. Author names in bold designate shared co-first authorship. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/j.gastro.2017.08.034.

See Covering the Cover synopsis on page 458; see editorial on page 478. Editor's Notes Background and Context The diverse types of neurons of the enteric nervous system (ENS) are required for normal gastrointestinal functions, yet how cellular diversity is created during ENS development remains unknown. New Findings RNA profiles and histochemical patterns identified regulatory genes in developing enteric neuron subtypes in mouse and human embryos. The transcription factor SOX6 was found essential for gastric dopamine neuron formation. Limitations Although many of the identified gene expressions were described in different regions and stages of the gastrointestinal tract by immunohistochemical stainings, the large majority need validating analyses. Impact This resource of gene expression patterns can be implemented in molecular research on ENS development and in studies aimed at differentiating specific types of enteric neurons from embryonic stem cells.

Limitations Although many of the identified gene expressions were described in different regions and stages of the gastrointestinal tract by immunohistochemical stainings, the large majority need validating analyses. Impact This resource of gene expression patterns can be implemented in molecular research on ENS development and in studies aimed at differentiating specific types of enteric neurons from embryonic stem cells. Vital gastrointestinal (GI) functions, including bowel motility, blood flow, and fluid exchange, are regulated by the enteric nervous system (ENS). To accomplish these complex tasks, the ENS is organized into full neural circuits composed of a multitude of different cell types, including intrinsic sensory neurons, motor neurons, interneurons, and enteric glia.1 In the congenital ENS disorder Hirschsprung disease, children are typically born with the distal bowel lacking motility due to local absence of enteric neurons.2 In other enteric neuropathies, including achalasia and gastroparesis, only subsets of neuronal subtypes are affected.1, 2 Hence, the diversity of enteric neuronal subtypes is critical for normal gut function and selective dysfunction or local neuronal loss leads to GI disorders. Recent progress in producing stem cell–derived ENS cells raises hope for development of novel cell-based treatments of ENS disorders,2, 3 but functional recovery rests on recreating enteric circuits with diverse cellular composites. To date, knowledge of the molecular mechanisms underlying the generation of different ENS cell types in correct numbers and proportions is incomplete,4 but would be instrumental in the engineering of functional ENS networks.

but functional recovery rests on recreating enteric circuits with diverse cellular composites. To date, knowledge of the molecular mechanisms underlying the generation of different ENS cell types in correct numbers and proportions is incomplete,4 but would be instrumental in the engineering of functional ENS networks. The ENS is primarily derived from neural crest cells, which emigrate from the vagal (and partly sacral) neural tube and colonize the foregut at embryonic day (E) 9.0 in mice and week (W) 4 in humans. Extensive proliferation and migration of these enteric neural stem cells (ENSCs) eventually results in an interconnected ganglionic network along the GI tract.5 Asynchronous differentiation into neurons starts at E10 and continues into postnatal stages,6 resulting in 16 mature neuronal subtypes as defined by axonal projection patterns and neurotransmitter expression.4 However, detailed investigation of enteric neuronal subtypes and their lineage development is challenged by the lack of a clear functional architecture in the ENS. In contrast, the patterned germinal zones generating groups of functionally related neurons in the central nervous system has allowed extensive research on its development. There, establishment of distinct neurons relies on the interplay between instructive signaling factors and intrinsic programs of transcription factor networks, sequentially regulating specification and differentiation.7 On account of this, the stage of maturity and identity of a neural cell can be inferred from its transcription factor signature. Many transcription factors have been detected during ENS development, but only a few have been described in detail and shown to influence neuronal subtype commitment.8, 9 The bowel mesenchyme secretes several growth factors essential for ENS development, but our understanding is mainly restricted to the roles of these factors in general differentiation and proliferation.8 Intricate cell-to-cell communication also is required for the subsequent establishment of neuron-specific connections to form functional circuits, but remains uncharacterized. Expression profiles of transcriptional and signaling regulators in ENS sublineages at different stages could thus be deployed to uncover the molecular mechanisms underlying neuronal diversification and network formation during ENS development.

ecific connections to form functional circuits, but remains uncharacterized. Expression profiles of transcriptional and signaling regulators in ENS sublineages at different stages could thus be deployed to uncover the molecular mechanisms underlying neuronal diversification and network formation during ENS development. Birth-dating studies demonstrate that the different enteric neuronal subclasses are born (undergo neurogenesis) in a temporally sequential manner.10, 11 Enteric serotonergic neurons (5-HT+) are born early (E10–E12.5), whereas for instance, dopaminergic neurons (TH+) appear slightly later (peaks at E13–E15.5). The timed order of neuronal subtype generation indicates a progressive change in the differentiation competence of ENSCs. To reveal intrinsic and extrinsic regulatory genes at discrete steps during differentiation of enteric neuronal subclasses, we performed RNA expression analysis of ENS subpopulations and surrounding gut tissue at developmentally distinct stages. Adding to the transcriptome analysis, we present a spatiotemporal histochemical expression pattern map of transcriptional and signaling factors in both mouse and human developing gut. Demonstrating the strength of our strategy to find determining genes, we show that one candidate transcription factor, Sox6, is essential for the selective development of gastric dopamine neurons. Implementation of this novel gene platform would significantly refine basic and translational studies of the successive processes in which ENSCs differentiate into various enteric neurons and assemble into discrete ENS circuitries to control bowel physiology.

l for the selective development of gastric dopamine neurons. Implementation of this novel gene platform would significantly refine basic and translational studies of the successive processes in which ENSCs differentiate into various enteric neurons and assemble into discrete ENS circuitries to control bowel physiology. Material and Methods Mouse and Human Embryos The generation of Sox10CreERT2,6 Wnt1Cre,12 and R26ReYFP13 mouse strains has been described. Sox6fl mice were kindly provided by V. Lefebvre (Cleveland Clinic Lerner Research Institute, Cleveland, OH).14 Histochemical analysis was performed on C57 and MF1 mice. Animal experiments adhered to Animal Research: Reporting In Vivo Experiments standards, and were approved by the local National Institute for Medical Research ethical review panel or by the local ethics committee in Stockholm (N87/13). Remains of human embryos and fetuses (5.5–10.0 weeks after conception) were obtained after elective routine abortions with written consent given by the pregnant women. Collection of human tissue for research was approved by the Regional Human Ethics Committee, Stockholm (2013/564–32) and conducted by the Karolinska Institutet Stem Cell and Tissue Bank. Induction of Inducible Reporter Time-mated Sox10CreERT2 × R26ReYFP mice received a single intraperitoneal injection of 4-hydroxytamoxifen (4-OHT; Sigma-Aldrich, Saint Louis, MO) dissolved in ethanol/corn oil (1:9) (0.1 mg/g body weight) at E10.5 or E14.5. After 20 hours, embryos were harvested.

Material and Methods Mouse and Human Embryos The generation of Sox10CreERT2,6 Wnt1Cre,12 and R26ReYFP13 mouse strains has been described. Sox6fl mice were kindly provided by V. Lefebvre (Cleveland Clinic Lerner Research Institute, Cleveland, OH).14 Histochemical analysis was performed on C57 and MF1 mice. Animal experiments adhered to Animal Research: Reporting In Vivo Experiments standards, and were approved by the local National Institute for Medical Research ethical review panel or by the local ethics committee in Stockholm (N87/13). Remains of human embryos and fetuses (5.5–10.0 weeks after conception) were obtained after elective routine abortions with written consent given by the pregnant women. Collection of human tissue for research was approved by the Regional Human Ethics Committee, Stockholm (2013/564–32) and conducted by the Karolinska Institutet Stem Cell and Tissue Bank. Induction of Inducible Reporter Time-mated Sox10CreERT2 × R26ReYFP mice received a single intraperitoneal injection of 4-hydroxytamoxifen (4-OHT; Sigma-Aldrich, Saint Louis, MO) dissolved in ethanol/corn oil (1:9) (0.1 mg/g body weight) at E10.5 or E14.5. After 20 hours, embryos were harvested. Preparation of Cell Populations Dissected guts were digested in 1 mg/mL dispase/collagenase (Roche, Basel, Switzerland) for 3 to 5 minutes (E11.5) at room temperature (RT) or 1 hour (E15.5) at 37°C. Tissue was then dissociated in OPTIMEM (Life Technologies, Carlsbad, CA) supplemented with 10% fetal calf serum, collected by centrifugation and passed through a cell strainer cap (Falcon; 352235). Fluorescent-activated cell sorting (FACS) of cells was performed using a MoFlo Cell Sorter (BeckmanCoulter, Brea, CA) or a FACSAria II (BD Biosciences, San Jose, CA) (Supplementary Figure 1). Purity-checked cell samples were counted using a cytometer, spun at 200g for 12 minutes at 4°C and resuspended in lysis buffer from RNeasy Micro kit (Qiagen, Hilden, Germany). Samples were vortexed or run through a shredder (Qiagen), snap-frozen, and kept at −80°C.

D Biosciences, San Jose, CA) (Supplementary Figure 1). Purity-checked cell samples were counted using a cytometer, spun at 200g for 12 minutes at 4°C and resuspended in lysis buffer from RNeasy Micro kit (Qiagen, Hilden, Germany). Samples were vortexed or run through a shredder (Qiagen), snap-frozen, and kept at −80°C. RNA Preparation Samples from different FACS sortings were pooled to obtain appropriate final RNA concentrations. Total RNA was extracted from the samples using an RNeasy Micro kit (Qiagen). Purity of RNA was verified by Bioanalyzer Total RNA Nano (Agilent Technologies, Santa Clara, CA). Biotin-labeled complementary DNA was prepared using the Ovation Pico kit (Nugen, Manchester, UK) at the University College of London (UCL) Genomics (London, UK). For detailed information of cell samples and RNA analysis see Supplementary Figure 1. Microarray Analysis The samples were hybridized to GeneChip Mouse Gene 1.0 ST arrays at UCL Genomics and processed using the rma algorithm using the Partek software (Partek Inc., St. Louis, MO). Principal component analysis and box-plots showed normal distribution.

RNA Preparation Samples from different FACS sortings were pooled to obtain appropriate final RNA concentrations. Total RNA was extracted from the samples using an RNeasy Micro kit (Qiagen). Purity of RNA was verified by Bioanalyzer Total RNA Nano (Agilent Technologies, Santa Clara, CA). Biotin-labeled complementary DNA was prepared using the Ovation Pico kit (Nugen, Manchester, UK) at the University College of London (UCL) Genomics (London, UK). For detailed information of cell samples and RNA analysis see Supplementary Figure 1. Microarray Analysis The samples were hybridized to GeneChip Mouse Gene 1.0 ST arrays at UCL Genomics and processed using the rma algorithm using the Partek software (Partek Inc., St. Louis, MO). Principal component analysis and box-plots showed normal distribution. Bioinformatic Analysis Partek software was used at UCL Genomics to perform pairwise comparisons. Significant genes were based on 5% false discovery rate using the Benjamini & Hochberg correction. Functional annotation clustering was performed in the DAVID resource 6.7.15 Although filtered away by the applied false discovery rate, expression of a few genes was found by in situ hybridization (ISH), immunohistochemistry (IHC), or other published study. These genes were introduced into resulting gene lists.

ection. Functional annotation clustering was performed in the DAVID resource 6.7.15 Although filtered away by the applied false discovery rate, expression of a few genes was found by in situ hybridization (ISH), immunohistochemistry (IHC), or other published study. These genes were introduced into resulting gene lists. Tissue Preparation E11.5 to 12.5 mouse embryos were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C for 1.5 hours. Isolated E15.5 to 19 mouse guts and human guts were fixed for 2 hours. Tissue was incubated at 4°C overnight in 30% sucrose in PBS, embedded in optimum cutting temperature compound (Histolab, Leiden, The Netherlands) and stored at −80°C. Tissue was sectioned at 14 μm. Immunohistochemistry IHC was performed as described.9 Briefly, mouse sections were preincubated for 2 hours at RT with unconjugated donkey anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA). Mouse and human tissue were then blocked 1 hour (2% donkey or goat serum, 0.1% Triton x-100, PBS) and incubated overnight with primary antibodies and detected with secondary antibodies the following day (Supplementary Table 1). Before IHC, some antibodies required antigen retrieval: microwave heating of slides in antigen retrieval solution (Dako, Santa Clara, CA) followed by cooling to RT.

, 0.1% Triton x-100, PBS) and incubated overnight with primary antibodies and detected with secondary antibodies the following day (Supplementary Table 1). Before IHC, some antibodies required antigen retrieval: microwave heating of slides in antigen retrieval solution (Dako, Santa Clara, CA) followed by cooling to RT. Imaging Images were taken using a Zeiss (Oberkochen, Germany) LSM700 confocal microscope and processed in Adobe Photoshop CS6 (Adobe Systems Inc., San Jose, CA) or Image J (National Institutes of Health, Bethesda, MD). Counting of fluorescent cells was performed on images or directly under a Zeiss fluorescent microscope. The abundance of protein coexpression was estimated by visual inspection. For all ISH analysis, we used Allen Developing Mouse Brain Atlas (http://developingmouse.brain-map.org) and GenePaint (http://www.GenePaint.org). Gastric Emptying Liquid gastric emptying assay was essentially performed as described.16 Mice were fasted 6 hours and water withdrawn 1 hour before the experiment. Animals were killed 15 minutes after gavage (rhodamine B dextran: 100 μL; 10 mg/mL in 2% methylcellulose; Invitrogen, Carlsbad, CA), whereby stomach and small intestine (10 equal-length segments) were collected in 0.9% NaCl, homogenized, and centrifuged. Fluorescence was measured in 200-μL aliquots (triplets) of the supernatant (Fluostar Omega; BMG Labtech, Offenburg, Germany) and percentage of total fluorescence that emptied from the stomach was calculated. Stomach size assessment was made in mice culled at 9 AM with prior free access to food and drink.

Gastric Emptying Liquid gastric emptying assay was essentially performed as described.16 Mice were fasted 6 hours and water withdrawn 1 hour before the experiment. Animals were killed 15 minutes after gavage (rhodamine B dextran: 100 μL; 10 mg/mL in 2% methylcellulose; Invitrogen, Carlsbad, CA), whereby stomach and small intestine (10 equal-length segments) were collected in 0.9% NaCl, homogenized, and centrifuged. Fluorescence was measured in 200-μL aliquots (triplets) of the supernatant (Fluostar Omega; BMG Labtech, Offenburg, Germany) and percentage of total fluorescence that emptied from the stomach was calculated. Stomach size assessment was made in mice culled at 9 AM with prior free access to food and drink. Statistical Analysis Mutant embryos were compared with littermate controls. Student paired t test was performed for cell countings, and Student t test for the functional assay. Bars indicate means ± standard deviation. Significance levels for the tests were assumed at *P < .05; **P < .01; ***P < .001. Data Availability The microarray data have been submitted to the GEO database (http://www.ncbi.nlm.nih.gov/geo/) and assigned the identifier GSE100130.

Statistical Analysis Mutant embryos were compared with littermate controls. Student paired t test was performed for cell countings, and Student t test for the functional assay. Bars indicate means ± standard deviation. Significance levels for the tests were assumed at *P < .05; **P < .01; ***P < .001. Data Availability The microarray data have been submitted to the GEO database (http://www.ncbi.nlm.nih.gov/geo/) and assigned the identifier GSE100130. Results Transcriptome Analysis Identifies Differentially Expressed Genes in Enteric Progenitors, Neurons, and Non-ENS Gut Cells at Different Developmental Stages Aiming to identify intrinsic and extrinsic genes with regulatory functions at successive steps during differentiation of phenotypically distinct neurons, we designed a transcriptome analysis comparing ENS progenitor cells, the entire ENS (including also immature neurons), and surrounding gut tissue at 2 developmentally distinct stages (E11.5 and E15.5).

trinsic genes with regulatory functions at successive steps during differentiation of phenotypically distinct neurons, we designed a transcriptome analysis comparing ENS progenitor cells, the entire ENS (including also immature neurons), and surrounding gut tissue at 2 developmentally distinct stages (E11.5 and E15.5). As SOX10 specifically marks dividing ENS progenitor cells,8 we could retrieve this subpopulation by inducing reporter expression in Sox10CreERT2 × R26ReYFP embryos.6 The whole ENS could be collected from Wnt1Cre × R26ReYFP mouse embryos, owing to ENS-specific yellow fluorescence protein (YFP) reporter expression in the GI tract.12 YFP+ and YFP− cells from dissociated E11.5 and E15.5 guts of transgenic embryos were separated using FACS, after which RNA expression was determined using gene arrays. In total, the transcriptomes of 4 different ENS populations were analyzed: Sox10CreERT2 × R26ReYFP at E11.5 and E15.5 (hereinafter denoted S11 and S15) and Wnt1Cre × R26ReYFP at E11.5 and E15.5 (hereinafter denoted W11 and W15) (Figure 1A). Transcriptomes of YFP− cells, derived from Wnt1Cre × R26ReYFP gut at E11.5 and E15.5 representing non-ENS gut tissue also were included in the analysis (partly as a control and hereinafter denoted C11 and C15) (Figure 1A). Pairwise comparisons between the 6 populations can be found in Supplementary Table 2.Figure 1 Transcriptome screen design, verification, and identification of enriched genes. (A) Schematic drawing depicting the transcriptome analysis. ENS populations included SOX10+ progenitors (red boxes) and the whole ENS (blue boxes), each at 2 distinct development stages: E11.5 (S11 and W11) and E15.5 (S15 and W15). W11 contained immature neurons differentiating into, for example, 5-HT+ neurons, whereas W15 included immature neurons differentiating into other subtypes (eg, TH+). Non-ENS control gut tissue (gray boxes) at E11.5 (C11) and E15.5 (C15) was also included. (B) Heat map summarizing differentially expressed genes, compiled from the union of genes with top-10 absolute fold change (and P < .05) in the comparisons of S11vsC11, S15vsC15, W11vsC11, and W15vsC15. Genes and conditions are clustered by their hierarchical similarity. Color intensity represents the mean-centred log2 expression values. (C) Graphs comparing gene ontology (GO) term enrichment, ENS, and gut marker genes in the 4 ENS populations to control populations. (D) Heat map depicting cell cycle or neuronal genes.

enes and conditions are clustered by their hierarchical similarity. Color intensity represents the mean-centred log2 expression values. (C) Graphs comparing gene ontology (GO) term enrichment, ENS, and gut marker genes in the 4 ENS populations to control populations. (D) Heat map depicting cell cycle or neuronal genes. Genes are clustered by their hierarchical similarity. Color intensity represents the mean-centred log2 expression values. (E) Tables showing the number of transcription factors, signaling factors, and receptors found in pairwise comparisons (absolute fold change >1.2; P < .05) of the transcriptomes. n/a, not analyzed.

enes and conditions are clustered by their hierarchical similarity. Color intensity represents the mean-centred log2 expression values. (C) Graphs comparing gene ontology (GO) term enrichment, ENS, and gut marker genes in the 4 ENS populations to control populations. (D) Heat map depicting cell cycle or neuronal genes. Genes are clustered by their hierarchical similarity. Color intensity represents the mean-centred log2 expression values. (E) Tables showing the number of transcription factors, signaling factors, and receptors found in pairwise comparisons (absolute fold change >1.2; P < .05) of the transcriptomes. n/a, not analyzed. By distributing the top-10 enriched genes obtained from pairwise comparisons between ENS populations (S11, S15, W11, and W15) and non-ENS controls (C11 or C15) in a combined unsupervised chart, we confirmed clustering of each data set and an overall similarity between ENS populations (Figure 1B). Gene ontology analysis verified that enteric nervous system development (7.6–8.3) (Figure 1C) was highly enriched in the 4 ENS transcriptomes in comparison with non-ENS controls. Well-known ENS-expressed genes also displayed high expression levels in ENS populations, whereas gut regulatory genes were enriched in C11 and C15 (Figure 1C). Functional cluster analysis of the S15 versus W15 (S15vsW15) comparison confirmed a high enrichment score for the gene ontology term cell cycle (37.2) in S15 and neuron projection (21.6) in W15. The S11vsW11 comparison yielded too few genes for such analysis. However, several genes involved in cell cycle regulation were enriched in both S11 and S15, whereas genes associated with neurons were enriched in W11 and W15 (Figure 1D). In summary, this analysis verified that the 4 isolated ENS transcriptomes (S11, W11, S15, W15) represented ENS cells. S11 and S15 mostly contained dividing progenitors, whereas W11 and W15 also included differentiating neurons (Figure 1A).

whereas genes associated with neurons were enriched in W11 and W15 (Figure 1D). In summary, this analysis verified that the 4 isolated ENS transcriptomes (S11, W11, S15, W15) represented ENS cells. S11 and S15 mostly contained dividing progenitors, whereas W11 and W15 also included differentiating neurons (Figure 1A). As this study focused on revealing regulatory genes in cellular diversification of the developing ENS, we decided to mine the pairwise transcriptome comparisons for 3 sets of enriched genes: transcription factors, signaling factors, and receptors (summarized in Figure 1E; corresponding gene lists in Supplementary Tables 3–5).

study focused on revealing regulatory genes in cellular diversification of the developing ENS, we decided to mine the pairwise transcriptome comparisons for 3 sets of enriched genes: transcription factors, signaling factors, and receptors (summarized in Figure 1E; corresponding gene lists in Supplementary Tables 3–5). Identification of Novel Transcription Factors in the Developing ENS Out of the hundreds of enriched transcription factors (Figure 1E; Supplementary Table 3) we performed a confirmative in-depth analysis primarily including transcription factors already linked to developmental processes and omitting those associated with mitochondria or the general transcriptional machinery. The expression of these transcription factors was first examined using online ISH resources (Supplementary Table 6; Supplementary Figure 2). A total of 31 genes were then selected for an extensive IHC analysis. In this mapping, we determined the expression in relation to HUC/D+ neurons and SOX10+ progenitor cells in mouse stomach and intestine at E11-12, E15-16, and E18-19 (Figure 2A, Supplementary Figure 3). Taken together, we identified and confirmed the expression of 73 novel transcription factors in the developing ENS. The expression of additionally 39 previously found transcription factors (Supplementary Table 6) was also confirmed, and in some cases included in the IHC expression analysis.Figure 2 Expression dynamics of transcription factors in the developing ENS. (A) Table summarizing IHC expression analysis (Supplementary Figures 3 and 4) of transcription factors in relation to HUC/D+ neurons and SOX10+ progenitor cells in stomach and intestine of mouse and human embryos at different stages. Genes are grouped according to their expression dynamics. (B) Table showing transcription factors ordered according to their DNA binding domain and their onset of expression. Onset time was estimated based on IHC, ISH (Supplementary Figure 2), and/or the transcriptome analysis. (C) Examples of IHC from groups I to IV showing similar gene expression in mouse and human. Yellow arrowheads, expression in progenitors; white arrowheads, expression in neurons. (D) Expression of SOX proteins in the developing ENS together with SOX2/10+ progenitors or HUC/D+ neurons at E15.5 (SOX4 at E12.5). Arrowheads indicate double-positive cells. (E) Expression of Hox genes in the developing ENS. Note localization of Hoxa3, Hoxc5, Hoxb3, and Hoxc4 in HUC/D+ neurons (arrowheads).

n neurons. (D) Expression of SOX proteins in the developing ENS together with SOX2/10+ progenitors or HUC/D+ neurons at E15.5 (SOX4 at E12.5). Arrowheads indicate double-positive cells. (E) Expression of Hox genes in the developing ENS. Note localization of Hoxa3, Hoxc5, Hoxb3, and Hoxc4 in HUC/D+ neurons (arrowheads). To address evolutionary conservation and directly bridge our findings to translational implementations, IHC analysis was performed in human embryonic gut at W5-6 (equivalent to ∼E12–E14) and W7-10 (equivalent to ∼E14.5–E17.5) (Figure 2A and C; Supplementary Figure 4). With very few exceptions, the transcription factors found in mouse ENS were also detected in the developing human ENS and with very similar expression dynamics (Figure 2C, Supplementary Figure 4). To view a comprehensive list of transcription factors expressed in the developing ENS (novel and previously known), see Supplementary Table 6.

tions, the transcription factors found in mouse ENS were also detected in the developing human ENS and with very similar expression dynamics (Figure 2C, Supplementary Figure 4). To view a comprehensive list of transcription factors expressed in the developing ENS (novel and previously known), see Supplementary Table 6. Expression Dynamics of Transcription Factors of Key Developmental Gene Families Based on the spatiotemporal expression patterns revealed by IHC, we subdivided the transcription factors into 4 groups characterized by (I) early and abundant; (II) early onset, mainly neuronal; (III) late onset, mainly neuronal; or (IV) highly selective expression dynamics (Figure 2A and C; Supplementary Figures 3 and 4). Expression pattern I likely represented genes controlling neurogenesis or generic aspects of ENS lineage differentiation. Expression patterns II and III suggested roles in neuronal differentiation, perhaps of specific sublineages. The selective expression of group IV transcription factors indicated roles in neuronal subtype differentiation.

ikely represented genes controlling neurogenesis or generic aspects of ENS lineage differentiation. Expression patterns II and III suggested roles in neuronal differentiation, perhaps of specific sublineages. The selective expression of group IV transcription factors indicated roles in neuronal subtype differentiation. Taking all transcription factors into account, a wide range of gene families was represented (Figure 2B). In particular, we uncovered many genes of the Sox (HMG-box) and Hox (homeo-box) families, which play key roles in diverse developmental processes.17, 18 SOX2, SOX8, and SOX10 already have defined functions in the ENS (Supplementary Table 6). The present screen identified 5 additional SOX proteins with distinct expression patterns at various stages (Figure 2D; Supplementary Figures 3 and 4). SOX6 belonged to group IV and is revisited later in this study. Expression of SOX5 largely coincided with progenitors akin to the previously described SOX genes (Figure 2A and D; Supplementary Figures 3 and 4), whereas SOX4, SOX9, and SOX11 also were detected in neurons. Our screen identified 9 Hox genes in addition to the 5 already reported Hox genes (Figure 2A, B, and E; Supplementary Figures 2–4; Supplementary Table 6). Notably, our analysis revealed that many of the Hox genes were exclusively expressed in differentiating neurons (Figure 2E). In summary, our screen uncovered a large set of novel transcription factors, the gene family and expression dynamics indicating roles for generic or specific aspects within proliferating ENSCs and/or immature neurons.

Taking all transcription factors into account, a wide range of gene families was represented (Figure 2B). In particular, we uncovered many genes of the Sox (HMG-box) and Hox (homeo-box) families, which play key roles in diverse developmental processes.17, 18 SOX2, SOX8, and SOX10 already have defined functions in the ENS (Supplementary Table 6). The present screen identified 5 additional SOX proteins with distinct expression patterns at various stages (Figure 2D; Supplementary Figures 3 and 4). SOX6 belonged to group IV and is revisited later in this study. Expression of SOX5 largely coincided with progenitors akin to the previously described SOX genes (Figure 2A and D; Supplementary Figures 3 and 4), whereas SOX4, SOX9, and SOX11 also were detected in neurons. Our screen identified 9 Hox genes in addition to the 5 already reported Hox genes (Figure 2A, B, and E; Supplementary Figures 2–4; Supplementary Table 6). Notably, our analysis revealed that many of the Hox genes were exclusively expressed in differentiating neurons (Figure 2E). In summary, our screen uncovered a large set of novel transcription factors, the gene family and expression dynamics indicating roles for generic or specific aspects within proliferating ENSCs and/or immature neurons. Unique Combinatorial Expression of Transcription Factors in Enteric Neuronal Subtypes To gain insights into the correlation between the newly identified transcription factors and neuronal subtypes, we next determined the colocalization pattern of 25 transcription factors with common neurotransmitter/peptide markers. Analysis at E18-19 revealed that 10 transcription factors showed selective expression with subsets of the marker proteins (Figure 3B), whereas others colocalized with all (Figure 3A). Taken together, the transcription factors constituted unique combinatorial codes for each neurotransmitter/peptide marker.Figure 3 Coexpression analysis of transcription factors with enteric neurotransmitters/neuropeptides. (A and B) Tables summarizing IHC coexpression analysis of abundant (A) or selectively expressed (B) transcription factors together with ENS marker genes at E18.5. TH, NPY, 5-HT, and ChAT were analyzed in the stomach, and all other markers in the small intestine. Coexpression with SOX6 could be addressed only in the stomach. (C) Coexpression (arrowheads) between TH and 6 transcription factors in the stomach at E18.5. CALB1, calbindin; CGRP, calcitonin gene-related peptide; ChAT, choline acetyltransferase; 5-HT, 5-hydroxytryptamine; NOS1, nitric oxide synthase 1; NPY, neuropeptide Y; TH, tyrosine hydroxylase; VIP, vasoactive intestinal polypeptide.

pression (arrowheads) between TH and 6 transcription factors in the stomach at E18.5. CALB1, calbindin; CGRP, calcitonin gene-related peptide; ChAT, choline acetyltransferase; 5-HT, 5-hydroxytryptamine; NOS1, nitric oxide synthase 1; NPY, neuropeptide Y; TH, tyrosine hydroxylase; VIP, vasoactive intestinal polypeptide. Most enteric neurotransmitters are used by several functionally distinct neuronal subclasses; however, dopamine is selectively produced in only subsets of enteric neurons.4 Dopamine neurons mature slowly, and at E18.5, only dopamine neurons of the stomach and not intestines have started to robustly express the indicative marker gene TH. The combinatorial expression code of the gastric TH+ neurons (Figure 3A and B) might therefore specifically define a single neuronal subtype, and included, for example, EBF1, MEIS2, ETV1, SATB1, KLF7, and SOX6 (Figure 3C).

of the stomach and not intestines have started to robustly express the indicative marker gene TH. The combinatorial expression code of the gastric TH+ neurons (Figure 3A and B) might therefore specifically define a single neuronal subtype, and included, for example, EBF1, MEIS2, ETV1, SATB1, KLF7, and SOX6 (Figure 3C). Loss of Sox6 Impairs Development of Gastric TH+ Neurons SOX6 colocalized with gastric TH+ neurons and only 2 other markers: CALB1 and NPY (Figure 3B and C; Figure 4B and C). SOX6 expression initiated at E11.5 in a small subset of SOX10+ progenitor cells, but gradually increased its to a higher proportion of cells (it was identified in S15vsS11) and coincided with HUC/D+ neurons at E15.5 (Figure 4A). Expression of SOX6 thus correlated well with the development of gastric dopamine neurons.Figure 4 Loss of Sox6 results in selective reduction of gastric TH+ neurons and gastric motility. (A) IHC of SOX6 expression in progenitor cells (yellow arrowheads) or neurons (white arrowheads) in the developing stomach but not intestine. (B and C) IHC showing neurotransmitters that are coexpressed (arrowheads) (B), or not coexpressed (C) with SOX6 in enteric neurons at E18.5. (D) Pairwise IHC analysis showing expression of NPY and CALB1 with each other but not with TH in enteric neurons at E18.5. (E) IHC of SOX6 at E18.5 showing expression in ENS (arrowhead) and non-ENS tissue (stars). (F and G) Representative IHC images depicting expression of phenotypic marker proteins in the stomach of control embryos (F) and Sox6 mutant embryos (G). (H–J) Average percentage of neurons expressing specific markers in the stomach of Sox6 mutant and littermate control E18.5 embryos (H and I) or adults (J). n = 3–4. (K) Decreased weight of Sox6 mutant males compared with littermate control mice. n = 3–5. (L) Enlarged stomach with more residual food in a Sox6 mutant (left) in comparison with a control (right) mouse. n = 4. (M) Gastric emptying shown as percentage of administered rhodamine dextran that emptied from the stomach of Sox6 mutant and littermate control male adult mice. n = 3 *P < .05, **P < .01, *** P < .001.

(L) Enlarged stomach with more residual food in a Sox6 mutant (left) in comparison with a control (right) mouse. n = 4. (M) Gastric emptying shown as percentage of administered rhodamine dextran that emptied from the stomach of Sox6 mutant and littermate control male adult mice. n = 3 *P < .05, **P < .01, *** P < .001. The neuronal subtypes of the stomach are poorly characterized. To gain insights to the combinatorial expression of gastric phenotypic markers, we performed pairwise IHC at E18.5. The analysis revealed that most NPY+ neurons displayed CALB1 expression (95%), whereas TH+ cells neither expressed NPY nor CALB1 (Figure 4D). Therefore, we concluded that SOX6 is expressed in 2 gastric subpopulations characterized by either TH or NPY/CALB1 expression.

otypic markers, we performed pairwise IHC at E18.5. The analysis revealed that most NPY+ neurons displayed CALB1 expression (95%), whereas TH+ cells neither expressed NPY nor CALB1 (Figure 4D). Therefore, we concluded that SOX6 is expressed in 2 gastric subpopulations characterized by either TH or NPY/CALB1 expression. To address the possible role of SOX6 in the generation of these neuronal subtypes, we analyzed Wnt1Cre × Sox6fl/fl embryos at E18.5. A conditional knockout approach was crucial to attribute possible alterations in ENS development to cell autonomous effects, as SOX6 expression also localized to the serosa and mucosa (Figure 4E). Of the markers normally coexpressed with SOX6, TH expression was drastically reduced (by 70%), whereas CALB1+ and NPY+ neurons were found in similar numbers in mutant and littermate control stomachs (Figure 4F, G, and I). The total numbers of HUC/D+ neurons or the unrelated phenotypic marker, VIP, were unchanged in Sox6 mutants in comparison with controls (Figure 4F–I). The reduction of TH+ neurons remained in the stomach of adult Sox6 mutant mice (Figure 4J), which also weighed 20% less than littermate control mice (Figure 4K). To determine the functional impact of Sox6 deletion on gastric motility, we performed a liquid gastric transit test, which showed a reduction in the rate of gastric emptying in Sox6 mutants compared with littermate controls (Figure 4M). Moreover, in Sox6 mutant mice with free access to food and drink, stomachs were enlarged with more contents than stomachs of littermate controls (Figure 4L). In conclusion, our analysis demonstrates that SOX6 expression is required for the specific generation of gastric TH+ neurons, with possible importance for normal gastric motility.

mutant mice with free access to food and drink, stomachs were enlarged with more contents than stomachs of littermate controls (Figure 4L). In conclusion, our analysis demonstrates that SOX6 expression is required for the specific generation of gastric TH+ neurons, with possible importance for normal gastric motility. Identification of Signaling Ligands and Receptors in the Developing Gut Wall We next characterized the signaling factors and receptors found to be enriched in the transcriptome comparisons (Figure 1E, Supplementary Tables 4 and 5). From the combined identified genes, the expression of 157 receptors and signaling factors was first analyzed in ISH images (Supplementary Table 7, Supplementary Figure 5). Additionally, we determined the expression of 19 signaling genes in relation to HUC/D+ neurons and SOX10+ progenitors in the stomach and intestine of mouse (at E11–12, E15–16, and E18–19) and human (at W6–11) embryos (Figure 5B and C, Supplementary Figures 6 and 7, Supplementary Table 7). In total, this screen identified and confirmed the novel expression of 74 ligands and receptors. A total of 82 previously described receptors and ligands in or around the ENS were additionally found, and in a few cases included in the IHC expression analysis. With few exceptions, all signaling factors confirmed in mouse were also detected in the developing human ENS (Supplementary Figure 7). To view a comprehensive list of signaling factors/receptors expressed in the developing ENS (novel and previously known), see Supplementary Table 7.Figure 5 Novel cell-cell communication pathways during ENS development. (A) Table summarizing novel signaling pathways, including the identified ligands and receptors, signaling pathway associated genes (see Supplementary Table 8), and putative functions based on studies in other developing tissue. N/D, not determined; eg, indicates examples of binding partners found in the screen when ligand or receptor are incompletely studied or show promiscuous binding capacities. (B) Expression of ligand-receptor couples found within the ENS using IHC and analysis of ISH images. (C) IHC or ISH depicting ligands and receptors expressed in the ENS without the image of corresponding binding partners. (D) Expression of ligand-receptor couples using IHC and ISH, where ligands are expressed outside the ENS. Arrowheads indicate expression in HUC/D+ (yellow), SOX10+ (white), or ENS (black) cells. *Expression outside ENS.

ting ligands and receptors expressed in the ENS without the image of corresponding binding partners. (D) Expression of ligand-receptor couples using IHC and ISH, where ligands are expressed outside the ENS. Arrowheads indicate expression in HUC/D+ (yellow), SOX10+ (white), or ENS (black) cells. *Expression outside ENS. All images are shown in Supplementary Figures 5–7.

ting ligands and receptors expressed in the ENS without the image of corresponding binding partners. (D) Expression of ligand-receptor couples using IHC and ISH, where ligands are expressed outside the ENS. Arrowheads indicate expression in HUC/D+ (yellow), SOX10+ (white), or ENS (black) cells. *Expression outside ENS. All images are shown in Supplementary Figures 5–7. Novel Signaling Pathways During ENS Development Nearly all signaling ligands and receptors previously shown to operate during ENS development were picked up by our screening strategy and belonged for example to the signaling pathways of bone morphogenetic protein (BMP), netrin (NTN), semaphorin (SEMA), and Wnt (Supplementary Table 7). In addition to previously reported components, we detected numerous novel molecules in these major signaling families. Special effort was made to reveal ligand-receptor pairs known to interact in other contexts and included BMP7-ACVR1a/ACVR2a, GDF10-ACVR1b/ACVR2a, NTNG1-LRRC4C, SEMA4d-PLXNB1, and WNT2b-LRP5/FZD3 (Figure 5A and B; Supplementary Figures 5–7). Furthermore, we found receptor-ligand expression indicative of 9 unreported signaling pathways: fibroblast growth factor (FGF), Activin A (Inhba dimers), transforming growth factor beta (TGFβ), chemokine (CXCL12), ephrin/Eph, insulin-like growth factor (IGF), fibronectin leucin rich transmembrane protein (FLRT), connective tissue growth factor (CTGF), and midkine (MDK) (Figure 5A–C; Supplementary Figures 5–7). Novel receptors with incomplete signaling characterization also were found and included in the analysis: leukocyte receptor tyrosine kinase (LTK) and aryl hydrocarbon receptor (AHR) (Figure 5A and C; Supplementary Figures 5–7). Taken together, we present signaling components indicative of 16 cell-cell communication pathways with unexplored roles in the developing ENS.

rization also were found and included in the analysis: leukocyte receptor tyrosine kinase (LTK) and aryl hydrocarbon receptor (AHR) (Figure 5A and C; Supplementary Figures 5–7). Taken together, we present signaling components indicative of 16 cell-cell communication pathways with unexplored roles in the developing ENS. Candidate Regulatory Factors in Auto/Paracrine Signaling During ENS Development The temporally ordered formation of specific neurons in correct proportions may be controlled by auto/paracrine signaling within the developing ENS.16 In light of this, it is noteworthy that 9 secreted ligands and many membrane-bound ligands were found to be produced by ENS cells (Figure 5B and C; more in Supplementary Table 7 and Supplementary Figure 5). The IHC analysis showed that most novel ligands were expressed throughout development (eg, CTGF), but several factors also displayed temporally patterned expressions: early to mid-stages (GDF10), mid-stages only (FGF1), or from mid- to late-stages (eg, MDK, Activin A) (Figure 6). In addition, some receptors showed regulated spatiotemporal expressions: AHR expression was limited to the colon between E15 and 18, whereas FGFR1 was expressed at all regions except for the colon, and SLITRK3 was found in neurons only at E15 to E18 (Figure 6).Figure 6 Summary of IHC analysis of cell-cell communication components in the developing gut wall. Table summarizing the IHC expression analysis of receptors and ligands in the stomach and intestine at different developmental stages in mouse and human. Column to the right indicates non-ENS expression. n/a, antibody staining inconclusive or incompatible with species.

mmunication components in the developing gut wall. Table summarizing the IHC expression analysis of receptors and ligands in the stomach and intestine at different developmental stages in mouse and human. Column to the right indicates non-ENS expression. n/a, antibody staining inconclusive or incompatible with species. Discussion Recapitulation of developmental programs will be an important tool in future cell-engineering to treat enteric neuropathies. Through a comprehensive transcriptome and histochemical analysis in mouse and human developing ENS, we have here identified substantial numbers of transcription factors and signaling pathways with likely roles in stem cell maintenance/neurogenesis, neuronal specification/differentiation, and neural connectivity. A summary of key candidate genes with respect to plausible functions is depicted in Figure 7 and discussed as follows.Figure 7 Summary of genes with putative regulatory functions in the developing ENS. (A) Transcription factors in the developing ENS. Left column includes genes with putative functions in stem cell maintenance versus neurogenesis. Right column includes genes with putative functions in specification and/or differentiation of enteric neurons. (B) Signaling factors in the developing ENS. Ligands expressed in the ENS (left column) or in proximal gut tissue (right column) presented with putative receptors. Ligand/receptor couples are likely involved in proliferation/differentiation or migration/network formation as indicated. See Supplementary Tables 6 and 7 for more information.

s in the developing ENS. Ligands expressed in the ENS (left column) or in proximal gut tissue (right column) presented with putative receptors. Ligand/receptor couples are likely involved in proliferation/differentiation or migration/network formation as indicated. See Supplementary Tables 6 and 7 for more information. Candidate Transcription Factors in ENS Stem Cell Maintenance and Differentiation To ensure functional ganglia along the full extent of the gut, the relatively small number of ENSCs must balance extensive expansion with differentiation. The expression dynamics of SOX4, 5, 9, and 11 shown here suggest that they complement the currently known set of SOX genes (2, 8, and 10) in the sequential differentiation process of ENS stem cells to neurons, akin to their intricate functions in the developing spinal cord.17 Other novel genes with likely regulatory roles in proliferation/neurogenesis include, for instance, the Nzf-family (Myt1l, Myt1, and St18) (Figure 7A).

t of SOX genes (2, 8, and 10) in the sequential differentiation process of ENS stem cells to neurons, akin to their intricate functions in the developing spinal cord.17 Other novel genes with likely regulatory roles in proliferation/neurogenesis include, for instance, the Nzf-family (Myt1l, Myt1, and St18) (Figure 7A). Sox6, Dopamine Neurons, and Parkinson’s Disease In the central nervous system, the set of transcription factors expressed in progenitor cells as they commence neurogenesis initiate a molecular program of subtype-specific differentiation.19 Several lines of evidence show that enteric neuron specification follows a similar mechanistic logic. First, the subtype fate of an enteric neuron depends on the time point when it was generated during embryogenesis.10 Second, clonal lineage-tracing indicates that progenitors undergoing their last cell cycle are fixed to a particular neuronal subfate.20 Corroborating these findings, our study uncovered that SOX6 expression correlates with dopamine neuron birth11 and is necessary for acquisition of the dopaminergic trait. The transcription factor ASCL1 also regulates generation of gastric dopamine neurons; however, several other neuronal subtypes as well.9 SOX6 is thus the first transcription factor linked to the generation of a single neuronal subtype in the developing ENS.

11 and is necessary for acquisition of the dopaminergic trait. The transcription factor ASCL1 also regulates generation of gastric dopamine neurons; however, several other neuronal subtypes as well.9 SOX6 is thus the first transcription factor linked to the generation of a single neuronal subtype in the developing ENS. Within the midbrain, SOX6 is similarly important for the generation of substantia nigra dopamine neurons, and lower levels of SOX6 accompany the loss of these neurons in patients with Parkinson’s disease.21 Although Parkinson’s disease is classified as a movement disorder of the brain, most patients also suffer from GI dysfunction, including gastroparesis and constipation.22 Characteristic Parkinson pathology has been observed in TH+ (and vasoactive intestinal polypeptide–positive) enteric neurons, primarily in anterior gut regions.22 We note that the reduced numbers of TH+ gastric neurons and the gastric emptying phenotype in Sox6 mutant mice correlate well with these symptoms. Sophisticated solid gastric emptying tests and investigation of the lower-GI tract on large cohorts of Sox6 mutant mice could perhaps further delineate the role of TH+ neurons in gastrointestinal physiology. Mounting experimental evidence suggests that Parkinson pathology may initiate in the ENS and subsequently spread to the brain.22 TH+ gastric neurons may thus be relevant to understand early pathological events in Parkinson’s disease. We found a battery of genes, shared between gastric and brain dopamine neurons (Figure 3C),23, 24, 25, 26 indicating conservation of gene regulatory networks in different types of dopamine neurons, possibly forming a basis for further studies of dopamine neuron pathology.

rly pathological events in Parkinson’s disease. We found a battery of genes, shared between gastric and brain dopamine neurons (Figure 3C),23, 24, 25, 26 indicating conservation of gene regulatory networks in different types of dopamine neurons, possibly forming a basis for further studies of dopamine neuron pathology. Candidate Transcription Factors in ENS Cell Diversification We identified Hox genes with an unanticipated expression in enteric neurons. Reminiscent of this, sets of Hox proteins are differentially expressed in spinal motor neuron subclasses, where they function in neuronal subtype differentiation, morphogenesis, and synaptic specificity.18 To impose differential transcriptional read-out, Hox proteins depend on the interaction with members of the 3 amino acid loop extension (TALE) family. Four TALE genes were found in the developing ENS (Pbx2, Pbx3, Meis2, and Pknox1). A gene network analysis predicted a high level of interactions between the ENS-expressed TALE and Hox proteins (Supplementary Figure 8). We therefore propose that combinatorial expression of various HOX and TALE proteins may contribute to the acquisition of subtype-specific features in enteric neurons.

s2, and Pknox1). A gene network analysis predicted a high level of interactions between the ENS-expressed TALE and Hox proteins (Supplementary Figure 8). We therefore propose that combinatorial expression of various HOX and TALE proteins may contribute to the acquisition of subtype-specific features in enteric neurons. Our transcriptome screen also identified, for example, Foxd1, Cux2, and Tbx20 as plausible regulators of enteric neuronal specification, and Atoh1 and Runx1 in postmitotic subtype differentiation (Figure 7A). Even promiscuously expressed transcription factors (eg, Onecut2 and Hmx1) may contribute to differentiation programs in various enteric neuronal subtypes by forming distinct combinatorial expression patterns. Candidate Signaling Pathways Mediating Autonomous Development of the ENS This study demonstrates several signaling pathways with plausible roles in the fine-tuned orchestration of proliferation, neurogenesis, neuronal diversification, and network formation, which are needed for establishment of functional enteric networks (Figure 7B). IGF1, LTK, AHR, MDK, GDF10, Activin A, and NRG1 have been attributed neurotrophic, gliogenic, or neurogenic activities in developing nervous tissue.27, 28, 29, 30, 31, 32, 33 Thus, our identification of these signaling factors within the developing ENS indicates that immature enteric neurons could play an active role in controlling generation of the appropriate numbers and ratio of glia and neurons (Figure 7B).

or neurogenic activities in developing nervous tissue.27, 28, 29, 30, 31, 32, 33 Thus, our identification of these signaling factors within the developing ENS indicates that immature enteric neurons could play an active role in controlling generation of the appropriate numbers and ratio of glia and neurons (Figure 7B). The idea that phenotypically distinct neurons are generated in a temporally defined manner currently lacks a clear underlying molecular mechanism. One hypothesis states that ENS diversity is governed by self-regulatory mechanisms where early-born neurons would affect the genesis of later-born neurons.16 Among identified factors, FGF, TGFβ, and BMP7 have important instructive roles during the sequential generation of neuronal subtypes in other developing tissues,34, 35, 36 and are thus prime candidates as auto/paracrine regulators of ENS diversification (Figure 7B). Cell fate also can be influenced in a temporal manner by differential expression of modulatory proteins. CTGF is a modulatory factor that can adjust TGFβ and BMP7 signaling,37 whereas INHA forms inhibitory dimers with INHBA, thus preventing Activin A signaling.38 It is therefore possible that the differential levels of CTGF and INHA expression found in the ENS (Supplementary Figures 6 and 7) bestows ENS subpopulations a means to modify their response to signaling ligands.

d BMP7 signaling,37 whereas INHA forms inhibitory dimers with INHBA, thus preventing Activin A signaling.38 It is therefore possible that the differential levels of CTGF and INHA expression found in the ENS (Supplementary Figures 6 and 7) bestows ENS subpopulations a means to modify their response to signaling ligands. Establishment of fine-tuned connectivity patterns of appropriate synapses is required for the functionality of enteric networks. Recent studies of brain development have implicated SEMA4d-PLXNB1, NTNG1-LRRC4C, SLITRK2-PTPRO, SLITRK3-PTPRD, and FLRT–LPHN/UNC5D in the formation of selective synapses.39, 40, 41, 42, 43 Our identification of these signaling molecules in the developing ENS thus provides novel candidate genes with likely roles in mediating precise synaptic connections during enteric circuit formation.

TNG1-LRRC4C, SLITRK2-PTPRO, SLITRK3-PTPRD, and FLRT–LPHN/UNC5D in the formation of selective synapses.39, 40, 41, 42, 43 Our identification of these signaling molecules in the developing ENS thus provides novel candidate genes with likely roles in mediating precise synaptic connections during enteric circuit formation. Future Treatments of Enteric Neuropathies Cell-based regenerative medicine is considered a promising future therapeutic approach for Hirschsprung disease, but may initially be more applicable for disorders in which small phenotypic changes underlie ENS dysfunction.2 Substantial progress has been made in deriving ENSCs that can survive, proliferate, and migrate on transplantation; however, the challenge of controlling neuronal differentiation remains. We have here aspired to provide an extensive expression pattern resource of mouse and human developing ENS for basic understanding of ENS diversification and implementation in translational studies. The important role of Sox6 for gastric dopamine neurons encourages further exploration of this and other identified genes (Figure 7) for their possible capacity to influence subtype-specific or other aspects of ENS cell development; abilities that could advance the engineering of clinically relevant ENS cells. Notably, the expression signature of in vitro–derived ENS stem cells3 contained many of the genes we describe (eg, Tlx3, Hoxb2, and Tbx2). Cross-comparisons to our histological gene expression maps could therefore provide valuable insights into the cellular composition and maturity stage of human ENS cell cultures, being essential for further development of this tool.

ved ENS stem cells3 contained many of the genes we describe (eg, Tlx3, Hoxb2, and Tbx2). Cross-comparisons to our histological gene expression maps could therefore provide valuable insights into the cellular composition and maturity stage of human ENS cell cultures, being essential for further development of this tool. Supplementary Material Supplementary Table of Contents Supplementary Text Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Supplementary Figure 4 Supplementary Figure 5 Supplementary Figure 6 Supplementary Figure 7 Supplementary Figure 8 Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7 Supplementary Table 8 Acknowledgments We acknowledge Marjan Abbassi and Lynn Pieters for laboratory assistance, Graham Preece for technical expertise, and Dr Ruth Palmer for discussions. We also thank Prof Patrik Ernfors, Dr Marlene Hao, and Dr Reena Lasrado for constructive comments on the manuscript. Present affiliation of Rebecca Sadler: Institut für Schlaganfall- und Demenzforschung, Klinikum der Universität München, D-81377 Munich, Germany; and present affiliation of Catia Laranjeira: Fundacao para a Ciencia e Tecnologia, Unidade FCCN – Computacao Cientifica Nacional, 1700–066 Lisboa, Portugal. Conflicts of interest The authors disclose no conflicts.

Present affiliation of Rebecca Sadler: Institut für Schlaganfall- und Demenzforschung, Klinikum der Universität München, D-81377 Munich, Germany; and present affiliation of Catia Laranjeira: Fundacao para a Ciencia e Tecnologia, Unidade FCCN – Computacao Cientifica Nacional, 1700–066 Lisboa, Portugal. Conflicts of interest The authors disclose no conflicts. Funding This work was supported by the Knut and Alice Wallenberg FoundationKAW2008.0123, Swedish Research Council (VR) 521–2012–1676, EMBO, Swedish Medical Society (SLS), Ruth and Richard Julin Foundation, Magnus Bergvall Foundation, Ollie och Elof Ericssons Foundation, Åke Wiberg Foundation and Swedish Society for Medical Research. V.P. acknowledges the support of the UK Medical Research Council (MRC) and the Francis Crick Institute (which receives funding from MRC, Cancer Research UK, and the Wellcome Trust). Prof. Veronique Lefebvre generously contributed Sox6fl mice. Human embryonic and fetal tissue was provided by the KI Stem Cell and Tissue Bank. Author names in bold designate shared co-first authorship. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/j.gastro.2017.10.005.

Editor's Notes Background and Context The intestinal epithelium is thought to play a critical role in the pathogenesis of inflammatory bowel Diseases (IBD), yet evidence derived from primary human tissue remain scarce. New Findings Purified intestinal epithelium from children newly diagnosed with IBD display distinct epigenetic and transcriptional alternations, which are partly retained in organoid cultures and correlate with disease outcome. Limitations Relatively small patient numbers require validation in additional cohorts. Impact Stable epigenetic alterations in the intestinal epithelium of children with IBD may explain variations in disease outcome and have potential to be developed into disease prognostic biomarkers in the future. Inflammatory bowel diseases (IBD) cause chronic relapsing inflammation that can affect any segments of the digestive tract (ie, Crohn’s disease [CD]) or be restricted to the colon (ulcerative colitis [UC]).1, 2 Although these diseases can manifest at any age, approximately one quarter of patients3 are diagnosed in childhood or early adulthood, when the disease course and subsequent outcomes can be particularly severe.3

the digestive tract (ie, Crohn’s disease [CD]) or be restricted to the colon (ulcerative colitis [UC]).1, 2 Although these diseases can manifest at any age, approximately one quarter of patients3 are diagnosed in childhood or early adulthood, when the disease course and subsequent outcomes can be particularly severe.3 The etiology of IBD is multifactorial, although the interplay of factors is still poorly understood. Large-scale genome-wide association studies have helped to characterize the genetic risk, identifying over 200 disease-associated loci.4, 5 The striking overlap of genetic risk loci between CD, UC, and other immune-mediated diseases strongly suggests common immune regulatory pathways are affected in these conditions.4, 5 However, current estimates of the overall genetic contribution to IBD risk are still only 13% for CD and 8% for UC. The rapid increase in the incidence of IBD in recent decades,6, 7, 8 the stability of the human genome, the dysbiosis of the gut microbiome,9, 10, 11, 12 as well as epidemiologic evidence, all suggest an association between the rise in IBD and the recent changes in our environment.

still only 13% for CD and 8% for UC. The rapid increase in the incidence of IBD in recent decades,6, 7, 8 the stability of the human genome, the dysbiosis of the gut microbiome,9, 10, 11, 12 as well as epidemiologic evidence, all suggest an association between the rise in IBD and the recent changes in our environment. Epigenetic mechanisms operate at the interface between genetic predisposition and our environment, capable of causing stable, potentially heritable changes of cellular function in response to environmental triggers.13, 14 Consequently, epigenetics is being increasingly recognized as a highly plausible mechanism that may both initiate and then maintain intestinal mucosal inflammation in human IBD. A growing number of studies have reported IBD-associated alterations in epigenetic profiles, as well as associated changes in gene expression and/or cellular function. For example, DNA methylation (DNAm) changes in mucosal biopsies and peripheral blood mononuclear cells of both adults and children diagnosed with IBD have been demonstrated.15, 16 However, the vast majority of studies were performed on mixed cell tissue samples (eg, whole blood, peripheral blood mononuclear cells, or mucosal biopsies) and, possibly because of changes in cellular composition, demonstrated a strong effect of inflammation on the observed epigenetic changes. Importantly, advances made by epigenetic consortia such as ROADMAP,17 BLUEPRINT,18 as well as single-cell RNA sequencing, have all demonstrated the importance of studying individual, disease-relevant cell types to best identify molecular alterations involved in pathophysiology.

ation on the observed epigenetic changes. Importantly, advances made by epigenetic consortia such as ROADMAP,17 BLUEPRINT,18 as well as single-cell RNA sequencing, have all demonstrated the importance of studying individual, disease-relevant cell types to best identify molecular alterations involved in pathophysiology. Genetic and functional studies predominantly using mouse models and cell lines19, 20, 21 have provided strong evidence for impaired function of the intestinal epithelium in IBD. Yet, these models have done little to explain how the complex interplay between environmental factors, host genetics, intestinal cell function, and the adjacent microbiome lead to the development of the IBD phenotype and its subsequent evolution. To better elucidate specific alterations in this jigsaw, a genome-wide multi-layered omics approach of carefully selected primary cell samples is required. Importantly, in addition to unravelling novel aspects of disease pathogenesis, this approach in disease-relevant cell types (ie, the intestinal epithelium) could provide clinically relevant information. For example, in children and adults with IBD, it can be difficult to confidently distinguish CD from UC, with many patients remaining ‘unclassified’ despite disease progression. Intestinal epithelial cell (IEC)-specific ‘omics’ signatures have the potential to more rapidly and accurately diagnose the patient and, hence, improve the specificity of treatment management. Furthermore, variations of cell type-specific molecular profiles amongst IBD patients may be indicative of disease sub-phenotype and could therefore help to understand the large variations in disease behavior and outcome.

y and accurately diagnose the patient and, hence, improve the specificity of treatment management. Furthermore, variations of cell type-specific molecular profiles amongst IBD patients may be indicative of disease sub-phenotype and could therefore help to understand the large variations in disease behavior and outcome. Therefore, we simultaneously profiled genotype, epi-genotype (ie, DNAm), gene expression, and the adjacent gut microbiota of highly purified IEC, obtained from children newly diagnosed with IBD and a matched cohort of non-IBD controls. We analyzed genome-wide ‘omics’ layers for potential IBD-specific alterations and functional consequences, as well as cross-talk between layers. Additionally, we generated intestinal epithelial organoids from patient biopsy samples and investigated their epigenetic profiles. Lastly, we applied statistical models to genome-wide datasets to test their ability to distinguish between disease subtypes, as well as potential correlation with disease outcome measures.

Additionally, we generated intestinal epithelial organoids from patient biopsy samples and investigated their epigenetic profiles. Lastly, we applied statistical models to genome-wide datasets to test their ability to distinguish between disease subtypes, as well as potential correlation with disease outcome measures. Methods Patient Cohort A cohort of 66 treatment-naïve children at diagnosis of their IBD, along with 30 age- and sex-matched non-inflammatory control children, were recruited by the Paediatric Gastroenterology team at Addenbrooke’s Hospital during 2013–2016. This study was conducted with informed patient and/or carer consent as appropriate, and with full ethical approval (REC-12/EE/0482). Sample and patient details are provided in Supplementary Tables 1 and 2. Children with macroscopically and histologically normal mucosa who had a diagnosis of IBD ruled out served as the non-disease control group. Each patient’s final clinical diagnosis was based on the revised Porto criteria.22 At the diagnostic colonoscopy, additional mucosal biopsies were taken from the small bowel (ie, terminal ileum [TI]) and 2 large bowel sections (ie, ascending colon [AC] and sigmoid colon [SC]). A blood sample was taken for patient genotyping. Clinical phenotype and outcome data were prospectively recorded over a minimum of 18 months post-diagnosis (Supplementary Table 1). The inflammation status of a sample (inflamed vs non-inflamed) was based on the histology of a paired sample taken within 2 cm of samples at the time of the initial endoscopy. Longitudinal samples were taken from the TI and SC of a subset of patients that underwent repeat endoscopy (CD: n = 14; UC: n = 9).

ry Table 1). The inflammation status of a sample (inflamed vs non-inflamed) was based on the histology of a paired sample taken within 2 cm of samples at the time of the initial endoscopy. Longitudinal samples were taken from the TI and SC of a subset of patients that underwent repeat endoscopy (CD: n = 14; UC: n = 9). Purification of Intestinal Epithelium Biopsy samples were processed immediately and IECs purified using enzyme digestion and magnetic bead sorting for the epithelial cell adhesion molecule as described previously.16, 23 Mucus for the isolation of adjacent microbiota was collected during tissue processing from sieve and centrifugation supernatant, then pooled, pelleted, and stored at -80°C to extract DNA from the adjacent microbiota. Further information is provided in the Supplementary Methods. Human Intestinal Epithelial Organoid Culture Intestinal organoids were generated from mucosal biopsies by isolation of intestinal crypts and culturing as described previously and detailed in the Supplementary Methods section and Supplementary Table 4.24 DNA and RNA Extraction DNA and RNA were extracted simultaneously from the same sample using the AllPrep DNA/RNA mini kit (Qiagen, Hilden, Germany). DNA from the adjacent microbiota was extracted using QIAamp DNA Stool Mini Kit and from whole blood using the DNeasy Blood and Tissue Kit (both Qiagen). DNA was bisulfite-converted using Zymo DNA methylation Gold kit (Zymo Research, Irvine, CA, USA).

sample using the AllPrep DNA/RNA mini kit (Qiagen, Hilden, Germany). DNA from the adjacent microbiota was extracted using QIAamp DNA Stool Mini Kit and from whole blood using the DNeasy Blood and Tissue Kit (both Qiagen). DNA was bisulfite-converted using Zymo DNA methylation Gold kit (Zymo Research, Irvine, CA, USA). Arrays and Sequencing Genome-wide DNA methylation was profiled using the Illumina Infinium HumanMethylation450 and EPIC BeadChip platforms (Illumina, Cambridge, UK; Accession Number: E-MTAB-5463). Sample numbers are provided in Supplementary Table S3. Expression profiling was performed using RNA-sequencing (RNA-seq) at the University of Kiel, Germany using an established pipeline as described previously.10 (Project accession number: E-MTAB-5464). Patient genotyping was performed using the Illumina OmniExpressExome-8 BeadChip Kit. 16S rRNA gene profiling of the adjacent microbiota was performed at the Wellcome Trust Sanger Centre (Hinxton, Cambridge). The 16S microbiota data can be found under EBI study ID PRJEB6663. For further details of the arrays and sequencing, please see the Supplementary Methods. Locus-specific validation of DNA methylation profiles was performed on bisulfite-converted DNA after polymerase chain reaction amplification using the Pyromark Q24 (Qiagen) pyrosequencing system as described previously.16

Arrays and Sequencing Genome-wide DNA methylation was profiled using the Illumina Infinium HumanMethylation450 and EPIC BeadChip platforms (Illumina, Cambridge, UK; Accession Number: E-MTAB-5463). Sample numbers are provided in Supplementary Table S3. Expression profiling was performed using RNA-sequencing (RNA-seq) at the University of Kiel, Germany using an established pipeline as described previously.10 (Project accession number: E-MTAB-5464). Patient genotyping was performed using the Illumina OmniExpressExome-8 BeadChip Kit. 16S rRNA gene profiling of the adjacent microbiota was performed at the Wellcome Trust Sanger Centre (Hinxton, Cambridge). The 16S microbiota data can be found under EBI study ID PRJEB6663. For further details of the arrays and sequencing, please see the Supplementary Methods. Locus-specific validation of DNA methylation profiles was performed on bisulfite-converted DNA after polymerase chain reaction amplification using the Pyromark Q24 (Qiagen) pyrosequencing system as described previously.16 Bioinformatics Analyses Extensive details of the bioinformatics methods used in this publication are described and referenced in the Supplementary Methods. Briefly, DNAm analyses were performed using minfi,25 sva,26 DMRcate,27 and limma28 R packages. RNAseq data was processed using established workflows.10 Microbiota composition analysis was performed using QIIME and phyloseq.29 Differential analysis was performed using limma28 for DNAm data (false discovery rate [FDR] <0.01) and DESeq230 for gene expression data (FDR <0.01 and log fold change >±0.5). InnateDB31 and the Reactome pathways were used to perform pathway enrichment analysis for the disease signatures identified from omics data layers. The diagnostic potential of omics data layers was tested using random forest classification models. Diagnostic accuracy was assessed via area under the receiver operator characteristic curve (AUC), precision scores, and receiver operator characteristic (ROC) curves. Weighted Gene Co-expression Network Analysis (WGCNA)32 was applied to gene expression and DNAm data for each diagnosis (CD and UC) by gut segment to correlate omics datasets with clinical phenotypic variables.

iver operator characteristic curve (AUC), precision scores, and receiver operator characteristic (ROC) curves. Weighted Gene Co-expression Network Analysis (WGCNA)32 was applied to gene expression and DNAm data for each diagnosis (CD and UC) by gut segment to correlate omics datasets with clinical phenotypic variables. Results DNA Methylation and Gene Expression Profiling of Purified Intestinal Epithelium Reveals Gut Segment-specific and Disease-associated Alterations To investigate IEC pathophysiology in pediatric IBD, we first performed unsupervised analysis of genome-wide DNAm, gene expression, and 16S microbial profiles generated from a total of 170 samples (Figure 1A and Table 1). Multidimensional scaling (MDS) plots indicate sample similarity/differences based on all data points included. MDS plots of DNAm and gene expression profiles revealed distinct clustering of samples by gut segment separating all TI-derived epithelium from colonic (ie, AC and SC) samples (Figure 1Bi and 1Bii). Moreover, samples derived from controls clustered closely both on an epigenetic (DNAm) and transcriptomic level for each gut segment. Interestingly, IBD-derived samples displayed more variation, with a subset of IBD samples distinctly separating from controls (Figure 1Bi, Bii and Supplementary Figure 1). In contrast to DNAm and gene transcription, no clear separation was evident from the 16S microbial community profiles using the same MDS approach (Figure 1Biii). However, analysis of the bacterial operational taxonomic unit by family abundance and alpha-diversity did reveal variation by gut segment (Supplementary Figure 2) and reduction in species diversity for CD patients (Supplementary Figure 2).Figure 1 Overview of study design and multi-dimensional scaling (MDS) analysis of genome-wide datasets. (A) Outline of study design. (B) MDS plots for each dataset: (i) DNAm based on batch corrected M-values; (ii) r-log normalized RNAseq gene expression counts; (iii) gut microbiota 16S operational taxonomic units normalized counts. Samples are labelled according to diagnosis (CD, Crohn’s disease; UC, ulcerative colitis; control) and gut segment. Schematic in part A adapted from Tauschmann et al.41

batch corrected M-values; (ii) r-log normalized RNAseq gene expression counts; (iii) gut microbiota 16S operational taxonomic units normalized counts. Samples are labelled according to diagnosis (CD, Crohn’s disease; UC, ulcerative colitis; control) and gut segment. Schematic in part A adapted from Tauschmann et al.41 Table 1 Summary of Patients, Samples, and Generated Datasets DNAm RNAseq 16S sequencing Genotype TI AC SC TI AC SC TI AC SC Total samples at diagnosis 162 81 170 62 Total individuals 73 15 74 33 15 33 58 53 59 62 CD 31 5 32 11 5 11 21 21 22 24 450K cohort 13 5 13 EPIC cohort 17 18 Organoids 5 5 UC 18 5 18 11 5 11 18 16 18 18 450K cohort 13 5 13 EPIC cohort 5 5 Controls 24 5 24 11 5 11 19 16 19 20 450K cohort 14 5 14 EPIC cohort 3 3 Organoids 7 7 Repeat endoscopies UC 9 9 CD 14 14 AC, ascending colon; CD, Crohn’s disease; SC, sigmoid colon; TI, terminal ileum; UC ulcerative colitis. Taken together, initial unsupervised analysis of genome-wide intestinal epithelial profiles reveals highly gut segment-specific signatures and suggests disease-associated alterations of DNAm and gene expression.

DNAm RNAseq 16S sequencing Genotype TI AC SC TI AC SC TI AC SC Total samples at diagnosis 162 81 170 62 Total individuals 73 15 74 33 15 33 58 53 59 62 CD 31 5 32 11 5 11 21 21 22 24 450K cohort 13 5 13 EPIC cohort 17 18 Organoids 5 5 UC 18 5 18 11 5 11 18 16 18 18 450K cohort 13 5 13 EPIC cohort 5 5 Controls 24 5 24 11 5 11 19 16 19 20 450K cohort 14 5 14 EPIC cohort 3 3 Organoids 7 7 Repeat endoscopies UC 9 9 CD 14 14 AC, ascending colon; CD, Crohn’s disease; SC, sigmoid colon; TI, terminal ileum; UC ulcerative colitis. Taken together, initial unsupervised analysis of genome-wide intestinal epithelial profiles reveals highly gut segment-specific signatures and suggests disease-associated alterations of DNAm and gene expression. Disease-specific Alterations in IEC Epigenetic and Transcriptional Profiles are Partly Independent of Inflammation Status Given the distinct clustering patterns we observed on a genome-wide scale, we next used a variance component model to assess the relative contribution of diagnosis and inflammation to the observed variance within each data layer, by gut segment. As shown in Figure 2A, the variation explained by disease (ie, diagnosis) exceeds that of inflammation in the majority of datasets. This highlights the presence of disease-specific molecular alterations, which are partly independent of the current inflammatory activity. Full results of the variance decomposition analysis can be found in Supplementary Figure 3. To extend these findings, we performed separate differential analysis (gene expression and DNAm) of inflammation and disease in the colonic epithelium. This allowed us to identify, for each CpG site or gene, the relative significance of inflammation and disease. Results showed that a majority of the differentially methylated positions (DMPs) between CD or UC and controls are primarily driven by diagnostic status and not inflammation (FDR <0.01, Figure 2B and C). Similarly, diagnosis explained 74% and 82% of the differentially expressed genes (DEGs) for CD and UC, respectively (Figure 2D and E).Figure 2 Contribution of diagnosis and inflammation to variance within each data layer. (A) Bar chart of the explained variance by diagnosis and inflammation across each dataset separated by gut segment. (B–E) Scatterplot of P values derived from differential DNAm (I and I) and gene expression (I and I) in sigmoid colon (SC) samples. For each CpG or gene, P values were generated for the comparison between Crohn’s disease (CD)/ulcerative colitis (UC) and control, and inflammation status (ie, inflamed vs non-inflamed). CpGs and genes with significant P values are plotted in purple for inflammation, in red for diagnosis, and in green if significant for both comparisons. Adjusted P < .01 was considered as significant.

tween Crohn’s disease (CD)/ulcerative colitis (UC) and control, and inflammation status (ie, inflamed vs non-inflamed). CpGs and genes with significant P values are plotted in purple for inflammation, in red for diagnosis, and in green if significant for both comparisons. Adjusted P < .01 was considered as significant. Additionally, generating MDS plots by labelling samples according to gut segment, diagnosis, and inflammatory status did not show a clear separation between inflamed and non-inflamed samples (Supplementary Figure S1). Lastly, we also tested for the potential impact of inflammation on cellular composition in our purified samples by generating a gene expression heatmap of common epithelial and immune cell marker genes (Supplementary Figure S4). Although a number of genes were found to be differentially expressed between IBD and control samples (eg, DEFA5, DEFA6, LYZ, PLA2G2A, CD40, CD44), none of the marker genes correlated with inflammatory status, suggesting minimal impact of epithelial cell composition and immune cell contamination on the observed disease-specific molecular changes (Supplementary Figure S4). In summary, these analyses demonstrate the presence of clear epigenetic, transcriptomic, and adjacent microbial alterations in the intestinal epithelium of children newly diagnosed with IBD, with a proportion being independent of intestinal inflammation.

Additionally, generating MDS plots by labelling samples according to gut segment, diagnosis, and inflammatory status did not show a clear separation between inflamed and non-inflamed samples (Supplementary Figure S1). Lastly, we also tested for the potential impact of inflammation on cellular composition in our purified samples by generating a gene expression heatmap of common epithelial and immune cell marker genes (Supplementary Figure S4). Although a number of genes were found to be differentially expressed between IBD and control samples (eg, DEFA5, DEFA6, LYZ, PLA2G2A, CD40, CD44), none of the marker genes correlated with inflammatory status, suggesting minimal impact of epithelial cell composition and immune cell contamination on the observed disease-specific molecular changes (Supplementary Figure S4). In summary, these analyses demonstrate the presence of clear epigenetic, transcriptomic, and adjacent microbial alterations in the intestinal epithelium of children newly diagnosed with IBD, with a proportion being independent of intestinal inflammation. Differential Methylation Analysis Reveals Disease-specific Signatures That Affect Gene Transcription Next we performed differential DNAm and gene expression analyses by comparing control, CD-, and UC-derived datasets for each gut segment. When performing these analyses, inflammation was controlled for within the differential analysis thereby allowing us to focus on molecular alterations that occur in relative independence of mucosal inflammation. Additionally, in an attempt to connect epigenetic and transcriptomic signatures, we identified differentially methylated regions (DMRs) that were located within 10kb of the transcription start site of a DEG. Such regions were termed regulatory DMRs (rDMRs).

s that occur in relative independence of mucosal inflammation. Additionally, in an attempt to connect epigenetic and transcriptomic signatures, we identified differentially methylated regions (DMRs) that were located within 10kb of the transcription start site of a DEG. Such regions were termed regulatory DMRs (rDMRs). Analysis of ileal IECs revealed CD-specific changes in both DNAm (Figure 3Ai and Supplementary Tables 5 and 6) and gene expression (Figure 3Aii and Supplementary Tables 7 and 8), when compared with either controls or UC, with a proportion overlapping between the 2 comparisons. In contrast, no significant DMPs or DEGs were identified when comparing UC with controls. Importantly, amongst identified rDMRs, several have previously been reported to be associated with IBD (eg, CASP133 and APOA19) (Figure 3B).Figure 3 Differential DNAm and gene expression analysis were performed separately for terminal ileum (TI) (A and B) and sigmoid colon (SC) (C and D), taking mucosal inflammation into account. (A and C) Venn diagrams of significant differentially methylated positions (DMPs), differentially expressed genes (DEGs), and regulatory DMRs (rDMRs). (B and D) Example of disease-specific rDMRs displaying DNA methylation levels expressed as Beta value on the y-axis in the left panel separately for TI and SC samples in the upper and lower panel, respectively. Beta value of 0 represents un-methylated, while 1 represents fully methylated CpG site. Genomic location is indicated on the x-axis. The middle panel displays identified rDMR (enlarged). The right panel displays a boxplot of the respective gene expression according to diagnosis. (B) rDMR within the APOA1 identified in TI-derived epithelium of children diagnosed with CD. (D) rDMR within the BACH2 gene identified in colonic IEC.

dicated on the x-axis. The middle panel displays identified rDMR (enlarged). The right panel displays a boxplot of the respective gene expression according to diagnosis. (B) rDMR within the APOA1 identified in TI-derived epithelium of children diagnosed with CD. (D) rDMR within the BACH2 gene identified in colonic IEC. Contrary to the ileum, changes observed in the SC reflected a ‘common IBD’ signature, with a major overlap between UC and CD signatures and only a single significant DEG (RARRES3 [Retinoic Acid Receptor Responder 3]) identified between the 2 diagnoses (Figure 3C and Supplementary Tables 9–14). RARRES3 is thought to have growth inhibitory and cell differentiation activities. One example of an rDMR that jointly affects CD and UC in the colon is BACH2 (Figure 3D), a transcription regulator, where a decrease in DNAm matched the increase in gene expression levels in both CD and UC patients. Interestingly, a proportion of the CD-related changes identified in TI samples were also found to be present in SC samples (Supplementary Figure 5). Overall, these results indicate that CD-specific DNAm and gene-expression changes are present in ileal IECs. In contrast, molecular changes observed in the colonic epithelium revealed a major overlap between CD and UC, reflecting a ‘common IBD’ signature.

Contrary to the ileum, changes observed in the SC reflected a ‘common IBD’ signature, with a major overlap between UC and CD signatures and only a single significant DEG (RARRES3 [Retinoic Acid Receptor Responder 3]) identified between the 2 diagnoses (Figure 3C and Supplementary Tables 9–14). RARRES3 is thought to have growth inhibitory and cell differentiation activities. One example of an rDMR that jointly affects CD and UC in the colon is BACH2 (Figure 3D), a transcription regulator, where a decrease in DNAm matched the increase in gene expression levels in both CD and UC patients. Interestingly, a proportion of the CD-related changes identified in TI samples were also found to be present in SC samples (Supplementary Figure 5). Overall, these results indicate that CD-specific DNAm and gene-expression changes are present in ileal IECs. In contrast, molecular changes observed in the colonic epithelium revealed a major overlap between CD and UC, reflecting a ‘common IBD’ signature. Pathway Enrichment Analysis of Identified rDMRs Reveal Both Common IBD and Disease-specific Pathways The intestinal epithelium serves a wide range of functions as a physical, chemical and immunological barrier and a bridge between the innate and adaptive immune response.34 We used pathway enrichment analysis to investigate functional pathways (of rDMRs) that may be altered at diagnosis in a child with IBD. A wide variety of immune system-, metabolism-, and signal transduction-related pathways were significantly enriched (Figure 4). Many of the immune system-related pathways (eg, interferon signaling and immuno-regulatory interactions) are shared between the gut segments and diagnoses; suggesting common alterations are present in IBD. Moreover, several of the significantly enriched pathways have previously been implicated in either IBD pathogenesis or IEC function (Figure 4).Figure 4 Pathway enrichment analysis of disease-specific regulatory DMRs (rDMRs). Pathway enrichment analysis was performed on identified rDMRs derived from the 3 comparisons between Crohn’s disease (CD) vs controls in terminal ileum (TI) and sigmoid colon (SC) samples (left and middle panel) and ulcerative colitis (UC) vs controls in SC samples (right panel). Analysis was performed using InnateDB and Reactome database and significant enrichment of individual pathways is displayed as the -log10 (adjusted P value).

(CD) vs controls in terminal ileum (TI) and sigmoid colon (SC) samples (left and middle panel) and ulcerative colitis (UC) vs controls in SC samples (right panel). Analysis was performed using InnateDB and Reactome database and significant enrichment of individual pathways is displayed as the -log10 (adjusted P value). IBD-associated intestinal Epithelial-specific Epigenetic Alterations are Stable Over Time and Partly Retained in Ex-vivo Organoid Culture Next, we investigated the stability of IEC DNAm profiles in IBD patients over time. We obtained ileal and colonic biopsies (SC) from IBD patients both at diagnosis and at a later stage in their disease (n = 14 CD, n = 9 UC). Strikingly, CD- and UC-associated DMPs showed remarkable stability over time, demonstrated by the strong correlations of the methylation values at diagnosis and repeat endoscopy within each gut segment (Figure 5A and Supplementary Figure 6). This was in spite of changes to the underlying mucosal inflammatory status (see Supplementary Table 1). To further test the stability and potential inflammation independence of disease-specific epigenetic alterations in IBD-derived IEC, we generated patient-derived intestinal organoids from an additional cohort of children newly diagnosed with CD (n=5) and matched healthy controls (n=7). Expansion of mucosal crypts from TI and SC in culture gave rise to 3-dimensional organoids (Figure 5B). Organoids derived from CD patients did not differ in their microscopic appearance or culturing behavior from those derived from healthy controls (Figure 5B). However, assessing their genome-wide DNA methylation profiles revealed distinct alterations, suggesting that they retain a proportion of disease-associated epigenetic changes. Despite the relatively small sample number, CD-associated DMPs (ie, identified in Figure 3) showed a clear trend to be also differentially methylated in CD-derived compared with control organoids. This was indicated by the presence of inflated P values (larger difference between observed vs expected P values) of CD-associated DMPs compared with randomly selected CpGs (Figure 5C).

(ie, identified in Figure 3) showed a clear trend to be also differentially methylated in CD-derived compared with control organoids. This was indicated by the presence of inflated P values (larger difference between observed vs expected P values) of CD-associated DMPs compared with randomly selected CpGs (Figure 5C). Using locus-specific pyrosequencing, we were able to validate a subset of CD-specific DMPs that were retained in organoid cultures (Figure 5D and 5E).Figure 5 Stability of disease-associated intestinal epithelial DNA methylation changes: (A) Correlation plot of DNA methylation (Beta values) of disease-associated differentially methylated positions (DMPs) at diagnosis and at repeat endoscopy for each patient at the 2 time points. Shown are Crohn’s disease (CD)-associated DMPs (left) and ulcerative colitis (UC)-associated DMPs (right) in sigmoid colon (SC) epithelium (adjusted P <. 01). (B) Brightfield microscopic images of fully grown intestinal epithelial organoids derived from 2 gut segments (ie, terminal ileum [TI] and SC) of CD and control patients. (C) Quantile-quantile plot generated from organoid-derived genome-wide DNAm P values. Plotted are P values (observed vs expected) comparing specific CD-associated DMPs (from Figure 3) for each gut segment with randomly selected CpGs. (D) Examples of CD-associated DMPs being retained in patient-derived organoids. Plotted are beta values derived from genome-wide array data generated from purified colonic epithelium and respective organoids. GREB1, Growth Regulation By Estrogen In Breast Cancer 1; TMEM173, Transmembrane Protein 173; PDE1B, Phosphodiesterase 1B; CtrlP, Control purified IEC (n = 14); CDP, CD purified IEC (n = 13); CtrlO, Control organoids (n = 7); CDO, CD organoids (n = 5). (E) Validation of CpGs shown in D. Validation of genome-wide DNAm data using pyrosequencing. n = 5–7 per group; *P < .05; ***P < .001; unpaired, 2-tailed t-test between Ctrl and CD.

rase 1B; CtrlP, Control purified IEC (n = 14); CDP, CD purified IEC (n = 13); CtrlO, Control organoids (n = 7); CDO, CD organoids (n = 5). (E) Validation of CpGs shown in D. Validation of genome-wide DNAm data using pyrosequencing. n = 5–7 per group; *P < .05; ***P < .001; unpaired, 2-tailed t-test between Ctrl and CD. Together, these data demonstrate that disease-associated epigenetic alterations in the intestinal epithelium are stable over time and are at least in part retained in ex-vivo organoid cultures. Intestinal Epithelial DMRs, DEGs, and rDMRs are Enriched Around Genetic IBD Risk Loci Genome-wide association studies have successfully identified over 200 loci predisposing to IBD.4, 5 However, limited information is currently available on the potential mechanisms involved in mediating genetic risk and/or which cell types are particularly affected. Here we used our epithelial cell-derived molecular signatures to test for an enrichment of disease-specific DMRs, DEGs, and rDMRs within genomic IBD risk loci. We observed highly significant enrichment of DMRs, DEGs, and rDMRs in both colonic and ileal IECs for IBD risk loci, while limited enrichment was found for genetic variants that have been linked to other multifactorial diseases with an immune-mediated pathogenesis, such as Type 1 diabetes and multiple sclerosis (Supplementary Figure 7). Together these results suggest that interactions between the IBD risk loci and DNAm and/or transcription may occur in children carrying disease variants.

that have been linked to other multifactorial diseases with an immune-mediated pathogenesis, such as Type 1 diabetes and multiple sclerosis (Supplementary Figure 7). Together these results suggest that interactions between the IBD risk loci and DNAm and/or transcription may occur in children carrying disease variants. IEC DNA Methylation and Gene Expression Signatures Accurately Predict Disease Status and Correlate With Clinical Outcome Measures Given the striking IBD-associated changes observed in intestinal epithelial DNAm and gene expression, we went on to test the ability of these signatures to predict diagnosis. Additionally, we hypothesized that variation observed within IBD-derived patient samples could be indicative of future disease behavior and outcome. To address these hypotheses, we applied a machine-learning model (random forest) to the individual omics data layers. The model identified those data points (eg, CpGs, genes, operational taxonomic units) that could predict disease status for each patient with high precision and accuracy (see Supplementary Methods section for further details). As demonstrated in Figure 6, DNAm data derived from either gut segment produced a model with a high AUC (>0.8) (Figure 6A). The best model separating disease from control was based on DNAm data from the SC (AUC=0.94, cross-validation (CV)=40) with sensitivity of 75% and specificity of 100% (Figure 6Ai and 6Aii). Importantly, the use of ileal DNAm datasets allowed separation between CD and UC with high precision (77%), sensitivity (57%), and specificity (93%) (AUC=0.92, CV=24) (Figure 6B). The accuracy of the TI DNAm signatures in distinguishing CD and UC was confirmed in a follow-up patient cohort, analyzed using a second DNAm array platform (Illumina EPIC array, see Methods and Supplementary Figure 8). Full details of the models, including AUC, sensitivity, and specificity, can be found in Supplementary Table 15. In contrast, models built using the IBD risk loci from our patient genotyping data yielded the lowest model score (AUC=0.49) (Figure 6Ai).Figure 6 Correlation of intestinal epithelial cells (IEC)-specific molecular signatures with diagnosis and clinical outcome measures: IEC-derived epigenetic, transcriptomic, and microbial signatures were tested for their potential to predict diagnostic status (A and B) and correlation with disease outcome parameters (C–F).

re 6 Correlation of intestinal epithelial cells (IEC)-specific molecular signatures with diagnosis and clinical outcome measures: IEC-derived epigenetic, transcriptomic, and microbial signatures were tested for their potential to predict diagnostic status (A and B) and correlation with disease outcome parameters (C–F). (Ai) Bar chart indicating area under the curve (AUC) of the best model to accurately differentiate samples based on diagnosis (ie, IBD vs controls). (Aii) ROC curve for the best diagnostic model (inflammatory bowel disease [IBD] vs control) using colonic DNAm data. (Bi) Bar chart of the AUC of the best models to differentiate between Crohn’s disease (CD) and ulcerative colitis (UC). (Bii) Receiver operator characteristic (ROC) curve for the best model separating CD from UC using ileal IEC DNAm data. (C) Weighted Gene Co-Expression Network Analysis (WGCNA) of CD terminal ileum (TI)-derived RNA-Seq data showing correlations between key gene-expression modules and clinical parameters. Each cell on the heatmap displays Pearson correlation coefficient and corresponding P value. Outlined cells indicate significant correlations. (D) Heatmap and hierarchical clustering of patients based on gene expression (ie, RNAseq counts) for strongest module. (Dii and Diii) Kaplan-Meier curves based on patient grouping derived from 7Di, ie, top gene expression module for use of biologics and time to third treatment escalation during 75 weeks of follow-up (n = 10 patients, P = .049 and P = .032, log-rank test). (Ei) Heat-map and hierarchical clustering of CpGs within strongest module identified by applying WCGNA to CD TI DNA methylation profiles. (Eii and Eiii) Kaplan Meier curves based on patient grouping derived from 7Ei for use of biologics (Eii) and time to third treatment escalation (Eiii) during 75 weeks follow-up (n = 29 patients, P = .025 and P = .043, log-rank test). (F) Venn-diagram showing the overlap between annotated genes that were present in the top modules for both gene expression and DNA methylation.

ing derived from 7Ei for use of biologics (Eii) and time to third treatment escalation (Eiii) during 75 weeks follow-up (n = 29 patients, P = .025 and P = .043, log-rank test). (F) Venn-diagram showing the overlap between annotated genes that were present in the top modules for both gene expression and DNA methylation. To correlate genome-wide IEC profiles with clinical outcome measures (including binary, numerical, and categorical parameters) we used a Weighted Gene Co-expression Network Analysis (WGCNA) approach. WGCNA identifies patterns within a given dataset (ie, RNAseq data) and combines genes or CpGs that vary similarly across samples into modules. Each of these modules was then tested for a significant correlation with clinical outcomes. The application of WGCNA to RNAseq data derived from the TI of CD patients led to the identification of several gene modules (ie, groups of genes) that correlate significantly (correlation >±0.6, P < .05) with a number of disease outcome measures including the requirement for treatment with biologics and number of treatment escalations within the first 18 months following diagnosis (Figure 6C). Interestingly, modules correlating with disease outcome measures did not show any correlation with gender, age, or disease phenotype at diagnosis (eg, abdominal pain, diarrhea, and Pediatric Crohn’s Disease Activity Index). Clustering all samples according to expression levels of genes within the strongest modules separated CD patients in 2 groups (Figure 6Di). Kaplan Meier curves for these groups demonstrated striking differences in both the requirement for biologics and time to third treatment escalation (Figure 6Dii and 6Diii). In addition, we applied WGCNA to the DNAm data. Although the overall correlation of identified modules was less striking, separating samples according to DNAm profiles of the strongest module still demonstrated a significant difference in outcome measures between resulting patient groups (Figure 6Ei–6Eiii). Similar results were obtained from UC patient-derived signatures in SC samples (Supplementary Figure 9). Finally, comparing annotated genes from the top modules identified in RNAseq and DNAm datasets revealed an overlap of 57% and 79%, respectively, suggesting that expression signatures might be in part underpinned by stable epigenetic changes (Figure 6F).

patient-derived signatures in SC samples (Supplementary Figure 9). Finally, comparing annotated genes from the top modules identified in RNAseq and DNAm datasets revealed an overlap of 57% and 79%, respectively, suggesting that expression signatures might be in part underpinned by stable epigenetic changes (Figure 6F). Based on these preliminary results, both DNAm and RNAseq data contain signatures that accurately predict disease status and correlate with selected disease outcome parameters. Discussion Substantial evidence suggests that impaired function of the intestinal epithelium plays a major role in IBD pathogenesis. However, our current understanding of the exact mechanisms remains limited. It is also becoming increasingly clear that functional alterations in complex disease are likely to be caused by and/or result in a multifaceted interplay between several layers of cellular regulation. Specific to the GI tract, the intestinal microbiota adds further complexity because it has been shown to influence cellular function of the intestinal epithelium both in health and IBD.9, 10, 12 Given the wide range of phenotypes and diverse spectrum of disease behavior within the conditions we currently label as CD and UC, we urgently require novel molecular signatures to allow better classification of clinically relevant disease entities.

function of the intestinal epithelium both in health and IBD.9, 10, 12 Given the wide range of phenotypes and diverse spectrum of disease behavior within the conditions we currently label as CD and UC, we urgently require novel molecular signatures to allow better classification of clinically relevant disease entities. Here, we applied a multi-omics profiling approach to a highly purified IEC sample set obtained from a prospectively recruited, treatment-naïve, pediatric patient cohort. Unsupervised analysis revealed fundamental differences in the methylation and gene expression profiles by gut segment.9, 10, 16 Moreover, within each gut segment, we observed distinct disease-specific variation in both DNAm and gene expression, which were found to be partly independent of the presence or absence of microscopic mucosal inflammation. Specifically, we found that the majority of DNAm and RNAseq disease signatures from the SC were not primarily explained by inflammation status. This strongly suggests that there are underlying epigenetic and transcriptomic changes within IECs in IBD patients, which are present irrespective of inflammatory activity. Our findings further expand on previous studies using whole gut biopsies, which reported major transcriptional or epigenetic changes that were primarily associated with the presence of mucosal inflammation.9, 10 Although we also observed a strong, inflammation-associated signal, purification of the intestinal epithelium (and thereby removal of infiltrating immune cells) has allowed identification of a cell type-specific signature that does not seem to be exclusively driven by mucosal inflammation.

of mucosal inflammation.9, 10 Although we also observed a strong, inflammation-associated signal, purification of the intestinal epithelium (and thereby removal of infiltrating immune cells) has allowed identification of a cell type-specific signature that does not seem to be exclusively driven by mucosal inflammation. In contrast to genome-wide DNAm and gene transcription profiles, unsupervised MDS analysis of our 16S data did not show any specific sample clusters and/or clear association with key phenotypes such as gut segment, disease entity or inflammation. This is most likely because of the large inter-individual and intra-individual variation; an observation that has been previously reported by others.9, 11, 12 However, supervised analyses revealed gut segment as well as disease-associated changes in microbial composition. Analysis of the 16S data in combination with the epithelial omics data was unable to identify strong correlations between DEGs and 16S abundances or dose-dependent relationships for subgroups of patients. Nevertheless, we consider the fact that our 16S data was generated from microbes isolated from individual gut segments as novel and therefore potentially highly valuable as reference for future work in this rapidly evolving field.

etween DEGs and 16S abundances or dose-dependent relationships for subgroups of patients. Nevertheless, we consider the fact that our 16S data was generated from microbes isolated from individual gut segments as novel and therefore potentially highly valuable as reference for future work in this rapidly evolving field. Further investigating disease-specific DNAm and gene expression changes, we were able to identify a number of significant DEGs and DMRs, a proportion of which overlapped (rDMRs), indicating a functional interconnection between the 2 data layers. Reassuringly, a number of identified genes had previously been reported, including APOA112 and CASP1.9 When comparing identified DMRs, DEGs, and rDMRs, we discovered that significant changes in the TI were only present in CD-derived samples. In contrast, analysis in the colonic epithelium showed both CD- and UC-specific changes, which also displayed a major overlap likely reflecting common phenotypic features shared between the 2 conditions. Together these data suggest the presence of a CD-specific signature in the TI epithelium and a common IBD signature in the SC. The identification of shared, enriched pathways for the 2 diagnoses further supports this hypothesis. Additionally, enrichment for pathways implicated in the cross-talk between cells of the innate and adaptive immune response highlights the important role of the intestinal epithelium in orchestrating intestinal host defense and suggests that alterations in these key processes may lead to the initiation and/or persistence of gut inflammation in IBD.

athways implicated in the cross-talk between cells of the innate and adaptive immune response highlights the important role of the intestinal epithelium in orchestrating intestinal host defense and suggests that alterations in these key processes may lead to the initiation and/or persistence of gut inflammation in IBD. The impact of the observed IEC-specific epigenetic alterations on IBD pathogenesis will depend at least in part on the stability of such molecular signatures. Investigating IEC DNAm profiles of the same patient at 2 time points (ie, at diagnosis and at later disease stage) allowed us to demonstrate the strikingly high stability of disease-associated methylation signatures in small bowel and colonic IEC. This was despite changes in medication and mucosal inflammation over a period of up to 20 months. These findings suggest that stable epigenetic alterations may contribute to chronic relapsing inflammation by mediating altered IEC function. Interestingly, CD-derived epithelium appeared to retain a degree of disease-specific alterations even when cultured ex-vivo as organoids, further highlighting both their stability and relative independence of mucosal inflammation. Additionally, our findings add further support to recent reports on patient-derived intestinal organoids to be used as novel translational research tools.35

disease-specific alterations even when cultured ex-vivo as organoids, further highlighting both their stability and relative independence of mucosal inflammation. Additionally, our findings add further support to recent reports on patient-derived intestinal organoids to be used as novel translational research tools.35 Despite the major success of genome-wide association studies in identifying disease-predisposing genetic loci, information on the functional consequences and cell specificity remain limited. Expression quantitative trait loci have been identified for a subset of the IBD risk loci from whole biopsies36 and blood cell subsets.37 More recently, differences in DNAm and chromatin conformation38 were also identified for a subset of the IBD risk loci in immune cells.39 Our study adds further detail and specificity by demonstrating enrichment of disease-specific DMRs, DEGs, and rDMRS within IBD susceptibility loci4, 5 in both gut segments.

ubsets.37 More recently, differences in DNAm and chromatin conformation38 were also identified for a subset of the IBD risk loci in immune cells.39 Our study adds further detail and specificity by demonstrating enrichment of disease-specific DMRs, DEGs, and rDMRS within IBD susceptibility loci4, 5 in both gut segments. An additional major strength of our prospectively recruited pediatric patient cohort was the availability of detailed phenotype and disease outcome data, allowing us to test for potential correlation between molecular signatures and clinical phenotypes. Despite the relatively small sample numbers included in these analyses, both the potential diagnostic and prognostic value of our IEC signatures is evident. While current diagnostic approaches are sufficient for most patients, a minority of patients requires repeated and prolonged investigations to confirm a diagnosis. Additionally, it is frequently challenging to differentiate UC from CD in children, both at diagnosis and later in the disease course. Therefore, a diagnostic model to reliably differentiate CD from UC, such as the model built using DNAm data from the ileum with high sensitivity and specificity, could be of clinical value. Correlating genome-wide molecular signatures with clinical outcome measures continues to be a major challenge and a wide range of bioinformatics tools have been developed. We decided to utilize WGCNA, which has been successfully applied to both RNA-seq and DNAm datasets, allowing identification and correlation of individual gene modules with clinical parameters.40 Results are highly encouraging because we discovered a number of gene expression modules that correlated strongly with the number of relapses and the requirement for treatment with biologics. Interestingly, overlapping modules derived from applying WGCNA to RNAseq data with those derived from DNAm data revealed a major overlap, suggesting prognostic expression signatures maybe at least in part underpinned by epigenetic changes.

ly with the number of relapses and the requirement for treatment with biologics. Interestingly, overlapping modules derived from applying WGCNA to RNAseq data with those derived from DNAm data revealed a major overlap, suggesting prognostic expression signatures maybe at least in part underpinned by epigenetic changes. As a limitation to our study, we acknowledge that the total number of patients included is relatively low and hence some of the analyses performed, particularly those that correlate signatures with clinical outcome, should be considered as preliminary. However, to the best of our knowledge this is the first and largest study applying a multi-omics profiling approach to a unique sample collection of highly purified intestinal epithelium. The fact that all patients were recruited at diagnosis (treatment-naïve) also represents an important strength. Last but not least, although our study was performed on a pediatric patient cohort, we consider our findings to be equally relevant to adult-onset IBD given the similarities in disease phenotype (particularly in teenage onset) and common concepts of disease pathogenesis. In summary, our study is the first to apply a multi-omics profiling approach to a large collection of purified intestinal epithelial samples. The findings clearly demonstrate disease-specific abnormalities in epithelial cell function in children with IBD. We also highlight how specific data signatures might be indicative of disease status and behavior and therefore have a potential to be of clinical relevance in the future.

ied intestinal epithelial samples. The findings clearly demonstrate disease-specific abnormalities in epithelial cell function in children with IBD. We also highlight how specific data signatures might be indicative of disease status and behavior and therefore have a potential to be of clinical relevance in the future. Supplementary Material Supplementary Methods Supplementary Figures S1–S9 Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7 Supplementary Table 8 Supplementary Table 9 Supplementary Table 10 Supplementary Table 11 Supplementary Table 12 Supplementary Table 13 Supplementary Table 14 Supplementary Table 15

able 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7 Supplementary Table 8 Supplementary Table 9 Supplementary Table 10 Supplementary Table 11 Supplementary Table 12 Supplementary Table 13 Supplementary Table 14 Supplementary Table 15 Acknowledgments The authors would like to express their gratitude towards all patients and their parents that participated in this study. They would like to thank the clinical team of Paediatric Gastroenterology at Addenbrooke’s Hospital,Cambridge University Hospitals, with special emphasis on Dr Torrente and Dr Salvestrini for the collection of biopsies. Furthermore, they would like to thank Hans Clevers (Hubrecht Institute, NL) and Calvin Kuo (Stanford University, CA, USA) for providing cell lines for the production of conditioned medium for organoids. Lastly, they would like to thank the Wellcome Trust-MRC Stem Cell Institute Tissue Culture facility and the Cambridge NIHR BRC Cell Phenotyping Hub for technical assistance. Transcript profiling 16S Gut microbiota data can be found under EBI study ID PRJEB6663. RNAseq: E-MTAB-5464 and DNA methylation arrays: E-MTAB-5463. Conflicts of interest The authors disclose no conflicts.

Acknowledgments The authors would like to express their gratitude towards all patients and their parents that participated in this study. They would like to thank the clinical team of Paediatric Gastroenterology at Addenbrooke’s Hospital,Cambridge University Hospitals, with special emphasis on Dr Torrente and Dr Salvestrini for the collection of biopsies. Furthermore, they would like to thank Hans Clevers (Hubrecht Institute, NL) and Calvin Kuo (Stanford University, CA, USA) for providing cell lines for the production of conditioned medium for organoids. Lastly, they would like to thank the Wellcome Trust-MRC Stem Cell Institute Tissue Culture facility and the Cambridge NIHR BRC Cell Phenotyping Hub for technical assistance. Transcript profiling 16S Gut microbiota data can be found under EBI study ID PRJEB6663. RNAseq: E-MTAB-5464 and DNA methylation arrays: E-MTAB-5463. Conflicts of interest The authors disclose no conflicts. Funding This work was supported by funding from the following organizations: Crohn’s in Childhood Research Association (CICRA), the Evelyn Trust, Crohn’s and Colitis in Childhood (“3Cs”), Addenbrookes Charitable Trust (ACT), and Crohn’s and Colitis UK (CCUK; M16-5). K.H. was funded by an EBPOD EMBL-EBI/Cambridge Biomedical Research Centre Postdoctoral Fellowship. P. R. was supported by the Deutsche Forschungsgemeinschaft EXC306, BMBF DEEP IHEC TP5.2, EMED SysINFLAME CP4 and EU SYSCID DLV-733100. Author names in bold designate shared co-first authorship.

Funding This work was supported by funding from the following organizations: Crohn’s in Childhood Research Association (CICRA), the Evelyn Trust, Crohn’s and Colitis in Childhood (“3Cs”), Addenbrookes Charitable Trust (ACT), and Crohn’s and Colitis UK (CCUK; M16-5). K.H. was funded by an EBPOD EMBL-EBI/Cambridge Biomedical Research Centre Postdoctoral Fellowship. P. R. was supported by the Deutsche Forschungsgemeinschaft EXC306, BMBF DEEP IHEC TP5.2, EMED SysINFLAME CP4 and EU SYSCID DLV-733100. Author names in bold designate shared co-first authorship. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/j.gastro.2017.10.007.

The inflammatory bowel diseases (IBDs) Crohn’s disease (CD) and ulcerative colitis (UC) are an important health problem, with an incidence among European adults of 12.7 and 24.3 per 100,000 person-years, respectively, and a prevalence of 0.5% to 1.0%.1 Moreover, the incidence of IBD is increasing among adults and children and in the developed and developing world.1–5 In the United Kingdom, IBDs cost the National Health Service approximately £720 million (approximately $1.1 billion) per annum.6,7 The pathogenesis of IBD is believed to involve an aberrant immune response to intestinal microbiota in genetically susceptible individuals.8 Genetic studies have provided many candidate loci in the past decade, and the innate and acquired immune responses have been implicated in pathogenesis. However, identified genetic factors account for only a modest proportion of the disease variance: 13.6% for CD and 7.5% for UC.9 These figures highlight the need for critical evaluation of genetic discoveries to date and indicate the importance of environmental factors in the pathogenesis of IBD; in addition, the intriguing possibility arises that epigenetics may partially account for the “hidden heritability” in IBD. We review recent genetic discoveries in IBD and introduce readers to epigenetic factors that could be involved in pathogenesis.

The pathogenesis of IBD is believed to involve an aberrant immune response to intestinal microbiota in genetically susceptible individuals.8 Genetic studies have provided many candidate loci in the past decade, and the innate and acquired immune responses have been implicated in pathogenesis. However, identified genetic factors account for only a modest proportion of the disease variance: 13.6% for CD and 7.5% for UC.9 These figures highlight the need for critical evaluation of genetic discoveries to date and indicate the importance of environmental factors in the pathogenesis of IBD; in addition, the intriguing possibility arises that epigenetics may partially account for the “hidden heritability” in IBD. We review recent genetic discoveries in IBD and introduce readers to epigenetic factors that could be involved in pathogenesis. Genetics in IBD: A Generation of Progress In the past 25 years, there has been intense interest in identifying genetic and more recently epigenetic changes that relate to the pathogenesis of IBD. Few other complex diseases have been the subject of such extensive genetic and epigenetic research.

We review recent genetic discoveries in IBD and introduce readers to epigenetic factors that could be involved in pathogenesis. Genetics in IBD: A Generation of Progress In the past 25 years, there has been intense interest in identifying genetic and more recently epigenetic changes that relate to the pathogenesis of IBD. Few other complex diseases have been the subject of such extensive genetic and epigenetic research. National consortia and subsequently large international collaborative research groups, such as the International IBD Genetics Consortium, have led the way in performing large-scale appraisals of the genome of patients with IBD (http://www.ibdgenetics.org/). The assumption-free approach of genome-wide association studies (GWAS) has helped to support established etiologic roles of the innate and acquired immune system in IBD and identified interesting new mechanisms such as autophagy.10

large-scale appraisals of the genome of patients with IBD (http://www.ibdgenetics.org/). The assumption-free approach of genome-wide association studies (GWAS) has helped to support established etiologic roles of the innate and acquired immune system in IBD and identified interesting new mechanisms such as autophagy.10 Findings from the past 25 years of genetic discovery in IBD have been put into context by the latest meta-analysis of the GWAS and ImmunoChip data.9 The ImmunoChip (Illumina, Inc, San Diego, CA) was developed following GWAS of IBD and other immune diseases; it contains 200,000 single nucleotide polymorphisms (SNPs) relevant to IBD and other immune-mediated diseases. The aims of the ImmunoChip experiments are to replicate and fine map the known IBD susceptibility loci and to identify common links with other immune disorders. The meta-analysis comprised more than 75,000 cases and controls and more than 1.23 million SNPs from several centers worldwide. It identified a further 64 loci, bringing the total number of IBD-associated loci to 163; this is significantly more than for any other complex disease.9

See editorial on page 2045; CME quiz on page 1999. It is widely recognized that knowledge regarding the genetic basis of inflammatory bowel disease (IBD) and other complex diseases will provide key insights into pathogenic mechanisms. It is this fact that has spurred efforts to identify disease susceptibility genes. Of the many complex diseases investigated using molecular genetic techniques, Crohn’s disease (CD) is exceptional in that specific genetic variants unequivocally associated with disease susceptibility have been successfully identified.1,2 Nonetheless, characterization of the unknown number of remaining CD genes is required to complete the picture and remains a priority. CD is one of the 2 common and related forms of IBD, the other being ulcerative colitis (UC). Within the United Kingdom, they have a combined prevalence of approximately 4/1000.3 Both are known to have a significant genetic contribution to their etiology, but this is stronger for CD than UC.4 The epidemiologic evidence also suggests that CD and UC share some susceptibility genes. In 2001, fine mapping of a widely replicated linkage region on chromosome 16 led to the identification of CARD15 as a major CD susceptibility gene, with mutations leading to dysregulation of innate immune pathways.1,2 CARD15 genes have subsequently been shown in meta-analysis to predominantly determine susceptibility to ileal CD. Variants within a number of other genes have been associated with CD, UC, or both,5–9 although their exact roles in IBD susceptibility require clarification and, in some cases, replication.

une pathways.1,2 CARD15 genes have subsequently been shown in meta-analysis to predominantly determine susceptibility to ileal CD. Variants within a number of other genes have been associated with CD, UC, or both,5–9 although their exact roles in IBD susceptibility require clarification and, in some cases, replication. To date, pinpointing of disease genes has depended on detailed evaluation of candidates implicated by their function or patterns of expression or by fine mapping within large regions identified in the course of genome-wide linkage scans. Across the range of common diseases, productivity of such approaches has been limited. Most complex disease genetic studies, including many in IBD, have been beset by poor reproducibility of results and slow progress in identifying disease genes. This has been attributed to a range of factors, some of the most important being the low resolution of sib-pair linkage analysis, use of inappropriate statistical thresholds for significance, and poor matching of controls due to population admixture.10 One powerful new method for the identification of complex disease genes is genome-wide association scanning, genotyping large panels of affected individuals and appropriately matched population controls for hundreds of thousands of polymorphic markers across the genome and using appropriately stringent statistical thresholds for significance.11 Within the past year, such studies have become technically and financially possible using sets of markers that capture most of the common variation across the genome using knowledge regarding human haplotype structure available from the International HapMap Project (http://www.hapmap.org).12 Systematic whole-genome association studies, in comparison with the previous gold standard of linkage analysis, should provide substantially increased power and resolution for detection of complex disease susceptibility genes.13

lotype structure available from the International HapMap Project (http://www.hapmap.org).12 Systematic whole-genome association studies, in comparison with the previous gold standard of linkage analysis, should provide substantially increased power and resolution for detection of complex disease susceptibility genes.13 Recently, the results of a 308,332-marker genome scan in a North American panel of 547 non-Jewish case patients with CD and 548 controls were reported. Case patients were selected as having ileal CD to reduce heterogeneity.14 Three markers showed a highly significant association with CD, 2 of which were in CARD15. The third marker was a rare coding variant rs11209026c (1142G→A; Arg381Gln) found in the interleukin 23 receptor (IL23R) gene on chromosome 1 (P = 5.05 × 10−9). Nine other markers showed association with P < .0001 either within IL23R or in the intergenic area with the adjacent IL12RB2 gene. Internal replication was achieved in the index study using both a Jewish CD case-control cohort (peak P value, 3.36 × 10−13) and family-based methodologies, the latter in addition suggesting association with UC in a small non-Jewish cohort. This finding indicates that IL23R may have a general role in the etiology of IBD.14

rnal replication was achieved in the index study using both a Jewish CD case-control cohort (peak P value, 3.36 × 10−13) and family-based methodologies, the latter in addition suggesting association with UC in a small non-Jewish cohort. This finding indicates that IL23R may have a general role in the etiology of IBD.14 The aims of the current study were to seek replication of the association between IL23R and IBD in a large independent North European cohort representing the full range of CD and UC phenotypes, examine in detail genotype-phenotype relationships, explore evidence for epistasis with the known CD susceptibility gene CARD15, and provide accurate estimates of disease risk for associated variants. Replication of the association in an independent cohort would serve 2 important purposes. First, it is key to confirming the veracity of the original finding and the applicability of these findings in populations outside North America. Further, strong independent replication of the key finding of one of the first published genome-wide association scans would provide proof of principle that this novel methodology can be used to identify risk variants for complex diseases. Subjects and Methods Subjects A total of 2877 individuals with IBD (1902 with CD and 975 with UC) were recruited in 5 centers across England and Scotland. The study was approved by the research ethics committees at each center.

The aims of the current study were to seek replication of the association between IL23R and IBD in a large independent North European cohort representing the full range of CD and UC phenotypes, examine in detail genotype-phenotype relationships, explore evidence for epistasis with the known CD susceptibility gene CARD15, and provide accurate estimates of disease risk for associated variants. Replication of the association in an independent cohort would serve 2 important purposes. First, it is key to confirming the veracity of the original finding and the applicability of these findings in populations outside North America. Further, strong independent replication of the key finding of one of the first published genome-wide association scans would provide proof of principle that this novel methodology can be used to identify risk variants for complex diseases. Subjects and Methods Subjects A total of 2877 individuals with IBD (1902 with CD and 975 with UC) were recruited in 5 centers across England and Scotland. The study was approved by the research ethics committees at each center. Standard clinical, radiologic, and histologic diagnostic criteria were applied.15 Phenotypic details were obtained by retrospective case notes review. CD phenotype was classified by age at diagnosis, location, and behavior of disease. Only one member of multiply affected families was included. A total of 1.75% were of Jewish origin, and 2.25% were nonwhite. Demographic and subphenotype data are presented in Table 1.

s were obtained by retrospective case notes review. CD phenotype was classified by age at diagnosis, location, and behavior of disease. Only one member of multiply affected families was included. A total of 1.75% were of Jewish origin, and 2.25% were nonwhite. Demographic and subphenotype data are presented in Table 1. Control allele frequencies were obtained from 1345 individuals recruited across Britain as part of the 1958 British birth cohort.16 Cases and controls were categorized into 12 broad geographical regions within Great Britain to minimize confounding due to variation in allele frequencies across the country.17 Genotyping Genotyping of cases was undertaken with iPLEX chemistry on a matrix-assisted laser desorption/ionization time-of-flight MassARRAY platform (Sequenom, San Diego, CA). Cases were genotyped for 8 IL23R markers reported in the index study, including the nonsynonymous single nucleotide polymorphism (SNP) rs11209026 encoding amino acid change Arg381Gln (primer sequences in Supplementary Table 1; see supplemental material online at www.gastrojournal.org). Two of the North American markers (rs7517847, rs2201841) were omitted due to their location within a sequence of interspersed low-complexity repeats.

polymorphism (SNP) rs11209026 encoding amino acid change Arg381Gln (primer sequences in Supplementary Table 1; see supplemental material online at www.gastrojournal.org). Two of the North American markers (rs7517847, rs2201841) were omitted due to their location within a sequence of interspersed low-complexity repeats. Genotyping of controls was undertaken at the Wellcome Trust Sanger Institute using the Illumina 550K chip (Illumina, San Diego, CA). Concordance of genotype calls between the different platforms was confirmed by genotyping 87 control DNAs for all 8 markers using the MassARRAY platform with strong concordance of calls between technologies—98.99% for the 8 markers overall. There was 100% concordance for 3 markers, including the coding variant Arg381Gln (Supplementary Table 2; see supplemental material online at www.gastrojournal.org). The data for 1594 cases of CD genotyped for CARD15 mutations in earlier studies were used to undertake analysis for evidence of interaction between CARD15 and IL23R.18–21

concordance for 3 markers, including the coding variant Arg381Gln (Supplementary Table 2; see supplemental material online at www.gastrojournal.org). The data for 1594 cases of CD genotyped for CARD15 mutations in earlier studies were used to undertake analysis for evidence of interaction between CARD15 and IL23R.18–21 Statistical Methods Allele frequencies were compared between cases and controls and between phenotypic subgroups using χ2 tests of 2 × 2 tables. Odds ratios were calculated for the minor allele at each SNP; confidence intervals (CIs) were calculated using Woolf’s method.22 Pairwise SNP linkage disequilibrium coefficients were estimated using Haploview.23 Conditional association analysis was implemented using COCAPHASE, a module of the UNPHASED program.24 This method tests for equality of odds ratios for haplotypes identical at conditioning loci. The Mantel–Haenszel test for association conditioning on geographical region was implemented using PLINK (http://pngu.mgh.harvard.edu/∼purcell/plink/). Median age at disease diagnosis between groups was compared using the Wilcoxon rank sum test. Age at diagnosis was dichotomized according to the Montreal classification.25 Unless specified otherwise, all analyses were performed using R version 2.2 for Windows (http://www.R-project.org).

gh.harvard.edu/∼purcell/plink/). Median age at disease diagnosis between groups was compared using the Wilcoxon rank sum test. Age at diagnosis was dichotomized according to the Montreal classification.25 Unless specified otherwise, all analyses were performed using R version 2.2 for Windows (http://www.R-project.org). Results All genotypes were in Hardy–Weinberg equilibrium in both cases and controls (P > .05). A highly significant association with CD was observed across the region (Table 2). The strongest association was observed at the nonsynonymous SNP Arg381Gln, where the frequency of the A allele was 2.5% in CD compared with 6.2% in controls (P = 1.1 ×10−12). The odds ratio for this protective allele was 0.38 (95% CI, 0.29–0.50). Alternatively, the common wild-type homozygous GG genotype can be considered as the risk genotype with an odds ratio of 2.70. To minimize potential confounding from regional differences in allele frequencies, a Mantel–Haenszel test was performed across 12 regional strata. Mantel–Haenszel odds ratios was very similar to those obtained from pooled data for all SNPs. For example, the Mantel–Haenszel odds ratio was 0.36 (95% CI, 0.25–0.51) for Arg381Gln.

ize potential confounding from regional differences in allele frequencies, a Mantel–Haenszel test was performed across 12 regional strata. Mantel–Haenszel odds ratios was very similar to those obtained from pooled data for all SNPs. For example, the Mantel–Haenszel odds ratio was 0.36 (95% CI, 0.25–0.51) for Arg381Gln. Several SNPs also showed significant association with UC (Table 2). The strongest signal was observed with common SNPs rs1004819 (P = .0071) and rs10889677 (P = .0042). The frequency of Arg381Gln was only marginally different between cases and controls (UC, 0.046; controls, 0.062; P = .029), with an odds ratio of 0.73 (95% CI, 0.55–0.96). The nonsynonymous SNP Arg381Gln was in tight linkage disequilibrium with one other SNP (rs11465804, r2 = 0.85) but weak linkage disequilibrium with all 6 other SNPs (r2 = 0.03–0.1). A separate test for CD association was performed for each SNP conditioning on Arg381Gln by conditional regression modeling. This showed a significant association at all SNPs (P < .001) except rs11465804, with the strongest residual association detected at rs10889677 (P = 4.6 × 10−8). Hence, the nonsynonymous SNP does not account for all the association signal at this locus.

h SNP conditioning on Arg381Gln by conditional regression modeling. This showed a significant association at all SNPs (P < .001) except rs11465804, with the strongest residual association detected at rs10889677 (P = 4.6 × 10−8). Hence, the nonsynonymous SNP does not account for all the association signal at this locus. Data were then analyzed for evidence of significant genotype-phenotype correlations based on age at onset of CD, disease location, and disease behavior (Table 3). No significant subgroup association was observed. In particular, the subgroup of subjects with CD affecting the colon only without small bowel disease (n = 539) appeared to be as strongly associated as those with exclusively ileal/small bowel involvement (n = 668) (minor allele frequencies, 2.3% and 2.0%, respectively). The age at disease onset ranged from 12 to 67 years in patients with CD who carried the A allele of Arg381Gln and from 0 to 80 years in wild-type GG cases. There was no difference in the median age of onset between these 2 groups (AA/AG: median, 28 years [n = 85]; GG: median, 26 years [n = 1650]; P = .26). Stratification of cases by age at diagnosis according to the Montreal classification25 revealed similar genotype frequencies in all groups (Table 3). For UC, subgroup analysis by disease extent, smoking history, and sex also revealed no significant subgroup association. Age at onset of UC ranged from 14 to 79 years in cases who carried the A allele of Arg381Gln and from 2 to 81 years in wild-type GG cases, with no difference in the median age of onset between the 2 groups (AA/AG: median, 34 years [n = 72]; GG: median, 33 years [n = 708]; P = .14) (Table 4). A total of 1540 subjects with CD were fully genotyped for the 3 CARD15 mutations (G908R, L1007fs, R702W) (Table 3). The frequency of Arg381Gln in 460 cases carrying at least one CARD15 mutation (2.2%) was not significantly different from that in 1081 cases who carried none (2.7%; P = .47). None of the 3 cases who were homozygous for the rare A allele also carried a CARD15 mutation.

CARD15 mutations (G908R, L1007fs, R702W) (Table 3). The frequency of Arg381Gln in 460 cases carrying at least one CARD15 mutation (2.2%) was not significantly different from that in 1081 cases who carried none (2.7%; P = .47). None of the 3 cases who were homozygous for the rare A allele also carried a CARD15 mutation. Discussion This study provides unequivocal confirmation of association between variants in the IL23R gene and IBD, suggesting a major effect on overall susceptibility to CD and a more modest effect on UC. Importantly, this study also shows the association at IL23R for the first time in a non-American population. The strength of this association at IL23R and the fact that it reaches such a magnitude in 2 independent data sets leaves no doubt that it is a true finding. In addition, this is one of the first instances of highly significant, independent replication of data derived from a genome-wide association scan and provides important validation of this technique as a hypothesis-free method for the identification of complex disease genes.

s leaves no doubt that it is a true finding. In addition, this is one of the first instances of highly significant, independent replication of data derived from a genome-wide association scan and provides important validation of this technique as a hypothesis-free method for the identification of complex disease genes. As with the North American genome-wide scan, the strongest evidence for association was seen at the nonsynonymous SNP Arg381Gln, where the frequency of the A allele was 2.5% in CD compared with 6.2% in controls (P = 1.1 × 10−12). These allele frequencies are similar to those seen in the North American panel.14 There was no evidence that IL23R variants associate with any particular subphenotype of CD based on disease behavior or location. Hence, there was no difference in minor allele frequency even between the extremes of pure ileal/small bowel CD and pure colonic CD (2.7% and 2.3%, respectively). Likewise, analysis based on disease behavior did not show any specific subgroup associations (Table 3). This negative result is interesting because it contrasts with the other confirmed CD susceptibility locus CARD15, which seems to have definite associations with ileal disease.26 These findings are extended by the observation of association with UC overall but not with any known UC subphenotype group, suggesting that IL23R variants may exert a rather generic effect on chronic intestinal inflammation, although the effect size in UC does appear to be smaller than in CD. It is noteworthy that the odds ratio confidence interval at Arg381Gln for UC (0.73 [95% CI, 0.55–0.96]) does not overlap with that for CD (0.38 [95% CI, 0.29–0.50]), suggesting a significantly less marked protective effect of the rare allele for UC compared with CD.

fect size in UC does appear to be smaller than in CD. It is noteworthy that the odds ratio confidence interval at Arg381Gln for UC (0.73 [95% CI, 0.55–0.96]) does not overlap with that for CD (0.38 [95% CI, 0.29–0.50]), suggesting a significantly less marked protective effect of the rare allele for UC compared with CD. Based on data from our large, independent panel of CD cases, it is possible to provide an accurate estimate of the size of the effect conferred by IL23R variants with regard to the risk of CD. We estimated an odds ratio of 0.38 (95% CI, 0.29–0.49) for Arg381Gln. This is likely to be a more accurate estimate than that provided in the index report from the North American study (odds ratio, 0.26; 95% CI, 0.15–0.43) due to the well-recognized bias of the so-called “winner’s curse,” which leads to overestimation of effect size in discovery panels.27 Characterizing the exact effect size is important to permit sample size calculation for any further attempts at replication. Where the effect size is overestimated, there is a risk that apparently appropriately powered studies will fail to observe the effect and erroneously conclude that it is a false positive.

nels.27 Characterizing the exact effect size is important to permit sample size calculation for any further attempts at replication. Where the effect size is overestimated, there is a risk that apparently appropriately powered studies will fail to observe the effect and erroneously conclude that it is a false positive. In some previous reports of genetic association, effect sizes have been quantitated as a population-attributable risk in addition to the odds ratio. This figure is intended to estimate the proportion of disease incidence attributable to a specific variant. However, it cannot be calculated for a protective minor allele as in the case of Arg381Gln. Further, while it is possible to think of the effect at Arg381Gln as an increased risk conferred by the common G allele, calculations of population-attributable risk based on this assumption lead to an implausibly high figure due to the very high carriage rate of the G allele in the control population.

se of Arg381Gln. Further, while it is possible to think of the effect at Arg381Gln as an increased risk conferred by the common G allele, calculations of population-attributable risk based on this assumption lead to an implausibly high figure due to the very high carriage rate of the G allele in the control population. One important question is whether the nonsynonymous variant accounts for all of the association signal at IL23R. This was tested by conditional regression modeling, looking for evidence of association while controlling for the effect at Arg381Gln. From this analysis, it is clear that there is a strong residual signal, maximal at rs10889677 (P = 4.6 × 10−8), and hence that variation at loci in addition to Arg381Gln, either within or adjacent to IL23R, exert an influence on IBD susceptibility. Whether this reflects a functional impact of the noncoding variants themselves or the fact that they are in linkage disequilibrium with other functionally significant or coding variants is yet to be established. Analysis within Haploview (http://www.broad.mit.edu/mpg/haploview/) of data available from the International HapMap Project28 (http://www.hapmap.org) shows that this selection of 8 tag SNPs captures only 18 out of 83 informative SNPs, within IL23R and the 3′ intergenic region covered by these markers, genotyped in the CEPH (Utah residents with ancestry from northern and western Europe) panel at r2 > 0.8. Any additional coding variation is likely to be rare because interrogation of Ensembl database release 41 (October 2006) (http://www.ensembl.org/) revealed the presence of only 2 additional nonsynonymous coding variants (rs1884444 and rs7530511) in IL23R with minor allele frequency >1% in healthy European populations, both of which were investigated in the North American study with neither showing evidence of association. However, it is known that different splice isoforms of IL23R exist and it is possible that their expression is determined by some of the documented noncoding variation.29 To clarify these issues, future studies will need to include resequencing of IL23R in a CD panel and fine mapping across the gene using markers identified as a result, as well as studies to assess the potential functional impact of variants identified.

sion is determined by some of the documented noncoding variation.29 To clarify these issues, future studies will need to include resequencing of IL23R in a CD panel and fine mapping across the gene using markers identified as a result, as well as studies to assess the potential functional impact of variants identified. The data with regard to Arg381Gln provide evidence of a very common variant being a disease risk allele, or conversely protection from CD being conferred by the rare allele. The explanations are likely to be complex but for immune-mediated conditions may include the fact that genetic variation at a particular locus confers a spectrum of risk, being protective against some diseases, such as infections, while increasing the risk of others, such as autoimmunity or inflammatory conditions. These variations will have been subject to differing selection pressures in diverse populations as a result of different environmental exposures. It is also noteworthy that for some of the markers showing evidence of association, it is the rarer allele that is associated with increased risk of CD. This further supports the argument for more than one variant in IL23R with different effects on gene function.

ations as a result of different environmental exposures. It is also noteworthy that for some of the markers showing evidence of association, it is the rarer allele that is associated with increased risk of CD. This further supports the argument for more than one variant in IL23R with different effects on gene function. Recent studies have identified IL-23, the cognate ligand of IL23R, as a key player in both innate and adaptive immune systems. Most IL-23 is secreted by activated dendritic cells, monocytes, and macrophages following their exposure to pathogen-derived molecules that bind at toll-like receptors.30 IL-23 stimulates a unique CD4+ helper T-cell population characterized by the production of IL-17, tumor necrosis factor, and IL-6 and known as Th17 cells. These cells play a central role in driving autoimmune inflammation in a number of animal models. IL-17 stimulates monocytes and endothelial cells to produce proinflammatory mediators, which in turn promote rapid neutrophil recruitment.30 The effect of IL-23 has recently been distinguished from that of the related heterodimer IL-12, with which it shares a common p40 subunit.31 Importantly in this regard, 2 studies in knockout mice lacking the p19 subunit of IL-23 showed marked attenuation of T cell–mediated colitis, while knockout of the p35 subunit of IL-12 produces no such attenuation, suggesting that IL-23 but not IL-12 is essential for the development of colitis.32,33 The identification of different roles for IL-12 and IL-23 in control of immune pathways together with the current genetic data suggest that targeting IL-23 (and components of its downstream effector pathway) may be a useful and specific strategy to inhibit IBD while sparing systemic host protective immunity.34

is.32,33 The identification of different roles for IL-12 and IL-23 in control of immune pathways together with the current genetic data suggest that targeting IL-23 (and components of its downstream effector pathway) may be a useful and specific strategy to inhibit IBD while sparing systemic host protective immunity.34 As well as focusing attention on the IL-23 pathway in the pathogenesis of IBD, the current study also provides key validation of genome-wide association scanning as a means of identifying complex disease susceptibility genes. The North American study group applied an appropriate, genome-wide significance level, and use of such a stringent threshold has immediately led to replication in our independent panel with a level of significance that makes the association indisputable.

g as a means of identifying complex disease susceptibility genes. The North American study group applied an appropriate, genome-wide significance level, and use of such a stringent threshold has immediately led to replication in our independent panel with a level of significance that makes the association indisputable. To date, complex disease genetic studies have been beset by poor study design, particularly use of nonconservative thresholds for significance, resulting in publication of many unreplicated false-positive results across the spectrum of common disease, hence the importance of the current study in providing unequivocal early replication in an independent panel of the principal findings from one of the first reported genome-wide association scans. There are recent reports in another complex disease (age-related macular degeneration) that also provide grounds for optimism that this technique produces replicable genetic association data.35–37 The clear message is that genome-wide association scanning works and that this study design, which is being so vigorously applied across a number of common diseases, is likely to be highly productive. The hope is that with use of appropriately stringent statistical thresholds and appropriately powered data sets, the success seen here in CD will be generally applicable without the plethora of false-positive results that have vexed the field of complex disease genetics to date. Supplementary Data Supplementary Table 1 Primer Sequences for all 8 SNPs Using iPlex Chemistry on a MassARRAY Platform

To date, complex disease genetic studies have been beset by poor study design, particularly use of nonconservative thresholds for significance, resulting in publication of many unreplicated false-positive results across the spectrum of common disease, hence the importance of the current study in providing unequivocal early replication in an independent panel of the principal findings from one of the first reported genome-wide association scans. There are recent reports in another complex disease (age-related macular degeneration) that also provide grounds for optimism that this technique produces replicable genetic association data.35–37 The clear message is that genome-wide association scanning works and that this study design, which is being so vigorously applied across a number of common diseases, is likely to be highly productive. The hope is that with use of appropriately stringent statistical thresholds and appropriately powered data sets, the success seen here in CD will be generally applicable without the plethora of false-positive results that have vexed the field of complex disease genetics to date. Supplementary Data Supplementary Table 1 Primer Sequences for all 8 SNPs Using iPlex Chemistry on a MassARRAY Platform Supplementary Table 2 Results of Genotyping to Confirm Concordance Between Illumina and MassARRAY Technologies

To date, complex disease genetic studies have been beset by poor study design, particularly use of nonconservative thresholds for significance, resulting in publication of many unreplicated false-positive results across the spectrum of common disease, hence the importance of the current study in providing unequivocal early replication in an independent panel of the principal findings from one of the first reported genome-wide association scans. There are recent reports in another complex disease (age-related macular degeneration) that also provide grounds for optimism that this technique produces replicable genetic association data.35–37 The clear message is that genome-wide association scanning works and that this study design, which is being so vigorously applied across a number of common diseases, is likely to be highly productive. The hope is that with use of appropriately stringent statistical thresholds and appropriately powered data sets, the success seen here in CD will be generally applicable without the plethora of false-positive results that have vexed the field of complex disease genetics to date. Supplementary Data Supplementary Table 1 Primer Sequences for all 8 SNPs Using iPlex Chemistry on a MassARRAY Platform Supplementary Table 2 Results of Genotyping to Confirm Concordance Between Illumina and MassARRAY Technologies The authors acknowledge use of DNA from the 1958 British Birth Cohort collection (R. Jones, S. Ring, W. McArdle, and M. Pembrey), funded by Medical Research Council grant G0000934 and Wellcome Trust grant 068545/Z/02, the National Association for Colitis and Crohn’s Disease and the Wellcome Trust for supporting the case DNA collections, and the Wellcome Trust Case Control Consortium, for which the Crohn’s disease panel was originally assembled.

, funded by Medical Research Council grant G0000934 and Wellcome Trust grant 068545/Z/02, the National Association for Colitis and Crohn’s Disease and the Wellcome Trust for supporting the case DNA collections, and the Wellcome Trust Case Control Consortium, for which the Crohn’s disease panel was originally assembled. The authors have no conflicts of interest to declare. M.T., F.C., and S.A.F. contributed equally to this work. The authors thank all the subjects who contributed samples, as well as the consultants and nursing staff across the United Kingdom who helped with recruitment of study subjects: C. Todhunter, A. Sutherland, K. Mohiuddin, N. Thompson, M. Hudson, J. Barbour, P. Donaldson, S. J. Middleton, J. Woodward, J. Hunter, R. S. Harvey, J. H. Saunders, A. Douds, D. Sharpstone, S. Whalley, A. Nicolson, S. M. Greenfield, P. B. McIntyre, M. J. Carter, I. Barrison, H. J. Kennedy, I. W. Fellows, R. Tighe, M. G. Phillips, C. Jamieson, I. Beales, A. Hart, A. Prior, J. Wyke, S. Williams, Y. Miao, M. Ninkovic, M. Dronfield, P. Nair, R. Dickinson, P. Roberts, C. P. Willoughby, I. Dunkley, D. Morris, M. Twist, N. Fisher, D. Kelf A. Nightingale, C. W. Lees, G. T. Ho, I. D. Arnott, T. Ahmad, D. McGovern, J. Beckly, R. Cooney, L. Hancock, A. Geramia, S. Goldthorpe, and S. Patham. Appendix Supplementary data associated with this article can be found, in the online version, at doi:10.1053/j.gastro.2007.02.051. Table 1 Demographic Details of 2877 Individuals With IBD Used in Case-Control Panel

The authors thank all the subjects who contributed samples, as well as the consultants and nursing staff across the United Kingdom who helped with recruitment of study subjects: C. Todhunter, A. Sutherland, K. Mohiuddin, N. Thompson, M. Hudson, J. Barbour, P. Donaldson, S. J. Middleton, J. Woodward, J. Hunter, R. S. Harvey, J. H. Saunders, A. Douds, D. Sharpstone, S. Whalley, A. Nicolson, S. M. Greenfield, P. B. McIntyre, M. J. Carter, I. Barrison, H. J. Kennedy, I. W. Fellows, R. Tighe, M. G. Phillips, C. Jamieson, I. Beales, A. Hart, A. Prior, J. Wyke, S. Williams, Y. Miao, M. Ninkovic, M. Dronfield, P. Nair, R. Dickinson, P. Roberts, C. P. Willoughby, I. Dunkley, D. Morris, M. Twist, N. Fisher, D. Kelf A. Nightingale, C. W. Lees, G. T. Ho, I. D. Arnott, T. Ahmad, D. McGovern, J. Beckly, R. Cooney, L. Hancock, A. Geramia, S. Goldthorpe, and S. Patham. Appendix Supplementary data associated with this article can be found, in the online version, at doi:10.1053/j.gastro.2007.02.051. Table 1 Demographic Details of 2877 Individuals With IBD Used in Case-Control Panel CD (n = 1902) UC (n = 975) Median age at diagnosis (y) 26 38.9 Gender (F/M) 1153/745 480/495 Smoking at diagnosis (%) Never 58.4 55.0 Ex 9.4 30.3 Current 32.2 14.7 Jewish ancestry (%) 1.75 1.9 Nonwhite (%) 2.25 3.25 Surgery (%) 61.8 Location/extent (%) 32.7 ileal 16.5 rectum only 31.8 colonic 35.0 distal to 35.5 ileocolonic  splenic flexure 27.1 perianal 48.5 proximal to splenic flexure Behavior (%) 36.5 stenosing 17.15 penetrating Table 2 Case-Control Allele Frequencies and Disease Odds Ratios (95% Confidence Intervals) for CD and UC

ery (%) 61.8 Location/extent (%) 32.7 ileal 16.5 rectum only 31.8 colonic 35.0 distal to 35.5 ileocolonic  splenic flexure 27.1 perianal 48.5 proximal to splenic flexure Behavior (%) 36.5 stenosing 17.15 penetrating Table 2 Case-Control Allele Frequencies and Disease Odds Ratios (95% Confidence Intervals) for CD and UC SNP Allele Controls CD P Odds ratio (95% CI) UC P Odds ratio (95% CI) rs1004819 T 0.307 0.383 1.1 × 10−8 1.41 (1.23–1.56) 0.348 .00713 1.20 (1.05–1.37) rs10489629 G 0.448 0.372 1.8 × 10−8 0.73 (0.66–0.82) 0.43 .26 0.93 (0.82–1.05) rs11465804 G 0.058 0.025 7.2 × 10−11 0.41 (0.31–0.53) 0.046 .081 0.77 (0.58–1.02) rs11209026 A 0.062 0.025 1.1 × 10−12 0.38 (0.29–0.50) 0.046 .0291 0.73 (0.55–0.96) rs1343151 T 0.332 0.266 1.1 × 10−7 0.73 (0.65–0.82) 0.315 .26 0.92 (0.81–1.06) rs10889677 A 0.315 0.398 3.4 × 10−10 1.45 (1.28–1.61) 0.358 .0042 1.22 (1.07–1.39) rs11209032 A 0.320 0.390 1.3 × 10−7 1.35 (1.22–1.52) 0.3524 .032 1.16 (1.01–1.32) rs1495965 G 0.447 0.517 3.4 × 10−7 1.32 (1.19–1.47) 0.457 .57 1.04 (0.92–1.18) Table 3 Arg381Gln Genotype and Allele Frequencies in CD Cases Stratified by Known Phenotypic Subgroups and CARD15 Status

.28–1.61) 0.358 .0042 1.22 (1.07–1.39) rs11209032 A 0.320 0.390 1.3 × 10−7 1.35 (1.22–1.52) 0.3524 .032 1.16 (1.01–1.32) rs1495965 G 0.447 0.517 3.4 × 10−7 1.32 (1.19–1.47) 0.457 .57 1.04 (0.92–1.18) Table 3 Arg381Gln Genotype and Allele Frequencies in CD Cases Stratified by Known Phenotypic Subgroups and CARD15 Status AA AG GG Total Freq(A) Sex Male 1 36 690 727 0.026 Female 2 49 1068 1119 0.024 Smoking history No 1 28 729 758 0.020 Yes 1 21 414 436 0.026 Ex 0 1 127 128 0.004 Disease location Pure colorectal disease 1 23 515 539 0.023 Pure ileal disease 0 31 533 564 0.027 Ileocolonic disease 1 25 642 668 0.020 Any colorectal disease 2 46 1101 1149 0.022 Any ileal disease 1 54 1115 1170 0.024 Perianal disease Yes 2 20 454 476 0.025 No 1 59 1205 1265 0.024 Disease behavior Stenosing 1 32 604 637 0.027 Penetrating 2 15 248 265 0.036 Inflammatory only 0 30 714 744 0.020 Surgery Yes 2 51 1035 1088 0.025 No 1 31 645 677 0.024 Age at diagnosis (y) 16 or younger 1 8 197 206 0.024 17–40 2 60 1129 1191 0.027 Older than 40 0 14 324 338 0.021 CARD15 statusa −/ − 3 52 1025 1080 0.027 −/+ 0 15 342 357 0.021 +/+ 0 5 98 103 0.024 a Samples are subdivided by CARD15 status into those homozygous wild-type (−/ −), those heterozygous for CD-associated variants (−/+), and those homozygous or compound heterozygous for CD-associated variants (+/+). Table 4 Arg381Gln Genotype and Allele Frequencies in UC Cases Stratified by Known Phenotypic Subgroups

AA AG GG Total Freq(A) Sex Male 1 36 690 727 0.026 Female 2 49 1068 1119 0.024 Smoking history No 1 28 729 758 0.020 Yes 1 21 414 436 0.026 Ex 0 1 127 128 0.004 Disease location Pure colorectal disease 1 23 515 539 0.023 Pure ileal disease 0 31 533 564 0.027 Ileocolonic disease 1 25 642 668 0.020 Any colorectal disease 2 46 1101 1149 0.022 Any ileal disease 1 54 1115 1170 0.024 Perianal disease Yes 2 20 454 476 0.025 No 1 59 1205 1265 0.024 Disease behavior Stenosing 1 32 604 637 0.027 Penetrating 2 15 248 265 0.036 Inflammatory only 0 30 714 744 0.020 Surgery Yes 2 51 1035 1088 0.025 No 1 31 645 677 0.024 Age at diagnosis (y) 16 or younger 1 8 197 206 0.024 17–40 2 60 1129 1191 0.027 Older than 40 0 14 324 338 0.021 CARD15 statusa −/ − 3 52 1025 1080 0.027 −/+ 0 15 342 357 0.021 +/+ 0 5 98 103 0.024 a Samples are subdivided by CARD15 status into those homozygous wild-type (−/ −), those heterozygous for CD-associated variants (−/+), and those homozygous or compound heterozygous for CD-associated variants (+/+). Table 4 Arg381Gln Genotype and Allele Frequencies in UC Cases Stratified by Known Phenotypic Subgroups AA AG GG Total Freq(A) Sex Male 1 41 447 489 0.044 Female 2 43 436 481 0.049 Smoking history No 0 28 301 329 0.043 Yes 0 7 81 88 0.040 Ex 1 20 160 181 0.061 Disease extent Rectum only 1 12 134 147 0.048 Distal to splenic flexure 1 24 286 311 0.042 Proximal to splenic flexure 1 43 387 431 0.052 Age at diagnosis (y) 16 or younger 0 2 39 41 0.025 17–40 1 41 443 485 0.044 Older than 40 0 28 228 256 0.055

Dear Sir: In their recent article, Castellanos-Rubio et al1 use genome-wide expression profiling combined with published data on genomic regions showing modest linkage to celiac disease as a strategy to identify inherited genetic variants predisposing to disease. They investigated 361 identified variants in a small case-control collection (262 cases, 214 controls) of Spanish origin. After quality control procedures 330 single nucleotide polymorphisms (SNPs) were available for association testing: 10 SNPs from 6 different genes/regions were stated to show evidence of association with celiac disease (uncorrected P < .005), with 4 SNPs remaining significant after statistics were corrected for multiple testing. They suggested these preliminary findings should be validated and tested in different populations.1

ting: 10 SNPs from 6 different genes/regions were stated to show evidence of association with celiac disease (uncorrected P < .005), with 4 SNPs remaining significant after statistics were corrected for multiple testing. They suggested these preliminary findings should be validated and tested in different populations.1 We recently carried out a genome-wide association study of 310,605 SNPs in 778 celiac cases and 1422 population controls from the British population.2 Such studies have recently been shown to be highly effective to identify risk variants for common disease.3 We previously directly genotyped rs365836 in our UK celiac/control genome-wide study,2 and have now imputed data for the other 9 SNP markers analyzed by Castellanos-Rubio et al1 using algorithms implemented in PLINK v1.0.0.4 We found no evidence (Table 1, uncorrected P>.05) in the UK collection for 9 of the 10 associations reported in the Spanish collection. Although rs6747096 was borderline significant (uncorrected P = .016) in the UK collection, the effect is in the opposite direction (rs6747096 G allele is more frequent in UK cases versus controls, yet less frequent in Spanish cases versus controls).1

tion for 9 of the 10 associations reported in the Spanish collection. Although rs6747096 was borderline significant (uncorrected P = .016) in the UK collection, the effect is in the opposite direction (rs6747096 G allele is more frequent in UK cases versus controls, yet less frequent in Spanish cases versus controls).1 There are a number of possible reasons for these apparent discrepancies. First, the results of the study by Castellanos-Rubio et al1 could be due to type 1 statistical error arising through the multiple hypothesis testing of the 330 SNPs investigated. The authors did correctly apply a Bonferroni correction. Second, UK data may represent a type 2 error. However, the UK collection is of much larger sample size, and power calculations suggest adequate power (>80% at P < .05 for all markers, using a multiplicative model and assuming celiac disease population prevalence of 1%) to detect effects of the allele frequencies (as observed in UK controls) and odds ratios as reported by Castellanos-Rubio et al.1 A third possible explanation is the presence of genetic heterogeneity between the British and Spanish populations. However, in our genetic investigations of celiac disease to date across Irish, British, and Dutch populations (∼8000 samples genotyped for >1000 SNPs) we have not yet observed any evidence for heterogeneity at 8 non-HLA celiac disease associated regions.5 Broadly similar disease prevalence and clinical features are seen across European populations suggesting (with the possible exception of HLA-DQ) that genetic heterogeneity is at most a minor issue.

d for >1000 SNPs) we have not yet observed any evidence for heterogeneity at 8 non-HLA celiac disease associated regions.5 Broadly similar disease prevalence and clinical features are seen across European populations suggesting (with the possible exception of HLA-DQ) that genetic heterogeneity is at most a minor issue. Multiple risk variants for common human diseases have recently been identified by genome-wide association studies.6 Most of these findings do not map to regions previously identified by genetic linkage studies. In celiac disease, consistent findings from linkage studies have not been obtained. We feel the strategy pursued by Castellanos-Rubio et al1 (prioritizing linkage regions for association studies) is suboptimal. As Castellanos-Rubio et al discuss,1 there are often discrepancies between the findings of candidate gene association studies. We highlight the importance of carrying out large, well-designed, genome-wide studies with findings replicated in at least 1 other population. (Table 1) Support and funding provided by Coeliac UK and The Wellcome Trust. Table 1 Analysis of 10 Celiac Disease-Associated SNPs Reported Castellanos-Rubio et al1 in a UK Collection

As Castellanos-Rubio et al discuss,1 there are often discrepancies between the findings of candidate gene association studies. We highlight the importance of carrying out large, well-designed, genome-wide studies with findings replicated in at least 1 other population. (Table 1) Support and funding provided by Coeliac UK and The Wellcome Trust. Table 1 Analysis of 10 Celiac Disease-Associated SNPs Reported Castellanos-Rubio et al1 in a UK Collection SNP Spanish (Castellanos-Rubio et al1) UK genome wide association study Allele 1/2 MAF cases MAF controls P value Method Allele minor/major MAF cases MAF controls P value rs12619019 G/C 0.18 0.11 .0015 Imputation INFO = 0.90 G/C 0.173 0.162 .33 rs6747096 G/A 0.16 0.29 2.38 × 10−5 Imputation INFO = 0.97 G/A 0.211 0.182 .016 rs11954744 A/T 0.13 0.21 .0016 Imputation INFO = 0.93 A/T 0.179 0.164 .19 rs6887645 A/G 0.13 0.21 .0022 Imputation INFO = 0.93 A/G 0.180 0.164 .17 rs365836 G/A 0.24 0.33 .0016 Illumina Hap300 G/A 0.256 0.267 .40 rs1048251 T/G 0.53 0.40 3.02 × 10−4 Imputation INFO = 1.03 T/G 0.464 0.444 .21 rs7019234 G/A 0.53 0.40 3.08 × 10−4 Imputation INFO = 1.03 G/A 0.464 0.444 .21 rs459311 T/G 0.52 0.39 1.38 × 10−4 Imputation INFO = 0.99 T/G 0.434 0.434 .21 rs458046 T/A 0.52 0.39 1.35 × 10−4 Imputation INFO = 0.99 T/A 0.435 0.435 .21 rs7040561 T/A 0.15 0.06 6.55 × 10−5 Imputation INFO = 0.77 T/A 0.157 0.141 .12 MAF, minor allele frequency.

0 3.08 × 10−4 Imputation INFO = 1.03 G/A 0.464 0.444 .21 rs459311 T/G 0.52 0.39 1.38 × 10−4 Imputation INFO = 0.99 T/G 0.434 0.434 .21 rs458046 T/A 0.52 0.39 1.35 × 10−4 Imputation INFO = 0.99 T/A 0.435 0.435 .21 rs7040561 T/A 0.15 0.06 6.55 × 10−5 Imputation INFO = 0.77 T/A 0.157 0.141 .12 MAF, minor allele frequency. SNP imputation was performed using celiac UKGWAS data2 merged with 2,244,775 high-quality SNPs from 60 CEU founders from the HapMap project (http://pngu.mgh.harvard.edu/∼purcell/plink/pimputation.shtml). Quality (INFO, information) scores for all 9 imputed markers were high.

mon links with other immune disorders. The meta-analysis comprised more than 75,000 cases and controls and more than 1.23 million SNPs from several centers worldwide. It identified a further 64 loci, bringing the total number of IBD-associated loci to 163; this is significantly more than for any other complex disease.9 The Genetic Architecture of IBD Early GWAS identified IBD loci common and unique to CD and UC.11 The latest data show the increasing proportion of loci common to both diseases, with relatively fewer CD- or UC-specific loci.9 Of the 163 identified loci, 110 are associated with both diseases, 30 are CD specific, and 23 are UC specific.9 Studies of gene loci shared by UC and CD may provide insight into their common pathogenic mechanisms. The T-helper (Th)17 and interleukin (IL)-12/IL-23 pathway is well established in the pathogenesis of IBD, with susceptibility gene loci IL23R, IL12B, JAK2, and STAT3 identified in both UC and CD.12,13 Variants in IL12B, which encodes the p40 subunit of IL-12 and IL-23, have been associated with IBD and other immune disorders. Defects in the function of IL-10, an immunosuppressive cytokine, have also been associated with CD and UC.14,15 A severe, childhood-onset, CD-like form of enterocolitis is associated with rare mutations in IL10R. However, this disorder could be a separate entity from idiopathic IBD.16,17 Other susceptibility genes that regulate immune function include CARD9, IL1R2, REL, SMAD3, and PRDM1.12 Interestingly, the well-established CD risk variants of NOD2 and PTPN22 appear to protect against UC.

sociated with rare mutations in IL10R. However, this disorder could be a separate entity from idiopathic IBD.16,17 Other susceptibility genes that regulate immune function include CARD9, IL1R2, REL, SMAD3, and PRDM1.12 Interestingly, the well-established CD risk variants of NOD2 and PTPN22 appear to protect against UC. CD-Specific Susceptibility Loci and Pathways CD has a greater genetic component than that of UC, and several CD-specific susceptibility loci have been delineated. The latest genetic data increasingly highlight the relationship between the host innate immune system and the intestinal microbiota in CD. GWAS have indicated that intracellular bacterial processing by autophagy is an important pathogenic mechanism. Importantly, the association between CD and NOD2 has been consistently replicated at the genome-wide significance level18; NOD2 has been mechanistically linked with autophagy.19,20 Cigarette smoking, a strong environmental factor in risk of CD, might affect NOD2 function.21 Furthermore, the product of the CD susceptibility gene ATG16L1 is recruited to the plasma membrane by NOD2, where it initiates bacterial internalization by autophagosomes.15,18,20

en mechanistically linked with autophagy.19,20 Cigarette smoking, a strong environmental factor in risk of CD, might affect NOD2 function.21 Furthermore, the product of the CD susceptibility gene ATG16L1 is recruited to the plasma membrane by NOD2, where it initiates bacterial internalization by autophagosomes.15,18,20 Another gene involved in autophagy-induced bacterial killing is IRGM. CD-associated polymorphisms in IRGM lead to reduced protein expression. A different SNP of IRGM protects against Mycobacterium tuberculosis.22,23 The most recent data from ImmunoChip studies indicated an overlap between IBD loci and complex mycobacterial disease loci9; 7 CD susceptibility genes overlap with leprosy susceptibility genes, and 6 mycobacterium susceptibility genes overlap with IBD loci. However, for several of these diseases, the genetic associations have opposite effects.9 Genes involved in the host response to mycobacteria that were previously associated with CD include CARD9 and LTA.11,15 Other CD-specific loci identified related to the immune system include PTPN22, IL2RA, IL27, TNFSF11, and VAMP3.15,18

er, for several of these diseases, the genetic associations have opposite effects.9 Genes involved in the host response to mycobacteria that were previously associated with CD include CARD9 and LTA.11,15 Other CD-specific loci identified related to the immune system include PTPN22, IL2RA, IL27, TNFSF11, and VAMP3.15,18 UC-Specific Susceptibility Loci and Pathways Although UC susceptibility loci have primarily included genes that regulate intestinal epithelial barrier function, there is recent evidence that HLA variants are involved in the development of UC.9 HLA-DQA1 was the locus most strongly associated with UC (odds ratio, 1.44),with no corresponding increased risk in CD.9 The HLA class II genes are tremendously diverse and control antigen presentation to T cells; they have been implicated in other immune diseases. Hepatocyte nuclear factor 4A (HNF4A) regulates expression of cell junction proteins in the intestinal epithelial barrier; variants have been associated with UC and with colorectal cancer, a complication of chronic inflammation in patients with IBD.24 Rare SNPs at the HNF4A gene locus, not implicated in UC, are associated with maturity-onset diabetes, inherited in an autosomal dominant fashion.25 Other UC-associated genes that affect epithelial barrier function include CHD1, which encodes E-cadherin, and LAMB1, which encodes the lamina β subunit 1. Many UC risk alleles encode cytokines and inflammatory mediators, including tumor necrosis factor (TNF) receptor superfamily members (TNFRSF14, TNFRSF9), ILs, and IL receptors (IL1R2, IL8Ra/RB, IL7R).12

epithelial barrier function include CHD1, which encodes E-cadherin, and LAMB1, which encodes the lamina β subunit 1. Many UC risk alleles encode cytokines and inflammatory mediators, including tumor necrosis factor (TNF) receptor superfamily members (TNFRSF14, TNFRSF9), ILs, and IL receptors (IL1R2, IL8Ra/RB, IL7R).12 Relationships With Other Diseases Jostins et al reported that 70% of IBD loci overlap with loci associated with other complex immune diseases, such as IL23R variants associated with psoriasis and ankylosing spondylitis.26–29 However, these polymorphisms sometimes have opposite effects in different diseases. For example, a variant of PTPN22 protects against CD but is a risk factor for type 1 diabetes and rheumatoid arthritis.30 Extraintestinal manifestations of IBD also share common loci, which may explain their co-occurrence. For example, variants of REL, IL2, and CARD9 are associated with UC and primary sclerosing cholangitis.11,31

, a variant of PTPN22 protects against CD but is a risk factor for type 1 diabetes and rheumatoid arthritis.30 Extraintestinal manifestations of IBD also share common loci, which may explain their co-occurrence. For example, variants of REL, IL2, and CARD9 are associated with UC and primary sclerosing cholangitis.11,31 Current Agenda for Genetic Studies of IBD Many of the IBD loci identified so far have not been accurately characterized or fine mapped, and the candidate genes commonly used to describe them are only putative. Moreover, the biological functions of their products, and their complex interactions, in most cases require delineation. Studies are under way to use the greater detail afforded by the ImmunoChip data to fine map loci, and functional studies are needed. Further work is required to determine how specific variants affect levels of messenger RNA (mRNA) and consequently protein, which could provide further insight into mechanisms of pathogenesis. This is likely to take considerable time; NOD2 was identified more than 10 years ago, and there is still uncertainty about its function.29

k is required to determine how specific variants affect levels of messenger RNA (mRNA) and consequently protein, which could provide further insight into mechanisms of pathogenesis. This is likely to take considerable time; NOD2 was identified more than 10 years ago, and there is still uncertainty about its function.29 GWAS have excelled in identifying moderate-risk genetic variants with at least 5% prevalence in the population. Novel approaches are needed to discover lower-prevalence variants with higher effect size. Whole-exome sequencing, which covers only coding areas of the genome, costs less than whole-genome sequencing and tends to afford higher-depth coverage and therefore greater certainty about novel discoveries. It has been successfully used to identify single mutations in very early-onset IBD32 and is perhaps most likely to produce results in individuals with a strong family history or early age of disease onset. However, many polymorphisms that affect disease susceptibility are located in noncoding areas of the genome; the ENCODE project has highlighted the importance of noncoding regions in disease risk.33 Exome sequencing and whole-genome sequencing are each under way, with large-scale endeavors at the Sanger Centre likely to report results in mid-2013 (http://www.ibdresearch.co.uk/).

y are located in noncoding areas of the genome; the ENCODE project has highlighted the importance of noncoding regions in disease risk.33 Exome sequencing and whole-genome sequencing are each under way, with large-scale endeavors at the Sanger Centre likely to report results in mid-2013 (http://www.ibdresearch.co.uk/). Other researchers are looking for associations between specific genotypes and disease phenotypes. To date, NOD2 mutations have been associated with stricturing ileal CD, and DRB1*0103 has been associated with severe extensive UC.34–36 The International IBD Genetics Consortium recently presented preliminary data from the core phenotyping project, which used ImmunoChip data to identify genetic factors that correspond to disease phenotypes. The International IBD Genetics Consortium confirmed the associations of variants of NOD2 and HLA with CD and UC, respectively, as well as associated HLA variants with location of CD and reported the effects of NOD2 and HLA genotypes on age of disease onset.37 Most IBD genetic analyses have been performed in the white populations of Northern Europe and America. More recently, there has been a push to expand this work to other ethnic populations.38 IBD-associated variants of NOD2, for example, are less prevalent in black populations, and CD-associated mutations have not been detected in Asian or Sub-Saharan African populations.39,40

populations of Northern Europe and America. More recently, there has been a push to expand this work to other ethnic populations.38 IBD-associated variants of NOD2, for example, are less prevalent in black populations, and CD-associated mutations have not been detected in Asian or Sub-Saharan African populations.39,40 Pharmacogenomics, the study of how genomic factors affect the efficacy, tolerability, and side effects of a therapeutic agent, remains high on the research agenda. Patients are evaluated for thiopurine S-methyltransferase (encoded by TPMT) genotype and phenotyping before initiation of thiopurine therapy is recommended by the US Food and Drug Administration.41,42 There are ongoing attempts to predict patients’ response to other agents based on genetic factors. Studies supported by the Serious Adverse Events Consortium aim to predict mesalamine-induced nephrotoxicity (http://www.ibdresearch.co.uk/5asa/) and serious complications of anti-TNF therapies and thiopurines (http://www.ibdresearch.co.uk/pred4/). A main goal of IBD research is to develop disease-specific therapeutics. Many researchers are developing reagents to alter activities of genes and pathways identified through GWAS, and the IL-12/IL-23 signaling pathway is one promising target. Ustekinumab, a monoclonal antibody that binds to the shared p40 subunit encoded by IL12B, has undergone phase 2b induction and maintenance trials in patients with CD.43 Apilimod mesylate, briakinumab (ABT-874), and SCH-900222, also target components of the IL-12/IL-23 signaling pathway, and are currently under evaluation.44

inumab, a monoclonal antibody that binds to the shared p40 subunit encoded by IL12B, has undergone phase 2b induction and maintenance trials in patients with CD.43 Apilimod mesylate, briakinumab (ABT-874), and SCH-900222, also target components of the IL-12/IL-23 signaling pathway, and are currently under evaluation.44 From the Environment to Genetics via Epigenetics The challenge remains to measure patients’ duration, intensity, and frequency of exposure to the many environmental factors that potentially could contribute to IBD, making the environmental impact on disease difficult to disentangle.45 Epigenetic factors could mediate gene-environment interactions involved in pathogenesis. Epigenetic programming begins at fertilization and continues throughout life. Studies of Agouti mice and the offspring of post–World War II Dutch famine survivors revealed how the environment can affect epigenetic factors. Dietary intake during pregnancy was shown to affect the epigenetic reprogramming step in offspring during early development, an effect that persisted for up to 2 generations.46–50 Moreover, there is evidence for acquired epigenetic changes with aging caused by a range of environmental factors.51 Epigenetics could therefore play a central role in the pathogenesis of IBD and other diseases, affecting interactions among genetic and environmental factors such as the intestinal microbiome (Figure 1).

Moreover, there is evidence for acquired epigenetic changes with aging caused by a range of environmental factors.51 Epigenetics could therefore play a central role in the pathogenesis of IBD and other diseases, affecting interactions among genetic and environmental factors such as the intestinal microbiome (Figure 1). In IBD research, several key developments in molecular studies have led us from genetics to explore epigenetics. GWAS have identified key epigenetic regulatory enzymes such as DNA methyltransferase (DNMT) 3a and more recently DNMT3b as CD susceptibility genes.9,15 Dendritic cells that express CD-associated variants of NOD2 fail to up-regulate microRNA (miR) clusters that regulate Th1 and Th17 cell–mediated immune responses.52 Epigenetic mechanisms have also been shown to regulate the immune system. For example, differentiation of Th2 cells requires epigenetic silencing of the IFNG locus.53 What Is Epigenetics? Epigenetics may be defined as mitotically heritable changes in gene function not explained by changes in the DNA sequence. Gene expression can be altered by changes to the structure and function of chromatin (Figure 2). The main epigenetic mechanisms include DNA methylation, histone modification, RNA interference, and the positioning of nucleosomes (which will not covered in depth in this review).

explained by changes in the DNA sequence. Gene expression can be altered by changes to the structure and function of chromatin (Figure 2). The main epigenetic mechanisms include DNA methylation, histone modification, RNA interference, and the positioning of nucleosomes (which will not covered in depth in this review). The epigenome can be regarded as both stable and plastic. The epigenome can be regarded as stable as epigenetic marks are passed onto daughter cells during mitosis.54 However, stochastic and environmental factors can cause dynamic changes to the epigenome over time.55 During mitosis, the level of fidelity of epigenomic replication is much lower than that of the genetic sequence (error rate of 1 × 106 for the DNA sequence compared with 1 × 103 for DNA modifications), leading to an accumulation of epigenetic changes over time.45,56 Similarly, several environmental factors produce epimutations (epigenetic changes associated with disease); factors relevant to IBD include smoking, the microbiota, and diet. Epigenetic marks are reset during meiosis; the epigenome is established early in embryogenesis after undergoing several reprogramming steps, during which the epigenome is most subject to modification.54,57 Given that the epigenome is reset during meiosis, it was believed that epigenetic marks were not passed between generations.

marks are reset during meiosis; the epigenome is established early in embryogenesis after undergoing several reprogramming steps, during which the epigenome is most subject to modification.54,57 Given that the epigenome is reset during meiosis, it was believed that epigenetic marks were not passed between generations. Although environmental exposures in utero can lead to epigenetic changes that persist for up to 2 generations, there is increasing interest in true epigenetic inheritance, which lasts multiple generations.58–60 Transgenerational epigenetic inheritance has attracted both excitement and skepticism in the scientific community and pertains to epigenetic marks resistant to the major reprogramming steps.45,59 The most compelling evidence for transgenerational epigenetic inheritance comes from studies of plants; DNA methylation-mediated silencing of the Lcyc promoter causes variations in floral symmetry in Linaria vulgaris (toadflax) that are stably inherited over many generations.61 Although additional examples have been reported from studies of plants, insects, and mammals, the concept is still met with some reservation.62 Incomplete erasure of epigenetic mutations across generations could contribute to familial predisposition to diseases such as IBD.45

e stably inherited over many generations.61 Although additional examples have been reported from studies of plants, insects, and mammals, the concept is still met with some reservation.62 Incomplete erasure of epigenetic mutations across generations could contribute to familial predisposition to diseases such as IBD.45 DNA Methylation DNA methylation is the most widely studied epigenetic modification; in this process, a methyl group is covalently added to cytosines that are part of cytosine-guanine dinucleotides (CpG). Full methylation occurs when cytosine residues on both DNA strands are methylated. CpG dinucleotides are generally sparse in the genome (∼1%) but are relatively concentrated in specific regions called “CpG islands.” CpG islands are defined as a 200-base sequence containing greater than 50% CpG dinucleotides at an observed to statistically expected ratio of 0.6.63 The areas where most tissue-specific methylation appears to border CpG islands have been termed “CpG shores.”64 Transcriptionally repressive activity generally occurs where a gene has methylation of CpG islands in promoter areas and is an important mechanism of gene silencing.65 DNA methylation may lead to transcriptional repression by hindering access of transcription factors to promoter regions, although many researchers believe the reverse is true: that gene silencing subsequently leads to DNA methylation.66,67

CpG islands in promoter areas and is an important mechanism of gene silencing.65 DNA methylation may lead to transcriptional repression by hindering access of transcription factors to promoter regions, although many researchers believe the reverse is true: that gene silencing subsequently leads to DNA methylation.66,67 DNA is methylated by enzymes called the DNMTs. There are 5 members of the DNMT family: DNMT1 (maintenance of methylation), DNMT2 (involved in RNA methylation), DNMT3a, DNMT3b, and DNMT3L (involved in new methylation). There is evidence that these DNMTs interact and that other epigenetic mechanisms can recruit DNMTs to specific gene loci.63 DNA methylation passes from the mother to the daughter cells during mitosis; DNMT1 mediates full methylation of hemimethylated CpG sites during DNA replication. DNA methylation is part of normal genetic imprinting (hypermethylation of one parental allele leads to monoallelic expression) and inactivation of the X autosome in females. Inherited deficiency of DNMT3B leads to immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome, whereas complete lack of DNMT enzymes leads to embryonic lethality.65

mprinting (hypermethylation of one parental allele leads to monoallelic expression) and inactivation of the X autosome in females. Inherited deficiency of DNMT3B leads to immunodeficiency, centromeric instability, and facial anomalies (ICF) syndrome, whereas complete lack of DNMT enzymes leads to embryonic lethality.65 Histone Modification Histones can undergo a range of complex modifications; a complete discussion is beyond the scope of this review. Histones have N-terminal amino acid tails that protrude and can be modified by acetylation, methylation, ubiquitination, and phosphorylation.68 Different posttranscriptional modifications to histone ends are believed to recruit different coactivators or corepressors, which determines whether chromatin is in its relaxed or condensed form.69 Histone acetylation is the most well-described posttranslational modification and is regulated by the levels and activity of histone acetyl transferase (HAT) and histone deacetylase (HDAC).70 In a simple model, chromatin is transcriptionally active when lysines on histones H3 and H4 are acetylated. Although it is not exactly clear how acetylated histones affect transcription, they might change the structure of chromatin (acetylation of lysine neutralizes the positive electrostatic charge of the histone, facilitating the opening of chromatin) and thereby reveal binding sites for important coactivators.68 Overexpression or increased activity of HDACs can lead to hypoacetylation and gene silencing.

might change the structure of chromatin (acetylation of lysine neutralizes the positive electrostatic charge of the histone, facilitating the opening of chromatin) and thereby reveal binding sites for important coactivators.68 Overexpression or increased activity of HDACs can lead to hypoacetylation and gene silencing. There are 18 subtypes of HDAC in mammalian cells. Classes I and II are simple hydrolases, whereas class III HDACs require cofactors. HDAC enzymes can be inhibited by a range of natural (eg, lactate, butyrate) and synthetic compounds.71 RNA Interference miRs are single-stranded noncoding RNAs typically 22 nucleotides in length that are highly conserved throughout evolution.72 miRs with members of the Argonaut (Ago) family form the RNA interference-silencing complex. This complex regulates translation by binding to 3′ regions of untranslated mRNAs by directly inhibiting mRNA translation or by causing mRNA degradation (depending on the degree of complementarity between the miR and the mRNA target).73

h members of the Argonaut (Ago) family form the RNA interference-silencing complex. This complex regulates translation by binding to 3′ regions of untranslated mRNAs by directly inhibiting mRNA translation or by causing mRNA degradation (depending on the degree of complementarity between the miR and the mRNA target).73 After their description in the mid-1990s, a large number of miRs have been described (>1600 in humans, see http://mirbase.org). miRs are transcribed by RNA polymerase II into hairpin structures called pre-miR. Pre-miR is processed in the nucleus (by enzyme Drosha) and then the cytoplasm by the Dicer enzyme.72 After processing, each miR may show complementarity with many different mRNAs, and each mRNA may be targeted by many different miRs.74,75 miRs regulate gene expression and thereby numerous biological processes, including cell proliferation, differentiation, and death. Altered miR expression has been associated with many diseases.

ssing, each miR may show complementarity with many different mRNAs, and each mRNA may be targeted by many different miRs.74,75 miRs regulate gene expression and thereby numerous biological processes, including cell proliferation, differentiation, and death. Altered miR expression has been associated with many diseases. Potential Pitfalls A major hurdle in interpreting results from epigenetic studies is to determine causality, that is, whether a particular epigenetic profile is cause or consequence of disease. Furthermore, many of these studies provide only a snapshot of the epigenetic profile after the disease process has been established, rather than describing a temporal relationship between an epigenetic alteration and subsequent disease development. The epigenetic profile changes over time, and apparent associations may be a consequence of the disease itself or other environmental factors such as therapy.76,77 Conditional correlation models have attempted to determine how DNA methylation and genetic factors interact to cause diseases.78 Epigenetic marks are tissue and cell type specific, and therefore selection of a disease-relevant tissue type is crucial. In IBD research, it has been a major challenge to identify disease-relevant cell types. Currently, interest is focused on immune cells in the blood (such as CD4+ and CD8+ T cells) and the intestine (intraepithelial lymphocytes).

d cell type specific, and therefore selection of a disease-relevant tissue type is crucial. In IBD research, it has been a major challenge to identify disease-relevant cell types. Currently, interest is focused on immune cells in the blood (such as CD4+ and CD8+ T cells) and the intestine (intraepithelial lymphocytes). Results from studies of whole tissues, such as whole blood or colon biopsy samples, are difficult to interpret because of the heterogeneity among cell types, each with their own epigenetic signature. Early epigenetic studies of IBD were mostly performed with whole tissues, making results difficult to interpret; epigenetic features of individual cells can be masked by those of heterogeneous cell groups. Some researchers have used statistical methods to adjust for differing cell proportions.78,79 Studies are in progress to address this issue.

f IBD were mostly performed with whole tissues, making results difficult to interpret; epigenetic features of individual cells can be masked by those of heterogeneous cell groups. Some researchers have used statistical methods to adjust for differing cell proportions.78,79 Studies are in progress to address this issue. Epigenetics and IBD DNA Methylation Initial DNA methylation studies largely focused on the predisposition to cancer in IBD. DNA methylation changes in colonic epithelial cells that normally occur with aging are accelerated in IBD because of higher cell turnover in inflammation.80 Increased age-related DNA methylation, observed in colon cells of patients with colitis, could lead to genetic instability and development of cancer.80 Increased DNA methylation has been shown in dysplastic and the surrounding nondysplastic colon tissues from patients with UC compared with control subjects or patients with UC who do not have dysplasia.80 Four of 15 loci associated with development of cancer (CDH1, GDNF, HPP1, and MYOD1) were differentially methylated in surgical resection samples from patients with active UC compared with those with quiescent mucosa.81 CDH1 encodes the cell adhesion molecule E-cadherin, which is associated with IBD-associated cancer. Hypermethylation of the CDH1 promoter region has been shown in dysplastic and cancerous gut tissue of patients with UC compared with nondysplastic samples.24,82,83

h active UC compared with those with quiescent mucosa.81 CDH1 encodes the cell adhesion molecule E-cadherin, which is associated with IBD-associated cancer. Hypermethylation of the CDH1 promoter region has been shown in dysplastic and cancerous gut tissue of patients with UC compared with nondysplastic samples.24,82,83 The increasing interest in the role of DNA methylation in the pathogenesis of IBD has coincided with advances in platform-based DNA methylation array technologies, which have superseded candidate gene methylation profiling techniques. Initial IBD epigenome-wide methylation association studies (EWAS) used platform-based arrays to analyze peripheral blood samples (Table 1). Nimmo et al analyzed the methylation profile of peripheral blood from women and children with CD.79 Whole-blood DNA was analyzed using the Illumina 27K chip (Illumina, Inc). Fifty genes showed significantly different levels of methylation between patients with IBD and controls, including some involved in immune system activation (MAPK, RPIK3, and IL21R). Ontology analysis highlighted several pathways associated with IBD, including immune system processes, immune response, and host response to bacteria, whereas canonical pathway analysis indicated the involvement of Th17 cell pathways.79

rols, including some involved in immune system activation (MAPK, RPIK3, and IL21R). Ontology analysis highlighted several pathways associated with IBD, including immune system processes, immune response, and host response to bacteria, whereas canonical pathway analysis indicated the involvement of Th17 cell pathways.79 Another study showed the tissue-specific nature of epigenetic marks. No significant differences in DNA methylation were observed between children with IBD and controls based on methylation-specific amplification microarray analysis of peripheral blood. However, they found that peripheral blood mononuclear cells (PBMCs) from patients with IBD showed hypermethylation at the TEPP locus, which encodes testes, prostate, and placenta-expressed protein and is of uncertain relevance in IBD.84 A study that analyzed DNA methylation in Epstein–Barr virus—transformed B cells from 18 patients with IBD versus nonaffected siblings identified 49 differentially methylated CpG sites. More than half of the differentially methylated loci contained genes that regulate immune functions, including several (BCL3, STAT3, OSM, STAT5) involved in the IL-12 and IL-23 pathways.85

virus—transformed B cells from 18 patients with IBD versus nonaffected siblings identified 49 differentially methylated CpG sites. More than half of the differentially methylated loci contained genes that regulate immune functions, including several (BCL3, STAT3, OSM, STAT5) involved in the IL-12 and IL-23 pathways.85 DNA methylation has also been studied in colonic tissue (Table 1). An EWAS of intestinal biopsy samples from 20 monozygotic twins discordant for UC identified 61 differentially methylated loci, with several containing genes that regulate inflammation (CFI, SPINKK4, THY1/CD90). This study had an interesting design in that after the loci were identified in the analysis of discordant monozygotic twins (to exclude differences in genetic factors), they were validated in an independent cohort.86

ly methylated loci, with several containing genes that regulate inflammation (CFI, SPINKK4, THY1/CD90). This study had an interesting design in that after the loci were identified in the analysis of discordant monozygotic twins (to exclude differences in genetic factors), they were validated in an independent cohort.86 To overcome the heterogeneity of cell types in tissues, a methylation-wide profiling study of whole rectal biopsy specimens from patients with active and quiescent UC and CD was validated using isolated epithelial cells from rectal biopsy specimens.87 Many differentially methylated genes were identified in whole tissue, encoding proteins including DOK2 (involved in IL-4–mediated cell proliferation), Tap1 (a major histocompatibility complex class I transport molecule), and members of the TNF family (TNFSF4 and TNFSF12). ULK1 was methylated only in patients with CD; its product has a role in autophagy. Genes identified as being differentially methylated in this study, replicated findings from other EWAS,79 and have also been identified as susceptibility genes in GWAS,12,24 including CDH1, ICAM3, IL8RA, and CARD9. Histone Modification Histone modification is a complex process and the least studied epigenetic mechanism in IBD research. Although histone modifications have been shown to regulate genes that control inflammation,88 much of our understanding of histone modifications in the context of IBD has come from experimental and therapeutic trials of histone deacetylase inhibitors (HDACi).

and the least studied epigenetic mechanism in IBD research. Although histone modifications have been shown to regulate genes that control inflammation,88 much of our understanding of histone modifications in the context of IBD has come from experimental and therapeutic trials of histone deacetylase inhibitors (HDACi). Patterns of histone acetylation in colon tissues from rats with colitis, induced by administration of dextran sulfate sodium and 2,4-trinitrobenzene sulfonic acid, and biopsy specimens from patients with CD have been described. Inflamed tissue and Peyer’s patches from rats with colitis and patients were found to have increased acetylation of H4 (at lysines residues 8 and 12).89

ith colitis, induced by administration of dextran sulfate sodium and 2,4-trinitrobenzene sulfonic acid, and biopsy specimens from patients with CD have been described. Inflamed tissue and Peyer’s patches from rats with colitis and patients were found to have increased acetylation of H4 (at lysines residues 8 and 12).89 Several mechanisms have been proposed to link histone modification with inflammation, involving the innate immune response to microbiota. Butyrate, an endogenous metabolite formed during fermentation of dietary fibers by the intestinal microbiota, is an HDAC inhibitor. Butyrate increases expression of NOD2 by increasing histone acetylation in its promoter region.90 Histone acetylation has a role in production of intestinal alkaline phosphatase, an endogenous protein responsible for detoxification of bacterial lipopolysaccharide. Sodium butyrate increased intestinal alkaline phosphatase production via its ability to inhibit HDAC.91 Toll-like receptor 4 regulates intestinal homeostasis by preventing excessive inflammatory responses to commensal bacteria and could be regulated by histone deacetylation.92 Expression of a gingival antimicrobial protein, β-defensin 2, is also increased by histone acetylation.93

ction via its ability to inhibit HDAC.91 Toll-like receptor 4 regulates intestinal homeostasis by preventing excessive inflammatory responses to commensal bacteria and could be regulated by histone deacetylation.92 Expression of a gingival antimicrobial protein, β-defensin 2, is also increased by histone acetylation.93 HDAC inhibitors can be classified according to structural class. These classes include carboxylates (sodium butyrate and valproate), hydroxamic acids (trichostatin A, SAHA, KBH-A42, and ITF2357), benzamides (entinostat or MS-275 and mocetinostat or MGCD-0103), and cyclic peptides (α-apicidin and depsipeptides).94 HDAC inhibitors have primarily been investigated in cancer research but also have anti-inflammatory effects.95 It is worth noting that the enzymes that affect histone acetylation status (HAT, HDAC) do not act exclusively on histones but affect acetylation of a range of proteins, including p53, STAT3, and HIF1α.96 Therefore, HDAC inhibitors act not only through epigenetic mechanisms but also on multiple histone-independent targets, including the transcription factor nuclear factor κB (NF-κB) pathway, cytoskeletal proteins, and cell cycle and apoptosis regulators.97,98

on of a range of proteins, including p53, STAT3, and HIF1α.96 Therefore, HDAC inhibitors act not only through epigenetic mechanisms but also on multiple histone-independent targets, including the transcription factor nuclear factor κB (NF-κB) pathway, cytoskeletal proteins, and cell cycle and apoptosis regulators.97,98 Several anti-inflammatory mechanisms of HDACi have been proposed. The expression of T-regulatory cells, which mediate immune tolerance and abrogate excessive inflammation, is linked to Foxp3 gene expression. Administration of HDACi in mice leads to increased T-regulatory cell differentiation and suppression of bowel inflammation, potentially as a result of acetylation of lysines in the forkhead domain of Foxp3.99 Another mechanism could involve their effects on acetylation of proteins in the NF-κB pathway.96,100 In models of murine colitis, the HDACi ITF2357 increases acetylation of histone 3 and reduces activation of NF-κB.101 HDAC2 is associated with corticosteroid responsiveness, affecting NF-κB regulation of gene expression.102

her mechanism could involve their effects on acetylation of proteins in the NF-κB pathway.96,100 In models of murine colitis, the HDACi ITF2357 increases acetylation of histone 3 and reduces activation of NF-κB.101 HDAC2 is associated with corticosteroid responsiveness, affecting NF-κB regulation of gene expression.102 Butyrate enemas have been used to treat patients with colitis, although inhibition of HDAC may not be their predominant mechanism of action. Butyrate has several effects on the gastrointestinal tract, including maintenance of barrier function and a homeostatic reduction in epithelial cell production of IL-8.103–105 Butyrate reduces the disease activity index of patients as well as nuclear translocation of NF-κB in lamina propria macrophages.106 Other HDACi have also been shown to ameliorate dextran sulfate sodium–induced colitis in mice.101,104 RNA Interference Studies in animals have shown that intestinal miRs regulate gut homeostasis. Mice deficient in intestinal Dicer1, an miR-processing enzyme, have disorganized intestinal epithelial crypts with increased goblet cells, rapid jejunal epithelial migration, and accelerated apoptosis. Additionally, mice deficient in intestinal Dicer1 have increased inflammation and neutrophil and lymphocyte migration, and reduced epithelial barrier function, compared with mice not deficient in Dicer1.107

nal epithelial crypts with increased goblet cells, rapid jejunal epithelial migration, and accelerated apoptosis. Additionally, mice deficient in intestinal Dicer1 have increased inflammation and neutrophil and lymphocyte migration, and reduced epithelial barrier function, compared with mice not deficient in Dicer1.107 A number of studies have investigated differences in miRs between patients with and without IBD (Table 2). Changes in miRs in human IBD were first described in 2008.108 In sigmoid biopsy specimens from patients with active UC, levels of 8 miRs were significantly increased and 3 were decreased compared with samples from patients without UC. miR-192, normally expressed in colonic epithelial cells, was significantly reduced in tissues of patients with active UC.108 miR-192 reduces expression of macrophage inhibitory peptide 2α, a CXC chemokine expressed by epithelial cells; its levels are increased in colon tissues of patients with UC.108 miR-150 is up-regulated in mice with dextran sulfate sodium–induced colitis in colon tissues from patients with UC; its levels correlate inversely with those of its target c-Myb, which has a role in apoptosis.109 Up-regulation of miR-21, which promotes inflammation, has been reported in several studies of patients with active UC and CD colitis (but not ileitis), along with miR-155.110–112 miR-196 is overexpressed in the inflamed epithelium of patients with CD and may reduce IRGM-mediated autophagy.113

as a role in apoptosis.109 Up-regulation of miR-21, which promotes inflammation, has been reported in several studies of patients with active UC and CD colitis (but not ileitis), along with miR-155.110–112 miR-196 is overexpressed in the inflamed epithelium of patients with CD and may reduce IRGM-mediated autophagy.113 Distinct miR signatures have been identified in peripheral blood samples from patients with IBD compared with controls and in patients with CD compared with those with UC.114 Several miRs have been found to be significantly up-regulated or down-regulated in 2 or more studies, including miRs-16, -21, -28-5p, -149, -151-5p, -199-a, and -532-3p.114–116 Eleven miRs were also found to be differentially expressed between serum samples from pediatric patients with CD and healthy children.115 Further adequately powered studies are required to identify IBD-associated miR profiles in intestinal tissues and serum, plasma, and separated blood cells. Specific miR profiles might be able to predict IBD susceptibility, progression, and response to therapy. Moreover, identifying the targets of these miRs will provide additional insight into the pathogenesis of IBD.

identify IBD-associated miR profiles in intestinal tissues and serum, plasma, and separated blood cells. Specific miR profiles might be able to predict IBD susceptibility, progression, and response to therapy. Moreover, identifying the targets of these miRs will provide additional insight into the pathogenesis of IBD. Interaction Between Genetics and Epigenetics in Complex Disease An intriguing field of investigation is the relationship between genetic and epigenetic factors. There is evidence of colocalization of differentially methylated CpGs at predisposing SNPs identified at GWAS. In our own EWAS of CD, we showed enrichment of methylation changes within 50 kilobases from GWAS-identified susceptibility loci, including IL-19, IL-27, TNF, and NOD2.79 In a recent large methylation study of patients with rheumatoid arthritis, in 5 of 9 MHC genes, a specific genotype was associated with differential methylation.78 This phenomena has also been observed in studies of patients with type 2 diabetes mellitus, where a specific allele, rs8050136, within the obesity and diabetes susceptibility gene FTO is associated with increased DNA methylation.117 However, Toperoff et al associated a different SNP, rs1121980, with hypomethylation of FTO.118

enomena has also been observed in studies of patients with type 2 diabetes mellitus, where a specific allele, rs8050136, within the obesity and diabetes susceptibility gene FTO is associated with increased DNA methylation.117 However, Toperoff et al associated a different SNP, rs1121980, with hypomethylation of FTO.118 It is not clear how these SNPs affect methylation of the gene. They could increase the numbers of CpG dinucleotides or alter the access of the methylation machinery to the gene.117 Allele- or haplotype-specific methylation occurs more commonly with cis-acting polymorphisms.119 A potential cofounder of the Illumina BeadArray, used in most DNA methylation studies, is that certain probes contain SNPs or repetitive elements that can affect methylation analysis.120 Variants in STAT4 have also been reported to alter its methylation (Figure 3A). Additional evidence of haplotype-specific methylation has been shown in the promoter regions of IL8RA and IL8RB. Rectal biopsy specimens from patients with IBD were shown to have increased methylation of the CpG island closest to the transcriptional start site of IL8RA: the proposed binding site of transcription factor PU.1 (SPI-1).87 The risk allele rs11676348 alters a CpG, is located between IL8RA and IL8RB coding sequences, and contains a binding site for the transcription factor STAT3.87 Although not specifically probed itself by the Illumina 27K, differential methylation was observed on either side of rs11676348.87

iption factor PU.1 (SPI-1).87 The risk allele rs11676348 alters a CpG, is located between IL8RA and IL8RB coding sequences, and contains a binding site for the transcription factor STAT3.87 Although not specifically probed itself by the Illumina 27K, differential methylation was observed on either side of rs11676348.87 Similarly, SNPs can affect the complementarity of miR binding. IBD-associated variants of IL23R, which have a role in IL-12 and IL-23 signaling, may show altered binding with miRs Let-7e and Let-7f, leading to altered expression of IL23R and inappropriate Th17 activation (Figure 3B). IRGM mediates innate immune defense against intracellular organisms, including Mycobacterium tuberculosis.23 Variants of IRGM alter the binding site for miR-196 (Figure 4B).113 Another study evaluated the risk of UC conferred by 3 common allelic variants of 3 pre-miRs (miR-146a, -196a, and -499). Three SNPs (rs11614913, rs2910164, and rs3746444) were genotyped in 170 patients with UC and 403 control patients. The AG heterozygous genotype of rs3746444, encoding miR-499, was significantly associated with an increased risk of UC (odds ratio, 1.51). The same genotype was also associated with older age of onset, left-sided colitis, hospitalization, and dependence on corticosteroids.121

patients with UC and 403 control patients. The AG heterozygous genotype of rs3746444, encoding miR-499, was significantly associated with an increased risk of UC (odds ratio, 1.51). The same genotype was also associated with older age of onset, left-sided colitis, hospitalization, and dependence on corticosteroids.121 Clinical Implications A number of potential clinical applications of epigenetics in diagnostics and therapeutics are receiving attention. The diagnostic applications of epigenetics include the use of biomarkers to confirm diagnosis, stratify disease course and response to chemotherapy, and predict development of cancer.122 Particularly pertinent for IBD, methylation changes in SFRP2, measured in fecal DNA samples, have been used to identify patients with colorectal cancer with approximately 75% sensitivity and specificity.123 Biomarkers have been found in a range of body fluids, including sputum, urine, and saliva for lung, bladder, and head and neck cancers, respectively.124–126 DNA methylation is a quantitative trait and therefore an attractive biomarker. A panel of relevant hypomethylated or hypermethylated CpGs might someday be used to distinguish between UC and CD, enable disease stratification, and predict treatment response.

lung, bladder, and head and neck cancers, respectively.124–126 DNA methylation is a quantitative trait and therefore an attractive biomarker. A panel of relevant hypomethylated or hypermethylated CpGs might someday be used to distinguish between UC and CD, enable disease stratification, and predict treatment response. The tissue-specific nature of epigenetic changes becomes especially relevant when considering their use as biomarkers. In oncology, tumor-specific aberrant DNA methylation might be used to identify patients with cancer.127 DNA methylation might also be useful in identifying patients with UC who are most likely to develop cancer. DNA methylation was increased in dysplastic and nondysplastic tissues from patients with colitis-associated cancer compared with those without cancer.81 More recently, changes in neoplasia-associated DNA methylation in SLIT2 and TMEFF2 were found in mucosal biopsy specimens and fecal DNA from patients with IBD who are at high risk for developing dysplasia or cancer.128 Disease-specific DNA methylation might be difficult to detect because of the heterogeneity of cell types in whole tissue samples and only become evident when isolated cell types are analyzed.84 The additional resources required to separate cells may limit the applicability of DNA methylation as a biomarker, although early studies of whole blood have produced promising results for biomarker development.

neity of cell types in whole tissue samples and only become evident when isolated cell types are analyzed.84 The additional resources required to separate cells may limit the applicability of DNA methylation as a biomarker, although early studies of whole blood have produced promising results for biomarker development. Likewise, miRs have been advocated as possible biomarkers. When analyzing more than 300 miRNAs present in colonic tissue biopsy specimens, distinct profiles were identified in IBD versus control samples (miR-26a, -29b, -126, -127-3p, -324-3p) and between quiescent UC and CD (miR-196b, 199a-3p, -199-5b, -150, -223).111 Distinct IBD profiles of miRs have also been described in peripheral blood, providing appealing, minimally invasive biomarkers. A miR panel from peripheral blood was able to distinguish between patients with UC and CD; levels of 10 miRs were increased and the level of one was decreased in patients with UC compared with CD.108 miRs derived from platelets, microvesicles, and PBMCs have also been measured in attempts to differentiate patients with and without UC. Seven miRs were differentially expressed in patients with UC compared with controls.129 The specificity of miRs for IBD is not known and may limit their applicability as diagnostic biomarkers, but they could be used to monitor disease activity. For example, the proinflammatory miR-21 is increased in peripheral blood of patients with IBD but is also increased in patients with other diseases, including colorectal cancer.130,131

Rs for IBD is not known and may limit their applicability as diagnostic biomarkers, but they could be used to monitor disease activity. For example, the proinflammatory miR-21 is increased in peripheral blood of patients with IBD but is also increased in patients with other diseases, including colorectal cancer.130,131 Therapeutics Further studies of epigenetic factors associated with IBD could lead to new therapeutic strategies, whether they specifically target epigenetic mechanisms or affect the pathways they control. Pharmacologic agents that affect epigenetic processes include HDACi, HAT inhibitors, and DNMT inhibitors. However, these have not been tested in patients with IBD. HDACi were first licensed in the United States for treatment of T-cell lymphoma and are now being evaluated for inflammatory disorders such as rheumatoid arthritis, multiple sclerosis, and juvenile arthritis.132,133 A controlled trial of butyrate, conducted more than 10 years ago, found that a combination of butyrate and mesalamine compounds induced remission significantly more frequently than mesalamine alone in patients with refractory UC.134 These so-called “pan-HDAC inhibitors” lack specificity for individual HDAC enzymes and are consequently accompanied by side effects such as worsening of atherosclerosis and immunosuppression. The ultimate objective will be to develop specific HDACi that target only enzymes involved in intestinal inflammation.96 For example, inhibition of HDAC9 alone could increase the function of FoxP3 T-regulatory cells and help to ameliorate colitis.135

e effects such as worsening of atherosclerosis and immunosuppression. The ultimate objective will be to develop specific HDACi that target only enzymes involved in intestinal inflammation.96 For example, inhibition of HDAC9 alone could increase the function of FoxP3 T-regulatory cells and help to ameliorate colitis.135 Conclusions Great strides have been made in understanding the genetic basis for IBD, providing insight into new pathogenic mechanisms and expanding existing ones. Further studies are required to fine map the 163 known IBD susceptibility loci and determine how they contribute to disease risk. Epigenomics is an emerging field that adds an extra layer of complexity to genetic analyses. Epigenetic studies could provide exciting clues into the pathogenesis of IBD but, like genetics, are unlikely to address all outstanding questions. It is more likely that epigenetics analyses will be integrated into larger bioanalytical models alongside other emerging IBD research disciplines, such as transcriptomics, metagenomics, glycomics, glycoproteomics, and metabolomics. Epigenetic research could provide biomarkers for use in diagnosis of IBD, along with predicting disease progression and response to therapy. Prospective studies supported by the European Union that follow the course of disease in newly diagnosed patients and relate phenotype to biomarker profile could deliver clinically useful results in the next decade to help in personalizing care.136,137 Conflicts of interest The authors disclose no conflicts.

Epigenetic research could provide biomarkers for use in diagnosis of IBD, along with predicting disease progression and response to therapy. Prospective studies supported by the European Union that follow the course of disease in newly diagnosed patients and relate phenotype to biomarker profile could deliver clinically useful results in the next decade to help in personalizing care.136,137 Conflicts of interest The authors disclose no conflicts. Funding N.T.V. is funded through the EU FP7 grant “IBD-BIOM,” and N.A.K. is funded through the Wellcome Trust. Figure 1 Roles for epigenetics in pathogenesis. Epigenetics could mediate between the genetic environment and environmental factors to help determine the phenotype of IBD. The classic paradigm of genotype leading to phenotype and disease (A) has been expanded to embrace key etiologic factors in IBD (B). Epigenetics (purple) may interact with both genetic factors (blue) and environmental factors (green) in affecting the immune system (orange). The subsequent immune response has consequences on whether insults are tolerated or chronic inflammation is initiated and propagated (red).

race key etiologic factors in IBD (B). Epigenetics (purple) may interact with both genetic factors (blue) and environmental factors (green) in affecting the immune system (orange). The subsequent immune response has consequences on whether insults are tolerated or chronic inflammation is initiated and propagated (red). Adapted with permission from Macmillan Publishers Ltd: Nature Immunology Renz et al, copyright 2011.138Figure 2 The structure and function of chromatin. The structure of chromatin helps determine whether genes are transcribed or not. Chromatin comprises DNA strands that entwine an octamer of histone proteins, comprising histone subtypes H2a, H2b (2× dimer), H3, and H4 (1× tetramer).139 In a simple model, chromatin exists as heterochromatin or euchromatin. (A) Heterochromatin is a condensed form of chromatin that does not allow access of transcription factors such as RNA polymerase and therefore prevents gene transcription. In its condensed form, heterochromatin is linked with various corepressors, methylated DNA, low levels of histone acetylation, and methylation of key lysine residues (H3 lysine 9 [H3K9Me] and 27 [H3K27Me]). (B) Euchromatin is a relaxed form of chromatin that allows access to DNA and RNA polymerases and transcription to occur. Euchromatin is associated with coactivators and specific histone acetylation (acetylation on lysine 16 of histone 4 [H4K16Ac]) that together form a large protein complex called an enhanceosome. This enhanceosome then recruits RNA polymerase to perform transcription.68,140

and RNA polymerases and transcription to occur. Euchromatin is associated with coactivators and specific histone acetylation (acetylation on lysine 16 of histone 4 [H4K16Ac]) that together form a large protein complex called an enhanceosome. This enhanceosome then recruits RNA polymerase to perform transcription.68,140 Figure 3 The relationship between genetic polymorphisms and epigenetic factors. Epigenetic features of T cells in patients with IBD affect Th1 and Th17 cell differentiation. (A) STAT4 is associated with several immune diseases, acting as a transcription factor for IL-12 and IL-23 that leads to Th1 and Th17 cell differentiation.141 An SNP in STAT4, rs7574865, is associated with several immune disorders, including IBD, rheumatoid arthritis, type 1 diabetes, and lupus.142–146 The rs7574865 risk variants (T/T + G/T) are associated with promoter region hypomethylation in colon tissues and PBMCs of patients with IBD. STAT4 promoter hypomethylation was associated with increases in STAT4 mRNA and could promote the Th1 phenotype and interferon γ production.147 In T cells from patients with asthma, STAT4 expression is also regulated by DNA methylation at promoter regions. Interestingly, STAT4 expression was markedly increased after treatment with a DNMT inhibitor.148 (B) An IBD-associated SNP in IL23-R, rs10889677, is associated with increased levels of IL-23R mRNA and protein. This could result from reduced binding of microRNAs Let-7e and Let-7f at the regulatory 3′ untranslated region of the rs10889677 risk variant (A) compared with cells from patients without IBD (C).149 Reduced binding of Let-7e and Let-7f to rs10889677 is associated with increased levels of IL23R mRNA and protein, potentially leading to sustained activation of Th17 cells and the chronic inflammation associated with IBD.149

region of the rs10889677 risk variant (A) compared with cells from patients without IBD (C).149 Reduced binding of Let-7e and Let-7f to rs10889677 is associated with increased levels of IL23R mRNA and protein, potentially leading to sustained activation of Th17 cells and the chronic inflammation associated with IBD.149 Figure 4 Loci identified in GWAS indicating roles for the innate immune response to the microbiota and autophagy in the pathogenesis of CD. Several CD risk alleles were identified in GWAS in genes that control autophagy, including TLR4, ATG16L1, IRGM, and ULK1. (A) The ULK1 locus has been associated with susceptibility to CD; it encodes a serine-threonine kinase involved in the autophagy response to starvation. ULK1 is hypermethylated in cells from patients with CD compared with controls.87 (B) IRGM encodes a gene that regulates the innate response to intracellular organisms, including Mycobacterium tuberculosis. The CD risk allele rs10065172 is associated with a deletion upstream of IRGM.150 This SNP had been termed noncausative due to an absence in alteration of protein sequence or splice sites. However, the risk variant has an altered binding site for microRNA-196. Individuals with this SNP down-regulate IGRM. The consequence is a functional reduction of autophagy and processing of the adhesive invasive Escherichia coli, which has been associated with CD.113 Table 1 DNA Methylation Studies in IBD in Peripheral Blood and Intestinal Biopsy Specimens

Figure 4 Loci identified in GWAS indicating roles for the innate immune response to the microbiota and autophagy in the pathogenesis of CD. Several CD risk alleles were identified in GWAS in genes that control autophagy, including TLR4, ATG16L1, IRGM, and ULK1. (A) The ULK1 locus has been associated with susceptibility to CD; it encodes a serine-threonine kinase involved in the autophagy response to starvation. ULK1 is hypermethylated in cells from patients with CD compared with controls.87 (B) IRGM encodes a gene that regulates the innate response to intracellular organisms, including Mycobacterium tuberculosis. The CD risk allele rs10065172 is associated with a deletion upstream of IRGM.150 This SNP had been termed noncausative due to an absence in alteration of protein sequence or splice sites. However, the risk variant has an altered binding site for microRNA-196. Individuals with this SNP down-regulate IGRM. The consequence is a functional reduction of autophagy and processing of the adhesive invasive Escherichia coli, which has been associated with CD.113 Table 1 DNA Methylation Studies in IBD in Peripheral Blood and Intestinal Biopsy Specimens Authors Subjects Study design Samples Techniques Highlighted differentially methylated loci Number of loci showing differential DNA methylation Peripheral blood DNA methylation studies Harris et al, 201284 Discordant monozygotic twins (4 CD, 7 UC), childhood IBD control (14 CD, 8 UC) Training set discordant monozygotic twins, testing set childhood IBD control Peripheral leukocytes PBMCs Methylation-specific amplification array 450K Illumina BeadChip Bisulfite pyrosequencing TEPP 1 Lin et al, 201285 18 patients with IBD (9 CD, 9 UC) Case control Epstein–Barr virus—transformed B cells Illumina GoldenGate Restriction length polymorphisms Bcl3, PPARG, STAT3, OSM, STAT5, IL12RB, SOX1, COL18A1 49 Nimmo et al, 201279 21 ileal CD 19 controls Case control, testing set on childhood IBD Whole blood Illumina 27K BeadChip MAPK13, FASLG, PRF1, S100A13, RIPK3, and IL-21R 50 Intestinal biopsy DNA methylation studies Cooke et al, 201287 8 with active UC, 8 with quiescent UC, 8 with active CD, 8 with quiescent CD, 8 without IBD Case control Active versus quiescent Pyrosequencing validation Quantitative reverse-transcriptase polymerase chain reaction mRNA quantification Rectal biopsy specimens (whole tissue and separated epithelial cells) Illumina 27K BeadChip THRAP2, FANCC, TNFSF4, TNFSF12, FUT7, CARD9, ICAM3, and IL8RB >500 Hasler et al, 201286 20 UC discordant monozygotic twins 135 unrelated subjects 3-layer EWAS: 1. Training set 2. Methylation variable positions 3. Differentially methylated regions Intestinal biopsy specimens (whole tissue) 1. Affymetrix array 2. Illumina 27K BeadChip 3. MeDip 385K CFI, SPINK4, and THY1/CD90 61 Lin et al 2012151 9 CD, 17 UC and 26 nondisease Case control Training set (14 vs 14) and testing set (12 vs 12) Intestinal tissue from surgery (whole tissue) Illumina GoldenGate Restriction length polymorphisms BGN, SERPINA, TNFSF1A, AATK, GABRA5, MAPK10, and STAT5A 7 Table 2 MicroRNA Studies in IBD in Peripheral Blood and Intestinal Biopsy Specimens

d 26 nondisease Case control Training set (14 vs 14) and testing set (12 vs 12) Intestinal tissue from surgery (whole tissue) Illumina GoldenGate Restriction length polymorphisms BGN, SERPINA, TNFSF1A, AATK, GABRA5, MAPK10, and STAT5A 7 Table 2 MicroRNA Studies in IBD in Peripheral Blood and Intestinal Biopsy Specimens Authors Subjects Samples Techniques Increased microRNA expression Decreased microRNA expression Peripheral blood microRNA studies Duttagupta et al, 2012129 20 active UC vs 20 healthy controls Peripheral blood qRT-PCR miR-188-5p, -378, -422a, -500, -501-5p, -769-5p, and -874 Paraskevi et al, 2012130 128 CD vs 162 healthy controls Peripheral blood qRT-PCR miR-16, -23a, -29a, 106a, -107, -126, -191, -199a-5p, -200c, 362-3p, and 532-3p 88 active UC vs 162 healthy controls miR-16, -21, -28-5p, -151-5p, -155, and 199a-5p Wu et al, 2011114 14 active CD vs 13 healthy controls Peripheral blood Microarray and qRT-PCR miR-199a-5p, -340, -363-3p, -532-3p, and miRplus-E1271 miR-149* and miRplus-F1065 5 quiescent CD vs 13 healthy controls miR-340* miR149* 13 active UC vs 13 healthy controls miR-28-5p, -151-5p, -103-2*, -199a-5p, -340*, -362-3p, -532-3p, and miRplus-E1271 miR-505* 10 active UC vs 10 active CD miR-28-5p, 103-2*, 149*, 151-5p, -340, -532-3p, and miRplus-E1153 miR-505* Zahm et al, 2011115 46 active CD vs 32 healthy controls Serum LDA qRT-PCR miR-16, -20a, -21, -30e, -93, -106a, -140, -192, -195, -484, and let-7b Colonic biopsy microRNA studies Bian et al, 2011109 5 active UC vs 4 healthy controls Colonic biopsy specimens qRT-PCR miR-150 Brest et al, 2011113 83 active CD vs 67 healthy controls Colonic biopsy specimens qRT-PCR and in situ hybridization miR 196 Fasseu et al, 2010111 8 active UC vs 8 healthy controls Colon biopsy specimens qRT-PCR miR-7, -31, -135b, 223, 29a, 29b, -126, -127-3p, and -324-3p miR-188-5p, -215, -320a, and -346 8 quiescent UC vs 8 healthy controls miR-196a, -29a, 29b, -126, -127-3b, and -324-3p miR-188-5p, -215, -320a, and 346 8 active CD vs 8 healthy controls miR-9, -21, -22, -26a, -29a, 29c, 30b, -31, -34c-5p, -106a, -126, -126*, -127-3p, -130a, -133b, -146a, -146b-3p, -150 ,155, -181c, -196a, -324-3p, -375 8 quiescent CD vs 8 healthy controls miR-9*, -21, -22, -26a, 29b, 29c, 30a*, -30b, -30c -31, -34c-5p, 106a, -126, -127-3p, -133b, -146a, 146b-3p, -150, -155, -196a -223, and -324-3p 8 quiescent UC vs 8 quiescent CD miR-150, 196b, -199a-3p, -199-5p, -223, and 320a Nguyen et al, 2010152 8 active CD vs 6 healthy controls Colonic biopsy

vs 8 healthy controls miR-9*, -21, -22, -26a, 29b, 29c, 30a*, -30b, -30c -31, -34c-5p, 106a, -126, -127-3p, -133b, -146a, 146b-3p, -150, -155, -196a -223, and -324-3p 8 quiescent UC vs 8 quiescent CD miR-150, 196b, -199a-3p, -199-5p, -223, and 320a Nguyen et al, 2010152 8 active CD vs 6 healthy controls Colonic biopsy specimens qRT-PCR miR-7 Olaru et al, 2011153 IBD-associated dysplasia vs active IBD Colonic biopsy specimens Microarray and qRT-PCR miR-31, 31*, -96, -135b, -141, -183, -192, -192*, -194, -194*, -200a, -200a*, -200b, -200b*, -200c, -203, -215, -224, -375, -424*, -429, and -552 miR -122, -139-5p, -142-3p, -146b-5p, -155, -223, -490-3p, 501-5p, -892b, and -1288 Pekow et al, 2012154 8 active UC vs 8 healthy controls Colonic biopsy specimens qRT-PCR miR-143 and -145 Takagi et al, 2010110 12 active UC vs 12 healthy controls Sigmoid colon biopsy Microarray and qRT-PCR miR-21 and -155 Wu et al, 2008108 15 active UC vs 15 healthy controls Sigmoid colon biopsy Microarray and qRT-PCR miR-16, -21, 23a, 24, 29a, 126, 195, and left-7f miR-192, 375, and 422b Wu et al, 2010112 5 active colonic CD vs 13 healthy controls Sigmoid colon biopsy specimens Microarray and qRT-PCR miR-23b, -106a, and -191 miR-19b and -629 6 active small bowel CD vs 13 healthy controls Terminal ileal biopsy specimens Microarray and qRT-PCR miR-16, -21, -223, and 594 qRT-PCR, quantitative reverse-transcriptase polymerase chain reaction.

Adaptation to different states, such as exercise, rest, and starvation or overnutrition, is essential for life. In turn, dysfunction and perturbation of these networks can lead to metabolic imbalances, which if uncorrected induce diseases such as obesity or diabetes. Metabolic adaptation is largely controlled by transcriptional co-regulators and transcription factors responsible, respectively, for sensing metabolic disturbances and fine-tuning the transcriptional response.1 During starvation, this adaptive response is essential for species survival, and the liver plays a central role in this process as a main site for gluconeogenesis and energy production.2 At early stages, the liver mobilizes glucose from its glycogen stores; as fasting progresses, it oxidizes fat to provide both energy for gluconeogenesis and substrate for ketogenesis. Generation of sugar from nonsugar carbon substrates (gluconeogenesis) involves several enzyme-catalyzed reactions that take place in both cytosol and mitochondria.

mobilizes glucose from its glycogen stores; as fasting progresses, it oxidizes fat to provide both energy for gluconeogenesis and substrate for ketogenesis. Generation of sugar from nonsugar carbon substrates (gluconeogenesis) involves several enzyme-catalyzed reactions that take place in both cytosol and mitochondria. Iron is essential for vital redox activities in the cell, in particular it is required for respiration and energy production in mitochondria (which are also the unique site for heme synthesis and the major site for Fe-S cluster biosynthesis), and likewise is important for mitochondria biogenesis.3 A number of iron abnormalities, ranging from low serum iron/iron-restricted anemia to hepatic/systemic iron overload, have been reported in human disorders with activated gluconeogenic signaling pathways, including obesity,4 metabolic syndrome,5–7 and diabetes.8,9 Interestingly, iron excess has been associated with worsened insulin sensitivity and disease progression, whereas iron removal has been found to be beneficial.6,8,10 Based on these premises, we asked whether iron status could be regulated directly by gluconeogenic signals.

sity,4 metabolic syndrome,5–7 and diabetes.8,9 Interestingly, iron excess has been associated with worsened insulin sensitivity and disease progression, whereas iron removal has been found to be beneficial.6,8,10 Based on these premises, we asked whether iron status could be regulated directly by gluconeogenic signals. Systemic and local iron status are under the control of hepcidin, a defensin-like circulating peptide that degrades the iron exporter ferroportin (FPN1), thereby dictating the extent of iron release or retention in the cell.11 Hepcidin expression is transcriptionally controlled by a number of factors that deliver the relevant stimulatory or inhibitory signals to the nuclear machinery and turn on or off the hepcidin (HAMP) gene. The main stimulatory transcription factors include small mother against decapentaplegic (SMAD) proteins, which bind the bone morphogenetic protein responsive element and deliver the “iron signal,”12,13 STAT-3, mainly involved in the inflammatory signal,14–16 and cyclic adenosine monophosphate (cAMP) response element binding protein 3–like 3, CREB3L3 (also known as CREBH), more recently found to mediate hepcidin induction by endoplasmic reticulum (ER) stress17 triggered by a variety of physiological and pathophysiological states.18–20 Therefore, we focused on investigating the regulation of hepcidin expression in the liver in response to gluconeogenic stimuli. To this end, we studied mice undergoing prolonged starvation, a classic model of persistently activated gluconeogenesis and insulin resistance.

Systemic and local iron status are under the control of hepcidin, a defensin-like circulating peptide that degrades the iron exporter ferroportin (FPN1), thereby dictating the extent of iron release or retention in the cell.11 Hepcidin expression is transcriptionally controlled by a number of factors that deliver the relevant stimulatory or inhibitory signals to the nuclear machinery and turn on or off the hepcidin (HAMP) gene. The main stimulatory transcription factors include small mother against decapentaplegic (SMAD) proteins, which bind the bone morphogenetic protein responsive element and deliver the “iron signal,”12,13 STAT-3, mainly involved in the inflammatory signal,14–16 and cyclic adenosine monophosphate (cAMP) response element binding protein 3–like 3, CREB3L3 (also known as CREBH), more recently found to mediate hepcidin induction by endoplasmic reticulum (ER) stress17 triggered by a variety of physiological and pathophysiological states.18–20 Therefore, we focused on investigating the regulation of hepcidin expression in the liver in response to gluconeogenic stimuli. To this end, we studied mice undergoing prolonged starvation, a classic model of persistently activated gluconeogenesis and insulin resistance. Materials and Methods Animal Studies The starvation experiment was as follows: 8- to 10-week-old male C57BL/6Crl, 129S2/SvPas, BALB/c wild-type mice, and Creb3l3-/- null mice (The Jackson Laboratory, Bar Harbor, ME) were allowed free access to water and fed a standard, iron-balanced chow diet in pellets (2018S Teklad Global 18% Protein Rodent Diet; Harlan Laboratories, (San Pietro Al Natisone, UD, Italy); iron content, 225 mg/kg) or starved up to 48 hours starting at the beginning of the light cycle.

on Laboratory, Bar Harbor, ME) were allowed free access to water and fed a standard, iron-balanced chow diet in pellets (2018S Teklad Global 18% Protein Rodent Diet; Harlan Laboratories, (San Pietro Al Natisone, UD, Italy); iron content, 225 mg/kg) or starved up to 48 hours starting at the beginning of the light cycle. Iron-deficient diet experiments were as follows: 8-week-old male C57BL/6Crl wild-type mice were fed an iron-deficient diet (ssniff EF R/M Iron Deficient; Charles River, Calco, LC, Italy; iron content, <10 mg/kg) for 9 days before death, or for 6 days before the 24- to 48-hour starvation period. All animals received humane care according to the criteria outlined by the Federation of European Laboratory Animal Science Associations. The study was approved by the Ethics Committee for Animal Studies at the University of Modena and Reggio Emilia. Blood Measurements and Tissue Iron Content Serum iron, serum ferritin (Tina-quant Ferritin kit; Roche Diagnostics, Milan, Italy), hemoglobin, and glucose were determined using an automated COBAS C501 counter (Roche, Milan, Italy) at the clinical-chemical laboratory of the University Hospital of Modena. Serum hepcidin was determined using an enzyme-linked immunosorbent assay kit (USCN Life Science, Hubei, China) according to the manufacturer's instructions, as previously reported.9 Serum ketone bodies were analyzed using a β-Hydroxybutyrate Assay Kit (Sigma-Aldrich, Milan, Italy) following the manufacturer's instructions. Liver and spleen tissue specimens were analyzed for non-heme iron content as previously reported.21

Blood Measurements and Tissue Iron Content Serum iron, serum ferritin (Tina-quant Ferritin kit; Roche Diagnostics, Milan, Italy), hemoglobin, and glucose were determined using an automated COBAS C501 counter (Roche, Milan, Italy) at the clinical-chemical laboratory of the University Hospital of Modena. Serum hepcidin was determined using an enzyme-linked immunosorbent assay kit (USCN Life Science, Hubei, China) according to the manufacturer's instructions, as previously reported.9 Serum ketone bodies were analyzed using a β-Hydroxybutyrate Assay Kit (Sigma-Aldrich, Milan, Italy) following the manufacturer's instructions. Liver and spleen tissue specimens were analyzed for non-heme iron content as previously reported.21 Real-Time Quantitative Reverse-Transcription Polymerase Chain Reaction and Semiquantitative Reverse-Transcription Polymerase Chain Reaction Total cellular RNA was obtained by incubating cells in iScript quantitative reverse-transcription polymerase chain reaction (qRT-PCR) Sample Preparation Reagent (Bio-Rad, Milan, Italy) according to the manufacturer's instructions. Total hepatic RNA was extracted as described.17 Complementary DNA was generated by reverse transcription of 2 μL of iScript buffer (for cultured cells) or 1 μg (for liver) with 200 U ImProm-II Reverse Transcriptase (Promega, Milan, Italy) following the manufacturer's instructions. Expression of mRNA was analyzed using SsoFast EvaGreen Supermix (Bio-Rad). Primer sequences are listed in Supplementary Table 1. Cycling conditions were as follows: 30 seconds at 98°C, followed by 40 cycles of 2 seconds at 98°C and 10 seconds at 60°C. After 40 amplification cycles, threshold cycle values were calculated automatically using the default settings of the CFX Manager software (version 2.0; Bio-Rad), and femtograms of starting complementary DNA were calculated from a standard curve covering a range of 5 orders of magnitude. At the end of the PCR run, melting curves of the amplified products were obtained and used to determine the specificity of the amplification reaction. In each experiment, the change of specific mRNA expression was reported as the fold increase as compared with that of control cells or mice. Normalization of qRT-PCR data was based on RPL19 housekeeping mRNA expression after validation using the target stability value obtained from the CFX Manager software (version 2.0; Bio-Rad).22 X-box binding protein 1 (Xbp1) splicing was analyzed as described by Vecchi et al.17 Primer sequences are listed in Supplementary Table 1. The Hamp oligos detects total Hamp mRNA (Hamp1 and Hamp2 mRNA).

on after validation using the target stability value obtained from the CFX Manager software (version 2.0; Bio-Rad).22 X-box binding protein 1 (Xbp1) splicing was analyzed as described by Vecchi et al.17 Primer sequences are listed in Supplementary Table 1. The Hamp oligos detects total Hamp mRNA (Hamp1 and Hamp2 mRNA). Western Blot Analyses For the FPN1 assay, mouse liver specimens were homogenized in lysis buffer (150 mmol/L NaCl, 10 mmol/L Tris, pH = 8, 1 mmol/L EDTA, 0.5% Triton X-100) containing 1:100 protease inhibitor cocktail (Sigma-Aldrich). After centrifugation at 13,000 × g at 4°C for 15 minutes, the supernatant was collected and the protein concentration was assayed by the Bradford method. A total of 60 μg of liver extracts were loaded without boiling on 10% acrylamide gels with Laemmli sample buffer, and run in sodium dodecyl sulfate–polyacrylamide gel electrophoresis buffer. Membranes were probed with specific antibodies: rabbit anti-FPN1 (1:1000; Alpha Diagnostic, Inc, San Antonio, TX), as previously reported,23 and mouse anti-tubulin (1:3000; Sigma-Aldrich), followed by appropriate horseradish-peroxidase–conjugated secondary antibodies. Western blot analysis was performed by Western Lightning Ultra substrate (PerkinElmer, Waltham, MA) according to the manufacturer's instructions. Chemiluminescence was detected and quantified using the Molecular Imager ChemiDoc XRS+ with Image Lab Software (Bio-Rad).

seradish-peroxidase–conjugated secondary antibodies. Western blot analysis was performed by Western Lightning Ultra substrate (PerkinElmer, Waltham, MA) according to the manufacturer's instructions. Chemiluminescence was detected and quantified using the Molecular Imager ChemiDoc XRS+ with Image Lab Software (Bio-Rad). Cell Cultures and Primary Hepatocyte Isolation Human hepatoma HepG2 cells were cultured in Modified Eagle's Medium (MEM) (containing 1 g/L glucose), supplemented with 1 mmol/L glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% heat-inactivated fetal bovine serum, in a 5% CO2 atmosphere at 37°C. Mouse primary hepatocytes from 8- to 10-week-old male C57BL/6Crl mice were isolated as previously described.21 HepG2 cells and mouse primary hepatocytes were incubated for 8 hours in the presence of 1 mmol/L of 8Br cAMP (Sigma-Aldrich) or for 6 hours in the presence of 100 nmol/L of glucagon (Sigma-Aldrich), both in 2% fetal bovine serum culture medium. Plasmids, Small Interfering RNAs, Transfection, and Luciferase Assay Hepcidin promoter construct, plasmid encoding Flag-tagged CREB3L3-N (the active form of the factor), CREB3L3 small interfering RNA (siRNA) transfection, and luciferase analysis have been reported elsewhere.17 Plasmid encoding peroxisome proliferator-activated receptor gamma coactivator 1-α (PPARGC1A) was kindly provided by Dr Chang Liu (Nanjing, China). PPARGC1A siRNA were obtained from Invitrogen (Life Technologies Italia, Monza, Italy) (PPARGC1AHSS116799).

nsfection, and luciferase analysis have been reported elsewhere.17 Plasmid encoding peroxisome proliferator-activated receptor gamma coactivator 1-α (PPARGC1A) was kindly provided by Dr Chang Liu (Nanjing, China). PPARGC1A siRNA were obtained from Invitrogen (Life Technologies Italia, Monza, Italy) (PPARGC1AHSS116799). Chromatin Immunoprecipitation Chromatin immunoprecipitation (ChIP) was described elsewhere17 with the following modifications. Briefly, HepG2 cells were transfected using X-tremeGENE transfection reagent (Roche Applied Science, Milan, Italy) with plasmid encoding Flag-tagged CREB3L3-N. Forty-eight hours after transfection, cells were treated with 1 mmol/L 8Br cAMP for 8 hours and fixed for formaldehyde cross-linking and ChIP. Protein–DNA complexes were immunoprecipitated overnight using the following antibodies: anti-Flag (Sigma-Aldrich), anti-PPARGC1A (anti-PGC1A; Santa Cruz Biotechnology, Dallas, TX), or anti-green fluorescent protein (GFP) (Abcam, Cambridge, UK) as negative control.

ours and fixed for formaldehyde cross-linking and ChIP. Protein–DNA complexes were immunoprecipitated overnight using the following antibodies: anti-Flag (Sigma-Aldrich), anti-PPARGC1A (anti-PGC1A; Santa Cruz Biotechnology, Dallas, TX), or anti-green fluorescent protein (GFP) (Abcam, Cambridge, UK) as negative control. Statistical Analyses All data were controlled for normal distribution (Kolmogorov–Smirnov and Shapiro–Wilk tests). When comparing a variable in 2 groups, a paired t test or the Wilcoxon–Mann–Whitney test was used, depending on the presence or absence of normal data distribution and/or small sample size. When making multiple statistical comparisons on a single data set, for normally distributed data a 1-way analysis of variance with the Tukey or Dunnett post hoc tests, depending on the presence or absence of homoscedasticity, was used. For skewed data, the Kruskal–Wallis test was used. In all statistical analyses, a P value less than .05 was considered significant. Data presented in Figures are mean ± SEM. All analyses were conducted using Prism 5 for mac OS X version 5.0a software (GraphPad Software, Inc, La Jolla, CA).

f homoscedasticity, was used. For skewed data, the Kruskal–Wallis test was used. In all statistical analyses, a P value less than .05 was considered significant. Data presented in Figures are mean ± SEM. All analyses were conducted using Prism 5 for mac OS X version 5.0a software (GraphPad Software, Inc, La Jolla, CA). Results In starving mice, phosphoenolpyruvate carboxykinase 1 (Pck1) mRNA, known to be readily responsive to gluconeogenic stimuli, rapidly increased at 2 hours (Figure 1A), whereas Hamp mRNA increased at 5 hours, in concomitance with a marked serum glucose decrease, and remained increased for up to 48 hours (Figure 1B). In addition, serum hepcidin showed a sharp increase at 5 hours, although slightly decreased at later time points (Figure 1C). Hamp induction led to a decrease of serum iron, and a progressive increase of serum ferritin and iron content in the spleen and the liver (Table 1). In agreement with the hepcidin model of iron regulation, which implies a post-translational down-regulation of ferroportin protein by hepcidin, hepatic Fpn1 mRNA was unchanged (Figure 1D) whereas FPN1 protein was degraded in a time-dependent manner starting at 5 hours' starvation (Figure 1E and F). A visible ferroportin down-regulation in the spleen was not detected (data not shown). As discussed earlier, the main stimuli for hepcidin transcription in vivo are increased serum and hepatic iron,24 and cytokines produced during inflammation and infection, particularly interleukin 6 (IL6),25 IL22,26 tumor necrosis factor-α,27 and ER stress.17 In mice undergoing prolonged starvation, we were unable to detect up-regulation of cytokines such as IL6 and tumor necrosis factor-α, whereas IL22 actually was depressed by food withdrawal (Supplementary Figure 1A–C). IL1β was induced by short-term fasting but returned to normal at 48 hours (Supplementary Figure 1D), when hepcidin mRNA expression was still increased markedly. Similar negative results were found when analyzing inflammation marker C-reactive protein (Crp) mRNA (Supplementary Figure 1E) and ER stress markers (namely, Xbp1 mRNA splicing; Supplementary Figure 1F). To address whether hypoferremia in starving mice was caused by lower iron intake associated with food deprivation, we studied mice premaintained on an iron-deprived diet for 1 week.

arker C-reactive protein (Crp) mRNA (Supplementary Figure 1E) and ER stress markers (namely, Xbp1 mRNA splicing; Supplementary Figure 1F). To address whether hypoferremia in starving mice was caused by lower iron intake associated with food deprivation, we studied mice premaintained on an iron-deprived diet for 1 week. After the iron-deficient diet, this group of mice showed normal serum iron levels (Figure 2A), but almost halved spleen iron stores compared with fed mice maintained on an iron-balanced diet (Figure 2B), suggesting a marked iron redistribution from the storage site toward the bloodstream to sustain red cell production and maintain normal hemoglobin levels (Figure 2C). However, even under this circumstance, starvation led to a progressive decrease of serum iron (Figure 2A). Moreover, hepcidin mRNA expression, although depressed in control mice (iron-deficient group) likely because of the latent iron-deficiency state and active marrow activity, still dramatically was induced by starvation (Figure 2D). Activation of hepcidin and perturbation of iron homeostasis during starvation-induced gluconeogenesis also was found in other tested mouse strains, such as BALB/c (Supplementary Figure 2A–C) or 129S2 (Supplemental Figure 2D–F). Overall, these data suggested that, in starving mice, stimuli that are independent of inflammation and/or stress may be responsible for hepcidin induction. To identify the molecular basis for this novel hepcidin regulatory mechanism, we used an in vitro approach.

After the iron-deficient diet, this group of mice showed normal serum iron levels (Figure 2A), but almost halved spleen iron stores compared with fed mice maintained on an iron-balanced diet (Figure 2B), suggesting a marked iron redistribution from the storage site toward the bloodstream to sustain red cell production and maintain normal hemoglobin levels (Figure 2C). However, even under this circumstance, starvation led to a progressive decrease of serum iron (Figure 2A). Moreover, hepcidin mRNA expression, although depressed in control mice (iron-deficient group) likely because of the latent iron-deficiency state and active marrow activity, still dramatically was induced by starvation (Figure 2D). Activation of hepcidin and perturbation of iron homeostasis during starvation-induced gluconeogenesis also was found in other tested mouse strains, such as BALB/c (Supplementary Figure 2A–C) or 129S2 (Supplemental Figure 2D–F). Overall, these data suggested that, in starving mice, stimuli that are independent of inflammation and/or stress may be responsible for hepcidin induction. To identify the molecular basis for this novel hepcidin regulatory mechanism, we used an in vitro approach. The hepatic expression of genes encoding gluconeogenic enzymes, such as PCK1, is regulated by a network of transcription factors and cofactors, including CREB proteins28,29 and PPARGC1A.30 We recently found that a member of the CREB family, CREBH, is engaged constitutively on the hepcidin promoter and readily transactivates it during ER stress.17 Both Ppargc1a and Creb3l3 mRNA are induced by hepatic gluconeogenesis in vivo during starvation (Figure 3A and B). We hypothesized that CREBH is a target for PPARGC1A coactivation during hepcidin induction by active gluconeogenesis. In line with this hypothesis, PPARGC1A silencing in HepG2 cells led to a 60% decrease of hepcidin mRNA expression, similar to the effect obtained by CREB3L3 knockdown (Figure 3C).

ion (Figure 3A and B). We hypothesized that CREBH is a target for PPARGC1A coactivation during hepcidin induction by active gluconeogenesis. In line with this hypothesis, PPARGC1A silencing in HepG2 cells led to a 60% decrease of hepcidin mRNA expression, similar to the effect obtained by CREB3L3 knockdown (Figure 3C). Gluconeogenesis induced by food deprivation involves cAMP as the main intracellular second messenger in response to hormonal stimuli.31,32 HepG2 cells exposed to 8Br cAMP, a cAMP analog, showed a significant increase of both PCK1 and HAMP mRNA in a time-dependent manner (Figure 4A). A similar trend of hepcidin activation also was found in primary hepatocytes exposed to either glucagon or 8Br cAMP. Both treatments induced Pck1 and Hamp mRNA expression in cultured hepatocytes, although Hamp response was significantly but appreciably lower than in HepG2 cells (Figure 4B). Hepcidin stimulation by 8Br cAMP in HepG2 cells transfected with siRNA for either PPARGC1A or CREB3L3 was appreciably lower as compared with 8Br cAMP-treated control cells (Figure 4C). A similar effect was documented when we tested the response of Hamp promoter to 8Br cAMP in the presence of PPARGC1A or CREB3L3 siRNAs (Figure 4D). To prove that PPARGC1A cooperates with CREBH to turn on hepcidin in response to gluconeogenesis, we assessed if the coactivator PPARGC1A/CREBH transduces and binds the hepcidin promoter in response to gluconeogenic stimuli. Overexpression of PPARGC1A in HepG2 cells led to a significant transactivation of the Hamp promoter, indicating that the transcription factor is involved in hepcidin promoter regulation (Figure 4E). In a previous study we showed that CREBH constitutively occupies the HAMP promoter and transactivates it in response to ER stress.17 Here, the ChIP assay showed that, in addition to the known constitutive hepcidin promoter occupancy by CREBH (Figure 4F, αFlag, control cells), PPARGC1A also constitutively binds to the same region (Figure 4F, αPGC1A, control cells). In agreement with the studies reported earlier, after exposure of HepG2 cells to 8Br cAMP, more CREBH was stabilized on the HAMP promoter in the presence of stable PPARGC1A binding (Figure 4F, 8Br cAMP-treated cells).

trol cells), PPARGC1A also constitutively binds to the same region (Figure 4F, αPGC1A, control cells). In agreement with the studies reported earlier, after exposure of HepG2 cells to 8Br cAMP, more CREBH was stabilized on the HAMP promoter in the presence of stable PPARGC1A binding (Figure 4F, 8Br cAMP-treated cells). In Creb3l3 null mice, in agreement with the in vitro studies, starvation correctly induced Pck1 mRNA (Figure 5A), but was unable to activate hepcidin mRNA (Figure 5B), modify serum hepcidin levels (Figure 5C), or cause hypoferremia (Figure 5D). Of note, Ppargc1a mRNA was still induced by starvation (Figure 5E), but it apparently was unable to stimulate hepcidin expression in the absence of CREBH. These data support a role for CREBH in hepcidin activation by gluconeogenic stimuli in the liver. Interestingly, serum glucose levels were significantly lower in starving Creb3l3 null mice as compared with starving wild-type mice (Table 2). Seemingly, the increase of serum ketone bodies during starvation was more pronounced in the Creb3l3 null mice (4.7- to 5.6-fold) as compared with control mice (3.1- to 3.5-fold) (Table 2).

ly, serum glucose levels were significantly lower in starving Creb3l3 null mice as compared with starving wild-type mice (Table 2). Seemingly, the increase of serum ketone bodies during starvation was more pronounced in the Creb3l3 null mice (4.7- to 5.6-fold) as compared with control mice (3.1- to 3.5-fold) (Table 2). Discussion Hepcidin is constitutively produced by the liver to maintain plasma iron levels within a narrow physiologic range. To do so it senses a variety of physiologic and pathophysiologic stimuli that tend to alter blood iron levels, and responds by inhibiting ferroportin, the main iron-exporter in mammals.33 In this study we showed that hepcidin is regulated transcriptionally also by gluconeogenic signals through PPARGC1A/CREBH. Induction of this regulatory pathway in a classic model of insulin resistance/activated gluconeogenesis, ie, starvation, leads to tissue iron retention and circulatory iron deficiency. Hypoferremia is clearly secondary to increased tissue iron retention after hepcidin induction and not to reduced food iron intake because it still is preserved in mice premaintained on an iron-deprived diet (Figure 2). Activation of hepcidin and perturbation of iron homeostasis during starvation-induced gluconeogenesis seem to represent a general defensive response in rodents because it was found in other tested mouse strains. However, differences in terms of the time course of hepcidin induction and the extent of iron status modifications were detected clearly among various starving mice strains. This could be explained by the fact that both the gluconeogenic response/gluconeogenic gene expression and iron status/iron gene expression may vary appreciably among mouse strains, as also documented by the significantly higher expression of the Pck1 gene in C57BL/6 mice (an optimal mouse model for studying gluconeogenesis/insulin resistance34,35 and the model that most closely parallels the gluconeogenic response to starvation seen in human beings) as compared with 129S2, BALB/c, and Creb3l3 null mice (which actually display a mixed genetic background of 129S1, 129X1, C57BL/6, FVB/N). A close look at the time course induction of Pck1/Hamp (Figure 1A and B) and Ppargc1a/Creb3l3 RNAs (Figure 3A and B) suggests that the initial 5-hour burst of Hamp transcription largely depends on increased Creb3l3 expression.

ull mice (which actually display a mixed genetic background of 129S1, 129X1, C57BL/6, FVB/N). A close look at the time course induction of Pck1/Hamp (Figure 1A and B) and Ppargc1a/Creb3l3 RNAs (Figure 3A and B) suggests that the initial 5-hour burst of Hamp transcription largely depends on increased Creb3l3 expression. Later, the increase in Ppargc1A expression likely sustains hepcidin transcription by enhancing and further stabilizing CREBH binding on the Hamp promoter (Figure 4F, ChIP study). We were able to reproduce the effect of starvation in vitro, in a hepatoma cell line and cultured primary hepatocytes, using different gluconeogenic stimuli (Figure 4). However, the Hamp gene response to gluconeogenic signals in primary hepatocytes was lower than in hepatoma cells. This may depend on the fact that in primary hepatocytes the gluconeogenic signals may be attenuated, as indicated by the lower Pck1 induction in primary hepatocytes exposed to 8Br cAMP as compared with HepG2 cells (Figure 4B vs A), and/or that additional factors essential for the hepcidin transcriptional machinery are lost in primary hepatocytes after the disruption of liver architecture/microenvironment. The newly identified regulatory pathway links glucose and iron metabolism in the liver and identifies hepcidin, the iron hormone, as a gluconeogenic sensor.

Later, the increase in Ppargc1A expression likely sustains hepcidin transcription by enhancing and further stabilizing CREBH binding on the Hamp promoter (Figure 4F, ChIP study). We were able to reproduce the effect of starvation in vitro, in a hepatoma cell line and cultured primary hepatocytes, using different gluconeogenic stimuli (Figure 4). However, the Hamp gene response to gluconeogenic signals in primary hepatocytes was lower than in hepatoma cells. This may depend on the fact that in primary hepatocytes the gluconeogenic signals may be attenuated, as indicated by the lower Pck1 induction in primary hepatocytes exposed to 8Br cAMP as compared with HepG2 cells (Figure 4B vs A), and/or that additional factors essential for the hepcidin transcriptional machinery are lost in primary hepatocytes after the disruption of liver architecture/microenvironment. The newly identified regulatory pathway links glucose and iron metabolism in the liver and identifies hepcidin, the iron hormone, as a gluconeogenic sensor. PPARGC1A is a transcriptional coactivator that regulates the genes involved in energy metabolism. During starvation, PPARGC1A readily is activated to turn on the gluconeogenic machinery, but also to stimulate mitochondrial biogenesis and respiration,36 which are essential to support the increased energy demands. Interestingly, in osteoclasts, mitochondrial biogenesis involves CREB/PPARGC1A proteins, but requires iron uptake and supply to mitochondrial respiratory proteins.37 Here, we found that PPARGC1A constitutively occupies the hepcidin promoter and, in response to gluconeogenic stimuli, stabilizes CREBH binding and transactivates HAMP promoter. CREBH is an ER stress–associated liver-specific transcription factor originally involved in the induction of acute-phase response genes (such as serum amyloid protein and C-reactive protein38), and subsequently has been found to activate the transcription of HAMP.17 Based on recent publications and this report, CREBH now emerges as a key metabolic regulator in the liver: it is activated by fatty acids and PPARα,39,40 and regulates the expression of genes involved in hepatic lipogenesis, fatty acid oxidation, and lipolysis under metabolic stress.20 Interestingly, CREBH also has been found to transcriptionally regulate Pck1 and glucose-6-phosphatase, the critical genes in hepatic gluconeogenic response.41 Here, we report that CREBH is engaged constitutively on the hepcidin promoter to sense metabolic gluconeogenic stress and modify, accordingly, iron traffic. Of note is that starving Creb3l3 null mice show reduced glucose and increased ketone body output.

sphatase, the critical genes in hepatic gluconeogenic response.41 Here, we report that CREBH is engaged constitutively on the hepcidin promoter to sense metabolic gluconeogenic stress and modify, accordingly, iron traffic. Of note is that starving Creb3l3 null mice show reduced glucose and increased ketone body output. Adaptation to starvation is essential for species survival.42 Seemingly, defense against pathogens represents a priority in species evolution. The liver, as the main source for hepcidin, seems to play a central role in both processes. During infection, hepcidin limits vital iron that is needed by invading microorganisms, thus contributing to host defense.25 During prolonged starvation, hepcidin likely preserves tissue iron and helps to maintain energy balance and support gluconeogenesis in the liver (this report). Most likely, this response originally evolved to protect human beings during food withdrawal.

d by invading microorganisms, thus contributing to host defense.25 During prolonged starvation, hepcidin likely preserves tissue iron and helps to maintain energy balance and support gluconeogenesis in the liver (this report). Most likely, this response originally evolved to protect human beings during food withdrawal. Paradoxically, in human disorders associated with food excess and storage, such as type 2 diabetes, obesity, and the metabolic syndrome, persistently activated gluconeogenesis may result in overstimulation of hepcidin, iron accumulation, and potential damage. Cases of unexplained hepatic iron excess, characterized by high serum ferritin levels with normal or subnormal transferrin saturation, and associated with metabolic abnormalities, originally were reported by Moirand et al,43 who also introduced the term iron overload-associated insulin resistance (recently renamed dysmetabolic iron overload syndrome). Hepatocellular and/or mesenchymal iron deposition, usually slight or mild, has been reported since then in nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis.44 The clinical relevance of iron excess in these disorders, in terms of fibrosis development and cancer risk, is actively debated,45 but increasing data indicate that iron may sustain disease activity and/or contribute to its progression.46–49 Interestingly, NAFLD patients with mixed or mesenchymal iron overload (a pattern of iron deposition consistent with a “hepcidin-excess model”) seem more likely to develop fibrosis than those with pure parenchymal iron deposits (a pattern of iron deposition consistent with a “hepcidin-deficient model”).47,49 The mechanism of iron deposition in NAFLD/dysmetabolic iron overload syndrome likely is multifactorial: sex, diet, disease activity, genetic background (HFE hemochromatosis gene mutations), ethnicity, and (micro)inflammation all may account for the variability of both iron excess and its pattern of distribution. We hypothesize that a fraction of dysmetabolic/NAFLD patients with normal-low transferrin saturation and mixed/mesenchymal hepatic iron deposits may represent a subgroup of patients with prominent insulin resistance and hepcidin induction via the gluconeogenic PPARGC1A/CREBH-driven pathway described here.

distribution. We hypothesize that a fraction of dysmetabolic/NAFLD patients with normal-low transferrin saturation and mixed/mesenchymal hepatic iron deposits may represent a subgroup of patients with prominent insulin resistance and hepcidin induction via the gluconeogenic PPARGC1A/CREBH-driven pathway described here. In these patients, hepcidin, depending on the degree and duration of its induction, may modify iron traffic locally or systemically and lead, respectively, to simple hepatic iron retention with marginal systemic reflections (ie, mesenchymal/mixed hepatic iron accumulation with normal or subnormal transferrin saturation), or substantial tissue iron retention, hypoferremia, and iron-restricted anemia. Further studies are needed to prove that the gluconeogenic signal-driven induction of hepcidin in starving mice also takes place in other instances of activated gluconeogenesis and insulin resistance, such as diabetes, obesity, or NAFLD. If so, because of the increasingly recognized negative effect of iron excess on the progression of these disorders, the novel regulatory pathway reported here may offer potential new therapeutic targets to prevent or correct iron disturbances in common metabolic disorders.

sistance, such as diabetes, obesity, or NAFLD. If so, because of the increasingly recognized negative effect of iron excess on the progression of these disorders, the novel regulatory pathway reported here may offer potential new therapeutic targets to prevent or correct iron disturbances in common metabolic disorders. Supplementary Material Supplementary Figure 1 Hepatic expression of inflammation or ER stress markers in mice during starvation. Total liver mRNA analysis in liver of C57BL/6 mice fed a standard diet (white bar) and starved for the indicated time points (gray bars). (A–D) Real-time qRT-PCR analysis of cytokine mRNA expression relative to housekeeping Rpl19 mRNA: (A) Il6, (B) Il22, (C) Tnf, and (D) Il1β. (E) Crp mRNA expression, as an inflammatory marker, and (F) PCR analysis of Xbp1 mRNA splicing analysis, as an ER stress marker. Results are mean ± SEM of 6–8 mice per group. For mRNA expression analysis, mean control values for the fed mice group are set to 1. In the Xbp1 splicing analysis, 3 representative mice per group are shown. MW, molecular weight, PC positive control. P values are reported for comparisons between fed mice and fasted mice at each time point. *P < .05.

mice per group. For mRNA expression analysis, mean control values for the fed mice group are set to 1. In the Xbp1 splicing analysis, 3 representative mice per group are shown. MW, molecular weight, PC positive control. P values are reported for comparisons between fed mice and fasted mice at each time point. *P < .05. Supplementary Figure 2 Fasting induces hepcidin gene expression and hypoferremia in vivo in BALB/c and 129S2/SvPas (129S2) wild-type mice. Eight- to 10-week-old (A–C) BALB/c and (D–F) 129S2/SvPas wild-type mice were fasted for 24-48 hours. Real-time qRT-PCR analysis of (A and D) Pck1 mRNA, (B and E) Hamp mRNA, and (C and F) serum iron in fed and fasted mice. Results are expressed as the mean ± SEM of 6–8 mice per group. For mRNA expression analysis in panels A and B and in D and E, the mean control values are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between control fed mice and fasted mice. *P < .05, **P < .01, ***P < .001. Supplementary Table 1 List of Primers Used for Real-Time qRT-PCR

Supplementary Figure 2 Fasting induces hepcidin gene expression and hypoferremia in vivo in BALB/c and 129S2/SvPas (129S2) wild-type mice. Eight- to 10-week-old (A–C) BALB/c and (D–F) 129S2/SvPas wild-type mice were fasted for 24-48 hours. Real-time qRT-PCR analysis of (A and D) Pck1 mRNA, (B and E) Hamp mRNA, and (C and F) serum iron in fed and fasted mice. Results are expressed as the mean ± SEM of 6–8 mice per group. For mRNA expression analysis in panels A and B and in D and E, the mean control values are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between control fed mice and fasted mice. *P < .05, **P < .01, ***P < .001. Supplementary Table 1 List of Primers Used for Real-Time qRT-PCR Accession number Gene name Forward Reverse Murine oligonucleotides NM_032541.1 Hamp 5′-GCCTGTCTCCTGCTTCTCCT-3′ 5′-GCTCTGTAGTCTGTCTCATCTGTT-3′ NM_011044.2 Pck1 5′-AACTGTTGGCTGGCTCTC-3′ 5′-GAACCTGGCGTTGAATGC-3′ NM_145365.3 Creb3l3 5′-GATACCCTGTACCCGGAGGAG-3′ 5′-CGGACAGCAGCAGTTCCTTC-3′ NM_008904.2 Ppargc1a 5′-CCGTAAATCTGCGGGATGATG-3′ 5′-CAGTTTCGTTCGACCTGCGTAA-3′ NM_031168.1 Il6 5′-ATGGATGCTACCAAACTGGAT-3′ 5′-TGAAGGACTCTGGCTTTGTCT-3′ NM_008361.3 Il1b 5′-CCTTTTCGTGAATGAGCAGACAG-3′ 5′-TCTCTTTGAACAGAATGTGCCATG-3′ NM_016971.2 Il22 5′-ATGAGTTTTTCCCTTATGGGGAC-3′ 5′-GCTGGAAGTTGGACACCTCAA-3′ NM_013693.2 Tnf 5′-ATGGCCTCCCTCTCATCAGTT-3′ 5′-GGCTACAGGCTTGTCACTCG-3′ NM_007768.4 Crp 5′-CTGCACAAGGGCTACACTGT-3′ 5′-TCTCCCACCAAAGACTGCTTT-3′ NM_009078.2 Rpl19 5′-ATGAGTATGCTCAGGCTACAGA-3′ 5′-GCATTGGCGATTTCATTGGTC-3′ NM_016917.2 Fpn1 5′-AAGACTCCAACATCCGTGAACTT-3′ 5′-GACCCATCCATCTCGGAAAGTG-3′ Human oligonucleotides NM_021175.2 HAMP 5′-TGTTTTCCCACAACAGACGGG-3′ 5′-CGCAGCAGAAAATGCAGATGG-3′ NM_000981.3 RPL19 5-′GGGCATAGGTAAGCGGAAGG-3′ 5′-TCAGGTACAGGCTGTGATACA-3′ NM_032607.1 CREB3L3 5′-CCTCTGTGACCATAGACCTGG-3′ 5′-ACGGTGAGATTGCATCGTGG-3′ NM_013261.3 PPARGC1A 5′-GCTTTCTGGGTGGACTCAAGT-3′ 5′-TCTAGTGTCTCTGTGAGGACTG-3′ NM_002591.3 PCK1 5′-GCGGCTGAAGAAGTATGA-3′ 5′-GGAACCTGGCATTGAACG-3′

5′-CGCAGCAGAAAATGCAGATGG-3′ NM_000981.3 RPL19 5-′GGGCATAGGTAAGCGGAAGG-3′ 5′-TCAGGTACAGGCTGTGATACA-3′ NM_032607.1 CREB3L3 5′-CCTCTGTGACCATAGACCTGG-3′ 5′-ACGGTGAGATTGCATCGTGG-3′ NM_013261.3 PPARGC1A 5′-GCTTTCTGGGTGGACTCAAGT-3′ 5′-TCTAGTGTCTCTGTGAGGACTG-3′ NM_002591.3 PCK1 5′-GCGGCTGAAGAAGTATGA-3′ 5′-GGAACCTGGCATTGAACG-3′ Conflicts of interest The authors disclose no conflicts. Funding This work was supported by Telethon grant GGP10233 and PRIN grant 2010REYFZH_005 to AP. Author names in bold designate shared co-first authorship. Figure 1 In vivo time course of hepcidin and ferroportin expression in mice during starvation. (A) Real-time qRT-PCR analysis of Pck1 mRNA and (B) Hamp mRNA expression relative to housekeeping Rpl19 mRNA in C57BL/6 mice fed a standard diet (white bar) or starved for the indicated time periods (gray bars). (C) Enzyme-linked immunosorbent assay quantification of serum hepcidin levels. (D) Fpn1 mRNA expression relative to housekeeping Rpl19 mRNA. (E) Western blot analysis of FPN1 protein expression in the liver, with tubulin as loading control. The arrow indicates the specific FPN1 band, whereas the nonspecific upper band is owing to the secondary antibody. (F) Densitometric quantification of FPN1 protein expression relative to tubulin. Results are mean ± SEM of 6–8 mice per group. In Western blot analysis, 3 representative mice per group are shown. For mRNA expression analysis, mean control values for the fed mice group are set to 1. P values are reported for comparisons between fed mice and mice fasted at each time point. *P < .05, ***P < .001.

lin. Results are mean ± SEM of 6–8 mice per group. In Western blot analysis, 3 representative mice per group are shown. For mRNA expression analysis, mean control values for the fed mice group are set to 1. P values are reported for comparisons between fed mice and mice fasted at each time point. *P < .05, ***P < .001. Figure 2 Fasting induces hepcidin gene expression also in mice premaintained on an iron-deficient diet. Eight- to 10-week-old male C57BL/6Crl mice were fed an iron-balanced diet or an iron-deficient diet for 9 days before death (IB and ID, respectively), or for 6 days before the 24- to 48-hour starvation period (ID fast 24-hr and 48 hr). (A) Serum iron quantification, (B) spleen iron content, (C) hemoglobin (Hb) levels, and (D) Hamp mRNA expression relative to housekeeping Rpl19 mRNA expression. Results are expressed as the mean ± SEM of 5–6 mice per group. P values are reported for comparisons between the indicated groups. *P < .05, **P < .01, ***P < .001.

erum iron quantification, (B) spleen iron content, (C) hemoglobin (Hb) levels, and (D) Hamp mRNA expression relative to housekeeping Rpl19 mRNA expression. Results are expressed as the mean ± SEM of 5–6 mice per group. P values are reported for comparisons between the indicated groups. *P < .05, **P < .01, ***P < .001. Figure 3 Ppargc1a and Creb3l3 are induced by starvation and are involved in hepcidin expression. (A) Real-time qRT-PCR analysis of Ppargc1a mRNA and (B) Creb3l3 mRNA expression in liver of C57BL/6 mice fed an iron-standard diet (white bar) and starved for the indicated time points (gray bars). (C) Basal expression of HAMP mRNA in HepG2 cells transfected with siRNAs against PPARGC1A, CREB3L3, or both. Results are mean ± SEM of 6–8 mice per group or 3–4 independent experiments each repeated in triplicate. Mean control values for the fed mice group for in vivo experiments or unspecific (US) RNA interference (RNAi) for in vitro experiments are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between control and treated cells, or between indicated groups. *P < .05, **P < .01, ***P < .001.

for the fed mice group for in vivo experiments or unspecific (US) RNA interference (RNAi) for in vitro experiments are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between control and treated cells, or between indicated groups. *P < .05, **P < .01, ***P < .001. Figure 4 Hepcidin is induced by gluconeogenic signals through PPARGC1A/CREBH. (A) HepG2 cells were cultured in the presence of a cAMP analog (8Br cAMP) and analyzed at different time points for PCK1 and HAMP mRNA expression by real-time qRT-PCR. (B) Pck1 and Hamp mRNA expression in primary mouse hepatocytes isolated from C57BL/6 mice and exposed to glucagon or 8Br cAMP. (C) HAMP mRNA expression and (D) Hamp-promoter luciferase activity in HepG2 cells after silencing of PPARGC1A and CREB3L3. (E) Hamp-promoter luciferase activity in HepG2 cells transfected with control plasmid (empty) or construct encoding PPARGC1A protein. (F) ChIP assay of HepG2 cells transfected with Flag-tagged CREB3L3-N vector and exposed to 8Br cAMP. CREBH (αFlag) and PPARGC1A (αPGC1A) occupancy of CREBH site on hepcidin endogenous promoter was evaluated by real-time qRT-PCR and expressed as a percentage of the input signal. αGFP (green fluorescent protein) antibody is used as control, unrelated antibody. Results are mean ± SEM of 3–4 independent experiments, each repeated in triplicate. For mRNA and luciferase analysis, mean control values are set to 1. ChIP data are mean ± SEM representative of 2 separate experiments. P values are reported for comparisons (A and B) between control and treated cells, (C–E) between indicated groups, or between αGFP and specific antibodies. *P < .05, **P < .01, ***P < .001.

mRNA and luciferase analysis, mean control values are set to 1. ChIP data are mean ± SEM representative of 2 separate experiments. P values are reported for comparisons (A and B) between control and treated cells, (C–E) between indicated groups, or between αGFP and specific antibodies. *P < .05, **P < .01, ***P < .001. Figure 5 Starvation fails to induce hepcidin gene expression in Creb3l3-/- mice. Eight- to 10-week-old wild type (WT) or Creb3l3-/- male mice were starved for 24 or 48 hours before death. (A) Pck1 mRNA and (B) Hamp mRNA expression were assessed by real-time qRT-PCR. (C) Enzyme-linked immunosorbent assay quantification of serum hepcidin levels and (D) serum iron levels. (E) Ppargc1a mRNA expression in starved mice. Results are mean ± SEM of 6–8 mice per group. (A, B, and E) Mean control values for the fed mice group are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between fed and 24- or 48-hour fasted mice, within each genotype. *P < .05, **P < .01, ***P < .001. Table 1 In Vivo Effects of Starvation on Biochemical Parameters in Mice

Figure 5 Starvation fails to induce hepcidin gene expression in Creb3l3-/- mice. Eight- to 10-week-old wild type (WT) or Creb3l3-/- male mice were starved for 24 or 48 hours before death. (A) Pck1 mRNA and (B) Hamp mRNA expression were assessed by real-time qRT-PCR. (C) Enzyme-linked immunosorbent assay quantification of serum hepcidin levels and (D) serum iron levels. (E) Ppargc1a mRNA expression in starved mice. Results are mean ± SEM of 6–8 mice per group. (A, B, and E) Mean control values for the fed mice group are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between fed and 24- or 48-hour fasted mice, within each genotype. *P < .05, **P < .01, ***P < .001. Table 1 In Vivo Effects of Starvation on Biochemical Parameters in Mice Fed 2-h Fast 5-h Fast 16-h Fast 24-h Fast 48-h Fast Serum glucose, mg/dL 249.7 ± 36.83 276.4 ± 29.31 188.0 ± 9.09a 193.2 ± 13.61a 146.7 ± 39.11b 180.7 ± 38.53b Serum iron, μg/dL 136.3 ± 3.33 143.6 ± 2.85 138.0 ± 8.12 84.00 ± 2.96b 86.62 ± 3.06b 87.08 ± 4.41b Serum ferritin, ng/mL 204.6 ± 8.78 190.0 ± 25.61 193.5 ± 10.78 286.5 ± 32.53 240.0 ± 12.58 277.0 ± 13.58a Spleen iron, μg/gdry weight 776.1 ± 29.31 837.5 ± 64.33 706.7 ± 27.62 905.2 ± 13.66 852.0 ± 56.25 1418 ± 58.28b Liver iron, μg/gdry weight 224.3 ± 7.39 226.4 ± 9.42 214.9 ± 11.12 231.3 ± 19.73 262.8 ± 16.56 367.1 ± 26.99b NOTE. Serum glucose, serum iron, serum ferritin, and tissue iron levels were analyzed in spleen and liver tissue of wild-type starving and nonstarving mice. Results are mean ± SEM of 6–8 mice per group. P values are reported for comparisons between fed and each time-point fasted mice.

19.73 262.8 ± 16.56 367.1 ± 26.99b NOTE. Serum glucose, serum iron, serum ferritin, and tissue iron levels were analyzed in spleen and liver tissue of wild-type starving and nonstarving mice. Results are mean ± SEM of 6–8 mice per group. P values are reported for comparisons between fed and each time-point fasted mice. a P < .01. b P < .001. Table 2 In Vivo Effects of Starvation on Glucose and Ketone Body Status in Creb3l3 Null Mice WT Creb3l3 -/- Fed 24-h Fast 48-h Fast Fed 24-h Fast 48-h Fast Serum glucose, mg/dL 243.7 ± 8.32 146.2 ± 26.91a 102.2 ± 12.78a 237.6 ± 8.26 95.2 ± 9.24a 64.8 ± 9.41a Serum ketone bodies, mmol/L 0.54 ± 0.13 1.68 ± 0.10a 1.88 ± 0.24a 0.39 ± 0.02 1.85 ± 0.17a 2.20 ± 0.17a NOTE. Serum glucose and ketone bodies levels were analyzed in WT and Creb3l3 null mice starved for 24 or 48 hours before sacrifice. Results are mean ± SEM of 6–8 mice per group. P value are reported for comparisons between fed and 24- or 48-hour fasted mice, within each genotype. a P < .001.

Inflammatory bowel disease (IBD), comprising Crohn’s disease (CD) and ulcerative colitis (UC), is an increasingly common immune-mediated disease of the gut of unknown cause.1,2 The genetic architecture of IBD is complex, with more than 130 significantly associated susceptibility loci identified to date,3 indicating that multiple mechanisms of disease may exist. Nevertheless, prominent roles for innate immunity and particular immune response pathways, including the interleukin (IL) 23/IL17 axis, strongly are implicated. Innate lymphoid cells (ILCs) are emerging as important players in mucosal immunity. Although recognized to perform protective roles against mucosal pathogens,4,5 they also contribute to chronic intestinal inflammation, which is particularly apparent in mice lacking conventional T and B cells.6,7 This is in part dependent on their capacity to produce inflammatory cytokines, including interferon-γ, IL17A, and IL22.4–8 ILCs can be subdivided into discrete populations, which accumulate in mucosal tissues in different pathologic settings.9 At least 3 subsets exist, including ILC1s, which produce interferon-γ; ILC2s, which produce IL5/IL13; and ILC3, which can be subdivided further based on differential expression of natural cytotoxicity receptors (NCRs), CD4, and production of IL17 and/or IL22.9

ulate in mucosal tissues in different pathologic settings.9 At least 3 subsets exist, including ILC1s, which produce interferon-γ; ILC2s, which produce IL5/IL13; and ILC3, which can be subdivided further based on differential expression of natural cytotoxicity receptors (NCRs), CD4, and production of IL17 and/or IL22.9 Tbx21-/- Rag2-/- ulcerative colitis (TRUC) mice spontaneously develop severe colitis with striking similarities to some aspects of human UC.10 Colon lesions histologically resemble UC with goblet cell depletion, crypt abscess formation, epithelial hyperplasia, and infiltration of colonic lamina propria with neutrophils and mononuclear cells.7,10 TRUC mice develop inflammation-associated epithelial dysplasia, which frequently progresses to frank adenocarcinoma,11 one of the most severe complications in human forms of IBD. TRUC disease is dependent on interactions between intestinal CD11c+ mononuclear phagocytes and CD90+ IL7R-receptor–positive (IL7R+) ILCs.7 Depletion of ILCs or genetic deficiency of the common γ-chain cytokine receptor, which is necessary for ILC survival, prevents disease.7 Similarly, blockade of IL23 or IL17A significantly attenuates disease.7 ILCs accumulate in gut lesions from IBD patients12–14 and it has been speculated that targeting these cells might represent a viable therapeutic approach in IBD.15 IL236,7 and IL1β16 contribute to ILC activation, although curiously TRUC mice that additionally are deficient for either IL23R or IL1R are not fully protected from colitis,17 consistent with a possible role for alternative ILC activation pathways contributing to disease. The purpose of this study was to investigate the proximal signals responsible for driving intestinal ILC activation and to determine whether similar pathways might exist in human disease.

not fully protected from colitis,17 consistent with a possible role for alternative ILC activation pathways contributing to disease. The purpose of this study was to investigate the proximal signals responsible for driving intestinal ILC activation and to determine whether similar pathways might exist in human disease. Materials and Methods Mice Balb/C Rag2-/- and wild-type mice were sourced commercially (Jackson Laboratories; Bar Harbor, ME). TRUC mice were a gift from Laurie Glimcher. Animal experiments were performed in accredited facilities in accordance with the UK Animals (Scientific Procedures) Act 1986 (Home Office License Number PPL: 70/6792 and PPL: 70/7869 from November 2013). Human Studies Studies in human tissues received ethical approval from the City and Hackney Local Research Ethics Committee (REC reference: 10/H0704/74 and 10/H0804/65). Colonic lamina propria mononuclear cells (cLPMCs) were isolated as described previously18 from colectomy specimens and endoscopically acquired biopsy specimens. Normal colonic mucosal samples were collected from macroscopically unaffected areas of patients undergoing intestinal resection for colon cancer or polyps. Informed written consent was obtained in all cases.

) were isolated as described previously18 from colectomy specimens and endoscopically acquired biopsy specimens. Normal colonic mucosal samples were collected from macroscopically unaffected areas of patients undergoing intestinal resection for colon cancer or polyps. Informed written consent was obtained in all cases. Flow Cytometry and Cell Sorting Intracellular cytokine expression was measured as described previously.7 Cells were stimulated with IL23 (10–20 ng/mL), IL6 (10–100 ng/mL), or phorbol 12-myristate 13-acetate (PMA) (50 ng/mL) and ionomycin (1 μmol/L) for 4–6 hours at 37°C with monensin (3 μmol/L) added for the last 2 hours. In human work, antibodies used to stain cell surface antigens were incubated with unstimulated cells for 25 minutes and then fixed in 2% paraformaldehyde pending analysis. For fluorescence-activated cell sorter purification of murine ILCs, CD45+ cells first were sorted immunomagnetically from unfractionated cLPMCs using anti-CD45 beads (Miltenyi) and LS columns (Miltenyi). CD45+ cells were stained with CD90, NKp46, and IL7R. Antibodies used in flow cytometry experiments are listed in Supplementary Table 1. Ex Vivo Organ Culture Colon explants cultures from murine experiments were performed as described previously.7 Three biopsy punches from the distal colon were cultured in 500 μL of complete medium for 24 hours at 37°C. In human studies explant cultures were set up as described previously.18 Cytokine production in culture supernatants was measured by enzyme-linked immunosorbent assay (ELISA).

performed as described previously.7 Three biopsy punches from the distal colon were cultured in 500 μL of complete medium for 24 hours at 37°C. In human studies explant cultures were set up as described previously.18 Cytokine production in culture supernatants was measured by enzyme-linked immunosorbent assay (ELISA). Cell Culture Unfractionated murine splenocytes (2 × 106/mL) and mesenteric lymph node (mLN) cells (1 × 106/mL) or cLPMCs (1 × 106/mL) were cultured in complete medium for 24 hours at 37°C as described previously.7 cLPMCs from IBD and noninflammatory control patients were cultured with recombinant human IL6 (R&D) (0–100 ng/mL) overnight at 37°C, 5% CO2, and then restimulated with PMA (50 ng/mL) and ionomycin (1 μmol/L). In some experiments, cLPMCs were cultured with IL6 (100 ng/mL) for 6 hours in the presence of monensin (3 μmol/L). Fluorescence-activated cell sorter–purified NCR- ILC3s (CD45+ CD90+ IL7R+ NKp46-) from TRUC mice were cultured at 5 × 104/mL for 24 hours. Cytokine concentrations in culture supernatants were measured by ELISA (R&D Systems and eBioscience). Histology Colon histology was processed, stained (H&E), and colitis scores were calculated as described previously.7 Proximal and distal colitis scores from individual mice were averaged, unless otherwise stated. ELISA and Cytokine Bead Arrays Cytokine concentrations were measured in culture supernatants by ELISA or T-helper cell (Th)1, Th2, or Th17 CBA (BD Biosciences).

Histology Colon histology was processed, stained (H&E), and colitis scores were calculated as described previously.7 Proximal and distal colitis scores from individual mice were averaged, unless otherwise stated. ELISA and Cytokine Bead Arrays Cytokine concentrations were measured in culture supernatants by ELISA or T-helper cell (Th)1, Th2, or Th17 CBA (BD Biosciences). Microarray and Real-Time Polymerase Chain Reaction RNA was extracted from 3 Rag2-/- and 3 TRUC mice, aged 10 weeks, using TRIzol reagent (Invitrogen Carlsbad, CA). Transcript expression was analyzed with Mouse Genome 430 2.0 Affymetrix Expression Array. For real-time polymerase chain reaction (PCR) experiments cells were lysed in TRIzol reagent (Invitrogen) and RNA was extracted. Complementary DNA was generated with the complementary DNA synthesis kit (Bioline, Taunton, MA). Quantitative PCR was used to quantify messenger RNA transcripts using TaqMan gene expression assays (Applied Biosystems). Gene expression was normalized to the expression of β-actin to generate ΔCT values and relative abundance was quantified using the 2-ΔCT method. Human RORC (Hs01076112_m1) and β-actin (4326315E) TaqMan quantitative PCR primer sets were used. In Vivo Antibody Treatment Intraperitoneal injections of anti-CD4 (1 mg, GK1.5), anti-CD90 (1 mg, 30H12), anti-IL6 (750 μg, MP5-20F3), or isotype-matched control antibodies (LTF-2 or HRPN) (Bio X Cell, West Lebanon, NH) were administered on days 0, 7, 14, 21, and 28 (anti-CD4, anti-CD90) or days 0, 4, 9, 14, 18, 23, and 27 (anti-IL6).

Treatment Intraperitoneal injections of anti-CD4 (1 mg, GK1.5), anti-CD90 (1 mg, 30H12), anti-IL6 (750 μg, MP5-20F3), or isotype-matched control antibodies (LTF-2 or HRPN) (Bio X Cell, West Lebanon, NH) were administered on days 0, 7, 14, 21, and 28 (anti-CD4, anti-CD90) or days 0, 4, 9, 14, 18, 23, and 27 (anti-IL6). Microbiota Analysis See the Supplementary Materials and Methods section for more detail.

Treatment Intraperitoneal injections of anti-CD4 (1 mg, GK1.5), anti-CD90 (1 mg, 30H12), anti-IL6 (750 μg, MP5-20F3), or isotype-matched control antibodies (LTF-2 or HRPN) (Bio X Cell, West Lebanon, NH) were administered on days 0, 7, 14, 21, and 28 (anti-CD4, anti-CD90) or days 0, 4, 9, 14, 18, 23, and 27 (anti-IL6). Microbiota Analysis See the Supplementary Materials and Methods section for more detail. Results NCR- CD4- ILC3 Cells Are the Predominant Colonic ILC Subset in Chronic Intestinal Inflammation in TRUC Mice We validated the phenotype of ILCs in TRUC mice, confirming excessive accumulation of IL17A- and IL22-producing CD90+ IL7R+ NCR- ILC3 in diseased colons (Figure 1A and 1B, Supplementary Figure 1A–D). CD4-expressing NCR- ILC3s resembling lymphoid tissue inducer cells participate in mucosal immune responses in the gut,5 therefore, we considered the possibility that CD4+ ILCs might be the NCR- ILC3 subset responsible for mediating chronic inflammation in TRUC mice. CD4+ cells were present in mLNs of TRUC mice (many of which co-expressed CD90); however, very few CD4+ cells were present in the colon (Figure 1C). Given the low frequency of intestinal CD4+ ILCs in TRUC mice we considered it unlikely that these cells would play a major role in disease. To test this assumption we depleted CD4-expressing cells in vivo. The administration of anti-CD4 antibodies successfully depleted CD4-expressing cells in mLNs and colon of TRUC mice (Figure 1C). However, many CD90+ cells still remained in the colon and there was no reduction in the number of IL17A- or IL22-producing cells (Figure 1D). Depleting anti-CD4 treatment did not alter the severity of TRUC disease significantly (Figure 1E). In contrast, anti-CD90 treatment depleted both CD90- and CD4-expressing ILCs, reduced the number of IL17- and IL22-producing cells in the colon, and significantly attenuated disease (Figure 1C–E). Taken together, these data indicate IL17A/IL22 producing CD90+ IL7R+ NCR- CD4- ILC3 are the key ILC population in the colon responsible for causing disease in TRUC mice.

CD90- and CD4-expressing ILCs, reduced the number of IL17- and IL22-producing cells in the colon, and significantly attenuated disease (Figure 1C–E). Taken together, these data indicate IL17A/IL22 producing CD90+ IL7R+ NCR- CD4- ILC3 are the key ILC population in the colon responsible for causing disease in TRUC mice. IL6 Is Expressed Highly and Augments Pathogenic Cytokine Production in TRUC Mice We sought to define the proximal immune signals responsible for triggering effector function of colonic NCR- ILC3 in TRUC mice. IL1β and IL6 were among the most highly expressed (>2-fold induction) cytokine transcripts in the colon of TRUC mice in comparison with Rag2-/- mice (Figure 2A). The other IL1 family member, Il1a, and the IL23 subunit transcripts (Il23a and Il12b) also were increased. Proximal cytokines responsible for driving ILC1 (IL12, IL15, and IL18) or ILC2 (IL25 and IL33) responses were not up-regulated, and, indeed in most instances were down-regulated in the colon of TRUC mice in comparison with Rag2-/- controls.

, Il1a, and the IL23 subunit transcripts (Il23a and Il12b) also were increased. Proximal cytokines responsible for driving ILC1 (IL12, IL15, and IL18) or ILC2 (IL25 and IL33) responses were not up-regulated, and, indeed in most instances were down-regulated in the colon of TRUC mice in comparison with Rag2-/- controls. IL23 and IL1 have been described to play an important role triggering ILCs in TRUC disease, therefore, in this study we focussed our attention on IL6. In addition to increased Il6 transcripts in the colon, there were very high concentrations of IL6 in serum and significantly increased production of IL6 in colon explant cultures from TRUC mice (Supplementary Figure 2A). Transcripts of genes known to be regulated by IL619 were up-regulated in the colon of TRUC mice in comparison with Rag2-/- controls (Supplementary Figure 2B), including well-recognized immune genes (Socs1, Socs3, and Icam1), IL6 signaling components (Stat1 and Stat3), and anti-apoptotic genes (Bcl3, Bcl6, and Bcl-x1). The most highly expressed IL6-regulated gene in the colon of TRUC mice (12-fold enrichment) was Pou2af1, which encodes a transcriptional co-activator responsible for IL6-mediated regulation of IL17 responses in T cells.20

1), IL6 signaling components (Stat1 and Stat3), and anti-apoptotic genes (Bcl3, Bcl6, and Bcl-x1). The most highly expressed IL6-regulated gene in the colon of TRUC mice (12-fold enrichment) was Pou2af1, which encodes a transcriptional co-activator responsible for IL6-mediated regulation of IL17 responses in T cells.20 To determine whether IL6 might trigger ILC-derived cytokines we stimulated unfractionated cLPMCs and mLN cells from TRUC mice with recombinant IL6. Strikingly, IL6 triggered IL17A production by both cLPMCs and mLN cells (Figure 2B). We also performed flow cytometry with intracellular cytokine staining after IL6 stimulation of unfractionated mLN cells. Although less potent than IL23, IL6 induced expression of IL17A in NCR- ILC3s (Figure 2C). To determine whether this was a cell-intrinsic phenomenon we purified CD90+ IL7R+ NCR- ILC3s from TRUC colons by fluorescence-activated cell sorting (Supplementary Figure 3A). To our surprise, neither IL6, IL23, nor IL1α by themselves induced significant cytokine production by purified colonic NCR- ILC3s (Figure 2D). However, the combination of IL23 and IL1α was a potent trigger for ILC production of IL17A and IL22. The addition of IL6 together with IL23 and IL1α was the most potent trigger of all. Purified intestinal NCR- ILCs from TRUC mice produced little tumor necrosis factor α or interferon-γ under these conditions (Supplementary Figure 3B). IL23 and IL1α were weak inducers of IL6 by colonic NCR- ILCs (Supplementary Figure 3B). Taken together, these data showed that IL6 augments IL23/IL1α-induced pathogenic cytokine production by intestinal ILCs in TRUC mice in a cell-intrinsic manner.

tor α or interferon-γ under these conditions (Supplementary Figure 3B). IL23 and IL1α were weak inducers of IL6 by colonic NCR- ILCs (Supplementary Figure 3B). Taken together, these data showed that IL6 augments IL23/IL1α-induced pathogenic cytokine production by intestinal ILCs in TRUC mice in a cell-intrinsic manner. IL6 signals through a heterodimeric receptor comprising ubiquitously expressed gp130 and selectively expressed IL6Rα. However, IL6Rα also exists as a soluble form, which can complex with IL6 in solution and then bind to cells expressing gp130, enabling cells, which do not usually express IL6Rα, to respond to IL6 stimulation. Therefore, we investigated IL6Rα and soluble IL6Rα (sIL6Rα) expression in TRUC mice. IL6Rα expression by ILCs was highly variable in the colon of TRUC mice, but typically was less than 10% (Supplementary Figure 4A, and data not shown). However, sIL6Rα was abundant in the serum of TRUC mice and was detected in supernatants from cultured colon explants and unfractionated splenocytes (Supplementary Figure 4B). Therefore, it is likely that ILCs respond to IL6 stimulation directly, but also potentially through trans-signaling given the abundance of sIL6R in TRUC mice.

Rα was abundant in the serum of TRUC mice and was detected in supernatants from cultured colon explants and unfractionated splenocytes (Supplementary Figure 4B). Therefore, it is likely that ILCs respond to IL6 stimulation directly, but also potentially through trans-signaling given the abundance of sIL6R in TRUC mice. IL6 Blockade Attenuated TRUC Disease Independently of Changes to Intestinal Microbiota Community Profiles To determine whether IL6-mediated activation of innate immunity was functionally important in TRUC disease, mice were treated with monoclonal antibodies that neutralize the biological activity of IL6. Treatment with anti-IL6 resulted in loss of IL6 bioavailability (Supplementary Figure 5A). IL17A production by unfractionated cLPMCs and splenocytes was reduced significantly in anti-IL6–treated TRUC mice, although was not abolished completely (Figure 3A). IL6 neutralization significantly attenuated TRUC disease, including reduced colitis scores and reduced splenomegaly (Figure 3B and C).

mentary Figure 5A). IL17A production by unfractionated cLPMCs and splenocytes was reduced significantly in anti-IL6–treated TRUC mice, although was not abolished completely (Figure 3A). IL6 neutralization significantly attenuated TRUC disease, including reduced colitis scores and reduced splenomegaly (Figure 3B and C). Similar to the situation in human IBD, TRUC disease is associated with perturbation of intestinal microbial communities. Because IL6 directly influences the success of mucosal colonization by some intestinal bacteria,21 we considered the possibility that attenuation of chronic TRUC disease after IL6 blockade might have occurred secondarily to changes in key components in the composition of the intestinal microbiota. To address this question we sequenced 16S ribosomal RNA genes that were PCR-amplified from fecal samples from anti-IL6 or control antibody–treated TRUC mice. Overall, we identified 2642 different operational taxonomic units (OTUs). Treatment with anti-IL6 antibody appeared to have a relatively minor impact on the microbiota (Figure 4A). At the phylum level, Firmicutes were reduced slightly in proportional abundance in anti-IL6–treated mice (P = .035) (Figure 4A) but, at finer taxonomic levels, anti-IL6 treatment did not impact the proportional abundance of the most common 150 OTUs significantly, which cumulatively accounted for more than 96% of the total amount of sequence data generated (Supplementary Table 2). Cluster analysis, using the Bray Curtis calculator, confirmed that there was no signature microbiota profile associated with anti-IL6 treatment (Supplementary Figure 5B). Helicobacter typhlonius was ubiquitously present in anti-IL6– or isotype control–treated mice, however, the proportional abundance did not differ significantly between the 2 groups either before or after treatment (Figure 4B). There was a tendency for increased bacterial diversity in the gut of anti-IL6–treated mice in comparison with control antibody–treated mice, although this did not achieve statistical significance (P < .095) (Supplementary Figure 5C).

iffer significantly between the 2 groups either before or after treatment (Figure 4B). There was a tendency for increased bacterial diversity in the gut of anti-IL6–treated mice in comparison with control antibody–treated mice, although this did not achieve statistical significance (P < .095) (Supplementary Figure 5C). IL6 Augments Pathogenic Cytokine Production by Colonic CD3- IL7R+ Cells From IBD Patients Our preclinical data support the possibility that targeting IL6 may be therapeutically tractable in chronic gut inflammation. Therefore, we aimed to verify whether this pathway was relevant in human disease. As expected, stimulation of unfractionated intestinal immune cells with PMA and ionomycin resulted in production of pathogenic cytokines, including IL17A, IL22, and interferon-γ by CD3+ T cells (Figure 5A and Supplementary Figure 6A). However, we also observed production of these cytokines in the non–T-cell (CD3-) fraction, particularly in IBD patients. Within the non–T-cell fraction (CD3-), we could identify a population of IL7R-expressing cells in the colon of patients with CD, UC, and noninflammatory control patients (Figure 5B). Although the frequency of these cells was variable, their proportional abundance within the lymphocyte population was increased in IBD patients in comparison with noninflammatory controls (Figure 5B). Consistent with ILC3s being present among the CD3- IL7R+ population, there was enriched expression of RORγt and c-kit (CD117) (Figure 5C and Supplementary Figure 6C). RORC transcripts also were enriched in fluorescence-activated cell sorter–purified CD3- IL7R+ cells analyzed by real-time PCR (Supplementary Figure 6B), corroborating the likelihood of ILCs being present in the CD3- IL7R+ population. Analysis of the CD3- IL7R+ population according to NCR expression showed the presence of 3 discrete populations, comprising NKp46+ NKp44- cells, NKp44+ NKp46- cells, and NCR- (NKp44- NKp46-) cells (Figure 5D and Supplementary Figure 6C), indicating that NCR+ and NCR- ILCs are present within this CD3- IL7R+ population. In most patients, including IBD and noninflammatory control patients, CD3- IL7R+ NKp44+ NKp46- cells were the predominant subset present (Figure 5D and Supplementary Figure 6C).

Kp44- NKp46-) cells (Figure 5D and Supplementary Figure 6C), indicating that NCR+ and NCR- ILCs are present within this CD3- IL7R+ population. In most patients, including IBD and noninflammatory control patients, CD3- IL7R+ NKp44+ NKp46- cells were the predominant subset present (Figure 5D and Supplementary Figure 6C). To determine whether CD3- IL7R+ cells present in diseased mucosa of IBD patients were responsive to IL6, cLPMCs were incubated overnight with recombinant human IL6 before being restimulated with PMA and ionomycin. Production of IL17A, IL22, and interferon-γ by CD3- IL7R+ cells was increased significantly when cLPMCs were cultured in the presence of IL6 (Figure 6A and B). In addition, some samples were stimulated with IL6 directly without mitogen, which showed induction of IL17A by CD3- IL7R+ cells in a dose-dependent manner (Figure 6C).

on of IL17A, IL22, and interferon-γ by CD3- IL7R+ cells was increased significantly when cLPMCs were cultured in the presence of IL6 (Figure 6A and B). In addition, some samples were stimulated with IL6 directly without mitogen, which showed induction of IL17A by CD3- IL7R+ cells in a dose-dependent manner (Figure 6C). Finally, we analyzed IL6 production by diseased mucosa from IBD patients to see whether blocking IL6 might be a reasonable therapeutic strategy in some or all IBD patients. IL6 was produced by colon explant cultures in CD, UC, and noninflammatory control patients (Supplementary Figure 6D). However, IL6 production was variable, especially in IBD patients, ranging from 72.2 to 8426.4 pg/mg colonic tissue. IBD patients could be stratified according to mucosal production of IL6, with half of IBD patients producing relatively low levels comparable with noninflammatory control patients and the other half producing high amounts (>1000 pg/mg tissue). Taken together, these data indicate that IL6, which is produced in very high quantities in approximately 50% of CD and UC patients drives pathologically relevant immune pathways in chronic intestinal inflammation.

able with noninflammatory control patients and the other half producing high amounts (>1000 pg/mg tissue). Taken together, these data indicate that IL6, which is produced in very high quantities in approximately 50% of CD and UC patients drives pathologically relevant immune pathways in chronic intestinal inflammation. Discussion CD4- NCR- ILC3s are the predominant CD90+ IL7R+ ILC population in the colon of TRUC mice responsible for causing disease. Purified intestinal NCR- ILC3 from TRUC mice produced IL17A and IL22, but were poor producers of interferon-γ and tumor necrosis factor α. They were also a modest source of IL6. Few NCR+ ILCs were present in the colon of TRUC mice, consistent with data from other groups reporting a requirement for T-bet in NKp46+ ILC development and differentiation.22 Intestinal CD4+ ILCs are important in host resistance to intestinal pathogens, such as Citrobacter rodentium.5 Here, we show that CD4+ ILCs, which are abundant in mLN, but infrequent in the colon, do not play a major role in TRUC disease. Depletion of CD4+ ILCs had no impact on pathogenic cytokine production or disease outcome. In TRUC mice, highly purified colonic NCR- ILC3 did not respond to IL23 or IL1α in isolation. Instead, combinations of IL23 together with IL1α were required for production of effector cytokines by ILC3. Furthermore, additional exposure to IL6 was required for optimal IL17A and IL22 production, showing a novel role for IL6 in the innate immune system in chronic intestinal inflammation.

IL23 or IL1α in isolation. Instead, combinations of IL23 together with IL1α were required for production of effector cytokines by ILC3. Furthermore, additional exposure to IL6 was required for optimal IL17A and IL22 production, showing a novel role for IL6 in the innate immune system in chronic intestinal inflammation. IL17A-, IL22-, and interferon-γ–producing CD3- IL7R+ cells also were identified in the colonic lamina propria of patients with IBD. Although this population is heterogeneous, there was enrichment of RORγt and c-kit, confirming the likelihood that ILC3 were present within this compartment. Most CD3- IL7R+ cells were NKp44+, although NKp46+ and NCR- (NKp44- NKp46-) cells also were present. These data are broadly consistent with previous reports of ILC populations in human gut.12–14 Crucially, IL6 increased pathogenic cytokine production by CD3- IL7R+ cLPMCs from IBD patients in a dose-dependent manner, consistent with our preclinical data showing IL6-responsive colonic ILC3s.

6-) cells also were present. These data are broadly consistent with previous reports of ILC populations in human gut.12–14 Crucially, IL6 increased pathogenic cytokine production by CD3- IL7R+ cLPMCs from IBD patients in a dose-dependent manner, consistent with our preclinical data showing IL6-responsive colonic ILC3s. IL6 is a pleiotropic cytokine that may be important in IBD. Peripheral blood and cLPMCs produce excess IL6 in IBD,23,24 often at levels correlating with disease activity.25 Genetic variation at the IL6 locus is linked with early onset IBD26 and polymorphisms at loci encoding IL6R signaling components are associated with increased IBD risk.3 IL6 blockade is therapeutic in some preclinical models of IBD, although it has been assumed that the therapeutic mechanism likely was attributable to limitation of T-cell–mediated pathology27–30 because IL6 contributes to intestinal Th17 differentiation.31 We show a novel role of IL6 in innate immune-mediated chronic intestinal pathology. It is interesting that cytokines contributing to CD4+ Th17 differentiation, including IL1, IL23, and IL6, have conserved roles promoting innate IL17 production. Our data build on other work implicating ILCs as potentially important mediators in IBD,12–14 and confirm NCR- ILC3 as a source of pathogenic cytokines in IBD. Polymorphisms at multiple susceptibility loci in IBD that previously were considered to impact adaptive immunity, similarly could impact ILC phenotype, including RORC, IL23R, IL12RB2, IL12B, IL22, IFNG, STAT1, STAT3, STAT4, CCR6, IL1R1, IL15RA, and IL6ST.3 Accordingly, it is possible that genetic variation at these loci in IBD could impact disease susceptibility by altering the activation and effector function of mucosal ILCs. However, the relative contribution of ILC to the initiation and propagation of chronic intestinal inflammation in IBD remains to be determined. Polyclonal stimulus of unfractionated cLPMCs from IBD patients showed that most cytokine-expressing cells reside within the CD3+ cell fraction. It should be remembered that cytokine responses induced by polyclonal stimuli may overestimate T-cell contribution because under physiological conditions few of these tissue-trafficking T cells would be encountering their relevant antigen, so would unlikely be triggered to produce cytokine. By contrast, despite their numeric inferiority to T cells, mucosal ILC are likely to be activated directly by cytokine signals abundant in chronically inflamed tissue, such as IL6.

s few of these tissue-trafficking T cells would be encountering their relevant antigen, so would unlikely be triggered to produce cytokine. By contrast, despite their numeric inferiority to T cells, mucosal ILC are likely to be activated directly by cytokine signals abundant in chronically inflamed tissue, such as IL6. IL6-induced stimulation of ILC effector function may prove to be especially pertinent in UC because IL23, the canonical ILC-activating cytokine, is produced at low levels in UC in comparison with CD.32 In this study, IL6 neutralization reduced innate production of IL17A in TRUC mice and significantly attenuated disease severity, although the magnitude of impact was less than seen with ILC depletion or IL23 blockade.7 This is in keeping with our observation that although IL6 is required for optimal activation of ILC effector function, other proximal cytokine signals, including IL23 and/or IL1 stimulation, additionally are required. IL6 blockade had a minimal impact on the intestinal microbiota, other than a minor shift in the proportional abundance of Firmicutes and a tendency for increased intestinal bacterial diversity. It is possible that this latter change occurred secondary to reduced intestinal inflammation in anti-IL6–treated mice. Indeed, IBD activity/severity is recognized to correlate inversely with bacterial diversity in the gut.33,34

rtional abundance of Firmicutes and a tendency for increased intestinal bacterial diversity. It is possible that this latter change occurred secondary to reduced intestinal inflammation in anti-IL6–treated mice. Indeed, IBD activity/severity is recognized to correlate inversely with bacterial diversity in the gut.33,34 Our data support extending biological therapies targeting IL6 in IBD. The IL6R blocking antibody tocilizumab is efficacious in other inflammatory diseases, including arthritis35,36 and lupus,37 and a pilot study in CD showed promising initial results.38 In this study mucosal IL6 production was highly variable, however, only half of IBD patients produced more IL6 than non-IBD control patients. Similarly, the frequency of CD3- IL7R+ cells, and the magnitude of IL6-induced cytokine responses by these cells also markedly was variable. With the promise of personalized medicine on the horizon,39 it is tempting to speculate that treatment strategies targeting IL6 might be favored in patient subsets defined by high mucosal expression of IL6 and/or high frequencies of IL6 responsive effector cells in diseased tissue. In summary, we have shown that IL6 augments pathogenic cytokine production by intestinal ILCs in chronic intestinal inflammation and that this pathway may be operational in human IBD. Novel therapeutic strategies targeting ILC or their proximal cytokine signals may offer a new treatment paradigm in IBD.

Our data support extending biological therapies targeting IL6 in IBD. The IL6R blocking antibody tocilizumab is efficacious in other inflammatory diseases, including arthritis35,36 and lupus,37 and a pilot study in CD showed promising initial results.38 In this study mucosal IL6 production was highly variable, however, only half of IBD patients produced more IL6 than non-IBD control patients. Similarly, the frequency of CD3- IL7R+ cells, and the magnitude of IL6-induced cytokine responses by these cells also markedly was variable. With the promise of personalized medicine on the horizon,39 it is tempting to speculate that treatment strategies targeting IL6 might be favored in patient subsets defined by high mucosal expression of IL6 and/or high frequencies of IL6 responsive effector cells in diseased tissue. In summary, we have shown that IL6 augments pathogenic cytokine production by intestinal ILCs in chronic intestinal inflammation and that this pathway may be operational in human IBD. Novel therapeutic strategies targeting ILC or their proximal cytokine signals may offer a new treatment paradigm in IBD. Supplementary Materials and Methods Microbiota Analysis DNA was extracted from mouse fecal samples using the FastDNA SPIN Kit for Soil and a FastPrep 24 machine (MP Biomedicals Santa Ana, CA) according to the protocol provided by the manufacturer. Bacterial 16S ribosomal RNA genes were PCR-amplified using barcoded primers MiSeq 27F (5’AATGATACGGCGACCACCGAGATCTACACTATGGTAATTCCAGMGTTYGATYMTGGCTCAG-3’) and MiSeq-338R (5’-CAAGCAGAAGACGGCATACGAGAT-barcode-AGTCAGTCAGAAGCTGCCTCCCGTAGGAGT-3’), which target variable regions V1–V2 of the 16S ribosomal RNA gene. Q5 Taq polymerase (New England Biolabs, Ipswich, MA) was used for the PCR step, and 4 PCR reactions were performed per sample. Cycling conditions were as follows: 98°C for 2 minutes, followed by 20 cycles of 98°C for 30 seconds, 50°C for 30 seconds, 72°C for 90 seconds, and then a final extension step at 72°C for 5 minutes.

q polymerase (New England Biolabs, Ipswich, MA) was used for the PCR step, and 4 PCR reactions were performed per sample. Cycling conditions were as follows: 98°C for 2 minutes, followed by 20 cycles of 98°C for 30 seconds, 50°C for 30 seconds, 72°C for 90 seconds, and then a final extension step at 72°C for 5 minutes. The 4 PCR reactions from each DNA extraction then were pooled and concentrated down to 25 μL volumes per sample. PCR amplicons then were quantified using a Qubit 2.0 Fluorometer (Life Technologies, Ltd, Carlsbad, CA) and equimolar concentrations of each were added to a final mastermix for sequencing, using an Illumina MiSeq (San Diego, CA) machine with 2 × 250 bp read length. Sequence data have been deposited in the European Nucleotide Archive under Study Accession Number ERP005850 and Sample Accession numbers ERS459682–ERS459702.

and equimolar concentrations of each were added to a final mastermix for sequencing, using an Illumina MiSeq (San Diego, CA) machine with 2 × 250 bp read length. Sequence data have been deposited in the European Nucleotide Archive under Study Accession Number ERP005850 and Sample Accession numbers ERS459682–ERS459702. The sequence data were processed by following the MiSeq SOP of the mothur software package (Available at: http://www.mothur.org/wiki/MiSeq_SOP).1 Paired-read contigs were created from the forward and reverse read sequence data, and preliminary quality processing was performed by removing all contigs that were shorter than 260 bp, longer than 450 bp, or those that contained any ambiguous bases or homopolymeric stretches longer than 7 bases. Perseus,2 as implemented in mothur, was used to remove putative chimeras, and any reads mapping to chloroplasts, mitochondria, eukarya, or archaea also were removed. After these steps, more than 486,000 sequences were left in the final data set (range, 447–51,571 sequences per sample). Next, after a preclustering step (diffs, 3), 97% similarity OTUs were generated. Taxonomic classifications for each of these OTUs were created using the RDP taxonomy provided at the mothur web page. Metastats,3 as implemented in mothur, was used to test for significant differences in the proportional abundance of each of the phyla present, and the 150 most abundant OTUs between anti-IL6 and control mice. Bacterial diversity was measured by generating Shannon and inverse Simpson indices in mothur. Before these calculations each sample was subsampled randomly down to 447 reads to ensure equal sequencing depth for each. Kruskall–Wallis and Mann–Whitney U tests, implemented in Minitab v16, were used to test for significant differences in diversity measures between the anti-IL6 and control mouse groups. A cluster dendrogram, using the Bray Curtis calculator, also was generated in mothur from the subsampled data set, and subsequently was visualized using the iTOL online software resource.4Supplementary Table 1 Flow Cytometry Antibodies Used

ifferences in diversity measures between the anti-IL6 and control mouse groups. A cluster dendrogram, using the Bray Curtis calculator, also was generated in mothur from the subsampled data set, and subsequently was visualized using the iTOL online software resource.4Supplementary Table 1 Flow Cytometry Antibodies Used Antigen Clone Supplier Anti-mouse antibodies CD45 30-F11 eBioscience CD90.2 53-2.1 eBioscience CD127 A7R34 eBioscience NKp46 29A1.4 eBioscience CD126 D7715A7 Biolegend CCR6 140706 R&D Systems ICOS 7E.17G9 eBioscience CD4 RM4.5 eBioscience Hematopoietic lineage cocktail (CD3, CD45R, B220, CD11b, TER-119, Gr-1) 17A2, RA3-6B2, M1/70, TER-119 eBioscience CD62L MEL-14 eBioscience IL17RB 752101 R&D Systems CD69 H1.2F3 eBioscience RORγt AFKJS-9 eBioscience IL17A eBio17B7 eBioscience IL22 1H8PWSR eBioscience Interferon-γ XMG1.2 eBioscience IL4 11B11 eBioscience Anti-human antibodies CD3 UCHT-1 Biolegend CD127 A019D5 (eBioRDR5) Biolegend (eBioscience) CD117 104D2 eBioscience NKp46 9E2 eBioscience NKp44 44.189 eBioscience IL17RB I170220 R&D Systems RORγt AFKJS-9 eBioscience IL17A eBio64DEC17 eBioscience IL22 22URTI eBioscience Interferon-γ 4S.B3 eBioscience Supplementary Table 2 Metastats Comparison Between the Proportional Abundance of the Top 150 Most Abundant OTUs in the Microbiota Data Set Between Anti-IL6 and Control Mice Groups OTU no.

Antigen Clone Supplier Anti-mouse antibodies CD45 30-F11 eBioscience CD90.2 53-2.1 eBioscience CD127 A7R34 eBioscience NKp46 29A1.4 eBioscience CD126 D7715A7 Biolegend CCR6 140706 R&D Systems ICOS 7E.17G9 eBioscience CD4 RM4.5 eBioscience Hematopoietic lineage cocktail (CD3, CD45R, B220, CD11b, TER-119, Gr-1) 17A2, RA3-6B2, M1/70, TER-119 eBioscience CD62L MEL-14 eBioscience IL17RB 752101 R&D Systems CD69 H1.2F3 eBioscience RORγt AFKJS-9 eBioscience IL17A eBio17B7 eBioscience IL22 1H8PWSR eBioscience Interferon-γ XMG1.2 eBioscience IL4 11B11 eBioscience Anti-human antibodies CD3 UCHT-1 Biolegend CD127 A019D5 (eBioRDR5) Biolegend (eBioscience) CD117 104D2 eBioscience NKp46 9E2 eBioscience NKp44 44.189 eBioscience IL17RB I170220 R&D Systems RORγt AFKJS-9 eBioscience IL17A eBio64DEC17 eBioscience IL22 22URTI eBioscience Interferon-γ 4S.B3 eBioscience Supplementary Table 2 Metastats Comparison Between the Proportional Abundance of the Top 150 Most Abundant OTUs in the Microbiota Data Set Between Anti-IL6 and Control Mice Groups OTU no. RDP taxonomic classifications NCBI MegaBLAST ID P value Q value Phylum Class Order Family Genus Otu0001 Firmicutes(100) Bacilli(100) Lactobacillales(100) Lactobacillaceae(100) Lactobacillus(100) Lactobacillus animalis/murinus .287595 1 Otu0002 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) unclassified(100) Butyrivibrio species P79 (86% similarity) .022376 1 Otu0003 Firmicutes(100) Bacilli(100) Lactobacillales(100) Lactobacillaceae(100) Lactobacillus(100) Lactobacillus taiwanensis/johnsonii/acidophilus .947484 1 Otu0004 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVb(100) Clostridium lactatifermentans (90% similarity) .15343 1 Otu0005 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Eubacterium plexicaudatum (92% similarity) .82004 1 Otu0006 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Unclassified(100) Unclassified(100) Butyricimonas species JCM 18677 (80% similarity) .012182 1 Otu0007 Proteobacteria(100) Epsilonproteobacteria(100) Campylobacterales(100) Helicobacteraceae(100) Helicobacter(100) Helicobacter typhlonius .173889 1 Otu0008 Deferribacteres(100) Deferribacteres(100) Deferribacterales(100) Deferribacteraceae(100) Mucispirillum(100) Mucispirillum colimuris .063365 1 Otu0009 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) Alistipes onderdonkii (90% similarity) .874014 1 Otu0010 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) Alistipes senegalensis (91% similarity) .800781 1 Otu0011 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) Clostridium scindens (88% similarity) .766219 1 Otu0012 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Coprococcus catus (88% similarity) .571687 1 Otu0013 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) Oscillibacter valericigenes (91% similarity) .03155 1 Otu0014 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) Bacteroides acidofaciens/uniformis .082054 1 Otu0015 Fir

.571687 1 Otu0013 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) Oscillibacter valericigenes (91% similarity) .03155 1 Otu0014 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) Bacteroides acidofaciens/uniformis .082054 1 Otu0015 Fir micutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) Anaerotruncus colihominis (92% similarity) .489023 1 Otu0016 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Clostridium hathewayi (92% similarity) .861923 1 Otu0017 Firmicutes(100) Bacilli(100) Lactobacillales(100) Lactobacillaceae(100) Lactobacillus(100) Lactobacillus reuteri .644746 1 Otu0018 Proteobacteria(100) Deltaproteobacteria(100) Unclassified(100) Unclassified(100) Unclassified(100) Desulfocurvus vexinensis (87% similarity) .44467 1 Otu0019 Deferribacteres(100) Deferribacteres(100) Deferribacterales(100) Deferribacteraceae(100) Mucispirillum(100) Mucispirillum colimuris (97% similarity) .039672 1 Otu0020 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Unclassified(99) Unclassified(99) Tannerella forsythensis (82% similarity) .262652 1 Otu0021 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) Bacteroides acidifaciens .733786 1 Otu0022 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Prevotellaceae(100) Paraprevotella(100) Prevotella species (87% similarity) .137208 1 Otu0023 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(52) Unclassified(52) Coprococcus catus (84% similarity) .501202 1 Otu0024 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Pseudobutyrivibrio ruminis (88% similarity) .510185 1 Otu0025 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Paludibacter(77) Prevotella dentalis (81% similarity) .162484 1 Otu0026 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) Bacteroides uniformis (96% similarity) .195279 1 Otu0027 Firmicutes(100) Erysipelotrichia(99) Erysipelotrichales(99) Erysipelotrichaceae(99) Erysipelotrichaceae_incertae_sedis(99) Eubacterium cylindroides (88% similarity) .605205 1 Otu0028 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) Clostridium aminophilum (86% similarity) .246769 1 Otu0029 TM7(100) TM7_class_incertae_sedis(100) TM7_order_incertae_sedis(100) TM7_family_incertae_sedis

_incertae_sedis(99) Eubacterium cylindroides (88% similarity) .605205 1 Otu0028 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) Clostridium aminophilum (86% similarity) .246769 1 Otu0029 TM7(100) TM7_class_incertae_sedis(100) TM7_order_incertae_sedis(100) TM7_family_incertae_sedis (100) TM7_genus_incertae_sedis(100) TM7 phylum species (93% similarity) .222701 1 Otu0030 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Clostridium phytofermentans (90% similarity) .574396 1 Otu0031 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Clostridium celerecrescens (93% similarity) .664006 1 Otu0032 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) Clostridium scindens (84% similarity) .737231 1 Otu0033 Proteobacteria(100) Gammaproteobacteria(100) Enterobacteriales(100) Enterobacteriaceae(100) Unclassified Escherichia/Enterobacter/Citrobacter/Shigella species .497501 1 Otu0034 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) Alistipes senegalensis (92% similarity) .127302 1 Otu0035 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) Segmented filamentous bacterium .798342 1 Otu0036 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) Eubacterium oxidoreducens (87%) .334068 1 Otu0037 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .263709 1 Otu0038 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Unclassified(100) Unclassified(100) .951701 1 Otu0039 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) .178105 1 Otu0040 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Lachnospiracea_incertae_sedis(79) .987598 1 Otu0041 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .948299 1 Otu0042 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Bacteroidaceae(100) Bacteroides(100) .334148 1 Otu0043 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Dorea(93) .850598 1 Otu0044 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .622208 1 Otu0045 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .948849 1 Otu0046 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100

0) Dorea(93) .850598 1 Otu0044 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .622208 1 Otu0045 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .948849 1 Otu0046 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100 ) Alistipes(100) .211071 1 Otu0047 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVb(100) .449205 1 Otu0048 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) .101219 1 Otu0049 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Butyricicoccus(100) .597761 1 Otu0050 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Odoribacter(100) .824846 1 Otu0051 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(72) .889006 1 Otu0052 Bacteroidetes(100) Unclassified(99) Unclassified(99) Unclassified(99) Unclassified(99) .352835 1 Otu0053 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .420749 1 Otu0054 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .807154 1 Otu0055 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Roseburia(100) .549208 1 Otu0056 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) unclassified(100) .795695 1 Otu0057 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Acetanaerobacterium(100) .795544 1 Otu0058 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .594923 1 Otu0059 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .319484 1 Otu0060 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .815476 1 Otu0061 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) .245678 1 Otu0062 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Roseburia(100) .642752 1 Otu0063 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .866248 1 Otu0064 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) .684305 1 Otu0065 Actinobacteria(100) Actinobacteria(100) Coriobacteriales(100) Coriobacteriaceae(100) Enterorhabdus(100) .470026 1 Otu0066 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .781442 1 Ot

tes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) .684305 1 Otu0065 Actinobacteria(100) Actinobacteria(100) Coriobacteriales(100) Coriobacteriaceae(100) Enterorhabdus(100) .470026 1 Otu0066 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .781442 1 Ot u0067 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .075425 1 Otu0068 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .23604 1 Otu0069 Proteobacteria(100) Deltaproteobacteria(100) Desulfovibrionales(100) Desulfovibrionaceae(100) Desulfovibrio(93) .122282 1 Otu0070 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Dorea(100) .335755 1 Otu0071 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .275564 1 Otu0072 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Unclassified(100) .6759 1 Otu0073 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(98) Unclassified(98) .739009 1 Otu0074 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .099638 1 Otu0075 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Unclassified(100) .142259 1 Otu0076 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .446379 1 Otu0077 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Butyrivibrio(91) .048125 1 Otu0078 Actinobacteria(100) Actinobacteria(100) Coriobacteriales(100) Coriobacteriaceae(100) Unclassified(100) .020051 1 Otu0079 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .995336 1 Otu0080 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Dorea(100) .580909 1 Otu0081 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Butyrivibrio(96) .231131 1 Otu0082 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .144752 1 Otu0083 Actinobacteria(100) Actinobacteria(100) Coriobacteriales(100) Coriobacteriaceae(100) Asaccharobacter(100) .683732 1 Otu0084 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .059598 1 Otu0085 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .557724 1 Otu0086 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .829296

0084 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .059598 1 Otu0085 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .557724 1 Otu0086 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .829296 1 Otu0087 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Unclassified(100) .147604 1 Otu0088 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .325188 1 Otu0089 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .270353 1 Otu0090 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVa(98) .893607 1 Otu0091 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Unclassified(97) Unclassified(97) .064746 1 Otu0092 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Syntrophococcus(100) .465783 1 Otu0093 Proteobacteria(100) Deltaproteobacteria(100) Desulfovibrionales(100) Desulfovibrionaceae(100) Desulfocurvus(73) .719197 1 Otu0094 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(100) .223923 1 Otu0095 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(65) .323615 1 Otu0096 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .220402 1 Otu0097 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .755524 1 Otu0098 Firmicutes(100) Erysipelotrichia(100) Erysipelotrichales(100) Erysipelotrichaceae(100) Clostridium_XVIII(100) .167966 1 Otu0099 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .658952 1 Otu0100 Firmicutes(100) Bacilli(100) Lactobacillales(100) Streptococcaceae(100) Streptococcus(100) .271556 1 Otu0101 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .529912 1 Otu0102 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Unclassified(98) .239692 1 Otu0103 Actinobacteria(100) Actinobacteria(100) Coriobacteriales(100) Coriobacteriaceae(100) Unclassified(100) .438692 1 Otu0104 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Flavonifractor(97) .884299 1 Otu0105 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Parabacteroides(100) .604151 1 Otu0106 Firmicutes(100) Clostridia(100) Clostridiales(100)

e(100) Unclassified(100) .438692 1 Otu0104 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Flavonifractor(97) .884299 1 Otu0105 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Parabacteroides(100) .604151 1 Otu0106 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .052835 1 Otu0107 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Lactonifactor(100) .688385 1 Otu0108 Firmicutes(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .808444 1 Otu0109 Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) Unclassified(100) .580963 1 Otu0110 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) .590556 1 Otu0111 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .70702 1 Otu0112 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .629385 1 Otu0113 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVa(97) .690915 1 Otu0114 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .659852 1 Otu0115 Firmicutes(100) Clostridia(100) Clostridiales(100) Unclassified(100) Unclassified(100) .052071 1 Otu0116 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Clostridium_IV(100) .496749 1 Otu0117 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Lachnospiracea_incertae_sedis(64) .783083 1 Otu0118 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .450217 1 Otu0119 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .474766 1 Otu0120 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Roseburia(100) .677578 1 Otu0121 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(100) .764818 1 Otu0122 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Roseburia(71) .221575 1 Otu0123 Proteobacteria(100) Deltaproteobacteria(100) Desulfovibrionales(100) Desulfovibrionaceae(100) Bilophila(100) .759601 1 Otu0124 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Hydrogenoanaerobacterium(100) .641479 1 Otu0125 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(100) .015658 1 Otu0126 Firmicutes(100) Clostridia(100)

ovibrionaceae(100) Bilophila(100) .759601 1 Otu0124 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Hydrogenoanaerobacterium(100) .641479 1 Otu0125 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(100) .015658 1 Otu0126 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Flavonifractor(100) .017547 1 Otu0127 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Butyrivibrio(100) .106234 1 Otu0128 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .090769 1 Otu0129 Firmicutes(100) Erysipelotrichia(100) Erysipelotrichales(100) Erysipelotrichaceae(100) Erysipelotrichaceae_incertae_sedis(100) .304972 1 Otu0130 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Syntrophococcus(100) .351336 1 Otu0131 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Flavonifractor(100) .733134 1 Otu0132 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Pseudoflavonifractor(100) .651459 1 Otu0133 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .336219 1 Otu0134 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) .364453 1 Otu0135 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .297057 1 Otu0136 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Anaerotruncus(100) .255234 1 Otu0137 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(66) .06594 1 Otu0138 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVa(100) .934883 1 Otu0139 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(100) Unclassified(100) .364453 1 Otu0140 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .019222 1 Otu0141 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .486003 1 Otu0142 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .714399 1 Otu0143 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .641689 1 Otu0144 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(69) Unclassified(69) .511826 1 Otu0145 Bacteroidetes(100) Flavobacteria(97) Flavobacteriales(97) Unclassified(97) Unclassified(97) .718342 1 Otu0146 Ba

0) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .641689 1 Otu0144 Bacteroidetes(100) Bacteroidia(100) Bacteroidales(100) Porphyromonadaceae(69) Unclassified(69) .511826 1 Otu0145 Bacteroidetes(100) Flavobacteria(97) Flavobacteriales(97) Unclassified(97) Unclassified(97) .718342 1 Otu0146 Ba cteroidetes(100) Bacteroidia(100) Bacteroidales(100) Rikenellaceae(100) Alistipes(100) .977493 1 Otu0147 Firmicutes(100) Clostridia(100) Clostridiales(100) Ruminococcaceae(100) Oscillibacter(100) .345778 1 Otu0148 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .08833 1 Otu0149 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Clostridium_XlVb(100) .548781 1 Otu0150 Firmicutes(100) Clostridia(100) Clostridiales(100) Lachnospiraceae(100) Unclassified(100) .633538 1 NOTE. Taxonomic classifications for each OTU were generated using the reference taxonomy database from the Ribosomal Database Project. The numbers in parenthesis following each classification show the consistency of the classification assignments (in percent) for all of the sequence reads within a given OTU. p values and q values (false discovery rate adjusted p values, to allow for multiple comparisons) were generated using Metastats software (White JR et al. PLoS Comput Biol. 2009:e1000352).

lassification show the consistency of the classification assignments (in percent) for all of the sequence reads within a given OTU. p values and q values (false discovery rate adjusted p values, to allow for multiple comparisons) were generated using Metastats software (White JR et al. PLoS Comput Biol. 2009:e1000352). Supplementary Figure 1 (A) Representative flow cytometry plot and statistical analysis (B) showing differential expression of NKp46 and CCR6 in live CD45+ CD90+ IL7R+ cells in the colon of Rag2-/- (n = 6, white bar) and TRUC (n = 5, red bar) mice. Error bars depict SEM. *P < .01, **P < .005. (C) Flow cytometric analysis of the phenotype of CD90+ IL7R+ ILCs in the colons of TRUC mice. Grey histograms show staining with isotype-matched control antibody in comparison with staining with specific antibody (white histograms with black lines). Data are representative of more than 3 individual experiments. (D) Proportion of cytokine-expressing CD45+ CD90+ IL7R+ ILCs in the colons of Rag2-/- and TRUC mice after stimulation of unfractionated cLPMCs with PMA and ionomycin. Each dot/square represents an individual mouse.

ibody (white histograms with black lines). Data are representative of more than 3 individual experiments. (D) Proportion of cytokine-expressing CD45+ CD90+ IL7R+ ILCs in the colons of Rag2-/- and TRUC mice after stimulation of unfractionated cLPMCs with PMA and ionomycin. Each dot/square represents an individual mouse. Supplementary Figure 2 (A) IL6 concentration in serum (left panel) and colon explant culture (right panel) of Rag2-/- and TRUC mice measured by ELISA. Dots/squares represent individual mice. Line depicts the median. (B) Microarray analysis showing abundance of transcripts encoded by genes known to be regulated by IL6 in the colon of TRUC mice relative to Rag2-/- mice. Blue bars represent transcripts up-regulated in the colons of TRUC mice and red bars represent down-regulated genes (in comparison with Rag2-/- mice). Supplementary Figure 3 (A) FACs gating strategy used to purify ILCs from the colon of TRUC mice. CD45+ cells were first enriched from cLPMCs using anti-CD45 immunomagnetic beads. ILCs were FACs purified more than 98%. (B) Cytokine production by FACs purified CD90+ IL7R+ NKp46- colonic ILCs from TRUC mice. Cells were cultured in the presence of combinations of IL1α, IL23, IL6 (as depicted), or medium alone (unstimulated). IL6 was measured in culture supernatant by CBA. Bars show the mean cytokine production and error bars depict the SEM. Data are from 2 independent experiments.

+ IL7R+ NKp46- colonic ILCs from TRUC mice. Cells were cultured in the presence of combinations of IL1α, IL23, IL6 (as depicted), or medium alone (unstimulated). IL6 was measured in culture supernatant by CBA. Bars show the mean cytokine production and error bars depict the SEM. Data are from 2 independent experiments. Supplementary Figure 4 (A) Flow cytometry plots showing IL6R expression by CD4+ T cells, ILCs (lineage- IL7R+) in the colons of wild-type (WT) and TRUC mice. Data are representative of 3 independent experiments. (B) Concentration of sIL6R in serum (n = 6) and supernatants of cultured colon explants (n = 5) and unfractionated splenocytes (n = 3) measured by ELISA. Bars represent the mean sIL6R concentration and error bars depict the SEM.

lons of wild-type (WT) and TRUC mice. Data are representative of 3 independent experiments. (B) Concentration of sIL6R in serum (n = 6) and supernatants of cultured colon explants (n = 5) and unfractionated splenocytes (n = 3) measured by ELISA. Bars represent the mean sIL6R concentration and error bars depict the SEM. Supplementary Figure 5 (A) Concentration of IL6 in serum of TRUC mice treated with anti-IL6 or control isotype antibody. Dots/squares represent individual mice. Line depicts median. (B) Bray Curtis cluster dendrogram showing that anti-IL6 treatment is not associated with a distinct microbiota profile. “Pre” indicates samples before anti-IL6 treatment and “Post” indicates samples after treatment. Bacterial families colored in shades of green belong to the Firmicutes phylum, blue belong to the Bacteroidetes phylum, yellow belong to the Deferribacteres phylum (Mucispirillum genus), and maroon/red belong to the Proteobacteria phylum. (C) Box and whisker plots of Simpson diversity index of the intestinal microbiota from TRUC mice treated with anti-IL6 or isotype-matched control antibodies. Diversity indices are highly sensitive to differential sequencing depth. Therefore, analyses were confined to 477 reads per sample.

to the Proteobacteria phylum. (C) Box and whisker plots of Simpson diversity index of the intestinal microbiota from TRUC mice treated with anti-IL6 or isotype-matched control antibodies. Diversity indices are highly sensitive to differential sequencing depth. Therefore, analyses were confined to 477 reads per sample. Supplementary Figure 6 (A) Representative flow cytometry plots of intracellular cytokine and surface CD3 expression in noninflammatory control, CD, and UC patients. Cells were stimulated with PMA and ionomycin. (B) Real-time PCR analysis of RORC expression in sorted colonic CD3- IL7R+ cells in comparison with CD14+ monocytes (immunomagnetically selected from peripheral blood monocytes). Histogram shows the mean expression of RORC in purified colonic CD3- IL7R+ cells (n = 2 IBD patients) relative to monocytes. (C) Proportion of colonic CD3- IL7R+ cells expressing c-kit (CD117) in noninflammatory control, CD, and UC patients. (D) Proportion of colonic CD3- IL7R+ cells expressing NKp46, NKp44, or double-negative cells (NKp46- NKp44-) in noninflammatory control, CD, and UC patients. (E) IL6 production in colon explant cultures from IBD patients. In graphs, each dot represents an individual patient and a line depicts the median. Acknowledgments The authors are grateful to the Wellcome Trust Sanger Institute’s core sequencing team for performing 16S ribosomal RNA gene sequencing, and to Chris Evagora and support staff at the Pathology Core at Queen Mary University of London.

Supplementary Figure 6 (A) Representative flow cytometry plots of intracellular cytokine and surface CD3 expression in noninflammatory control, CD, and UC patients. Cells were stimulated with PMA and ionomycin. (B) Real-time PCR analysis of RORC expression in sorted colonic CD3- IL7R+ cells in comparison with CD14+ monocytes (immunomagnetically selected from peripheral blood monocytes). Histogram shows the mean expression of RORC in purified colonic CD3- IL7R+ cells (n = 2 IBD patients) relative to monocytes. (C) Proportion of colonic CD3- IL7R+ cells expressing c-kit (CD117) in noninflammatory control, CD, and UC patients. (D) Proportion of colonic CD3- IL7R+ cells expressing NKp46, NKp44, or double-negative cells (NKp46- NKp44-) in noninflammatory control, CD, and UC patients. (E) IL6 production in colon explant cultures from IBD patients. In graphs, each dot represents an individual patient and a line depicts the median. Acknowledgments The authors are grateful to the Wellcome Trust Sanger Institute’s core sequencing team for performing 16S ribosomal RNA gene sequencing, and to Chris Evagora and support staff at the Pathology Core at Queen Mary University of London. Microbiota sequence data were deposited in the European Nucleotide Archive under Study Accession Number ERP005850 and Sample Accession Numbers ERS459682–ERS459702.

Acknowledgments The authors are grateful to the Wellcome Trust Sanger Institute’s core sequencing team for performing 16S ribosomal RNA gene sequencing, and to Chris Evagora and support staff at the Pathology Core at Queen Mary University of London. Microbiota sequence data were deposited in the European Nucleotide Archive under Study Accession Number ERP005850 and Sample Accession Numbers ERS459682–ERS459702. Conflicts of interest These authors disclose the following: Nick Powell has received honoraria for acting in an advisory capacity or speaking on behalf of Actavis UK, Ferring and AstraZeneca; Peter Irving has received honoraria for acting in an advisory capacity or speaking on behalf of AbbVie, MSD, Actavis UK, Shire, Ferring, Falk, Genentech, Tillotts, Takeda, Vifor Pharma, Pharmacosmos, and Symprove; Thomas MacDonald receives support from Glaxo Smith Kline, Janssen Pharmaceuticals, Grunenthal, VH2, and Topivert; and Bu Hayee has received honoraria for acting in an advisory capacity or speaking on behalf of AbbVie, Takeda, and Actavis UK. The remaining authors disclose no conflicts.

arma, Pharmacosmos, and Symprove; Thomas MacDonald receives support from Glaxo Smith Kline, Janssen Pharmaceuticals, Grunenthal, VH2, and Topivert; and Bu Hayee has received honoraria for acting in an advisory capacity or speaking on behalf of AbbVie, Takeda, and Actavis UK. The remaining authors disclose no conflicts. Funding Supported by grants awarded from the Wellcome Trust (WT101159AIA to N.P.; WT088747MA to N.P., G.M.L., and T.T.M.; 091009 to G.M.L.; and 098051 to A.W.W., P.S., and J.P.), and the Medical Research Council (G0802068 to G.M.L. and T.T.M., and MR/K002996/1 to G.M.L. and J.K.H.), by the Scottish Government Rural and Environmental Science and Analysis Service (A.W.W.), and by the National Institute for Health Research Biomedical Research Centre at Guy’s and St Thomas’ and King’s College London. The views expressed are those of the author(s) and not necessarily those of the National Health Service, the National Institute for Health Research, or the Department of Health. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at http://dx.doi.org/10.1053/j.gastro.2015.04.017.

Funding Supported by grants awarded from the Wellcome Trust (WT101159AIA to N.P.; WT088747MA to N.P., G.M.L., and T.T.M.; 091009 to G.M.L.; and 098051 to A.W.W., P.S., and J.P.), and the Medical Research Council (G0802068 to G.M.L. and T.T.M., and MR/K002996/1 to G.M.L. and J.K.H.), by the Scottish Government Rural and Environmental Science and Analysis Service (A.W.W.), and by the National Institute for Health Research Biomedical Research Centre at Guy’s and St Thomas’ and King’s College London. The views expressed are those of the author(s) and not necessarily those of the National Health Service, the National Institute for Health Research, or the Department of Health. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at http://dx.doi.org/10.1053/j.gastro.2015.04.017. Figure 1 IL17A/IL22-producing CD4- NCR- ILC3 mediate colitis in TRUC mice. (A) Flow cytometry dot plots of live, CD45+ cells according to expression of CD90 and IL7R in the colons of Rag2-/- and TRUC mice, and (B) absolute numbers of live CD45+ CD90+ IL7R+ ILCs in the colons of TRUC and Rag2-/- mice. Each dot/square represents an individual mouse. Line depicts the median. (C) Representative flow cytometry dot plots (left) and statistical analysis (right) of the proportion of CD4+ and CD90+ cells (gated on live CD45+ cells) in mLNs and colons of TRUC mice treated with isotype-matched control antibodies (n = 5), anti-CD4 (n = 4), or anti-CD90 (n = 4) antibodies. Statistical analyses were performed on colonic cells and show the proportion of colonic ILCs (cILCs) after treatment. *P < .019, **P < .04. (D) Representative flow cytometry dot plots (left) and statistical analysis (right) of the proportion of IL17A+ and IL22+ cells (gated on live CD45+ cells) in the colon of TRUC mice treated with isotype-matched control antibodies (n = 5), anti-CD4 (n = 4), or anti-CD90 (n = 24) antibodies. Cells were stimulated with PMA and ionomycin before intracellular cytokine staining *P < .01. (E) Representative colon micrographs (H&E stained) (left) and statistical analysis of colitis scores (right) of TRUC mice treated with isotype-matched control antibodies, anti-CD4, or anti-CD90 antibodies. *P < .03 (for both anti-CD90 vs control antibody and anti-CD90 vs anti-CD4). Each dot/square represents an individual mouse. Lines depict the median.

s (H&E stained) (left) and statistical analysis of colitis scores (right) of TRUC mice treated with isotype-matched control antibodies, anti-CD4, or anti-CD90 antibodies. *P < .03 (for both anti-CD90 vs control antibody and anti-CD90 vs anti-CD4). Each dot/square represents an individual mouse. Lines depict the median. Figure 2 IL6 promotes cytokine production by NCR- ILC3s in a cell-intrinsic manner. (A) Microarray analysis showing an abundance of cytokine transcripts in the colon of TRUC mice relative to Rag2-/- mice. Blue dotted line depicts 2-fold induction. (B) IL17A production by unfractionated cLPMCs and mLN cells isolated from TRUC mice in medium alone (-) or after supplementation with recombinant IL6 or IL23. Columns represent mean cytokines and error bars depict SEM. Analysis of cLPMCs comprised 4 biological replicates. Analysis of mLN included 9 biological replicates for the unstimulated condition and 7 biological replicates for each of the stimulated conditions. *P < .02; **P < .003. (C) Flow cytometry plots of intracellular IL17A expression by CD90+ NKp46- cells after stimulation of unfractionated mLN cells with IL6, IL23, or unstimulated cells (-), which were incubated with monensin alone. Data are representative of 3 separate experiments. (D) Cytokine production by fluorescence-activated cell sorted CD45+ CD90+ IL7R+ NKp46- ILCs purified from the colons of TRUC mice (the gating strategy for cell sorting is illustrated in Supplementary Figure 3A). Purified NCR- ILCs were stimulated with combinations of IL1α, IL6, and IL23 as depicted. After 24 hours cytokine concentrations were measured in culture supernatant by ELISA or CBA. Data are representative of 2 individual experiments with ILCs pooled from 10–15 colons. Bars show the mean cytokine production and error bars depict SEM.

LCs were stimulated with combinations of IL1α, IL6, and IL23 as depicted. After 24 hours cytokine concentrations were measured in culture supernatant by ELISA or CBA. Data are representative of 2 individual experiments with ILCs pooled from 10–15 colons. Bars show the mean cytokine production and error bars depict SEM. Figure 3 IL6 blockade reduces IL17A production and attenuates TRUC disease. (A) IL17A concentration in culture supernatants of unfractionated cLPMCs and splenocytes from TRUC mice treated with anti-IL6 (n = 8) or isotype-matched control antibodies (n = 8). (B) Representative colon micrographs (H&E stained) (left panel) and statistical analysis (right panel) of colitis scores of distal colons of TRUC mice treated with anti-IL6 or isotype-matched control antibodies. (C) Spleen and colon mass of TRUC mice treated with anti-IL6 or isotype-matched control antibodies. Each dot/square represents an individual mouse. Lines represent medians. Results from 2 separate antibody blockade experiments conducted under the same experimental conditions were pooled.

ype-matched control antibodies. (C) Spleen and colon mass of TRUC mice treated with anti-IL6 or isotype-matched control antibodies. Each dot/square represents an individual mouse. Lines represent medians. Results from 2 separate antibody blockade experiments conducted under the same experimental conditions were pooled. Figure 4 IL6 blockade does not significantly impact the composition of the intestinal microbiota in TRUC mice. (A) The mean percentage of sequences of particular phyla present in the intestinal microbiota of TRUC mice before (top panel) and after (bottom panel) treatment with anti-IL6 (red bars) or isotype-matched control antibodies (white bars). (B) The mean proportional abundance of H typhlonius in the intestinal microbiota of TRUC mice before and after treatment with anti-IL6 (red bars) or isotype-matched control antibodies (white bars). Error bars depict the SEM.

m panel) treatment with anti-IL6 (red bars) or isotype-matched control antibodies (white bars). (B) The mean proportional abundance of H typhlonius in the intestinal microbiota of TRUC mice before and after treatment with anti-IL6 (red bars) or isotype-matched control antibodies (white bars). Error bars depict the SEM. Figure 5 CD3- IL7R+ cells are expanded in IBD patients. (A) Flow cytometry plots showing CD3 and intracellular IL17A expression in unfractionated cLPMCs after stimulation with PMA and ionomycin. Additional representative flow cytometry plots are illustrated in Supplementary Figure 6A. (B) Representative flow cytometry plots (left panel) and statistical analysis (right panel) of CD3 and IL7R staining by LPMCs in the colon of noninflammatory control and IBD patients. Individual dots represent individual patients. *P < .04. (C) Flow cytometric analysis of the phenotype of colonic CD3- IL7R+ cells. Grey histograms show isotype control antibody staining. White histograms show staining with specific antibody. Data are representative of more than 3 independent experiments in IBD patients. (D) Flow cytometry dot plots showing expression of NKp46 and NKp44 by colonic CD3- IL7R+ cells in noninflammatory control and IBD patients. Further analyses of additional patient replicates are shown in Supplementary Figure 6C and D.

ibody. Data are representative of more than 3 independent experiments in IBD patients. (D) Flow cytometry dot plots showing expression of NKp46 and NKp44 by colonic CD3- IL7R+ cells in noninflammatory control and IBD patients. Further analyses of additional patient replicates are shown in Supplementary Figure 6C and D. Figure 6 IL6-responsive CD3- IL7R+ cells are present in the colon of IBD patients. (A) Flow cytometry histograms and (B) statistical analyses of intracellular cytokine expression by CD3- IL7R+ cells after overnight culture in the presence or absence of IL6 (100 ng/mL). Cells were restimulated with PMA and ionomycin before staining. (B) Each connected pair of dots represents an individual patient. (C) Flow cytometry histograms showing the number of CD3- IL7R+ IL17A+ cells after culture with increasing doses of IL6.

Barrett’s esophagus (BE) is a common premalignant condition that affects up to 2% of the adult population in the Western world.1 BE comprises the second stage in the esophagitis–metaplasia–dysplasia–adenocarcinoma sequence. BE confers a 2%–4% lifetime risk of esophageal adenocarcinoma (EAC).1 Chronic gastric acid reflux is the predominant etiologic factor for BE. In addition, BE co-occurs with conditions such as intestinal metaplasia, hiatal hernia, obesity, and hypercholesterolemia.2–5 Several factors, including the degree of acid reflux, hiatal hernia size, and the percentage of intestinal metaplasia–positive glands, can affect the progression to cancer. A role for genetics in the pathogenesis of gastroesophageal reflux disease, including BE and EAC, has been implicated on the basis of 3 observations: concordance in monozygous and dizygous twins6–8; the increased risk of disease in those with a positive family history9,10; and, recently, the identification of single nucleotide polymorphisms (SNPs) associated with BE in Genome-Wide Association Studies (GWAS).11,12 The proportion of variation in BE risk explained by common variants has been estimated to be 35%.13

8; the increased risk of disease in those with a positive family history9,10; and, recently, the identification of single nucleotide polymorphisms (SNPs) associated with BE in Genome-Wide Association Studies (GWAS).11,12 The proportion of variation in BE risk explained by common variants has been estimated to be 35%.13 Our GWAS previously identified 2 SNPs, on chromosomes 6p21 (rs9257809; P = 4.1 × 10−9) and 16q24 (rs9936833; P = 2.7 × 10−10), that are associated with BE.11 One of these loci lies within the HLA region and the other, close to FOXF1, which is involved in esophageal structure and development. Both SNPs have also been shown to be associated with risk of EAC.14 More recently, in a combined analysis of BE and EAC cases, the Barrett's and Esophageal Adenocarcinoma Consortium (BEACON) identified susceptibility SNPs within CRTC1 and BARX1, and near FOXP1.12 The last 2 of these genes are known to be involved in esophageal development.15,16 We aimed to identify further BE predisposition SNPs from our GWAS by performing wider and deeper independent replication of SNPs that already had promising disease associations.

Our GWAS previously identified 2 SNPs, on chromosomes 6p21 (rs9257809; P = 4.1 × 10−9) and 16q24 (rs9936833; P = 2.7 × 10−10), that are associated with BE.11 One of these loci lies within the HLA region and the other, close to FOXF1, which is involved in esophageal structure and development. Both SNPs have also been shown to be associated with risk of EAC.14 More recently, in a combined analysis of BE and EAC cases, the Barrett's and Esophageal Adenocarcinoma Consortium (BEACON) identified susceptibility SNPs within CRTC1 and BARX1, and near FOXP1.12 The last 2 of these genes are known to be involved in esophageal development.15,16 We aimed to identify further BE predisposition SNPs from our GWAS by performing wider and deeper independent replication of SNPs that already had promising disease associations. Methods Patient and Sample Collection Criteria Figure 1 outlines this study and the numbers of samples that contributed to each phase. As described previously,11 Discovery Phase cases were diagnosed with histologically confirmed BE and ascertained through the UK-based Aspirin Esomeprazole Chemoprevention Trial (AspECT) in Barrett's Metaplasia, a clinical trial of proton-pump inhibitor (esomeprazole) and aspirin as preventive agents for progression of BE to EAC.17 Replication Phase UK, Irish, Dutch, and Belgian patient samples were obtained from the Chemoprevention Of Premalignant Intestinal Neoplasia (ChOPIN) genetic study and the Esophageal Adenocarcinoma GenEtics (EAGLE) consortium.1 Replication Phase patients were diagnosed with BE with lengths of ≥1 cm (C1M1) circumferential disease or ≥2 cm tongue patterns (C0M2), according to the Prague criteria.18 Patient collection was in accordance with British Society of Gastroenterology criteria for BE19 and followed verification of endoscopic findings and proven BE on histopathologic records. Presence of EAC at presentation or subsequently was recorded, but was not an inclusion criterion.

according to the Prague criteria.18 Patient collection was in accordance with British Society of Gastroenterology criteria for BE19 and followed verification of endoscopic findings and proven BE on histopathologic records. Presence of EAC at presentation or subsequently was recorded, but was not an inclusion criterion. Sample Sets Discovery Phase BE patients (n = 1852) were UK participants in the AspECT study (Chief Investigator: Jankowski), HANDEL study (Chief Investigator: Jankowski), ChOPIN study (Chief Investigator: Jankowski), and population controls of white Caucasian origin (n = 5172) were from the common Wellcome Trust Case Control Consortium 2 (WTCCC2) set.11 Replication Phase 1 UK Replication 1 totaled 1105 BE patients from ChOPIN and 6819 controls. The controls comprised People of the British Isles (Chief Investigator: W. Bodmer) (n = 2578) and WTCCC2 (Chief Investigator: P. Donnelly) samples (n = 4241) that were not genotyped in the Discovery Phase. The Dutch replication samples consisted of 473 BE patients and 1780 controls from the University Medical Centre, Groningen. An additional 64 Dutch cases and 206 controls, provided since 2012 from Nijmegen and Rotterdam as part of EAGLE, were genotyped for the 7 SNPs taken into Replication Phase 3.

the Discovery Phase. The Dutch replication samples consisted of 473 BE patients and 1780 controls from the University Medical Centre, Groningen. An additional 64 Dutch cases and 206 controls, provided since 2012 from Nijmegen and Rotterdam as part of EAGLE, were genotyped for the 7 SNPs taken into Replication Phase 3. Replication Phase 2 UK Replication 2 comprised 1765 BE patients from the ChOPIN study. Controls (n = 1586) were from the Colorectal Tumour Gene Identification (CoRGI) Consortium20 (Chief Investigator: Tomlinson), comprising spouses or partners unaffected by cancer and without a family history (to 2nd-degree relative level) of colorectal neoplasia. All were of white UK ethnic origin. The Irish replication samples were 245 BE patients and 473 controls of white Caucasian origin from St James’s Hospital and Mater Misericordiae University Hospital, Dublin. Healthy donor controls were provided by Trinity Biobank.

(to 2nd-degree relative level) of colorectal neoplasia. All were of white UK ethnic origin. The Irish replication samples were 245 BE patients and 473 controls of white Caucasian origin from St James’s Hospital and Mater Misericordiae University Hospital, Dublin. Healthy donor controls were provided by Trinity Biobank. Replication Phase 3 UK Replication 3 comprised 997 BE patients from the ChOPIN study and 974 female controls from the Genetics of Lobular Carcinoma In Situ in Europe (GLACIER) study (Chief Investigators: Sawyer, Roylance) with no personal or family history of breast cancer and of white Caucasian origin.21 The Belgian replication samples consisted of 362 cases and 848 controls from Leuven. Finally, 3295 BE patients and 3204 controls predominantly of northern European descent from the BEACON consortium GWAS (Chief Investigators: Vaughan, Whiteman, Levine) were included.12 All studies received ethical board approval (details in Supplementary Material). Two SNPs described in Su et al11 had been genotyped previously in Replication Phase 2 and BEACON/BEAGESS samples. All other Replication Phase 3 samples were new to this study.

hief Investigators: Vaughan, Whiteman, Levine) were included.12 All studies received ethical board approval (details in Supplementary Material). Two SNPs described in Su et al11 had been genotyped previously in Replication Phase 2 and BEACON/BEAGESS samples. All other Replication Phase 3 samples were new to this study. Genotyping For all samples, genomic DNA was extracted from peripheral blood. Various genotyping methods were used, depending on the phase of the study and on pre-existing data from some sample sets. In brief, Discovery Phase genotyping was performed using the Illumina 660W-Quad array for cases and a custom Human 1.2M-Duo array for controls at the Wellcome Trust Sanger Institute.11 Replication Phase 1 genotyping was performed using the Illumina Immunochip at the Wellcome Trust Sanger Institute11 or as described in Trynka et al.22

very Phase genotyping was performed using the Illumina 660W-Quad array for cases and a custom Human 1.2M-Duo array for controls at the Wellcome Trust Sanger Institute.11 Replication Phase 1 genotyping was performed using the Illumina Immunochip at the Wellcome Trust Sanger Institute11 or as described in Trynka et al.22 In Replication Phase 2, samples underwent custom genotyping for SNPs that met one of the following criteria: Passociation < 10−4 in combined Discovery and Replication Phase 1 analysis (n = 63); Passociation < 10−4 in Discovery Phase, but not included in Replication Phase 1 (n = 12); and Passociation < 10−4 in a sex-stratified analysis of the Discovery phase (n = 5); and candidate polymorphisms previously reported as associated with BE and not well tagged by the Discovery Phase or Immunochip arrays, specifically, MSR1 p.Arg293Gly,23 and variants in IGF1R and GHR24 (Supplementary Table 1). Sequenom iPLEX assays were successfully designed for 65 of these SNPs and genotyping was performed at the Wellcome Trust Sanger Institute. Genotypes were assigned using MassArray TyperAnalyzer 4.0 (Sequenom). Samples with sex discrepancies between manifests or with overall call rates <95% were excluded, as were SNPs with call rates of <95%. Where a SNP was in the top 40 of the prioritized SNPs (by P value) and had failed at the design stage of the iPLEX (n = 3), genotyping was performed by KASPar in the full Replication Phase 2 sample set. Seventy-seven of two hundred and forty-five Irish cases passing quality control were genotyped using iPLEX assays, and the other cases were genotyped using the Immunochip. Eighteen iPLEX SNPs were not analyzed in the Irish cohort, as the SNPs were not present on the Immunochip. The samples were also genotyped for 4 SNPs after publication of Levine et al.12

-five Irish cases passing quality control were genotyped using iPLEX assays, and the other cases were genotyped using the Immunochip. Eighteen iPLEX SNPs were not analyzed in the Irish cohort, as the SNPs were not present on the Immunochip. The samples were also genotyped for 4 SNPs after publication of Levine et al.12 Replication Phase 3 samples (Supplementary Table 2) were genotyped using KASPar for 7 SNPs prioritized after analysis of the previous phases. The samples were also genotyped for 4 SNPs after publication of Levine et al.12 All SNPs had call rates >95%. Sample exclusions were as for Replication Phase 2. We had previously demonstrated >99% concordance between genome-wide array, Immunochip, and KASPar assays for other SNPs.11 Sequenom call rate was >96%. For samples analyzed only by KASPar, genotyping QC was tested using duplicate DNA samples within studies and SNP assays, together with direct sequencing of subsets of samples to confirm genotyping accuracy. For all SNPs, >98% concordant results were obtained.

KASPar assays for other SNPs.11 Sequenom call rate was >96%. For samples analyzed only by KASPar, genotyping QC was tested using duplicate DNA samples within studies and SNP assays, together with direct sequencing of subsets of samples to confirm genotyping accuracy. For all SNPs, >98% concordant results were obtained. Association Analysis Case-control analysis was performed using frequentist tests under a missing data logistic regression model, as implemented in SNPTEST (version 2.4.1). Principal component analysis was performed for all samples typed on GWAS arrays (Discovery Phase) and has been described in Su et al.11 As described previously, principal component 1 (PC1) was included as a covariate in all analyses of the Discovery Phase. Each SNP was tested as a quantitative explanatory variable, coded as 0, 1, 2. We used GWAMA (version 2.1) to implement fixed inverse variance-based methods for meta-analysis.25 The software tests for heterogeneity of effects between studies26 and enables sex-specific meta-analysis.

all analyses of the Discovery Phase. Each SNP was tested as a quantitative explanatory variable, coded as 0, 1, 2. We used GWAMA (version 2.1) to implement fixed inverse variance-based methods for meta-analysis.25 The software tests for heterogeneity of effects between studies26 and enables sex-specific meta-analysis. Replication and Validation of Single Nucleotide Polymorphisms From BEACON/BEAGESS Meta-analysis In order to examine the 4 genome-wide significant BE + EAC SNPs and 83 other SNPs with Passoc < 10−4 in the BEACON/BEAGESS data, we performed association testing using AspECT Discovery Phase cases. Because our Discovery Phase controls overlapped with those used by Levine et al,12 we used 1898 white European controls from colorectal cancer GWAS studies CoRGI and Colon Cancer Family Registry.27 Genotypes were imputed where necessary, with strict cut-offs for imputation quality.28,29 Other Analyses Details of imputation, fine mapping, pathway analyses, estimation of heritability, and URLs are provided in the Supplementary Material.

Replication and Validation of Single Nucleotide Polymorphisms From BEACON/BEAGESS Meta-analysis In order to examine the 4 genome-wide significant BE + EAC SNPs and 83 other SNPs with Passoc < 10−4 in the BEACON/BEAGESS data, we performed association testing using AspECT Discovery Phase cases. Because our Discovery Phase controls overlapped with those used by Levine et al,12 we used 1898 white European controls from colorectal cancer GWAS studies CoRGI and Colon Cancer Family Registry.27 Genotypes were imputed where necessary, with strict cut-offs for imputation quality.28,29 Other Analyses Details of imputation, fine mapping, pathway analyses, estimation of heritability, and URLs are provided in the Supplementary Material. Results Identification of Two New Barrett’s Esophagus Predisposition Single Nucleotide Polymorphisms In order to identify further loci associated with BE, we prioritized 65 SNPs (Supplementary Table 1) with the best evidence of association with BE from our previous GWAS Discovery Phase and Replication Phase 1 (details in Methods).11 These SNPs were genotyped in an additional 1765 cases and 1586 controls from the UK and in an Irish cohort of 245 cases and 473 controls (Replication Phase 2, previously used to genotype rs9257809 and rs9936833 described in Su et al11). After meta-analysis of these new data together with Discovery Phase and Replication Phase 1, seven SNPs showing evidence of associations with BE risk at Pmeta <5 × 10−6 were identified and genotyped in Replication Phase 3 samples (Table 1). After Replication Phase 3, two SNPs—rs3072 and rs2701108 on chromosome 2p24 and 12q24, respectively—reached the level of significance conventionally used for GWAS (P = 5 × 10−8) (Table 1). Combined Pmeta values were 1.8 × 10−11 for rs3072 (OR = 1.14; 95% CI: 1.09–1.18) and 7.5 × 10−9 for rs2701108 (OR = 0.90; 95% CI: 0.86–0.93), derived from a total sample of 10,158 BE cases and 21,062 controls (Supplementary Table 3). The associations remained at or near genome-wide significance upon restricting the analysis to the 8521 cases with histologically proven intestinal metaplasia (rs3072: P = 1.3 × 10−9; OR = 1.13; 95% CI: 1.09–1.17; rs2701108: P = 6.2 × 10−8; OR = 0.90; 95% CI: 0.86–0.94). There was no evidence of sex heterogeneity for either SNP (Supplementary Table 4).

ar genome-wide significance upon restricting the analysis to the 8521 cases with histologically proven intestinal metaplasia (rs3072: P = 1.3 × 10−9; OR = 1.13; 95% CI: 1.09–1.17; rs2701108: P = 6.2 × 10−8; OR = 0.90; 95% CI: 0.86–0.94). There was no evidence of sex heterogeneity for either SNP (Supplementary Table 4). In Silico Fine Mapping and Annotation of the Chromosome 2p24 and 12q24 Loci rs3072 lies between 2 genes, mapping 7.5 kb downstream of GDF7 (also known as BMP12) and 6.5 kb downstream of C2orf43 (Figure 2). rs2701108 is 117 kb downstream of TBX5 and 270 kb upstream of RBM19. We imputed in our Discovery Phase all SNPs in 1-Mb regions flanking each of the lead SNPs. At chromosome 2p24, rs3072 remained the most strongly associated SNP, but at chromosome 12q24, rs1920562 was more strongly associated with disease risk (PDiscovery = 1.4 × 10−5; OR = 0.84) than the lead genotyped SNP (PDiscovery = 1.4 × 10−3; OR = 0.88). rs1920562 (linkage disequilibrium [LD] with rs2701108; r2 = 0.6) lies 131 kb downstream of TBX5 and 256 kb upstream of RBM19. Nonsynonymous SNPs in the genes flanking the signals on chromosomes 2 and 12 were not in strong LD (r2 < 0.4; D’ < 0.8) with the lead genotyped or imputed SNPs, suggesting that the functional variants may have effects on gene expression and regulation rather than protein sequence. Haploregv230 and Annovar31 were used to annotate SNPs in strong LD (r2 > 0.4) with the 2 lead tagging SNPs.

and 12 were not in strong LD (r2 < 0.4; D’ < 0.8) with the lead genotyped or imputed SNPs, suggesting that the functional variants may have effects on gene expression and regulation rather than protein sequence. Haploregv230 and Annovar31 were used to annotate SNPs in strong LD (r2 > 0.4) with the 2 lead tagging SNPs. rs3072, which may alter a GATA binding motif, lies within a region of histone modifications, such as H3K4Me1, which mark enhancers (data from lymphoblastoid cell line (LCL) GM12878). Three other SNPs in LD with rs3072 map to the enhancer region detected in GM12878. One of these, rs7255, maps to a site of high evolutionary conservation/constraint; another SNP, rs9306894, whilst not at a conserved site (Supplementary Table 5), is predicted as “likely to affect protein binding and linked to expression of a gene target” according to RegulomeDB.32 We examined associations between SNPs in this region and gene expression in The Cancer Genome Atlas (TCGA) EAC data.33 Genotypes were only available for rs9306894 in the chromosome 2 locus and gene expression data had been obtained using RNASeq. After correcting for copy number, we determined associations between rs9306894 genotype and total RNA levels for expression quantitative trait locus (eQTL) analysis and bias in allelic expression of coding SNPs (allele-specific expression [ASE] analysis). There was no significant association with expression of the closest genes GDF7, HS1BP3 and C2orf43 (PeQTL > .20; PASE > .38; n = 62) and no genome-wide association with expression of any other gene was present (q > .05, details not shown). In public data sets based on monocytes34 and on lymphoblastoid cell lines and adipose tissue,35 C2orf43 is the suggested target of rs9306894 following eQTL studies (GENevar; P = 7 × 10−4). rs9306894 genotype was not associated with GDF7 expression in these cell types.

ther gene was present (q > .05, details not shown). In public data sets based on monocytes34 and on lymphoblastoid cell lines and adipose tissue,35 C2orf43 is the suggested target of rs9306894 following eQTL studies (GENevar; P = 7 × 10−4). rs9306894 genotype was not associated with GDF7 expression in these cell types. rs2701108 itself is not likely to be a functionally regulatory SNP, but rs1920562, which showed the strongest regional association after imputation, is a more promising candidate (Supplementary Table 6). This SNP maps to a highly conserved base and a region containing enhancer marks in human embryonic stem cells (h1-ESC) and lung fibroblasts (NHLF). rs1920562 and an additional SNP (rs1247938) in moderate LD (r2 = 0.52) with rs2701108, are highlighted by Regulome DB as being the most likely SNPs in this region to affect protein binding. CTCF and RAD21 binding are predicted to be affected by rs1247938 and the ability of IKZF1 to bind is predicted to be altered by rs1920562. Expression analyses were performed for the rs2701108 region, in the same way as for rs9306894. However, none of the three rs2701108 region SNPs was associated with TBX5, TBX3 or RBM19 expression in the TCGA data (PeQTL > .39; PASE > .43; n = 62), was an eQTL in whole-transcriptome analysis, or was an eQTL in the public databases (details not shown).

med for the rs2701108 region, in the same way as for rs9306894. However, none of the three rs2701108 region SNPs was associated with TBX5, TBX3 or RBM19 expression in the TCGA data (PeQTL > .39; PASE > .43; n = 62), was an eQTL in whole-transcriptome analysis, or was an eQTL in the public databases (details not shown). Pathway/Geneset Enrichment Analysis Improved Gene Set Enrichment Analysis for Genome Wide Association Study (iGSEA4GWAS) and SNP ratio test respectively found 26 and 34 pathways significantly enriched in cases at False Discovery Rate–corrected P < .05. Genetic Genomics Analysis of Complex Data (Gengen) did not identify any pathways with corrected P < .05, but 10 pathways had P < .25 (Supplementary Table 7). Three pathways (type 1 diabetes mellitus, KEGG antigen processing and presentation, and KEGG autoimmune thyroid disease) were identified by all methods and the SNPs mapping to each pathway were subjected to set-based association tests using PLINK, producing empirical P values of .0021, .025, and .0317, respectively. The SNPs within these pathways that showed replication at P < .05 in Replication Phase 1 all mapped to chromosome 6p21, either close to or within HLA genes. Upon removal of the HLA genes from all the pathways, 20/26 pathways originally with False Discovery Rate P < .05 according to the iGSEA4GWAS approach remained significant, but only 1/34 pathways identified by SNP ratio test remained significant. No HLA-depleted pathways were even suggestive of enrichment (all P > .25) by Gengen. The top networks identified by Ingenuity Pathway Analysis were Cardiovascular System Development and Function, Embryonic Development and Organ Development (Supplementary Figure 1). The 5 genes implicated by the BE susceptibility SNPs here are all involved in development at a cellular, embryonic, organ, and organism level. Bone morphogenetic protein 4 was the most significant upstream regulator (Poverlap = 1.99 × 10−6).

unction, Embryonic Development and Organ Development (Supplementary Figure 1). The 5 genes implicated by the BE susceptibility SNPs here are all involved in development at a cellular, embryonic, organ, and organism level. Bone morphogenetic protein 4 was the most significant upstream regulator (Poverlap = 1.99 × 10−6). Barrett’s Esophagus Heritability Genome-wide haplotype-tagging SNP data on the 1852 cases and 5172 controls in the Discovery Phase of this study were used in Genome-wide Complex Trait Analysis to estimate the proportion of variation in risk of BE that can be explained by common genetic variants. In line with our previous, disease score test analysis,11 we found a statistically significant component of BE risk to be polygenic (9.99% [SE 1.2%]). This was a lower estimate than that recently derived by Ek et al.13

estimate the proportion of variation in risk of BE that can be explained by common genetic variants. In line with our previous, disease score test analysis,11 we found a statistically significant component of BE risk to be polygenic (9.99% [SE 1.2%]). This was a lower estimate than that recently derived by Ek et al.13 Replication Testing of Previously Reported Barrett’s Esophagus Susceptibility Single Nucleotide Polymorphisms at Candidate Loci Using a systematic review, we identified 26 polymorphisms reported in the literature to be associated with BE (Supplementary Table 8). In our Discovery Phase samples, 20 of 26 SNPs were directly genotyped or were in strong LD (r2 > 0.7) with a directly genotyped SNP. Only one of these SNPs showed a nominally significant association (P < .05) in our data (rs909253, proxy for rs1041981, r2 = 0.93; OR = 1.12; P = .005). This SNP was also present on the Immunochip and showed additional evidence of replication (Pmeta = 3.1 × 10−4, OR = 1.07). rs909253 maps to a highly conserved base (based on SiPhy score) in an intron of LTA (tumor necrosis factor–β) where histone marks associated with both promoters and enhancers are present in lymphoblastoid cell lines (LCLs).30 PBX3, PU1, POL2, YY1, and nuclear factor–κB have all been found to bind here in ENCODE ChIP-seq experiments. No data were available for this SNP in RegulomeDB. We were able to genotype 3 other candidate SNPs previously reported for BE susceptibility (rs41341748, rs2715425, rs6898743) on the Sequenom iPLEX panel used to genotype UK Replication 2. None of these SNPs showed associations with BE risk (Supplementary Table 1). Because rs41341748 (MSR1 p.Arg293Gly) has a low minor allele frequency (<5%) and consequently our power to detect an association was relatively low, we additionally genotyped it in UK Replication 3. After meta-analysis of Replication 2 and 3, we remained unable to replicate the previously reported association23 between this SNP and BE risk (OR = 1.07; 95% CI: 0.70–1.43; P = .79).

requency (<5%) and consequently our power to detect an association was relatively low, we additionally genotyped it in UK Replication 3. After meta-analysis of Replication 2 and 3, we remained unable to replicate the previously reported association23 between this SNP and BE risk (OR = 1.07; 95% CI: 0.70–1.43; P = .79). Assessment of Previously Reported Barrett’s Esophagus + Esophageal Adenocarcinoma Single Nucleotide Polymorphisms and Meta-analysis With BEAGESS Data Three new genome-wide significant BE + EAC loci (4 SNPs) were recently identified by Levine et al12 in a combined analysis of EAC and BE: rs10419226 and rs10423674 in CRTC1, rs11789015 in BARX1 and rs2687201 within 100 kb of FOXP1. None of these associations was genome-wide significant at P < 5 × 10−8 when Levine et al restricted their analysis to BE cases alone, although one SNP, rs10419226, within CRTC1, reached P = 5.5 × 10−8. In our datasets, only rs10423674 had been directly genotyped, but the remaining SNPs were all reliably imputed in our Discovery Phase samples (Info scores >0.95). However, the controls used in the replication phase of the Levine et al12 study overlapped entirely with the controls used in our Discovery phase (WTCCC2 controls) and we therefore used alternative UK controls from the CoRGI study (see Methods).

e all reliably imputed in our Discovery Phase samples (Info scores >0.95). However, the controls used in the replication phase of the Levine et al12 study overlapped entirely with the controls used in our Discovery phase (WTCCC2 controls) and we therefore used alternative UK controls from the CoRGI study (see Methods). Of the 4 Levine SNPs, rs10423674, 1 of 2 SNPs in CRTC1, and rs2687201, near FOXP1, were supported in this study (P = 0.02; OR = 1.14; 95% CI: 1.03–1.27 and P = 0.05; OR = 0.94; 95% CI: 0.88–1.00, respectively). There was also some support for rs11789015, near BARX1 (P = 0.07; OR = 0.90; 95% CI: 0.81–1.01). However, the association at rs10419226, within CRTC1, was not replicated in our data (P = 0.87; OR = 1.01; 95% CI: 0.91–1.11). All 4 SNPs still reached genome-wide significance (P < 5 × 10−8) upon meta-analysis of our BE and Levine’s BE + EAC datasets. In a BE-only meta-analysis, the associations improved with the inclusion of our data for 3 out of the 4 SNPs, with 1 (rs2687201, FOXP1) reaching genome-wide significance for BE (Table 2).

1.11). All 4 SNPs still reached genome-wide significance (P < 5 × 10−8) upon meta-analysis of our BE and Levine’s BE + EAC datasets. In a BE-only meta-analysis, the associations improved with the inclusion of our data for 3 out of the 4 SNPs, with 1 (rs2687201, FOXP1) reaching genome-wide significance for BE (Table 2). We then addressed the other 87 other SNPs with Passoc < 10−4 in the Levine data (Supplementary Table 3 of Levine et al12). Of these, 73 were directly genotyped in our samples or could be imputed with an IMPUTE2 info score of >0.95. Of the 10 SNPs that could not be imputed with high quality, only one had Passoc < 10−5 in the original Levine data; we therefore genotyped this SNP (rs11771429) using KASPar in our cases and controls. We did not obtain genotypes for the remaining 9 SNPs. On performing a meta-analysis of the Levine BE + EAC cases with our UK Discovery Phase, 4 SNPs (rs1497205, rs254348, rs3784262, and rs4523255) showed Pmeta < 10−5 and were not strongly correlated with 1 of the 4 BE + EAC SNPs reported previously. We therefore genotyped these 4 SNPs in our Replication Phase samples; rs3784262 (within ALDH1A2) was associated with BE + EAC (OR = 0.90; 95% CI: 0.87–0.93; P = 3.72 × 10−9). No SNP was formally associated with BE alone (Table 2). eQTL and ASE analysis (see Results - In Silico Fine Mapping and Annotation of the Chromosome 2p24 and 12q24 Loci) did not show associations for the rs3784262 proxy rs7165247 in TCGA data or other public data sets (details not shown).

: 0.87–0.93; P = 3.72 × 10−9). No SNP was formally associated with BE alone (Table 2). eQTL and ASE analysis (see Results - In Silico Fine Mapping and Annotation of the Chromosome 2p24 and 12q24 Loci) did not show associations for the rs3784262 proxy rs7165247 in TCGA data or other public data sets (details not shown). Discussion We have added 2 new BE predisposition SNPs, rs3072 on chromosome 2p24 and rs2701108 on chromosome 12q24, to the 2 BE SNPs on chromosome 6p21 (HLA region) and chromosome 16q23 (near FOXF1) that we reported previously.11 Both of the new SNPs remained at or very near genome-wide significance when analysis was restricted to cases with intestinal metaplasia. In silico fine mapping provided evidence that rs3072 and/or 1 of 3 nearby SNPs might be functional because they map to putative enhancer regions. The nearby gene, GDF7, is the best functional candidate, because this encodes the BMP12 protein and the BMP pathway has previously been implicated in the development of BE.36 The importance of this pathway in BE is also suggested in Ingenuity Pathway Analysis, where bone morphogenetic protein 4 acts upstream of proteins encoded by genes close to the BE predisposition SNPs. GDF7 plays a role in the neural system and tendon/ligament development and repair,37,38 and also regulates Hedgehog and Wnt signaling pathways that impact on esophageal development through FOXF1 and TBX5. In the chromosome 12q24 region, rs1920562 (the top imputed SNP) provided the strongest association signal and maps to a possible enhancer. Gene expression analysis did not suggest the target of the chromosome 12q24 variation, although TBX5 is a very strong functional candidate. It is involved in cardiac development and its deficiency causes thoracic malformations and abnormalities of the diaphragmatic musculature,39,40 which could predispose patients to hiatus hernia and acid reflux, 2 subphenotypes of BE.

of the chromosome 12q24 variation, although TBX5 is a very strong functional candidate. It is involved in cardiac development and its deficiency causes thoracic malformations and abnormalities of the diaphragmatic musculature,39,40 which could predispose patients to hiatus hernia and acid reflux, 2 subphenotypes of BE. Messenger RNA expression analysis using TCGA EAC data and public data from leukocytes and adipocytes provided little evidence that rs3072 or rs2701108 (or other SNPs in strong LD) were eQTLs or influenced ASE. The absence of these associations is typical for GWAS or cancer or precancerous traits. Even for the “prototypic” multicancer SNP rs6983267, convincingly demonstrating the effects of SNP alleles on gene expression has required considerable additional work in a variety of systems, and even now, consistent eQTL and ASE associations have not been shown.41–44 The likely major reason for the lack of eQTLs at GWAS SNPs is that the SNPs have their effects in a restricted set of cells or at a particular time. There is, for example, evidence that the forkhead box (FOX) proteins are most strongly expressed during embryogenesis, and that the levels of these transcription factors are critical for proper development.45–47 Given this, our first choice in searching for eQTLs would be cells in the developing human thorax. Unfortunately, such sample collections do not currently exist.

(FOX) proteins are most strongly expressed during embryogenesis, and that the levels of these transcription factors are critical for proper development.45–47 Given this, our first choice in searching for eQTLs would be cells in the developing human thorax. Unfortunately, such sample collections do not currently exist. We showed rs2687201 (FOXP1) to be associated with disease in a BE-only analysis. Our data generally support the report by Levine et al12 of associations between BE + EAC and SNPs on chromosome 3 (FOXP1), chromosome 9 (BARX1), and one of the SNPs on chromosome 19 (CRTC1). However, we were not able to replicate the association observed for another SNP (rs10419226) in CRTC1. For this last SNP, the meta-analysis showed evidence of significant heterogeneity between the BEACON/BEAGESS data and our data (Table 2), and in the absence of clear reasons for this difference, we caution against drawing firm conclusions here. We found another SNP, rs3784262 (ALDH1A2), to be formally associated with BE + EAC upon meta-analysis of our data with the Levine BE + EAC dataset. ALDH1A2 encodes retinaldehyde dehydrogenase 2, which catalyzes the synthesis of retinoic acid and may also be involved in alcohol metabolism.48 Of the candidate SNPs we assessed (Supplementary Table 8), we found supporting evidence, albeit short of genome-wide significance, for rs909253 (P = 3.1 × 10−4), mapping to an intronic region of LTA within the HLA region, but not in LD with rs9257809, the other HLA BE SNP.

may also be involved in alcohol metabolism.48 Of the candidate SNPs we assessed (Supplementary Table 8), we found supporting evidence, albeit short of genome-wide significance, for rs909253 (P = 3.1 × 10−4), mapping to an intronic region of LTA within the HLA region, but not in LD with rs9257809, the other HLA BE SNP. We previously reported that our original GWAS provided evidence that multiple common variants, each with small effects contribute to BE susceptibility.11 Ek et al13 recently estimated that the heritability of BE is 35% (SE 6%). We also found that the heritability of BE is highly significant, but explains only 9.99% of BE risk (SE 1.2%). Our GWAS consisted of UK cases and controls, and the BEACON/BEAGESS samples used by Ek et al originated from 3 continents (Europe, North America, and Australia). Cryptic population stratification could perhaps explain the larger estimate of heritability obtained using the BEACON/BEAGESS GWAS. In addition, we used software to calculate LD-adjusted kinships, such that the SNPs used in the heritability analysis were weighted according to local LD structure. It has been found that heritability estimation from genome-wide SNPs is highly sensitive to uneven LD; causal SNPs in regions of high LD can lead to overestimation of heritability and conversely causal SNPs in regions of low LD can result in an underestimation of heritability.49

eighted according to local LD structure. It has been found that heritability estimation from genome-wide SNPs is highly sensitive to uneven LD; causal SNPs in regions of high LD can lead to overestimation of heritability and conversely causal SNPs in regions of low LD can result in an underestimation of heritability.49 Although the BE GWAS have not yet identified the functional SNPs in each region or their gene targets, the information generated already permits the generation of hypotheses regarding processes that may be involved in BE. First, transcription factors involved in development and structure of the thorax, diaphragm, and esophagus may be important: the SNPs near FOXF1, FOXP1, BARX1, and TBX5 might act in this way and the genes appear to be functionally related (Supplementary Figure 1). Second, the inflammatory response may be important: the SNPs within the HLA region (rs9257809 and, perhaps, rs909253) might act in this way and pathway analysis provided suggestive evidence of a role for type 1 diabetes genes in BE etiology. A plausible, testable hypothesis is that these 2 groups of SNPs respectively influence the tendency to gastroesophageal reflux disease, perhaps through thoracic and diaphragmatic structure (hiatal hernia defect), and the inflammatory response to the refluxed gastric acid. Given the limited scope for clinical intervention in the former processes, we await with interest the outcome of trials such as AspECT that target the inflammatory response to gastric reflux.1,17 Supplementary Material Supplementary Data

Although the BE GWAS have not yet identified the functional SNPs in each region or their gene targets, the information generated already permits the generation of hypotheses regarding processes that may be involved in BE. First, transcription factors involved in development and structure of the thorax, diaphragm, and esophagus may be important: the SNPs near FOXF1, FOXP1, BARX1, and TBX5 might act in this way and the genes appear to be functionally related (Supplementary Figure 1). Second, the inflammatory response may be important: the SNPs within the HLA region (rs9257809 and, perhaps, rs909253) might act in this way and pathway analysis provided suggestive evidence of a role for type 1 diabetes genes in BE etiology. A plausible, testable hypothesis is that these 2 groups of SNPs respectively influence the tendency to gastroesophageal reflux disease, perhaps through thoracic and diaphragmatic structure (hiatal hernia defect), and the inflammatory response to the refluxed gastric acid. Given the limited scope for clinical intervention in the former processes, we await with interest the outcome of trials such as AspECT that target the inflammatory response to gastric reflux.1,17 Supplementary Material Supplementary Data Acknowledgments The authors would like to thank the WTCCC2 consortium, the BEACON consortium and the AspECT, BOSS, and ChOPIN trial teams (comprising the EAGLE consortium) and participants in all these studies. The Discovery Phase and Immunochip replication were funded by the Wellcome Trust IPOD grant (084722/Z/08/Z). In addition, sample collection for all phases was core supported by the Cancer Research UK funded AspECT, ChOPIN and Handel trials. CoRGI and GLACIER were also funded by Cancer Research UK. In addition, the authors thank AstraZeneca for an educational grant for the tissue and blood collection. Core funding to the Wellcome Trust Centre for Human Genetics was provided by the Wellcome Trust (090532/Z/09/Z). We also thank Liam J. Murray and Wong-Ho Chow, who provided samples as part of the BEACON consortium. The authors would also like to thank the Experimental Cancer Medicine Centre–supported University of Southampton Faculty of Medicine Tissue Bank. The Cancer Genome Atlas (TCGA) esophageal carcinoma data were also gratefully used for expression analyses. Finally, the authors want to thank Dr Vincent Plagnol of University College London for advice on statistical analyses and presentation.

Centre–supported University of Southampton Faculty of Medicine Tissue Bank. The Cancer Genome Atlas (TCGA) esophageal carcinoma data were also gratefully used for expression analyses. Finally, the authors want to thank Dr Vincent Plagnol of University College London for advice on statistical analyses and presentation. Conflicts of interest This author discloses the following: Janusz Jankowski is Chief Investigator of AspECT and ChOPIN trials and was an AstraZeneca consultant from 2002 to 2012. The remaining authors disclose no conflicts. Funding This work was supported by Cancer Research UK (Chopin Grant C548, AsPECT Grant A4584, AZ Educational Grant JJ2), Peninsula Schools of Medicine and Dentistry School (Grant code - ‘PUPSMED -JJ2013-1’), Wellcome Trust (084722/Z/08/Z and 090532/Z/09/Z), and an AstraZeneca UK educational grant. Author names in bold designate shared co-first authorship.

Funding This work was supported by Cancer Research UK (Chopin Grant C548, AsPECT Grant A4584, AZ Educational Grant JJ2), Peninsula Schools of Medicine and Dentistry School (Grant code - ‘PUPSMED -JJ2013-1’), Wellcome Trust (084722/Z/08/Z and 090532/Z/09/Z), and an AstraZeneca UK educational grant. Author names in bold designate shared co-first authorship. Figure 1 Outline of the phases of this study and the SNPs analyzed. Two SNPs described in Su et al11 had previously been genotyped in Replication Phase 2 and BEACON/BEAGESS samples. All other replication phase 3 samples are new to this study, as is the genotyping of additional SNPs in phases 2 and 3. Dutch Replication (Phase 1 Replication) and the Dutch extension (Phase 3 Replication) is one cohort in our analyses for the SNPs taken through to Replication Phase 3. ∗11 SNPs: Our SNPs: rs3072, rs6751791, rs2731672, rs2701108, rs189247, rs2043633, and rs12985909 and Levine et al12 SNPs: rs1497205, rs254348, rs3784262, and rs4523255. +8 SNPs: Our SNPs: rs3072, rs6751791, rs2731672, rs2701108, rs189247, rs2043633, and Levine et al SNP: rs3784262. Δ7 SNPs: Our SNPs: rs3072, rs6751791, rs2731672, rs2701108, rs189247, rs2043633. Figure 2 Regional plots of association (left y-axis) and recombination rates (right y-axis) for the chromosomes 2p24 and 12q24 loci after imputation. The lead genotyped SNP is marked with a purple square. Imputed SNPs are plotted as circles and genotyped SNPs as squares. Table 1 Meta-analysis of Discovery and Replication Phase Sample Sets for SNPs Taken Into Replication Phase 3

Figure 2 Regional plots of association (left y-axis) and recombination rates (right y-axis) for the chromosomes 2p24 and 12q24 loci after imputation. The lead genotyped SNP is marked with a purple square. Imputed SNPs are plotted as circles and genotyped SNPs as squares. Table 1 Meta-analysis of Discovery and Replication Phase Sample Sets for SNPs Taken Into Replication Phase 3 SNP Chr Position (build 37) Minor/ major Discovery MAF (cases /controls) Discovery Replication phase 1 (rep 1) Meta 1 (discovery + rep 1) Replication phase 2 (rep 2) Meta 2 (meta 1+ rep 2) Replication phase 3 (rep 3) Final meta (meta 2 + rep 3) Final meta I2 N rs3072 2 20878406 G/A 0.41/0.36 1.23 (1.14–1.33) 2.64 × 10−7 1.10 (1.02–1.19) 1.33 × 10−2 1.16 (1.10–1.23) 8.13 × 10−8 1.13 (1.03–1.24) 7.00 × 10−3 1.16 (1.10–1.21) 2.27 × 10−9 1.11 (1.04–1.17) 1.02 × 10−3 1.14 (1.09–1.18) 1.75 × 10−11 0.42 8 rs6751791 2 35581997 A/G 0.51/0.48 1.15 (1.06–1.23) 5.03 × 10−4 1.16 (1.07−1.25) 1.45 × 10−4 1.15 (1.09−1.21) 2.68 × 10−7 1.07 (0.97−1.16) 1.64 × 10−1 1.13 (1.08−1.18) 2.91 × 10−7 0.99 (0.93−1.05) 7.99 × 10−1 1.08 (1.04−1.12) 7.65 × 10−5 0.60 8 rs2731672 5 176842474 A/G 0.27/0.24 1.18 (1.09−1.28) 1.64 × 10−4 1.14 (1.04−1.24) 3.87 × 10−3 1.16 (1.09−1.23) 2.54 × 10−6 1.09 (0.99−1.21) 8.20 × 10−2 1.14 (1.08−1.20) 8.33 × 10−7 0.95 (0.89−1.02) 1.81 × 10−1 1.07 (1.03−1.12) 1.66 × 10−3 0.63 8 rs2701108 12 114674261 G/A 0.38/0.41 0.88 (0.81–0.95) 1.00 × 10−3 0.87 (0.81–0.94) 4.40 × 10−4 0.87 (0.83–0.92) 1.51 × 10−6 0.89 (0.81–0.97) 1.10 × 10−2 0.88 (0.84–0.92) 5.68 × 10−8 0.93 (0.87–0.99) 1.42 × 10−2 0.90 (0.86–0.93) 7.48 × 10−9 0.14 8 rs189247 15 97586630 A/G 0.41/0.37 1.18 (1.09−1.27) 5.67 × 10−5 1.14 (1.05−1.23) 1.25 × 10−3 1.15 (1.09−1.22) 2.91 × 10−7 1.10 (1.00−1.20) 4.90 × 10−2 1.14 (1.09−1.19) 6.12 × 10−8 0.96 (0.90−1.02) 1.73 × 10−1 1.10 (1.06−1.14) 3.55 × 10−7 0.20 8 rs2043633 16 5819274 C/A 0.37/0.41 0.85 (0.79−0.92) 6.04 × 10−5 0.88 (0.82−0.95) 9.80 × 10−4 0.87 (0.82−0.92) 2.49 × 10−7 0.88 (0.80−0.96) 5.00 × 10−3 0.87 (0.83−0.91) 4.74 × 10−9 0.99 (0.94−1.05) 8.39 × 10−1 0.92 (0.88−0.95) 2.25 × 10−6 0.58 8 rs12985909 19 18439383 G/A 0.48/0.45 1.12 (1.04−1.21) 2.94 × 10−3 1.12 (1.04−1.21) 2.73 × 10−3 1.12 (1.06−1.18) 2.45 × 10−5 1.11 (1.01−1.21) 2.70 × 10−2 1.12 (1.07−1.17) 1.99 × 10−6 1.07 (1.01−1.13) 2.63 × 10−2 1.10 (1.06−1.14) 3.28 × 10−7 0.00 8 NOTE. For each phase, association data show (top to bottom) OR, (95% CI), and Passoc. Results are presented with respect to the minor allele.

12 (1.04−1.21) 2.73 × 10−3 1.12 (1.06−1.18) 2.45 × 10−5 1.11 (1.01−1.21) 2.70 × 10−2 1.12 (1.07−1.17) 1.99 × 10−6 1.07 (1.01−1.13) 2.63 × 10−2 1.10 (1.06−1.14) 3.28 × 10−7 0.00 8 NOTE. For each phase, association data show (top to bottom) OR, (95% CI), and Passoc. Results are presented with respect to the minor allele. rs3072 and rs2701108 reached genome-wide significance and thus are shown in bold. In BEACON, rs7598399 was used as a proxy for rs6751791 (r2 = 1) and rs189247 was imputed from 4 genotyped SNPs (rs991757, rs2670927, rs2670930, and rs234540) with accuracy approximately 98%. The Dutch extension samples were analyzed with the previously described Dutch replication samples as part of Rep 1. The P value threshold for including a SNP in Phase 2 was 1 × 10−4 and that for inclusion in Phase 3 was 5 × 10−6. Chr, chromosome; I2, I2 heterogeneity index; MAF, minor allele frequency; N, number of studies. Table 2 Meta-analysis With Our Data for 4 BE/EAC SNPs and 4 Other Selected SNPs with P < 1 × 10−4 from Levine et al12

rs3072 and rs2701108 reached genome-wide significance and thus are shown in bold. In BEACON, rs7598399 was used as a proxy for rs6751791 (r2 = 1) and rs189247 was imputed from 4 genotyped SNPs (rs991757, rs2670927, rs2670930, and rs234540) with accuracy approximately 98%. The Dutch extension samples were analyzed with the previously described Dutch replication samples as part of Rep 1. The P value threshold for including a SNP in Phase 2 was 1 × 10−4 and that for inclusion in Phase 3 was 5 × 10−6. Chr, chromosome; I2, I2 heterogeneity index; MAF, minor allele frequency; N, number of studies. Table 2 Meta-analysis With Our Data for 4 BE/EAC SNPs and 4 Other Selected SNPs with P < 1 × 10−4 from Levine et al12 SNP Chr Position Nearby genes Minor/ major allele BE+/−EAC Levine et al meta OR (95% CI) PLevineet al This study meta OR (95% CI) PThis Study This study + Levine et al12 meta OR (95% CI) Pmeta Meta I2 No. of studies rs2687201 3 70928930 FOXP1 T/G BE 1.18 (1.10–1.26) 2.00 × 10−6 1.14 (1.03–1.27) 1.18 × 10−2 1.16 (1.10−1.23) 4.61 × 10−8 0.00 3 BE/EAC 1.18 (1.12−1.25) 5.47 × 10−9 1.17 (1.11−1.23) 6.70 × 10−10 0.00 3 rs11789015 9 96716028 BARX1 G/A BE 0.85 (0.79−0.91) 5.08 × 10−6 0.90 (0.81−1.01) 6.63 × 10−2 0.86 (0.81−0.92) 1.38 × 10−6 0.00 3 BE/EAC 0.83 (0.79−0.88) 1.02 × 10−9 0.85 (0.81−0.89) 1.14 × 10−10 0.00 3 rs10419226 19 18803172 CRTC1 A/C BE 1.19 (1.12−1.26) 5.54 × 10−8 1.01 (0.91−1.11) 8.65 × 10−1 1.13 (1.08−1.20) 2.14 × 10−6 0.82 3 BE/EAC 1.18 (1.12−1.24) 3.55 × 10−10 1.14 (1.09−1.19) 1.17 × 10−8 0.82 3 rs10423674 19 18817903 CRTC1 T/G BE 0.85 (0.80−0.91) 1.92 × 10−6 0.94 (0.88−1.00) 4.88 × 10−2 0.89 (0.85−0.93) 2.99 × 10−7 0.40 5 BE/EAC 0.84 (0.80−0.89) 1.75 × 10−9 0.88 (0.84−0.91) 4.87 × 10−11 0.49 5 rs1497205 4 76169067 PARM1, RCHY1 C/T BE 0.86 (0.80−0.92) 2.86 × 10−5 0.92 (0.87−0.98) 7.59 × 10−1 0.90 (0.86−0.94) 2.57 × 10−6 0.00 6 BE/EAC 0.87 (0.82−0.93) 1.28 × 10−5 0.90 (0.86−0.94) 3.68 × 10−7 0.00 6 rs254348 16 65980789 T/C BE 0.88 (0.83−0.94) 1.15 × 10−4 0.95 (0.91−1.01) 8.88 × 10−2 0.93 (0.89−0.97) 5.49 × 10−4 0.53 6 BE/EAC 0.89 (0.84−0.94) 1.40 × 10−5 0.92 (0.89−0.96) 2.81 × 10−5 0.53 6 rs3784262 15 58253106 ALDH1A2 G/A BE 0.85 (0.80−0.90) 3.62 × 10−7 0.93 (0.89−0.98) 5.13 × 10−3 0.91 (0.87−0.94) 1.37 × 10−6 0.12 9 BE/EAC 0.88 (0.83−0.92) 6.72 × 10−7 0.90 (0.87−0.93) 3.72 × 10−9 0.16 9 rs4523255 8 8713038 MFHAS1 A/G BE 1.13 (1.06−1.21) 2.46 × 10−4 1.07 (1.01−1.12) 2.11 × 10−2 1.09 (1.05−1.14) 2.48 × 10−5 0.36 6 BE/EAC 1.13 (1.07−1.20) 4.15 × 10−5 1.09 (1.05−1.14) 9.24 × 10−6 0.46 6 NOTE. The minimum meta-analysis comprised the Levine et al12 Discovery and Replication Phases and our Discovery Phase (with amended controls, as described in Methods). rs10423674 was additionally genotyped in our UK and Dutch Replication Phase 1.

6 BE/EAC 1.13 (1.07−1.20) 4.15 × 10−5 1.09 (1.05−1.14) 9.24 × 10−6 0.46 6 NOTE. The minimum meta-analysis comprised the Levine et al12 Discovery and Replication Phases and our Discovery Phase (with amended controls, as described in Methods). rs10423674 was additionally genotyped in our UK and Dutch Replication Phase 1. rs1497205, rs254348, rs3784262, and rs4523255 were genotyped in our UK and Dutch Replication Phase 1 and UK Replication Phase 2. rs3784262 was also genotyped in Irish Replication Phase 2 samples and UK and Belgium Replication Phase 3 samples. For rs10419226, which shows evidence of inter-study heterogeneity, random effects model P values for BE and BE/EAC are .10 and .04, respectively. Chr, chromosome.

Rudolf Valenta Heidrun Hochwallner Birgit Linhart Sandra Pahr There are several mechanisms by which people develop adverse reactions to foods also termed food intolerance.1 These reactions can be considered toxic or nontoxic (Figure 1).2 Among the nontoxic reactions, those that are not immune-mediated, such as those involving enzyme defects (eg, vasoactive amines) or reactions to certain substances (eg lactose intolerance), are far more common than immune-mediated reactions.2 Nevertheless, immune-mediated reactions affect millions of people, are responsible for significant morbidity and health care costs, and can cause severe life-threatening reactions that lead to death.3–5 Food allergy was defined by an expert panel of the National Institute of Allergy and Infectious Diseases as “an adverse health effect arising from a specific immune response that occurs reproducibly on exposure to a given food.” This response comprises basically all types of immune-mediated reactions, including those caused by the adaptive and innate immune system (Figure 1).6

Institute of Allergy and Infectious Diseases as “an adverse health effect arising from a specific immune response that occurs reproducibly on exposure to a given food.” This response comprises basically all types of immune-mediated reactions, including those caused by the adaptive and innate immune system (Figure 1).6 The term allergy was coined in 1906 by the Austrian pediatrician Clemens von Pirquet,7 who described cases of serum sickness in children treated with antibody preparations. According to Coombs and Gell,8 there are 4 major types of allergic reactions based on pathogenesis mechanisms. The most common forms of immune-mediated adverse reactions to foods (type I reactions) always are characterized by the development of IgE against food allergens. It can be accompanied by inflammation, induced by cellular components, and mediated by T cells and eosinophils. Patients with IgE-associated food allergy can be identified based on the detection of food allergen–specific IgE in serum and body fluids, and by measuring IgE-mediated cellular and in vivo responses.4 Although it is tempting to speculate that food antigen–specific IgG can cause adverse reactions via type II or type III hypersensitivity, there is no solid experimental evidence to support the relevance of these reactions to food allergies that develop in patients (Figure 1). Accordingly, several position papers strongly recommend against testing for food antigen–specific IgG in the diagnosis of food allergy.9,10

via type II or type III hypersensitivity, there is no solid experimental evidence to support the relevance of these reactions to food allergies that develop in patients (Figure 1). Accordingly, several position papers strongly recommend against testing for food antigen–specific IgG in the diagnosis of food allergy.9,10 Type IV hypersensitivity, which mainly involves food antigen–specific T-cell responses and can damage the gut mucosa, is associated with disorders such as celiac disease. Celiac disease is characterized by a hypersensitivity reaction against the wheat gluten fraction comprising alcohol soluble gliadins and acid-, alkali-soluble glutenins, accompanied by an autoimmune component.11 Type IV hypersensitivity reactions also might be involved in food protein–induced enterocolitis (Figure 1).12 Studies have shown that certain food proteins can induce inflammation via direct activation of the innate immune system. For example, wheat amylase trypsin inhibitors and certain milk oligosaccharides can cause intestinal inflammation via activation of Toll-like receptor 4,13,14 and certain allergens have been shown to stimulate the innate immune system.15 Innate immune mechanisms might mediate nonceliac gluten sensitivity.16

une system. For example, wheat amylase trypsin inhibitors and certain milk oligosaccharides can cause intestinal inflammation via activation of Toll-like receptor 4,13,14 and certain allergens have been shown to stimulate the innate immune system.15 Innate immune mechanisms might mediate nonceliac gluten sensitivity.16 In developed countries, IgE-associated food allergy affects 3%–8% of children and 1%–3% of adults.3–5 It not only is common, but often is a serious and life-threatening health condition that requires accurate diagnosis and has strong effects on an individual’s dietary habits and social life. Milk, eggs, wheat, peanuts, nuts, sesame, fish, fruits, and vegetables are common inducers of IgE-associated food allergy.4 Allergies to foods such as milk, egg, and wheat often are outgrown (patients acquire tolerance), whereas allergies to peanuts, tree nuts, and fish allergies often persists over a lifetime.3 The exact incidence of food allergies has not been fully established because there are discrepancies among findings from studies in which food allergies were self-reported vs those diagnosed by various assays (eg, provocation, skin test, or serologic tests).4,17

nuts, and fish allergies often persists over a lifetime.3 The exact incidence of food allergies has not been fully established because there are discrepancies among findings from studies in which food allergies were self-reported vs those diagnosed by various assays (eg, provocation, skin test, or serologic tests).4,17 The prevalence and severity of food allergies seem to be increasing. In addition to genetic factors, a number of environmental, cultural, and behavioral factors affect the frequency, severity, and type of allergic manifestations in patients.18–21 A recent study identified epigenetic differences in CD4+ T cells from children with IgE-mediated food allergies, compared with children without food allergies—differences such as these might contribute to the development of a food allergy.22 According to the hygiene hypothesis, decreases in family size and improvements in personal hygiene have contributed to the increased prevalence of IgE-mediated allergies. On the other hand, factors such as an anthroposophic lifestyle (eating organic foods that contain lactobacilli and restrictive use of antibiotics, antipyretics, and vaccines) have been associated with a reduced incidence of allergies.23,24 It has been proposed that insufficient exposure to dietary and bacterial metabolites might have contributed to increases in inflammatory disorders in Western countries.25

contain lactobacilli and restrictive use of antibiotics, antipyretics, and vaccines) have been associated with a reduced incidence of allergies.23,24 It has been proposed that insufficient exposure to dietary and bacterial metabolites might have contributed to increases in inflammatory disorders in Western countries.25 Allergic Sensitization and Secondary Immune Responses The term allergic sensitization describes the first induction of an allergic immune response upon allergen encounter.26,27 Two routes of allergic sensitization are well established (Figure 2A). Class 1 food allergens (eg, milk, egg, or peanut) are oral allergens that cause sensitization via the gastrointestinal tract.28 Class 2 food allergens are aeroallergens (eg, major birch pollen allergen Bet v 1) that cause sensitization via the respiratory tract. Immune responses against these allergens can cross-react with homologous food allergens (eg, major apple allergen Mal d 1) to cause symptoms.29–31 It recently was proposed that people become sensitized to food allergens via skin contact, but there have been few studies of this process.18 Interestingly, studies of animal models have indicated that epicutaneous sensitization leads to expansion of IgE-dependent intestinal mast cells and food-induced allergic reactions.32,33 For an overview of food allergen sources that may cause sensitization via the respiratory tract and skin, see the article by Asero and Antonicelli.34

tudies of animal models have indicated that epicutaneous sensitization leads to expansion of IgE-dependent intestinal mast cells and food-induced allergic reactions.32,33 For an overview of food allergen sources that may cause sensitization via the respiratory tract and skin, see the article by Asero and Antonicelli.34 Determinants of allergic sensitization include features of the epithelial barrier, the allergen itself (whether allergens are stable and not degraded in the environment or gastrointestinal tract), nonallergenic components of the food matrix, and substances that act as adjuvants (Figure 2A).35 For example, food allergens have been proposed to have greater stability during digestion than other molecules in food.36 Intrinsic factors (eg, genetic factors such as mutations in the filaggrin gene) and exogenous factors (eg, alcohol, anti-inflammatory drugs, pathogens, or stress) have been proposed to reduce the barrier function of the intestinal epithelium and facilitate sensitization.37–39 On the other hand, secretory antibodies, particularly secretory IgA (SIgA), have important roles in reinforcing the epithelial barrier. Mice deficient in SIgA and secretory IgM are prone to develop food allergen–induced anaphylactic shock, which can be overcome by induction of tolerance with T-regulatory cells.40,41

39 On the other hand, secretory antibodies, particularly secretory IgA (SIgA), have important roles in reinforcing the epithelial barrier. Mice deficient in SIgA and secretory IgM are prone to develop food allergen–induced anaphylactic shock, which can be overcome by induction of tolerance with T-regulatory cells.40,41 Many environmental and genetic factors contribute to the atopic predisposition of individuals. These determine their susceptibility to develop allergic immune responses against allergens.42 In atopic individuals who have a predisposition toward developing IgE-associated allergies, encounters with allergen activate, after processing by antigen-presenting cells (eg, dendritic cells or B cells), allergen-specific T-helper 2 (Th2) cells, which produce cytokines such as interleukin (IL)4 and IL13. These cytokines induce class switching and production of allergen-specific IgE.43–45 Primary allergic sensitization (such as a class switch toward IgE production) occurs early in life and leads to T-cell and IgE memory, which can be boosted with repeated allergen contact (secondary immune response).46–49 Upon contact with a primary food allergen, nonallergic individuals produce allergen-specific IgG and IgA, which do not induce allergic reactions. The formation of food allergen–specific IgE is a main feature of IgE-associated food allergy and its diagnosis.

th repeated allergen contact (secondary immune response).46–49 Upon contact with a primary food allergen, nonallergic individuals produce allergen-specific IgG and IgA, which do not induce allergic reactions. The formation of food allergen–specific IgE is a main feature of IgE-associated food allergy and its diagnosis. Analyses of the time courses of allergic sensitization to respiratory and food allergen sources in large birth cohort studies have shown that food allergies and their associated symptoms develop before respiratory allergies.50 In later life, there is a reverse trend—food allergies often are outgrown and respiratory allergies increase and dominate.50 Interestingly, the prevalence of food allergies is approximately 10-fold lower than that of respiratory allergies.4,51 This could be because oral exposure to allergens activates tolerance mechanisms (via regulatory T cells) and less frequently results in allergic sensitization than respiratory exposure to allergens.52,53

terestingly, the prevalence of food allergies is approximately 10-fold lower than that of respiratory allergies.4,51 This could be because oral exposure to allergens activates tolerance mechanisms (via regulatory T cells) and less frequently results in allergic sensitization than respiratory exposure to allergens.52,53 Several cellular mechanisms seem to influence primary allergic sensitization vs tolerance in the intestine. Tolerance can be mediated by antigen presentation by dendritic cells, which interact with C-type lectin receptors54; dendritic cell–bound IgE can down-regulate allergic inflammation at mucosal sites.55 Children with an egg allergy were reported to have reduced function of neonatal T-regulatory cells compared with children without an egg allergy.56 On the other hand, children who outgrew a cow’s milk allergy had increased T-regulatory cell responses.57 These findings indicate that T-regulatory cells modulate the development of food allergies.58 After primary sensitization, the allergic immune response is boosted with repeated exposure to allergen, increasing activation of allergen-specific T cells and production of IgE.

Several cellular mechanisms seem to influence primary allergic sensitization vs tolerance in the intestine. Tolerance can be mediated by antigen presentation by dendritic cells, which interact with C-type lectin receptors54; dendritic cell–bound IgE can down-regulate allergic inflammation at mucosal sites.55 Children with an egg allergy were reported to have reduced function of neonatal T-regulatory cells compared with children without an egg allergy.56 On the other hand, children who outgrew a cow’s milk allergy had increased T-regulatory cell responses.57 These findings indicate that T-regulatory cells modulate the development of food allergies.58 After primary sensitization, the allergic immune response is boosted with repeated exposure to allergen, increasing activation of allergen-specific T cells and production of IgE. In persons with a respiratory allergy, the IgE response is boosted by contact with a mucosal allergen48 and, interestingly, does not seem to require T-cell help.59–61 Another interesting feature of the established secondary IgE response is that in adults with an allergy, the profile of allergens recognized by IgE does not change substantially, whereas it seems that young children can be sensitized to new allergens.62,63 In the case of a respiratory allergy, allergen contact through the respiratory mucosa strongly boosts IgE production, but has little influence on the other classes of allergen-specific antibodies (eg, IgA or IgG).48 The responding B ε memory cells may reside in the respiratory mucosa or the adjacent lymphoid tissues,64 but little is known about the precise location of the cells involved in secondary IgE responses in allergic patients.65

t has little influence on the other classes of allergen-specific antibodies (eg, IgA or IgG).48 The responding B ε memory cells may reside in the respiratory mucosa or the adjacent lymphoid tissues,64 but little is known about the precise location of the cells involved in secondary IgE responses in allergic patients.65 The mechanisms by which food allergen–specific IgE responses are boosted in patients with food allergies are poorly understood. When food allergens were administered orally, patients had strong increases in the production of allergen-specific IgG, accompanied by an initial boost of IgE.66–68 These findings indicated that allergen ingestion can boost allergen-specific production of IgG as well as IgE. It is possible that oral allergens can boost production of disease-causing IgE, as well as that of potentially protective IgG; this might explain why elimination or continued intake of food allergens can benefit patients.69,70

dicated that allergen ingestion can boost allergen-specific production of IgG as well as IgE. It is possible that oral allergens can boost production of disease-causing IgE, as well as that of potentially protective IgG; this might explain why elimination or continued intake of food allergens can benefit patients.69,70 It is clear that avoidance of food allergens over a prolonged period of time reduces levels of allergen-specific IgE below the threshold level for symptoms (Figure 2B). However, there is controversy about whether intake of food allergens is beneficial. If food allergen intake mainly stimulates production of protective IgG antibodies, tolerance could be induced and allergen-specific IgE production could be reduced. However, insufficient IgG production could cause IgE levels to increase, leading to increased sensitivity and symptoms (Figure 2B). Several recent studies have shown that the induction of allergen-specific IgG antibodies, which block IgE recognition of food allergens, is associated with the successful immunotherapy for food allergy.68

cient IgG production could cause IgE levels to increase, leading to increased sensitivity and symptoms (Figure 2B). Several recent studies have shown that the induction of allergen-specific IgG antibodies, which block IgE recognition of food allergens, is associated with the successful immunotherapy for food allergy.68 Pathogenesis and Manifestations of Food Allergy Upon interaction with food antigens, IgE becomes cross-linked and binds to mast cells and basophils via the high-affinity receptor FcεRI (Figure 2C).71 This process activates these cells, leading to the release of granules that contain preformed inflammatory mediators (eg, histamine), as well as de novo synthesis and/or release of inflammatory mediators (eg, leukotrienes), proteases (eg, tryptase), inflammatory cytokines (eg, IL4), and chemotactic molecules. Mast cells and basophils are activated within a few minutes of IgE cross-linking, therefore this process it called an immediate allergic reaction; symptoms occur shortly after allergen contact. Because food allergens enter the blood via the gastrointestinal tract, symptoms can develop directly at the sites of allergen contact (eg, mouth, esophagus, and/or intestine), or in other organs. Systemic reactions occur when allergens capable of cross-linking effector cell-bound IgE pass the barrier of the mucosa into the circulation (Figure 2C, right). Allergen uptake also may affect the circulatory and nervous systems.

he sites of allergen contact (eg, mouth, esophagus, and/or intestine), or in other organs. Systemic reactions occur when allergens capable of cross-linking effector cell-bound IgE pass the barrier of the mucosa into the circulation (Figure 2C, right). Allergen uptake also may affect the circulatory and nervous systems. Factors that contribute to the type and severity of reactions include the amount of ingested allergen, the stability of the allergen against digestion, and the permeability of the epithelial barrier (Figure 2C). The immediate allergic reaction leads to intense inflammation that can become life-threatening. The release of vasoactive mediators into the circulation can lead to vascular collapse and anaphylactic shock.72 Supplementary Table 1 summarizes the clinical manifestations of food allergies, the organ systems affected by IgE-mediated mast cell and basophil degranulation, and the clinical aspects of gastrointestinal food allergy.3,6

vasoactive mediators into the circulation can lead to vascular collapse and anaphylactic shock.72 Supplementary Table 1 summarizes the clinical manifestations of food allergies, the organ systems affected by IgE-mediated mast cell and basophil degranulation, and the clinical aspects of gastrointestinal food allergy.3,6 Studies performed with well-defined reagents (eg, monoclonal IgE, in vitro cellular systems, defined allergens, and IgE epitopes) have shown that the degranulation of effector cells, and therefore the intensity of the immediate-type reaction, increases with the number of IgE epitopes on an allergen, high levels of allergen-specific IgE, and high-affinity allergen-specific IgE.73–75 The fact that high levels of allergen-specific IgE cause up-regulation of FcεRI on mast cells and basophils, and thereby a more dense loading of these cells with IgE, could account for the association between levels of allergen-specific IgE against stable food allergens and the severity of allergic reactions.76–78 In addition to the levels of FcεRI on mast cells, the number of intestinal mast cells and basophils and (probably related to mast cell numbers and activation) the levels of intestinal tryptase also seem to be related to the severity of reactions to food allergens.79–81 Interestingly, studies performed in animal models have shown that cytokines such as IL4 can induce expansion of intestinal mast cells.82

and basophils and (probably related to mast cell numbers and activation) the levels of intestinal tryptase also seem to be related to the severity of reactions to food allergens.79–81 Interestingly, studies performed in animal models have shown that cytokines such as IL4 can induce expansion of intestinal mast cells.82 In addition to the immediate allergic reaction (the most frequent pathogenic mechanism of IgE-associated allergies), late-phase allergic reactions also occur after allergen contact; there are 2 types. The late-phase response to allergens has been studied in mainly cutaneous models, such as skin blister and skin chamber models. Several hours after allergen contact and the immediate reaction, there is an influx of basophils and eosinophils.83 This influx is steroid-sensitive and seems to involve granulocyte-macrophage colony-stimulating factor.84 Relatively little is known about the importance of late-phase reactions in food allergy, but it is tempting to speculate that they could be involved in food allergen–induced forms of eosinophilic gastroenteritis.85 In fact, data from experimental animal models have indicated that thymic stromal lymphopoetin–induced basophil responses promote eosinophilic esophagitis.86 Interestingly, it also has been shown that enteric eosinophils not only contribute to inflammation, but control dendritic cells to initiate primary Th2 cell–mediated immune responses, indicating a complex interaction among cells in food allergies.87

opoetin–induced basophil responses promote eosinophilic esophagitis.86 Interestingly, it also has been shown that enteric eosinophils not only contribute to inflammation, but control dendritic cells to initiate primary Th2 cell–mediated immune responses, indicating a complex interaction among cells in food allergies.87 In addition to the late-phase responses, delayed-type reactions can occur 24–48 hours after allergen contact. These resemble features of a type IV hypersensitivity reaction, involving allergen-specific T cells. Allergen-specific T cells can be activated via IgE-dependent and IgE-independent pathways (Figure 2). In fact, in patients with allergies, antigen-presenting cells express FcεRI as well as the low-affinity receptor for IgE (FcεRII also known as CD23). The cells use this receptor for IgE-facilitated allergen presentation—a process found to be more effective for T-cell activation than allergen presentation without IgE.88,89 Studies performed with allergen peptides that do not react with IgE and recombinant allergen derivatives showed that activation of allergen-specific T cells also can occur without IgE, and lead to delayed-type allergic reactions in patients.90,91 Induction of atopic dermatitis by food allergens has been shown to require not only Th2 cells (and Th2 cytokines such as IL4, IL13, and IL5—a cytokine that activates eosinophils), but also Th1 cells, which mediated delayed allergic inflammation.92 Interferon-γ, secreted by allergen-specific Th1 cells, was shown to induce epithelial damage in a model of respiratory allergy.93

shown to require not only Th2 cells (and Th2 cytokines such as IL4, IL13, and IL5—a cytokine that activates eosinophils), but also Th1 cells, which mediated delayed allergic inflammation.92 Interferon-γ, secreted by allergen-specific Th1 cells, was shown to induce epithelial damage in a model of respiratory allergy.93 The immediate- and delayed-type allergic inflammation that occurs during IgE-associated food allergy has been studied extensively in patients with oral allergy syndrome (OAS) (Supplementary Table 1). OAS is caused by sensitization to respiratory allergens that structurally are similar to allergens in foods, leading to a cross-reactive immune response. The most common form of OAS develops with sensitization to the major birch pollen allergen, Bet v 1. In patients with this form of OAS, the immune response cross-reacts with allergens in plant-derived food such as apples, nuts, carrots, and celery.94 This leads to local allergy symptoms of the immediate type (local itching and swelling of the lips or tongue), caused by IgE-mediated mast cell degranulation.95 However, Bet v 1–related plant food allergens are digested by the gastrointestinal tract, therefore systemic reactions occur only in exceptional cases (such as after consumption of large amounts of the plant food allergens during exercise), and anaphylactic shock does not occur.96 Likewise, cooking destroys IgEs, but leaves peptides recognized by allergen-specific T cells intact.97 Ingestion of Bet v 1–related plant food allergens therefore can activate allergen-specific T cells by IgE-independent mechanisms, and induce late-phase and chronic allergic inflammation to cause disorders such as atopic dermatitis in sensitized patients.98

aves peptides recognized by allergen-specific T cells intact.97 Ingestion of Bet v 1–related plant food allergens therefore can activate allergen-specific T cells by IgE-independent mechanisms, and induce late-phase and chronic allergic inflammation to cause disorders such as atopic dermatitis in sensitized patients.98 Information on the time until onset of allergic reactions after food allergen intake (minutes–hours vs hours–days) and allergy phenotype (eg, urticaria vs atopic dermatitis) can help to determine whether the symptoms involve immediate IgE-mediated mast cell or basophil activation, or late-phase or chronic allergic inflammation caused by T-cell or eosinophil activation. Then, it is possible to select the most appropriate therapy. For example, immediate reactions would be treated with antihistamines, antileukotrienes, epinephrine, mast cell stabilizers, or anti-IgE, whereas late-phase chronic inflammation would be treated with steroids or anti-IL5.

d by T-cell or eosinophil activation. Then, it is possible to select the most appropriate therapy. For example, immediate reactions would be treated with antihistamines, antileukotrienes, epinephrine, mast cell stabilizers, or anti-IgE, whereas late-phase chronic inflammation would be treated with steroids or anti-IL5. Preventing Food Allergies According to the current food allergy and anaphylaxis guidelines of the European Academy of Allergy and Clinical Immunology,99 there are no restrictions regarding diet for mothers during pregnancy and lactation. Exclusive breastfeeding is recommended for the first 4–6 months of life, which prevents the development of allergies. If breastfeeding is not possible, hypoallergenic formulas with documented preventive effects are recommended for high-risk children. Breastfeeding transfers protective SIgA to the child, which may prevent allergic sensitization, and avoids early exposure to potential food allergens. This process may involve the uptake of SIgA-allergen complexes via receptors on M cells. Although certain studies have shown that early feeding of probiotics can reduce the development of allergic manifestations, particularly eczema, other studies have found that probiotic supplementation during early childhood did not prevent the development of allergic diseases.100,101 Likewise, there is no clear evidence that the administration of prebiotics or lipopolysaccharides can prevent the development of allergies.102 Accordingly, there is currently no evidence to support the use of prebiotics or probiotics in the prevention of food allergies. However, research is underway to identify specific probiotics or prebiotics that affect allergy development.99

of prebiotics or lipopolysaccharides can prevent the development of allergies.102 Accordingly, there is currently no evidence to support the use of prebiotics or probiotics in the prevention of food allergies. However, research is underway to identify specific probiotics or prebiotics that affect allergy development.99 Diagnosis and Management Once a patient is diagnosed with a food allergy, it becomes important to identify the allergen(s) that cause the disorder and determine if it is mediated by IgE. If so, treatments for IgE-associated allergies can be selected (Figure 3). In early studies of food allergies,103 diagnoses were based on careful analyses of case histories and diaries to document symptoms and offending foods.6 For more information on the diagnosis of food allergies, see Figure 3 and articles by Sicherer and Sampson,3 De Silva et al,5 and Boyce et al.6

be selected (Figure 3). In early studies of food allergies,103 diagnoses were based on careful analyses of case histories and diaries to document symptoms and offending foods.6 For more information on the diagnosis of food allergies, see Figure 3 and articles by Sicherer and Sampson,3 De Silva et al,5 and Boyce et al.6 Results from serologic and in vitro tests alone are not sufficient for the diagnosis of IgE-associated food allergies because the presence of allergen-specific IgE is not always associated with symptoms.104 This might be because certain antigens in food react with IgE but do not have allergenic activity. Furthermore, certain allergens easily are degraded and/or do not pass the epithelial barrier in sufficient quantities. For this reason, provocation tests often are useful and required to confirm IgE-associated food allergy. Skin tests are relatively easy to perform, but just as for IgE tests, results are not always associated with symptoms. Skin tests can detect IgE-mediated mast cell degranulation; these include the skin-prick test, prick-to-prick test, and intradermal test. Atopy patch tests can detect delayed-type reactions mediated by T-cell reactions.

atively easy to perform, but just as for IgE tests, results are not always associated with symptoms. Skin tests can detect IgE-mediated mast cell degranulation; these include the skin-prick test, prick-to-prick test, and intradermal test. Atopy patch tests can detect delayed-type reactions mediated by T-cell reactions. Oral provocation tests are the most accurate in the diagnosis of clinically relevant IgE-associated food allergies once allergen-specific IgE has been detected. These involve placing patients on elimination diets, and re-introducing foods or providing an open oral challenge. A double-blind, placebo-controlled food challenge is the standard for antigen identification.105 However, the double-blind, placebo-controlled food challenge can induce severe reactions and requires careful planning and well-equipped clinical facilities. Several provocation tests involve the application of mucosal allergens.106 Major advances in the diagnosis of food allergies and identification of disease-causing allergens include new in vitro multiplex allergy tests, which involve purified allergens.

Oral provocation tests are the most accurate in the diagnosis of clinically relevant IgE-associated food allergies once allergen-specific IgE has been detected. These involve placing patients on elimination diets, and re-introducing foods or providing an open oral challenge. A double-blind, placebo-controlled food challenge is the standard for antigen identification.105 However, the double-blind, placebo-controlled food challenge can induce severe reactions and requires careful planning and well-equipped clinical facilities. Several provocation tests involve the application of mucosal allergens.106 Major advances in the diagnosis of food allergies and identification of disease-causing allergens include new in vitro multiplex allergy tests, which involve purified allergens. IgE-associated food allergies are managed with allergen-specific treatments such as avoidance of the disease-causing allergens via diets that ensure balanced nutrition with the least possible effects on quality of life. The elimination diet is the most important and relevant long-term management strategy for food allergies.5 Once the offending food allergens have been identified the allergenic food must be avoided. For a summary of the management paths and guidelines for treatment, see the articles by Sicherer and Sampson,3 De Silva et al,5 and Boyce et al.6 For patients with a cow’s milk allergy, milk can be replaced with extensively hydrolyzed milk formulas.107

allergens have been identified the allergenic food must be avoided. For a summary of the management paths and guidelines for treatment, see the articles by Sicherer and Sampson,3 De Silva et al,5 and Boyce et al.6 For patients with a cow’s milk allergy, milk can be replaced with extensively hydrolyzed milk formulas.107 Allergen-specific immunotherapy (SIT) is currently the only allergen-specific and disease-modifying treatment that has long-term effects.108 SIT is used mainly to treat respiratory allergies, and less frequently to treat food allergies because standardized vaccines are not available. In the case of food allergies, SIT often is performed orally, by administration of the offending food instead of a vaccine.68,109 Progress in the molecular characterization of food allergens will lead to the development of defined vaccines for the treatment of food allergies, as for respiratory allergens, which may become available in the future.110–112 Patients diagnosed with IgE-mediated food allergies can be given medications to reduce their symptoms.4,5 These can be selected based on the involvement of IgE-mediated mast cell or basophil degranulation (antihistamines, antileukotrienes, epinephrine, anti-IgE), or T cells or eosinophil activation (steroids, anti-IL5) (Figure 3).3,6 This procedure is beginning to enter clinical practice.

tions to reduce their symptoms.4,5 These can be selected based on the involvement of IgE-mediated mast cell or basophil degranulation (antihistamines, antileukotrienes, epinephrine, anti-IgE), or T cells or eosinophil activation (steroids, anti-IL5) (Figure 3).3,6 This procedure is beginning to enter clinical practice. Food Allergen Structure, Pathogenesis, and Diagnosis Since the 1980s, we have learned much about the structure and immunologic characteristics of allergens, and the clinical reactions they can cause.110 Researchers have produced recombinant allergens comprising repertoires of the most common antigens. Instead of ill-defined allergen extracts, which are prepared from the allergen sources (eg, wheat, apple, milk, or peanuts) (Figure 4) and consist of mixtures of various allergens and nonallergenic materials, pure allergen molecules are available for diagnosis and allergen-specific therapy. Purified recombinant allergens can be used to determine a patient’s IgE reactivity profile. Many allergen sources contain antigens that have little or no clinical relevance because they are poor inducers of allergic reactions. These include IgE-reactive carbohydrate epitopes without allergenic activity,113 or molecules that induce only mild or local symptoms. Other molecules can induce severe systemic allergic reactions (Figure 4).

ontain antigens that have little or no clinical relevance because they are poor inducers of allergic reactions. These include IgE-reactive carbohydrate epitopes without allergenic activity,113 or molecules that induce only mild or local symptoms. Other molecules can induce severe systemic allergic reactions (Figure 4). Marker allergens have been identified from the most common food allergen sources (eg, apple, peanut, milk, and wheat).114–118 Marker allergens are those found only in specific sources, and can be used to confirm sensitizations to these sources. Other allergens are present in different food sources. Patients who are sensitized to these can develop symptoms after ingestion of seemingly unrelated foods. Individual allergen molecules are named and listed by the nomenclature committee of the World Health Organization and International Union of Immunological Societies.119 For an overview of the general biochemical characterization and features of allergens (see the article by Valenta120). Several databases also store information on allergens, including names, classifications, and characteristics (eg, http://www.allergen.org/; http://farrp.unl.edu/resources/farrp-databases; http://allergen.nihs.go.jp/ADFS/; http://www.allergome.org/; http://www.meduniwien.ac.at/allergens/allfam/; and http://www.allergenonline.org/).

bases also store information on allergens, including names, classifications, and characteristics (eg, http://www.allergen.org/; http://farrp.unl.edu/resources/farrp-databases; http://allergen.nihs.go.jp/ADFS/; http://www.allergome.org/; http://www.meduniwien.ac.at/allergens/allfam/; and http://www.allergenonline.org/). Table 1 shows some important plant food and animal food allergen families. According to a recent classification, allergens from different sources can be grouped into families with similar biologic functions, primary structure, and immunologic cross-reactivity.121 A few examples for allergens from different sources that can be attributed to these structurally related allergen families are shown in Table 1. The diagnosis of food and other allergies has transitioned from the identification of allergen sources without knowledge of the molecules that cause the symptoms, to the precise identification of allergy-inducing molecules. These processes are called “component-resolved allergy diagnosis” and “molecular allergy diagnosis.”122

gnosis of food and other allergies has transitioned from the identification of allergen sources without knowledge of the molecules that cause the symptoms, to the precise identification of allergy-inducing molecules. These processes are called “component-resolved allergy diagnosis” and “molecular allergy diagnosis.”122 Our ability to test a patient’s reactivity to a growing number of well-characterized allergen molecules has required the development of new diagnostic tests. We now can test small volumes of serum for IgE reactivity against multiple allergens simultaneously. Allergen chips containing micro-arrayed allergen molecules have been developed for this purpose123; they can be used to analyze serum samples for reactivity with a comprehensive set of molecules. This approach is ideal for analysis of children, or studies of differences or changes in allergic immune responses in large groups, such as birth cohorts.124 Based on component-resolved allergy diagnosis, the type and severity of symptoms can be predicted. Furthermore, the most relevant components can be identified to aid in the development of allergen-specific treatments and preventative strategies.125

in allergic immune responses in large groups, such as birth cohorts.124 Based on component-resolved allergy diagnosis, the type and severity of symptoms can be predicted. Furthermore, the most relevant components can be identified to aid in the development of allergen-specific treatments and preventative strategies.125 Allergen-Specific Prevention and Therapy A major limitation of SIT is the difficulty in preparation of effective and safe vaccines from natural allergen sources.126 However, based on the knowledge of the structure of the disease-causing allergens, it has become possible to produce new forms of allergy vaccines based on purified allergen molecules (Figure 5).110 Clinical trials have shown the efficacy of immunotherapies that include recombinant allergens in the wild-type, folded forms.127 Vaccines can be developed based on these allergens, under standardized conditions.

become possible to produce new forms of allergy vaccines based on purified allergen molecules (Figure 5).110 Clinical trials have shown the efficacy of immunotherapies that include recombinant allergens in the wild-type, folded forms.127 Vaccines can be developed based on these allergens, under standardized conditions. However, wild-type allergens still contain epitopes that activate IgE and T cells, and therefore might induce allergic reactions in patients. Three approaches, based on modified allergens, have been developed to make allergy vaccines more safe, effective, and convenient. These allow for selective targeting of different facets of the allergic immune response. They include synthetic allergen-derived peptides that contain allergen-specific T-cell epitopes without IgE reactivity.128 Because of their small size, peptide vaccines can induce T-cell tolerance without allergen-specific IgG responses. Recombinant hypoallergenic allergen derivatives are characterized by strongly reduced IgE reactivity, and contain allergen-specific T-cell epitopes. After internalization, they can induce allergen-specific IgG responses.129 Carrier-bound peptides that contain B-cell epitopes are fusion proteins that consist of an allergen-unrelated carrier protein and nonallergenic peptides from the IgE binding sites of allergens. They lack IgE reactivity and most allergen-specific T-cell epitopes, but can induce allergen-specific IgG antibodies.130