Browse the corpus

Respiratory viruses are associated with prolonged shedding, higher rates of lower respiratory tract disease, and mortality in hematopoietic cell transplant (HCT) recipients. Development of novel therapeutics and effective infection prevention has been critically important, especially for well-established respiratory viruses such as respiratory syncytial virus, influenza virus, and parainfluenza viruses [1–4]. With new molecular diagnostics widely available, similar concerns have been raised for other respiratory viruses including human coronavirus (HCoV) [5]. In addition to the demonstration of frequent prolonged shedding of HCoV after HCT [6], recent data suggest that common HCoVs (229E, OC43, NL63, and HKU1) are important respiratory pathogens related to significant mortality in HCT recipients [7]. Data on host and virologic factors associated with prolonged shedding, including genome evolution within a host, may provide a rationale for the development of antiviral therapy at various stages, but are currently lacking. Therefore, we examined HCT recipients to define viral and host factors associated with prolonged HCoV shedding in the upper respiratory tract and examine evolution of viral genomic sequences over time by metagenomic next-generation sequencing (mNGS).

nt of antiviral therapy at various stages, but are currently lacking. Therefore, we examined HCT recipients to define viral and host factors associated with prolonged HCoV shedding in the upper respiratory tract and examine evolution of viral genomic sequences over time by metagenomic next-generation sequencing (mNGS). METHODS Study Design We reviewed HCT recipients with HCoV detected in nasal samples by multiplex respiratory viral polymerase chain reaction (PCR) at the Fred Hutchinson Cancer Research Center. Subjects were required to have a negative viral PCR test within 2 weeks of the last positive virology testing performed. If the interval between consecutive positive tests was beyond 2 weeks, strain identification was performed using both samples to confirm the strains were the same. The subjects were identified from 2 cohorts (Supplementary Figure). The first cohort included patients whose nasal samples were collected and tested for clinical purposes when respiratory symptoms were present from March 2009 through June 2016. The second cohort came from a prospective surveillance study of HCT recipients undergoing transplant from December 2005 and February 2010 [8]. Standardized respiratory symptom surveys and multiplex respiratory PCR tests were performed weekly during the first 100 days posttransplant, then every 3 months through year 1 posttransplant and whenever respiratory symptoms occurred between days 100 and 365 posttransplant. Only subjects with respiratory symptoms were selected for the current study, and no duplicated subjects were analyzed. Separately, mNGS was conducted when ≥4 positive samples with cycle threshold (Ct) values of <28 were available irrespective of presence of respiratory symptoms from the above-mentioned prospective surveillance study of HCT recipients. Demographic and clinical data were collected from the database and medical chart review. The study was approved by the Institutional Board Review at Fred Hutchinson Cancer Research Center.

ailable irrespective of presence of respiratory symptoms from the above-mentioned prospective surveillance study of HCT recipients. Demographic and clinical data were collected from the database and medical chart review. The study was approved by the Institutional Board Review at Fred Hutchinson Cancer Research Center. Laboratory Testing and Definitions HCoV detection and viral load were determined from nasal specimens by quantitative reverse-transcription PCR as part of a multiplex PCR used to detect 12 respiratory viruses. Strain-specific PCR was performed using saved nasal samples according to a previously published protocol [9]. mNGS was performed on DNAse I-treated RNA extracts from 0.45-uM filtered nasal specimens using “tagmented” (transposon-mediated fragmentation) cDNA libraries with 15–20 cycles of PCR amplification when ≥4 samples with Ct values of <28 were available [10]. Sequence reads were trimmed using cutadapt and aligned to a concatenation of the 4 HCoV reference genomes (NC_002645, NC_005831, NC_006577, and KF530069) using Geneious version 9.1 [11]. The duration of shedding was defined as time between the first positive and first negative sample. Prolonged shedding was defined as the duration of shedding ≥21 days, which was described to be a median shedding duration of HCoV during the first 100 days after HCT [6]. Highest daily steroid dose and lowest cell count in the 2 weeks prior to first HCoV detection were recorded. Conditioning regimen was categorized into myeloablative and nonmyeloablative/reduced intensity based on the definition previously described [12].

hedding duration of HCoV during the first 100 days after HCT [6]. Highest daily steroid dose and lowest cell count in the 2 weeks prior to first HCoV detection were recorded. Conditioning regimen was categorized into myeloablative and nonmyeloablative/reduced intensity based on the definition previously described [12]. Statistical Analysis Univariable and bivariable logistic regression analyses were performed to evaluate associations between virologic and host factors and prolonged shedding. Only the first episode of HCoV infection per subject was used for the outcome analyses. Variables with P ≤ .2 in the univariable models were candidates for bivariable models. Kruskal-Wallis test was performed to compare continuous values among more than 2 groups. Two-sided P values <.05 were considered statistically significant. All statistical analyses were performed using SAS 9.4 for Windows (SAS Institute, Cary, North Carolina). RESULTS Host and Virological Characteristics We identified 20 and 24 HCT recipients with respiratory HCoV infection from cohort 1 and cohort 2, respectively (42 adult and 2 pediatric patients) (Table 1 and Supplementary Figure). The median duration of shedding was 14 days (4–60 days), and 17 patients had prolonged shedding (≥21 days). Among 31 available nasal samples, 35% were OC43, 32% were NL63, 19% were HKU1, and 13% were 229E. The median shedding duration of HCoV in nasal samples did not differ between strains (Figure 1; P = .79). Table 1. Clinical Features of Patients With Human Coronavirus Upper Respiratory Tract Disease

RESULTS Host and Virological Characteristics We identified 20 and 24 HCT recipients with respiratory HCoV infection from cohort 1 and cohort 2, respectively (42 adult and 2 pediatric patients) (Table 1 and Supplementary Figure). The median duration of shedding was 14 days (4–60 days), and 17 patients had prolonged shedding (≥21 days). Among 31 available nasal samples, 35% were OC43, 32% were NL63, 19% were HKU1, and 13% were 229E. The median shedding duration of HCoV in nasal samples did not differ between strains (Figure 1; P = .79). Table 1. Clinical Features of Patients With Human Coronavirus Upper Respiratory Tract Disease Characteristic Total (N = 44) Patients With Prolonged Shedding (n = 17) Patients With Short-term Shedding (n = 27) Female sex 18 (41) 7 (41) 11 (41) Age, y, median (range) 54 (7–73) 54 (7–67) 55 (14–73) Transplant number ≥2 12 (27) 6 (35) 6 (22) Cell source Cord 3 (7) 2 (12) 1 (4) Bone marrow 4 (9) 2 (12) 2 (7) PBSC 37 (84) 13 (76) 24 (89) Donor type Autologous 3 (7) 1 (6) 2 (7) Related 21 (48) 6 (35) 15 (56) Unrelated 20 (45) 10 (59) 10 (37) Conditioning regimen Myeloablative 18 (41) 10 (59) 8 (30) NMA or RIC 26 (59) 7 (41) 19 (70) Onset of shedding relative to transplant Pretransplant 4 (9) 4 (23) 0 0–100 days posttransplant 22 (50) 6 (35) 16 (59) >100 days posttransplant 18 (41) 7 (41) 11 (41) Human coronavirus strains OC43 11 (25) 3 (18) 8 (30) NL63 10 (23) 5 (29) 5 (19) 229E 4 (9) 2 (12) 2 (7) HKU1 6 (14) 2 (12) 4 (15) Unknown 13 (30) 5 (29) 8 (30) Ct value, median (range) 28.3 (19.2–39.4) 26.1 (19.2–39.4) 28.8 (19.6–39.4) Lowest WBC count <1000 × 106 cells/La 14/36 (39) 3/15 (20) 11 /21 (52) Lowest lymphocyte count <300 × 106 cells/La 18/36 (50) 6 /15 (40) 12/21 (57) Lowest neutrophil count <500 × 106 cells/La 13/36 (36) 3/15 (20) 10 /21 (48) Lowest monocyte count <100 × 106 cells/La 16/36 (44) 5/15 (33) 11/21 (52) Highest daily steroid dosea None 26 (59) 8 (47) 18 (67) ≤1 mg/kg 12 (27) 5 (29) 7 (26) >1 mg/kg 6 (14) 4 (24) 2 (7) Data are presented as No. (%) unless otherwise indicated.

st neutrophil count <500 × 106 cells/La 13/36 (36) 3/15 (20) 10 /21 (48) Lowest monocyte count <100 × 106 cells/La 16/36 (44) 5/15 (33) 11/21 (52) Highest daily steroid dosea None 26 (59) 8 (47) 18 (67) ≤1 mg/kg 12 (27) 5 (29) 7 (26) >1 mg/kg 6 (14) 4 (24) 2 (7) Data are presented as No. (%) unless otherwise indicated. Abbreviations: Ct, cycle threshold; NMA, nonmyeloablative; PBSC, peripheral blood stem cell; RIC, reduced intensity; WBC, white blood cell. aIn the 2 weeks prior to first human coronavirus detection. Figure 1. Duration of shedding according to human coronavirus strain. The bars indicate median values and first and third quartiles (P = .79 by Kruskal-Wallis test). Outcome Analyses Initial high viral load (Ct value below the median) was associated with prolonged shedding with the lowest P value (<.01) by univariable analysis. Univariable and bivariable logistic regression analyses indicated that initial high viral load was associated with prolonged shedding consistently in all models (Table 2). High-dose steroid use (≥1 mg/kg/day) prior to HCoV diagnosis and myeloablative conditioning regimen were associated with prolonged shedding in the bivariable analyses. Four patients started viral shedding prior to transplant; therefore, we separately analyzed 40 patients who started shedding after transplant, and the results remained similar (data not shown.) Table 2. Univariable and Bivariable Logistic Regression Analyses for Prolonged Shedding (n = 44)a

Outcome Analyses Initial high viral load (Ct value below the median) was associated with prolonged shedding with the lowest P value (<.01) by univariable analysis. Univariable and bivariable logistic regression analyses indicated that initial high viral load was associated with prolonged shedding consistently in all models (Table 2). High-dose steroid use (≥1 mg/kg/day) prior to HCoV diagnosis and myeloablative conditioning regimen were associated with prolonged shedding in the bivariable analyses. Four patients started viral shedding prior to transplant; therefore, we separately analyzed 40 patients who started shedding after transplant, and the results remained similar (data not shown.) Table 2. Univariable and Bivariable Logistic Regression Analyses for Prolonged Shedding (n = 44)a Covariates Categories Univariable Model Bivariable Model 1 Bivariable Model 2 Bivariable Model 3 Bivariable Model 4 OR (95% CI) P Value Adjusted OR (95% CI) P Value Adjusted OR (95% CI) P Value Adjusted OR (95% CI) P Value Adjusted OR (95% CI) P Value Ct value <28.3 vs ≥28.3 6.5 (1.6–26) <.01 11.6 (2.1–64.7) <.01 11.0 (2.1–58.8) <.01 5.1 (1.0–25.2) .05 5.5 (1.1–26.8) .03 Conditioning regimen MA vs NMA/RIC 3.4 (.95–12) .06 6.9 (1.3–38) .03 Highest steroid doseb (mg/kg/ day) ≥1 vs <1 3.85 (.6–24) .15 10.1 (1.1–96) .05 Lowest WBC countb (×106 cells/L) <1.0 vs ≥1.0 0.23 (.05–1.05) .06 0.33 (.06–1.7) .18 Lowest neutrophil countb (×106 cells/L) <0.5 vs ≥0.5 0.28 (.06–1.3) .10 0.37 (.07–1.9) .23 Lowest lymphocyte countb (×106 cells/L) <0.3 vs ≥0.3 0.5 (.13–1.9) .31 Lowest monocyte countb (×106 cells/L) <0.1 vs ≥0.1 0.45 (.12–1.8) .26 Transplant number ≥2 vs ≤1 1.91 (.5–7.3) .35 Age at diagnosis As continuous 0.98 (.94–1.03) .48 Abbreviations: CI, confidence interval; Ct, cycle threshold; MA, myeloablative; NMA, nonmyeloablative; OR, odds ratio; RIC, reduced intensity; WBC, white blood cell.

cyte countb (×106 cells/L) <0.1 vs ≥0.1 0.45 (.12–1.8) .26 Transplant number ≥2 vs ≤1 1.91 (.5–7.3) .35 Age at diagnosis As continuous 0.98 (.94–1.03) .48 Abbreviations: CI, confidence interval; Ct, cycle threshold; MA, myeloablative; NMA, nonmyeloablative; OR, odds ratio; RIC, reduced intensity; WBC, white blood cell. aVariables with P ≤ .2 in the univariable models were candidates for bivariable models where data only support inclusion of 2 factors per model. bIn the 2 weeks prior to first human coronavirus detection.

cyte countb (×106 cells/L) <0.1 vs ≥0.1 0.45 (.12–1.8) .26 Transplant number ≥2 vs ≤1 1.91 (.5–7.3) .35 Age at diagnosis As continuous 0.98 (.94–1.03) .48 Abbreviations: CI, confidence interval; Ct, cycle threshold; MA, myeloablative; NMA, nonmyeloablative; OR, odds ratio; RIC, reduced intensity; WBC, white blood cell. aVariables with P ≤ .2 in the univariable models were candidates for bivariable models where data only support inclusion of 2 factors per model. bIn the 2 weeks prior to first human coronavirus detection. Whole-Genome Sequencing Whole genomes of OC43, NL63, and HKU1 were consecutively sequenced in samples from 4 HCT adult subjects and 1 pediatric subject where samples were available for 19 to 132 days following the first positive sample (Table 3). Engraftment occurred in 4 patients prior to the start of shedding. No majority consensus variants were recovered for any patient <30 days after onset of shedding. Single-nucleotide variants accumulated at a rate of approximately 1 variant per 3–4 weeks, consistent with previous estimates of the HCoV molecular clock (Figure 2) [13, 14]. No single-nucleotide polymorphisms (SNPs) of OC43 and HKU1 were recovered in patients 4 and 5, respectively. One adult patient (patient 3) developed lower respiratory tract disease in the setting of high-dose steroid use for acute graft-vs-host disease during prolonged shedding. Bronchoalveolar lavage was performed at day 73 after starting the shedding, from which Aspergillus fumigatus was detected in addition to HCoV OC43. One 18-year-old pediatric patient (patient 4) had 3 different HCoV strains detected in succession over a period of 5 months.

raft-vs-host disease during prolonged shedding. Bronchoalveolar lavage was performed at day 73 after starting the shedding, from which Aspergillus fumigatus was detected in addition to HCoV OC43. One 18-year-old pediatric patient (patient 4) had 3 different HCoV strains detected in succession over a period of 5 months. Table 3. Specimens Sequenced in this Study Patient Strain Species Daya Ct Values HCoV Reads Total Reads Accession 1 N07-196B NL63 –10 26.1 291765 465084 KY554969 N07-262B NL63 4 25.9 72640 249121 KY829118 N07-468B NL63 42 26.2 59658 191697 KY554971 2 N06-1144B NL63 48 25.7 341740 428487 KY554967 N07-6B NL63 64 27.7 67223 342554 KY674915 N07-64B NL63 79 24.5 136330 239866 KY674916 N07-185B NL63 107 24.1 30368 79584 KY554968 N07-324B NL63 135 25 56685 542950 KY554970 3 N09-33B OC43 9 21.5 128622 213882 KY554974 N09-382B OC43 77 27.1 8063 137198 KY554975 N09-595B OC43 118 27.6 6297 1044785 KY674920 4 N07-1541B OC43 116 23.8 160033 580450 KY554972 N07-1609B OC43 130 24.7 43692 221874 KY674917 N07-1647B OC43 137 22.7 107905 147986 KY674918 N08-87B HKU1 174 27 23772 737630 KY674921 N08-434B 229E 248 27.2 12019 2981384 KY674919 5 N09-1605B HKU1 29 20.8 1636757 2936518 KY674943 N09-1627B HKU1 36 23.5 491949 877155 KY674942 N09-1663B HKU1 47 27.8 9164 266311 KY674941 The number of HCoV reads, total reads, Ct values, and accession number are depicted for each of the specimens for which whole genomes were recovered. Abbreviations: Ct, cycle threshold; HCoV, human coronavirus. aDay is relative to engraftment.

Patient Strain Species Daya Ct Values HCoV Reads Total Reads Accession 1 N07-196B NL63 –10 26.1 291765 465084 KY554969 N07-262B NL63 4 25.9 72640 249121 KY829118 N07-468B NL63 42 26.2 59658 191697 KY554971 2 N06-1144B NL63 48 25.7 341740 428487 KY554967 N07-6B NL63 64 27.7 67223 342554 KY674915 N07-64B NL63 79 24.5 136330 239866 KY674916 N07-185B NL63 107 24.1 30368 79584 KY554968 N07-324B NL63 135 25 56685 542950 KY554970 3 N09-33B OC43 9 21.5 128622 213882 KY554974 N09-382B OC43 77 27.1 8063 137198 KY554975 N09-595B OC43 118 27.6 6297 1044785 KY674920 4 N07-1541B OC43 116 23.8 160033 580450 KY554972 N07-1609B OC43 130 24.7 43692 221874 KY674917 N07-1647B OC43 137 22.7 107905 147986 KY674918 N08-87B HKU1 174 27 23772 737630 KY674921 N08-434B 229E 248 27.2 12019 2981384 KY674919 5 N09-1605B HKU1 29 20.8 1636757 2936518 KY674943 N09-1627B HKU1 36 23.5 491949 877155 KY674942 N09-1663B HKU1 47 27.8 9164 266311 KY674941 The number of HCoV reads, total reads, Ct values, and accession number are depicted for each of the specimens for which whole genomes were recovered. Abbreviations: Ct, cycle threshold; HCoV, human coronavirus. aDay is relative to engraftment. Figure 2. Genome evolution of human coronavirus over time. Single-nucleotide variants in the consensus genome for patient 1 (A), 2 (B), and 3 (C) are depicted. Patients 4 and 5 showed no variants over time for the same coronavirus species (Table 3). Both nonsynonymous and synonymous variants are depicted relative to the day 0 genome recovered. Majority consensus nonsynonymous changes at allele frequency >50% are indicted in red, while majority consensus synonymous changes at allele frequency >50% are shown in green. Allele frequencies for variant sites are depicted when the majority allele had <95% frequency. Nonsynonymous changes are shown for the amino acid of a given gene, while synonymous changes are shown for the nucleotide of that gene. Allele frequencies alone are shown when the majority consensus base is not a variant relative to the day 0 genome but the majority allele frequency at that base is <95% (in flux). Abbreviations: HE, hemagglutinin esterase; M, membrane protein; N, nucleocapsid protein.

changes are shown for the nucleotide of that gene. Allele frequencies alone are shown when the majority consensus base is not a variant relative to the day 0 genome but the majority allele frequency at that base is <95% (in flux). Abbreviations: HE, hemagglutinin esterase; M, membrane protein; N, nucleocapsid protein. DISCUSSION In this study, we demonstrated a significant association between prolonged shedding of HCoV and initial high viral load in transplant recipients. In addition, prior high-dose steroid use and myeloablative conditioning regimen appear to be associated with prolonged shedding. The duration of shedding appeared to be similar across all 4 HCoV strains. No drastic intrahost evolution of viral genomes occurred in this immunocompromised population with prolonged shedding.

addition, prior high-dose steroid use and myeloablative conditioning regimen appear to be associated with prolonged shedding. The duration of shedding appeared to be similar across all 4 HCoV strains. No drastic intrahost evolution of viral genomes occurred in this immunocompromised population with prolonged shedding. Severe acute respiratory syndrome and Middle East respiratory syndrome (MERS) coronaviruses are recognized as highly human-pathogenic coronaviruses causing fatal lower respiratory tract disease [15–17]; however, there are no established antiviral therapies [18, 19]. Recent data suggest that lower respiratory tract disease caused by 4 other HCoV strains (229E, OC43, NL63, and HKU1) was also associated with significant respiratory support and mortality in immunocompromised hosts [7]. The unmet need for the development of antiviral therapy against HCoV is expected to expand as immunocompromised populations grow. We found that initial high viral load, prior high-dose steroid use, and myeloablative conditioning were important factors associated with prolonged HCoV shedding in HCT recipients. Duration of viral shedding is often used as an endpoint at early stages of clinical trials for new antiviral drugs [20, 21]. Stratification based on risk factors is critical to avoid imbalances due to host and viral factors in randomized trials, which might otherwise mask true differences of experimental agents.

ients. Duration of viral shedding is often used as an endpoint at early stages of clinical trials for new antiviral drugs [20, 21]. Stratification based on risk factors is critical to avoid imbalances due to host and viral factors in randomized trials, which might otherwise mask true differences of experimental agents. Genetic variability of HCoV OC43 at the community level and intrahost heterogeneity of MERS coronavirus have been reported [22–25]. Such variability might have important implications in viral disease pathogenesis, and the study of viral genome evolution within a host can provide vital information important in developing and assessing antiviral agents [26]. For example, development of antiviral drug resistance during individualized therapy that is associated with poor outcome has been described extensively with influenza virus [27–29]. Grad et al reported intrahost genome evolution of respiratory syncytial virus over time in an infant with severe combined immune deficiency who underwent a bone marrow transplant [30]. The viral population diversity dramatically increased after engraftment, which appeared to reflect dynamic response to immune pressure from host immunity. To our knowledge, no previous data exist describing how the HCoV genome evolves within a host over time. In the current study, engraftment occurred in 4 of 5 patients sequenced prior to the onset of shedding. Interestingly, no SNPs were recovered <30 days after the onset of shedding even after immune reconstitution. Variants accumulated starting at 1 month after the onset of shedding (1–6 changes over time), consistent with the previously estimated evolution rates of HCoV [13, 14]. Given the relatively slow evolution rate of coronaviruses, these observations could be promising from the standpoint of antiviral resistance and therapeutic development [13, 14]. Due to their exceptionally large RNA genomes, coronaviruses are known to encode highly processive polymerases as well as proofreading exoribonucleases that temper viral genome evolution relative to other RNA viruses [31–33]. Further epidemiological and biochemical work is required to characterize the functional impact of the variants recovered here.

e RNA genomes, coronaviruses are known to encode highly processive polymerases as well as proofreading exoribonucleases that temper viral genome evolution relative to other RNA viruses [31–33]. Further epidemiological and biochemical work is required to characterize the functional impact of the variants recovered here. The main limitation of this study was the relatively small sample size; thus, bivariable logistic regression analyses were performed instead of multivariable analyses to evaluate risk factors for prolonged shedding. Similarly, although no particular strain appeared to be associated with prolonged shedding, strain identification using saved samples was successful in only 70% of the patients, which limited our ability to detect small difference of shedding duration among each HCoV strain. Further studies with larger sample sizes will help to clarify the distinct association between particular HCoV strains and prolonged shedding. Finally, our cohort included 4 patients who had documented HCoV shedding prior to transplant. Considering the unmeasured influence of their different backgrounds on our analyses, we separately analyzed 40 patients who started shedding after transplant. Only univariable logistic regression analysis could be performed due to the small sample size, with similar results.

ocumented HCoV shedding prior to transplant. Considering the unmeasured influence of their different backgrounds on our analyses, we separately analyzed 40 patients who started shedding after transplant. Only univariable logistic regression analysis could be performed due to the small sample size, with similar results. This is the first study to evaluate risk factors associated with prolonged shedding of HCoV by quantitative and strain-specific reverse transcription PCR as well as intrahost genomic evolutions by metagenomic RNA sequencing in transplant recipients. Our study provides critical information to develop antiviral therapies and design randomized trials with viral load endpoints. In addition, as the duration of shedding is an important determinant of viral infectivity and transmissibility, predictive factors for prolonged shedding may provide useful information for effective infection control, such as the expected duration of isolation. Further studies are needed to validate the risk factors including particular HCoV strain for prolonged shedding. Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material Supplementary Figure 1 Click here for additional data file. Notes

Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Supplementary Material Supplementary Figure 1 Click here for additional data file. Notes Acknowledgments. We thank Zachary Stednick for database services, Reigran Sampoleo and Isabel Palileo for laboratory assistance, and Laurel Joncas-Shronce for sample management assistance. Financial support. This work was supported by the National Institutes of Health (K24HL093294 to M. B., K23 AI114844 to A. W., CA18029 to W. L., CA15704 to H. X., T32HD00723332 to C. O.); the Fred Hutchinson Cancer Research Center Vaccine and Infectious Disease Division (biorepository); and a Pediatric Infectious Diseases Society Fellowship Award funded by Horizon Pharma to C.O.. Potential conflicts of interest. M. B. and J. A. E. have received research support and served as a consultant for Gilead Sciences. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Mother-to-child transmission (MTCT) of human immunodeficiency virus (HIV) is a natural setting in which the effect of HIV-specific antibodies on HIV infection risk can be studied. Antibodies circulating in the mother and passively acquired maternal antibodies circulating in the infant may both influence MTCT risk. In the case of breastfeeding transmissions, infants who are uninfected at birth have pre-existing passively acquired HIV-specific antibodies present before breastfeeding HIV exposure. Studies have suggested both enhancing and protective effects of maternal HIV antibodies on MTCT risk [1–8]. Some studies of binding antibodies have shown that antibodies specific for the CD4 binding site (CD4bs), V3, gp120, p24, or gp41 are associated with reduced MTCT, whereas others have shown an association with increased MTCT or no association [1–6]. It is unclear which, if any, of these epitopes are targeted by protective binding antibodies in MTCT. The majority of these studies included in utero and peripartum transmissions where it is impossible to sample infant antibody responses to investigate pre-existing passively acquired antibodies present at the time of HIV exposure before infection [1–6]. Most studies included maternal samples but did not investigate passively acquired antibodies in infants [1–4]. Measuring both maternal and passively acquired antibody responses from mother-infant pairs where transmissions are due to breastfeeding could identify protective characteristics of pre-existing antibodies and help clarify results from prior studies.

ut did not investigate passively acquired antibodies in infants [1–4]. Measuring both maternal and passively acquired antibody responses from mother-infant pairs where transmissions are due to breastfeeding could identify protective characteristics of pre-existing antibodies and help clarify results from prior studies. We screened plasma from a cohort of antiretroviral therapy (ART)-naive breastfeeding mother-infant pairs for binding against a panel of HIV antigens to determine whether maternal and/or passively acquired antibodies targeting specific epitopes were associated with MTCT risk. MATERIALS AND METHODS Study Design and Plasma Samples This study used samples from ART-naive mother-infant pairs from the Nairobi Breastfeeding Clinical Trial conducted from 1992 to 1998. All pairs meeting the selection criteria defined by Milligan et al [9] were included. Infants tested HIV-negative at birth, were breastfed for at least 3 months, and infant samples were collected from the first week of life, before the estimated time of infection (Supplementary Table S1). Paired maternal samples were also tested. The cohort included 70 paired samples (50 nontransmitting, 20 transmitting), 1 unpaired transmitting maternal sample, and 1 unpaired HIV-exposed uninfected infant sample. Cohort characteristics are listed in Supplementary Table S1. This research has been approved by the Kenyatta National Hospital Ethics and Research Committee and the Institutional Review Boards of the University of Nairobi, University of Washington, and the Fred Hutchinson Cancer Research Center.

infected infant sample. Cohort characteristics are listed in Supplementary Table S1. This research has been approved by the Kenyatta National Hospital Ethics and Research Committee and the Institutional Review Boards of the University of Nairobi, University of Washington, and the Fred Hutchinson Cancer Research Center. Binding Antibody Multiplex Assay The binding antibody multiplex assay (BAMA) was performed on maternal and infant plasma as described previously using a panel of 20 HIV antigens (Supplementary Table S2) [10]. Median fluorescence intensity (MFI) of plasma antibody binding to each antigen was measured and averaged across duplicate wells. Plasma antibody binding to an appropriate negative control antigen (described in Supplemental Methods) was considered background and subtracted from the plasma MFI for each HIV antigen. To average data from biological replicates, we normalized the background subtracted MFI to that of the positive control HIV immunoglobulin (HIVIG) and converted it to a percent: percent binding = (plasma MFIHIV antigen – plasma MFInegative control antigen)/(HIVIG MFIHIV antigen – HIVIG MFInegative control antigen) × 100%.

average data from biological replicates, we normalized the background subtracted MFI to that of the positive control HIV immunoglobulin (HIVIG) and converted it to a percent: percent binding = (plasma MFIHIV antigen – plasma MFInegative control antigen)/(HIVIG MFIHIV antigen – HIVIG MFInegative control antigen) × 100%. V1V2 Enzyme-Linked Immunosorbent Assay Plasma was added to plates coated with clade B V1V2 caseA2-mulvgp70. Immunoglobulin (Ig)G was detected with goat antihuman IgG conjugated to horseradish peroxidase (HRP). Horseradish peroxidase was detected with 3,3’,5,5’-tetramethylbenzidine substrate. The reaction was stopped with 1 N H2SO4, and the absorbance (optical density [OD]) was read at 450 nm. The background absorbance (OD without plasma present) was subtracted from the absorbance of each well. The background-subtracted OD was normalized to that of HIVIG and converted to a percentage. Soluble CD4 Blocking Assay The soluble CD4 (sCD4) blocking assay was performed similar to that described previously [3]. Clade A BG505.W6M.C2.T332N.L111A or clade D C2-94UG114 gp120 were used as antigens. Plasma samples were added to the plates followed by sCD4. Soluble CD4 binding was detected with biotinylated anti-OKT4 antibody followed by incubation with streptavidin-HRP. Horseradish peroxidase was detected as described above. Background absorbance (without sCD4) at 450 nm was subtracted from the OD of each well. Percent inhibition of sCD4 was calculated as follows: 100% − [(background-subtracted ODplasma/background-subtracted ODno plasma) × 100%].

d by incubation with streptavidin-HRP. Horseradish peroxidase was detected as described above. Background absorbance (without sCD4) at 450 nm was subtracted from the OD of each well. Percent inhibition of sCD4 was calculated as follows: 100% − [(background-subtracted ODplasma/background-subtracted ODno plasma) × 100%]. Statistical Analysis Statistical analyses were performed with Stata v15 SE (StataCorp, College Station, TX). To measure the association of plasma antibody binding to each HIV antigen with MTCT risk, we performed a logistic regression analysis on the percentage of plasma antibody binding or sCD4 inhibition adjusted for maternal plasma ribonucleic acid viral load. Statistical significance was prespecified as P < .05. P values between .05 and .1 were considered nonsignificant statistical trends. The Holm-Bonferoni method was used to account for multiple comparisons.

percentage of plasma antibody binding or sCD4 inhibition adjusted for maternal plasma ribonucleic acid viral load. Statistical significance was prespecified as P < .05. P values between .05 and .1 were considered nonsignificant statistical trends. The Holm-Bonferoni method was used to account for multiple comparisons. RESULTS Infant plasma collected before estimated infection and paired maternal plasma were screened for binding to HIV antigens via BAMA (Supplementary Table S2). Maternal and infant samples showed similar binding patterns across antigens (Supplementary Figure S1). Logistic regression analysis showed that maternal plasma antibody binding to the clade C gp41 ectodomain antigen, clade A1 gp140, and clade B gp41 protein were significantly associated with increased odds of MTCT (ectodomain adjusted odds ratio [aOR] = 1.04, P = .006; gp140 aOR = 1.03, P = .023; gp41 aOR = 1.03, P = .047). Passively acquired IgG binding to these antigens likewise increased risk of transmission but failed to reach statistical significance (ectodomain aOR = 1.03, P = .058; gp140 aOR = 1.02, P = .071; gp41 aOR = 1.04; P = .08). Maternal plasma antibody binding to 6-Helix (provided by Peter Kim and Nitya Ramadoss), in which all 6 helices forming the gp41 trimer-of-hairpins have been linked into a single polypeptide [11], was also associated with a trend towards increased odds of MTCT (aOR = 1.03, P = .075). Plasma antibody binding to the remaining antigens showed no association with odds of MTCT. After correcting for multiple comparisons, plasma antibody binding showed no significant or trending associations with odds of MTCT for any antigen.

o associated with a trend towards increased odds of MTCT (aOR = 1.03, P = .075). Plasma antibody binding to the remaining antigens showed no association with odds of MTCT. After correcting for multiple comparisons, plasma antibody binding showed no significant or trending associations with odds of MTCT for any antigen. For most BAMA antigens, there was a wide dynamic range of plasma antibody binding above the limit of detection (100 MFI after background subtraction). However, for 2 V1V2 antigens (clade C ZM109 and clade B case A2), consensus D V3 peptide (conDV3), and the CD4bs core, over 40% percent of samples had binding below the limit of detection in multiple biological replicates, making it difficult to compare binding across transmission groups. To measure plasma antibody binding to these regions, we performed a V1V2 enzyme-linked immunosorbent assay (ELISA) and sCD4 blocking assays. We used V1V2 caseA2 as the antigen in the V1V2 ELISA because IgG binding to this antigen was correlated with protection in the RV144 trial [12]. Because the BAMA antigen panel contained 3 other V3 peptides, we did not use an alternative assay to measure plasma antibody binding to conDV3. Plasma antibody binding to V1V2caseA2 and the CD4bs as measured by ELISA and sCD4 blocking assays, respectively, showed a wide dynamic range (data not shown) but were not associated with odds of MTCT (Figure 1).

ained 3 other V3 peptides, we did not use an alternative assay to measure plasma antibody binding to conDV3. Plasma antibody binding to V1V2caseA2 and the CD4bs as measured by ELISA and sCD4 blocking assays, respectively, showed a wide dynamic range (data not shown) but were not associated with odds of MTCT (Figure 1). Figure 1. Association of plasma binding with odds of mother-to-child transmission (MTCT). The association of plasma binding to each antigen with odds of MTCT was measured using a logistic regression analysis adjusted for maternal plasma ribonucleic acid viral load. Results are shown as forest plots. Adjusted odds ratios ([OR] diamonds), 95% confidence intervals ([CI] horizontal lines), and P values are shown for infant samples (left) and maternal samples (right). Statistical significance was defined as P < .05 (*). Shown in the rightmost column is the assay used for each antigen: binding antibody multiplex assay (BAMA), enzyme-linked immunosorbent assay (ELISA), or soluble CD4 (sCD4) blocking assay. Figure 2 illustrates common features of the gp41 antigens. All 4 antigens that showed significant or trending associations with increased odds of MTCT contain the gp41 ectodomain. The common feature is the presence of the N heptad repeat (NHR) and C-terminal heptad repeat (CHR) that form the 6-Helix peptide.

Figure 1. Association of plasma binding with odds of mother-to-child transmission (MTCT). The association of plasma binding to each antigen with odds of MTCT was measured using a logistic regression analysis adjusted for maternal plasma ribonucleic acid viral load. Results are shown as forest plots. Adjusted odds ratios ([OR] diamonds), 95% confidence intervals ([CI] horizontal lines), and P values are shown for infant samples (left) and maternal samples (right). Statistical significance was defined as P < .05 (*). Shown in the rightmost column is the assay used for each antigen: binding antibody multiplex assay (BAMA), enzyme-linked immunosorbent assay (ELISA), or soluble CD4 (sCD4) blocking assay. Figure 2 illustrates common features of the gp41 antigens. All 4 antigens that showed significant or trending associations with increased odds of MTCT contain the gp41 ectodomain. The common feature is the presence of the N heptad repeat (NHR) and C-terminal heptad repeat (CHR) that form the 6-Helix peptide. Figure 2. gp41-containing binding antibody multiplex assay (BAMA) antigens. Schematics of the gp41 domains present in BAMA antigens. The BG505 SOSIP trimer and 6-helix are both trimeric. 6-helix is a continuous trimer of heterodimers of N heptad repeat (NHR) and C-terminal heptad repeat (CHR) regions connected with glycine-serine linkers. The gp41 ectodomain antigen contains a C-terminal His-tag. The gp41 ectodomain is occluded in the BG505 SOSIP trimer as denoted by the light orange color and gray text [13]. CT, cytoplasmic tail; FP, fusion peptide; loop, immunodominant C-C loop; MPER, membrane proximal external region; PR, proximal region; TM, transmembrane domain.

omain antigen contains a C-terminal His-tag. The gp41 ectodomain is occluded in the BG505 SOSIP trimer as denoted by the light orange color and gray text [13]. CT, cytoplasmic tail; FP, fusion peptide; loop, immunodominant C-C loop; MPER, membrane proximal external region; PR, proximal region; TM, transmembrane domain. DISCUSSION This study used the unique setting of breastfeeding MTCT, in which maternal antibodies and infant pre-existing HIV-specific antibodies present at the time of breastfeeding HIV exposure can be measured, to assess whether antibodies targeting certain HIV epitopes are associated with reduced risk of transmission. Antibody binding to V3, the CD4bs, gp41, gp120, and p24 have been identified as correlates of reduced risk of MTCT [1–6]. We were surprised to find that there were no correlates of reduced risk of MTCT for plasma antibody binding to our antigen panel; rather, we found an association with increased MTCT for plasma antibody binding to 4 antigens. Even though the odds ratios were small (possibly because the unit of percentage binding is small, and there is a wide dynamic range of percentage binding among samples), not all of these associations were statistically significant, and none were significant after correcting for multiple comparisons. What is remarkable is that all of these associations, whether maternal or infant, were with gp41-containing antigens (Figure 2). Furthermore, all of the associations were with increased odds of MTCT. The consistency of these results, along with multiple statistically significant associations, supports a biologically relevant association rather than random chance. Studies that included in utero and peripartum transmissions have also shown gp41-specific binding antibodies to be associated with increased risk of MTCT [2, 8]. Although one of these studies was conducted before methods to define the timing of infant infection were available [2], they nonetheless provide additional support for a role for gp41 antibodies in MTCT.

transmissions have also shown gp41-specific binding antibodies to be associated with increased risk of MTCT [2, 8]. Although one of these studies was conducted before methods to define the timing of infant infection were available [2], they nonetheless provide additional support for a role for gp41 antibodies in MTCT. The 3 antigens associated with MTCT for both infant and maternal plasma antibody binding contained the fusion peptide, fusion peptide proximal region, NHR, the immunodominant C-C loop, CHR, and membrane proximal external region (MPER). Although MPER is exposed in the clade B gp41 protein and clade A gp140 antigens, MPER is likely occluded in the gp41 ectodomain antigen by a C-terminal His-tag because MPER-positive control antibodies do not bind this antigen (data not shown). Plasma antibody binding to the clade C gp41 ectodomain antigen showed the strongest association with MTCT, suggesting that plasma antibody binding to MPER is not driving the association with increased MTCT. A fourth antigen, 6-Helix, containing NHR and CHR regions linked in trimeric form, showed a trend for an association with MTCT for maternal plasma antibody binding. This suggests that the key target of these antibodies is the heptad repeat regions. The BG505.SOSIP.664.D7342 trimer includes part of the gp41 ectodomain, but the ectodomain is known to be occluded in this antigen (Figure 2, light orange) [13]. Plasma antibody binding to this antigen was not associated with odds of MTCT. None of the gp120-only antigens were associated with odds of MTCT upon plasma antibody binding. This suggests that plasma antibody binding to the gp41 ectodomain, most likely the heptad repeat regions, is driving the association with increased odds of MTCT.

binding to this antigen was not associated with odds of MTCT. None of the gp120-only antigens were associated with odds of MTCT upon plasma antibody binding. This suggests that plasma antibody binding to the gp41 ectodomain, most likely the heptad repeat regions, is driving the association with increased odds of MTCT. The association of plasma antibody binding to the gp41 ectodomain with MTCT could be direct, by which gp41 ectodomain-specific antibodies are risk-enhancing, perhaps by binding to gp41 on the virus and promoting infection through Fc-mediated mechanisms [14]. On the other hand, the association may reflect an indirect mechanism, perhaps where the gp41 ectodomain is acting as an immune decoy. In contrast to the native-like trimeric antigen (where the ectodomain is occluded [13]) that was not associated with MTCT upon plasma antibody binding, gp41 antigens associated with increased MTCT have conformations similar to nonnative forms of envelope such as monomers, uncleaved envelope, and gp41 stumps [15], which may have the ectodomain exposed.

imeric antigen (where the ectodomain is occluded [13]) that was not associated with MTCT upon plasma antibody binding, gp41 antigens associated with increased MTCT have conformations similar to nonnative forms of envelope such as monomers, uncleaved envelope, and gp41 stumps [15], which may have the ectodomain exposed. A limitation of our study is that 7 of the 21 infants that acquired HIV had an estimated time of infection after 6 months. Passively acquired HIV-specific antibodies wane in the infant over time and are often gone by 6 months [9]; therefore, it is unclear whether passively acquired antibodies were present near the time of transmission. The associations of maternal binding antibodies with odds of transmission were similar, but weaker, when transmissions after 6 months of age were excluded (data not shown). This may suggest a loss of statistical power or that gp41 ectodomain-specific antibodies are a marker of maternal disease progression or other risk factors.

iations of maternal binding antibodies with odds of transmission were similar, but weaker, when transmissions after 6 months of age were excluded (data not shown). This may suggest a loss of statistical power or that gp41 ectodomain-specific antibodies are a marker of maternal disease progression or other risk factors. CONCLUSIONS Ours is one of only a few studies to investigate the role of both maternal binding antibodies and infant passively acquired binding antibodies on risk of breastfeeding MTCT using samples with well timed infant infection data. The inconsistencies between our study and prior studies may be due to differences in infecting clades, availability of maternal and/or infant samples, ART treatment, or mode of transmission. Most of the aforementioned studies included in utero and peripartum transmissions, whereas our study focused on breastfeeding transmissions and cases in which the infants tested HIV negative at birth. This allowed us to assess the impact of passively acquired antibodies in HIV-uninfected infants and to observe their impact on incident infection risk. Our results suggest that antibody binding to the gp41 ectodomain is a correlate of increased risk of MTCT, rather than a correlate of decreased risk. Investigating the specific gp41 epitope(s) correlated with increased risk and the mechanism of this association has the potential to help guide future vaccine and therapeutic design.

suggest that antibody binding to the gp41 ectodomain is a correlate of increased risk of MTCT, rather than a correlate of decreased risk. Investigating the specific gp41 epitope(s) correlated with increased risk and the mechanism of this association has the potential to help guide future vaccine and therapeutic design. Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Figure S1. Correlation matrix of plasma binding to human immunodeficiency virus (HIV) antigens. Spearman correlation matrix of infant (left) or maternal (right) percentage plasma antibody binding to the panel of HIV antigens. Spearman rank correlation coefficients are color coded, with red denoting a negative correlation, blue denoting a positive correlation, and light gray denoting no correlation. Statistical significance was defined as P < .05 (*). Alternative non-BAMA assays (ELISAs or soluble CD4 blocking assays are noted in parentheses for appropriate antigens). jiz444_suppl_Supplementary_Figure_Legend Click here for additional data file. jiz444_suppl_Supplementary_Figure_S1 Click here for additional data file. jiz444_suppl_Supplementary_Material Click here for additional data file. jiz444_suppl_Supplementary_Table_1 Click here for additional data file. jiz444_suppl_Supplementary_Table_2 Click here for additional data file.

jiz444_suppl_Supplementary_Figure_Legend Click here for additional data file. jiz444_suppl_Supplementary_Figure_S1 Click here for additional data file. jiz444_suppl_Supplementary_Material Click here for additional data file. jiz444_suppl_Supplementary_Table_1 Click here for additional data file. jiz444_suppl_Supplementary_Table_2 Click here for additional data file. Notes Acknowledgments. We thank Peter Kim and Nitya Ramadoss for sharing 6-Helix, Xiangpeng Kong and Xunqing Jiang for providing the V1V2 peptides and scaffolds, the Duke Protein Production facility for providing the gp140 antigen and Mulvgp70 scaffold, and Marit van Gils, Rogier Sanders, and John Moore for providing the BG505 SOSIP trimer. We also thank Keshet Ronen and Adam Dingens for their help in optimizing the binding antibody multiplex assays and developing the antigen panel. Finally, we thank the participants and investigators of the Nairobi Breastfeeding Clinical Trial. Financial support. This work was funded by the National Institutes of Health (Grants R01AI076105 and F30AI136636). Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Presented in part: HIVR4P, October 2018, Madrid, Spain; CROI, March 2019, Seattle, WA; Annual Meeting of the University of Nairobi Collaborative Centre for Research and Training in HIV/AIDS/STIs, January 2019, Nairobi, Kenya.