Browse the corpus

Introduction Skeletal muscle dysfunction is an important complication of chronic obstructive pulmonary disease (COPD) occurring in mild as well as more advanced disease [1, 2]. It is largely driven by physical inactivity and is most apparent in muscles of the lower limb, in particular the quadriceps [3, 4]. Quadriceps weakness is associated with reduced exercise performance [5], poor health status [6], increased healthcare utilisation [7] and mortality, independent of airflow obstruction [8]. Changes in the quadriceps include muscle atrophy, loss of strength and endurance [9, 10], and a shift towards a less aerobic phenotype with reduced type I fibre proportions, capillarity and oxidative enzymes [11, 12]. The degree and nature of impairment varies widely between patients and cannot currently be predicted accurately without muscle biopsy [13]. The need for biomarkers of muscle phenotype has recently been highlighted, underscored by the prospect of new pharmacotherapies that selectively target certain muscle characteristics [14]. For example, peroxisome proliferator-activated receptor-δ agonists or 5′ AMP-activated protein kinase activators, which change the fuel preference within skeletal muscle away from glycolysis towards fat oxidation.

ored by the prospect of new pharmacotherapies that selectively target certain muscle characteristics [14]. For example, peroxisome proliferator-activated receptor-δ agonists or 5′ AMP-activated protein kinase activators, which change the fuel preference within skeletal muscle away from glycolysis towards fat oxidation. Computed tomography (CT) is an established method to assess muscle mass or quantity [15]. Images can also be analysed to assess muscle quality by the level of fatty infiltration of muscle, or muscle adiposity. Skeletal muscle adiposity increases with age [16] and, independent of mass, is associated with muscle triglyceride content [17], oxidative capacity [18], insulin resistance [19], muscle strength and power [20, 21], as well as immobility and hip fracture [22, 23]. Little is known about muscle adiposity in patients with COPD, but given the above, we predicted that CT-based measures could be valuable biomarkers of muscle phenotype in this group. In particular, their use may help identify patients with type I to type II muscle fibre shift as candidates for trials of targeted therapies. In this study, we hypothesised that mid-thigh intramuscular fat and muscle attenuation would be associated with physical activity, exercise capacity and fibre shift.

this group. In particular, their use may help identify patients with type I to type II muscle fibre shift as candidates for trials of targeted therapies. In this study, we hypothesised that mid-thigh intramuscular fat and muscle attenuation would be associated with physical activity, exercise capacity and fibre shift. Methods Study design and subjects Data were collated from patients with a diagnosis of COPD consistent with the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria [24] who had previously participated in two clinical studies at the Royal Brompton, King’s College or St Thomas’ Hospitals (London, UK). Some of the phenotypic data for some of these patients has been previously reported [6, 25]. Exclusion criteria were: diagnoses of heart, renal or liver failure; systemic inflammatory, metabolic or neuromuscular disorders; warfarin (bleeding risk from biopsy); or an exacerbation within the preceding 4 weeks. Healthy age-matched controls with no history or symptoms of respiratory or cardiovascular disease, or unresolved musculoskeletal injury were recruited by local advertisement. All participants had provided written informed consent for studies approved by the Joint University College London Committees on the Ethics of Human Research (Committee Alpha) and the Ethics Committee of the Royal Brompton and Harefield NHS Foundation Trust.

musculoskeletal injury were recruited by local advertisement. All participants had provided written informed consent for studies approved by the Joint University College London Committees on the Ethics of Human Research (Committee Alpha) and the Ethics Committee of the Royal Brompton and Harefield NHS Foundation Trust. Mid-thigh composition CT was performed on a 64-slice CT scanner (SOMATOM Sensation 64; Siemens, Erlangen, Germany) with the patient in a supine position. A single section of both mid-thighs, at a predefined level, was obtained using the following acquisition parameters: 50 mA·s and 120 kVp. The protocol was modified to deliver a reduced amount of radiation per scan. Images were viewed and analysed using SliceOMatic software (version 4.3; TomoVision, Magog, QC, Canada).

A single section of both mid-thighs, at a predefined level, was obtained using the following acquisition parameters: 50 mA·s and 120 kVp. The protocol was modified to deliver a reduced amount of radiation per scan. Images were viewed and analysed using SliceOMatic software (version 4.3; TomoVision, Magog, QC, Canada). Skeletal muscle and intramuscular fat were identified and quantified by use of standard Hounsfield unit thresholds, which represent the physical properties of tissues expressed in numerical form (skeletal muscle -29–150 HU; intramuscular fat -190– -30 HU) [26]. By specifying Hounsfield unit range, the software filters by tissue type and allows each to be quantified in turn. Tissue perimeters can be altered manually where necessary (fig. 1). Intramuscular fat was normalised to mid-thigh cross-sectional area (MTCSA) and expressed as a percentage. Skeletal muscle attenuation was measured using the mean radiation attenuation in Hounsfield units, which reflects macroscopic fatty muscle infiltration. A lower mean indicates less attenuation and greater fat infiltration. These methods are highly reliable with reported intra- and inter-class coefficients of variation between 0.2% and 4.8% [27, 28]. Figure 1– Mid-thigh compositional analysis based on a single slice computed tomography image. Tissues were differentiated according to standardised Hounsfield unit thresholds; skeletal muscle (red) -29–150 HU, adipose tissue (green) -190– -30 HU.

Skeletal muscle and intramuscular fat were identified and quantified by use of standard Hounsfield unit thresholds, which represent the physical properties of tissues expressed in numerical form (skeletal muscle -29–150 HU; intramuscular fat -190– -30 HU) [26]. By specifying Hounsfield unit range, the software filters by tissue type and allows each to be quantified in turn. Tissue perimeters can be altered manually where necessary (fig. 1). Intramuscular fat was normalised to mid-thigh cross-sectional area (MTCSA) and expressed as a percentage. Skeletal muscle attenuation was measured using the mean radiation attenuation in Hounsfield units, which reflects macroscopic fatty muscle infiltration. A lower mean indicates less attenuation and greater fat infiltration. These methods are highly reliable with reported intra- and inter-class coefficients of variation between 0.2% and 4.8% [27, 28]. Figure 1– Mid-thigh compositional analysis based on a single slice computed tomography image. Tissues were differentiated according to standardised Hounsfield unit thresholds; skeletal muscle (red) -29–150 HU, adipose tissue (green) -190– -30 HU. Additional measurements Lung volumes, transfer factor of the lung for carbon monoxide (TLCO), fat mass and fat free mass index (FFMI), daily step count and physical activity level (PAL), quadriceps maximum voluntary contraction (QMVC) and twitch tension, vastus lateralis fibre type and shift, exercise capacity as assessed by incremental shuttle walk test (ISWT) or 6-min walking distance (6MWD), and health-related quality of life were assessed as described previously [6, 25] and in the online supplementary material.

ximum voluntary contraction (QMVC) and twitch tension, vastus lateralis fibre type and shift, exercise capacity as assessed by incremental shuttle walk test (ISWT) or 6-min walking distance (6MWD), and health-related quality of life were assessed as described previously [6, 25] and in the online supplementary material. Statistical analysis Normally distributed continuous data are presented as mean±sd. Between-group comparisons were performed using unpaired t-test, Chi-squared tests or one-way ANOVA, as appropriate. Associations between measures of skeletal muscle adiposity and other variables were analysed using Pearson’s correlation coefficients, and univariate and multivariate linear regression models incorporating age, MTCSA, QMVC and TLCO % predicted as independent variables. For receiver operating characteristic (ROC) analysis, percentage intramuscular fat and skeletal muscle attenuation were used alone or in combination to discriminate subjects according to the presence of fibre shift. Statistical analysis and graphical presentations were performed using SPSS version 19 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 5 (GraphPad Software Inc., San Diago, CA, USA), respectively. No attempt at imputation of missing data was made. A p-value of <0.05 was regarded as statistically significant.

ibre shift. Statistical analysis and graphical presentations were performed using SPSS version 19 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 5 (GraphPad Software Inc., San Diago, CA, USA), respectively. No attempt at imputation of missing data was made. A p-value of <0.05 was regarded as statistically significant. Results Subjects CT data were available for 111 subjects: 101 COPD patients and 10 healthy age-matched controls (table 1). Consistent with a diagnosis of COPD, patients had impaired lung function and reduced arterial blood oxygen tension compared with controls. There were no significant differences in age, body mass index, FFMI or arterial blood carbon dioxide tension between the patient and healthy control groups, but patients had significantly reduced quadriceps strength, and mid-thigh and lean tissue cross-sectional area (table 1).

od oxygen tension compared with controls. There were no significant differences in age, body mass index, FFMI or arterial blood carbon dioxide tension between the patient and healthy control groups, but patients had significantly reduced quadriceps strength, and mid-thigh and lean tissue cross-sectional area (table 1). Table 1– Subject characteristics Healthy COPD p-value Subjects 10 101 Males/females 4/6 60/41 0.20 Age years 68.0±7.3 64.8±7.6 0.21 Height cm 166±9 169±9 0.42 Weight kg 75.8±14.6 70.0±15.1 0.25 Body mass index kg·m−2 27.2±3.7 24.5±4.8 0.09 Fat mass kg 22.4±9.7 27.0±6.0 0.15 Fat-free mass kg 48.8±12.1 47.7±9.0 0.71 Fat-free mass index kg·m−2 16.9±2.7 16.8±2.4 0.85 Current/former/never-smoker 1/4/5 25/76/0 <0.01 Smoking pack-years 10±15 49±24 <0.01 FEV1 L 2.59±0.64 1.08±0.54 <0.01 FEV1 % predicted 106±13 41±20 <0.01 VC L 3.72±0.95 3.00±0.79 <0.01 VC % predicted 121±19 89±19 <0.01 FEV1/VC 69.9±5.7 35.8±13.8 <0.01 TLC L 5.97±1.32 7.34±1.74 0.02 TLC % predicted 105±11 123±19 <0.01 RV L 2.31±0.55 4.26±1.49 <0.01 RV % predicted 102±12 189±60 <0.01 RV/TLC 39±6 56±12 <0.01 TLCO L 6.79±1.52 3.49±1.58 <0.01 TLCO % predicted 87±13 42±19 <0.01 PCO2 kPa 5.34±0.47 5.20±0.70 0.55 PO2 kPa 11.28±1.02 9.61±1.53 <0.01 GOLD stage I/II/III/IV 4/26/31/40 GOLD grade A/B/C/D 9/4/19/69 Exacerbations in the past year 1 (0–2) Medication n (%) Long-acting β-agonist 85 (84) Inhaled corticosteroid 85 (84) Oral steroid >5 mg·day−1 8 (8) Oxygen 18 (18) Nocturnal NIV 8 (8) SGRQ 8.2±9.9 49.9±19.0 <0.01 QMVC kg 31.5±9.2 26.7±7.4 0.05 Quadriceps twitch force kg 7.4±2.1 9.7±3.1 0.04 Body composition (both thighs) MTCSA cm2 221.8±44.9 190.9±47.6 0.04 Lean tissue cross-sectional area cm2 212.1±42.5 178.0±44.4 0.02 Lean tissue HU 42.3±4.0 41.5±4.8 0.63 Intramuscular fat cm2 11.8±5.1 13.0±7.4 0.64 Intramuscular fat % 4.30±1.23 6.64±3.32 0.03 Fibre type %# I 48.9±9.7 28.2±12.5 <0.01 IIa 44.3±8.5 57.0±13.4 0.01 IIx 3.2±4.6 12.1±13.1 0.02 II 47.5±10.1 67.9±15.2 <0.01 Exercise capacity# 6MWD m 572±81 341±141 <0.01 ISWT m 250±132 Physical activity# Wear time h per day 23.6±0.5 Wear time % 98.3±0.2 TEE kcal per day 1976±332 Steps per day 4501±3273 PAL 1.41±0.17 Data are presented as n, mean±sd or median (interquartile range), unless otherwise stated.

0.02 II 47.5±10.1 67.9±15.2 <0.01 Exercise capacity# 6MWD m 572±81 341±141 <0.01 ISWT m 250±132 Physical activity# Wear time h per day 23.6±0.5 Wear time % 98.3±0.2 TEE kcal per day 1976±332 Steps per day 4501±3273 PAL 1.41±0.17 Data are presented as n, mean±sd or median (interquartile range), unless otherwise stated. COPD: chronic obstructive pulmonary disease; FEV1: forced expiratory volume in 1 s; VC: vital capacity; TLC: total lung capacity; RV: residual volume; TLCO: transfer factor of the lung for carbon monoxide; PCO2: carbon dioxide tension; PO2: oxygen tension; GOLD: Global Initiative for Chronic Obstructive Lung Disease; NIV: noninvasive ventilation; SGRQ: St George’s respiratory questionnaire; QMVC: quadriceps maximum voluntary contraction; MTCSA: mid-thigh cross-sectional area; 6MWD: 6-min walking distance; ISWT: incremental shuttle walk test; TEE: total energy expenditure; PAL: physical activity level. #: COPD fibre type n = 48, ISWT n = 30, 6MWD and physical activity n = 69.

e’s respiratory questionnaire; QMVC: quadriceps maximum voluntary contraction; MTCSA: mid-thigh cross-sectional area; 6MWD: 6-min walking distance; ISWT: incremental shuttle walk test; TEE: total energy expenditure; PAL: physical activity level. #: COPD fibre type n = 48, ISWT n = 30, 6MWD and physical activity n = 69. Percentage intramuscular fat was significantly elevated in patients with COPD compared with healthy controls, with mean±sd values of 6.6±3.3% and 4.3±1.2%, respectively (p = 0.03). In the patient group, values stratified by GOLD stage revealed that fat infiltration tended to be greater in those with advanced disease, although the between group differences were not significant (GOLD I/II/III/IV 5.0±0.5/5.9±0.5/7.9±0.7/6.4±0.5%) (table 1). Skeletal muscle attenuation was numerically, but not significantly, different between patient and control groups (table 1), and was significantly reduced in the patients with most fat infiltration (table 2).

he between group differences were not significant (GOLD I/II/III/IV 5.0±0.5/5.9±0.5/7.9±0.7/6.4±0.5%) (table 1). Skeletal muscle attenuation was numerically, but not significantly, different between patient and control groups (table 1), and was significantly reduced in the patients with most fat infiltration (table 2). Table 2– Clinical characteristics of chronic obstructive pulmonary disease patients stratified by upper and lower quartile of percentage intramuscular fat Lower quartile Upper quartile p-value Subjects 25 25 Males/females 17/8 15/10 0.38 Age years 63.4±7.3 66.4±6.3 0.14 Height cm 169±8 168.2±8.7 0.60 Weight kg 63.9±15.2 75.9±15.5 0.01 Body mass index kg·m−2 22.0±4.2 26.7±5.2 0.001 Fat mass kg 16.1±6.4 26.0±9.5 0.001 Fat-free mass kg 47.7±10.5 49.8±8.0 0.43 Fat-free mass index kg·m−2 16.7±2.6 17.1±2.6 0.61 Smoking pack-years 49±19 57±29 0.32 FEV1 L 1.16±0.62 0.98±0.43 0.25 FEV1 % predicted 42±21 37±15 0.42 VC L 3.37±0.84 2.83±0.85 0.03 VC % predicted 94±18 85±17 0.08 FEV1/VC 32.6±13.4 35.2±11.8 0.49 TLC L 7.98±1.5 6.97±1.73 0.04 TLC % predicted 129±14 116±16 0.005 RV L 4.54±1.25 4.09±1.17 0.21 RV % predicted 201±58 179±42 0.14 RV/TLC 52±15 59±7 0.08 TLCO L 3.66±1.72 2.98±1.49 0.18 TLCO % predicted 42±18 37±18 0.35 PCO2 kPa 5.21±0.43 5.27±0.92 0.76 PO2 kPa 10.25±1.38 9.11±1.72 0.02 GOLD stage I/II/III/IV 1/7/6/11 0/5/13/7 0.19 GOLD grade A/B/C/D 2/2/3/18 1/0/2/17 0.27 Exacerbations in the past year 1 (0–2) 1 (0–2) 0.45 Medication n (%) Long-acting β-agonist 21 (84) 22 (88) 0.46 Inhaled corticosteroid 21 (84) 22 (88) 0.20 Oral steroid >5 mg·day−1 3 (12) 2 (8) 0.50 Oxygen 4 (16) 8 (32) 0.16 Nocturnal NIV 0 (0) 4 (16) 0.03 SGRQ 47.2±21.2 56.5±13.5 0.09 QMVC kg 25.4±8.2 26.1±7.6 0.76 Quadriceps twitch force kg 10.3±3.0 9.0±2.7 0.21 Body composition (both thighs) MTCSA cm2 190.9±49.2 188.2±50.4 0.85 Lean tissue cross-sectional area cm2 185.0±48.4 167.1±45.0 0.19 Lean tissue HU 44.9±4.7 37.6±4.9 <0.001 Intramuscular fat cm2 5.8±2.1 21.1±7.1 <0.001 Intramuscular fat % 2.94±0.80 11.22±2.20 <0.001 Fibre type %# I 29.3±13.3 21.8±10.5 0.14 IIa 60.9±16.3 59.3±11.6 0.79 IIx 8.4±8.7 21.0±18.4 0.04 II 69.3±14.1 76.8±11.4 0.68 Exercise capacity# 6MWD m 404±166 188±83 0.03 ISWT m 295±138 172±88 0.003 Physical activity# Wear time h per day 23.7±0.4 23.7±0.3 0.87 Wear time % 98.7±1.6 98.8±1.1 0.81 TEE kcal per day 1961±412 1925±296 0.76 Steps per day 4632±3407 3073±2167 0.11 PAL 1.44±0.17 1.33±0.12 0.04 Data are presented as n, mean±sd or median (interquartile range), unless otherwi

3 0.03 ISWT m 295±138 172±88 0.003 Physical activity# Wear time h per day 23.7±0.4 23.7±0.3 0.87 Wear time % 98.7±1.6 98.8±1.1 0.81 TEE kcal per day 1961±412 1925±296 0.76 Steps per day 4632±3407 3073±2167 0.11 PAL 1.44±0.17 1.33±0.12 0.04 Data are presented as n, mean±sd or median (interquartile range), unless otherwi se stated. FEV1: forced expiratory volume in 1 s; VC: vital capacity; TLC: total lung capacity; RV: residual volume; TLCO: transfer factor of the lung for carbon monoxide; PCO2: carbon dioxide tension; PO2: oxygen tension; GOLD: Global Initiative for Chronic Obstructive Lung Disease; NIV: noninvasive ventilation; SGRQ: St George’s respiratory questionnaire; QMVC: quadriceps maximum voluntary contraction; MTCSA: mid-thigh cross-sectional area; 6MWD: 6-min walking distance; ISWT: incremental shuttle walk test; TEE: total energy expenditure; PAL: physical activity level. #: comparison limited to n = 14 versus n = 11 for fibre type, n = 6 versus n = 5 for 6MWD and n = 19 versus n = 18 for ISWT and physical activity measures.

CSA: mid-thigh cross-sectional area; 6MWD: 6-min walking distance; ISWT: incremental shuttle walk test; TEE: total energy expenditure; PAL: physical activity level. #: comparison limited to n = 14 versus n = 11 for fibre type, n = 6 versus n = 5 for 6MWD and n = 19 versus n = 18 for ISWT and physical activity measures. Reproducibility For all measures based on a standard CT image, the mean differences between two researchers, or repeat assessments by the same researcher were less than 1% of the mean value for both assessments. The lower 95% confidence intervals of correlation coefficients for all assessments exceeded 0.99 and can be interpreted as excellent (tables S1a and S1b). A subset of 29 patients had an additional CT scan following 3 months participation in the placebo arm of a clinical trial. Overall, repeat assessments demonstrated good reproducibility with all lower 95% confidence intervals of correlation coefficients exceeding 0.70 (table S1c). Skeletal muscle adiposity and physical activity Percentage intramuscular fat was negatively correlated with step count (r = -0.27, p = 0.03) and PAL (r = -0.35, p = 0.003) (fig. 2), although only PAL was significantly reduced in patients in the upper compared with the lower quartile of this measure (p = 0.04) (table 2). In a regression model, percentage intramuscular fat was significantly associated with step count (β coefficient±se -258±121, p = 0.04) and PAL (β±se -0.02±0.01, p = 0.007) independent of age, MTCSA and QMVC, but not TLCO % pred.

atients in the upper compared with the lower quartile of this measure (p = 0.04) (table 2). In a regression model, percentage intramuscular fat was significantly associated with step count (β coefficient±se -258±121, p = 0.04) and PAL (β±se -0.02±0.01, p = 0.007) independent of age, MTCSA and QMVC, but not TLCO % pred. Figure 2– Relationship between a, c) mid-thigh percentage intramuscular fat and b, d) skeletal muscle attenuation and physical inactivity as assessed by a, b) mean daily step count and c, d) physical activity level (PAL). Skeletal muscle attenuation was positively correlated with step count (r = 0.36, p = 0.003) and PAL (r = 0.35, p = 0.004) (fig. 2) and was significantly associated with these variables independent of age, MTCSA and QMVC, but not TLCO % pred (step count β±se 206±100, p = 0.02; PAL β±se 0.13±0.01, p = 0.008). Skeletal muscle adiposity and exercise capacity Percentage intramuscular fat was negatively correlated with ISWT (r = -0.45, p<0.001), and 6MWD when analysed with (r = -0.49, p = 0.001) and without (r = -0.42, p = 0.02) control subjects (fig. 3). ISWT and 6MWD were significantly reduced among patients in the upper quartile of percentage intramuscular fat compared with those in the lower quartile (table 2). Percentage intramuscular fat was significantly associated with ISWT (β±se -16.0±4.5, p<0.001) and 6MWD (β±se -19.5±6.2, p = 0.004) independent of age, MTCSA, QMVC and TLCO % pred.

nificantly reduced among patients in the upper quartile of percentage intramuscular fat compared with those in the lower quartile (table 2). Percentage intramuscular fat was significantly associated with ISWT (β±se -16.0±4.5, p<0.001) and 6MWD (β±se -19.5±6.2, p = 0.004) independent of age, MTCSA, QMVC and TLCO % pred. Figure 3– Relationship between a, c) mid-thigh percentage intramuscular fat and b, d) skeletal muscle attenuation and exercise capacity as assessed by the a, b) incremental shuttle walk test (ISWT) and c, d) 6-min walking distance (6MWD). Skeletal muscle attenuation was positively correlated with ISWT distance (r = 0.37, p = 0.002) and 6MWD when analysed with (r = 0.57, p = 0.001) or without (r = 0.60, p = 0.001) control subjects (fig. 3), and was significantly associated with exercise performance independent of age, MTCSA, QMVC and TLCO % pred (ISWT β±se 9.8±3.8, p = 0.001; 6MWD β±se 12.7±3.8, p = 0.002).

with ISWT distance (r = 0.37, p = 0.002) and 6MWD when analysed with (r = 0.57, p = 0.001) or without (r = 0.60, p = 0.001) control subjects (fig. 3), and was significantly associated with exercise performance independent of age, MTCSA, QMVC and TLCO % pred (ISWT β±se 9.8±3.8, p = 0.001; 6MWD β±se 12.7±3.8, p = 0.002). Skeletal muscle adiposity and fibre shift Percentage intramuscular fat was negatively correlated with the percentage of type I fibres when analysed with (r = -0.36, p = 0.006) and without (r = -0.27, p = 0.05) control subjects (fig. S1). Percentage of type I fibres was not significantly different between patients in the upper and lower quartile of percentage intramuscular fat, but percentage of type IIx fibres was significantly greater in patients with most fat infiltration (21.8±18.4% versus 8.4±8.7%, p = 0.04). Percentage intramuscular fat was independently associated with percentage of type I fibres when considered with age, MTCSA and QMVC, but not TLCO % pred (β±se -1.3±0.6, p = 0.04). In a similar analysis, skeletal muscle attenuation was positively correlated with percentage of type I fibres (r = 0.30, p = 0.02) (fig. S1b) but was not independently associated when the other factors were considered.

ype I fibres when considered with age, MTCSA and QMVC, but not TLCO % pred (β±se -1.3±0.6, p = 0.04). In a similar analysis, skeletal muscle attenuation was positively correlated with percentage of type I fibres (r = 0.30, p = 0.02) (fig. S1b) but was not independently associated when the other factors were considered. To determine whether measures of skeletal muscle adiposity could predict fibre shift, a ROC analysis was performed with fibre shift, defined as percentage of type I fibres ≤27%, as the dependent variable. TLCO % pred was a stronger predictor of fibre shift than percentage intramuscular fat and skeletal muscle attenuation, with area under the curve (AUC) values (95% CI) of 0.825 (0.709–0.940) and 0.648 (0.7494–0.802), respectively (fig. 4). Combining measures of skeletal muscle adiposity with TLCO % pred, the AUC increased to 0.833 (fig. 4). The curve profile also changed such that, with a greater emphasis on sensitivity than specificity, measures of skeletal muscle adiposity could be used with TLCO % pred to identify >80% of patients with fibre shift with >65% specificity (fig. 4). Figure 4– Receiver operating characteristic curves for transfer factor of the lung for carbon monoxide (TLCO) % predicted and measures of skeletal muscle adiposity and areas under the curve (AUC) (95% CI). Optimal sensitivity/specificity values for the dark solid line: 88.9/50.0%; 83.3/66.7%; 77.8/82.2%; 72.2/91.1%.

To determine whether measures of skeletal muscle adiposity could predict fibre shift, a ROC analysis was performed with fibre shift, defined as percentage of type I fibres ≤27%, as the dependent variable. TLCO % pred was a stronger predictor of fibre shift than percentage intramuscular fat and skeletal muscle attenuation, with area under the curve (AUC) values (95% CI) of 0.825 (0.709–0.940) and 0.648 (0.7494–0.802), respectively (fig. 4). Combining measures of skeletal muscle adiposity with TLCO % pred, the AUC increased to 0.833 (fig. 4). The curve profile also changed such that, with a greater emphasis on sensitivity than specificity, measures of skeletal muscle adiposity could be used with TLCO % pred to identify >80% of patients with fibre shift with >65% specificity (fig. 4). Figure 4– Receiver operating characteristic curves for transfer factor of the lung for carbon monoxide (TLCO) % predicted and measures of skeletal muscle adiposity and areas under the curve (AUC) (95% CI). Optimal sensitivity/specificity values for the dark solid line: 88.9/50.0%; 83.3/66.7%; 77.8/82.2%; 72.2/91.1%. Skeletal muscle adiposity and COPD characteristics Neither percentage intramuscular fat nor skeletal muscle attenuation were significantly associated with sex, body mass index, FFMI, MTCSA or QMVC. The latter measures were not as consistently related to physical activity, exercise capacity or fibre shift as measures of skeletal muscle adiposity (table S2). Percentage intramuscular fat was not significantly associated with the other characteristics assessed except for fat mass (r = 0.38, p<0.001). Skeletal muscle attenuation was associated with age (r = -0.30, p = 0.002), TLCO % pred (r = 0.30, p = 0.003), quadriceps twitch force (r = 0.37, p = 0.01), MTCSA (r = 0.22, p = 0.02) and fat mass (r = -0.19, p = 0.04).

ficantly associated with the other characteristics assessed except for fat mass (r = 0.38, p<0.001). Skeletal muscle attenuation was associated with age (r = -0.30, p = 0.002), TLCO % pred (r = 0.30, p = 0.003), quadriceps twitch force (r = 0.37, p = 0.01), MTCSA (r = 0.22, p = 0.02) and fat mass (r = -0.19, p = 0.04). Discussion We have investigated CT-based measures of skeletal muscle adiposity in patients with COPD. Patients had a significantly elevated percentage of intramuscular fat compared with age-matched healthy controls. Moderate significant associations were found between percentage intramuscular fat and skeletal muscle attenuation, and measures of physical activity, exercise capacity and muscle fibre type independent of age, quadriceps strength and MTCSA. By combining measures of skeletal muscle adiposity with TLCO, >80% of subjects with quadriceps fibre shift could be identified with >65% specificity. Significance of the findings Our findings extend knowledge regarding the extent and nature of quadriceps muscle dysfunction in COPD. Reductions in muscle mass, strength and endurance are well known phenomena, as are fibre shift and loss of oxidative capacity [8, 9, 13]. Abnormally high levels of lipids have also been observed microscopically in the quadriceps of patients with COPD both under the fascia [29] and within muscle fibres [30, 31].

n COPD. Reductions in muscle mass, strength and endurance are well known phenomena, as are fibre shift and loss of oxidative capacity [8, 9, 13]. Abnormally high levels of lipids have also been observed microscopically in the quadriceps of patients with COPD both under the fascia [29] and within muscle fibres [30, 31]. Image-based measures of quadriceps muscle structure may help to identify patients with specific muscle phenotypes. Muscle biopsy data confirm that COPD muscle abnormalities are heterogeneous, unrelated to muscle strength and that, on average, only 50% of unselected patients will have pathological fibre shift or fibre atrophy [13]. For emerging pharmacotherapies that have fibre-type specific modes of action, such as those that affect the transcriptional coactivator [32] or stimulate fatty acid oxidation [33], practical biomarkers to identify relevant populations for study are desirable. Other biomarkers have been suggested for this purpose, including 31P-magnetic resonance spectroscopy (MRS) [34] and plasma microRNA-1 [35]. CT measures are less expensive to undertake than 31P-MRS, and are likely to be more rapidly available and more stable than any biomarker based on blood analysis.

udy are desirable. Other biomarkers have been suggested for this purpose, including 31P-magnetic resonance spectroscopy (MRS) [34] and plasma microRNA-1 [35]. CT measures are less expensive to undertake than 31P-MRS, and are likely to be more rapidly available and more stable than any biomarker based on blood analysis. Our findings also support data suggesting that emphysema (evaluated here via TLCO) rather than airflow obstruction is the more closely associated with skeletal muscle abnormalities [36] and strength [37] than FEV1, although in this it differs from the review by Gosker et al. [11], which found FEV1 was more strongly associated than TLCO at the study level. For study stratification, CT has the advantage that it can be scored centrally and is less patient- and operator-dependent than gas transfer measurement.

trength [37] than FEV1, although in this it differs from the review by Gosker et al. [11], which found FEV1 was more strongly associated than TLCO at the study level. For study stratification, CT has the advantage that it can be scored centrally and is less patient- and operator-dependent than gas transfer measurement. The accumulation of fat in the skeletal muscle of patients with COPD may reflect an intermediate between a type I to type II fibre shift and insulin resistance. Oxidative muscle fibres consume fats [38] and these results are consistent with the observation that transgenic animal models with an increased skeletal muscle oxidative fibre proportion have a reduced intramuscular triglyceride content, compared to wild-type controls, when exposed to a high-fat diet [39]. In humans, type I to type II fibre shift is associated with both insulin resistance and obesity [40, 41], and intramuscular fat is associated with [42] and may exacerbate insulin resistance [43]. Fat within skeletal muscle may also reflect fat tissue elsewhere in the body exceeding its storage capacity and the subsequent release of fatty acids into the circulation. Thus, intramuscular fat may also facilitate a stratified medicine approach as a biomarker of insulin resistance and cardiovascular risk in COPD more generally. The former could be investigated by measures of fasting insulin levels, for example, skeletal muscle adiposity with euglycaemic insulin clamp studies [44], and the latter by prospective observation of cardiovascular events in patients with COPD whose levels of intramuscular fat are known.

r risk in COPD more generally. The former could be investigated by measures of fasting insulin levels, for example, skeletal muscle adiposity with euglycaemic insulin clamp studies [44], and the latter by prospective observation of cardiovascular events in patients with COPD whose levels of intramuscular fat are known. Functional exercise performance was also associated with percentage intramuscular fat and skeletal muscle attenuation. Given the link between physical inactivity and muscle fibre shift, this probably reflects reduced quadriceps oxidative capacity and endurance, leading to early fatigue. This consolidates previous data in which our group demonstrated patients with fibre shift had poorer exercise performance than those without, independent of differences in MTCSA and quadriceps strength [25].

bre shift, this probably reflects reduced quadriceps oxidative capacity and endurance, leading to early fatigue. This consolidates previous data in which our group demonstrated patients with fibre shift had poorer exercise performance than those without, independent of differences in MTCSA and quadriceps strength [25]. Relationship to other work Two small studies have previously demonstrated greater levels of intramuscular fat in patients with COPD compared with healthy controls using CT and magnetic resonance imaging [45, 46]. Neither demonstrated an association between measures of skeletal muscle adiposity and functional performance. Sheilds et al. [47] recently demonstrated increased intermuscular fat, as assessed by magnetic resonance imaging, in the quadriceps of patients with COPD compared with healthy controls (mean increase 32%). As in the present study, fat infiltration, but not muscle cross-sectional area, was associated with increased anaerobic metabolism during exercise and reduced exercise performance [47]. Bioenergetic changes were not observed in the biceps brachii despite patients having been selected on the basis of weakness in the quadriceps [47].

sent study, fat infiltration, but not muscle cross-sectional area, was associated with increased anaerobic metabolism during exercise and reduced exercise performance [47]. Bioenergetic changes were not observed in the biceps brachii despite patients having been selected on the basis of weakness in the quadriceps [47]. Critique of the method Interobserver and interoccasion agreement for measures of skeletal muscle adiposity based on a standard CT image were excellent, with narrower limits of agreement than those reported during measurement of rectus femoris cross-sectional area with ultrasound [48]. Studies using CT report coefficients of variation between 1.5% and 2.5% for tissue cross-sectional area and <1% for muscle attenuation [49], with each HU increase equivalent to a 1% increase in lipid concentration when assessed microscopically [17]. Differences in serial measurements over 3 months in patients with stable COPD were also small and comparable to those found with CT-based measures of muscle or adipose cross-sectional area as confirmed by axial cadaver sections [50, 51]. The small measurement differences may be ascribed to operator error, particularly in the manual correction of tissue margins, which are required in the absence of a fully automated analysis procedure. Error relating to image acquisition should be minimal, particularly in comparison to other modalities that rely on surface land-marking, for example ultrasound. Nonetheless, the marginally greater differences between, as compared with within, observers underscores a need for training and standardisation of observers for longitudinal measurements.

isition should be minimal, particularly in comparison to other modalities that rely on surface land-marking, for example ultrasound. Nonetheless, the marginally greater differences between, as compared with within, observers underscores a need for training and standardisation of observers for longitudinal measurements. Muscle biopsies were obtained in a large subset of subjects in this study. Samples were used for fibre typing, and to measure inflammatory markers and other mediators to fulfil the aims of primary studies. Unfortunately, insufficient tissue remained to stain for intermyocellular and intramyocellular fat deposition, which would have served as criterion measures to validate the CT-based measures as surrogate markers of muscle adiposity. Direct measures of intramyocellular fat and myofilament density, and other measures of muscle phenotype such as capillarisation, oxidative enzyme content and glucose handling, would be helpful to examine in this population.

s criterion measures to validate the CT-based measures as surrogate markers of muscle adiposity. Direct measures of intramyocellular fat and myofilament density, and other measures of muscle phenotype such as capillarisation, oxidative enzyme content and glucose handling, would be helpful to examine in this population. Future longitudinal work should explore the accumulation and mechanisms of intramuscular fat, the responsiveness to change of these CT-based measurements and their ability to help predict clinical outcome, for example treatment response, risk of admission or cardiovascular events. In older adults, 12 weeks of high-intensity resistance training has been shown to increase quadriceps and hamstring muscle attenuation by 5.4% and 5.5%, respectively, with concurrent changes in strength and mass [52]. Skeletal muscle attenuation has also been used to predict community-based falls, hip fracture and functional mobility, independent of strength [22, 23]. Conclusion In conclusion, skeletal muscle adiposity assessed by CT imaging reflects important aspects of muscle dysfunction in patients with COPD and may help to identify patients with fibre shift. Future work should identify its responsiveness to intervention, both in proven therapies, for example exercise training, and with novel drugs where specific muscle phenotypes may be expected to respond differently. We thank all patients who kindly agreed to participate in the studies and members of the Lung Function Department at the Royal Brompton Hospital (London, UK) for their testing of study participants.

Conclusion In conclusion, skeletal muscle adiposity assessed by CT imaging reflects important aspects of muscle dysfunction in patients with COPD and may help to identify patients with fibre shift. Future work should identify its responsiveness to intervention, both in proven therapies, for example exercise training, and with novel drugs where specific muscle phenotypes may be expected to respond differently. We thank all patients who kindly agreed to participate in the studies and members of the Lung Function Department at the Royal Brompton Hospital (London, UK) for their testing of study participants. This article has supplementary material available from erj.ersjournals.com Support statement: M. Maddocks is a National Institute for Health Research (NIHR) Post-Doctoral Research Fellow. N.S. Hopkinson is a Higher Education Funding Council for England (HEFCE) Clinical Senior Lecturer. The study was supported by the NIHR Respiratory Biomedical Research Unit of the Royal Brompton and Harefield NHS Foundation Trust, and Imperial College London who part fund M.I. Polkey’s salary. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Conflict of interest: None declared.

To the Editor: Patients with chronic obstructive pulmonary disease (COPD) often suffer from acute exacerbations (AECOPD) of their disease, which have a significant impact on their health status [1]. Evidence suggests that ∼50% of these exacerbations are attributable to bacteria [2]. Pseudomonas aeruginosa can cause AECOPD and is associated with reduced survival in cystic fibrosis (CF) and bronchiectasis [3, 4]. Some studies have found that the presence of P. aeruginosa is also associated with mortality in COPD, but these findings have been based on patients hospitalised with exacerbations [5, 6] or those hospitalised with multidrug-resistant organisms [7]. The impact of P. aeruginosa identified in sputum from COPD outpatients is less clear but is an important issue for determining how aggressive strategies to attempt eradication should be. Therefore, we conducted a nested case–control study to investigate whether the isolation of P. aeruginosa in the sputum of a general COPD population was associated with long-term mortality.

PD outpatients is less clear but is an important issue for determining how aggressive strategies to attempt eradication should be. Therefore, we conducted a nested case–control study to investigate whether the isolation of P. aeruginosa in the sputum of a general COPD population was associated with long-term mortality. We first identified all sputum specimen results from Royal Brompton Hospital (London, UK) microbiology records between 2000 and 2012. These were cross-correlated with patients listed on our COPD research audit database. The reason for obtaining a sputum culture was not recorded systematically, but included repeated exacerbations and deterioration in symptom severity and/or clinical status, as well as opportunistic collection from patients with chronic sputum production. The laboratory threshold for a culture to be considered P. aeruginosa positive was 200 CFU·mL−1 (a semi-quantitative method). For patients with repeated positive sputum cultures the date of the first culture was recorded. Demographic variables, lung function measurements, gas transfer data, arterial blood gases and exacerbation frequency during the year prior to entering the study were recorded for all patients.

(a semi-quantitative method). For patients with repeated positive sputum cultures the date of the first culture was recorded. Demographic variables, lung function measurements, gas transfer data, arterial blood gases and exacerbation frequency during the year prior to entering the study were recorded for all patients. All analyses were performed using the Predictive Analytics Software (version 18; SPSS Inc., Chicago, IL, USA). Group comparisons were conducted utilising a t-test or the Chi-squared test as appropriate. Proportional Cox hazard analysis was utilised to assess: 1) the impact of parameters that differed between P. aeruginosa culture-positive and culture-negative groups and 2) the impact of antibiotic treatment on mortality. The proportionality hazard assumption was tested using partial residual plots (Schoenberg residuals proportionality hazard test). The Kaplan–Meier survival curve with the Log rank comparison was used to assess the impact of P. aeruginosa isolation on mortality and estimate the median survival of the P. aeruginosa culture-positive and culture-negative groups. A p-value <0.05 was considered significant.

nberg residuals proportionality hazard test). The Kaplan–Meier survival curve with the Log rank comparison was used to assess the impact of P. aeruginosa isolation on mortality and estimate the median survival of the P. aeruginosa culture-positive and culture-negative groups. A p-value <0.05 was considered significant. The initial cohort consisted of 380 subjects (247 male and 133 female), with a mean±sd age of 66.3±10.3 years and a forced expiratory volume in 1 s (FEV1) of 36.7±18.3 % predicted. 95 (25%) patients from the initial population had at least one positive sputum culture for P. aeruginosa, while the rest had either negative cultures or cultured bacteria other than P. aeruginosa. 66 P. aeruginosa culture-positive patients from the initial cohort were matched to another 66 P. aeruginosa culture-negative patients for sex, age and FEV1 % predicted.

had at least one positive sputum culture for P. aeruginosa, while the rest had either negative cultures or cultured bacteria other than P. aeruginosa. 66 P. aeruginosa culture-positive patients from the initial cohort were matched to another 66 P. aeruginosa culture-negative patients for sex, age and FEV1 % predicted. Differences in demographics and clinical characteristics between the initial COPD cohort and the study population, and P. aeruginosa culture-positive and culture-negative patients are presented in table 1. Although the initial matching was done according to three baseline variables, the two final groups were found to be similar in all recorded parameters of lung function (forced vital capacity, FEV1/forced vital capacity, diffusing capacity of the lung for carbon monoxide, transfer coefficient of the lung for carbon monoxide, total lung capacity, residual volume and residual volume/total lung capacity), as well as body mass index and arterial oxygen tension. Arterial carbon dioxide tension (p=0.018) and exacerbation rate (p=0.008) were higher in the P. aeruginosa culture-positive patients compared to the P. aeruginosa culture-negative patients (table 1). However, proportional Cox hazard analysis, which was conducted separately for each group, indicated that exacerbation rate had no impact on mortality (P. aeruginosa culture-positive: hazard ratio 1.334 (95% CI 0.351–5.067), p=0.672; P. aeruginosa culture-negative: hazard ratio 2.161 (95% CI 0.576–8.112), p=0.253). The results were similar for arterial carbon dioxide tension (P. aeruginosa culture-positive: hazard ratio 1.134 (95% CI 0.758–1.697), p=0.539; P. aeruginosa culture-negative: hazard ratio 1.074 (95% CI 0.842–1.369), p=0.572), indicating that it was not associated with survival.

io 2.161 (95% CI 0.576–8.112), p=0.253). The results were similar for arterial carbon dioxide tension (P. aeruginosa culture-positive: hazard ratio 1.134 (95% CI 0.758–1.697), p=0.539; P. aeruginosa culture-negative: hazard ratio 1.074 (95% CI 0.842–1.369), p=0.572), indicating that it was not associated with survival. Table 1– Baseline differences between the initial cohort and the study population, and the Pseudomonas aeruginosa culture-positive and culture-negative groups Initial cohort Study population p-value# P. aeruginosa culture-positive group P.aeruginosa culture-negative group p-value¶ Subjects n 380 132 66 66 Age years 65.4±10.9 68±8.8 0.011 68±8.8 68.3±9 0.969 Sex % Male 51.1 51.1 0.999 Female 47.4 47.4 BMI m·kg−2 24.7±5.2 24±6.3 0.257 23.6±5.8 24.4±6.8 0.497 FEV1 % predicted 38.6±20.1 (9–81.1) 33.6±14.7 (12.6–72.4) 0.008 33.4±14.5 (12.6–72.4) 33.8±14.9 (12.9–71) 0.882 FEV1/FVC % 40.6±21.2 (13–69.8) 36.1±16.8 (13.5–69.8) 0.029 35±12.1 (14.1–69.8) 37.2±20.7 (13.5–69.5) 0.456 DLCO % predicted 43.7±18.1 40.8±17.4 0.157 41.5±18.2 40.1±16.7 0.663 KCO % predicted 54.8±21.6 54±22.1 0.747 55.5±21.8 52.5±22.4 0.448 TLC % predicted 123.7±19.6 122.7±17 0.662 121.1±18.1 124.5±15.6 0.286 RV % predicted 205.1±61.9 203.3±53 0.780 206±52.8 200.5±53.6 0.569 RV/TLC 155.2±32.7 157±32.1 0.639 60.3±9.8 59.4±11.6 0.647 PaO2 kPa 9.5±1.4 9.0±1.5 0.014 9.1±1.4 8.9±1.8 0.440 PaCO2 kPa 5.3±0.8 5.5±1.1 0.116 5.3±0.9 5.8±1.4 0.018 Exacerbations per year 1.9±0.7 2.1±0.7 0.086 2.3±0.7 1.9±0.7 0.008 Data are presented as mean±sd or mean±sd (range), unless otherwise stated. Bold signifies statistical significance. BMI: body mass index; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DLCO: diffusing capacity of the lung for carbon monoxide; KCO: transfer coefficient of the lung for carbon monoxide; TLC: total lung capacity; RV: residual volume; PaO2: arterial oxygen tension; PaCO2: arterial carbon dioxide tension. #: level of significance for comparison between the initial cohort and the study population; ¶: level of significance for comparison between the P. aeruginosa culture-positive and culture-negative groups.

C: total lung capacity; RV: residual volume; PaO2: arterial oxygen tension; PaCO2: arterial carbon dioxide tension. #: level of significance for comparison between the initial cohort and the study population; ¶: level of significance for comparison between the P. aeruginosa culture-positive and culture-negative groups. Median (range) survival for the study population (n=132) was 81.2 (50.8–109.5) months. During this period 51 (38.6%) patients died, of these 52.9% (n=27) were P. aeruginosa culture-positive and 47.1% (n=24) were P. aeruginosa culture-negative. The Kaplan–Meier survival curve, using the Log rank comparison, indicated that P. aeruginosa positive sputum culture was not associated with mortality in this population. Median (range) survival for P. aeruginosa culture-positive patients was 80.1 (40.1–120.1) months compared to 88.6 (52–125.3) months in the P. aeruginosa culture-negative patients (p=0.49) (fig. 1). Furthermore, no differences in survival as a function of the presence of Pseudomonas would have been identified if the entire population of 380 had been analysed (data not shown). Figure 1– Kaplan–Meier survival curve for the Pseudomonas aeruginosa culture-positive and P. aeruginosa culture-negative groups. p=0.49 for P. aeruginosa culture-negative patients.

Median (range) survival for the study population (n=132) was 81.2 (50.8–109.5) months. During this period 51 (38.6%) patients died, of these 52.9% (n=27) were P. aeruginosa culture-positive and 47.1% (n=24) were P. aeruginosa culture-negative. The Kaplan–Meier survival curve, using the Log rank comparison, indicated that P. aeruginosa positive sputum culture was not associated with mortality in this population. Median (range) survival for P. aeruginosa culture-positive patients was 80.1 (40.1–120.1) months compared to 88.6 (52–125.3) months in the P. aeruginosa culture-negative patients (p=0.49) (fig. 1). Furthermore, no differences in survival as a function of the presence of Pseudomonas would have been identified if the entire population of 380 had been analysed (data not shown). Figure 1– Kaplan–Meier survival curve for the Pseudomonas aeruginosa culture-positive and P. aeruginosa culture-negative groups. p=0.49 for P. aeruginosa culture-negative patients. A secondary analysis was also undertaken to identify whether Pseudomonas eradication treatment among P. aeruginosa culture-positive patients had any impact on survival. Sputum samples were ordered on a clinical basis and the decision of whether to attempt to eradicate Pseudomonas was at the discretion of the treating physician. Out of 66 P. aeruginosa culture-positive patients, 34 (25.8%) did not receive any treatment while 32 (24.2%) were treated either with oral ciprofloxacin or appropriate intravenous antibiotics, depending on resistance pattern. These two patient subgroups were similar regarding age, sex, body mass index, pulmonary function testing variables and exacerbation rate (data not shown). Cox proportional hazard analysis indicated that P. aeruginosa treatment was not associated with mortality (hazard ratio 1.452 (95% CI 0.666–3.163), p=0.348) among P. aeruginosa culture-positive patients.

ar regarding age, sex, body mass index, pulmonary function testing variables and exacerbation rate (data not shown). Cox proportional hazard analysis indicated that P. aeruginosa treatment was not associated with mortality (hazard ratio 1.452 (95% CI 0.666–3.163), p=0.348) among P. aeruginosa culture-positive patients. This study indicates that a single isolate of P. aeruginosa in sputum from a general COPD population is not associated with worse survival. Recently, three different patterns of P. aeruginosa infection in COPD patients have been described: 1) short carriage and clearance of a P. aeruginosa strain; 2) acquisition of a new strain associated with an AECOPD; and 3) persistence of P. aeruginosa colonisation with unknown clinical significance [8]. Although the specific pattern of P. aeruginosa infection was not investigated in the current study, our results indicate that even though a single positive P. aeruginosa sputum culture was correlated with a higher exacerbation frequency, it was not associated with a worse long-term outcome in COPD outpatients.

nce [8]. Although the specific pattern of P. aeruginosa infection was not investigated in the current study, our results indicate that even though a single positive P. aeruginosa sputum culture was correlated with a higher exacerbation frequency, it was not associated with a worse long-term outcome in COPD outpatients. The results of our study may, to some extent, be limited by its retrospective design and by the relatively small number of patients included, although a formal power estimation was not conducted. However, most of the published studies in the field have used similar sample sizes to investigate the potential impact of P. aeruginosa infection on COPD survival [2, 6, 9]. Another limitation is the lack of baseline data on exercise capacity, dyspnoea severity and frequency of previous hospitalisations, all of which could have affected mortality. Moreover, high-resolution computed tomography of the chest was not systematically conducted in all patients so the prevalence of bronchiectasis, a factor which could have favoured P. aeruginosa infection, could not be estimated; although it should be noted that our institution runs separate clinics for patients known to have either CF or non-CF bronchiectasis. However, several of these prognostic factors [6, 7] and imaging data [6, 7, 9] are also lacking in previously published studies that investigated the effect of P. aeruginosa on COPD mortality. Furthermore, careful patient matching has controlled for most of the lung function parameters that could affect mortality [10], so these negative results could not be attributed to a selected patient subgroup.

also lacking in previously published studies that investigated the effect of P. aeruginosa on COPD mortality. Furthermore, careful patient matching has controlled for most of the lung function parameters that could affect mortality [10], so these negative results could not be attributed to a selected patient subgroup. In conclusion, this study indicated that a single P. aeruginosa sputum isolation is not a predictor of long-term mortality in a general COPD outpatient population. Current clinical practice usually targets P. aeruginosa eradication using radical treatment, after its isolation in the sputum of a COPD patient with clinical deterioration or frequent exacerbations. Although eradicating Pseudomonas may be justified to improve health status, patients and clinicians can, to an extent, be reassured by these data that if this is not possible the impact on survival of a single P. aeruginosa isolation is not as significant as it is in CF [3] or bronchiectasis [4]. Future prospective, case–control studies are needed in order to define exact criteria for chronic P. aeruginosa infection in COPD patients and the best stratification criteria for a clinical trial of P. aeruginosa eradication in COPD. Support statement: This study was supported by the NIHR Respiratory Biomedical Research Unit (Royal Brompton and Harefield NHS Foundation Trust and Imperial College, London, UK). Conflict of interest: None declared.

Introduction Chronic obstructive pulmonary disease (COPD) is characterised by airway remodelling that involves airway smooth muscle thickening, possibly caused by airway smooth muscle cell (ASMC) hypertrophy and/or hyperplasia [1]. ASMC dysfunction is caused, at least in part, by chronic exposure to inflammation-derived mediators, such as transforming growth factor (TGF)-β [2]. ASMCs from COPD patients show enhanced proliferation in response to TGF-β and fetal bovine serum (FBS), compared to ASMCs from healthy subjects [3]. However, the molecular mechanisms underlining ASMC dysfunction in COPD are not well understood.

ammation-derived mediators, such as transforming growth factor (TGF)-β [2]. ASMCs from COPD patients show enhanced proliferation in response to TGF-β and fetal bovine serum (FBS), compared to ASMCs from healthy subjects [3]. However, the molecular mechanisms underlining ASMC dysfunction in COPD are not well understood. Mitochondria are key regulators of metabolism, redox homeostasis and cell survival and proliferation [4]. Impaired mitochondrial function has been demonstrated in the large airways [5, 6] and lungs [7, 8] of patients with COPD, and may drive lung inflammation and remodelling [6–8]. Importantly, we have shown defective mitochondrial respiration in cultured COPD ASMCs [6]. Mitochondrial dysfunction associated with metabolic changes such as increased glycolysis and glutamine catabolism contribute to aberrant cellular growth in diseases such as pulmonary arterial hypertension (PAH) and cancer [9, 10]. Glycolytic intermediates feed into amino acid and fatty acid synthesis, and into the pentose phosphate pathway (PPP) to produce reduced nicotinamide adenine diphosphate (NADPH) required for redox homeostasis, and ribose-5-phosphate for nucleotide synthesis. Glutamine catabolism provides nitrogen for nucleotide and amino acid synthesis and glutamate for glutathione synthesis [11]. Therefore, these changes support macromolecule synthesis and maintain cellular redox balance, thereby facilitating cell growth and survival.

asis, and ribose-5-phosphate for nucleotide synthesis. Glutamine catabolism provides nitrogen for nucleotide and amino acid synthesis and glutamate for glutathione synthesis [11]. Therefore, these changes support macromolecule synthesis and maintain cellular redox balance, thereby facilitating cell growth and survival. The metabolomic profile of serum, urine, bronchoalveolar lavage fluid and exhaled breath condensates from COPD patients has been investigated in order to identify novel biomarkers for disease diagnosis and classification [12–20]. However, this approach does not indicate whether a different metabolic profile in lung structural cells, such as ASMCs, contributes to cellular dysfunction in COPD. We hypothesised that the mitochondrial dysfunction in COPD ASMCs is accompanied by metabolic and redox changes that may contribute to the increased capacity of COPD ASMCs to proliferate. To identify changes in metabolic pathways associated with the hyperproliferative phenotype of COPD ASMCs we investigated the global intracellular metabolome of ASMCs from healthy nonsmokers, healthy smokers and patients with COPD, at baseline and under the growth conditions of TGF-β and FBS. Materials and methods Additional details on the methods used in the study are provided in the online supplementary material.

We hypothesised that the mitochondrial dysfunction in COPD ASMCs is accompanied by metabolic and redox changes that may contribute to the increased capacity of COPD ASMCs to proliferate. To identify changes in metabolic pathways associated with the hyperproliferative phenotype of COPD ASMCs we investigated the global intracellular metabolome of ASMCs from healthy nonsmokers, healthy smokers and patients with COPD, at baseline and under the growth conditions of TGF-β and FBS. Materials and methods Additional details on the methods used in the study are provided in the online supplementary material. Subject demographics ASMCs were isolated from patients with mild/moderate COPD as defined by GOLD criteria, while healthy nonsmokers and healthy smokers, both current and ex-smokers, were used as controls. COPD patients showed significant airflow obstruction, as indicated by the forced expiratory volume in 1 s (FEV1) and the FEV1/forced vital capacity (FVC) ratio, had no history of asthma, gave a classical history of shortness of breath on exertion and were all smokers. The mean age of COPD patients was significantly higher than that of controls and smoking pack-year history was greater (tables 1 and 2). TABLE 1 Clinical characteristics of subjects who provided airway smooth muscle cells (ASMCs) used for metabolomics analysis

Subject demographics ASMCs were isolated from patients with mild/moderate COPD as defined by GOLD criteria, while healthy nonsmokers and healthy smokers, both current and ex-smokers, were used as controls. COPD patients showed significant airflow obstruction, as indicated by the forced expiratory volume in 1 s (FEV1) and the FEV1/forced vital capacity (FVC) ratio, had no history of asthma, gave a classical history of shortness of breath on exertion and were all smokers. The mean age of COPD patients was significantly higher than that of controls and smoking pack-year history was greater (tables 1 and 2). TABLE 1 Clinical characteristics of subjects who provided airway smooth muscle cells (ASMCs) used for metabolomics analysis Healthy nonsmokers Healthy smokers COPD Subjects 6 6 6 Age years 44.83±8.63 54.67±4.15 68.33±2.32*,# Male/female 4/2 4/2 6/0 Smoking (current/ex-smokers) NA 3/3 6/0 Smoking pack-years NA 31.67±5.80 61.20±11.10 FEV1 L 3.81±0.41 2.88±0.22 2.14±0.22* FEV1 % predicted 109.6±3.28 87.55±7.54 67.50±6.97** FVC L 4.87±0.50 3.69±0.30 3.66±0.18 FEV1/FVC % 78.17±2.91 78.28±1.72 58.39±5.03**,## Data are presented as n or mean±sem. COPD: chronic obstructive pulmonary disease; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; NA: not applicable. *: p<0.05, **: p<0.01 compared to healthy nonsmokers; #: p<0.05, ##: p<0.01 compared to healthy smokers. TABLE 2 Clinical characteristics of subjects who provided airway smooth muscle cells (ASMCs) used for the whole study

Healthy nonsmokers Healthy smokers COPD Subjects 6 6 6 Age years 44.83±8.63 54.67±4.15 68.33±2.32*,# Male/female 4/2 4/2 6/0 Smoking (current/ex-smokers) NA 3/3 6/0 Smoking pack-years NA 31.67±5.80 61.20±11.10 FEV1 L 3.81±0.41 2.88±0.22 2.14±0.22* FEV1 % predicted 109.6±3.28 87.55±7.54 67.50±6.97** FVC L 4.87±0.50 3.69±0.30 3.66±0.18 FEV1/FVC % 78.17±2.91 78.28±1.72 58.39±5.03**,## Data are presented as n or mean±sem. COPD: chronic obstructive pulmonary disease; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; NA: not applicable. *: p<0.05, **: p<0.01 compared to healthy nonsmokers; #: p<0.05, ##: p<0.01 compared to healthy smokers. TABLE 2 Clinical characteristics of subjects who provided airway smooth muscle cells (ASMCs) used for the whole study Healthy nonsmokers Healthy smokers COPD Subjects 7 8 8 Age years 40.71±6.33 56.50±3.34 66.63±2.13**,# Male/female 6/1 5/3 7/1 Smoking (current/ex-smokers) NA 4/4 6/2 Smoking pack-years NA 29.25±4.16 46.00±8.30 FEV1 L 4.22±0.30 2.68±0.28* 1.97±0.21** FEV1 % predicted 107.5±4.33 82.98±4.98* 62.25±5.86**,# FVC L 5.51±0.24 3.49±0.38** 3.44±0.25** FEV1/FVC % 76.40±3.54 76.91±1.59 56.71±3.87**,### Data are presented as n or mean±sem. COPD: chronic obstructive pulmonary disease; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; NA: not applicable. *: p<0.05, **: p<0.01 compared to healthy nonsmokers; #: p<0.05, ###: p<0.001 compared to healthy smokers.

FVC % 76.40±3.54 76.91±1.59 56.71±3.87**,### Data are presented as n or mean±sem. COPD: chronic obstructive pulmonary disease; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; NA: not applicable. *: p<0.05, **: p<0.01 compared to healthy nonsmokers; #: p<0.05, ###: p<0.001 compared to healthy smokers. ASMC isolation and culture ASMCs were isolated from endobronchial biopsies and from second- to fourth-generation segmental airways obtained during lung resection surgery from healthy nonsmoker and healthy smoker subjects and patients with COPD (tables 1 and 2), and placed in culture as described previously [21, 22]. The study was approved by the local ethics committee and informed consent was obtained from all participants. Untargeted metabolomics analysis Following treatment, ASMCs were detached, pelleted by centrifugation and stored at −80°C until processed. Sample preparation and analysis using ultra-high performance liquid chromatography-mass spectrometry or gas chromatography-mass spectrometry was performed by Metabolon (Durham NC, USA), as described previously [23]. Determination of differentially expressed metabolites Data preprocessing and normalisation was performed by Metabolon. Data are presented as “scaled intensity” and were re-scaled to have a median equal to one. Missing values were imputed with the minimum observed value. Differential expression analysis was performed using the Bioconductor R package limma (http://bioconductor.org/packages/release/bioc/html/limma.html).

med by Metabolon. Data are presented as “scaled intensity” and were re-scaled to have a median equal to one. Missing values were imputed with the minimum observed value. Differential expression analysis was performed using the Bioconductor R package limma (http://bioconductor.org/packages/release/bioc/html/limma.html). Supervised learning algorithm for phenotype classification Determination of the optimal number of differentially expressed metabolites was performed using the nearest shrunken centroid method [24], using an algorithm available in the Comprehensive R Archive Network (CRAN-pamr package, https://cran.r-project.org/). Data were adjusted for sex and age using the surrogate variable analysis package in Bioconductor, and principal component analysis was applied. Pathway analysis Pathway analysis was performed using the Pathway Activity Profiling algorithm, as previously described [25]. Determination of mitochondrial reactive oxygen species levels Mitochondrial reactive oxygen species (ROS) levels were determined using the mitochondrial-targeted, redox-sensitive fluorescent probe MitoSOX Red (Invitrogen, Paisley, UK) as previously described [6]. Determination of ASMC proliferation Changes in cell proliferation were determined by measuring BrdU incorporation using the Cell Proliferation ELISA kit (Roche Diagnostics, Burgess Hill, UK) according to the manufacturer's instructions. Alternatively, the numbers of live cells were determined by Trypan blue staining and haemocytometer counting.

iferation Changes in cell proliferation were determined by measuring BrdU incorporation using the Cell Proliferation ELISA kit (Roche Diagnostics, Burgess Hill, UK) according to the manufacturer's instructions. Alternatively, the numbers of live cells were determined by Trypan blue staining and haemocytometer counting. Statistical analysis Statistical analysis was performed using the GraphPad Prism v.5 software (GraphPad Software, San Diego, CA, USA). Unless specified otherwise, intragroup comparisons were performed using the Friedman test followed by Dunn's post hoc test, and intergroup comparisons used the Mann–Whitney test. Correlations were determined using Spearman's correlation coefficient. p<0.05 was considered as statistically significant. Results COPD ASMCs show a distinct metabolic profile Metabolomic analysis was performed directly after serum-starvation (0 h; baseline) and following 48 h incubation in the absence (unstimulated) and presence of TGF-β/FBS (growth conditions). Under these conditions, COPD ASMCs showed a distinct phenotype compared to ASMCs from healthy smokers, displaying increased proliferation in response to TGF-β/FBS, an effect inversely correlated with the subjects' lung function, and a lower α-smooth muscle actin mRNA expression (online supplementary figure E1).

tions). Under these conditions, COPD ASMCs showed a distinct phenotype compared to ASMCs from healthy smokers, displaying increased proliferation in response to TGF-β/FBS, an effect inversely correlated with the subjects' lung function, and a lower α-smooth muscle actin mRNA expression (online supplementary figure E1). Under unstimulated conditions, healthy nonsmoker and healthy smoker samples were separated from COPD samples in principal component (PC) 1 analysis (figure 1a and b). Following TGF-β/FBS treatment, healthy nonsmoker and COPD samples were separated along PC1 (figure 1c and d). The number and identities of differentially regulated metabolites are shown in online supplementary figure E2 and tables E2 and E3. The top differentially regulated metabolic pathways between COPD and healthy nonsmoker and smoker ASMCs included purine and pyrimidine metabolism, amino acid and fatty acid biosynthesis and degradation, pentose and glucuronate interconversions, glutathione metabolism and oxidative phosphorylation (online supplementary tables E4−E7). FIGURE 1 Plots of principal component analysis scores along principal component (PC) 1 and PC2 (a and c), or PC1 and PC3 (b and d) of differentially expressed metabolites, between chronic obstructive pulmonary disease (COPD) airway smooth muscle cell (ASMCs) and healthy nonsmoker and healthy smoker ASMCs, under unstimulated (a and b) and transforming growth factor-β/fetal bovine serum-stimulated (c and d) conditions.

d c), or PC1 and PC3 (b and d) of differentially expressed metabolites, between chronic obstructive pulmonary disease (COPD) airway smooth muscle cell (ASMCs) and healthy nonsmoker and healthy smoker ASMCs, under unstimulated (a and b) and transforming growth factor-β/fetal bovine serum-stimulated (c and d) conditions. Altered energy balance in COPD ASMCs ATP levels were not measured in this study; however, the ADP/AMP (figure 2a) and creatine phosphate (PCr)/creatine (Cr) ratios (figure 2b) were reduced, and inorganic phosphate levels were increased (figure 2c) in COPD ASMCs, compared to healthy nonsmoker and smoker ASMCs, under both unstimulated and TGF-β/FBS-stimulated conditions. These findings suggest lower ATP levels in COPD ASMCs, both in the absence and presence of mitogenic stimulation.

figure 2b) were reduced, and inorganic phosphate levels were increased (figure 2c) in COPD ASMCs, compared to healthy nonsmoker and smoker ASMCs, under both unstimulated and TGF-β/FBS-stimulated conditions. These findings suggest lower ATP levels in COPD ASMCs, both in the absence and presence of mitogenic stimulation. FIGURE 2 Relative ADP/AMP and creatine phosphate (PCr)/creatine (Cr) ratios and inorganic phosphate levels. Airway smooth muscle cells isolated from healthy nonsmokers (n=6), healthy smokers (n=6) and patients with chronic obstructive pulmonary disease (COPD) (n=6) were serum-starved overnight. Cell pellets were collected immediately after starvation (t=0) or after incubation in the absence or presence of transforming growth factor (TGF)-β (1 ng·mL−1) and fetal bovine serum (FBS) (2.5%) for 48 h. The ratios of the scaled intensities of a) ADP/AMP and b) PCr/Cr, and the scaled intensities of c) inorganic phosphate were determined in cell lysates using liquid chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01.

P/AMP and b) PCr/Cr, and the scaled intensities of c) inorganic phosphate were determined in cell lysates using liquid chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01. Altered glucose and nucleotide metabolism in COPD ASMCs Glucose levels were not significantly different across the three study groups (figure 3a). No differences were observed in any glycolytic intermediates (data not shown); however, the glycolytic products lactate (figure 3b) and alanine (figure 3c) were significantly increased in COPD ASMCs compared to healthy smoker ASMCs at baseline and after culture under unstimulated and TGF-β/FBS-stimulated conditions.

). No differences were observed in any glycolytic intermediates (data not shown); however, the glycolytic products lactate (figure 3b) and alanine (figure 3c) were significantly increased in COPD ASMCs compared to healthy smoker ASMCs at baseline and after culture under unstimulated and TGF-β/FBS-stimulated conditions. FIGURE 3 Relative levels of metabolites of glycolysis, pentose phosphate pathway and nucleotide metabolism. Airway smooth muscle cells isolated from healthy nonsmokers (n=6), healthy smokers (n=6) and patients with chronic obstructive pulmonary disease (COPD) (n=6) were serum-starved overnight. Cell pellets were collected immediately after starvation (t=0) or after incubation in the absence or presence of transforming growth factor (TGF)-β (1 ng·mL−1) and fetal bovine serum (FBS) (2.5%) for 48 h. The scaled intensities of a) glucose, b) lactate, c) alanine, d) ribose-5-phosphate, e) uridine, f) cytidine, g) thymidine and h) adenosine were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01.

idine, g) thymidine and h) adenosine were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01. In line with these findings, COPD ASMCs showed reduced baseline mRNA expression of peroxisome proliferator-activated receptor-γ coactivator (PGC)-1β, a key driver of mitochondrial respiration, and an increase in the baseline mRNA of pyruvate dehydrogenase kinase (PDK)-1, an enzyme that directs pyruvate away from the mitochondrion and towards glycolysis (online supplementary figure E3A−D) [4]. TGF-β/FBS stimulation reduced the mRNA of the mitochondrial gene activators PGC-1α and PGC-1β and increased the glycolytic genes PDK1 and lactate dehydrogenase A in healthy smoker ASMCs. In COPD ASMCs the TGF-β/FBS-mediated shift towards glycolytic gene activation was less pronounced, possibly due to their already high baseline glycolytic activity (online supplementary figure E3E−H). These findings suggest a metabolic shift towards glycolysis in COPD ASMCs. The glycolytic inhibitor 2-deoxy-d-glucose (2-DG) reduced TGF-β/FBS-induced DNA synthesis in both COPD and healthy smoker ASMCs, suggesting that glycolysis plays a key role in ASMC proliferation (online supplementary figure E3I).

figure E3E−H). These findings suggest a metabolic shift towards glycolysis in COPD ASMCs. The glycolytic inhibitor 2-deoxy-d-glucose (2-DG) reduced TGF-β/FBS-induced DNA synthesis in both COPD and healthy smoker ASMCs, suggesting that glycolysis plays a key role in ASMC proliferation (online supplementary figure E3I). Ribose-5-phosphate levels (figure 3d) were increased in TGF-β/FBS-stimulated COPD ASMCs, compared to healthy smoker ASMCs, suggesting an increased flow of glycolytic intermediates through the PPP. In line with this finding, the nucleosides uridine (figure 3e), cytidine (figure 3f), thymidine (figure 3g) and adenosine (figure 3h) were higher in COPD ASMCs compared to healthy nonsmoker and/or healthy smoker ASMCs under TGF-β/FBS stimulation. In addition, nucleoside levels in TGF-β/FBS-stimulated ASMCs correlated negatively with the FEV1/FVC ratio (online supplementary figure E4A−C). Nucleotide biosynthesis intermediates such as guanosine monophosphate, AMP and uridine monophosphate were also found to be elevated in TGF-β/FBS-stimulated COPD ASMCs (online supplementary tables E2 and E3), suggesting that the PPP may support increased nucleotide biosynthesis under growth conditions.

ry figure E4A−C). Nucleotide biosynthesis intermediates such as guanosine monophosphate, AMP and uridine monophosphate were also found to be elevated in TGF-β/FBS-stimulated COPD ASMCs (online supplementary tables E2 and E3), suggesting that the PPP may support increased nucleotide biosynthesis under growth conditions. Altered glutamine metabolism in COPD ASMCs Glutamine levels at baseline and after culture in the absence or presence of TGF-β/FBS were increased in COPD ASMCs compared to healthy smokers and nonsmokers (figure 4a), and negatively correlated with the FEV1/FVC ratio (online supplementary figure E5A−C). Glutamate (figure 4b) and γ-aminobutyrate (figure 4c), a glutamate metabolite, were significantly increased in TGF-β/FBS-treated COPD ASMCs compared to healthy nonsmokers, suggesting increased glutamine catabolism under growth conditions. Glutamine depletion partially attenuated TGF-β/FBS-induced DNA synthesis, suggesting a role of glutamine metabolism in ASMC proliferation (online supplementary figure E5D).

cantly increased in TGF-β/FBS-treated COPD ASMCs compared to healthy nonsmokers, suggesting increased glutamine catabolism under growth conditions. Glutamine depletion partially attenuated TGF-β/FBS-induced DNA synthesis, suggesting a role of glutamine metabolism in ASMC proliferation (online supplementary figure E5D). FIGURE 4 Relative levels of glutamine and glutamine catabolites. Airway smooth muscle cells isolated from healthy nonsmokers (n=6), healthy smokers (n=6) and patients with chronic obstructive pulmonary disease (COPD) (n=6) were serum-starved overnight. Cell pellets were collected immediately after starvation (t=0) or after incubation in the absence or presence of transforming growth factor (TGF)-β (1 ng·mL−1) and fetal bovine serum (FBS) (2.5%) for 48 h. The scaled intensities of a) glutamine, b) glutamate and c) γ-aminobutyrate (GABA) were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01.

tamate and c) γ-aminobutyrate (GABA) were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01. Altered fatty acid and amino acid metabolism in COPD ASMCs Most of the medium- and long-chain fatty acids (online supplementary table E8) detected, including caproate (figure 5a), myristoleate (figure 5b), caprylate (figure 5c) and vaccenate (figure 5d), were increased after culture under unstimulated and TGF-β/FBS-stimulated conditions in COPD ASMCs, compared to healthy nonsmokers and/or healthy smokers, indicating increased fatty acid synthesis or uptake. The ratios of acetylcarnitine (C2) to free carnitine (C0) (figure 5e) and the sum of C2 and propionylcarnitine (C3) to free carnitine ((C2+C3)/C0) (figure 5f), indices of fatty acid oxidation capacity and hexanoylcarnitine (C6) (figure 5g) levels were all reduced at baseline and under unstimulated conditions in healthy smoker and COPD ASMCs compared to healthy nonsmoker cells. In addition, the baseline C2/C0 ratio (figure 5e) and C6 (figure 5g) levels, and the (C2+C3)/C0 ratio in unstimulated cells (figure 5f) were lower in COPD ASMCs compared to healthy smoker ASMCs.

l reduced at baseline and under unstimulated conditions in healthy smoker and COPD ASMCs compared to healthy nonsmoker cells. In addition, the baseline C2/C0 ratio (figure 5e) and C6 (figure 5g) levels, and the (C2+C3)/C0 ratio in unstimulated cells (figure 5f) were lower in COPD ASMCs compared to healthy smoker ASMCs. FIGURE 5 Relative levels of fatty acids and intermediates of carnitine metabolism. Airway smooth muscle cells (ASMCs) isolated from healthy nonsmokers (n=6), healthy smokers (n=6) and patients with chronic obstructive pulmonary disease (COPD) (n=6) were serum-starved overnight. Cell pellets were collected immediately after starvation (t=0) or after incubation in the absence or presence of transforming growth factor (TGF)-β (1 ng·mL−1) and fetal bovine serum (FBS) (2.5%) for 48 h. The scaled intensities of a) caproate, b) myristoleate, c) caprylate, d) vaccenate and g) hexanoylcarnitine, as well as the e) ratio of acetylcarnitine (C2) to free carnitine (C0) and f) the ratio of the sum of C2 and propionylcarnitine (C3) to free carnitine (C0) were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01.

arnitine (C3) to free carnitine (C0) were determined in cell lysates by liquid chromatography mass spectrometry or gas chromatography mass spectrometry. Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01. The (C2+C3)/C0 ratios in untreated ASMCs correlated positively with the FEV1/FVC ratio and correlated negatively with age, suggesting an association of attenuated fatty acid oxidation with lung dysfunction and increasing age (online supplementary figure E6A−B). TGF-β/FBS restored the C2/C0 (figure 5e) and (C2+C3)/C0 (figure 5f) ratios and hexanoylcarnitine (figure 5g) levels in healthy smoker and COPD ASMCs, suggesting an increase in fatty acid oxidation during proliferation. Enhanced glutathione biosynthesis and reduced mitochondrial ROS levels in COPD ASMCs The ratio of reduced (GSH) to oxidised (GSSG) glutathione was similar between the study groups at baseline. However, after culture under unstimulated conditions, the GSH/GSSG ratio was lower in healthy smoker ASMCs and showed a statistically nonsignificant reduction in COPD ASMCs (figure 6a), reflecting oxidant–antioxidant imbalance in healthy smoker and COPD cells in the absence of mitogenic stimulation. In contrast, the GSH/GSSG ratio in TGF-β/FBS-treated COPD ASMCs was higher compared to healthy nonsmoker ASMCs, while healthy smoker ASMCs showed an increasing trend (figure 6a).

in COPD ASMCs (figure 6a), reflecting oxidant–antioxidant imbalance in healthy smoker and COPD cells in the absence of mitogenic stimulation. In contrast, the GSH/GSSG ratio in TGF-β/FBS-treated COPD ASMCs was higher compared to healthy nonsmoker ASMCs, while healthy smoker ASMCs showed an increasing trend (figure 6a). FIGURE 6 Relative ratios of reduced/oxidised glutathione and mitochondrial reactive oxygen species (ROS) levels. a) Airway smooth muscle cells (ASMCs) isolated from healthy nonsmokers (n=6), healthy smokers (n=6) and patients with chronic obstructive pulmonary disease (COPD) (n=6) were serum-starved overnight. Cell pellets were collected immediately after starvation (t=0) or after incubation in the absence or presence of transforming growth factor (TGF)-β (1 ng·mL−1) and fetal bovine serum (FBS) (2.5%) for 48 h. The ratio of the scaled intensities of reduced to oxidised glutathione (GSH/GSSG) was determined in cell lysates by liquid chromatography mass spectrometry; b) ASMCs isolated from healthy nonsmokers (n=7), healthy smokers (n=8) and patients with COPD (n=8) were serum-starved overnight and incubated in the absence or presence of TGF-β/FBS for 48 h. Mitochondrial ROS levels were determined using MitoSOX (Invitrogen, Paisley, UK) staining and expressed as median fluorescence intensity (MFI). Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01.

or 48 h. Mitochondrial ROS levels were determined using MitoSOX (Invitrogen, Paisley, UK) staining and expressed as median fluorescence intensity (MFI). Whiskers represent the spread of the data points; horizontal lines indicate the median value; and the + symbols indicate the mean of the values. *: p<0.05; **: p<0.01. In addition, TGF-β/FBS-treated COPD ASMCs had significantly lower mitochondrial ROS levels, and healthy smoker ASMCs showed a trend towards reduced levels, compared to healthy nonsmoker ASMCs (figure 6b). The glutathione synthesis inhibitor buthionine sulfoximine (10–25 µM) increased mitochondrial ROS levels both in the absence and presence of TGF-β/FBS (online supplementary figure E7A), and inhibited the increase in COPD ASMC number in response to TGF-β/FBS (online supplementary figure E7B). Thus, COPD, and to a lesser extent, healthy smoker ASMCs show improved redox homeostasis under growth conditions, which may contribute to their increased survival and proliferation.

line supplementary figure E7A), and inhibited the increase in COPD ASMC number in response to TGF-β/FBS (online supplementary figure E7B). Thus, COPD, and to a lesser extent, healthy smoker ASMCs show improved redox homeostasis under growth conditions, which may contribute to their increased survival and proliferation. Discussion We have demonstrated that COPD ASMCs exhibit a hyperproliferative phenotype associated with an altered metabolic profile in vitro. COPD cells show lower ATP levels, indicated by lower ADP/AMP and PCr/Cr ratios, both under unstimulated and growth conditions. In addition, fatty acid oxidation capacity was reduced in COPD ASMCs compared to healthy nonsmoker and smoker ASMCs under unstimulated conditions, but it was restored under growth conditions. COPD ASMCs showed increased levels of glutamine and the glycolytic products lactate and alanine, compared to healthy nonsmoker and/or smoker ASMCs, under both unstimulated and growth conditions. Additionally, TGF-β/FBS-stimulated COPD ASMCs showed higher levels of ribose-5-phosphate, indicating increased flow of glycolytic intermediates through the PPP, and accumulation of glutamine catabolites. Glycolysis, PPP and glutamine catabolism generate intermediates required for the biosynthesis of macromolecules and the maintenance of redox balance [11]. Indeed, fatty acid and amino acid levels were elevated in COPD ASMCs compared to healthy nonsmoker and/or smoker cells under unstimulated and growth conditions. Moreover, TGF-β/FBS-stimulated COPD ASMCs maintained higher levels of nucleotide biosynthesis intermediates, and a higher reduced to oxidised glutathione ratio and lower mitochondrial oxidant levels. Increased availability of macromolecules and maintenance of redox balance may support increased proliferation in COPD ASMCs.

over, TGF-β/FBS-stimulated COPD ASMCs maintained higher levels of nucleotide biosynthesis intermediates, and a higher reduced to oxidised glutathione ratio and lower mitochondrial oxidant levels. Increased availability of macromolecules and maintenance of redox balance may support increased proliferation in COPD ASMCs. Low ADP/AMP and PCr/Cr ratios and elevated inorganic phosphate levels indicate lower ATP levels in COPD ASMCs both in the absence and presence of mitogenic stimulation. This possibly reflects a reduction in mitochondrial respiration in COPD ASMCs, as previously described [6], and is consistent with a lower PGC-1β mRNA expression. Fatty acids interact with carnitine molecules, forming long-chain acylcarnitines that transport fatty acids to the mitochondrion and peroxisomes where they undergo fatty acid oxidation to produce acetyl-coenzyme A, NADH and FADH2 required for mitochondrial respiration [26]. Decreased baseline ratios of even-numbered (C2) and total (C2+C3) acylcarnitines to free carnitine (C0) suggest an impaired fatty acid oxidation capacity in COPD ASMCs, which may also contribute to the attenuated mitochondrial respiration. Carnitine levels are reduced in an elastase-induced mouse model of emphysema [27], while impaired fatty acid oxidation and lipid accumulation have been reported in ageing mice [28]. The accumulation of fatty acids observed in COPD ASMCs under unstimulated and growth conditions may result from reduced fatty acid oxidation and/or increased uptake or biosynthesis of fatty acid in these cells.

sema [27], while impaired fatty acid oxidation and lipid accumulation have been reported in ageing mice [28]. The accumulation of fatty acids observed in COPD ASMCs under unstimulated and growth conditions may result from reduced fatty acid oxidation and/or increased uptake or biosynthesis of fatty acid in these cells. Lactate, alanine and glutamine levels are elevated in COPD ASMCs, suggesting increased glycolytic activity and increased glutamine uptake or biosynthesis. COPD ASMCs showed elevated baseline mRNA expression of PDK1, which mediates the redirection of pyruvate towards lactate and alanine production [11], suggesting that the glycolytic shift possibly occurs downstream of pyruvate. This may explain our observation that COPD and healthy smoker ASMCs had the same sensitivity to the antiproliferative effect of 2-DG, an inhibitor of the first step of glycolysis [11]. Increased use of glycolysis and glutamine for energy production may be an adaptive response to mitochondrial dysfunction [9]. Reduced mitochondrial respiration in cigarette smoke extract-exposed lung epithelial cells has been shown to be associated with a shift towards glycolysis [29]. Glycolysis and glutamine catabolism support hyperproliferation and survival in cancer cells by providing precursors for biosynthesis and antioxidant protection. Glycolytic intermediates feed into fatty acid and amino acid biosynthesis, and into the PPP to generate ribose-5-phosphate for nucleotide synthesis, and NADPH to maintain redox balance [30]. Glutamine is catabolised to glutamate, donating its amide nitrogen for nucleotide synthesis. In addition, glutamate feeds into the Kreb's cycle through its conversion to α-ketoglutarate leading to the production of NADPH and lactate, and acts as a precursor for glutathione synthesis [11].

o maintain redox balance [30]. Glutamine is catabolised to glutamate, donating its amide nitrogen for nucleotide synthesis. In addition, glutamate feeds into the Kreb's cycle through its conversion to α-ketoglutarate leading to the production of NADPH and lactate, and acts as a precursor for glutathione synthesis [11]. In addition to elevated fatty acid levels, COPD ASMCs showed an increase in the majority of amino acids (online supplementary table E9) under both unstimulated and growth conditions. This increased availability of fatty acid and amino acids may be a result of increased biosynthesis; however, autophagy may also contribute to this effect [31]. Moreover, under growth conditions COPD ASMCs showed evidence of enhanced nucleotide biosynthesis and augmented antioxidant protection, reflected by a higher GSH/GSSG ratio and lower mitochondrial ROS levels. The increased PPP activity and glutamine catabolism observed in COPD ASMCs under growth conditions possibly drives these processes through the production of nucleotide precursors and NADPH. Enhanced glutathione biosynthesis may also be involved in the enhanced antioxidant response. Glutamate, a constituent of glutathione, and S-adenosyl methionine and cystathionine intermediates of the methionine cycle and transulfuration [32], which provide cysteine for glutathione synthesis, are increased in TGF-β/FBS-stimulated COPD ASMCs (online supplementary figure E7C−D).

the enhanced antioxidant response. Glutamate, a constituent of glutathione, and S-adenosyl methionine and cystathionine intermediates of the methionine cycle and transulfuration [32], which provide cysteine for glutathione synthesis, are increased in TGF-β/FBS-stimulated COPD ASMCs (online supplementary figure E7C−D). COPD ASMCs show evidence of reduced mitochondrial respiration accompanied by increased glycolysis and glutamine utilisation, processes that support biosynthesis and antioxidant responses. The greater availability of biosynthetic intermediates and antioxidant protection may help drive the associated enhanced proliferation seen in COPD cells [3]. A similar metabolic phenotype, involving reduced mitochondrial respiration and increased glycolysis, PPP activity and glutamine utilisation, associated with increased biosynthetic activity, has been shown to contribute to increased vascular smooth muscle cell and endothelial cell growth in PAH [33–35]. Thus, the metabolic reprogramming observed in COPD ASMCs may contribute to their hyperproliferative phenotype. This is supported by our findings showing attenuation of TGF-β/FBS-mediated COPD ASMC proliferation by glycolysis and glutathione synthesis inhibition, and glutamine depletion. These mechanisms merit further investigation.

abolic reprogramming observed in COPD ASMCs may contribute to their hyperproliferative phenotype. This is supported by our findings showing attenuation of TGF-β/FBS-mediated COPD ASMC proliferation by glycolysis and glutathione synthesis inhibition, and glutamine depletion. These mechanisms merit further investigation. The molecular mechanisms underlying the metabolic shift in COPD ASMCs are currently unknown. In line with our findings, studies in COPD lung tissue and airway epithelial cells have reported downregulation of genes involved in mitochondrial function, including oxidative phosphorylation, and increased expression of genes involved in glycolysis, PPP and glutathione synthesis [36–38]. Prolonged exposure to cigarette smoke may play a role in these changes as ASMCs from healthy smokers show distinct metabolic differences such as in fatty acid oxidation and methionine metabolism compared to healthy nonsmokers. Metabolic reprogramming is known to be driven by mitochondrial dysfunction and pathways such as the PI3K/Akt, mTOR and hypoxia-inducible factor-1α, which play a key role in COPD pathogenesis [10, 11, 39]. We cannot exclude the possibility that some of these changes may be epiphenomena rather than direct causes of the aberrant phenotype of COPD ASMCs. Future studies will aim to validate and elucidate these mechanisms and investigate their possible role as drivers of the defective airway smooth muscle function in COPD.

]. We cannot exclude the possibility that some of these changes may be epiphenomena rather than direct causes of the aberrant phenotype of COPD ASMCs. Future studies will aim to validate and elucidate these mechanisms and investigate their possible role as drivers of the defective airway smooth muscle function in COPD. A limitation of our study is the limited number of subjects. Nevertheless, in this preliminary study, we were able to show significant differences in the metabolomic profile of COPD ASMCs. Another limitation is the higher mean age of the COPD patients, which may be a confounding factor in our study, as age is associated with impaired cellular metabolic activity [40]. We cannot exclude the possibility that some of the metabolic changes we observe in COPD ASMCs are age-related; however, age cannot entirely explain the differences we observed between COPD and controls. In conclusion, we demonstrate that COPD ASMCs demonstrate a distinct metabolic and redox profile compared to those from healthy nonsmokers and smokers. This involves a shift in glucose and glutamine metabolism that may support increased biosynthesis and enhanced antioxidant levels. These metabolic changes are associated with increased cellular growth, and thus may be molecular targets for reversing airway smooth muscle dysfunction in COPD. Supplementary material 10.1183/13993003.00202-2017.Supp1Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author. Online Supplement ERJ-00202-2017_Supplement

In conclusion, we demonstrate that COPD ASMCs demonstrate a distinct metabolic and redox profile compared to those from healthy nonsmokers and smokers. This involves a shift in glucose and glutamine metabolism that may support increased biosynthesis and enhanced antioxidant levels. These metabolic changes are associated with increased cellular growth, and thus may be molecular targets for reversing airway smooth muscle dysfunction in COPD. Supplementary material 10.1183/13993003.00202-2017.Supp1Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author. Online Supplement ERJ-00202-2017_Supplement Disclosures 10.1183/13993003.00202-2017.Supp2I.M. Adcock ERJ-00202-2017_Adcock K.F. Chung ERJ-00202-2017_Chung D.K. Finch ERJ-00202-2017_Finch P. Kirkham ERJ-00202-2017_Kirkham

Supplementary material 10.1183/13993003.00202-2017.Supp1Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author. Online Supplement ERJ-00202-2017_Supplement Disclosures 10.1183/13993003.00202-2017.Supp2I.M. Adcock ERJ-00202-2017_Adcock K.F. Chung ERJ-00202-2017_Chung D.K. Finch ERJ-00202-2017_Finch P. Kirkham ERJ-00202-2017_Kirkham Acknowledgements The COPDMAP collaborators are as follows. Peter J. Barnes, Airways Disease Division, National Heart and Lung Institute, Imperial College London, London, UK; Christopher E. Brightling, Institute of Lung Health, Dept of Infection, Inflammation and Immunity, University of Leicester, Leicester, UK; Donna E. Davies, The Brooke Laboratories, Division of Infection, Inflammation and Repair, School of Medicine, University of Southampton, Southampton General Hospital, Southampton, UK; Andrew J. Fisher, Institute of Cellular Medicine, Newcastle University, and Cardiopulmonary Transplantation, Institute of Transplantation, Freeman Hospital, Newcastle upon Tyne, UK; Alasdair Gaw, Innovate UK, Technology Strategy Board, Swindon, UK; Alan J. Knox, Centre for Respiratory Research, City Hospital, Nottingham, UK; Ruth J. Mayer, GlaxoSmithKline, King of Prussia, PA, USA; Michael Polkey, NIHR Respiratory Biomedical Research Unit, Royal Brompton and Harefield NHS Foundation Trust and Imperial College London, London; Michael Salmon, Biology Discovery, Merck Research Laboratories, Boston, MA, USA; Yolanda Sanchez, GlaxoSmithKline; Dave Singh, University of Manchester, Medicines Evaluation Unit, University Hospital of South Manchester NHS Foundation Trust, Manchester, UK; Ruth Tal-Singer, GlaxoSmithKline.

rial College London, London; Michael Salmon, Biology Discovery, Merck Research Laboratories, Boston, MA, USA; Yolanda Sanchez, GlaxoSmithKline; Dave Singh, University of Manchester, Medicines Evaluation Unit, University Hospital of South Manchester NHS Foundation Trust, Manchester, UK; Ruth Tal-Singer, GlaxoSmithKline. This article has supplementary material available from erj.ersjournals.com Support statement: This study was supported by the MRC-ABPI COPD-MAP consortium (G1001367/1) and a Dunhill Medical Trust grant (R368/0714). It was also supported by the NIHR Respiratory Disease Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College London. The Canadian Respiratory Research Network (CRRN) is supported by grants from the Canadian Institutes of Health Research (CIHR), Institute of Circulatory and Respiratory Health; Canadian Lung Association (CLA); British Columbia Lung Association; and industry partners Boehringer-Ingelheim Canada, AstraZeneca Canada, Novartis Canada and GlaxoSmithKline. The funders had no role in the study design, data collection and analysis, or preparation of the manuscript. Funding information for this article has been deposited with the Crossref Funder Registry. Conflict of interest: Disclosures can be found alongside this article at erj.ersjournals.com