Annals of the American Thoracic Society

Rationale: The mechanism by which low-dose macrolide therapy reduces exacerbations in non–cystic fibrosis bronchiectasis is not known. Pseudomonas aeruginosa quorum sensing controls the expression of a range of pathogenicity traits and is inhibited by macrolide in vitro. Quorum sensing inhibition renders P. aeruginosa less pathogenic, potentially reducing its contribution to airway damage.

Objectives: The aim of this study was to determine whether long-term low-dose erythromycin inhibits P. aeruginosa quorum sensing within the airways of patients with non–cystic fibrosis bronchiectasis.

Methods: Analysis was performed on induced sputum from P. aeruginosa–positive subjects at recruitment to the BLESS (Bronchiectasis and Low-Dose Erythromycin Study) trial and after 48 weeks of treatment with erythromycin or placebo. To avoid changes in gene expression during culture, bacterial mRNA was extracted directly from sputum, and the relative expression of functionally critical quorum sensing genes was determined by quantitative polymerase chain reaction.

Measurements and Main Results: In keeping with the BLESS study, a significant reduction in total exacerbations was seen in this subgroup (placebo: 6, [interquartile range (IQR), 4–8]; erythromycin: 3, [IQR, 3–4]; P = 0.008; Mann-Whitney test). Erythromycin therapy did not change P. aeruginosa bacterial load determined by polymerase chain reaction. A significant reduction was observed in the expression of the quorum sensing genes, lasR (erythromycin: fold change, 0.065 [IQR, 0.01–0.85], n = 11; placebo: fold change, 1.000 [IQR, 0.05–3.05]; P = 0.047, Mann-Whitney U test) and pqsA (erythromycin: fold change, 0.07 [IQR, 0.02–0.25]; placebo: fold change, 1.000 [IQR, 0.21–4.31], P = 0.017, Mann-Whitney U test), after 48 weeks of erythromycin, compared with placebo.

Conclusions: We demonstrate inhibition of P. aeruginosa quorum sensing within the airways of patients with non–cystic fibrosis bronchiectasis receiving long-term, low-dose erythromycin, without a reduction in bacterial load, representing a potential mechanism of therapeutic impact beyond a classical antimicrobial or antiinflammatory pathway.

Long-term, low-dose macrolide therapy has been shown to be effective in the reduction of exacerbations in an increasing number of chronic inflammatory lung diseases, particularly non–cystic fibrosis bronchiectasis (1, 2). Although the mechanism of efficacy has not been clearly demonstrated, on the basis of in vitro data it has been proposed that macrolides achieve clinical benefit through an antiinflammatory effect (3). However, whether such an effect exists in vivo, and whether it accounts for the clinical benefit associated with macrolide therapy, has not been demonstrated.

Macrolides are bacteriostatic through interference with bacterial protein biosynthesis. Although effective against a broad range of species (particularly gram-positive bacteria) the doses used in the treatment of chronic lung diseases result in airway concentrations that are subinhibitory to the key airway pathogen, Pseudomonas aeruginosa. However, patients whose airway infections are dominated by P. aeruginosa are particularly responsive to macrolide therapy (1, 4). This benefit is independent of a reduction in P. aeruginosa load (4), suggesting a mechanism other than a classical antimicrobial effect.

In vitro studies have shown macrolides to interfere with P. aeruginosa quorum sensing (58), a central regulatory system that controls the expression of a wide range of virulence and pathogenicity traits (9). These traits include the production of secreted factors and secondary metabolites, such as elastase and pyocyanin, which contribute directly to airway damage and lung function decline (10). Inhibition of quorum sensing (via multiple mechanisms, including inhibition of gene expression [11]) suppresses these traits, rendering P. aeruginosa substantially less pathogenic and less fit in vitro (12, 13). Such an effect is likely to reduce both the direct contribution of P. aeruginosa to airway damage and its stimulation of a chronically elevated, deleterious host immune response.

Here, we hypothesized that the macrolide erythromycin would lead to inhibition of P. aeruginosa quorum sensing gene expression within the airways of patients with non–cystic fibrosis bronchiectasis.

Study Population

Analysis was performed on a subgroup of induced sputum samples collected as part of the BLESS (Bronchiectasis and Low-Dose Erythromycin Study) trial (1), a 12-month, double-blind, randomized, placebo-controlled trial comparing placebo with low-dose erythromycin ethylsuccinate (400 mg twice daily; equivalent to 250 mg twice daily of erythromycin base) in patients with non–cystic fibrosis bronchiectasis.

The subgroup for this study was defined as those subjects whose sputum was polymerase chain reaction (PCR)-positive for P. aeruginosa after 48 weeks of treatment with either erythromycin or placebo. Those subjects were screened for the presence of amplifiable bacterial mRNA, and only those samples from which we could reliably amplify two housekeeping genes (rpoD and rpsL) were further analyzed. After this screening process, 15 patients in the placebo arm and 11 patients in the treatment arm were analyzed (as per the flow chart in Figure 1). This subgroup was indistinguishable from the original BLESS cohort at baseline (see online supplement).

Baseline demographic and disease characteristics of patients contributing to the current study are shown in Table 1. Lung function, sputum weight, and quality of life (St. George’s Respiratory Questionnaire) assessments were assessed as in the BLESS (1) trial.

Table 1. Baseline clinical characteristics of the placebo- and erythromycin-treated groups

 Placebo (n = 15)Erythromycin (n = 11)P Value
Age, mean ± SD, yr63.47 ± 7.8863.64 ± 6.250.95
Sex, % female ± SD47 ± 5163 ± 500.41
Ex-smokers, % ± SD33 ± 4818 ± 400.41
Postbronchodilator FEV1, % predicted, mean ± SD63.5 ± 19.157.8 ±  18.40.46
24-h Sputum weight, mean ± SD, g/d17.7 ± 12.318.4 ± 10.40.89
C-reactive protein, mean ± SD, mg/L5.7 ± 5.65.4 ± 4.50.86

Baseline characteristics of the cohort tested. These subjects were Pseudomonas aeruginosa–positive and had amplifiable RNA. There was no statistically significant difference between the two groups at baseline.

Sample Collection

Induced sputum samples were collected after 48 weeks of low-dose erythromycin. Sputum induction was performed according to the standardized protocol recommended by the European Respiratory Society Task Force (14). Sputum samples were aliquoted into 2-ml collecting tubes and immediately frozen at −80°C.

Assessment of P. aeruginosa Load in Induced Sputum

DNA was extracted from sputum using the QIAamp DNA mini kit (Qiagen, Germantown, MD) according to manufacturer’s instructions, but with an extended initial lysis step (3 h). P. aeruginosa levels in induced sputum were determined by quantitative PCR (qPCR) as described previously (15). Levels were determined by comparison with a standard curve.

RNA Extraction

RNA was extracted from cell pellets using the High Pure RNA Isolation Kit (Roche, Indianapolis, IN) with minor modifications (a detailed protocol is provided in the online supplement). A more stringent lysis process was required for sputum, whereby sputum cell pellets were resuspended in Trizol (Life Technologies, Scoresby, Australia) and underwent physical cell disruption. An extended on-column DNase step was also required to ensure the removal of contaminating bacterial DNA.

RNA concentration and purity were determined using the NanoDrop 1000 spectrophotometer (Thermo Scientific, Scoresby, Australia). A total of 200 ng of total RNA extract was used as template for cDNA generation, using the SensiFAST cDNA Synthesis Kit (Bioline, London, UK). The absence of amplification of contaminating bacterial genomic DNA was determined by concomitant PCR amplification of cDNA reactions lacking reverse transcriptase.

qPCR

qPCR assays were performed in duplicate using Sybr Green (Takara Bio Inc., Seoul, Korea) (rpoD, rpsL, lasR, lasI, rhlR, rhlI, pqsA, mexB, mexX) and a custom Taqman gene expression assay (oprL). PCR primers are shown in Table E1 in the online supplement. In-house primer design was performed using Primer3 (http://bioinfo.ut.ee/primer3-0.4.0/primer3/). Primer sequences can be found in the online supplement. All primers were assessed for uniqueness to P. aeruginosa by comparison with the Nucleotide BLAST database and by confirming lack of amplification of DNA from common respiratory pathogens and human cells. For each gene, expression levels were normalized to the expression levels of two housekeeper genes, rpoD and rpsL, and then expressed as fold change relative to the median of the housekeeper-corrected results of the placebo group.

Measurement of in vivo Interleukin Levels

IL-8 and IL-1β concentrations in induced sputum samples were measured using ELISA as per the manufacturer’s instruction (Set II; BD Biosciences, San Jose, CA).

Determination of Absolute and Percentage Neutrophil Count

Sputolysin (Merck Millipore, Darmstadt, Germany) was added to the sputum and the sample homogenized by vortexing and rolling. The sample was centrifuged at 500 × g for 10 min, the supernatant removed, and the cell pellet resuspended in phosphate-buffered saline for cell counting in the presence of trypan blue and the Rapid Giemsa Stain System (Amber Scientific, Midvale, WA, Australia).

Statistical Analysis

We estimated that 26 patients would need to be enrolled for this substudy to have 80% power to detect a 90% reduction in gene expression, assuming a two-sided α level of 0.05.

The significance of variance between placebo and treatment was determined using t tests and Mann-Whitney U tests for parametric and nonparametric data, respectively. The normality of data distribution was determined by the Shapiro-Wilk and the D’Agostino and Pearson omnibus normality tests. All analyses were two sided, and P values < 0.05 were considered significant. Statistical analyses were performed using GraphPad Prism statistical software (version 6.0, CA).

Induced sputa were assessed from individuals recruited to the BLESS randomized controlled trial who were P. aeruginosa positive (as determined by PCR) and who had received 12 months of either erythromycin (n = 11) or placebo (n = 15). There was no statistically significant difference in baseline demographic data between these two groups (Table 1) or between these subjects and the wider patient population in the original BLESS cohort. To determine whether erythromycin therapy results in a reduction in P. aeruginosa load in vivo, DNA was extracted from induced sputum and assessed by P. aeruginosa–specific PCR. No significant reduction in P. aeruginosa levels was observed in patients receiving low-dose erythromycin compared with placebo, either as assessed by P. aeruginosa load at the study end (erythromycin: 6.25 Log10 cfu/ml ± 1.48 [SD]; placebo: 5.38 Log10 cfu/ml ± 1.32; P = 0.129, unpaired t test) (Figure 2) or load change during the study period.

As observed in the wider BLESS cohort, there was a significant reduction in total exacerbations in the erythromycin group (placebo: 6; interquartile range [IQR], 4–8; erythromycin: 3; IQR, 3–4; P = 0.008, Mann-Whitney test). In particular, there were fewer of the more moderate exacerbations requiring oral antibiotics (Figure 3).

Similarly, there was a significant reduction in sputum weight (erythromycin group, Week 48: 6.60 g/d ± 4.67 [SD]; baseline, Week 0: 18.36 g/d ± 10.38 [SD]; P = 0.007, paired t test; Figure 4) and quality of life (St. George’s Respiratory Questionnaire) (erythromycin group, Week 48: 28.75 ± 14.52 [SD]; baseline (Week 0): 33.94 ± 14.48 [SD]; P = 0.030, paired t test, Figure 3) in the erythromycin-treated group, not observed in the placebo group.

Despite erythromycin therapy having no significant impact on P. aeruginosa bacterial load in sputum in vivo, it did change P. aeruginosa gene expression. Expression of quorum sensing regulatory genes, lasR, lasI, rhlR, rhlI, and pqsA, was assessed in induced sputum samples. A significant reduction was observed in the expression of both lasR (erythromycin: fold change, 0.07 [IQR, 0.01–0.85], n = 11; placebo: fold change 1.00 [IQR, 0.05–3.05]; P = 0.047, Mann-Whitney U test) (Figure 5) and pqsA (erythromycin: fold change 0.07 [IQR, 0.02–0.25]; placebo: fold change 1.000 [IQR, 0.21–4.31], P = 0.017, Mann-Whitney U test) (Figure 5) in patients who had received erythromycin, compared with control subjects. LasI and rhlR did not differ significantly but lasI showed a trend to a reduction with erythromycin (Figure 5). This effect on P. aeruginosa gene expression appeared to be specific for the central quorum sensing genes tested, an inference supported by testing other, non–quorum sensing–controlled genes.

An expected increase in the expression of mexX (the P. aeruginosa erythromycin efflux pump gene) was observed, although this did not reach statistical significance (erythromycin: fold change 1.55 [IQR, 0.13–4.00], n = 11; placebo: fold change, 1.000 [IQR, 0.10–2.04], n = 15; P = 0.604, Mann-Whitney U test). In contrast, mexB, a gene encoding an efflux pump that is not under quorum sensing regulation, showed no difference in expression level (erythromycin: fold change 0.661 [IQR, 0.36–1.64], n = 11; placebo: fold change, 1.000 [IQR, 0.38–3.60], n = 15; P = 0.602, Mann-Whitney U test).

Despite a significant effect of erythromycin on P. aeruginosa quorum sensing, no significant effect of erythromycin on inflammation was observed in this subgroup. Sputum levels of IL-8, IL-1β, and neutrophils did not differ significantly between the placebo and erythromycin arms of the trial (not shown). There was no significant correlation between gene expression levels or inflammatory markers with specific clinical outcomes in this subgroup.

The aim of this study was to investigate the inhibition of P. aeruginosa quorum sensing by macrolide directly within the airways of patients with non–cystic fibrosis bronchiectasis receiving long-term, low-dose erythromycin. We have demonstrated that central P. aeruginosa quorum sensing genes are inhibited in the presence of erythromycin, despite there being no reduction in total P. aeruginosa bacterial load as measured by qPCR amplification of bacterial DNA.

Previous investigations have relied on P. aeruginosa isolates taken from clinical samples, cultured and assessed in vitro (16).

The use of such approaches raises questions about how representative individual isolates are of the wider pseudomonal populations in the airways (17) as well as the extent to which the P. aeruginosa response is altered in the artificial conditions of the culture environment (18, 19). Here, for the first time, we have addressed these issues through the analysis of gene expression levels on the basis of mRNA transcripts extracted directly from induced sputum from patients administered low-dose macrolide for an extended period under trial conditions.

Importantly, we were able to show a similar clinical benefit in this subgroup of patients with respect to exacerbations and sputum weight, despite there being no detectable difference in P. aeruginosa load between the treatment and placebo groups. We were able to demonstrate that, although erythromycin was administered in doses previously shown to be subinhibitory to P. aeruginosa (20), these levels are relevant to P. aeruginosa physiology, as demonstrated by the significant down-regulation of the key quorum sensing genes lasR and pqsA, whereas an efflux pump gene not controlled by quorum sensing, mexB, remained unchanged. These findings support a differential and sustained effect of erythromycin on P. aeruginosa gene expression in vivo, with a significant inhibitory effect on genes controlled by quorum sensing.

It is important to reiterate that this effect on gene expression was independent of an effect on P. aeruginosa bacterial density in the airways. We observed no significant correlation between these changes in gene expression and specific clinical outcomes; however, we did report a significant reduction in exacerbations, sputum weight, and QoL in the erythromycin group in association with the observed reduction in quorum sensing gene expression.

Analysis of this subset of patients demonstrated a lack of a significant difference in airway inflammation in those patients receiving erythromycin compared with control subjects, despite a reduction in exacerbation frequency (in keeping with the reported reduction in exacerbation rate reported in the wider BLESS study). This observation suggests that a specific antiinflammatory effect is unlikely to contribute substantially to clinical benefit in this study and is in keeping with the absence of a detectable decrease in P. aeruginosa content in the airway secretions, a principal proinflammatory stimulus. A more subtle effect cannot be excluded, however.

Limitations

This study had a number of limitations. Samples were obtained from the BLESS cohort (1), which was powered to detect a difference in exacerbation rate, not P. aeruginosa gene expression. Furthermore, the study cohort necessarily includes only the subgroup of the BLESS cohort positive for P. aeruginosa at the end point of the study. It must be noted, however, that the BLESS randomization was stratified for P. aeruginosa, and there were no demographic or baseline clinical differences between the subgroups analyzed here. Furthermore, the primary clinical outcomes were the same in this cohort as in BLESS. We acknowledge that alterations in gene expression do not always lead to a relevant change in protein production.

However, a number of in vitro studies (5, 21), including our own (online supplement), have demonstrated a strong correlation between down-regulation of gene expression and an associated reduction in pathogenicity traits. Our analysis was focused on the expression of a discrete group of genes, on the basis of their recognized role in quorum sensing regulation. Global gene expression analysis may provide further insight into the complexities of the P. aeruginosa quorum sensing system. Finally, the analysis of a wider group of inflammatory markers may identify antiinflammatory effects of erythromycin, which were not assessed in this study.

Conclusions

This study demonstrates the inhibition by macrolide of P. aeruginosa quorum sensing within the airways of patients with non–cystic fibrosis bronchiectasis receiving long-term, low-dose erythromycin. These findings hint at a possible mechanism of benefit of macrolide in P. aeruginosa–colonized inflammatory airway diseases, beyond classical antimicrobial or antiinflammatory pathways. Although we were not able to show a significant correlation between clinical outcomes and gene-expression measures, these data lend weight to the hypothesis that inhibition of bacterial virulence and behavior may underpin some of the efficacy of macrolide in the treatment of these diseases and warrants further assessment in a larger randomized controlled trial. What is clear is that the mechanism of benefit of macrolide in inflammatory airway diseases is unlikely to be explained by a single pathway and most likely represents a sum of beneficial effects, one of which is inhibition of bacterial virulence in P. aeruginosa–colonized subjects.

The authors dedicate this manuscript to the memory of A/Prof. David Serisier, who tragically died in May 2015, and whose loss will be deeply felt by the respiratory field. The authors thank Prof. Miguel Cámara, University of Nottingham, for generous technical support and advice. The authors also thank Dr. Rohan Lourie and A/Prof. Rachel Thomson for their critical appraisal of this manuscript and Ms. Megan Martin and Dr. Martina Proctor, who assisted in the application of laboratory procedures and sample collection.

1 . Serisier DJ, Martin ML, McGuckin MA, Lourie R, Chen AC, Brain B, Biga S, Schlebusch S, Dash P, Bowler SD. Effect of long-term, low-dose erythromycin on pulmonary exacerbations among patients with non-cystic fibrosis bronchiectasis: the BLESS randomized controlled trial. JAMA 2013;309:12601267.
2 . Wong C, Jayaram L, Karalus N, Eaton T, Tong C, Hockey H, Milne D, Fergusson W, Tuffery C, Sexton P, et al. Azithromycin for prevention of exacerbations in non-cystic fibrosis bronchiectasis (EMBRACE): a randomised, double-blind, placebo-controlled trial. Lancet 2012;380:660667.
3 . Hiwatashi Y, Maeda M, Fukushima H, Onda K, Tanaka S, Utsumi H, Hirano T. Azithromycin suppresses proliferation, interleukin production and mitogen-activated protein kinases in human peripheral-blood mononuclear cells stimulated with bacterial superantigen. J Pharm Pharmacol 2011;63:13201326.
4 . Rogers GB, Bruce KD, Martin ML, Burr LD, Serisier DJ. The effect of long-term macrolide treatment on respiratory microbiota composition in non-cystic fibrosis bronchiectasis: an analysis from the randomised, double-blind, placebo-controlled BLESS trial. Lancet Respir Med 2014;2:988996.
5 . Tateda K, Comte R, Pechere JC, Köhler T, Yamaguchi K, Van Delden C. Azithromycin inhibits quorum sensing in Pseudomonas aeruginosa. Antimicrob Agents Chemother 2001;45:19301933.
6 . Wagner T, Soong G, Sokol S, Saiman L, Prince A. Effects of azithromycin on clinical isolates of Pseudomonas aeruginosa from cystic fibrosis patients. Chest 2005;128:912919.
7 . Wagner VE, Bushnell D, Passador L, Brooks AI, Iglewski BH. Microarray analysis of Pseudomonas aeruginosa quorum-sensing regulons: effects of growth phase and environment. J Bacteriol 2003;185:20802095.
8 . Nalca Y, Jänsch L, Bredenbruch F, Geffers R, Buer J, Häussler S. Quorum-sensing antagonistic activities of azithromycin in Pseudomonas aeruginosa PAO1: a global approach. Antimicrob Agents Chemother 2006;50:16801688.
9 . Williams P, Camara M, Hardman A, Swift S, Milton D, Hope VJ, Winzer K, Middleton B, Pritchard DI, Bycroft BW. Quorum sensing and the population-dependent control of virulence. Philos Trans R Soc Lond B Biol Sci 2000;355:667680.
10 . Hunter RC, Klepac-Ceraj V, Lorenzi MM, Grotzinger H, Martin TR, Newman DK. Phenazine content in the cystic fibrosis respiratory tract negatively correlates with lung function and microbial complexity. Am J Respir Cell Mol Biol 2012;47:738745.
11 . Kai T, Tateda K, Kimura S, Ishii Y, Ito H, Yoshida H, Kimura T, Yamaguchi K. A low concentration of azithromycin inhibits the mRNA expression of N-acyl homoserine lactone synthesis enzymes, upstream of lasI or rhlI, in Pseudomonas aeruginosa. Pulm Pharmacol Ther 2009;22:483486.
12 . Hentzer M, Wu H, Andersen JB, Riedel K, Rasmussen TB, Bagge N, Kumar N, Schembri MA, Song Z, Kristoffersen P, et al. Attenuation of Pseudomonas aeruginosa virulence by quorum sensing inhibitors. EMBO J 2003;22:38033815.
13 . Heurlier K, Dénervaud V, Haas D. Impact of quorum sensing on fitness of Pseudomonas aeruginosa. Int J Med Microbiol 2006;296:93102.
14 . Paggiaro PL, Chanez P, Holz O, Ind PW, Djukanović R, Maestrelli P, Sterk PJ. Sputum induction. Eur Respir J Suppl 2002;37:3s8s.
15 . Rogers GB, Cuthbertson L, Hoffman LR, Wing PA, Pope C, Hooftman DA, Lilley AK, Oliver A, Carroll MP, Bruce KD, et al. Reducing bias in bacterial community analysis of lower respiratory infections. ISME J 2013;7:697706.
16 . Köhler T, Guanella R, Carlet J, van Delden C. Quorum sensing-dependent virulence during Pseudomonas aeruginosa colonisation and pneumonia in mechanically ventilated patients. Thorax 2010;65:703710.
17 . Wilder CN, Allada G, Schuster M. Instantaneous within-patient diversity of Pseudomonas aeruginosa quorum-sensing populations from cystic fibrosis lung infections. Infect Immun 2009;77:56315639.
18 . Diggle SP, Griffin AS, Campbell GS, West SA. Cooperation and conflict in quorum-sensing bacterial populations. Nature 2007;450:411414.
19 . Wilder CN, Diggle SP, Schuster M. Cooperation and cheating in Pseudomonas aeruginosa: the roles of the las, rhl and pqs quorum-sensing systems. ISME J 2011;5:13321343.
20 . Krasniqi S, Matzneller P, Kinzig M, Sörgel F, Hüttner S, Lackner E, Müller M, Zeitlinger M. Blood, tissue, and intracellular concentrations of erythromycin and its metabolite anhydroerythromycin during and after therapy. Antimicrob Agents Chemother 2012;56:10591064.
21 . Imamura Y, Yanagihara K, Mizuta Y, Seki M, Ohno H, Higashiyama Y, Miyazaki Y, Tsukamoto K, Hirakata Y, Tomono K, et al. Azithromycin inhibits MUC5AC production induced by the Pseudomonas aeruginosa autoinducer N-(3-Oxododecanoyl) homoserine lactone in NCI-H292 Cells. Antimicrob Agents Chemother 2004;48:34573461.
Correspondence and requests for reprints should be addressed to Lucy Burr, M.B.B.S., B.Sc., Mater Research Institute, Level 3 Aubigny Place, Raymond Terrace, South Brisbane, QLD 4101, Australia. E-mail:

Supported by the Mater Respiratory Research Fund, Mater Foundation Betty McGrath Clinical Research Fellowship

Author Contributions: L.D.B., G.B.R., S.D.B., and M.A.M. had principal intellectual input into the design and writing of this study. L.D.B. performed all RNA extractions and PCR with the help of S.L.T. A.C.-H.C. assessed inflammatory cytokines. B.R.H., G.F.P., S.L.T., D.V., and S.B. assisted in processing and assessing samples. All authors have reviewed and approved the final version of this manuscript.

This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org

Author disclosures are available with the text of this article at www.atsjournals.org.

Related

No related items
Comments Post a Comment




New User Registration

Not Yet Registered?
Benefits of Registration Include:
 •  A Unique User Profile that will allow you to manage your current subscriptions (including online access)
 •  The ability to create favorites lists down to the article level
 •  The ability to customize email alerts to receive specific notifications about the topics you care most about and special offers
Annals of the American Thoracic Society
13
10

Click to see any corrections or updates and to confirm this is the authentic version of record