Association Between Pathogens Detected Using Quantitative Polymerase Chain Reaction With Airway Inflammation in COPD at Stable State and Exacerbations

BACKGROUND: Relationships between airway inflammation and respiratory potentially pathogenic microorganisms (PPMs) quantified using quantitative polymerase chain reaction (qPCR) in subjects with COPD are unclear. Our aim was to evaluate mediators of airway inflammation and their association with PPMs in subjects with COPD at stable state and during exacerbations. METHODS: Sputum from 120 stable subjects with COPD was analyzed for bacteriology (colony-forming units; total 16S; and qPCR targeting Haemophilus influenzae, Moraxella catarrhalis, and Streptococcus pneumoniae), differential cell counts, and inflammatory mediators using the Meso-Scale Discovery Platform. Subjects were classified as colonized if any PPM was identified above the threshold of detection by qPCR. Symptoms were quantified using the visual analog scale. RESULTS: At stable state, 60% of subjects were qPCR positive for H influenzae, 48% for M catarrhalis, and 28% for S pneumoniae. Elevated sputum concentrations of IL-1β, IL-10, and tumor necrosis factor (TNF)-α were detected in samples qPCR positive for either H influenzae or M catarrhalis. Bacterial loads of H influenzae positively correlated with IL-1β, IL-8, IL-10, TNF-α, and symptoms; and M catarrhalis correlated with IL-10 and TNF-α. H influenzae qPCR bacterial load was an independent predictor of sputum TNF-α and IL-1β. In 55 subjects with paired exacerbation data, qPCR bacterial load fold change at exacerbation in M catarrhalis but not H influenzae correlated to changes in sputum TNF-α and IL-1β concentrations. CONCLUSIONS: At stable state, H influenzae is associated with increased airway inflammation in COPD. The relationship between bacterial load changes of specific pathogens and airway inflammation at exacerbation and recovery warrants further investigation.

COPD is characterized by irreversible airfl ow obstruction and airway infl ammation. Th e disease course is punctuated by exacerbation episodes, 1 which are oft en associated with increased airway infl ammation, 2 viruses, and bacteria. 3 Bacteria are isolated from sputum cultures in 30% to 40% of subjects with stable COPD [3][4][5][6][7] and found in approximately 50% of subjects during exacerbation episodes. 7,8 Colonization with bacteria is associated with worsened health status, 6 reduced lung function, 9 and an increase in the frequency and severity of exacerbations. 10 Additionally, patients with positive sputum cultures have an associated increased infl ammatory response detected by elevated levels of sputum neutrophils, 11 IL-8, 12-14 IL-6, 13 tumor necrosis factor (TNF)-a , 13 myeloperoxidase, 14 and leukotriene B4. 6 Both culture-based and culture-independent molecular techniques have shown that Haemophilus infl uenzae is the commonest sputum pathogen in stable COPD. 15,16 Although studies using culture-based techniques have suggested that airway infl ammation is higher in those colonized with H infl uenzae , [15][16][17] relationships between pathogens quantifi ed using molecular techniques and airway infl ammation are unclear. In this study using quantitative polymerase chain reaction (qPCR) to measure pathogen-specifi c bacterial loads, we hypothesized that detection of the respiratory potentially pathogenic microorganisms (PPMs) ( H influenzae, Moraxella catarrhalis , Streptococcus pneumoniae , and Staphylococcus aureus ) in subjects with COPD is associated with increased airway infl ammation at stable state and during exacerbations.

Subjects
Sputum samples from subjects aged Ն 40 years and with postbronchodilator FEV 1/ FVC , 0.7 enrolled within an observational COPD exacerbation cohort study were analyzed. Th e study design and inclusion and exclusion criteria have been described previously. 3 Subjects with COPD on prophylactic antibiotic therapy were excluded. Th e study was conducted in accordance with the amended Declaration of Helsinki and was approved by the Leicestershire, Northamptonshire, and Rutland ethics committee (07/H0406/157). All patients gave informed written consent.

Measurements
Baseline demographic information, smoking history, medication history, and patient-reported history of exacerbations were collected. Subjects were reviewed when clinically stable and during exacerbation episodes; stable visits took place a minimum of 8 weeks aft er exacerbation episodes. Exacerbations were defi ned according to the criteria of Anthonisen et al 18 and health-care use 19 and treated according to guidelines. 20 Exacerbation testing and sampling was only performed in subjects who were treatment naive for the episode. At all visits, spirometry and symptom assessment using the St. George's Respiratory Questionnaire (SGRQ), 21 the Chronic Respiratory Questionnaire, 22 and the visual analog scale (VAS) 23 were undertaken. Spontaneous or induced sputum sampling was collected for analysis of microbiology, diff erential cell counts, and cytokine analysis, as described later. No diff erences in infl ammatory counts between spontaneous or induced sputum samples were identifi ed in this study, in keeping with other studies, 24,25 and . 95% of subjects provided spontaneous sputum samples. CT imaging, to investigate bronchiectasis, was not performed as part of the study protocol; however, CT scans performed as part of a routine clinical investigation were interrogated in subjects who entered the study.

Sputum Assessments
Bacterial load was measured by colony-forming units (CFU) as per standard methods 26 and quantitative polymerase chain reaction (qPCR) as previously described. 3 Th e CFU is a semiquantitative analysis of total live bacterial counts, and a quantitative analysis of both live and dead bacteria was quantifi ed using qPCR, estimating both the total bacterial load based on the abundance of 16S ribosomal subunit encoding genes (total 16S). Pathogen-specifi c bacterial 16S abundance using qPCR was measured for H infl uenzae , M catarrhalis , S pneumoniae , and S aureus ( Pseudomonas aeruginosa was not measured by qPCR in this study). Quantifi cation of the total bacterial load of H infl uenzae and S aureus was performed using the SYBR Green assay (Life Technologies ). Th e TaqMan assay (Life Technologies) was used to quantify M catarrhalis and S pneumoniae (target genes and primers listed in e- Table 1 ). S aureus was infrequently detected, and therefore any results relating to this pathogen were not analyzed further. The threshold of detection for pathogen-specifi c bacterial 16S qPCR analysis and CFU counts was taken as 1 3 10 4 genome copies/mL and 1 3 10 5 colonies/mL of sputum, respectively, refl ecting previous cutoff thresholds used in this fi eld. 27,28 Subjects were categorized as pathogen-specifi c bacterial 16S qPCR positive detection if any qPCR PPM (defi ned in this study as identifi cation of H infl uenzae , M catarrhalis , S pneumoniae , and S aureus ) was identifi ed above the threshold. Subjects were classifi ed as codetection if more than one pathogen-specifi c bacterial 16S qPCR PPM was identifi ed above the threshold of identifi cation. Sputum was simultaneously processed to obtain cytospins for diff erential cell counts and cell-free supernatants as previously described. 29 Th e sputum infl ammatory mediators IL-1 b , IL-5, IL-6, IL-8, IL-10, TNF-a , TNFRI, CXCL10, CCL2, CCL3, CCL4, CCL5, CCL13, and CCL17 were measured using the Meso Scale Discovery Platform (Meso Scale Diagnostics, LLC) from sputum supernatants. 25 All mediators, except IL-10, measured using the MSD were detectable and above the limit of detection in . 75% of samples (IL-10 was above the limit of detection in . 50% of samples; limits of detection presented in e- Table 2 ).

Statistical Analysis
Statistical analysis was performed using SPSS version 20 (IBM) and PRISM version 6 (GraphPad Software, Inc). Parametric data were expressed as mean (SEM), nonparametric data as median (interquartile range), and log-normally distributed data as geometric mean (95% CI). Unpaired parametric and nonparametric groups were compared using the Student t test and Mann-Whitney test, respectively. The paired t test was used to compare matched stable and exacerbation measures of airway bacterial load and sputum infl ammatory mediators. For comparison of three or more groups, the one-way analysis of variance was used, with repeated analysis of variance for paired data. Th e x 2 test was used to compare proportions between groups. Pearson correlation coefficient was used to assess correlations between qPCR-measured airway bacterial load and sputum infl ammatory mediators. Multivariate stepwise regression analysis was performed to model the eff ects of bacterial load on proinflammatory cytokine expression, namely sputum TNF-a and IL-1 b . Variables that demonstrated signifi cance at the P , .10 level using univariate analysis were entered into the model: H infl uenzae and M catarrhalis bacterial load, sputum total cell count, percentage sputum neutrophils, and CFU. Exacerbation frequency and % FEV 1 predicted were also entered into the model for clinical relevance. Th e regression model did not show violation of multicollinearity or homoscedasticity, and the residuals observed normality. No corrections for multiple mediator measurements were performed. A P value , .05 was taken as the threshold of signifi cance for all statistical testing.

Results
Stable sputum samples with full complement of infl ammatory mediators were obtained from 120 subjects (83 men). Th e clinical characteristics are presented in Table 1 . A CT scan was available in 93 subjects (77.5%), and bronchiectasis was detected in 18 (19.4%). Th ere were no significant differences in the clinical parameters between the subjects with COPD with or without detectable pathogen-specific bacterial 16S qPCR PPM. Subjects with PPMs on qPCR had more severe airfl ow obstruction and increased CFU but not total 16S ( Table 1 ). Th ere was no correlation between total 16S qPCR and inhaled corticosteroid dose, smoking pack-years, or exacerbation frequency. Th e distribution of qPCR pathogen codetection is presented in Figure 1 .

qPCR and Infl ammation During Stable State
Subjects with pathogen-specifi c bacterial 16S qPCR PPM had higher levels of IL-1 b and TNF-a and lower  Table 3 ); this was associated with a trend to a dose-response increase in infl ammation with increasing numbers of qPCR PPMs ( Table 3 ). Th ere was no correlation of total bacterial 16S qPCR with infl ammation. Pathogen-specifi c 16S qPCR bacterial load in subjects positive for H infl uenzae strongly correlated with levels of IL-1 b , IL-8, IL-10, and TNF-a ( r 5 0.64, P , .01; r 5 0.51, P , .01; r 5 0.59, P , .01; and r 5 0.71, P , .01, respectively). Pathogen-specifi c 16S qPCR bacterial load of M catarrhalis moderately correlated with IL-10 and TNF-a levels ( r 5 0.31, P 5 .02; and r 5 0.39, P , .01) ( Fig 2 ). Multivariate regression analysis identifi ed qPCR H infl uenzae bacterial load and CFU as independent predictors of sputum TNF-a and IL-1 b ( H infl uenzae , b 5 0.38 and 0.32 for TNF-a and IL-1 b , respectively; and CFU, b 5 0.36 and 0.31, respectively) ( e- Table 4 ). In subjects with either H infl uenzae or M catarrhalis as a single pathogen (n 5 21 and n 5 13, respectively), only H infl uenzae 16S qPCR bacterial load correlated with IL-1 b , IL-10, and TNF-a ( Fig 3 ); and only H infl uenzae qPCR bacterial load correlated with VAS symptoms of cough and sputum purulence and the symptom domain of the SGRQ ( e -Fig 1 ).

qPCR and Infl ammation During an Exacerbation
Paired stable and exacerbation sputum was available in 55 subjects (men, n 5 43; mean FEV 1 % predicted, 51%) with an average time of sampling between stable and exacerbation of 49 days. All subjects were treated with oral antibiotic and corticosteroid therapy at the onset of an exacerbation. During an exacerbation, pathogenspecifi c 16S qPCR for H infl uenzae , M catarrhalis , and S pneumoniae were detected in 32 (58%), 26 (47%), and 20 (36%) subjects, respectively; the majority of these subjects had the same PPM qPCR at stable state (66% H influenzae , 69% M catarrhalis , and 55% S pneumoniae ). In subjects who were pathogenspecifi c bacterial 16S qPCR PPM positive for H infl uenzae and M catarrhalis at exacerbation, the change in cytokine concentration between stable state and exacerbation was most marked in those who were also positive rather than negative in stable state ( e - Table 5 ). The change in pathogen-specifi c 16S H infl uenzae qPCR bacterial load between stable and exacerbation visits did not correlate with change in cytokines or change in symptoms. Th e change in pathogen-specifi c 16S M catarrhalis qPCR bacterial load between stable and exacerbation was positively   correlated with a change in IL-1 b and TNF-a ( r 5 0.37, P , .01; r 5 0.31, P 5 .02, respectively) ( e- Fig 2 ) but was not related to changes in symptoms or health status.

Discussion
In this study we have shown that the majority of subjects with COPD at stable state had pathogenic bacteria detected by qPCR. Th e detection of bacteria using qPCR was associated with increased sputum IL-1 b , IL-10, and TNF-a and decreased CCL13. In stable state, the strongest relationship between bacterial load, infl ammation, and symptoms was observed with pathogen-specifi c 16S H infl uenzae qPCR, whether in codetection with other bacteria or as a lone pathogen. Furthermore, we determined that H infl uenzae qPCR bacterial load was the only pathogen that was an independent predictor of sputum TNF-a and IL-1 b levels, both infl ammatory chemokines. In our study, we determined that both total 16S and CFU were elevated in subjects with detectable pathogen-specifi c 16S but only CFU bacterial load, which suggests live bacteria growth, was independently also associated with increased infl ammation. At exacerbation, we have also shown that change of M catarrhalis , but not H infl uenzae , bacterial load at exacerbation compared with stable state correlated with change in sputum TNF-a and IL-1 b concentrations. Together, these fi ndings suggest there is a complex dynamic relationship in COPD between bacterial load of specifi c pathogens, airway infl ammation, and clinical expression of disease.
Our work represents the largest study to date using qPCR techniques to describe codetection in COPD. 3,27,30 Prevalence of codetection was not described by Garcha et al, 27 and, similar to our fi ndings, Curran et al 30 described, in a small study of 30 subjects with COPD, the presence of codetection using qPCR in 80% of subjects. Consistent with studies using culture-based 15,16 and culture-independent 27 techniques, we report here that H infl uenzae was the most commonly identifi ed organism by qPCR. Th e presence of any detectable PPM was associated with increased airway infl ammation and increased proinfl ammatory cytokines. Our data suggest that sputum TNF-a and IL-1 b are more closely related to bacterial load, particularly with H infl uenzae , which was found to be an independent predictor of sputum TNF-a and IL-1 b levels and associated with increased symptoms. Although this association does not confi rm causality, there are several biologically plausible reasons that H infl uenzae may be of most signifi cance. First, up-regulation of the MRLP3 infl ammasome occurs during nontypeable H infl uenzae -induced infl ammation leading to secretion of IL-1 b 31 ; second, the outer membrane protein P6 of nontypeable H infl uenzae has been found to induce the stimulation of TNF-a and IL-10 from human alveolar macrophages. 32,33 Furthermore, alveolar macrophages have the greatest reduction of complement-independent phagocytosis of nontypeable H infl uenzae . 34 However, it must be noted that qPCR techniques measure both live and dead bacteria. Our observation that CFU bacterial load and H infl uenzae 16S pathogen-specifi c bacterial load were closely associated with sputum levels of TNF-a and IL-1 b may suggest that live H infl uenzae is driving this infl ammation, but further studies to investigate this are warranted.
We also observed that CCL13, which is increased in eosinophilic airway disease, 35 was reduced in subjects with pathogen-specifi c 16S qPCR PPM detection, irrespective of pathogen. Although we could not show that there is increased neutrophilic infl ammation, with elevated IL8 or sputum neutrophils percent as demonstrated in previous studies, 11,36 there was a trend to increase in total cell counts and absolute sputum neutrophil counts as codetection of pathogen increased and a positive correlation of IL-8 with bacterial load, further suggesting a diff erential airway pattern of infl ammation with qPCR PPM detection. IL-10, an antiinfl ammatory cytokine, was also found to be increased in subjects with pathogen-specifi c 16S qPCR PPM and correlated with H infl uenzae bacterial load. Specifi cally, the outer membrane protein P6 of H infl uenzae has been shown to be a potent macrophage inducer of infl ammation, 33 and chronic upregulation and stimula-tion of macrophages may contribute to further to the pathogenesis of COPD infections. 37 In contrast to earlier reports, we were unable to detect signifi cant diff erences in exacerbation frequency, smoking, or inhaled corticosteroid dose between colonized and noncolonized subjects. 27 However, we have demonstrated a positive association between symptoms and H infl uenzae bacterial load. We report here that the relationship between bacterial load and symptoms is pathogen specifi c and again supports a central role for sputum H infl uenzae bacterial load in chronic persistent disease. Whether therapy specifi cally targeted at reducing H infl uenzae bacterial load during stable state to reduce infl ammatory activity in COPD is clinically benefi cial is currently unknown and warrants further study.
Previous studies have demonstrated that airway infl ammation increases during exacerbation episodes 3,12 and that the presence of PPM during exacerbations is associated with increased IL-8, IL-1 b , and TNF-a . 3,38,39 However, limited data examining relationships between specifi c pathogens and infl ammation during exacerbations exist.  To our knowledge, this is the fi rst study to explore relationships between dynamic changes in pathogen-specifi c bacterial load and infl ammation at exacerbation. We found that change in infl ammatory mediators at exacerbation compared with stable state was greatest in subjects who were pathogen-specifi c bacterial 16S qPCR PPM positive for either H infl uenzae or M catarrhalis at stable state. Th is suggests that there is activation of infl ammation in stable state and colonized subjects with COPD, which is then further exaggerated at exacerbation. At stable state, H infl uenzae is closely related to airway inflammation and clinical outcomes; however, at exacerbation, changes in infl ammation are more closely related to changes (including increases and decreases) of pathogen-specifi c bacterial 16S qPCR M catarrhalis bacterial loads. Th e role of this pathogen during exacerbations of COPD needs to be investigated further.
One potential limitation of this study was the use of spontaneous sputum to determine sputum infl amma-tory mediators and pathogen detection. Sputum collection predominately arises from the larger airways; thus, our results cannot infer causality of detectable pathogens and increased infl ammation in the pathogenicity of COPD, a predominately small airways disease. However, only samples with a squamous cell contamination , 5% were used, and infl ammatory mediators measured from induced and spontaneous samples have not been shown to be diff erent, 24 suggesting that the strong associations found between airway bacterial load and sputum infl ammatory mediator concentrations were not clinically signifi cant. During this study, we also selected cutoff s to determine the presence or absence of pathogen-specifi c 16S qPCR. Although this is an arbitrary cutoff , we acknowledge that this is in part because of a paucity of the literature, which may be missing or overdetecting pathogens, but we have used cutoff s derived from previous studies using qPCR platforms in patients with COPD. 27 We have defi ned the presence of pathogen in sputum samples as detection and not colonization in this study, as the stable measurements were performed only at one time point, and we cannot comment on whether there is transient detection of pathogens using qPCR methods or persistence; it is clear that more work is needed to interrogate the stability and repeatability of pathogen-specifi c 16S qPCR in patients with COPD. A further limitation is that the presence of P aeruginosa was not assessed using qPCR techniques in this study. Although previous culture studies have suggested airway infl ammation is higher in subjects in whom P aeruginosa is detected, 14 its prevalence in other COPD cohorts has been low. 27 Moreover, within our cohort, only fi ve of 120 patients had grown P aeruginosa on sputum cultures within the preceding 12 months, and hence its exclusion from our analysis is unlikely to have signifi cantly infl uenced our results. In our study we chose primers targeting specifi c pathogens in the respiratory tract, as they have previously been deemed clinically significant. However, these techniques cannot refl ect the total microbiologic diversity within the airway, and newer techniques that can characterize the entire microbiome need to be studied further. [40][41][42] Finally, measuring sputum mediators can be fraught with diffi culty, with variability in detection [43][44][45] ; however, we have only analyzed mediators that were known to be detectable using a sputum-processing method that we have previously validated in patients with airway disease. 25

Conclusions
In summary, using qPCR we found that in patients with COPD, sputum pathogens are frequently detected. H infl uenzae was associated with increased airway infl ammation and symptoms in a dose-response relationship with sputum TNF-a and IL-1 b . M catarrhalis was more closely related to dynamic changes observed at exacerbation. Th e mechanisms by which H infl uenzae and M catarrhalis are related to airway infl ammation in COPD warrant further investigation.

Acknowledgments
Author contributions: M. B. and C. E. B. had full access to the data and are responsible for the integrity of the data and fi nal decision to submit. B. L. B., K. H., and H. P. contributed to data collection and data interpretation; I. D. P. and M. R. B. contributed to study design and data collection and interpretation; and C. E. B. and M. B. contributed to the design of the study, volunteer recruitment, data collection, data interpretation, and data analysis; and all authors contributed to the writing of the manuscript and have approved the final version for submission.
Financial/nonfi nancial disclosures: Th e authors have reported to CHEST the following confl icts of interest: Dr Bafadhel has received travel grants from Almirall, S.A.; Boehringer Ingelheim GmbH; and GlaxoSmithKline. Drs Pavord and Brightling have received grant support and consultancy fees from AstraZeneca, Medimmune LLC, Novartis Corporation, Roche Diagnostics, and GlaxoSmithKline. Mss Patel and Haldar and Drs Barker and Barer have reported that no potential confl icts of interest exist with any companies/organizations whose products or services may be discussed in this article .

Role of sponsors :
Th is article presents independent research funded by the National Institute for Health Research (NIHR) and the Wellcome Trust. Th e views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health. Th e Wellcome Trust, NIHR, and the European Regional Development Fund had no involvement in the design of the study, data collection, analysis, data interpretation, writing of the manuscript or the decision to submit the manuscript.

Other contributions:
We thank all the volunteers who took part in the study and the following people for their assistance in subject characterization: P.