|Home | About | Journals | Submit | Contact Us | Français|
Acute exacerbations of chronic obstructive pulmonary disease (COPD) are a major source of morbidity and contribute significantly to healthcare costs. Although bacterial infections are implicated in nearly 50% of exacerbations, only a handful of pathogens have been consistently identified in COPD airways, primarily by culture-based methods, and the bacterial microbiota in acute exacerbations remains largely uncharacterized. The aim of this study was to comprehensively profile airway bacterial communities using a culture-independent microarray, the 16S rRNA PhyloChip, of a cohort of COPD patients requiring ventilatory support and antibiotic therapy for exacerbation-related respiratory failure. PhyloChip analysis revealed the presence of over 1,200 bacterial taxa representing 140 distinct families, many previously undetected in airway diseases; bacterial community composition was strongly influenced by the duration of intubation. A core community of 75 taxa was detected in all patients, many of which are known pathogens. Bacterial community diversity in COPD airways is substantially greater than previously recognized and includes a number of potential pathogens detected in the setting of antibiotic exposure. Comprehensive assessment of the COPD airway microbiota using high-throughput, culture-independent methods may prove key to understanding the relationships between airway bacterial colonization, acute exacerbation, and clinical outcomes in this and other chronic inflammatory airway diseases.
Chronic obstructive pulmonary disease (COPD) affects more than 12 million individuals in the United States and is the fourth leading cause of chronic morbidity and mortality (Rabe et al., 2007). A significant proportion of COPD-related healthcare costs are attributable to hospitalization for respiratory exacerbations (Mannino and Braman, 2007), with severe exacerbations having been associated with high mortality rates (Connors et al., 1996; Nseir et al., 2006). Bacterial infections are implicated in approximately 50% of COPD exacerbations (Sethi and Murphy, 2008). However, in most studies to date, bacterial identification has primarily relied on culture-based methods (Papi et al., 2006; Rosell et al., 2005; Soler et al., 2007) or species-specific, targeted PCR approaches (Murphy et al., 2004). Therefore, the true depth of bacterial diversity present in COPD airways is unknown, and the potential role for mixed-species bacterial communities in the pathogenesis of chronic airway colonization and acute exacerbations have been largely overlooked.
It is increasingly recognized that the human host is colonized by diverse, site-specific microbial communities that constitute the human microbiome (Dethlefsen et al., 2007; Eckburg et al., 2005). Growing interest in characterizing these consortia and their interactions with the host, is exemplified by the formation in 2008 of the International Human Microbiome Consortium (IHMC), a collaborative effort to merge data generated through the U.S. NIH Human Microbiome Project (HMP; http://nihroadmap.nih.gov/hmp) and the E.U. Metagenomics of the Human Intestinal Tract (MetaHIT; www.metahit.eu) initiatives. Constant exposure of the respiratory tract with the external environment could conceivably lead to important microbe–microbe and microbe–host interactions, yet the human airway microbiota, particularly that present in the context of pulmonary disease, remains largely uncharacterized. Given the limitations of culture-based methods, culture-independent approaches that identify species, or groups of closely related species, based on sequence polymorphisms in conserved genes, such as the 16S ribosomal RNA (16S rRNA) gene, enable a more comprehensive assessment of microbial members present in a mixed-species population. Such methods have increasingly been applied to identify bacterial consortia in a variety of human niches with demonstration that community composition is related to disease states such as obesity, ventilator-associated pneumonia and cystic fibrosis airway disease (Flanagan et al., 2007; Harris et al., 2007; Rogers et al., 2004; Turnbaugh et al., 2006).
An alternative approach to traditional culture-independent methods of microbial community analysis is the 16S rRNA PhyloChip, a high-density microarray containing 500,000 probes that can detect approximately 8,500 bacterial taxa [taxa are defined as a group of bacteria sharing at least 97% sequence homology within the 16S rRNA gene sequence; (Brodie et al., 2006)] The PhyloChip has previously been shown to detect substantially greater bacterial diversity compared in parallel with traditional clone library sequencing approaches (Brodie et al., 2006; Desantis et al., 2007; Flanagan et al., 2007). Using this microarray, we analyzed airway specimens from eight COPD patients who were being managed for severe respiratory exacerbations to determine if a more diverse bacterial community is present during pulmonary exacerbation in the setting of antibiotic administration.
Potential subjects for this study were screened from a database of airway specimens collected between August 2004 and April 2006 from mechanically ventilated patients admitted to the intensive care units at Moffitt-Long Hospital (University of California, San Francisco), who were enrolled in a parent study of Pseudomonas aeruginosa in intubated patients (Flanagan et al., 2007). Subjects admitted to the ICU with a primary diagnosis of “COPD exacerbation” were identified for inclusion in our study. Available endotracheal aspirates (ETAs) from eight patients were processed for 16S rRNA PhyloChip analysis, as detailed below. To compare results from PhyloChip analysis with conventional clinical cultures, we obtained results of quantitative clinical laboratory bacterial cultures (blood agar, chocolate agar, and EMB media) performed on minibronchoalveolar lavage (m-BAL) airway samples, collected within 1–5 days of the ETA specimen analyzed by PhyloChip, as previously described (Flanagan et al., 2007). We have previously found m-BALs to possess a similar bacterial community composition to that of ETAs obtained concurrently from the same patient (Flanagan et al., 2007). Clinical data (Table 1) were recorded in a secure database, including whether a diagnosis of pneumonia by conventional clinical and radiologic criteria (Zhuo et al., 2008) was made during the patient's hospitalization and the time frame between diagnosis and collection of airway samples. The Committee on Human Research at UCSF approved all study protocols, and all patients or their surrogates provided written, informed consent.
Total DNA was extracted from ETAs (200μL) using a bead-beating step (5.5m s−1 for 30s, FastPrep system) (MP Biomedicals, Cleveland, OH) prior to nucleic acid extraction using the Wizard Genomic DNA Purification kit (Promega, Madison, WI). Twelve, 25-cycle PCR reactions, containing 100ng of DNA, 2.5mM each of dNTPs, 1.5μM each primer (Bact-27F and Bact-1492R) (Lane, 1991) and 0.02U/μL of ExTaq (Takara Bio, Japan), were performed for each sample across a gradient of annealing temperatures (48–58°C), to maximize the diversity recovered. The resulting PCR products were pooled and gel-purified using the MinElute Gel Extraction kit (Qiagen, Chatsworth, CA). Known concentrations of synthetic 16S rRNA gene fragments and non-16S rRNA gene fragments were spiked into the pooled, purified PCR product, which served as internal standards for normalization. A total of 250ng of purified PCR product per sample was fragmented, biotin-labeled, and hybridized to the microarray as previously described (Brodie et al., 2006). Washing, staining, and scanning of arrays were conducted according to standard Affymetrix protocols. Background subtraction, noise calculations and scaling were carried out as described previously (Brodie et al., 2006; Desantis et al., 2007).
Relatively conservative detection and quantification criteria for each taxon were applied, as previously described (Desantis et al., 2007). Briefly, probe-pairs consisting of a perfectly matched and mismatched cross hybridization control probe (containing a mismatch at the 13th nucleotide) were scored as positive if they met two criteria: (1) the fluorescence intensity of the perfectly matched probe was ≥1.3 times greater than that of the mismatched probe, and (2) the difference in intensity in each probe pair was 130 times greater than the squared noise value for that array. The positive fraction (pf) of probe sets (minimum of 11, median of 24 probe-pairs per taxon) was calculated, and a taxon was considered “present” if the calculated pf was ≥90%. Statistical analyses were performed in the R environment (www.R-project.org), using the ecological community analysis package vegan (version 1.16-1). Log-transformed fluorescence intensities were used to calculate Bray-Curtis dissimilarity measures of ecological distance. Nonmetric multidimensional scaling (NMDS), a nonparametric ordination method that maps community relatedness, in this case using the Bray-Curtis distance metric, was used to assess variability in bacterial community structure. The function adonis (Anderson, 2001), which conducts a matrix-based nonparametric analysis of variance, was applied to explore relationships between community composition and clinical variables, including age, gender, number of intubation days, presence of pneumonia, time frame between pneumonia diagnosis and sample collection, antibiotic and corticosteroid treatments, and survival to ICU discharge. Between-group differences in taxon abundance were assessed by two-tailed t-testing with significance adjusted for false discovery using q-values, as previously described (Storey and Tibshirani, 2003). Taxa exhibiting q values<0.05, a p-value ≤0.02 and a change of >1,000 fluorescence units (log-fold change in 16S rRNA copy number) were considered statistically and biologically significant. Phylogenetic trees were constructed using representative 16S rRNA sequences from the Greengenes database (Desantis et al., 2006). A neighbor-joining tree with nearest-neighbor interchange was produced using FastTree (Price et al., 2009) and uploaded to the Interactive Tree of Life project (http://itol.embl.de/) for annotation (Letunic and Bork, 2007).
To confirm that changes in array fluorescence intensities were reflective of changes in target organism abundance, triplicate, Q-PCR reactions were performed for selected taxa containing species of interest, using a Stratagene MxP3000 real-time system and the QuantiTect SYBR Green PCR kit (Qiagen). Primers for taxa containing selected species of interest were designed based on PhyloChip probes for the target taxon (Table 2). Reaction conditions were similar to those previously described (Brodie et al., 2007) with the exception that 10ng of DNA extract were used in 40 cycles of reaction using the annealing temperatures listed in Table 2 for each primer set. Regression analyses of inverse cycle threshold values plotted against PhyloChip fluorescence intensities were determined for each targeted taxon.
16S rRNA PhyloChip analysis identified a total of 1,213 bacterial taxa present in airway samples from COPD patients obtained during the course of acute exacerbations (complete list is provided in Table 3; see at end of article). Despite recent or ongoing exposure to antibiotics across the group, the mean number of taxa detected in each sample was 411±246 taxa (SD). Identified taxa represented a diverse group of species belonging to 38 bacterial phyla and 140 distinct families (Fig. 1A). Bacterial families detected included members of the Pseudomonadaceae, Pasteurellaceae, Helicobacteraceae, Enterobacteriaceae, Comamonadaceae, Burkholderiaceae, and Alteromonadaceae, among many others. In addition, recently described phyla such as the TM7 subgroup of Gram-positive uncultivable bacteria were also detected in the airways of these patients (Table 3).
Interpersonal variation in bacterial richness (number of taxa detected) was noted across the patient samples (Fig. 1B). Four subjects (patients 1, 4, 5, and 6) exhibited communities with significantly fewer taxa (p<0.002) compared with the other four subjects. Patients in which fewer bacterial taxa were detected tended to possess more members of the Pseudomonadaceae. In contrast, members of the Clostridiaceae, Lachnospiraceae, Bacillaceae, and Peptostreptococcaceae were detected more commonly in those patients with richer communities (patients 2, 3, 7). Patient 8 was observed to possess a large proportion of taxa belonging to the Enterobacteriaceae family, which have been associated with more advanced COPD lung disease (Sethi and Murphy, 2008). Interestingly, this patient had radiographic evidence of coexisting bronchiectasis, which was not present in the other patients.
Given the variation in bacterial richness among samples, which suggested differences in bacterial community composition, NMDS was used to assess variation in bacterial community structure (based on Bray-Curtis dissimilarity measures) across the sample cohort. This revealed two distinct groups of patient samples and confirmed that patient 8 represented a structurally distinct airway community (Fig. 2). Although the cohort was small, given this separation of subjects based on differing bacterial community structures, the influence of available clinical parameters on community composition was explored. Matrix-based, nonparametric multivariate analysis of variance (Anderson, 2001) revealed that across the cohort, the number of elapsed intubation days was significantly associated with bacterial community composition and structure, accounting for the greatest percentage of the observed variability (44%, p<0.03; Fig. 2). Group 1 patient samples (patients 2, 3, and 7) were characterized by a shorter intubation duration prior to ETA sample collection (≤6 days), while those in Group 2 were intubated for significantly longer periods of time (p<0.0007; patients 1, 4, 5, and 6; ≥16 days) and exhibited a significantly less rich community composition compared to that of Group 1 (p<0.025). Given the community variation between Group 1 and Group 2, differences in the relative abundance of all detected taxa were assessed between the groups, which identified 153 taxa with significantly different relative abundances (Table 4; see at end of article). All of these significant taxa were present in higher abundance in Group 1, the majority of which (77%) belonged to the phylum Firmicutes. These included species such as Lactobacillus kitasatonis, L. perolens, L. sakei, and Bacillus clausii, as well as known pathogenic species such as Streptococcus constellatus, which is a member of the Streptoccocus milleri group (SMG; Table 4). No other clinical variable [including diagnosis of pneumonia (n=6; p<0.4) or the number of days between pneumonia diagnosis and sample collection (range: 3–52 days; p<0.6)] demonstrated a significant association with bacterial community composition in this cohort.
A common “core” of 75 bacterial taxa representing 27 classified bacterial families was identified in all patients analyzed (Fig. 3). This core group included members of the Pseudomonadaceae, Enterobacteriaceae, Campylobacteraceae, and Helicobacteraceae, amongst others. In addition, taxa containing species of pathogenic potential, such as Arcobacter cryaerophilus, Brevundimonas diminuta, Leptospira interrogans, as well as P. aeruginosa, were detected in all patients (a complete list of the core taxa is provided in Table 5; see at end of article). We also analyzed the array data for organisms that have previously been associated with COPD airways (Sethi and Murphy, 2008). Haemophilus influenzae was detected by the array in two subjects (patients 2 and 8), although corresponding m-BAL cultures were negative for this organism. Moraxella catarrhalis was not detected by PhyloChip or culture in any patient sample. However, other phylogenetically related members in the Moraxellaceae family, including Moraxella oblonga, Acinetobacter haemolyticus, and Psychrobacter psychrophilus were identified by the array in 80–100% of subjects (Table 3). Streptococcus pneumoniae was detected in four subjects (patients 2, 3, 7 and 8) despite all m-BALs being culture-negative for this species. Finally, we also examined PhyloChip data for the presence of the atypical bacteria, Mycoplasma pneumoniae and Chlamydophila pneumoniae, which are associated with 3–5% percent of exacerbations (Sethi and Murphy, 2008). Neither was detected by the PhyloChip, although a related species, Mycoplasma pulmonis, was identified in a single individual (patient 3).
Quantitative PCR was performed to validate that changes in reported array fluorescence intensities for targeted taxa correlated with changes in target species copy number for a selection of known airway pathogens (P. aeruginosa and Stenotrophomonas maltophilia) and two characteristic gastrointestinal organisms (Campylobacter mucosalis and Helicobacter cetorum). Regression analysis of species abundance determined by Q-PCR and array fluorescence intensity demonstrated strong concordance between the two independent methods for each target organism (Table 6), confirming their presence in these COPD airway samples and the ability of the array to accurately reflect changes in organism relative abundance.
Although nearly half of acute COPD exacerbations are associated with bacterial infection, our knowledge of the microbial species associated with these events is limited to a handful of organisms detected primarily by culture-based methods (Papi et al., 2006; Rosell et al., 2005; Soler et al., 2007). The overall aim of this study was to begin to address the overarching question whether previously undetected bacterial species exist in the airways of COPD patients during acute exacerbations. High-resolution, culture-independent analysis using the 16S rRNA PhyloChip revealed a much greater diversity of bacteria than has previously been appreciated in the airways of COPD patients being managed for severe exacerbations, including members of the Pseudomonadaceae, Enterobacteriaceae, and Helicobacteraceae, among others. The potential for a diverse airway bacterial community to play a role in chronic airway colonization and inflammation, a feature of COPD, has not been previously considered.
The identification of a diversity of bacterial communities in respiratory samples from COPD patients experiencing severe exacerbations suggests that the pathogenesis of these events could involve a polymicrobial process. Future studies in a larger cohort of patients, including nonintubated COPD patients with a greater range of exacerbation severity, will be necessary to determine relationships between community composition, structure, and pulmonary health. Only a handful of bacterial species, such as H. influenzae and P. aeruginosa, have previously been associated with COPD exacerbations. The possible role of other bacterial community members with pathogenic potential identified in this study, e.g. A. cryaerophilus, B. diminuta, and L. interrogans, may merit further investigation. Many of these species have been implicated in other pathogenic processes such as endocarditis (Han and Andrade, 2005; Marques da Silva et al., 2006; Paster et al., 2002) and bacteremia (Hsueh et al., 1997; Woo et al., 2001). L. interrogans, the causative agent of human leptospirosis (Gaudie et al., 2008), has recently been shown to induce pulmonary lesions in an experimental animal model of airway infection (Marinho et al., 2009) and pulmonary hemorrhage in severe cases (Dall'Antonia et al., 2008). Their potential for pathogenesis suggests the possibility of a role for these organisms in COPD chronic airway disease. Future studies involving functional and mechanistic analyses will be necessary to further assess this.
Multiple oropharyngeal and gut-associated bacterial species were also identified by PhyloChip analysis, suggesting a potential role in COPD exacerbations. Although contamination by oropharyngeal secretions is always a possibility, samples were collected through an endotracheal tube, diminishing the degree of direct contamination. Ongoing microaspiration during intubation, however, cannot be completely prevented, and it has been suggested that the oral cavity and gastrointestinal tract act as a microbial reservoir for seeding the airways in vulnerable patient populations (Garrouste-Orgeas et al., 1997; Heyland and Mandell, 1992; Orozco-Levi et al., 2003). In our study the relative abundance of targeted gastrointestinal-associated species, H. cetorum and C. mucosalis (Figura et al., 1993; Garcia-Amado et al., 2007), were confirmed by independent Q-PCR in airway samples from these patients. The presence of oropharyngeal or gut-associated bacteria in the lower airways may have significant implications for a disease population with greater risks from pulmonary infections. For example, Duan et al. (2003) demonstrated in a rat model that coinfection of P. aeruginosa with oropharyngeal bacterial species (isolated from cystic fibrosis patient sputa) resulted in enhanced lung damage and upregulation of P. aeruginosa virulence gene expression. In our study, P. aeruginosa and oropharyngeal and gut-associated species were present in the airways of all patients studied, suggesting the potential for enhanced pathogenesis in this patient population.
We recognize that our study numbers are small and represent a severely ill group of COPD patients. Therefore, caution must be exercised in extrapolating these findings to a broader group of COPD patients, especially nonintubated individuals experiencing less severe exacerbations. Although facilitating access to lower respiratory specimens that are otherwise challenging to obtain during severe exacerbations, the intubation status of these patients is an important consideration in weighing these findings. Although protracted intubation was associated with decreased bacterial community richness, the possibility for a more diverse bacterial community to play a role at least at the onset of exacerbation-associated respiratory failure remains. Control samples from nonintubated COPD and non-COPD patients were not available for this study, nor were longitudinal samples from these patients. In a previous study, however, we have found that endotracheal samples obtained from patients briefly intubated for elective surgery produced no detectable 16S rRNA PCR product (Flanagan et al., 2007). The lack of an association between duration of active antimicrobial therapy and bacterial community structure is most likely due to both the small study numbers as well as the administration of previous antibiotic courses up to one month prior to sample collection in the study (e.g., patient 6).
Results of this study highlight the advantages of complementing culture-based methods with higher resolution approaches for bacterial detection to provide a more comprehensive assessment of the airway microbiota present in COPD patients. Culture-independent methods are particularly relevant to identify viable but nonculturable species that produce and exist in biofilms (Rayner et al., 1998), which are characteristic of chronic airway infections (Costerton et al., 1999; Singh et al., 2000) and have recently been implicated in COPD (Martinez-Solano et al., 2008). Because our PhyloChip analysis was based on DNA extracted from airway samples, it does not provide information on the viability of the species detected. However, we have previously noted that within 24h of antimicrobial administration to cystic fibrosis patients, bacterial richness decreased approximately 10-fold as detected by the PhyloChip (Lynch, unpublished data). This suggests that DNA turnover is relatively rapid in the airways of chronic pulmonary disease patients, and that the taxa detected during antimicrobial administration in our COPD patients largely represent the viable portion of the community. This is supported by the finding that the PhyloChip detected all species that were isolated by concurrent clinical laboratory culture. Independent Q-PCR analysis of selected taxa also demonstrated strong concordance between calculated 16S rRNA copy number and PhyloChip-based fluorescence intensities, validating the abundance of individual species detected by the array. Although both cultures and the microarray demonstrated low detection rates for H. influenzae and M. catarrhalis, two species commonly associated with COPD exacerbation (potentially due to antibiotic-mediated clearance), only the PhyloChip identified other Haemophilus species across the majority of patients in this study, as well as other members of the Moraxellaceae family. Although these molecular methods may identify potentially relevant species undetected by culture, they provide no information on the viability or activity of these species. Hence, the significance of detecting species phylogenetically related to known pathogens in COPD airways is unclear, but may merit further study.
Application of the 16S rRNA PhyloChip to airway secretions from COPD patients during severe respiratory exacerbations requiring ventilatory support and antibiotic administration, demonstrated the presence of diverse bacterial communities, whose structural variation in this cohort was related to the duration of intubation. A core community of bacterial taxa comprised of many known pathogens, some not previously associated with COPD, was common to all patients studied. Given that the disease model of COPD is generally characterized by chronic airway colonization, recurrent infection-related exacerbations, and a persistent state of chronic airway inflammation, these results highlight the need to consider the polymicrobial community present in COPD airways and the potential functional effects of these consortia on immune response and pulmonary health. Identification of relationships that exist between bacterial communities, their collective gene expression, concomitant host response, and clinical outcomes may ultimately lead to improved understanding of the pathogenesis of COPD.
The authors thank Yvette Piceno, Todd DeSantis, and Gary Andersen at Lawrence Berkeley National Laboratory for their advice and technical support in this study, and to Homer Boushey, M.D., for his critique of the manuscript. Part of this work was performed at Lawrence Berkeley National Laboratory under the auspices of the University of California under contract number DOE DE-AC02-05CH11231. Research was supported by a Tobacco-Related Disease Research Program (Univ. of California) fellowship to Y.J.H. and an American Lung Association grant to S.V.L. S.V.L. and E.L.B. are funded also by NIH Award (AI075410).
The authors declare that no competing financial interests exist.