Background
Chronic obstructive pulmonary disease (COPD) is defined by airflow limitation but encompasses several lung diseases. This heterogeneity includes differences in clinical characteristics, source of inflammation, response to therapies and causes of exacerbation [
1]. As COPD progresses, exacerbations become more frequent and more severe. Exacerbation rates reflect an independent susceptibility phenotype [
2], which could be mediated by host factors [
3], environmental factors [
4], viral infections and/or the bacterial microbiome [
5,
6].
Infections are predominant causes of COPD exacerbations, with approximately half reported to be caused by bacterial infections including non typeable
Haemophilus influenzae (NTHi),
Moraxella catarrhalis,
Streptococcus pneumoniae, or
Pseudomonas aeruginosa, and the other half by viral infections, primarily human Rhinovirus (HRV), but also Influenza virus, Coronavirus and Respiratory syncytial virus (RSV) to name a few [
5,
6]. Bacteria and viruses are also frequently isolated in the airways of stable COPD patients [
6‐
8]. The advent of culture-independent testing has suggested viral persistence [
9] and colonization of the lower airways with a resident bacterial microbiota [
10], implicating a role for the microbiota in disease pathogenesis, progression and treatment outcome of lung diseases [
7].
Studies of the microbiome provide a new framework to understand host-pathogen interactions, which can also yield new markers for patient diagnosis and management. Microbiota diversity is seen as a potential biomarker in cases where a single pathogenic organism reduces community complexity such as in bacterial vaginosis [
11], or Crohn’s disease [
12]. Lung microbial dysbiosis in COPD is characterized by decreased diversity [
10‐
12], which may contribute to altered immune response to environmental insults [
13]. Dysbiosis at the time of COPD exacerbation contributes to increased disease severity [
14] and higher 1-year mortality rates [
15].
Geography could also be a potential covariate in COPD patients microbiota. The gut microbiota has been shown to be geographically variable [
13]. Previous studies in different conntries have evaluated the COPD microbiota [
16,
17], but the effect of geographical variations has not yet been evaluated in a single study in relation to disease severity. The present cohort stems from a study on the incidence of viral infections in COPD [
5]. Patients were enrolled in Europe and North America and followed up for up to 3 RSV seasons, with scheduled wellness visits and unscheduled illness visits. To further our understanding of COPD exacerbation dynamics, we retrospectively evaluated the sputum bacterial microbiota from a subset of this study. The goals were to identify differences in patients with higher rates of exacerbations, to assess geographical differences in the microbiota between Europe and the USA, and to determine the influence of viral infections on microbiota diversity and the frequency of exacerbations.
Discussion
Understanding of the presence and role of both bacterial and viral pathogens over time in the heterogeneous and dynamic COPD disease [
38] is needed for patient treatment and management. The characterization of the 16S rRNA gene microbiota and respiratory viruses from a longitudinal and international cohort of severe COPD patients described in this study provides the largest survey to date on their complex associations with geography, exacerbation frequency and other demographic and clinical history.
COPD patients are particularly susceptible to respiratory infections [
6]. HRV was identified as the most prominent agent in respiratory tract infections in this cohort of COPD patients. Viral characterization is of particular importance as few reports exist on the diversity of respiratory viral agents in COPD [
6] and viral diversity should be accounted for when designing new treatments. In particular, the newly described HRV C was detected in 23 of 11,179 samples (21 of 200 patients) in the present cohort, which was reported only once previously [
39].
We detected viruses in 15% of stable samples. Asymptomatic viral infections in COPD patients are common [
40]. However, the role of these asymptomatic infections in disease progression is unclear. Most respiratory viruses tested here showed highly significant positive odds ratio with exacerbation events, but lower significance in regards to exacerbation frequency. This may indicate that viral infections alone do not sensitize the lung to repeat exacerbations as much as the bacterial microbiota. Interestingly, similar trends have been shown in asthma where respiratory viral infections in early life resulted in microbiome changes and hypersensitivity predisposition that can lead to asthma [
41,
42].
Repeat viral detection were more frequent in patients with frequent exacerbator phenotype, but the number was small, and contradicting reports exist on the repeated detection of a single virus in COPD [
9]. Further complete viral genomic characterization will be needed to understand the nature of viral infection.
With the advent of culture-independent techniques, it appears that all microbiomes harbor potential bacterial pathogens as characterized here and elsewhere [
14], but that only a portion of them will develop exacerbation-prone phenotypes, a phenotype that appears independent of disease GOLD stage yet linked to microbiota diversity [
13]. It was previously observed that microbiota diversity in the COPD lung correlated with disease severity but not disease state [
43]. Here, microbiota diversity alone was not correlated with frequent exacerbations, but was highly correlated with certain bacterial taxa dominating the microbiota. Microbiota predominant with Escherichia, Pseudomonas or Streptococcus, showed significantly lower alpha diversity and significant positive odds ratio with the frequent exacerbation phenotype, suggesting a role of the microbiota in sensitizing the COPD lung to acute exacerbations. This study was limited in disease severity metrics with the exception of baseline evaluation and sampling of events. Longitudinal monitoring of symptoms scales would help to better understand the relation of certain bacteria to not only exacerbation frequency, but also the symptom severity and COPD progression.
The use of biotyping has been seldom used in respiratory microbiota research [
44] and not yet explored in COPD. A complex resident bacterial community could be identified in all COPD sputum samples and categorized into 2, 3 and 6 biotypes at stable, acute exacerbation and exacerbation follow-up visits, respectively. Biotype 1 at stable state was associated with higher viral infections, while biotypes 2 and 3 at acute exacerbation were associated with high exacerbation frequency. These findings are interesting because they mirror another study showing the partitioning of COPD exacerbation samples into 3 cytokine profile clusters [
3] with associations to specific ratios of Proteobacteria, Firmicutes and Bacteroidetes that highlighted the heterogeneity of exacerbation profiles in COPD patients. During exacerbation follow-up visits, biotype 6 with a high relative abundance of Pseudomonas was found over-represented in samples associated with antibiotics use. Antibiotic treatment inadequacy is the cause for secondary infection or the emergence of multi-drug resistant
P.aeruginosa [
45]
. New targeted treatments, such as monoclonal antibodies, could be useful in such settings [
46].
The principal novelty of this study cohort was the long term patient follow-up. We were able to collect several sputum samples per patient at stable state over the course of 3 years and studied the COPD microbiota longitudinally. The lung microbiome is inherently variable, shaped by a process of inhalation and elimination [
47]. The lung microbiome is also personal, with large inter-patient variability [
48]. Previously, it was shown that microbial dysbiosis from stable to exacerbated state correlated with greater exacerbation severity [
14]. Here, patients with greater microbiota variability at stable state correlated with higher exacerbation frequency. Proteobacteria such as Pseudomonas and Moraxella were more abundant in patients with more variable microbiota at stable state. Interestingly,
P.aeruginosa and
M.cattharalis are prominent causes of exacerbations [
6], but their role in stable disease is less clear [
49]. It was previously shown that chronic colonization with
P. aeruginosa occurs more frequently in more severe COPD patients [
50] and that
M.catarrhalis asymptomatic colonization was associated with a greater frequency of a sputum IgA response than exacerbation [
51]. Our results suggest that dysbiotic burden at stable state by Pseudomonas, Moraxella and others might sensitize the lung to further exacerbations and viral infections. Pseudomonas and Moraxella, like many opportunistic Proteobacteria, are pro-inflammatory [
52,
53]. Imbalanced inflammation can improve
P.aeruginosa’s fitness [
52], allow the acquisition of new
M.cattharalis strains [
53], leading to exacerbation and possibly infections from other pathogens in a coupled cycle of inflammation and dysbiosis [
47]. Microbiology clinical testing in COPD patients is most often performed at exacerbation or follow-up visits. Patients might benefit from clinical monitoring of these bacteria at stable state to assess their presence and/or growth which could lead to potential future exacerbations.
We also noted geographical differences in COPD lung microbiota. Geographical differences in gut microbiota have previously been noted [
35], but not yet in the lung. There were significant differences in alpha and beta diversity between the USA and Europe, but not within countries or sites. Microbiota diversity in the USA was lower and although frequent exacerbator phenotypes were more common than in Europe, the difference was not significant. USA patients tended to have samples with high relative abundance of Streptococcus (biotype 2) and Haemophilus (biotype 3) associated with the frequent exacerbator phenotype.
S.pneumoniae and
H.influenzae are commonly associated with exacerbations [
54], and should also be considered as potential risk factors in the frequent exacerbator phenotype. Significantly more samples collected in the USA were associated with antibiotics use, but this alone did not explain differences in diversity. This observational study included a variety of standard-of-care medications, doses and timings precluding precise treatment effect modeling on the microbiota. Clinical trials exploring current and novel treatment modalities will lead to better patient management and antibiotics stewardship as reviewed elsewhere [
55].
Predominance of Haemophilus was over-represented in acute exacerbation samples (Figs.
2 and
6), as noted in previous studies [
6,
14,
17,
43,
56]. However, interestingly, using odds ratio (Fig.
3) or ANCOM (Suppplementary Fig.
4a) over Haemophilus abundance,
H.influenzae was high but not significantly associated with acute exacerbation event, whereas
H.parainfluenzae was significantly higher at stable state. Differences in Haemophilus abundance compared to other studies might be due to the type of sputum collected, transportation media, extraction protocol or an effect of the sample size. Although
H.parainfluenzae can be the cause of respiratory infection in healthy subjects, it has not been associated with exacerbation in COPD [
56].
H.parainfluenzae could compete for niche resources leading to overgrowth of the more pathogenic
H.influenzae. Previous work has shown that
H.influenzae competes with
S.pneumoniae [
57] and that patients colonized by NTHi and acquiring HRV have more frequent and severe exacerbations [
58]. Here, Streptococcus species could not be resolved using 16S rRNA V4 region, and although speciation of Haemophilus was attempted, further validation using targeted PCR or whole genome sequencing will be necessary to ensure correct discrimination. To note, constant improvements in 16S databases can also affects taxonomic resolution. SILVA database [
59] version 132 updated in 2017 classified reads into more genera (
n = 562) compared to Greengenes version 13.8 updated in 2013 (
n = 395). However, bacterial taxa discussed in this study showed less than 1% variations in read classification between the 2 database classifiers (data not shown), and conclusions were unchanged. Speciations and typing of bacteria and viruses are critical to understanding their pathogenicity and complex relationships. Greater taxonomic resolution will be achieved using updated databases and more comprehensive techniques like shotgun metagenomics.
Acknowledgments
We would like to thank the study participants and the site investigators listed as follows: Andreea Antonescu-Turcu/USA, Wesley Bray/USA, Ana Dancheva/Bulgaria, Stefan Denchev/Bulgaria, Anthony Dimarco/USA, Mark Dransfield/USA, Ali El Solh/USA, Ann Falsey/USA, Jonathan Ilowite/USA, Ivan Kiselov/Bulgaria, Petr Kolman/Czech Republic, Daniela Kopecka/Czech Republic, Radovan Kozel/Czech Republic, Peter Krumpe/USA, Camil Kreit/USA, DavidLaman/USA, Arturo Meade/USA, Frank Sciurba/USA, Milkana Simeonova/Bulgaria, Amir Sharafkhaneh/USA, Sotir Sotirov/Bulgaria, John Southard/USA, Anelia Stoyanova/Bulgaria, Patrick Sturm/USA, Keipp Talbot/USA, Josef Veverka/Czech Republic, and Jiri Vytiska/Czech Republic. We would also like to thank Alison Humbles, Paul Warrener, Taylor Cohen, Omari Jones, Fiona Fernandes and Outi Vaarala for helpful discussions.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.