Overview of sputum and oropharyngeal microbiome
A total of 249 sputum samples and 235 oropharyngeal swab samples were collected. For group comparisons, 58 vs 40, 58 vs 151, and 40 vs 151 sputum samples, and 74 vs 32, 74 vs 129, and 32 vs 129 oropharyngeal swabs were included for comparing heathy controls vs COPD, healthy control vs AECOPD, and COPD vs AECOPD, respectively. For within-household pairwise comparisons, paired groups included numbers of samples ranging from 12 to 48 per group (Table 1). Enrolled individuals were long-term residents in Xishuangbanna, Dali, Kunming, and Lijiang (see Supplementary Fig. 1 for sample distribution and Supplementary Table 1 for individual demographic and clinical characteristics). Medications for COPD maintenance and exacerbations were prescribed as per routine clinical practice.
Using 16S rRNA amplicon sequencing, 1137 and 970 OTUs were identified across all sputum and oropharyngeal swab samples. Prevotellaceae, Steptococcaceae, Neisseriaceae, Veilonellaceae, Fusobacteriaceae, Lachnospiraceae, Leptotrichiaceae, and Pasteurellaceae were the predominant bacterial families in both sputum and oropharyngeal microbiomes, whereas Porphyromonadaceae and Micrococcaceae were also common in sputum and oropharyngeal microbiomes, respectively (Fig. 1).
Alteration of sputum and oropharyngeal microbiome by COPD and resident regions
Although sputum microbiome vs oropharyngeal swab microbiome were not compared within individuals, between-group comparison showed that sputum samples had slightly higher alpha-diversity than oropharyngeal swabs (Fig. 2a, c). The alpha diversity of oropharyngeal swab microbiome was lowest in AECOPD (AECOPD vs COPD, p < 0.0001; AECOPD vs healthy control, p < 0.01; COPD vs healthy control, p < 0.01), and a similar trend was observed in sputum microbiome (healthy control vs COPD, p < 0.05; AECOPD vs healthy control, p < 0.001). Compared with healthy controls, alpha diversity was increased in oropharyngeal swab microbiome of COPD patients but decreased in their sputum microbiome. Differences in alpha-diversity of microbiome were observed across different geographical locations, with individuals from Lijiang consistently had the least diverse microbiome in both sputum and oropharyngeal swabs (Fig. 2b, d).
The distinction in overall microbiome composition was evaluated by principal coordinate analysis (PCoA) based on weighted UniFrac distance metrices (Fig. 3). The first and second coordinates captured about 50.3% and 48.3% of variations in the composition of oropharyngeal microbiome and sputum microbiome, respectively. However, both oropharyngeal microbiome and sputum microbiome could not be discriminated by disease group or region. Nevertheless, PERMANOVA showed that both COPD disease status and resident region had an impact on microbiome compositions, suggesting the existence of differences in relative abundances of certain microbial taxa (Supplementary Table 3).
Respiratory tract microbiome taxa associated with COPD
In order to identify which bacterial clades in the oropharyngeal and sputum microbiome were different in relative abundances between healthy vs COPD stable state vs exacerbations, we performed negative binomial Wald test using DESeq2. A total of 35 OTUs were significantly different in relative abundances with a fold change ≥ .1.5 and an adjusted p-value ≤ 0.05 in either group comparisons or within-household pair-wise comparisons (Table 2, Supplementary Table 4). We considered differences revealed by pair-wise comparisons more likely to reflect disease effects on microbiome. Compared with healthy controls, oropharyngeal microbiome under COPD stable state had relative higher relative abundances of Streptococcus, Actinomyces, Actinobacteria spp., Rothia, and Veillonella. During COPD exacerbations compared with healthy controls, oropharyngeal microbiome consisted of higher relative abundances of Raoultella, Actinobacillus, Enterobacteriaceae spp., and Haemophilus, and sputum microbiome comprised higher relative abundances of Anaerolineaceae spp., Desulfobacterota spp., Anaerolineae spp., Pedosphaeraceae spp., Rhodobacteraceae spp., gammaproteobacterial spp., and alphaproteobacteria spp.. No taxa showed significant differences in relative abundances between COPD stable state and exacerbations in pair-wise comparisons using above mentioned cutoffs.
Genus level core microbial community structure profile
The core microbial community structure of association networks was generated from the oropharyngeal and sputum microbiome dataset in Healthy, COPD and AECOPD groups, respectively (Fig. 4 and Fig. 5, to better annotated the relationship between OTU-OTU, only OTUs with degree > 1 were shown in the figure.).
In heathy controls, oropharyngeal microbiome contained one large network made up by 40 OTUs. However, under COPD stable state, members of oropharyngeal microbiome became less correlated, showing no large networks with more than 7 OUTs. However, at exacerbations, the complexity of oropharyngeal microbiome networks increased compared to that at the stable state, where the largest two networks contained 16 and 11 OUTs, respectively.
A similar pattern was found with sputum microbiome. In healthy controls, the two biggest networks contained 14 OTUs and 7 OTUs, respectively. However, under COPD stable state, the biggest network was reduced in size containing 10 OTUs, while the others contained less than five OUTs. At exacerbations, microbiome members inter-correlations increased compared with the stable state, with the largest network containing 29 OTUs (49.2%) and the second and third largest ones containing 11 and 10 OTUs.
Under the same physical status in upper/lower respiratory tract, the composition of the core microbial community was similar (Fig. 6, A, B and C). Among the core microbe 61 OTUs with degree > 1 in the Healthy group, 27 OTUs (44.3%) were found in upper/lower respiratory tract, 23 OTUs (37.7% were only found in upper respiratory tract, and 11 OTUs (18%) were found only in lower respiratory tract (Fig. 6.A, The detailed list is in Supplementary Table 5). In Fig. 6.B (Supplementary Table 6), showed 18 common OTUs (56.3%) were detected in upper/lower respiratory tract, and 7 OTUs (21.9%) were only found in the upper respiratory tract with COPD, and also 7 OTUs (21.9%) were only found in the lower respiratory tract with COPD. Among the 72 OTUs were detected in the AECOPD groups, 35 OTUs (48.6%) were found in upper/lower respiratory tract, and 13 OTUs (18.1%) were only detected in upper respiratory tract, and 24 OTUs (33.3%) were only in lower respiratory tract (Fig. 6.C, Supplementary Table 7).
Bacterial composition in upper respiratory tract in healthy, COPD and AECOPD groups, among the core microbe 57 OTUs detected in the upper respiratory tract, 22 OTUs (38.6%) were detected in all three groups. Only 1 OTU (1.8%) were found in COPD group, and 6 OTUs (10.5%) were only in Healthy, and 6 OTUs (10.5%) were found only in AECOPD, Two OTUs (3.6%) were detected in Healthy and COPD groups, and 20 OTUs (35.1%) were detected in Healthy and AECOPD groups, and there was no common OTU only detected in COPD and AECOPD (Fig. 6.D, Supplementary Table 8). In lower respiratory tract, there were 62 core microbes in all three groups, 17 OTUs (27.4%) were identical in all three groups. Only 1 OTU (1.6%) were found in COPD group, and 2 OTUs (3.2%) were only in Healthy, and 16 OTUs (25.8%) were found in AECOPD. There was no identical OTU only detected in COPD and Healthy, 19 OTUs (30.6%) were identified both in Healthy and AECOPD groups, and 7 OTUs were found in COPD and AECOPD (Fig. 6.E, Supplementary Table 9).
Differences in network complexity suggested that bacterium-bacterium interactions within respiratory tract microbiome was changed in COPD and influenced by exacerbations.