Enrichment of lung microbiome with supraglottic taxa is associated with increased pulmonary inflammation

Abstract

Background: The lung microbiome of healthy individuals frequently harbors oral organisms. Despite evidence that microaspiration is commonly associated with smoking-related lung diseases, the effects of lung microbiome enrichment with upper airway taxa on inflammation has not been studied. We hypothesize that the presence of oral microorganisms in the lung microbiome is associated with enhanced pulmonary inflammation. To test this, we sampled bronchoalveolar lavage (BAL) from the lower airways of 29 asymptomatic subjects (nine never-smokers, 14 former-smokers, and six current-smokers). We quantified, amplified, and sequenced 16S rRNA genes from BAL samples by qPCR and 454 sequencing. Pulmonary inflammation was assessed by exhaled nitric oxide (eNO), BAL lymphocytes, and neutrophils.

Results: BAL had lower total 16S than supraglottic samples and higher than saline background. Bacterial communities in the lower airway clustered in two distinct groups that we designated as pneumotypes. The rRNA gene concentration and microbial community of the first pneumotype was similar to that of the saline background. The second pneumotype had higher rRNA gene concentration and higher relative abundance of supraglottic-characteristic taxa (SCT), such as Veillonella and Prevotella, and we called it pneumotypeSCT. Smoking had no effect on pneumotype allocation, α, or β diversity. PneumotypeSCT was associated with higher BAL lymphocyte-count (P= 0.007), BAL neutrophil-count (P= 0.034), and eNO (P= 0.022).

Conclusion: A pneumotype with high relative abundance of supraglottic-characteristic taxa is associated with enhanced subclinical lung inflammation.

Figures

Figure 1

Comparison of bacterial loads in background, BAL, and supraglotic samples. Samples from background were obtained from either sterile saline (open square) or from sterile saline flushed through bronchoscope (closed square) and samples from BAL (open circles for never-smokers and closed circles for smokers) and supraglotic (open triangles for never-smokers and closed triangles for smokers) were obtained via bronchoscopy as described (see Additional file 1: Figure S1 for details). To detect bacterial load, universal primers for bacterial 16S rRNA were used in combination with a TaqMan Probe. Differences in bacterial loads were evaluated using Mann–Whitney U test.

Figure 2

Evaluation ofcarry-over of supraglottic microbiome compared with first and second BAL. To evaluate carry-over of supraglotic-characteristic taxa to the lower airways, we evaluated the microbiome of the first and second BAL return in 15 cases where a separate bronchoscope was used to obtain a supraglotic sample (see Additional file 1: Figure S1 for details). Paired comparison of UniFrac distances between supraglottic and first BAL samples compared with supraglottic and second BAL samples (Wilcoxon rank-sum test).

Figure 3

Comparison of supraglottic bacterial communities with first and second BAL microbiome. Relative abundances of Veillonella, Prevotella, Propionibacterium, and Staphylococcus in 15 subjects with paired supraglottic (S), first BAL (1), and second BAL (2). Veillonella and Prevotellaare the two most abundant taxa in supraglottic samples while Propionibacterium and Staphylococcus are the two most abundant taxa in background. UniFrac distance to supraglottic and bacterial load adjusted per mL of BALF (log10 16S qPCR) are shown below the bar graph of relative abundance. Overall, first BAL is not consistently closer to supraglottic than second BAL.

Figure 4

Clustering analysis of background, supraglotic, and BAL microbiota patterns. (A) Heat map of unsupervised hierarchical clustering of most abundant OTUs at a genus level (relative abundance ≥5% in any sample) in background and supraglottic samples. Background microbiome (sterile saline and saline through bronchoscope, see Additional file 1: Figure S1 for details) is enriched with Staphylococcus, Propionibacterium, and Corynebacterium, while the supraglottic microbiome is enriched with Prevotella and Streptococcus. (B) PCoA analysis based on weighted UniFrac distances clustered background samples separated from supraglotic samples. (C) Heat map of unsupervised hierarchical clustering of BAL samples. Never-smokers are indicated with green labels and smokers with blue labels. Dendrogram shows deep cleft that identified two major BAL microbiomes: one characterized by high relative abundance of Staphylococcus, Propionibacterium, and Corynebacterium which we called PneumotypeUN and a second with high relative abundance of Prevotella, Veillonella, and Streptococcus (pneumotypeSCT). (D) PCoA analysis based on weighted UniFrac distances differentiate the same BAL samples in the same two well-defined clusters (never-smokers in green dots and smokers in blue dots).

Figure 5

Classification of Streptococcus OTUs in background, supraglottic, and BAL microbiota. Twenty-one Streptococcus OTUs were aligned to the Greengenes database of 16S rRNA sequences. In each case, the top hit (>96%) was used for final classification. Perfect match was indicated (*) if found. (A) Background samples contained predominantly Streptococcus thermophilus. (B) Supraglottic samples contained predominantly Streptococcus mitis. (C) Relative abundances of all Streptococcus OTUs in all 29 BAL samples clustered as indicated in Figure 3 (never-smokers indicated with green label and smokers with blue label). PneumotypeUN is enriched with Streptococcus thermophilus while PneumotypeSCT has higher relative abundances of Streptococcus mitis.

Figure 6

Relationship between pneumotype/supraglottic-characteristic taxa and lung inflammation. Inflammatory cells (lymphocytes and neutrophils) and eNO were measured in BAL of asymptomatic subjects (never-smokers in open circles and smokers in closed circles). (A, B,C) Comparisons of inflammatory cells and eNO between pneumotypeUN and pneumotypeSCT (Mann–Whitney). PneumotypeSCT was associated with higher inflammatory BAL cells and higher levels of eNO. (D, E,F) Correlations between relative abundance of Veillonella, supraglottic-characteristic taxa, in BAL in relation to inflammatory cells/eNO (Spearman rho correlation). Veillonella showed a positive correlation with both BAL inflammatory cells and eNO.