Supraglottic Lung Microbiome Taxa Are Associated with Pulmonary Abnormalities in an HIV Longitudinal Cohort

To the Editor:

Lung microbiome analysis of acellular BAL suggests the presence of two distinct lung pneumotypes, background predominant taxa (BPT) and supraglottic predominant taxa (SPT) (1, 2). The latter has been shown to be associated with a distinct metabolic profile and a T-helper cell type 17 proinflammatory phenotype among healthy individuals (2). We have previously demonstrated an altered lung microbiome in HIV-infected subjects, including persistent increases in taxa associated with pneumotype_SPT (3). However, the lung inflammatory profile and associated lung function has not been elucidated. In this work, we investigated associations between lung function, inflammation, and lung microbiome dysbiosis in an HIV-positive population with relatively advanced disease who were studied at baseline and over a 1-year follow-up.

Our study population consisted of 30 HIV-infected, treatment-naive adults who underwent analysis of lung inflammatory mediators and pulmonary function testing before antiretroviral therapy (ART) and again at 4 weeks and 1 year after starting ART (3). As previously reported (3), these subjects had advanced HIV disease at baseline (median and interquartile range of CD4 is 280 and 92–385) and had sustained viral control and improved CD4 counts after starting ART. Forty-eight genera had a relative abundance ≥2% and were included in this analysis. Taxa were then classified as either BPT or SPT based on previous classification (1, 2). Taxa not classified in these prior reports were included in the SPT group if they were known common oropharyngeal organisms. Taxa not described in these reports were placed in the BPT group if prior published work demonstrated they were not common oral taxa or left as unclassified or not defined taxa if no data were available linking them to a human compartment. In general, pneumotype_SPT was dominated by supraglottic phyla such as Bacteroidetes, Firmicutes, and Fusobacterium, whereas pneumotype_BPT was enriched with background predominant phyla such as Proteobacteria and Actinobacteria. The pairwise correlation between genera at baseline (before ART) and at 4 weeks and 1 year after ART is shown in Figure 1. Most (>80%) pairwise correlations between taxa within the pneumotype_SPT subgroup were positive (Figures 1A–1C). Correlations between pneumotype_BPT taxa were weaker, and the majority of the correlations between pneumotype_SPT and pneumotype_BPT were negative for all three time points. The negative correlation between pneumotype_SPT and pneumotype_BPT increased after receiving ART. These data suggest that taxa within pneumotype_SPT are relatively uniform and stable over time, reflecting organisms found in the oral cavity. In contrast, taxa within pneumotype_BPT are much more variable and likely influenced by environmental factors. Consistent with prior reports (4), adjusting for smoking had no significant effects on the respiratory microbiome.

Figure 1.

Heat map of Spearman correlation between genera at (A) baseline. (B) Four weeks after baseline. (C) One year after baseline. Spearman’s ρ correlation matrix was reordered by the hierarchical clustering algorithm with complete agglomeration. Red color represents positive correlation, whereas blue color shows negative correlation. Genera in pneumotype_BPT are in dark cyan font, and genera in pneumotype_SPT are in dark pink font. Dark gray font represents not defined pneumotype taxa. (D) A cooccurrence network of genus-level taxa generated by the SparCC (Sparse Correlation for Compositional Data) package and visualized by Cytoscape 3.2.6. Cooccurrences were assessed by significant weighted Spearman’s ρ calculated using all time points with permutation P < 0.05 and ρ > 0.6. Genera identified as pneumotype_BPT are in light blue, and genera identified as pneumotype_SPT are in light pink. The undefined genera are in light gray. The lung function parameters and inflammation markers are in yellow and purple, respectively. The dark red line indicates the positive association; the green line indicates the negative association. BPT = background predominant taxa; DsbHb = Dl_CO corrected for hemoglobin; FEF = forced expiratory flow; IP-10 = IFN-inducible protein 10; MCP-1 = monocyte chemoattractant protein-1; SPT = supraglottic predominant taxa.

A weighted Spearman correlation for repeated measures, using the number of observations as weights, was used to assess the association of taxa abundance with inflammatory markers as well as with lung function measurements (5). With the exception for Granulicatella, pneumotype_SPT taxa were positively associated with BAL proinflammatory cytokines (Table 1, top panel). It is worth noting that Tropheryma, which would be classified as BPT as it is not a typical supraglottic organism (6), was associated with greater inflammation. The only other pneumotype_BPT organism associated with lung inflammation was Moraxella, a known respiratory tract pathogen. Eleven genera were significantly associated with at least one spirometry or diffusion capacity measurement (i.e., FEV₁; FVC; forced expiratory flow, midexpiratory phase; and Dl_CO corrected for hemoglobin shown in Table 1, bottom panel). Among them, six genera were pneumotype_SPT, three were pneumotype_BPT, and two did not belong to either group. Overall, pneumotype_BPT genera Burkholderia and Propionibacterium were significantly associated with better lung function in both spirometry and diffusion capacity. In contrast, except Neisseria, pneumotype_SPT taxa were associated with poorer lung function. Finally, we constructed a cooccurrence network using the SparCC (Sparse Correlation for Compositional Data) package (7) among genus-level taxa with significant associations with inflammatory cytokines, chemokines, and pulmonary function outcomes (Figure 1D). Genera within pneumotype_SPT and pneumotype_BPT cooccurred with each other, forming two well-defined clusters. Additionally, most genera in pneumotype_SPT (5 out of 8) were significantly correlated with chemokine/cytokine levels. In contrast, most SPT-characteristic genera (5 out of 6) were negatively correlated with lung function except for Neisseria. Similar results were observed if each time point was analyzed separately.

Table 1.

Genera Significantly Associated with Chemokine or Cytokine Levels and Lung Function

Pneumotype	Genus	Phylum	Chemokine or Cytokine*	Weighted Spearman’s ρ (95% CI)*
BPT	Moraxella	Proteobacteria	IL-8	0.485 (0.151 to 0.719)
	Tropheryma	Actinobacteria	MCP-1	0.438 (0.093 to 0.690)
			IL-6	0.415 (0.064 to 0.674)
SPT†‡	Filifactor	Firmicutes	IL-8	0.662 (0.396 to 0.825)
			IP-10	0.463 (0.123 to 0.705)
			MIG	0.673 (0.413 to 0.832)
			IL-6	0.384 (0.027 to 0.654)
	Granulicatella	Firmicutes	MCP-1	−0.369 (−0.644 to −0.011)
			IL-6	−0.448 (−0.696 to −0.105)
	Gemella	Firmicutes	MIG	0.363 (0.003 to 0.639)
	Selenomonas	Firmicutes	IL-6	0.415 (0.064 to 0.674)
	Treponema	Spirochaetes	IL-8	0.434 (0.088 to 0.687)
NDT	Sneathia	Fusobacteria	MIG	0.367 (0.008 to 0.643)

Pneumotype	Genus	Phylum	Lung Function	Weighted Spearman’s ρ (95% CI)*
BPT	Burkholderia	Proteobacteria	FEV1	0.494 (0.163 to 0.725)
			FVC	0.538 (0.221 to 0.752)
			DsbHb	0.529 (0.209 to 0.747)
	Corynebacterium	Actinobacteria	FVC	0.433 (0.086 to 0.686)
	Propionibacterium	Actinobacteria	FEV1	0.361 (0.001 to 0.638)
			FVC	0.445 (0.101 to 0.694)
			DsbHb	0.650 (0.379 to 0.819)
			DsbHb% predicted	0.569 (0.262 to 0.771)
SPT†§	Filifactor	Firmicutes	FEF25–75	−0.419 (−0.677 to −0.069)
	Neisseria	Proteobacteria	DsbHb	0.490 (0.158 to 0.723)
			DsbHb% predicted	0.596 (0.301 to 0.787)
	Oribacterium	Firmicutes	FEV1	−0.418 (−0.676 to −0.068)
	Porphyromonas	Bacteroidetes	FEV1% predicted	−0.392 (−0.659 to −0.037)
			FVC% predicted	−0.370 (−0.644 to −0.011)
	Rothia	Actinobacteria	FEV1/FVC	−0.422 (−0.679 to −0.072)
	Treponema	Spirochaetes	FEF25–75	−0.440 (−0.691 to −0.094)
			FEF% predicted	−0.475 (−0.713 to −0.139)
NDT	Moryella	Firmicutes	FEV1	−0.368 (−0.643 to −0.008)
			FVC	−0.410 (−0.671 to −0.058)
	Capnocytophaga	Bacteroidetes	DsbHb% predicted	0.411 (0.059 to 0.672)

Definition of abbreviations: BPT = background predominant taxa; CI = confidence interval; DsbHb = Dl_CO corrected for hemoglobin; FEF = forced expiratory flow; FEF_25–75 = forced expiratory flow, midexpiratory phase; IP-10 = IFN-inducible protein 10; MCP-1 = monocyte chemoattractant protein-1; NDT = not defined taxa; SPT = supraglottic predominant taxa.

^*Weighted Spearman’s ρ that are larger than 2 SDs from mean were shown in the table.

^†A permutation test was adopted to evaluate the enrichment of positive or negative associations (defined by Spearman’s ρ with larger than 2 SDs from mean) among SPT genera with chemokine/cytokine or lung function levels. Ten thousand permutation replicates were used.

^‡Permutation P = 0.082.

^§Permutation P < 0.000.

In this study, we defined lung microbiome pneumotypes based on prior publications (1, 2). Pneumotype_SPT taxa are likely derived from the oral cavity and supraglottic region, considered widely as representing a true lower respiratory tract microbiome that have the potential to cause local inflammation. All other taxa have traditionally been grouped in the BPT pneumotype and are felt to represent “background” organisms. Our results highlight strengths and caveats to this paradigm, particularly in the context of HIV. The significant positive correlations between oropharyngeal taxa in Figures 1A–1C strongly support the concept of a supraglottic pneumotype. The lack of correlation between BPT taxa emphasizes that in the absence of SPT taxa, the pneumotype_BPT tends to be unique in different individuals. However, the concept that the pneumotype_BPT represents just background taxa needs to be reexamined. First, Tropheryma whipplei, a pneumotype_BPT taxa, was associated with lung inflammation. Widespread colonization of Tropheryma has been reported in the lungs of HIV-infected subjects and is dramatically reduced with ART (6). Although we classified Tropheryma as pneumotype_BPT as it has not been found in the oral cavity when sampled from oral washings (6), this organism is known to be present in the gastrointestinal tract, where it may cause Whipple’s disease. As such, Tropheryma is likely entering the lung compartment through the blood after translocation from the gut. Moraxella is another example of a taxa that was included in the BPT group but is clearly associated with lung inflammation in our cohort. Moraxella is a known lower respiratory tract pathogen that can cause lower airway inflammation. Even though they do not fall into the SPT category, certain groups of Moraxella (e.g., Moraxella catarrhalis) are known lung pathogens (8). These results highlight the fact that although most background taxa do not cause lung inflammation, there are instances through which bacteria not usually found in the supraglottic space may gain access to the lower respiratory tract and be associated with inflammation. The beneficial effect of some BPT taxa with lung function is intriguing. Whether this represents a true protective effect or just the absence of SPT bacteria requires further study. Finally, this work supports the paradigm that the presence of supraglottic taxa in the lower respiratory tract resulting from chronic low-grade aspiration seen in virtually all humans may be a susceptibility marker to chronic lung diseases seen in both HIV-infected subjects and non–HIV-infected aging population. This paradigm is further supported by recent work (9).

In conclusion, we demonstrated that there are significant correlations among lung microbiome composition, lower respiratory tract inflammation, and lung function in patients living with HIV. Our findings contribute to a growing body of work demonstrating correlations between the host microbiome and chronic disease (10), with implications for both the HIV-infected and HIV-uninfected population.