Microbial contribution to metabolic niche formation varies across the respiratory tract

Abstract

Variations in the airway microbiome are associated with inflammatory responses in the lung and pulmonary disease outcomes. Regional changes in microbiome composition could have spatial effects on the metabolic environment, contributing to differences in the host response. Here, we profiled the respiratory microbiome (metagenome/metatranscriptome) and metabolome of a patient cohort, uncovering topographical differences in microbial function, which were further delineated using isotope probing in mice. In humans, the functional activity of taxa varied across the respiratory tract and correlated with immunomodulatory metabolites such as glutamic acid/glutamate and methionine. Common oral commensals, such as Prevotella, Streptococcus, and Veillonella, were more functionally active in the lower airways. Inoculating mice with these commensals led to regional increases in several metabolites, notably methionine and tyrosine. Isotope labeling validated the contribution of Prevotella melaninogenica in generating specific metabolites. This functional characterization of microbial communities reveals topographical changes in the lung metabolome and potential impacts on host responses.

Keywords: COPD; genomics; lower airway dysbiosis; lung dysbiosis; lung inflammation; metabolomics; metatranscriptomics; microbiome; oral commensals; transcriptomics.

Figure 1.. Topographical evaluation of metabolomes and functional microbiomes

(A) Principal-component analysis (PCA) of metabolomic data comparing sample types, bronchoalveolar lavage (BAL, blue), upper airway (UA, yellow), and environmental background (BKG, gray) samples with permutational multivariate analysis of variance (PERMANOVA) p value. (B) Partial least squares discriminant analysis (PLS-DA) was used to identify metabolites enriched in BAL (blue) and UA (yellow), displaying component 1 of PLS-DA analysis on the x axis. (C) Differentially enriched functional metagenomic (MG) data based on fold change versus log₁₀ adjusted p value (false discovery rate [FDR]) using edgeR comparing BAL and UA. Pathways with a positive log fold change are enriched in BAL and those with a negative fold change are enriched in UA. Only pathways with an FDR < 0.2 are displayed. Bubble size is based on relative abundance. (D) Differentially enriched functional metatranscriptomic (MT) data based on fold change versus FDR using edgeR comparing BAL and UA. Pathways with a positive log fold change are enriched in BAL and those with a negative fold change are enriched in UA. Only pathways with an FDR < 0.2 are displayed. Bubble size is based on relative abundance. (E) Relative abundance of top lower airway microbial pathways in metatranscriptome data for BAL and UA. (F) Microbial functions significantly upregulated in lower airways samples represented as a delta of MT to MG by Wilcoxon rank sum test, p < 0.05. Data is represented as box-whisker plot with interquartile range. (G) Relative abundance of top upper airway microbial pathways in metatranscriptome data for BAL and UA. (H) Microbial functions significantly upregulated in upper airways samples represented as a delta of MT to MG by Wilcoxon rank sum test, p < 0.05. Data is represented as box-whisker plot with interquartlie range.

Figure 2.. Multi-omic analysis

(A) Schematic of data used for network HeatWave analysis. (B) HeatWave analysis based on differential expressions for metabolomic and microbial functional metatranscriptomic datasets for human UA and BAL samples. Highlighted nodes include several metabolite nodes with microbial function correlations, such as inosine, adenosine, glutamic acid/glutamate, homocysteine, niacinamide, and tyrosine. Color gradient of light blue (extreme downregulation) to light red (extreme upregulation), with a limit of 8-fold change or more in either direction. Circles represent metabolites whereas round cornered squares represent enzymes (orthologs). An interactive version of the main network is available in the supplement (Data S1).

Figure 3.. Topographical evaluation of the microbial taxonomy

(A) Relative abundance of top lower airway taxa in metatranscriptome (MT) data for BAL (blue) and UA (yellow). (B) Taxa significantly upregulated in lower airways samples represented as a delta of MT to metagenome (MG) by Wilcoxon rank sum test, p < 0.05. Data is represented as a box-whisker plot with interquartlie range. (C) Relative abundance of top upper airway taxa in metatranscriptome data for BAL and UA. (D) Taxa significantly upregulated in upper airways samples represented as a delta of MT to MG by Wilcoxon rank sum test, p < 0.05. Data is represented as a box-whisker plot with interquartlie range.

Figure 4.. Pre-clinical mouse model

(A) Murine experimental model and exposures. (B) In vivo relative abundance in the metagenomic dataset of P. melaninogenica (orange), S. mitis (green), and V. parvula (purple) in tongue and lung samples for the three different inoculations, including PBS to the lung (PBS_Lung), MOC to the oral cavity (MOC_Oral), and MOC to the lung (MOC_Lung). Comparisons and p values based on Wilcoxon rank sum test. Only statistically significant comparisons are shown. Data is represented as a box-whisker plot with interquartile range. (C) In vivo relative abundance in the metatranscriptomic dataset of P. melaninogenica, S. mitis, and V. parvula in tongue and lung samples for the three different inoculations, including PBS_Lung, MOC_Oral, and MOC_Lung. Comparisons and p values based on Wilcoxon rank sum test. Only statistically significant comparisons are shown. Data is represented as a box-whisker plot with interquartlie range. (D) Differentially enriched functional metagenomic data from murine lung samples based on fold change versus FDR using edgeR comparing two different exposures. Pathways with a positive log fold change are enriched in mice inoculated with MOC_Lung (blue) and those with a negative fold change are mice inoculated with PBS_Lung (gray). Only pathways with an FDR < 0.2 are displayed. Bubble size is based on relative abundance. Pathways not associated with the three MOC taxa (P. melaninogenica, S. mitis, and V. parvula) were italicized and colored gray. (E) Differentially enriched functional metatranscriptomic data from murine lung samples based on fold change versus FDR using edgeR comparing two different exposures. Pathways with a positive log fold change are enriched in mice inoculated with MOC_Lung and those with a negative fold change are mice inoculated with PBS_Lung. Only pathways with an FDR < 0.2 are displayed. Bubble size is based on relative abundance. Pathways not associated with the three MOC taxa were italicized and colored gray.

Figure 5.. Metabolomic features in an in vivo mouse model of airway dysbiosis

(A) PLS-DA was used to identify metabolites, known to be produced by the three MOC taxa, enriched in murine lung samples inoculated with MOC_Lung (blue) and PBS_Lung (gray). (B) Taxonomic contribution of P. melaninogenica (orange), S. mitis (green), and V. parvula (purple) in the metagenomic dataset for niacinamide, methionine, inosine, adenosine 5-monophosphate, glutamic acid/glutamate, adenosine, and tyrosine. (C) Taxonomic contribution of P. melaninogenica,S. mitis, and V. parvula in the metatranscriptomic dataset for niacinamide, methionine, inosine, adenosine 5-monophosphate, glutamic acid/glutamate, adenosine, and tyrosine.

Figure 6.. ¹³C Isotope glucose labeling to identify microbial contribution to measured metabolites

(A) Experimental design for ex vivo ¹³C isotope glucose labeling model. (B) Ex vivo comparison of C13 labeling intensity for several metabolites, including fructose 1,6-bisphosonate, adenosine, inosine, glutamic acid/glutamate, and methionine. *indicates a significant (p < 0.05) difference between C12 and C13 experiments for each metabolite based on Wilcoxon signed ranged test. Data is represented as a barplot with an error bar. (C) Experimental design for in vivo model. (D–F) (D) PCA of in vivo murine metabolomic data comparing mice inoculated with (C13) and without (C12) a C13-labeled glucose P. melaninogenica, sacrificed at 1 and 6 h, with PERMANOVA p value. Comparison of C13-labeled percentage adenosine, inosine, glutamic acid/glutamate, and methionine in the C13 (turquoise) and C12 (green border) samples at (E) 1 h and (F) 6 h. * indicates a significant (p < 0.05) difference between C12 and C13 experiment for each metabolite based on Wilcoxon signed ranged test. Data is represented as a barplot with an error bar. (G) Mass spectrometry analysis of three metabolites, adenosine, inosine, and methionine, showing intensity and weight in the first panel followed by time and intensity curves for each peak identified.