Gut microbiome and metabolome profiling in Framingham heart study reveals cholesterol-metabolizing bacteria

Abstract

Accumulating evidence suggests that cardiovascular disease (CVD) is associated with an altered gut microbiome. Our understanding of the underlying mechanisms has been hindered by lack of matched multi-omic data with diagnostic biomarkers. To comprehensively profile gut microbiome contributions to CVD, we generated stool metagenomics and metabolomics from 1,429 Framingham Heart Study participants. We identified blood lipids and cardiovascular health measurements associated with microbiome and metabolome composition. Integrated analysis revealed microbial pathways implicated in CVD, including flavonoid, γ-butyrobetaine, and cholesterol metabolism. Species from the Oscillibacter genus were associated with decreased fecal and plasma cholesterol levels. Using functional prediction and in vitro characterization of multiple representative human gut Oscillibacter isolates, we uncovered conserved cholesterol-metabolizing capabilities, including glycosylation and dehydrogenation. These findings suggest that cholesterol metabolism is a broad property of phylogenetically diverse Oscillibacter spp., with potential benefits for lipid homeostasis and cardiovascular health.

Keywords: Cardiovascular disease; Cholesterol; Metabolome; Microbiome; Oscillibacter.

Figure 1.. Overview of FHS datasets.

(A) Schematic of study design and data collected. (B) Projection of MSP abundances using PCoA analysis (Bray-Curtis dissimilarity). MSPs correlated with the top two axes were visualized (axes scaled to display correlation coefficients). (C-D) Significant association between blood measurements and top two PCoA axes (Kruskal–Wallis test, P<0.05). (E) All 119,563 metabolite peaks detected in the cohort visualized using t-distributed stochastic neighbor embedding (t-SNE). Top 10 nominal HMDB metabolite classes are highlighted. (F) Composition of top abundant genera. (G) Fitted Locally Weighted Scatterplot Smoothing (LOWESS) curves for MSP diversity and richness (number), total peak richness (number of distinct detected peaks) and known peak richness. (H) LOWESS curve fitted to the composition of major HMDB classes (r = Spearman correlation between fraction of detected peaks belonging to a class and PCoA1 coordinates in B). Samples in (F-H) were sorted by first PCoA coordinate. P values adjusted for multiple comparisons with the Benjamini-Hochberg method.

Figure 2.. Significant associations between MSP abundances and CVD-relevant blood measurements.

(A) Associations between MSPs and blood measurements (P_adj <0.1). Bar direction indicates +/− association. The most specific taxonomy was annotated to each MSP (species shown, unless otherwise indicated: g, genus; p, phylum). (B-D) MSP-blood measurement associations shown with fitted linear regression line. P values adjusted for multiple comparisons with the Benjamini-Hochberg method.

Figure 3.. Significant associations between metabolites and CVD-relevant blood measurements.

(A) LC-MS peaks significantly associated with blood measurements visualized by t-SNE plot (P_adj<0.05). (B) Distribution of association counts across nominal metabolite classes with the most associations. (C-E) Regression estimates for significant associations between peak co-abundance clusters (Methods) and blood measurements. A cluster is displayed if ≥ one peak within a class associated with a given measurement (P_adj<0.05). Peaks with provided standards are highlighted. P values adjusted for multiple comparisons with the Benjamini-Hochberg method.

Figure 4.. Associations between FHS gut microbiome and metabolome.

(A) Spearman correlations between MSPs and known LC-MS peaks matched to internal standards (MSPs or metabolites with >5 associations with absolute correlation coefficient >0.3 were shown; cells where P_adj≥0.05 were omitted (gray). (B) Associations between F. plautii abundances and mono-phenolic acids (PPA: phenyl-propionic acid, HPAA: hydroxy-phenyl-acetic acid, DHPPA: dihydroxy-phenyl-propionic acid) with fitted linear regression line. (C) Strong positive associations between MSPs and unknown peaks (Spearman correlation >0.4; MSPs with >5 associations shown). Number of peaks associated with each MSP is shown in parentheses. (D) Strong associations between MSPs and unknown peaks with fitted linear regression line. (E) Network showing potential metabolites predicted using additional MS/MS data for QI29637 featured in (D). (F) Distribution of number of associated MSPs (absolute Spearman correlation >0.4) for known and unknown LC-MS peaks.

Figure 5.. Oscillibacter spp. are strongly associated with stool cholesterol and its derivatives.

(A) Cholesterol to coprostanol biotransformation pathway. (B) Spearman correlation coefficients for MSPs in most represented genera, IsmA encoders and Other (“+”: mean). (C) Scaled Oscillibacter MSP abundances in FHS samples grouped based on presence/absence of IsmA-encoding MSPs. Samples annotated by corresponding cholesterol, cholestenone, and coprostanol abundances (log10(peak intensity)) and Oscillibacter total abundance. (D) Stool cholesterol abundance in samples stratified by presence of selected Oscillibacter MSPs (Methods) and IsmA encoders. (E) Candidate cholesterol derivative metabolic peaks. Seven seed peaks with predicted formula C₂₇H₄₆O, including authentic standard and MS/MS matches are extended using mass shifts and absolute retention time difference <0.5. (F) Targeted MS/MS of predicted peaks yielded 36 formula predictions, 19 MS/MS curated identities and 6 peaks with standards. (G) Spearman correlation coefficients between Oscillibacter genus abundance and predicted peaks in ismA+ or ismA− samples shown with fitted regression line (intercept=0). Black points, cholestenone peaks. (H) Relationship between Oscillibacter genus abundance and stool cholesterol in ismA+ and ismA− samples with fitted linear regression line. (I) Mass shift network between LC-MS peaks and Oscillibacter associations for ismA+ and ismA− samples. Each node represents a metabolic peak and edges represent assumed biotransformations. Colors denote effect size of association with Oscillibacter genus. (J) Oscillibacter and IsmA encoders were associated with decreased plasma cholesterol in a combinatorial manner. For D,J: *P<0.05, **P<0.01, Wilcoxon test. n.s., non-significant.

Figure 6.. Functional potential of Oscillibacter in cholesterol metabolism.

(A) Predicted PROSE distance of Oscillibacter isolate proteins with key cholesterol metabolism proteins. Oscillibacter proteins with sequence similarity are highlighted in black. (B) Sequence identity and coverage between MSPs proteins and cholesterol metabolism candidates identified from RJX3347 in (A). (C) Phylogenetic tree of MSPs, RJX3347, RJX3711, J115 and references from GTDB for the main genera negatively associated with stool cholesterol. Dots outside the tree show sequence identity and coverage compared to cholesterol metabolism candidates from RJX3347 (A.ond: Alistipes onderdonkii, B.the: Bacteroides thetaiotaomicron, B.obe: Blautia obeum, E.cop: Eubacterium coprostanoligenes, P.cop: Prevotella copri, R.hom: Roseburia homini). (D) Predicted protein structures of candidates RJX3347_02204 (magenta) and RJX3711_01178 (green) superimposed on IsmA from Eubacterium coprostanoligenes (grey). (E) Zoomed-in view highlighting conserved catalytic triad. Residues from reference sequence ECOP170 shown in red. (F) Predicted protein structures of candidate RJX3347_02251 (yellow) and cholesterol-alpha-glucosyltransferase (CgT) from Helicobacter pylori (gray) superimposed on an experimentally solved structure (3qhp, light blue).

Figure 7.. Oscillibacter sp. J115, RJX3347 and RJX3711 metabolize cholesterol via multiple pathways.

(A) Cholesterol intensity in unspent media (left) and after exposure to J115, RJX3347, RJX3711 and Bacteroides thetaiotaomicron VPI-5482. (B) Log fold change in peak intensity for Oscillibacter isolates compared to B. thetaiotaomicron (B.theta). Only peaks detected in all isolates were shown. CHO, media with cholesterol; control, media with ethanol solvent only. (C-D) Intensity in cell pellets for cholesterol and Oscillibacter-produced derivatives in media with CHO (C) or ¹³C isotope-labeled cholesterol (CHO*) (D). Stars in chemical structures indicate isotope-labeled carbon atoms. (E) Intensity of peaks with a matching isotope-labeled counterpart in the labeled cholesterol experiments.