Metagenomic and metabolomic analyses unveil dysbiosis of gut microbiota in chronic heart failure patients

Abstract

Previous studies suggested a possible gut microbiota dysbiosis in chronic heart failure (CHF). However, direct evidence was lacking. In this study, we investigated the composition and metabolic patterns of gut microbiota in CHF patients to provide direct evidence and comprehensive understanding of gut microbiota dysbiosis in CHF. We enrolled 53 CHF patients and 41 controls. Metagenomic analyses of faecal samples and metabolomic analyses of faecal and plasma samples were then performed. We found that the composition of gut microbiota in CHF was significantly different from controls. Faecalibacterium prausnitzii decrease and Ruminococcus gnavus increase were the essential characteristics in CHF patients' gut microbiota. We also observed an imbalance of gut microbes involved in the metabolism of protective metabolites such as butyrate and harmful metabolites such as trimethylamine N-oxide in CHF patients. Metabolic features of both faecal and plasma samples from CHF patients also significantly changed. Moreover, alterations in faecal and plasma metabolic patterns correlated with gut microbiota dysbiosis in CHF. Taken together, we found that CHF was associated with distinct gut microbiota dysbiosis and pinpointed the specific core bacteria imbalance in CHF, along with correlations between changes in certain metabolites and gut microbes.

Figure 1

Compositional and structural shifts of gut microbiota in CHF patients. (a) Principal coordinates analysis of beta-diversity analysis based on Bray Curtis distances of 86 genera differentially enriched across controls, DCM- and ICM-induced CHF patients. The represents control. The represents DCM-induced CHF. The represents ICM-induced CHF. (b,c) Boxplot of top ten genera differentially enriched in CHF patients (b) and controls (b). Black, controls; blue, DCM-induced CHF patients; red, ICM- induced CHF patients.

Figure 2

Gut bacteria differentially enriched across DCM-induced CHF, ICM-induced CHF and controls at genus level. Heatmap of genera differentially enriched across controls, DCM- and ICM-induced CHF patients. The abundance profiles were transformed into Z scores. Black, enriched in controls; blue, enriched in DCM-induced CHF patients; red, enriched in ICM- induced CHF patients.

Figure 3

CAGs differentially enriched between CHF patients and controls. The direction of enrichment was determined by Wilcoxon rank sum test (p < 0.05). Sizes of the nodes were in proportion with each CAGs’ gene numbers. CAGs within the same family were in the same colour. Edges between nodes represented Spearman’s correlation >0.9 (green), between 0.8 and 0.9 (blue) or <−0.55 (red). The represented the presence of the genes encoding choline TMA-lyase, choline TMA-lyase-activating enzyme, betaine reductase or tryptophanase. The represented the presence of the genes encoding butyrate-acetoacetate CoA transferase, propionate CoA-transferase or formate-tetrahydrofolate ligase.

Figure 4

Overview of functional shifts of the gut microbiota in CHF patients. Heatmap and hierarchical clustering of KO modules enriched across controls, DCM- and ICM- induced CHF patients. Modules differentially enriched across groups were identified on the basis of the reporter score derived from each KO’s Z-score. Blue, enriched in controls; red, enriched in CHF patients.

Figure 5

Important functional shifts of the gut microbiota in CHF patients. (a,b) Modules in lipopolysaccharide biosynthesis (a), TMA, tryptophan and lipid metabolism (b). Black, enriched in controls; blue, enriched in DCM-induced CHF patients; red, enriched in ICM- induced CHF patients. (c,d) Group level abundance shifts of choline TMA-lyase (c), butyrate-acetoacetate CoA transferase (d) between CHF patients and controls by Wilcoxon rank test. Black, enriched in controls; blue, enriched in CHF patients.

Figure 6

Metabolomic analyses of faecal and plasma samples of CHF patients and controls. (a,b) The PCA scores plot based on faecal metabolic profiles in ES+ (a) and ES− (b). (c,d) The OPLS-DA scores plot based on faecal metabolic profiles in ES+ (c) and ES− (d). (e,f) The PCA scores plot based on plasma metabolic profiles in ES+ (e) and ES− (f). (g,h) The OPLS-DA scores plot based on plasma metabolic profiles in ES+ (g) and ES− (h). The represents metabolic profiles of CHF patients. The represents metabolic profiles of controls. ES+ = positive ion mode; ES− = negative ion mode.

Figure 7

Correlations between plasma metabolic patterns and genera. Spearman’s correlation coefficients between the abundance of top 35 differentially enriched genera and the level of plasmatic metabolic patterns were calculated. Those with low correlation (|r| < 0.6) were not shown. Red, negative correlation; blue, positive correlation, ⁺q < 0.05, *q < 0.01. The enriched type of each genera and metabolic patterns was coloured according to its direction of enrichment. Black, enriched in controls; blue, enriched in CHF patients.