Microbial metabolites in chronic heart failure and its common comorbidities

Figures

• 

  Figure EV1. CONSORT diagram for REM‐HF cohort CHF, chronic heart failure; CKD, chronic kidney disease; DCM, Dilated cardiomyopathy; NGT, normal glucose tolerance; T2D, type 2 diabetes.

• 

  Figure EV2. Absolute plasma levels of all 151 microbially associated metabolites in REM‐HF cohort The central band in each box represents the median, the top, and bottom of the box the 25th and 75th percentiles, and the whiskers 1.5 times the interquartile range (n = 260; biological replicates).

• 

  Figure EV3. Venn plots depicting numbers of shared metabolites across cohorts A–FNumbers of metabolites (A) and significantly changed metabolites (C) that were overlapping between REM‐HF and EPIC‐Norfolk cohorts and similarly for those (B, D) between REM‐HF and BPRHS cohorts. Numbers of increased (E) and decreased (F) metabolites across all three cohorts, respectively.

• 

  Figure 1. Microbially associated metabolites altered in CHF and its comorbidities First panel: the heatmap indicating the mean serum concentrations (scaled by row) of 96 metabolites in the REM‐HF cohort. P values were calculated by the Wilcoxon rank‐sum test. Symbols indicate the raw P values (− P < 0.1, *P < 0.05, + P < 0.01, # P < 0.001), and symbols with open circles indicate metabolite‐disease links that were potentially confounded by medications. Second panel: the boxplot showing the absolute serum concentrations of the corresponding metabolites with box colors for different pathway annotations. The central band in each box represents the median, the top, and bottom of the box the 25th and 75th percentiles, and the whiskers 1.5 times the interquartile range (n = 260; biological replicates). Third panel: the bar plot showing the fold‐changes (FC) of 39 metabolites associated with the cardiometabolic stress scores (group 1–3) in the BPRHS cohort. Forth panel: the bar plot showing the hazard ratio (HR) of each metabolite for the incidence of HF, T2D, or CKD identified in the EPIC‐Norfolk cohort. Last panel: the bar plot showing the hazard ratios (HRs) of 33 metabolites for CHF…

• 

  Figure 2. Metabolic distinctiveness among all three cardiometabolic diseases AVenn diagram showing the shared and unique metabolites associated with CHF, T2D, and CKD (Wilcoxon rank‐sum test, P < 0.05; ImP with a marginal signicance level (P = 0.055) was also included).BThe phenylalanine metabolism pathway. The color code of each metabolite shows their associations with different diseases.C–EBoxplots showing the serum level of phenylacetylglutamine, mandelic acid and homovanillic acid in different CHF subgroups in the REM‐HF cohort. NGT: n = 23; NGT + CHF: n = 48; Prediabetes+CHF: n = 83; T2D + CHF: n = 56; Prediabetes + CHF + CKD: n = 34; T2D + CHF + CKD: n = 16 (biological replicates).FThe boxplot showing the serum level of ImP in the Swedish prediabetes versus the Chinese REM‐HF cohorts. For Swedish prediabetes cohort, NGT: n = 125; prediabetes: n = 262; T2D: n = 17 (biological replicates). For Chinese REM‐HF cohort: NGT: n = 23; NGT + CHF: n = 48; Prediabetes+CHF: n = 83; T2D + CHF: n = 56; Prediabetes+CHF + CKD: n = 34; T2D + CHF + CKD: n = 16.GDifferences in ImP when stratified by left ventricular ejection fraction (LVEF; two‐tailed Wilcoxon rank‐sum tests). NGT: n = 23; HFpEF: LVEF…

• 

  Figure EV4. Metabolites that differed between different LVEF groups in REM‐HF cohort The central band in each box represents the median, the top, and bottom of the box the 25th and 75th percentiles, and the whiskers 1.5 times the interquartile range. NGT: n = 23, HFpEF: LVEF ≥ 50, n = 66; HFmEF: 40 < LVEF < 50, n = 68; HFrEF: LVEF ≤ 40, n = 103 (biologcial replicates).

• 

  Figure 3. Cardiomyoblast‐metabolite coculturing and induction of NPPB gene expression AThe relative expression of NPPB (scaled by row) upon different metabolites exposure in H9c2 cells pretreated with or without hypoxia/reoxygenation (H/R; six replicates per group). Symbols indicate the raw P values (− P < 0.1, *P < 0.05, + P < 0.01) calculated by the Wilcoxon rank‐sum test.BRepresentative JC‐1 staining images showing red fluorescence of JC‐1 aggregates and green signal of monomers. Scale bars: 20 μm.CQuantification of mitochondrial membrane potential (n = 6, biological replicates). The data are shown as mean ± s.d.; Wilcoxon rank‐sum test (*P < 0.05, + P < 0.01). Source data are available online for this figure.

• 

  Figure EV5. The relative expression of NPPB upon phenylacetylglutamine exposure The relative expression of NPPB was detected by RT‐PCR after 6‐h exposure to 100 μM phenylacetylglutamine (n = 6, biological replicates). The data are shown as mean ± s.d. Statistical analysis was performed by the Wilcoxon rank‐sum test.

• 

  Figure 4. Mediation linkages among metabolites, clinical biomarkers, and diseases AThe association between altered metabolites and clinical variables. The bar plot shows the variance of serum metabolites that were explained by clinical factors based on the random forest model. The bubble plot depicts the importance of each clinical factor to the variation of the corresponding metabolite.B, CThe Sankey plot indicates the mediation relationships among metabolites, clinical factors, and diseases, with the clinical factors (B) and metabolites (C) as the underlying mediators, respectively. The ribbons were colored according to the associated diseases and text colored by the changes of corresponding metabolites or clinical factors, with red indicating increased and blue decreased serum levels in CHF patients. Source data are available online for this figure.

• 

  Figure 5. Performance of the metabolites in CHF prognosis AThe Kaplan–Meier survival curve based on the Framingham and GWTG‐HF risk scores. In total, 240 participants with clinical variables for Framingham and GWTG‐HF risk scores calculation were included in survival analysis (biological replicates). P values were calculated based on the log‐rank test.BThe Kaplan–Meier survival curve based on each microbially associated metabolite (gray colors except for imidazole propionic acid) and a combined metabolite risk score. In total, 244 participants who completed the follow‐up study were included in survival analysis (biological replicates). P values were calculated based on the log‐rank test.CThe time‐dependent receiver operating characteristic (ROC) curves for the performance of GWTG‐HF scores, metabolite‐based risk score, and the combination of both scores in CHF prognosis. The AUC values were determined by bootstrap resampling with 1,000 iterations. Source data are available online for this figure.