Gut Microbial Genes and Metabolism for Methionine and Branched-Chain Amino Acids in Diabetic Nephropathy

Abstract

Diabetic mellitus nephropathy (DMN) is a serious complication of diabetes and a major health concern. Although the pathophysiology of diabetes mellitus (DM) leading to DMN is uncertain, recent evidence suggests the involvement of the gut microbiome. This study aimed to determine the relationships among gut microbial species, genes, and metabolites in DMN through an integrated clinical, taxonomic, genomic, and metabolomic analysis. Whole-metagenome shotgun sequencing and nuclear magnetic resonance metabolomic analyses were performed on stool samples from 15 patients with DMN and 22 healthy controls. Six bacterial species were identified to be significantly elevated in the DMN patients after adjusting for age, sex, body mass index, and estimated glomerular filtration rate (eGFR). Multivariate analysis found 216 microbial genes and 6 metabolites (higher valine, isoleucine, methionine, valerate, and phenylacetate levels in the DMN group and higher acetate levels in the control group) that were differentially present between the DMN and control groups. Integrated analysis of all of these parameters and clinical data using the random-forest model showed that methionine and branched-chain amino acids (BCAAs) were among the most significant features, next to the eGFR and proteinuria, in differentiating the DMN group from the control group. Metabolic pathway gene analysis of BCAAs and methionine also revealed that many genes involved in the biosynthesis of these metabolites were elevated in the six species that were more abundant in the DMN group. The suggested correlation among taxonomic, genetic, and metabolic features of the gut microbiome would expand our understanding of gut microbial involvement in the pathogenesis of DMN and may provide potential therapeutic targets for DMN. IMPORTANCE Whole metagenomic sequencing uncovered specific members of the gut microbiota associated with DMN. The gene families derived from the discovered species are involved in the metabolic pathways of methionine and branched-chain amino acids. Metabolomic analysis using stool samples showed increased methionine and branched-chain amino acids in DMN. These integrative omics results provide evidence of the gut microbiota-associated pathophysiology of DMN, which can be further studied for disease-modulating effects via prebiotics or probiotics.

Keywords: branched-chain amino acid; diabetic nephropathy; gut microbiota; metabolite; metagenome; methionine; microbiota.

FIG 1

Gut microbial differences between the diabetic mellitus nephropathy (DMN) and healthy control (CTL) groups. (a to c) Shannon diversity (a), richness (b), and principal-coordinate analysis (PCoA) (c) plots calculated with the Bray-Curtis method for the DMN and healthy control groups. P values for alpha diversity and beta diversity metrics were calculated by a Wilcoxon rank sum test and PERMANOMA, respectively. ns, not significant. (d) Six species showed significant differences between the DMN and healthy control groups by multivariable-adjusted MaAsLin2 analysis (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

FIG 2

Heat map plot of functional pathways. Significantly different functional KEGG pathways between the diabetic mellitus nephropathy (DMN) and healthy control (CTL) groups are listed in the rows. Sample names are listed in the columns. Colors in the heat map represent the abundance of each pathway in samples. Adjusted P values and group classifications are shown at the left and top margins of the heat map, respectively. TCA, tricarboxylic acid.

FIG 3

Stool metabolite profiling in the diabetic mellitus nephropathy (DMN) and healthy control (CTL) groups. (a and b) OPLS-DA score plots for water-soluble (a) and lipid-soluble (b) metabolites of the DMN and healthy control groups. Models were obtained using one predictive (P_p) and two orthogonal (P_o) components (n = 11 for the DMN group; n = 13 for the healthy control group). (c) Metabolites contributing to DMN from statistical total correlation spectroscopy (STOCSY). The model coefficients for each NMR variable from water-soluble metabolites are shown. As a discriminator between the two groups, a color scale based on the value of P_(corr)p according to weight is used. P_p represents the modeled covariant. Metabolite signals that differed significantly between the DMN and healthy control groups are depicted on the coefficient plot. (d) Levels of the six metabolites are significantly different between the DMN and healthy control groups. The levels represent binned peak integrals from the NMR spectrum normalized against the sum of the integrals of the entire spectrum. All bar plots and error bars represent means and standard errors, respectively (*, 0.01 < P < 0.05; ***, P < 0.001). A.U., arbitrary units.

FIG 4

Integrated analysis of the predictive potential of variables for DMN. (a) Feature importance values of the top 20 variables obtained from random-forest analysis on clinical factors (blue), KEGG gene families (yellow), stool metabolites (gray), and gut microbiota species (red). Clinical factors included in this analysis were age, sex, BMI, eGFR, and UPCR. KEGG gene families and microbiota species that showed significant results in the MaAsLin2 analysis and stool metabolites found using peak-level differences in the NMR study are included. (b) Receiver operating characteristic curves for the prediction of DMN. The C statistics were calculated from all conclusive taxonomy, genes, and metabolites (green); taxonomy alone (yellow); KEGG gene family alone (red); and stool metabolites alone (blue). AUC, area under the curve.

FIG 5

(a) Pathways and gene families related to the biosynthesis of methionine and BCAAs. Blue and yellow squares represent dominant metabolites in the diabetic mellitus nephropathy (DMN) and healthy control (CTL) groups, respectively. Bar plots represent differences in each gene between the DMN and healthy control groups. All genes with bar plots were significantly different in the multivariable-adjusted MaAsLin2 analysis, and the asterisks in the bar plots represent P values of <0.05 by Kruskal-Wallis tests. PEP, phosphoenolpyruvate. (b) Contribution of specific members of the gut microbiota to the significant genes associated with methionine and BCAA pathways. All bar plots and error bars represent means and standard errors, respectively.

FIG 6

Association between HbA1c and stool metabolites. (a to c) Among methionine, valine, and isoleucine, which are significantly elevated in DMN, methionine and isoleucine showed a significant linear association with HbA1c levels. (d) Acetate, which was decreased in DMN, showed a reverse association with HbA1c levels. Black solid lines and gray zones in each graph represent fitted regression lines and 95% confidence intervals, respectively.