Gut microbiota and fermentation-derived branched chain hydroxy acids mediate health benefits of yogurt consumption in obese mice

Abstract

Meta-analyses suggest that yogurt consumption reduces type 2 diabetes incidence in humans, but the molecular basis of these observations remains unknown. Here we show that dietary yogurt intake preserves whole-body glucose homeostasis and prevents hepatic insulin resistance and liver steatosis in a dietary mouse model of obesity-linked type 2 diabetes. Fecal microbiota transplantation studies reveal that these effects are partly linked to the gut microbiota. We further show that yogurt intake impacts the hepatic metabolome, notably maintaining the levels of branched chain hydroxy acids (BCHA) which correlate with improved metabolic parameters. These metabolites are generated upon milk fermentation and concentrated in yogurt. Remarkably, diet-induced obesity reduces plasma and tissue BCHA levels, and this is partly prevented by dietary yogurt intake. We further show that BCHA improve insulin action on glucose metabolism in liver and muscle cells, identifying BCHA as cell-autonomous metabolic regulators and potential mediators of yogurt's health effects.

Fig. 1. Yogurt consumption preserves glucose homeostasis and insulin sensitivity.

Metabolic results of mice in the three studies (see also Supplementary Fig. 1 for the experimental design, Supplementary Data 1 and Supplementary Fig. 2). a Body weight throughout the Study 1. b Pooled analysis of the three studies showing mice final body weight after 12 weeks of dietary treatment. c Energy intake was recorded from week 4 to week 12 in Study 1. d Pooled analysis of the three studies showing mice daily energy intake. e–g Pooled analysis of fasting glucose, insulin, and corresponding HOMA-IR index after 12 weeks of treatment. h Glucose tolerance test at week 11 in Study 1. i Insulin response following the oral glucose challenge in Study 1. j C-peptide response following the oral glucose challenge in Study 1. k–m In Study 3, mice underwent an hyperinsulinemic-euglycemic clamp at week 12. k Glucose infusion rate (GIR). l Glucose rate of appearance (Ra) depicting hepatic glucose production. m Whole-body glucose rate disappearance (Rd) depicting glucose disposal. Data are expressed as mean ± SEM. Study 1: n = 18–24 except for panel (i) (n = 9–12) and (j) (n = 6–10); Study 3: n = 23–27; Pooled analysis: n = 66–84 biologically independent mice. H versus C: *p < 0.05, **p < 0.01, ***p < 0.001. Y versus H: ^#p < 0.05, ^##p < 0.01. For clamps data, basal versus clamped condition: ^@@p < 0.01, ^@@@p < 0.001. C: low-fat low-sucrose control diet (C: soft gray, C1: white, C3: gray); H: high-fat high-sucrose diet with a protein mixture replacing casein (H: soft red, H1: pink, H3: red); Y: lyophilized yogurt incorporated in H diet (Y: soft blue, Y1: light blue, Y3: blue). Numbers (1, 2, 3) refer to the study affiliation. Two-way repeated-measures ANOVA or Mann–Whitney tests depending on data distributions with (a, h–j) or without (b–g, k–m) baseline adjustment. T-test at baseline (j). All tests were two-sided.

Fig. 2. Yogurt preserves hepatic steatosis and function.

Liver state in mice from Study 1. a Liver weight. b Hepatic triglyceride content. c Representative liver sections (hematoxylin and eosin staining) of n = 16–24 sections from independent mice. d, e Pathological analyses of liver sections. d Hepatic steatosis. e Global histologic diagnosis. f Fibrosis score. g Liver expression of genes involved in fatty acid uptake, de novo lipogenesis and fatty acid oxidation. Data are expressed as mean ± SEM. n = 18–24 biologically independent mice, except for panel (g) (n = 8 for C1 group, n = 20 for H1 and Y1 groups). H versus C: **p < 0.01, ***p < 0.001. Y versus H: ^#p < 0.05, ^##p < 0.01. For panel (g) expressed in percentage, C group is considered as a reference and represented by a dash line. C: low-fat low-sucrose control diet (C1: white); H: high-fat high-sucrose diet with a protein mixture replacing casein (H1: pink); Y: lyophilized yogurt incorporated in H diet (Y1: light blue). Number (1) refers to the study affiliation. Mann–Whitney tests or T-test depending on data distributions (a, b). One-way ANOVAs and Benjamini–Hochberg adjustment for multiple testing on the number of variables (g). All tests were two-sided.

Fig. 3. Yogurt impact is partly driven by the gut microbiota.

a–c Fecal gut microbiota assessed at 12 weeks for Study 1, 3, and 15 weeks for Study 2. a Alpha diversity-observed OTUs, Shannon index, and Simpson reciprocal index (indices adjusted by their baseline value, least square mean and 95% CI). b Beta diversity-Principal Coordinate Analyses (PCoA) of Bray-Curtis dissimilarity, weighted UniFrac distances, and unweighted UniFrac distances with 95% confidence ellipses by group. c Heatmap of DESeq log2 fold changes over C or H group for genera significantly altered by H diet or Y treatment, respectively. Non-significant fold changes set to 0. Ward’s clustering criterion applied on genera. d Fecal hyodeoxycholic acid content at week 3 in Studies 1 and 2. For panels (a–d): Study 1, n = 18–24; Study 2, n = 13–24 and Study 3, n = 23–27 biologically independent mice. e–i Germ-free mice gavaged with fecal material from Study 1 and fed H diet for 9 weeks. Three cages of n = 3–4 receiver mice were allocated to each group, named H1-transplanted (H1-T) and Y1-transplanted (Y1-T). At week 9 post fecal material transplantation, e Fasting glucose, insulin, and corresponding HOMA-IR, f glucose tolerance test (GTT) and g glucose-stimulated insulin secretion during GTT. h Liver weight and i Hepatic triglyceride content. For panels (e–i): n = 11–12 biologically independent mice. Data are expressed as mean ± SEM except if specified otherwise. H versus C: **p < 0.01, ***p < 0.001. Y versus H: ^#p < 0.05, ^##p < 0.01, ^###p < 0.001. Y1-T versus H1-T: ^$p < 0.05. C: low-fat low-sucrose control diet (C1: white, C2: dark gray, C3: gray); H: high-fat high-sucrose diet with a protein mixture replacing casein (H1: pink, H2: dark red, H3: red); Y: lyophilized yogurt incorporated in H diet (Y1: light blue, Y2: dark blue, Y3: blue); T: transplanted mice (H1-T: pink, hatched and Y1-T: light blue, hatched). Numbers (1, 2, 3) refer to the study affiliation. a Generalized least-squares model and Benjamini–Hochberg adjustment for multiple testing on the number of variables. One-way ANOVAs or Mann–Whitney tests depending on data distribution (d). e, h, i T-test or Mann–Whitney and (f, g): mixed model followed by Tukey post hoc test. All tests were two-sided.

Fig. 4. Yogurt intake increases levels of branched-chain hydroxy acids (BCHA) in metabolic tissues of H-fed mice which correlates with metabolic benefits.

a Hepatic BCHA levels in Study 1: alpha-hydroxyisocaproate (HICA, lawn green), 2-hydroxy-3-methylvalerate (HMVA, dark green), and alpha-hydroxyisovalerate (HIVA, green) (HD4 UPLC-MS/MS, n = 4–8 mice). b BCHA levels in lyophilized yogurt and milk (HD4 UPLC-MS/MS, n = 5 productions). c BCHA content in lyophilized milk and yogurts used in all studies (targeted NMR, n = 5 productions in Study 1 and 2, n = 1 in Study 3, average of 5 replicates). d BCHA content in various dairy products (targeted NMR, n = 1, average of 5 replicates unless indicated otherwise). e Pearson correlations between hepatic BCHA levels and fasting glucose (solid gray line) and hepatic triglycerides (dash gray line). n = 4–8. P-value and adjusted p-value indicated. f BCHA metabolic pathway. BCAA are metabolized by EC 2.6.1.42 to oxo-acids and then by EC 1.2.4.4; EC 1.8.1.4 and EC 2.3.1.168 complex. Circles: metabolites; rectangles: enzymes. Red: p ≤ 0.05 between H1 and Y1; pink: 0.05 < p < 0.10. g–i BCHA content in g plasma, h liver and i muscle (targeted LC‐MS/MS, n = 4–5 in C and 8–9 mice in H, Y groups). Data are expressed as mean ± SEM. a, b Thick black line is median, box spans from Q1 (25th percentile) to Q3 (75th percentile), whiskers extend to the most extreme observation within 1.5 times the interquartile range (Q3–Q1) from the nearest quartile. Data are scaled such that the median value measured across all samples was set to 1.0. Outliers-dots outside the whiskers of the plot. H vs C: *p < 0.05, **p < 0.01, ***p < 0.001. Y vs H: ^#p < 0.05, ^##p < 0.01. Lyophilized milk vs lyophilized yogurts products used in studies: ^&&&p < 0.001. H1 versus C1: *p < 0.05, **p < 0.01, ***p < 0.001. Y1 vs H1: ^#p < 0.05, ^##p < 0.01. Lyophilized yogurt products vs milk: ^§§§p < 0.001. C: low-fat low-sucrose control diet (C1: white); H: high-fat high-sucrose diet with a protein mixture replacing casein (H1: pink); L.: Lactobacillus; Lc.: Lactococcus; ND: non detected; S.: Streptococcus; Y: lyophilized yogurt incorporated in H diet (Y1: light blue). Number (1) refers to the study affiliation. a ANOVA, b, g–i T-tests, e Pearson correlations. Benjamini–Hochberg correction for multiple testing within each BCHA (e) or within tissues (g–i). FDR (q-value) correction for multiple testing over all tested metabolites (a, b). All tests were two-sided.

Fig. 5. BCHA are cell-autonomous modulators of liver and muscle glucose metabolism.

a HGP of FAO cells treated with 0.1 µM, 1 µM, 0.5 mM and 1 mM mixture of HICA: HIVA: HMVA (white) at molar ratio 1: 1: 0.5 in basal condition. b HGP of FAO cells treated with 0.1 µM, 1 µM, 0.5 mM and 1 mM mixture of HICA: HIVA: HMVA at molar ratio 1: 1: 0.5 in insulin-treated condition. c HGP of FAO cells treated with 0.5 and 1 mM of HICA (lawn green), leucine (white, hatched green), or a combination of both (lawn green, hatched green) in basal condition. d HGP of FAO cells treated with 0.5 and 1 mM of HICA, leucine, or a combination of both in insulin-treated conditions. e 2dg uptake of L6 cells treated with 0.1 and 1 µM mixture of HICA: HIVA: HMVA at molar ratio 1:1:0.5 in basal condition. f 2dg uptake of L6 cells treated with 0.1 and 1 µM mixture of HICA: HIVA: HMVA at molar ratio 1:1:0.5 in insulin-treated condition. g 2dg uptake of L6 cells treated with 0.01, 0.1 or 1 µM of HICA in basal condition. h 2dg uptake of L6 cells treated with 0.01, 0.1, or 1 µM of HICA in insulin-treated condition. Data are expressed as mean ± SEM. Vs control: *p < 0.05 versus control, ***p < 0.001. HICA 0.5 mM vs HIVA 1 mM: ^∂∂p < 0.01. Vs Leu 0.5 mM: ^Kp < 0.05, ^KKp < 0.01, ^KKKp < 0.001. Vs Leu 1 mM: ^Ωp < 0.05, ^ΩΩp < 0.01, ^ΩΩΩp < 0.001. The basal and insulin controls are represented by a dash line. n = 8–9 independent experiments. One-way ANOVA followed by Dunnett’s post hoc test vs Control.