Dietary fibre-adapted gut microbiome clears dietary fructose and reverses hepatic steatosis

Abstract

Excessive consumption of the simple sugar fructose, which induces excessive hepatic lipogenesis and gut dysbiosis, is a risk factor for cardiometabolic diseases. Here we show in male mice that the gut microbiome, when adapted to dietary fibre inulin, catabolizes dietary fructose and mitigates or reverses insulin resistance, hepatic steatosis and fibrosis. Specifically, inulin supplementation, without affecting the host's small intestinal fructose catabolism, promotes the small intestinal microbiome to break down incoming fructose, thereby decreasing hepatic lipogenesis and fructose spillover to the colonic microbiome. Inulin also activates hepatic de novo serine synthesis and cystine uptake, augmenting glutathione production and protecting the liver from fructose-induced lipid peroxidation. These multi-modal effects of inulin are transmittable by the gut microbiome, where Bacteroides acidifaciens acts as a key player. Thus, the gut microbiome, adapted to use inulin (a fructose polymer), efficiently catabolizes dietary monomeric fructose, thereby protecting the host. These findings provide a mechanism for how fibre can facilitate the gut microbiome to mitigate the host's exposure to harmful nutrients and disease progression.

Fig. 1. Reversal of HFCS-induced metabolic dysfunctions by inulin supplementation.

a, Chemical structures of HFCS and inulin. G, glucose; F, fructose. b, Experimental groups. Mice received a control (C) or inulin-supplemented (I) diet with or without HFCS (F) in drinking water. For the CIF group, mice first received a control diet with HFCS and then an inulin diet with HFCS from week 16. c, Representative liver H&E staining and quantitation of lipid accumulation. Scale bars, 200 μm (n = 4, 4 mice). d,e, Body weight (d) and fat mass (e) (n = 8, 8, 8, 8, 9 mice). f–h, Fasting insulin on weeks 14 and 26 (f) (n = 8, 7, 8 mice), fasting glucose on week 26 (g) (n = 8, 8, 9 mice) and HOMA-IR on week 26 (h) (n = 7, 7, 8 mice). i, Representative liver H&E staining and quantitation of lipid accumulation. Staining was performed in three mice per group, and lipid accumulation was quantified in four randomly selected areas per liver. Scale bars, 50 μm. j, Liver lipidomics (n = 7, 8, 9 mice). k, Abundances of the indicated hepatic lipid species normalized to the CF group (n = 7, 8, 9 mice). Cer, ceramide; SM, sphingomyelin; DG, diacylglycerol; TG, triacylglycerol. Numbers in brackets denote total number of carbon atoms and number of double bonds. l, Liver fibrosis marker gene expression (n = 8, 8, 7, 7, 9 mice). Data are means; error bars, s.e.m. P values determined by one-way ANOVA with Tukey’s honestly significant difference (HSD) test (e–l) or two-sided unpaired Student’s t-test (c). Illustrations in a and b created with BioRender.com. Source data

Fig. 2. Inulin supplementation suppresses hepatic lipogenesis and increases FAO.

a, Schematic of lipogenesis measurements using ²H₂O tracing. i.p., intraperitoneal b, ²H-labelled fatty acids in circulating lipids on week 14 (n = 9, 16, 9 mice). c, ²H-labelled fatty acids in circulating lipids before and after switching diets for the CIF group. Data are fold changes relative to CF (n = 9, 9 mice). d, DNL rate measured by ²H₂O tracing after normalization to body water enrichment (n = 12, 9, 10 mice). See Methods for more details. e, ²H-labelled palmitate concentration normalized to body water enrichment (n = 12, 10, 10 mice). f, Fructose catabolism and lipogenesis pathways. GA, glyceraldehyde; GA3P, glyceraldehyde-3-phosphate. PEP, phosphoenolpyruvate. g, Liver fructose catabolism gene expression (n = 8, 7, 8, 9 mice). h, Liver lipogenesis gene expression (n = 8, 7, 8, 9 mice). i, FAO pathway. j, Schematic of ¹³C-fructose tracing and hepatic cytosol and mitochondria fractionation. k, ¹³C-labelled C16:0 carnitine levels in cytosol versus mitochondria fraction of liver (n = 4, 4 mice). l, Schematic of whole-body fatty acid oxidation measurements using ¹³C-acetate and ¹³C-palmitate tracing. m, The ratio of ¹³CO₂ to unlabelled CO₂ over time after ¹³C-acetate administration (n = 8, 7 mice). n, The ratio of ¹³CO₂ to unlabelled CO₂ over time (left) and slope up to the maximum point (right) after ¹³C-palmitate administration (n = 8, 7 mice). o, ¹³CO₂ from ¹³C-palmitate after normalization to circulating ¹³C-palmitate (n = 8, 7 mice). Data are means; error bars, s.e.m. P values determined by one-way ANOVA with Tukey’s HSD test (b,d–h), two-way ANOVA with Tukey’s HSD test (m,o) or two-sided unpaired Student’s t-test (c,k,n). NS, not significant. Illustrations in a, j, and l created with BioRender.com. Source data

Fig. 3. Inulin-fed small intestinal microbiome suppresses dietary fructose spillover.

a, Schematic of dietary fructose catabolism by the host organs and gut microbiome. The small intestine first catabolizes fructose, and the leftover fructose spills over to the liver or colon and induces lipogenesis and gut dysbiosis. b, ¹³C-fructose levels in various intestinal contents 30 min after oral provision of HFCS with ¹³C-labelled fructose (n = 8, 8, 9 mice). c, ¹³C-labelled SCFAs in caecal contents (n = 8, 8, 8 mice). d, Comparison of labelled metabolite abundances in caecal contents between CF versus IF (left) (n = 8, 8 mice) or CF versus CIF (right) (n = 8, 8 mice). e, ¹³C-labelled metabolites in jejunal contents (n = 7, 7, 7 mice). f, Dietary intervention groups. XIF, antibiotics-treated group. g, Faecal 16S rDNA copy number (n = 3, 3, 3, 6 mice). h, ¹³C-labelled circulating saponified fatty acids normalized to hepatic ¹³C-acetyl CoA fraction, 1 h after provision of HFCS with ¹³C-labelled fructose (n = 7, 8, 8, 6 mice). i, Schematic of small intestinal microbiome transplantation experiments from donors (CF or IF) to recipients (CF after antibiotics). j,k, ¹³C-labelled acetate fraction in jejunal content (j) and liver lipogenesis gene expression (k) in recipient mice, 1 h after provision of HFCS with ¹³C-labelled fructose (n = 9, 9 mice). l, Schematic of faecal transplantation experiments. m,n, ¹³C-labelled butyrate in faeces (m) and ¹³C-labelled circulating saponified fatty acids (n) in recipient mice 1 h after provision of HFCS with ¹³C-labelled fructose (n = 8, 8 mice). Data are means; error bars, s.e.m. P values determined by one-way ANOVA with Tukey’s HSD test (b,c,e,g,h), two-sided unpaired Student’s t-test (d) or one-sided unpaired Student’s t-test (j,k,m,n). Illustrations in a, i and j created with BioRender.com. Source data

Fig. 4. Inulin rewires hepatic fructose carbon use toward serine and GSH biosynthesis.

a, Schematic of ¹³C-fructose tracing and untargeted metabolomics in liver. b, Comparison of hepatic ¹³C-labelling (%) of metabolites between IF and CF, 30 min after oral provision of HFCS with ¹³C-labelled fructose, by Student’s t-test followed by false discovery rate correction. Different colours indicate metabolite categories. CE, cholesteryl ester; FA, fatty acid; MG, monoacylglycerol (n = 8, 8 mice). c,d, ¹³C-labelled abundances (c) and labelling fractions (d) of serine and glycine in liver (n = 8, 8, 9 mice). M, isotopologue. e, Correlation coefficient (r) and P values between ¹³C-labelled serine ion count and the indicated labelled metabolite ion counts. f, ¹³C-labelled fractions of the indicated metabolites in liver. GSSG, oxidized GSH (n = 8, 8, 9 mice). g, Comparison of hepatic gene expression between CF and IF (left) (n = 5, 5 mice) or CF and CIF (right) (n = 5, 4 mice). h, Comparison of ¹³C-labelled ion counts of the indicated metabolites between CF and IF in liver, 30 min after oral provision of ¹³C-Cys. *P < 0.05, **P < 0.01 (n = 8, 9 mice). FC, fold change. i, Immunofluorescence staining and quantitation of liver 4-HNE, a lipid peroxidation marker. Scale bars, 20 μm (n = 6, 6 mice). j, Hepatic malondialdehyde levels measured by TBARS assay (n = 8, 9 mice). k, Schematic of the GSH synthesis pathway. Red arrows indicate that metabolites and Slc7a11 gene (which encodes xCT) are significantly upregulated in IF compared to CF. Hcy, homocysteine; Cth, cystathionine. Data are means; error bars, s.e.m. P values determined by one-way ANOVA with Tukey’s HSD test (c), Student’s t-test followed by false discovery rate correction (b,g) or two-sided unpaired Student’s t-test (h–j). Illustrations in a and h were created with BioRender.com. Source data

Fig. 5. Inulin induces liver serine synthesis via gut microbiome.

a, Experimental groups including the antibiotics-treated group (XIF). b,c, ¹³C-labelled abundances of serine and glycine in serum on week 4 (b) and liver on week 16 (c), 1 h after oral provision of HFCS with ¹³C-labelled fructose (n = 7, 8, 7, 6 mice). d, Serine synthesis gene expression in liver (n = 7, 8, 7, 6 mice). e, Schematic of jejunal microbiome transplantation experiments from donors (CF or IF) to recipients (CF after antibiotics). Abx, antibiotics. f, ¹³C-labelled serine abundances in liver of recipient mice (n = 8, 8 mice). Data are means; error bars, s.e.m. P values were determined by one-way ANOVA with Tukey’s HSD test (b–d) or one-sided unpaired Student’s t-test (f). Illustrations in a and e created with BioRender.com. Source data

Fig. 6. B. acidifaciens contributes to inulin’s effects on lipogenesis suppression and fructose catabolism in the small intestine.

a,b, Linear discriminant analysis effect size analysis of jejunal (a) and caecal (b) microbial taxa. The cladogram shows the taxa with significant differences in abundance (from phylum to genus level) (n = 8, 8, 8 mice). c–f, Pearson correlation analysis between bacterial abundances and hepatic lipogenesis or serine synthesis. g, Relative abundance of Bacteroides spp. in jejunal contents (n = 6, 7 mice) and B. acidifaciens in caecal contents (n = 8, 8, 8 mice). h, Schematic of single-bacteria inoculation experiments. Recipient mice were fed HFCS alone for 4 weeks, followed by antibiotic treatment for 1 week. Anaerobically cultured bacteria were orally delivered to the recipient mice every day for 2 weeks while the mice received HFCS with inulin to promote bacteria survival and inulin usage. i, ¹³C-labelled circulating saponified fatty acids normalized to hepatic ¹³C-acetyl CoA fraction in the recipient mice, 1 h after oral provision of HFCS with ¹³C-labelled fructose (n = 12, 12, 12 mice). j, ¹³C-labelled concentrations (left two) and carbon fractions (right two) of the indicated SCFAs in the jejunal contents of the recipient mice, 1 h after oral provision of HFCS with ¹³C-labelled fructose (n = 12, 11, 12 mice). k, Proposed model of inulin’s multi-modal effects on stimulating small intestinal microbial breakdown of dietary fructose, reducing fructose spillover to colon, reducing hepatic lipogenesis and augmenting hepatic serine and GSH synthesis. Data are means; error bars, s.e.m. P values determined by one-way ANOVA with Tukey’s HSD test (g, right panel, i,j) or by one-sided unpaired Student’s t-test (g, left panel). Illustrations in k were created with BioRender.com. Source data

Extended Data Fig. 1. Metabolic characterization of mice fed HFCS, inulin or combinations.

a-c, Daily food, water, and calorie intake (n = 3,3,3,3,3 cages). d, Lean mass normalized to body weight (n = 8,9,8,8,9 mice). e-i, Locomotor activity, heat production, O₂ consumption, CO₂ production, and respiratory exchange ratio (n = 8,9,8,8,9 mice). j, Hepatic mtDNA contents (n = 5,7,7 mice). k, Representative liver trichrome staining for fibrosis measurements from two independent experiments. Scale bars, 50 μm (n = 3,3,3,3,3 mice). l, body water enrichment measured 15 hours after administration of ²H₂O. m, Circulating total saponified palmitate concentration (n = 12,9,10 mice). n, PCA plot of metabolites in cytosol and mitochondria fractions of liver (n = 4,4 mice). o, Glucose-6-phosphate and α-ketoglutarate levels in cytosolic vs mitochondrial fractions of liver (n = 4,4 mice). p, ¹³C-labeled circulating palmitate fraction over time after oral provision of ¹³C-palmitate (n = 4,3 mice). Data are mean±s.e.m. P-values by one-way ANOVA with Tukey’s HSD test. Source data

Extended Data Fig. 2. Inulin does not affect fructose catabolism by the host small intestine.

a, ¹³C-labeled F1P abundances in small intestine, 30 min after oral provision of HFCS with fructose ¹³C-labeled (n = 8,8,9 mice). b, Small intestinal fructose transporter and catabolism gene expression (n = 8,8,9 mice). c, Glucose levels in caecal content, 30 min after oral provision of HFCS (n = 8,8,9 mice). d-e, Total and ¹³C-labeled SCFAs concentrations in portal blood, 1 h after oral provision of HFCS with fructose ¹³C-labeled (n = 7,7,7 mice, left), (n = 12,8,12 mice, right). Data are mean±s.e.m. P-values by one-way ANOVA with Tukey’s HSD test. ns=not significant. Source data

Extended Data Fig. 3. Antibiotics reverse inulin’s effects.

a, Body weight (n = 7,8,7,6 mice). b-d, Daily food, water, and calorie intake (n = 3,3,3,2 cages). e, ¹³C-labeled acetyl CoA in liver (n = 7,8,7,6 mice), 1 hr after provision of HFCS with fructose ¹³C-labeled in mice fed the indicated diet for 16 weeks (n = 7,8,8,6 mice). f, ¹³C-labeled circulating saponified fatty acids. g, Abundances of the indicated hepatic lipid species normalized to the C group (n = 8,7,6 mice), h-i, Liver lipogenesis (h) and fibrosis marker (i) gene expression (n = 7,8,7,6 mice). j, NMDS plot showing microbial diversity among donor and recipient mice of jejunal microbiome transplant experiment (n = 5-7 mice/group). Data are mean±s.e.m. P-values determined by one-way ANOVA with Tukey’s HSD test. Source data

Extended Data Fig. 4. Inulin activates hepatic serine and GSH synthesis.

a, Schematic of serine synthesis pathway from ¹³C-fructose. Red circles indicate labeled carbons. b-c, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of IF(b) and CIF(c) compared CF (n = 5,5,4 mice for CF, IF and CIF groups). P-values by Fisher’s exact test. d, Hepatic malondialdehyde abundances measured by chemical derivatization and LC-MS measurement (n = 8,9 mice). e, Immunofluorescence staining and quantitation of liver dihydroethidium to measure ROS. Each dot indicates a mean of quadruplicate values from two independent experiments. Scale bars, 20 μm (n = 6,6 mice). f, ¹³C-labeled serine in liver of intestine-specific Khk-C transgenic mice, 1 hr after provision of HFCS with fructose ¹³C-labeled (n = 3, 3 mice). Data are mean±s.e.m. P-values by two-sided unpaired Student’s t-test. Source data

Extended Data Fig. 5. Characterization of gut microbiome in mice fed HFCS with or without inulin.

a, Total bacterial amounts, Alpha-diversity, and Firmicutes to Bacteroidetes ratio in jejunum (n = 6,5,7 mice). b, Total bacterial amounts, Alpha-diversity, and Firmicutes to Bacteroidetes ratio in caecum (n = 8,8,8 mice). c, Relative abundance of B. pseudolongum in jejunal (n = 6,5,7 mice) and caecal content (n = 8,8,8 mice). d, ¹³C-labeled acetyl CoA enrichment (%) in liver, 1 hr after provision of HFCS with fructose ¹³C-labeled (n = 12,12,12 mice). e, ¹³C-labeled circulating saponified fatty acids (n = 12,12,12 mice). f, Relative abundances of the indicated hepatic lipid species (n = 12,12 mice). g, Gene expression of fibrosis markers in liver (n = 12,12 mice). h, Hepatic ¹³C-labeled serine and glycine abundances in mice that received vehicle or the indicated bacteria species, 1 hr after provision of HFCS with fructose ¹³C-labeled (n = 12,12,12 mice). i, Growth changes of B. acidifaciens in response to the addition of inulin or glucose (n = 3). j, Changes in ¹³C-fructose consumption and labeled short-chain fatty acid production of B. acidifaciens over incubation time (n = 3). Data are mean±s.e.m. P-values by one-way ANOVA with Tukey’s HSD test (a-e, k) or two-way ANOVA with Tukey’s HSD test (i-j). Source data