Circulating metabolite signatures indicate differential gut-liver crosstalk in lean and obese MASLD

Abstract

BACKGROUNDAlterations in circulating metabolites have been described in obese metabolic dysfunction-associated steatotic liver disease (MASLD), but data on lean MASLD are lacking. We investigated serum metabolites, including microbial bile acids and short-chain fatty acids (SCFAs), and their association with lean and obese MASLD.METHODSSerum samples from 204 people of European descent were allocated to groups: lean healthy, lean MASLD, obese healthy, and obese MASLD (n = 47). Liquid chromatography-mass spectrometry-based metabolomics and linear model analysis were performed. MASLD prediction was assessed based on least absolute shrinkage and selection operator regression. Functional effects of altered molecules were verified in organotypic 3D primary human liver cultures.RESULTSLean MASLD was characterized by elevated isobutyrate, methionine sulfoxide, propionate, and phosphatidylcholines. Patients with obese MASLD had increased sarcosine and decreased lysine and asymmetric dimethylarginine. Using metabolites, sex, and BMI, MASLD versus healthy could be predicted with a median AUC of 86.5% and 85.6% in the lean and obese subgroups, respectively. Functional experiments in organotypic 3D primary human liver cultures showed propionate and isobutyrate induced lipid accumulation and altered expression of genes involved in lipid and glucose metabolism.CONCLUSIONLean MASLD is characterized by a distinct metabolite pattern related to amino acid metabolism, lipids, and SCFAs, while metabolic pathways of lipid accumulation are differentially activated by microbial metabolites. We highlight an important role of microbial metabolites in MASLD, with implications for predictive and mechanistic assessment of liver disease across weight categories.FUNDINGRobert Bosch Stiftung, Swedish Research Council (2021-02801, 2023-03015, 2024-03401), ERC Consolidator Grant 3DMASH (101170408), Ruth and Richard Julin Foundation for Gastroenterology (2021-00158), SciLifeLab and Wallenberg National Program for Data-Driven Life Science (WASPDDLS22:006), Novo Nordisk Foundation (NNF23OC0085944, NNF23OC0084420), PMU-FFF (E-18/28/148-FEL).

Keywords: Hepatology; Metabolism; Obesity.

Figure 1. Overview of study cohorts, metabolomics measurements, and statistical analysis.

Lean and obese patients with MASLD and BMI-matched healthy controls were allocated from the Salzburg Colon Cancer Prevention Initiative study (SAKKOPI) to the indicated patient groups (6). The study design and details of the clinical and biochemical workup of included participants have been reported previously (85). MASLD was diagnosed based on ultrasound analysis. Quantitative data from BA and SCFA measurements were merged with previously acquired metabolomics data (6), and the resulting n = 179 metabolic traits were analyzed by linear model analysis in order to detect metabolites that differ between lean and obese MASLD and least absolute shrinkage and selection operator–based (LASSO-based) AUC receiver operating characteristic (ROC) analysis for MASLD versus healthy prediction.

Figure 2. Serum metabolite profiles vary by sex and exhibit distinct metabolic signatures in lean and obese MASLD.

(A) Volcano plot displaying metabolites found at significantly (adjusted P value < 0.05) higher (blue) or lower (red) levels in males compared with females. Statistical analyses were performed in the complete cohort (n = 204) irrespective of MASLD and BMI. Metabolites that exhibit a log₂ fold-change > 0.3125 and an adjusted P value < 0.05 are labeled with the corresponding names. The top 10 significant metabolites with the highest log₂ fold-changes are underlined and shown as box plots in C (see Supplemental Figure 2 for box plots of other significantly altered metabolites). The upper/lower borders of a box are defined by the first/third quartile while the line within a box represents the median. Whiskers extend to the highest or lowest values. (B) Venn diagram representing metabolites that are significantly (adjusted P values < 0.05, Supplemental Tables 2–5) changed for the indicated group comparisons. (D) Volcano plot displaying significantly (adjusted P value < 0.05) upregulated (blue) or downregulated (red) metabolites in lean MASLD compared with BMI-matched healthy controls (n = 110). (E) Volcano plot displaying significantly (adjusted P value < 0.05) upregulated (blue) or downregulated (red) metabolites in obese MASLD compared with BMI-matched healthy controls (n = 94). (F) Volcano plot displaying metabolites found at significantly (adjusted P value < 0.05) higher (blue) or lower (red) levels in obese compared with lean patients with MASLD (n = 96). Metabolites in D–F that exhibit a log₂ fold-change > 0.3125 and an adjusted P value < 0.05 are labeled with the metabolite names (corresponding box plots are displayed in Supplemental Figures 3–7). Data were analyzed by linear model analysis for volcano plot generation.

Figure 3. Assessment of the potential of serum metabolites for steatosis prediction.

MASLD versus healthy prediction was assessed in (A) lean (n = 110) and (B) obese (n = 94) subgroups considering metabolites only (yellow curve), sex + BMI (gray curve), or sex + BMI + metabolites (magenta curve). Displayed ROC curves correspond to respective models with median AUC values (within 100 LASSO repeats). Variable selection and regularization was applied using LASSO regression.

Figure 4. Pro and Ibu induce lipid accumulation in human 3D liver spheroids by altering gene expression in metabolic pathways.

(A) Spheroids (n = 3 donors, with 8–10 individual spheroids per donor) were treated with FFAs or increasing concentrations (i.e., 1×, 10×, and 100× of quantified serum level) of Ibu and Pro for 2 weeks. Triglyceride levels were quantified weeks 1 and 2 and normalized to the control and FFA groups, respectively. Data are presented as box-and-whisker plots with points. Each dot represents an individual spheroid. Statistical analysis was performed by 1-way ANOVA with post hoc Dunnett’s multiple-testing correction against control (Ctrl) of respective week. P values are shown on top of box plots for statistically significant lipid accumulation against Ctrl. (B) Representative bright-field images of the spheroids from the experiment shown in A. Scale bars = 100 μm. (C) FFAs, (D) Ibu, and (E) Pro alter the expression of genes involved in lipid and glucose metabolism, BA/farnesoid X receptor (FXR) signaling, and cytochrome P450 (CYP) enzymes. Data are presented as box-and-whisker plots representing log₂FC of gene expression compared with controls (n = 2 donors, each analyzed in triplicate). The red dotted line denotes a log₂FC threshold of ± 0.5.