Abstract
Disclosure: K. Stefanakis: None. N. Perakakis: None. O. Verrastro: None. M. Garcovich: None. L. Riccardi: None. M. Eslam: None. G.E. Markakis: None. G. Papatheodoridis: None. G. Mingrone: None. J. George: None. C.S. Mantzoros: None. The growing burden of metabolic dysfunction-associated steatotic liver disease (MASLD) and steatohepatitis (MASH), coupled with novel therapies targeting fibrotic MASH (F2–F3), underscores the need for deeper mechanistic insights and improved non-invasive test (NITs). In a multi-national biobank, we enrolled 443 biopsy-proven individuals (81 healthy controls, 362 patients, including 162 with MASH F≥2) across three countries (Italy, Greece, Australia) and two clinical center types (bariatric, gastroenterology-hepatology). A total of 839 metabolites and 840 lipid species were quantified in the serum of all 443 participants by ultra-high-performance liquid chromatography–tandem mass spectrometry. Beyond absolute changes, we employed supervised/unsupervised classifiers, clustering, and pathway enrichment analyses to map metabolites and lipids across the MASLD histological continuum. Early steatosis and MASLD were characterized by a pronounced surge in triglycerides conjugated with 16:0 (palmitic acid), together with elevated ceramides and inflammatory markers. Transition to MASH was marked by elevated BCAAs, bile acid metabolites, sphingolipids, urea cycle intermediates, and xenobiotic compounds (mannonate, hydroxybenzoate), alongside a further spike in triglycerides. In MASH F2–F3, 108 metabolites were markedly upregulated—notably 3-ureidopropionate (uracil pathway) and kynurenine (tryptophan pathway)—and 137 lipid species (chiefly glycerols with saturated fatty acids), whereas 21 metabolites (purines, n-acetylmethionine, steroidogenic intermediates) and 177 lipids (mainly phospholipids and ceramides) were significantly reduced. With advanced fibrosis (F3–F4), the triglyceride surge attenuated, yet inflammation persisted, coupled with dysregulated xenobiotic metabolism and depletion of >10 fibrinopeptide species, indicative of impaired coagulation. Weighted correlation analyses unveiled distinct modules of functionally and structurally similar metabolites and lipids co-varying with transaminases, GDF-15, activins and follistatins in significant fibrosis; separate from obesity-driven steatosis and early inflammation. Building on these comprehensive profiles, we developed categorical gradient-boosting models (4:1 training–validation split, repeated cross-validation with random resampling) using routine clinical variables (e.g., transaminases) and select metabolites such as 3-ureidopropionate, kynurenine, and sphingadienine. These models achieved cross-validated AUCs exceeding 88%—and up to 94%—in the validation cohort, outperforming over 20 established imaging- and biomarker-based NITs for detecting MASH, at-risk MASH, and F≥3. These findings define a novel paradigm for delineating mechanistic pathways that could reveal future therapeutic targets and streamline the development of accurate NITs. Presentation: Saturday, July 12, 2025