Gut microbial and metabolic biomarkers for early-onset colorectal cancer

Abstract

Early-onset colorectal cancer (EO-CRC) is a growing global health challenge with poorly understood etiology. While gut microbiota and metabolites are implicated in colorectal carcinogenesis, EO-CRC-associated signatures remain poorly characterized. This study presents the first large-scale multi-omics analysis in Japan cohort to characterize gut microbial and metabolic features of EO-CRC, with cross-cohort validation in an independent China cohort. We enrolled 510 participants (EO-CRC n = 120, late-onset CRC [LO-CRC] n = 282, young healthy controls [yHC] n = 59, old healthy controls [oHC] n = 49); after age- and sex-matching, 324 samples were analyzed using metagenomics, metabolomics, and quantitative polymerase chain reaction (qPCR). Fourteen species formed a shared CRC-associated microbiome core enriched in both EO-CRC and LO-CRC relative to age-matched controls, of which seven were reproducible across both Japan and China cohorts. Beyond this core, 34 and 21 species were enriched exclusively in EO-CRC and LO-CRC, respectively. Among EO-CRC-specific virulence findings, the bft gene of enterotoxigenic Bacteroides fragilis , fadA of Fusobacterium nucleatum , and bai operon genes were enriched in EO-CRC, with cross-cohort validation for fadA and bai operon, and stage-specific enrichement of deoxycholic acid at Stage 0 providing preliminary metabolomic support. Targeted clbP qPCR revealed enrichment of colibactin-producing Escherichia coli specifically in EO-CRC against a background of age-increasing carriage. Metabolomic profiling identified kynurenine enrichment and depletion of vitamin B-related metabolites (pyridoxine, riboflavin, and nicotinamide) as EO-CRC-specific findings; LO-CRC showed distinct enrichment of carnitine, succinate, and spermine with relative butyrate depletion. The choline degradation pathway was enriched specifically in EO-CRC based on Enteropathway analysis. Species-based random forest classifiers distinguished EO-CRC from controls (AUPRC = 0.828; SD = 0.013; 95% CI 0.739–0.909) and generalized robustly to an independent cohort (AUPRC = 0.817; 95% CI 0.717–0.894), substantially outperforming functional gene profiles in cross-cohort settings. Within the Japan cohort, adding metabolomics provided modest additional discriminative value. These findings identify a shared CRC microbiome core reproducible across age groups and cohorts, alongside EO-CRC-specific signatures encompassing enrichment of virulence-associated genes ( bft, fadA ), colibactin-producing E. coli ( clbP ), secondary bile acid biosynthetic genes ( bai operon), and choline degradation capacity, as well as distinct metabolic perturbations including kynurenine enrichment and vitamin B-related metabolite depletion. Species-based classifiers generalize well across age groups and cohorts, with metabolomics providing complementary discriminatory value. All findings represent cross-sectional associations; longitudinal studies are needed to establish temporal ordering and evaluate clinical utility as non-invasive biomarkers for EO-CRC. Video Abstract