A CAZymes atlas of the human gut microbiota: From 40,000+ genomes to bench validation

Human gut microbiota encode diverse carbohydrate-active enzymes (CAZymes) that are widely implicated in carbohydrate degradation and utilization, yet systematic comparisons across broad taxonomic lineages remain limited. Here, we analyzed a curated resource of 43,972 human gut bacterial genomes and profiled CAZymes repertoires using a standardized comparative genomics pipeline based on the CAZymes database. We identified 475 CAZymes families, including a conserved core of 73 families (15.4%) shared across lineages, suggesting a common enzymatic backbone underlying carbohydrate processing across diverse commensal taxa. Based on differences in CAZymes portfolios, we organized taxa into four functional clusters: M1 (fiber end trimming), M2 (fast carbohydrate utilization), M3 (broad-spectrum carbohydrate utilization), and M4 (oligosaccharide/terminal processing and cross-feeding-enriched). Bifidobacterium, a prominent member of M4, showed marked enrichment of GH30/GH30_9/GH79. Single-substrate growth assays on representative strains revealed utilization tendencies consistent with cluster assignments, and in vitro fecal fermentations further indicated that cluster-associated responses were reproducible within complex communities. Focusing on M4, flavonoid glycosides linked to these CAZymes features preferentially promoted Bifidobacterium growth in a species-dependent manner, highlighting substrate-taxon specificity within the framework. Overall, our results support an integrative gut microbiota-CAZymes-substrate framework that may help characterize carbohydrate metabolic potential, clarify possible contributions of CAZymes portfolios to community-level division of labor and metabolic cooperation, and provide a data-driven basis for hypothesis generation and microbiota-directed carbohydrate interventions with potential relevance to host health. This atlas also offers a reference for prioritizing taxa, enzyme families, and substrates for follow-up validation in continuous-flow and in vivo models.