MOTIVATION: A bacterial polysaccharide utilization locus (PUL) is a set of physically linked genes that orchestrate the breakdown of a specific glycan. PULs are prevalent in the Bacteroidetes phylum and are key to the digestion of complex carbohydrates, notably by the human gut microbiota. A given Bacteroidetes genome can encode dozens of different PULs whose boundaries and precise gene content are difficult to predict. RESULTS: Here, we present a fully automated approach for PUL prediction using genomic context and domain annotation alone. By combining the detection of a pair of marker genes with operon prediction using intergenic distances, and queries to the carbohydrate-active enzymes database (www.cazy.org), our predictor achieved above 86% accuracy in two Bacteroides species with extensive experimental PUL characterization. AVAILABILITY AND IMPLEMENTATION: PUL predictions in 67 Bacteroidetes genomes from the human gut microbiota and two additional species, from the canine oral sphere and from the environment, are presented in our database accessible at www.cazy.org/PULDB/index.php.

BACKGROUND: Gut microbiome metagenomics has revealed many protein families and domains found largely or exclusively in that environment. Proteins containing the GxGYxYP domain are over-represented in the gut microbiota, and are found in Polysaccharide Utilization Loci in the gut symbiont Bacteroides thetaiotaomicron, suggesting their involvement in polysaccharide metabolism, but little else is known of the function of this domain. RESULTS: Genomic context and domain architecture analyses support a role for the GxGYxYP domain in carbohydrate metabolism. Sparse occurrences in eukaryotes are the result of lateral gene transfer. The structure of the GxGYxYP domain-containing protein encoded by the BT2193 locus reveals two structural domains, the first composed of three divergent repeats with no recognisable homology to previously solved structures, the second a more familiar seven-stranded β/α barrel. Structure-based analyses including conservation mapping localise a presumed functional site to a cleft between the two domains of BT2193. Matching to a catalytic site template from a GH9 cellulase and other analyses point to a putative catalytic triad composed of Glu272, Asp331 and Asp333. CONCLUSIONS: We suggest that GxGYxYP-containing proteins constitute a novel glycoside hydrolase family of as yet unknown specificity.