Abstract
Xanthan gum is a widely used industrial polysaccharide employed as a thickening and stabilizing agent in food, pharmaceutical, and technological applications. Its biosynthesis involves membrane-associated glycosyltransferases that assemble the repeating unit at the cytoplasmic side of the inner membrane. Among them, GumH and GumI catalyze consecutive reactions using the same donor substrate, guanosine 5'-diphospho-alpha-D-mannose, but with opposite stereoselectivity. Despite their biochemical characterization, structural insights into their catalytic mechanisms and membrane interactions remain limited, hindering a detailed understanding of their function and future engineering efforts. In this work, we combined artificial intelligence-based structure prediction with atomistic molecular dynamics simulations to investigate the structural organization and substrate-binding modes of GumH (family GT4) and GumI (family GT94). The predicted apo structures exhibit a conserved GT-B fold but differ in interdomain flexibility and membrane-anchoring strategies. GumH displays a more structured interdomain linker and a defined clamp-like region in the acceptor-binding domain, consistent with stable membrane interaction, whereas GumI shows a more flexible linker and an open groove architecture. Modeling of the donor-bound complexes reveals distinct substrate-binding modes. In GumH, it adopts a geometry consistent with its retaining stereochemical outcome, positioning the sugar close to the conserved catalytic residue. In contrast, GumI exhibits a different donor orientation, lacking a clearly positioned catalytic base near the reactive center, suggesting a substrate-assisted catalytic mechanism. Although the predicted ternary complexes show limited stability in our simulations, they provide chemically reasonable conformations and offer structural insights into substrate recognition, membrane association, and stereochemical control in these two glycosyltransferase families. SIGNIFICANCE STATEMENT: Xanthan gum is an industrially important polysaccharide widely used in food and other technological products. Although several enzymes in its biosynthetic pathway have been studied, structural information remains limited. Using AI-based structure predictions and molecular simulations, we revealed how these enzymes sit in the membrane and bind sugar substrates. These structural insights clarify xanthan biosynthesis and could help improve or engineer its production.