Search

Carbohydrates play a key role in a variety of physiological and pathological processes and, hence, represent a rich source for the development of novel therapeutic agents. Being able to predict binding mode and binding affinity is an essential, yet lacking, aspect of the structure-based design of carbohydrate-based ligands. We assembled a diverse data set comprising 273 carbohydrate-protein crystal structures with known binding affinity and evaluated the prediction accuracy of a large collection of well-established scoring and free-energy functions, as well as combinations thereof. Unfortunately, the tested functions were not capable of reproducing binding affinities in the studied complexes. To simplify the complex free-energy surface of carbohydrate-protein systems, we classified the studied proteins according to the topology and solvent exposure of the carbohydrate-binding site into five distinct categories. A free-energy model based on the proposed classification scheme reproduced binding affinities in the carbohydrate data set with an r 2 of 0.71 and root-mean-squared-error of 1.25kcal/mol (N=236). The improvement in model performance underlines the significance of the differences in the local micro-environments of carbohydrate-binding sites and demonstrates the usefulness of calibrating free-energy functions individually according to binding-site topology and solvent exposure. Graphical Abstract

The most prevalent diseases manifested by Escherichia coli are acute and recurrent bladder infections and chronic inflammatory bowel diseases such as Crohn's disease. E. coli clinical isolates express the FimH adhesin, which consists of a mannose-specific lectin domain connected via a pilin domain to the tip of type 1 pili. Although the isolated FimH lectin domain has affinities in the nanomolar range for all high-mannosidic glycans, differentiation between these glycans is based on their capacity to form predominantly hydrophobic interactions within the tyrosine gate at the entrance to the binding pocket. In this study, novel crystal structures of tyrosine-gate mutants of FimH, ligand-free or in complex with heptyl α-d-O-mannopyranoside or 4-biphenyl α-d-O-mannopyranoside, are combined with quantum-mechanical calculations and molecular-dynamics simulations. In the Y48A FimH crystal structure, a large increase in the dynamics of the alkyl chain of heptyl α-d-O-mannopyranoside attempts to compensate for the absence of the aromatic ring; however, the highly energetic and stringent mannose-binding pocket of wild-type FimH is largely maintained. The Y137A mutation, on the other hand, is the most detrimental to FimH affinity and specificity: (i) in the absence of ligand the FimH C-terminal residue Thr158 intrudes into the mannose-binding pocket and (ii) ethylenediaminetetraacetic acid interacts strongly with Glu50, Thr53 and Asn136, in spite of multiple dialysis and purification steps. Upon mutation, pre-ligand-binding relaxation of the backbone dihedral angles at position 137 in the tyrosine gate and their coupling to Tyr48 via the interiorly located Ile52 form the basis of the loss of affinity of the FimH adhesin in the Y137A mutant., SCOPUS: ar.j, info:eu-repo/semantics/published

Carbohydrates play a key role in a variety of physiological and pathological processes and, hence, represent a rich source for the development of novel therapeutic agents. Being able to predict binding mode and binding affinity is an essential, yet lacking, aspect of the structure-based design of carbohydrate-based ligands. We assembled a diverse data set comprising 273 carbohydrate-protein crystal structures with known binding affinity and evaluated the prediction accuracy of a large collection of well-established scoring and free-energy functions, as well as combinations thereof. Unfortunately, the tested functions were not capable of reproducing binding affinities in the studied complexes. To simplify the complex free-energy surface of carbohydrate-protein systems, we classified the studied proteins according to the topology and solvent exposure of the carbohydrate-binding site into five distinct categories. A free-energy model based on the proposed classification scheme reproduced binding affinities in the carbohydrate data set with an r 2 of 0.71 and root-mean-squared-error of 1.25kcal/mol (N=236). The improvement in model performance underlines the significance of the differences in the local micro-environments of carbohydrate-binding sites and demonstrates the usefulness of calibrating free-energy functions individually according to binding-site topology and solvent exposure. Graphical Abstract