Knowledge-based homology modeling and experimental determination of amino acid side chain accessibility by the laser photo CIDNP (chemically induced dynamic nuclear polarization) approach in solution: lessons from the small sialidase of Clostridium perfringens

Abstract

The success of knowledge-based homology modelling is critically dependent on the predictive potency of the program structure-based calculations, which attempt to translate homologous sequences into three-dimensional structures, and on the actual relevance of the crystal structure for the protein topology. As quality control, experimental data for selected parameters of the proteins conformation are required. Using the crystal structure of the sialidase of Salmonella typhimurium as framework for model building of the homologous enzyme from Clostridium perfringens, a set of energy-minimised conformers is derived. These proteins present e.g. Tyr, Trp and His residues with an assessable area on the surface, since the side chains of these amino acid residues are responsive to chemically induced dynamic nuclear polarization (CIDNP), monitored by NMR. Hence, as first lesson, a comparative analysis for model-derived and experimentally determined values can be performed. The second lesson of this study concerns the notable impact of single amino acid substitutions (Tyr/Phe, Cys/Ser) on the surface accessibility of the CIDNP-reactive amino acid side chains in mutant forms of the sialidase. Corroborating the predictions from the theoretical calculations, the spectra of the engineered mutants reveal marked and non-uniform alterations. Thus, the effect of apparently rather conservative amino acid substitutions on a distinct conformational aspect of this protein, even at distant sites, should not be underestimated.