Mutatomics analysis of the systematic thermostability profile of Bacillus subtilis lipase A

Abstract

Use of point mutagenesis technique to improve protein thermostability is a routine strategy in the protein engineering community and directed evolution approach has been widely utilized to fulfill this. However, directed evolution often does not assure a minimalist design for obtaining a desired property in proteins, and other traditional methods such as error-prone PCR and iterative saturation mutagenesis are also too time-consuming and expensive to carry out a systemic search for protein mutation space. In the current study, we performed mutatomics analysis of the systematic thermostability profile of Bacillus subtilis lipase A (LipA) using a virtual scanning strategy. In the procedure, a new characterization method was proposed to describe structural variations upon protein residue mutation, and the generated descriptors were then statistically correlated with protein thermostability change associated with the mutation based on a large panel of structure-solved, melting temperature-known protein mutation data. As a result, linear and nonlinear quantitative structure-thermostability relationship (QSTR) models were built and their statistical quality was verified rigorously through internal cross-validation and external blind test. It is suggested that the nonlinear support vector machine (SVM) performed much better than linear partial least squares (PLS) regression in correlating protein structure and thermostability information. Thus, the SVM model was employed to systematically scan the complete mutation profile of LipA protein, resulting in a 181×19 matrix that characterizes the change in theoretical thermostability of LipA due to the mutation of wild-type residue at each of the 181 sequence sites to other 19 amino acid types. From the profile most mutations were predicted to (i) destabilize LipA structure and (ii) address modest effect on LipA thermostability. Satisfactorily, several known thermostable mutations such as G80V, G111D, M134D, and N161Y were identified properly and, expectedly, a number of mutations including L55Y, A75V, and S162P that have never been reported previously were inferred as hotspot mutations that have high potential to enhance LipA thermostability. The structural basis and energetic property of the five promising mutations were further examined in detail using atomistic molecular dynamics (MD) simulations and molecular mechanics Poisson-Boltzmann/surface area (MM-PB/SA) analysis, revealing intensive nonbonded interaction networks created by these mutations.