No abstract

The CRISPR-Cas system holds great promise in the treatment of diseases caused by genetic variations. As wildtype SpyCas9 is known to generate many off-target effects, its use in the clinic remains an elusive goal. To date, several high-fidelity Cas9 variants with improved specificity over wildtype SpyCas9 have been described. Each of the variants was engineered either by rational design or experimental directed evolution. However, the potential of improved specificity using both methods is constrained to specific residues due to the requirement of a priori knowledge and challenges in selecting engineered proteins, which arise from factors such as selection pressure, assay limitations, host organism compatibility, and library size and diversity. Therefore, SpyCas9 modifications through in-silico protein engineering allow an efficient specificity improvement while overcoming those limitations. Nevertheless, the lack of living models to test the evolved protein remains a major challenge of computational protein engineering methods. We recently demonstrated the advantage of normal mode analysis to simulate and predict the enzymatic function of SpyCas9 in the presence of mismatches. Here, we report ComPE, a novel computational protein engineering method to modify the protein and measure the vibrational entropy of wildtype or variant SpyCas9 in complex with its sgRNA and target DNA. Using this platform, we discovered novel high-fidelity Cas9 variants with improved specificity. We functionally validated the improved specificity of four variants, and the intact on-target activity in one of them. Lastly, we demonstrate their reduced off-target editing and non-specific gRNA-independent DNA damage, highlighting their advantages for clinical applications. The described method could be applied to a wide range of proteins, from CRISPR-Cas orthologs to distinct proteins in any field where engineered proteins can improve biological processes.