Discovery of Aminoacyl-tRNA Synthetase Mutants for the Incorporation of Non Canonical Amino Acids into Proteins

Abstract

Efficient in vivo incorporation of a noncanonical amino acid into proteins often requires engineering new aminoacyl-tRNA synthetase (AARS) activity into the cell, usually by modifying a natural aaRS. Although experimental methods, relying on mutagenesis and library screening, have identified many successful mutant aaRS-substrate pairs in the recent years, computational approaches have reported only a few successes. Here we compare the results of computational and experimental screens of an E. coli methionyl-tRNA synthetase (MetRS) saturation-mutagenesis library for binding (in silico), and activation and cell-surface display (in vivo) of azidonorleucine (Anl). Three positions (L13, Y260, and H301) in the methionine binding pocket of E. coli MetRS were randomized, and the resulting library was screened for MetRS activity toward Anl, based on a screening strategy previously established in our group. This strategy relies on the introduction of reactive side chains into surface-exposed sites on outer-membrane protein C (OmpC), and their subsequent labeling with reactive, fluorescent probes. We have discovered a large diversity of MetRS mutants that allow the incorporation of Anl into proteins in vivo. The extent of OmpC expression and the amount of available Anl during the screen have substantial effects on the outcome of the screens. In addition to displaying improved activities toward Anl, identified mutants also show an improved discrimination against Met. We have shown that the degree of cell-surface labeling in vivo correlates well with the measured rates of Anl activation in vitro, which reflects the success of the screen design. Computational analysis of the experimentally identified mutants revealed a good agreement between computed binding energies and in vitro activation data. To better test the computation model, we performed an in silico screen for Anl binding on a saturation-mutagenesis library comparable to the experimental library. Computational screen predominantly selects mutants that interact with Anl through hydrogen bonds, whereas the hydrophobic residues are selected more often by the experimental screen. We identify that experimental mutants try to optimize packing at the Y260 and H301 sites, but not at the L13 site. We discuss possible explanations for these results. Combined results from computation and experiments suggest the importance of various factors in ligand recognition and in vivo selection of MetRS mutants. We explore the implications of these factors to the future efforts in the engineering of new MetRS activities.