Evolution-guided Engineering of Alpha/Beta Hydrolases

Abstract

This work applies principles from evolution to engineering enzyme properties. Specifically, by examining the phylogeny and evolved sequence diversity in a group of α/β-hydrolase fold enzymes from plants, we are able to engineer proteins with broader chemoselectivity, altered enantioselectivity, and increased stability. A number of ancestral α/β-hydrolases fold proteins were reconstructed in one set of experiments. These were more likely than related modern proteins to have relaxed chemoselectivities and, in one case, was more useful for synthesizing medicinally important molecules. Relative to modern enzymes, ancestral enzymes near functional branch points could catalyze more esterase and hydroxynitrile lyase reactions, as well as a number of other types of reactions: decarboxylation, Michael addition, γ-lactam hydrolysis, and 1,5-diketone hydrolysis. This finding helps to demonstrate the important role that enzyme promiscuity plays in the evolution of new enzymes. Additional experiments and structural analysis on one of these reconstructed ancestral enzymes, the early hydroxynitrile lyase HNL1 found that it is both more thermostable and more promiscuous than its modern relatives, HbHNL and MeHNL. X-ray crystallographic studies revealed, counterintuitively, that larger amino acids in the active site of the ancestor actually increased the size of the substrate binding pocket relative to modern relatives. To take advantage of the promiscuity observed in HNL1, it was used in the asymmetric synthesis of a precursor for the important pharmaceutical propranolol. Another set of experiments altered enantioselectivity by making phylogenetically informed mutations. The active sites from two related hydroxynitrile lyases, HbHNL and AtHNL, were modified to resemble their last common ancestor. This resulted in altered enantioselectivity, and in the case of AtHNL, reversed enantioselectivity. Surprisingly modeling suggested that some of these mutants use a previously undescribed mechanism. This may have been the extinct ancestral mechanism that served as an evolutionary stepping stone that allowed descendant lineages to diverge to either the S-HNL mechanism used by HbHNL, or the R-HNL mechanism used by AtHNL. A final set of experiments used a variety of methods to identify stabilizing mutations in another plant α/β-hydrolase, SABP2. All of the methods were able to identify stabilizing mutations. The most stabilizing mutations were identified by methods that used no structural information. Random mutagenesis identified highly stabilizing mutations, but required screening thousands of mutants. The most efficient approaches were found to be those that used sequence information from either one stable homolog, or the consensus of many homologs, to identify potential stabilizing mutations. Residues that evolution has conserved are often important for stabilizing a protein. We created a software application, Consensus Finder, to automate the process of identifying stabilizing mutations by consensus.