Coevolution Network Models Predict the Impact of Multiple Mutations on Protein Function
Beleva Guthrie, Violeta
MetadataShow full item record
Proteins often evolve new functions by acquiring a small number of mutations in an ancestral sequence not containing the phenotype. Modeling the functional effect of a mutation is, however, a nontrivial task, due to strong functional interdependencies. Here, I used the recent evolution of the bacterial enzyme TEM β-lactamase under antibiotic selection as a model for genetic adaptation. I compiled a database of TEM β-lactamase sequences evolved under antibiotic resistance selective pressure and identified functional interactions between individual mutations/mutated residues. I built network models of coevolving residues (possible functional interactions), in which nodes are mutations and edges represent coevolution between two mutations. I reconstructed both the alignment and phylogeny-based mutation coevolution networks and assessed the utility of network-theoretical tools to derive information regarding role of individual mutations in the observed resistance. Coevolution network} analysis reveals key properties of mutations in evolution of antibiotic resistance, many of which were confirmed through extensive fitness measurements in the lab and by previous experimental studies of TEM β-lactamase function. One finding is that mutations form densely connected clusters in the network corresponding to selection to different main classes of antibiotics or to different adaptive strategies within the same antibiotic class. Mutations that are central in the network tend to be either adaptive or compensate for effects of many other mutations. By extending node centrality metrics to paths of mutations (connected nodes in the network) I was able to study properties of adaptive evolutionary trajectories in TEM. I found that central paths are enriched in non-negative functional interactions. Specifically, paths corresponding to triple mutants were experimentally shown to increase fitness from all or most of their constitutive single and double mutants. It was also shown that relative rankings of central paths and their constituent shorter paths can be used to predict the direction of fitness change in an evolutionary trajectory. In this way, this predictor of the effect of an evolutionary trajectory can be useful in anticipating evolution of antibiotic resistance. In summary, my analysis of the combined functional effects of mutations in producing new biological activities should help anticipate evolution driven by a variety of clinically-relevant selections such as drug resistance, virulence, and immunity.