Over the course of evolution, proteins families undergo sequence diversification via mutation accumulation, with extant homologs often sharing less than 25% sequence identity. The resulting diversity presents a complex view of sequence-structure-function relationships, as epistasis is prevalent, and deleterious mutations in one protein can be tolerated in homologous sequences through networks of intramolecular, compensatory interactions. Understanding these epistatic networks is crucial for understanding and predicting protein function, yet comprehensive analysis of such networks across protein families is limited. In this study, we combine computational and experimental approaches to examine epistatic networks in the class B1 metallo-β-lactamases, a diverse family of antibiotic-degrading enzymes. Using Direct Coupling Analysis, we assess global coevolutionary signatures across the B1 family. We also obtain detailed experimental data from deep mutational scanning on two distant B1 homologs, NDM-1 and VIM-2. There is good agreement between the two approaches, revealing both family-wide and homolog specific patterns that can be associated with 3D structure. However, specific interactions remain complex, and strong epistasis in evolutionarily entrenched residues are not easily compensated for by changes in nearby interactions.

Understanding epistatic networks in the B1 ꞵ-lactamases through coevolutionary statistical modeling and deep mutational scanning