Experimental and In Silico Analysis of TEM β-Lactamase Adaptive Evolution

Abstract

Multiple mutations often have non-additive (epistatic) phenotypic effects. Epistasis is of fundamental biological relevance but is not well understood mechanistically. Adaptive evolution, i.e., the evolution of new biochemical activities, is rich in epistatic interactions. To better understand the principles underlying epistasis during genetic adaptation, we studied the evolution of TEM-1 β-lactamase variants exhibiting cefotaxime resistance. We report the collection of a library of 487 observed evolutionary trajectories for TEM-1 and determine the epistasis status based on cefotaxime resistance phenotype for 206 combinations of 2-3 TEM-1 mutations involving 17 positions under adaptive selective pressure. Gain-of-function (GOF) mutations are gatekeepers for adaptation. To see if GOF phenotypes can be inferred based solely on sequence data, we calculated the enrichment of GOF mutations in the different categories of epistatic pairs. Our results suggest that this is possible because GOF mutations are particularly enriched in sign and reciprocal sign epistasis, which leave a major imprint on the sequence space accessible to evolution. We also used FoldX to explore the relationship between thermodynamic stability and epistasis. We found that mutations in observed evolutionary trajectories tend to destabilize the folded structure of the protein, albeit their cumulative effects are consistently below the protein's free energy of folding. The destabilizing effect is stronger for epistatic pairs, suggesting that modest or local alterations in folding stability can modulate catalysis. Finally, we report a significant relationship between epistasis and the degree to which two protein positions are structurally and dynamically coupled, even in the absence of ligand.

Keywords: antibiotic resistance; contingency; epistasis; evolution; fitness landscape; selection.

Figure 1

Representation of interactions between mutations in the CCED database of clinical and experimental isolates, broken down by type of interaction and number of mutations in the mutant where the pair was observed. The types of interactions are determined experimentally in the first part of this work (Table S2).

Figure 2

Identification of gain-of-function ESBL mutations using competition assays. [TEM-WT GFP^hi] and [TEM-mut GFP^lo] (1:1) co-cultures were grown in 96-well plates in the presence of increasing concentrations of cefotaxime. Similarly, 1:1 [TEM-WT GFP^lo] and [TEM-mut GFP^hi] co-cultures were also grown in the presence of cefotaxime. The fluorescence of both competition experiments at the highest concentration of cefotaxime with observable growth was measured. The bar plot shows the difference in average fluorescence signal between the two cultures for each individual mutant.

Figure 3

Relative frequency of different CCED mutants containing (A) 1, (B) 2, (C) 3, and (D) >3 mutations and their predicted change in the energy of folding relative to the wild type. Distributions of 1000 random mutants were generated as a baseline comparison (orange) using individual mutations observed in the CCED. In (D), n = 4 was used to generate the random distribution. For this analysis, the thermodynamic impact of individual mutations in complex mutants was treated as additive.

Figure 4

Estimated impact of the change in free energy of folding by mutations in TEM-1 on the fraction of folded protein at 298 K.

Figure 5

Relationship between the epistasis status of the pairs of mutations included in this study and their estimated effect on the free energy of folding of TEM-1 β-lactamase. The estimated thermodynamic impact of pairs of mutations was modeled using FoldX assuming it considers additivity of effects on free energy of folding (see Methods). The p-values indicate the significance of an unpaired single-sided t-test between the corresponding type of epistasis and the group with non-significant epistasis.

Figure 6

Modeling of epistasis observations based on protein molecular dynamics (MD) parameters. The magnitude of the epistasis between two mutations is quantified here by the absolute value of its z-statistic, introduced in this work. Each point in each panel corresponds to a pair of alpha carbons in the protein. For the amino acid positions considered (i and j), we evaluate their average distance, ⟨d_i,j⟩, as well the fluctuation of such distance over time, s(d_i,j), by carrying out MD simulations of the wild-type structure (PDB: 1ZG4). (a) Relationship between the distance between two residues over time and extent of epistasis. (b) Relationship between the variation of the distance between two residues over time and the extent of epistasis. (c) Relationship between the average distance between two residues and the variability of this distance over time.