Compensatory mutations restore fitness during the evolution of dihydrofolate reductase

Abstract

Whether a trade-off exists between robustness and evolvability is an important issue for protein evolution. Although traditional viewpoints have assumed that existing functions must be compromised by the evolution of novel activities, recent research has suggested that existing phenotypes can be robust to the evolution of novel protein functions. Enzymes that are targets of antibiotics that are competitive inhibitors must evolve decreased drug affinity while maintaining their function and sustaining growth. Utilizing a transgenic Saccharomyces cerevisiae model expressing the dihydrofolate reductase (DHFR) enzyme from the malarial parasite Plasmodium falciparum, we examine the robustness of growth rate to drug-resistance mutations. We assay the growth rate and resistance of all 48 combinations of 6 DHFR point mutations associated with increased drug resistance in field isolates of the parasite. We observe no consistent relationship between growth rate and resistance phenotypes among the DHFR alleles. The three evolutionary pathways that dominate DHFR evolution show that mutations with increased resistance can compensate for initial declines in growth rate from previously acquired mutations. In other words, resistance mutations that occur later in evolutionary trajectories can compensate for the fitness consequences of earlier mutations. Our results suggest that high levels of resistance may be selected for without necessarily jeopardizing overall fitness.

FIG. 1.

A plot of resistance (IC50) versus growth rate in the absence of pyrimethamine for each of the 29 functional DHFR alleles. No significant correlation exists between these two phenotypes (Pearson’s correlation: P = 0.7512).

FIG. 2.

Ten most frequent evolutionary pathways leading from the wild type to the quadruple mutant using the correlated fixation model. According to our model, evolution of pyrimethamine resistance follows one of these ten pathways nearly 99% of the time. Five digit numbers indicate allelic states at each evolutionary step where each digit corresponds to an amino acid site (from left to right: 16, 51, 59, 108, 164). Wild-type states are depicted as 0, whereas mutant states are depicted as 1 or 2 (site 108: S = 0, N = 1, T = 2). Bold arrows indicate the three most likely pathways and the thickest arrows depicting the most likely path.

FIG. 3.

Probability density function (PDF) for ten evolutionary pathways of greatest frequency. The solid line depicts the PDF based on equal fixation probabilities, whereas the dashed line depicts the PDF from correlated fixation probability. Landscapes were simulated based upon IC50 data in supplementary table S3, Supplementary Material online. Pathways are ranked according to mean frequency. Error bars represent 95% confidence intervals.

FIG. 4.

Growth rates (left) and resistance values (right) of alleles at each step in the three most likely evolutionary pathways. Plots in the same row (A and B, C and D, and E and F) display data from the same trajectory. Alleles at each step in the mutational trajectories are displayed above the plots. Five digit numbers indicate allelic states at each evolutionary step where each digit corresponds to an amino acid site (from left to right: 16, 51, 59, 108, 164). Wild-type states are depicted as 0, whereas mutant states are depicted as 1 or 2 (site 108: S = 0, N = 1, T = 2). Rates represent growth in the absence of drug and are relative to the wild-type growth rate. Error bars represent 95% confidence intervals.