Fitness Costs and the Rapid Spread of kelch13-C580Y Substitutions Conferring Artemisinin Resistance

Abstract

Fitness costs are key determinants of whether drug resistance alleles establish and how fast they spread within populations. More than 125 different kelch13 alleles, each containing a different amino acid substitution, have arisen in Southeast Asian malaria parasite (Plasmodium falciparum) populations under artemisinin selection over the past 15 years in a dramatic example of a soft selective event. However, just one of these alleles (C580Y) is now outcompeting other alleles in multiple different countries and is spreading toward fixation. Here we examine the fitness consequences of C580Y, relative to another less successful kelch13 mutation (R561H), to try to explain the distinctive dynamics of C580Y. We hypothesized that C580Y will show lower fitness costs than other kelch13 substitutions in the absence of artemisinin treatment. We used CRISPR/Cas9 methods to introduce single mutations (C580Y or R561H) or synonymous control edits into a wild-type parasite isolated on the Thailand-Myanmar border, conducted replicated head-to-head competition assays, and determined the outcome of competition using deep sequencing of kelch13 amplicons. Contrary to our predictions, these experiments reveal that C580Y carries higher fitness costs (s [selection coefficient] = 0.15 ± 0.008 [1 standard error {SE}]) than R561H (s = 0.084 ± 0.005). Furthermore, R561H outcompetes C580Y in direct competition (s = 0.065 ± 0.004). We conclude that fitness costs of C580Y in isolation are unlikely to explain the rapid spread of this substitution.

Keywords: CRISPR/Cas9; amplicon sequencing; fitness costs; selection coefficient.

FIG 1

Study design. (A) We generated 4 different edited parasites from a single ART-S parasite clone (NHP4302). Two of the edited parasites (gray shading) have edited bases conferring ART-R, while two control edits (white shading) have synonymous shield mutations only. The amino acids present in edited parasites are shown in red. The black two-headed arrows show the four head-to-head competition experiments conducted. (B) Overview of amplicon sequencing for parasite competition experiments. (C) Sequences used for assigning reads to different competing parasite lines. Raw reads were chopped and aligned by using anchor sequences at each end of the CRISPR-edited region. We then used the mutations present to determine the frequencies of the competing parasites. (D and E) Observed and expected allele frequencies of NHP4302^R561H (D) and NHP4302^C580Y (E) from artificial mixtures.

FIG 2

Head-to-head competition experiments. (Top) Trajectory of competing parasites observed in six replicate experiments (each color represents one replicate). (Bottom) Natural log of the parasite ratio against time between competing parasites. The slope for the least-squares fit provides an estimate of the selection coefficient (s).

FIG 3

Quantifying the outcomes of head-to-head competition. The plot shows selection coefficients (s) with 95% confidence intervals (CI) from 4 sets of competition experiments. Six replicate competition experiments were conducted for each set: gray points show s for each replicate, while red points show meta-analysis results for each experimental comparison. The “winning” parasite is shown on the right for each comparison.