Selection of artemisinin partial resistance Kelch13 mutations in Uganda in 2016-22 was at a rate comparable to that seen previously in South-East Asia

Abstract

Background: Artemisinin partial resistance, mediated by mutations in the Plasmodium falciparum Kelch13 protein (K13), rapidly spread in South-East Asia (SEA), undermining antimalarial efficacies of artemisinin-based combination therapies (ACT). Validated K13 mutations have recently arisen in Africa, but rates of increase are not well characterized.

Methods: We investigated K13 mutation prevalence at 16 sites in Uganda (2016-2022, 6586 samples), and five sites in SEA (2003-2018, 5465 samples) by calculating selection coefficients using Bayesian mixed-effect linear models. We then tested whether SEA K13 mutation prevalence could have been forecast accurately using up to the first five years of available data and forecast future K13 mutation prevalence in Uganda.

Findings: The selection coefficient for the prevalence of relevant K13 mutations (441L, 469F/Y, 561H, 675V) was estimated at s=0·383 (95% CrI: 0·247 - 0·528) per year, a 38% relative prevalence increase. Selection coefficients across Uganda were s=0·968 (0·463 - 1·569) for 441L, s=0·153 (-0·445 - 0·727) for 469F, s=0·222 (-0·011 - 0·398) for 469Y, and s=0·152 (-0·023 - 0·312) for 675V. In SEA, the selection coefficient was s=-0·005 (-0·852 - 0·814) for 539T, s=0·574 (-0·092 - 1·201) for 580Y, and s=0·308 (0·089 - 0·536) for all validated K13 mutations. Forecast prevalences for Uganda assuming constant selection neared fixation (>95% prevalence) within a decade (2028-2033) for combined K13 mutations.

Interpretation: The selection of K13 mutations in Uganda was at a comparable rate to that observed in SEA, suggesting K13 mutations may continue to increase quickly in Uganda.

Funding: NIH R01AI156267, R01AI075045, and R01AI089674.

Figure 1.. Evolution of ART-R mutation prevalence in Uganda.

(A) Prevalence of the indicated mutations from 2016– 2022. (B) The pie-charts represent the distribution and prevalence of mutations in 2022, with study districts shaded and radii of pie charts proportional to the overall prevalence of the five K13 mutations at a site.

Figure 2.. Prevalence trends of K13 mutations across Ugandan sites and model fits.

Points indicate the observed prevalence of each genotyped mutation, with sample size indicated by point size. Lines represent the posterior median of Bayesian mixed-effects linear models fitted to the weighted, logit-transformed mutation prevalence data. The shaded areas represent the 95% CrI of the model fit. Panels show the prevalence of the indicated mutations (A) and the prevalence of all five mutations (B, gray, including 561H) in districts with at least three years of non-zero mutation prevalence.

Figure 3.. Estimated selection coefficients per year in Uganda and SEA.

Per year point estimates for individual sites (colored circles) or all combined sites (black diamonds) are shown with lines representing the 95% CrI for the indicated mutations in Uganda (A) and SEA (B), including the combined mutations (K13 combined).

Figure 4.. Forecasting ART-R prevalence in SEA and Uganda.

Based on the first five years of non-zero prevalence, the prevalence was forecast for all 13 validated K13 markers (A) and 580Y, the most common mutation, in SEA (B). The forecasting was also conducted for all K13 mutations combined (C), 675V (D), 441L (E), and 469F (F) in Uganda. The shaded regions highlight the 95% CrI for the forecast selection.