Understanding the global rise of artemisinin resistance: Insights from over 100,000 Plasmodium falciparum samples

Abstract

Artemisinin partial resistance (ART-R) in Plasmodium falciparum is a major challenge to malaria control globally. Over the last two decades, ART-R has spread widely across Southeast Asia, undermining public health strategies and hindering elimination. As of 2024, ART-R has now emerged in East Africa, with the potential to dramatically impact current efforts to control malaria in the region. Mitigating its spread requires detailed genomic surveillance of point mutations in the kelch13 gene, the primary known determinant of artemisinin resistance. Although extensive surveillance data on these markers is available, it is distributed across many literature studies and open databases. In this review, we aggregate spatiotemporal data for 112,933 P. falciparum samples collected between 1980 and 2023 into a single resource, providing the most comprehensive overview of kelch13 markers to date. We outline the history and current status of these mutations globally, with particular focus on their emergence in Southeast Asia and East/Northeast Africa. Concerningly, we find the recent increases in ART-R in Africa mirror patterns observed in Southeast Asia 10-15 years ago. We examine factors that may influence its spread, including fitness costs, treatment strategies, and local epidemiological dynamics, before discussing potential scenarios for how resistance may spread in Africa in coming years. This review provides a comprehensive account of how the situation of ART-R has unfolded globally so far, highlighting insights for researchers and public health bodies which aim to reduce its negative effects.

Keywords: Plasmodium falciparum; artemisinin; artemisinin combination therapy; epidemiology; genetics; genomic surveillance; genomics; global health; malaria; plasmodium falciparum; resistance.

Figure 1.. Global trends in kelch13 propeller mutations over time.

Panel A shows the distribution of kelch13 mutations globally for samples collected between 1980 and 2023. Countries from which samples were included are coloured according to their geographic population assignment. All populations are labelled with a fraction, denoting the number of samples with any kelch13 propeller mutation (excluding A578S, which is known to be artemisinin susceptible) over the total number of samples from each population. No WHO mutations were detected in the North African samples. For each population, pie charts show the proportion of samples with a WHO validated/candidate mutation in red, the proportion of any other propeller domain mutation in blue, and the proportion of samples with a 3D7 reference kelch13 sequence or A578S mutation in white. Proportions of each category at 25% or more are labelled. Panel B shows the number of unique kelch13 propeller domain markers for each population against their total number of unique WHO validated/candidate mutations. Populations are labelled with their ratio of WHO validated/candidate markers to the total number of unique propeller domain markers. Populations falling in the blue shaded area have fewer WHO validated/candidate markers relative to their total count of kelch13 mutations than populations falling in red areas. The upper plot of panel C shows the total number of samples over time. Grey areas denote years with fewer than 25 samples. The bottom panel shows the proportion of samples from each population for each year, with years with no samples coloured white. Note the limited number of samples from Southeast Asia after 2019, which coincided with an increased proportion of resistant samples in East and Northeast Africa. Panel D illustrates a schematic representation of the kelch13 protein (residues 1–726), with functional domains—including N-terminal, coiled-coil, BTB/POZ, and six-blade kelch13 repeat propeller (KREP) domains—annotated using InterProScan (version 5.73). WHO-validated (green) and candidate markers (blue) are mapped to their respective amino acid positions.

Figure 2.. Cumulative count of unique WHO validated/candidate kelch13 mutations over time across the whole dataset (grey), Southeast Asia (blue) and Africa (red).

Figure 2—figure supplement 1.. Number of unique WHO validated/candidate kelch13 mutations and other propeller mutations over time.

(A) Global overview of unique mutations (B) Unique mutations over time split by population. Darker hue = WHO-validated/candidate mutations, lighter hue = other propeller domain mutations. Bars represent stacked mutation counts. North Africa is not shown as no mutations were reported there.

Figure 3.. Proportions of kelch13 mutations over time from 2000 to 2024 for ten populations, with African populations in the left-hand column.

Years with fewer than 25 samples are shaded in grey with white stripes. Three populations (North Africa, Western Asia, Far-Eastern South Asia) are not displayed due to lack of data. Several of the most common WHO validated/candidate markers are highlighted. Other WHO validated/candidate markers are denoted by ‘Other WHO-V/C’. Mutations in the kelch13 BTB/POZ or propeller domain without validated/candidate status are categorised as ‘Other’. The 3D7 reference kelch13 sequence and the A578S mutation are known to not confer ART-R, so are categorised together as ‘Susceptible’. The high percentage of ‘Other’ mutations in East Africa in 2000 is due to our aggregation of samples from 1996 to 2003 to the median year of 2000 from a single study by de Laurent et al., 2018. Similarly, the percentages of ‘Other’ mutations in West Africa (2005) and Central Africa (2007) result from single studies (Ouattara et al., 2015; Taylor et al., 2015) where multiple different mutations were identified.

Figure 4.. Proportion of samples from each country and geographical population with a WHO validated/candidate kelch13 mutation (red) or other kelch13 mutation in the BTB/POZ or propeller domain (blue).

Bars show the weighted average percentage of samples from the three most recent observed years with at least 25 samples. Averages were weighted by the total number of samples observed between the years. Date ranges show the range of years for the three most recent years, with at least 25 samples each. Countries with fewer than 3 years with at least 25 samples are marked with an asterisk.

Figure 5.. Regional prevalence and temporal changes of kelch13 propeller mutations in Southeast Asia.

Panel (A) shows the distribution of kelch13 propeller mutations across Southeast Asia. The percentage of samples with any kelch13 propeller mutation are included in the country labels, above the number of samples with any kelch13 propeller mutation and the total number of samples collected for each country, respectively. Among only the samples with any observed kelch13 propeller mutation, pie charts show the proportions with markers of interest, where proportions above 25% are labelled accordingly. ‘Other’ denotes low-frequency mutations in the propeller domain which are not WHO validated/candidate markers, aggregated into a single category. Samples with the 3D7 reference sequence for kelch13 or A578S are denoted as ‘Susceptible’ (World Health Organization, 2022b). Panel (B) shows all samples collected over time for each country as a stacked bar chart, where the proportion of samples with each marker is coloured. Years with fewer than 25 samples are highlighted with grey dashed lines.

Figure 5—figure supplement 1.. Scatter plot showing the results of TES monitoring three commonly used ACTs across populations, using the data from the WHO Malaria Threats Map (https://apps.who.int/malaria/maps/threats/).

The x-axis represents delayed clearance, while the y-axis represents treatment failure rates. Scattered data points are colour-coded by region. Red dashed lines indicate the 10% threshold established by the WHO for policy change.

Figure 6.. Regional prevalence and temporal changes of kelch13 propeller mutations in East and Northeast Africa.

Panel (A) shows the distribution of kelch13 propeller mutations across East and Northeast Africa. The percentage of samples with any kelch13 propeller mutation is included in the country labels, above the number of samples with any kelch13 propeller mutation and the total number of samples collected for each country, respectively. Among only the samples with any observed kelch13 propeller mutation, pie charts show the proportions with markers of interest, where proportions above 25% are labelled accordingly. ‘Other’ denotes low-frequency mutations in the propeller domain which are not WHO validated/candidate markers, pooled into a single category. Samples with the 3D7 reference sequence for kelch13 or A578S are denoted as ‘Susceptible’ (World Health Organization, 2022b). Panel (B) shows all samples collected over time for each country as a stacked bar chart, where the proportion of samples with each marker is coloured. Years with fewer than 25 samples are highlighted with grey dashed lines. Notably, Uganda and Rwanda have shown a consistent increase in diversity among kelch13 mutations between 2016–2023. (C) shows all mutations with at least 25 samples across Africa and Southeast Asia on the x axis. Coloured bars show which subpopulation these samples came from across Africa and Southeast Asia. WHO validated or candidate mutations are suffixed with an asterisk.

Figure 7.. Distribution of kelch13 propeller domain markers across West and Central Africa (A), South America (B) and Oceania (C).

The percentage of samples with any kelch13 mutation is included in the country labels, above the number of samples with any kelch13 mutation, and the total number of samples collected for each country, respectively. Among only the samples with any observed kelch13 propeller domain mutation, pie charts show the proportions with markers of interest, where proportions above 25% are labelled accordingly. ‘Other’ denotes several low-frequency mutations in the propeller domain which are not WHO validated/candidate markers, pooled into a single category.

Figure 8.. Proportion of samples with a WHO validated/candidate kelch13 mutations over time (Table 2).

Each country is shown using a logistic regression curve fitted to the observed data (see Figure 8—figure supplement 1 for model fits). Timepoints with no line indicate years with no available data. The black dashed line highlights 2006 - the year in which the WHO recommended ACT therapy as a first-line treatment measure in Africa (World Health Organization, 2006). Notably, the rapid increase in prevalence of WHO kelch13 markers in East and Northeast African countries is similar to that observed in Southeast Asian countries between 2007 and 2016. However, it is also important to note these trends have so far only been observed in a few countries in East/Northeast Africa and are based on smaller sample sizes (Assefa et al., 2024). While these curves indicate the frequency of resistance has increased, we have deliberately avoided quantitative extrapolation beyond the most recent observed time range. This analysis should therefore not be interpreted as a predictive model.

Figure 8—figure supplement 1.. Percentage of samples with WHO validated/candidate kelch13 propeller mutations over time.

The observed data for each country is shown as a dashed line, coloured by population. The black line is the curve of a logistic model fitted to the observed data, which is forced to begin at 0 and end at 100%. The size of each point represents the number of samples collected in that country for each year. Time points with no line or value indicate years with no available data. Countries with fewer than 3 years with at least 25 samples each were excluded. Countries which showed no clear increase during this time were also excluded (Mali, Guyana, India, Ethiopia, China, and Equatorial Guinea). The grey dashed line highlights the year in which the fitted curve reached its steepest point - that is the year with the fastest increase in frequency. The goodness-of-fit of the logistic curves varied substantially across countries (R²=0.22–1.00). Notably, Tanzania had a very high growth rate, although this was based on a single year where the frequency was above 2%. To assess the robustness of the descriptive fits, we performed sensitivity analyses using bootstrap resampling within country-year groupings and a jackknife procedure where each individual year was excluded and the curve refit accomplished. These consistently reproduced the overall pattern of rising ART-R frequencies. While exact slopes and inflection points varied, the trend of increased ART-R prevalence in East and Northeast Africa remained robust. We emphasise that these are descriptive summaries of observed data and are not intended as predictive models.

Box 2—figure 1.. Evolutionary scenarios and within-host dynamics under cycling, MFT and TACT strategies.

(A) There are three possible evolutionary scenarios during cycling, MFT or TACT strategies. In the idealised cycling/MFT scenario (top panel), resistance will cycle between two states of elevated resistance to a given ACT combination therapy, but not both, at any given time (red/blue points). However, these strategies could promote dual resistant escape mutants through stepwise increases over time (middle panel). This may occur even if resistance to any individual ACT decreases in the short term (red and blue lines). In contrast, triple artemisinin-based combination therapies (bottom panel) may directly select for triple resistant escape mutants (black solid line). While the likelihood of this occurring may be lower than under cycling or MFT strategies, this could result in mutants which are resistant to multiple partner drugs, leaving limited options to change strategy image adapted from Baym et al., 2016. (B) TACTs involve the use of two partner drugs throughout treatment. Here, a theoretical example of parasite density over time within a patient is shown for artemisinin-resistant and -susceptible parasites (red and blue lines respectively). In cases where parasites are resistant to artemisinin, their density remains higher after initial artemisinin exposure (~3 days), resulting in more parasites for the partner drug to treat (shaded area), and therefore stronger selection for resistance to that drug. TACTs use two partner drugs (black dashed line), which in theory means selection for any individual drug is lower, and the chance of triply resistant parasites emerging within any given infection is less likely Bushman et al., 2018 image adapted from Hanboonkunupakarn and White, 2022.