key: cord-0306401-3574lfaf
authors: Stokes, Barbara H.; Rubiano, Kelly; Dhingra, Satish K.; Mok, Sachel; Straimer, Judith; Gnädig, Nina F.; Bath, Jade R.; Deni, Ioanna; Ward, Kurt E.; Striepen, Josefine; Yeo, Tomas; Ross, Leila S.; Legrand, Eric; Ariey, Frédéric; Cunningham, Clark H.; Souleymane, Issa M.; Gansané, Adama; Nzoumbou-Boko, Romaric; Ndayikunda, Claudette; Kabanywanyi, Abdunoor M.; Uwimana, Aline; Smith, Samuel J.; Kolley, Olimatou; Ndounga, Mathieu; Warsame, Marian; Leang, Rithea; Nosten, François; Anderson, Timothy J.C.; Rosenthal, Philip J.; Ménard, Didier; Fidock, David A.
title: P. falciparum K13 mutations present varying degrees of artemisinin resistance and reduced fitness in African parasites
date: 2021-01-27
journal: bioRxiv
DOI: 10.1101/2021.01.27.428390
sha: 12d75da74829f723fe563462647c55c1267e045f
doc_id: 306401
cord_uid: 3574lfaf

The emergence of artemisinin (ART) resistance in Plasmodium falciparum parasites, driven by K13 mutations, has led to widespread antimalarial treatment failure in Southeast Asia. In Africa, our genotyping of 3,299 isolates confirms the emergence of the K13 R561H variant in Rwanda and reveals the continuing dominance of wild-type K13 across 11 countries. We show that this mutation, along with M579I and C580Y, confers varying degrees of in vitro ART resistance in African parasites. C580Y and M579I cause substantial fitness costs, which may counter-select against their dissemination in high-transmission settings. We also define the impact of multiple K13 mutations on ART resistance and fitness in multiple Southeast Asian strains. ART susceptibility is unaltered upon editing point mutations in ferrodoxin or mdr2, earlier resistance markers. These data point to the lack of an evident biological barrier to mutant K13 mediating ART resistance in Africa, while identifying their detrimental impact on parasite growth.

Despite recent advances in chemotherapeutics, diagnostics and vector control measures, malaria continues to exert a significant impact on human health. In 2019, cases were estimated at 229 million, resulting in 409,000 fatal outcomes, primarily in Sub-Saharan Africa as a result of Plasmodium falciparum infection (WHO, 2020) . This situation is predicted to rapidly worsen as a result of the ongoing SARS-CoV-2 pandemic that has crippled malaria treatment and prevention measures (Sherrard-Smith et al., 2020) . Absent an effective vaccine, malaria control and elimination strategies are critically reliant on the continued clinical efficacy of first-line artemisinin-based combination therapies (ACTs). These ACTs pair fastacting artemisinin (ART) derivatives with partner drugs such as lumefantrine, amodiaquine, mefloquine or piperaquine (PPQ). ART derivatives can reduce the biomass of drug-sensitive parasites by up to 10,000-fold within 48 h (the duration of one intra-erythrocytic developmental cycle); however, these derivatives are rapidly metabolized in vivo. Longer-lasting albeit slower-acting partner drugs are co-administered to reduce the selective pressure for ART resistance and to clear residual parasitemias. Nonetheless, P. falciparum partial resistance to ART derivatives has spread throughout Southeast Asia (SEA), having first emerged a decade ago in western Cambodia (Dondorp et al., 2009; Noedl et al., 2009; Ariey et al., 2014; Imwong et al., 2020) . Clinically, partial ART resistance manifests as delayed clearance of circulating asexual blood stage parasites following treatment with an ACT. The accepted threshold for resistance is a parasite clearance half-life (the time required for the peripheral blood parasite density to decrease by 50%) of >5.5 h. Sensitive parasites are typically cleared in <2-3 h (WHO, 2019) . Partial resistance can also be evidenced as parasite-positive blood smears on day three post initiation of treatment. 4 In vitro, ART resistance manifests as increased survival of cultured ring-stage parasites exposed to a 6 h pulse of 700 nM dihydroartemisinin (DHA, the active metabolite of all ARTs used clinically) in the ring-stage survival assay (RSA) (Witkowski et al., 2013; Ariey et al., 2014) . Recently, ART-resistant strains have also acquired resistance to PPQ, which is widely used in SEA as a partner drug in combination with DHA. Failure rates following DHA-PPQ treatment now exceed 50% in parts of Cambodia, Thailand and Vietnam (van der Pluijm et al., 2019) .

In vitro selections, supported by clinical epidemiological data, have demonstrated that ART resistance is primarily determined by mutations in the beta-propeller domain of the P.

falciparum Kelch protein K13 (Ariey et al., 2014; Ashley et al., 2014; MalariaGEN, 2016; Siddiqui et al., 2020) . Recent evidence suggests that these mutations result in reduced endocytosis of host-derived hemoglobin and thereby decreased release of the ART-activating moiety Fe 2+ -heme, thus reducing ART potency (Yang et al., 2019; Birnbaum et al., 2020) . Mutations in other genes including ferredoxin (fd) and multidrug resistance protein 2 (mdr2) have also been associated with ART resistance in K13 mutant parasites, suggesting that they either contribute to a multigenic basis of resistance or fitness or serve as genetic markers of founder populations (Miotto et al., 2015) .

In SEA, the most prevalent K13 mutation is C580Y, which associates with delayed clearance in vivo (Ariey et al., 2014; Ashley et al., 2014; MalariaGEN, 2016; Imwong et al., 2017) . This mutation also mediates ART resistance in vitro, as demonstrated by RSAs on gene-edited parasites (Ghorbal et al., 2014; Straimer et al., 2015; Uwimana et al., 2020) .

Other studies have since demonstrated the emergence of almost 200 K13 mutations both in 5 The K13 R561H, M579I and C580Y mutations can confer in vitro artemisinin resistance in African parasites

To test whether R561H can mediate ART resistance in African strains, we developed a CRISPR/Cas9-mediated K13 editing strategy (Supplementary file 1) to introduce this mutation into 3D7 and F32 parasites. On the basis of whole-genome sequence analysis of African isolates, 3D7 was recently shown to segregate phylogenetically with parasites from Rwanda (Ariey et al., 2014; Uwimana et al., 2020) . F32 was derived from an isolate from Tanzania (Witkowski et al., 2010) . We also tested the C580Y mutation that predominates in SEA, as well as the M579I mutation identified in a P. falciparum-infected individual in Equatorial Guinea who displayed delayed parasite clearance following ACT treatment (Lu et al., 2017) . The positions of these residues are highlighted in the K13 beta-propeller domain structure shown in Supplementary file 2. For 3D7, F32 and other lines used herein, the geographic origins and genotypes at drug resistance loci are described in Table 1 and Table   1 -table supplement 1. All parental lines were cloned by limiting dilution prior to transfection. Edited parasites were identified by PCR and Sanger sequencing, and cloned.

RSAs, used to measure in vitro ART susceptibility, revealed a wide range of mean survival values for K13 mutant lines. For 3D7 parasites, the highest RSA survival rates were observed with 3D7 R561H parasites, which averaged 6.6% RSA survival. For the 3D7 M579I and 3D7 C580Y lines, mean RSA survival rates were both 4.8%, a 3 to 4-fold increase relative to the 3D7 WT line. No elevated RSA survival was seen in a 3D7 control line (3D7 ctrl ) that expressed only the silent shield mutations used at the guide RNA cut site (Figure 2A; Figure 2 -source data 1). 8 In contrast to results with 3D7, the introduction of K13 mutations into F32 WT parasites yielded almost no increase in RSA survival. Mean RSA survival rates were 0.5%, 0.5% and 0.3% for F32 R561H , F32 M579I and F32 C580Y parasites, respectively, compared to 0.3% for F32 WT ( Figure   2B ). Previously we reported that introduction of M476I into F32 parasites resulted in a modest gain of resistance (mean survival of 1.7%) while this same mutation conferred RSA survival levels of ~10% in edited Dd2 parasites (Straimer et al., 2015) . These data suggest that while K13 mutations differ substantially in the level of resistance that they impart, there is an equally notable contribution of the parasite genetic background.

We next introduced M579I and C580Y into cloned Ugandan isolates UG659 and UG815.

Editing of both mutations into UG659 yielded moderate RSA survival rates (means of 6.3% and 4.7% for UG659 M579I or UG659 C580Y respectively, vs. 1.0% for UG659 WT ; Figure 2C ).

These values resembled our results with 3D7. Strikingly, introducing K13 M579I or C580Y into UG815 yielded the highest rates of in vitro resistance, with mean survival levels reaching ~12% in both UG815 M579I and UG815 C580Y . These results were confirmed in a second independent clone of UG815 M579I (Figure 2D ). M579I and C580Y also conferred equivalent levels of resistance in edited Dd2 parasites (RSA survival rates of 4.0% and 4.7%, respectively; Figure 2-source data 1). These data show that mutant K13-mediated ART resistance in African parasites can be achieved, in some but not all strains, at levels comparable to or above those seen in Southeast Asian parasites.

The K13 C580Y and M579I mutations are associated with an in vitro fitness defect in

To examine the relation between resistance and fitness in African parasites harboring K13 mutations, we developed an in vitro fitness assay that uses quantitative real-time PCR (qPCR) for allelic discrimination. Assays were conducted with the eight pairs of K13 WT and either C580Y or M579I isogenic parasites used for RSAs, namely 3D7 WT + either 3D7 M579I or 3D7 C580Y ; F32 WT + either F32 M579I or F32 C580Y ; UG659 WT + either UG659 M579I or UG659 C580Y ; and UG815 WT + either UG815 M579I or UG915 C580Y .

Assays were initiated with tightly synchronized trophozoites, mixed in 1:1 ratios of WT to mutant isogenic parasites, and cultures were maintained over a period of 40 days (~20 generations of asexual blood stage growth). Cultures were sampled every four days for genomic DNA (gDNA) preparation and qPCR analysis. TaqMan probes specific to the K13 WT or mutant (M579I or C580Y) alleles were used to quantify the proportion of each allele.

Results showed that both K13 mutations (M579I or C580Y) conferred a significant fitness defect in all backgrounds tested, with the percentage of the K13 mutant allele declining over time in all co-cultures (Figure 3; Figure 3 -source data 1). This fitness defect varied between parasite backgrounds. To quantify this impact, we calculated the fitness cost, which represents the percent reduction in growth rate per 48 h generation of the test line compared to the WT isogenic comparator. The fitness cost was calculated using day 0 and day 32 as start and end points, respectively, as these yielded the most consistent slopes across lines and time series. For 3D7 parasites, the fitness cost was 8.9% and 6.9% for both the M579I and C580Y mutations, respectively ( Figure 3A) . In F32 and UG659 parasites, the fitness cost for the M579I mutation was slightly higher (4.3% and 5.8%) than for C580Y (2.8% and 1.9%; Figure 3B , C). The largest growth defects for both mutations were seen in the UG815 line ( Figure 3D) , with fitness cost values for the M579I and C580Y mutations both being 12.0%.

A comparison of data across these four African strains revealed that high RSA survival rates were generally accompanied by high fitness costs, with M579I mostly having a more detrimental fitness impact than C580Y (Figure 3E, 3F) .

We next examined the spatiotemporal distribution of K13 alleles in Cambodia, the epicenter of ART resistance in SEA. In total, we sequenced the K13 propeller domains of 3,327 parasite isolates collected from western, northern, eastern and southern provinces of Cambodia nonsynonymous polymorphisms in K13 were identified across all regions and years. Of these, only three were present in >10 samples, Y493H (n=83), R539T (n=87) and C580Y (n=1,915).

Each of these mutations was previously shown to confer ART resistance in vitro (Straimer et al., 2015) . Rarer mutations included A418V, I543T, P553L, R561H, P574L, and D584V (Figure 4) . In western Cambodia, where ART resistance first emerged 11 (Dondorp et al., 2009; Noedl et al., 2009) , the WT allele percentage in 2001-2002 had already fallen to 56%. This is striking given that delayed parasite clearance following ACT or artesunate treatment was first documented in 2008 -2009 (Noedl et al., 2008 Noedl et al., 2009 ).

In all four regions, the frequency of the WT allele declined substantially over time and the diversity of mutant alleles contracted, with nearly all WT and non-K13 C580Y mutant parasites being replaced by parasites harboring the C580Y mutation (Figure 4) . This effect was particularly pronounced in the west and the south, where the prevalence of C580Y in 2016-17 effectively attained 100%, increasing from 22% and 58% respectively in the initial sample sets (Figure 4A, D) . In northern and eastern Cambodia, C580Y also outcompeted all other mutant alleles; however, 19-25% of parasites remained K13 WT in 2016-17 ( Figure 4B, C) .

These data show the very rapid dissemination of K13 C580Y across Cambodia.

Given that most K13 polymorphisms present in the field have yet to be characterized in vitro, we selected a panel of mutations to test by gene editing, namely E252Q, F446I, P553L, R561H and P574L (Supplementary file 2) . The F446I mutation is the predominant mutation in the Myanmar-China border region. P553L, R561H and P574L have each been shown to have multiple independent origins throughout SEA , and were identified at low frequencies in our sequencing study in Cambodia (Figure 4) . Lastly, the E252Q mutation was formerly prevalent on the Thai-Myanmar border, and, despite its occurrence 12 upstream of the beta-propeller domain, has been associated with delayed parasite clearance in vivo (Anderson et al., 2017; Cerqueira et al., 2017; Group, 2019 Cam3.II R539T parental line that expresses the R539T mutation (~20% mean survival; Figure   5B ; Figure 5 -source data 1). The E252Q mutation did not result in elevated RSA survival in the Cam3.II background, a result also observed with Dd2. Nonetheless, ART resistance was apparent upon introducing the R561H mutation into Cam3.II parasites, whose mean survival rates exceeded the Cam3.II C580Y line (13.2% vs 10.0%, respectively). No elevated survival was seen in the Cam3.II ctrl line expressing only the silent shield mutations used at the guide RNA cut site.

Prior studies with isogenic gene-edited Southeast Asian lines have shown that certain K13 mutations can exert fitness costs, as demonstrated by reduced intra-erythrocytic asexual blood stage parasite growth (Straimer et al., 2017; Nair et al., 2018) . To determine the fitness impact of the K13 mutations described above, we utilized an eGFP-based parasite competitive growth assay (Ross et al., 2018) . Dd2 E252Q , Dd2 F446I , Dd2 P553L , Dd2 R561H or Dd2 P574L were co-cultured in 1:1 mixtures with an isogenic K13 WT eGFP + Dd2 reporter line for 20 days (10 generations), and the proportion of eGFP + parasites was assessed every two days. As controls, we included Dd2 WT , Dd2 bsm and Dd2 C580Y . These data provided evidence of a minimal impact with the F446I, P553L and C580Y mutations, with E252Q, R561H and . Both C580Y and P553L displayed elevated RSA survival and minimal fitness cost in the Dd2 strain, providing optimal traits for dissemination ( Figure 5D ). We note that all fitness costs in Dd2 were considerably lower than those observed in our four African strains ( Figure 3 ).

Given the earlier abundance of the R561H and E252Q alleles in border regions of Thailand and Myanmar, we next tested the impact of introducing these mutations into five Thai K13

WT isolates (Thai1-5). For comparison, we also edited C580Y into several isolates. These studies revealed a major contribution of the parasite genetic background in dictating the level of mutant K13-mediated ART resistance, as exemplified by the C580Y lines whose mean survival rates ranged from 2.1% to 15.4%. Trends observed for individual mutations were maintained across strains, with the R561H mutation consistently yielding moderate to high levels of in vitro resistance at or above the level of C580Y. Consistent with results for Dd2, introduction of E252Q did not result in significant increases in survival rates relative to We also profiled two unedited culture-adapted Thai isolates (Thai6 E252Q and Thai7 E252Q ) that express the K13 E252Q mutation, but that are otherwise K13 WT. Notably, both lines exhibited mean RSA survival rates significantly above the 1% threshold for ART sensitivity (2.7% for Thai6 E252Q and 5.1% for Thai7 E252Q ; Figure 6F ). These data suggest that additional genetic factors present in these two Thai isolates are required for E252Q to manifest ART resistance.

Mutations in the P. falciparum multidrug resistance protein 2 and ferredoxin genes do not modulate resistance to artemisinin in vitro In a prior genome-wide association study of SE Asian parasites, K13-mediated ART resistance was associated with D193Y and T484I mutations in the ferredoxin (fd) and multidrug resistance protein 2 (mdr2) genes, respectively (Miotto et al., 2015) . To directly test the role of these mutations, we applied CRISPR/Cas9 editing to the Cambodian strains RF7 C580Y and Cam3.II C580Y , which both express K13 C580Y (Supplementary file 3) . Isogenic 

Mutant K13-mediated ART resistance has substantially compromised the efficacy of antimalarial treatments across SEA, and the relatively high prevalence of the R561H variant associated with delayed clearance in Rwanda now poses a threat to high-transmission settings in sub-Saharan Africa (Conrad and Rosenthal, 2019; Hanboonkunupakarn and White, 2020; Uwimana et al., 2020; Bergmann et al., 2021) . Using gene editing and phenotypic analyses, we provide definitive evidence that the K13 R561H, M579I and C580Y mutations can confer in vitro ART resistance in several African strains. In vitro resistance, as defined using the RSA, was comparable between gene-edited African K13 R561H 3D7 parasites and Asian C580Y Dd2 and Cam3.II parasites. We also observed that K13 mutant African strains differed widely in their RSA survival rates. As an example, when introduced into the Tanzanian F32 and Ugandan UG815 strains, the C580Y mutation yielded 0.3% (not resistant) and 11.8% (highly resistant) RSA survival rates, respectively. These data suggest that F32 parasites lack additional genetic determinants that are required for mutant K13 to confer ART resistance. Nonetheless, our results provide conclusive evidence that multiple African strains present no core biological obstacle to becoming ART resistant upon acquiring K13 mutations.

Our spatio-temporal analysis of K13 sequence diversity in Cambodia highlights the emergence of C580Y in western Cambodia and its progressive replacement of other variants (Imwong et al., 2020) . The success of this mutation in SEA cannot be explained by resistance alone, as we previously reported that the less common R539T and I543T variants conferred greater ART resistance in vitro (Straimer et al., 2015) . Similarly, we now report that R561H and P553L yield equivalent degrees of ART resistance when compared with C580Y. Lower levels of resistance were observed with F446I and P574L, with the former predominating recently on the Thai-Myanmar border (Imwong et al., 2020) . In a recent study, F446I yielded no significant in vitro resistance in 3D7 parasites, although similar to our data this mutation was fitness-neutral (Siddiqui et al., 2020) . Of note, all four of these mutations, and others including C469Y, R622I and A675V, have now been detected in Africa and merit gene editing experiments in African strains (Warsame et al., 2019; Asua et al., 2020; Kayiba et al., 2020) .

Here we observed that the C580Y mutation exerts less of a fitness cost relative to other K13 variants, as measured in K13-edited Dd2 parasites co-cultured with an eGFP reporter line.

These data suggest that C580Y might be favored in part because of a less detrimental impact on asexual blood stage growth rates. The most detrimental impact on growth was observed with E252Q, which earlier predominated near the Thailand-Myanmar border but was later overtaken by C580Y, as well as R561H, which progressively disappeared over time in SEA (Phyo et al., 2016) . In our study C580Y produced an optimal combination of no measurable fitness cost and relatively high RSA survival rates in Dd2 parasites. R561H, however, showed slightly improved fitness relative to C580Y in paired isogenic parasites from Thailand (the NHP4302 strain) (Nair et al., 2018) , providing evidence that both fitness and resistance are strain-dependent.

Consistent with these findings, we observed substantial fitness costs with the K13 C580Y mutation in four African strains. The largest growth defect was observed with the edited UG815 C580Y line that also yielded the highest level of ART resistance. These data suggest that K13 C580Y may not easily take hold in Africa where, unlike in SEA, infections are often highly polyclonal, generating intra-host competition that impacts a strain's ability to succeed at the population level. In addition, individuals in highly-endemic African settings generally have high levels of acquired immunity, potentially preventing infection by relatively unfit parasites, and often have asymptomatic infections that go untreated and are thus less subject to selective drug pressure, compared with individuals in SEA. This situation recalls the history of chloroquine use in Africa, where fitness costs caused by mutations in the primary resistance determinant PfCRT resulted in the rapid resurgence of wild-type parasites following the implementation of other first-line antimalarial therapies (Kublin et al., 2003; Laufer et al., 2006; Ord et al., 2007; Frosch et al., 2014) . It remains to be determined whether mutations such as R561H, emerging in Rwanda, can ameliorate the fitness cost observed with other K13 variants in African strains.

Further research is also required to define secondary genetic determinants that could augment mutant K13-mediated ART resistance and to explore other potential mediators of resistance. The latter include mutations in AP-2µ, UBP-1 and Pfcoronin, which can modulate P. falciparum ART susceptibility in vitro and merit further investigation (Demas et al., 2018; Henrici et al., 2019; Sutherland et al., 2020) . Data provided herein argue against a direct role for mutations in fd and mdr2, earlier associated with mutant K13-mediated resistance in SEA (Miotto et al., 2015) . We note that P. falciparum population structures in Africa tend to be far more diverse than in the epicenter of resistance in Cambodia, where parasite strains are highly sub-structured into a few lineages that can readily maintain complex genetic traits (Amato et al., 2018) . A requirement to transmit mutant K13 and additional determinants of resistance in African malaria-endemic settings, where genetic outcrossing is the norm, would predict that ART resistance will spread more gradually than in SEA.

Another impediment to the dissemination of ART resistance in Africa is the continued potent activity of lumefantrine, the partner drug in the first line treatment artemether-lumefantrine.

This situation contrasts with SEA where ART-resistant parasites also developed high-level resistance to the partner drug PPQ, with widespread treatment failures enabling the dissemination of multidrug-resistant strains (Conrad and Rosenthal, 2019; van der Pluijm et al., 2019) . These data call for continent-wide monitoring for the emergence and spread of mutant K13 in Africa, and for studies of whether its emergence in Rwanda is a harbinger of subsequent partner drug resistance and ACT treatment failure.

Samples were obtained as blood-spot filter papers from patients seeking treatment at sites involved in national surveys of antimalarial drug resistance, from patients enrolled in therapeutic efficacy studies, from asymptomatic participants who were enrolled in surveillance programs. Collection details from African and Cambodian samples are provided in completeness of the molecular data. DNA was extracted from dried blood spots using QIAmp Mini kits, as described . A nested PCR was performed on each sample to amplify the K13-propeller domain, corresponding to codons 440-680. PCR products were sequenced using internal primers and electropherograms analyzed on both strands, using the Pf3D7_1343700 3D7 sequence from 3D7 parasites as the reference sequence. Quality controls included adding six blinded quality-control samples to each 96-well sequencing plate 20 prepared from samples from each in-country partner and independently retesting randomly selected blood samples. Isolates with mixed alleles were considered to be mutated for the purposes of estimating the mutation frequencies.

Plasmodium falciparum asexual blood-stage parasites were cultured in human erythrocytes at 3% hematocrit in RPMI-1640 medium supplemented with 2 mM L-glutamine, 50 mg/L hypoxanthine, 25 mM HEPES, 0.225% NaHCO3, 10 mg/L gentamycin and 0.5% w/v Albumax II (Invitrogen). Parasites were maintained at 37ºC in 5% O2, 5% CO2, and 90% N2. Cultures were monitored by light microscopy of methanol-fixed, Giemsa-stained blood smears. The geographic origin and year of culture adaptation for lines employed herein are described in Table 1 and Table 1-table supplement 1 .

To define the genome sequences of our P. falciparum lines used for transfection, we lysed parasites in 0.05% saponin, washed them with 1´PBS, and purified genomic DNA (gDNA) using the QIAamp DNA Blood Midi Kit (Qiagen). DNA concentrations were quantified by NanoDrop (Thermo Scientific) and Qubit (Invitrogen) prior to sequencing. 200 ng of gDNA was used to prepare sequencing libraries using the Illumina Nextera DNA Flex library prep kit with dual indices. Samples were multiplexed and sequenced on an Illumina MiSeq to obtain 300 bp paired-end reads at an average of 50´ depth of coverage. Sequence reads were aligned to the P. falciparum 3D7 reference genome (PlasmoDB version 36) using Burrow-Wheeler Alignment. PCR duplicates and unmapped reads were filtered out using Samtools and Picard. Reads were realigned around indels using GATK RealignerTargetCreator and base quality scores were recalibrated using GATK BaseRecalibrator. GATK HaplotypeCaller (version 3.8) was used to identify all single nucleotide polymorphisms (SNPs). These SNPs were filtered based on quality scores (variant quality as function of depth QD > 1.5, mapping quality > 40, min base quality score > 18) and read depth (> 5) to obtain high-quality SNPs, which were annotated using snpEFF. Integrated Genome Viewer was used to visually verify the presence of SNPs. BIC-Seq was used to check for copy number variations using the Bayesian statistical model (Xi et al., 2011) . Copy number variations in highly polymorphic surface antigens and multi-gene families were removed as these are prone to copy number changes with in vitro culture.

These whole-genome sequencing data were used to determine the genotypes of the antimalarial drug resistance loci pfcrt, mdr1, dhfr and dhps (Haldar et al., 2018) . We also genotyped fd, arps10, mdr2, ubp1, and ap-2μ, which were previously associated with ART resistance (Henriques et al., 2014; Miotto et al., 2015; Cerqueira et al., 2017; Adams et al., 2018) . These results are described in Table 1-table supplement 1 .

Zinc-finger nuclease-meditated editing of select mutations in the K13 locus was performed as previously described (Straimer et al., 2015) . CRISPR/Cas9 editing of K13 mutations was achieved using the pDC2-cam-coSpCas9-U6-gRNA-hdhfr all-in-one plasmid that contains a P. falciparum codon-optimized Cas9 sequence, a human dihydrofolate reductase (hdhfr) gene expression cassette (conferring resistance to WR99210) and restriction enzyme insertion sites for the guide RNA (gRNA) and donor template (White et al., 2019) . A K13 propeller domain-specific guide gRNA was introduced into this vector at the BbsI restriction sites using the oligo pair p1+p2 (Supplementary file 5) using T4 DNA ligase (New England BioLabs). Oligos were phosphorylated and annealed prior to cloning. A K13 donor template consisting of a 1.5 kb region of the K13 coding region including the entire propeller domain was amplified using the primer pair p3+p4 and cloned into the pGEM T-easy vector system (Promega). This donor sequence was subjected to site-directed mutagenesis in the pGEM vector to introduce silent binding-site mutations at the Cas9 cleavage site using the primer pair p5+p6, and to introduce allele-specific mutations using the primer pairs (p7 to p20). K13 donor sequences were amplified from the pGEM vector using the primer pair p21+p22 and sub-cloned into the pDC2-cam-coSpCas9-U6-gRNA-hdhfr plasmid at the EcoRI and AatII restriction sites by In-Fusion® Cloning (Takara). The final plasmids were then sequenced using primers p23 to p25. A schematic showing the method of K13 plasmid construction can be found in Supplementary file 1.

CRISPR/Cas9 editing of fd and mdr2 was performed using a separate all-in-one plasmid, pDC2-cam-Cas9-U6-gRNA-hdhfr, generated prior to the development of the codon-optimized version used above for K13 (Lim et al., 2016) . Cloning was performed as for K13, except for gRNA cloning that was performed using In-Fusion® cloning (Takara) rather than T4 ligase.

Cloning of gRNAs was performed using primer pair p29/p30 for fd and p42/p43 for mdr2.

Donor templates were amplified and cloned into the final vector using the primer pairs p31/p32 for fd and p44+p45 for mdr2. Site-directed mutagenesis was performed using the allelespecific primer pairs p33+p34 or p35+p36 for fd, and p46+p47 or p48+p49 for mdr2. All final plasmids (both fd-and mdr2-specific) were sequenced using the primer pair p37+p38 

Gene-edited lines were generated by electroporating ring-stage parasites at 5-10% parasitemia with 50 μg of purified circular plasmid DNA resuspended in Cytomix. Transfected parasites were selected by culturing in the presence of WR99210 (Jacobus Pharmaceuticals) for six days post electroporation. Parental lines harboring 2-3 mutations in the P. falciparum dihydrofolate reductase (dhfr) gene were exposed to 2.5 nM WR99210, while parasites harboring four dhfr mutations were selected under 10 nM WR99210 (see Table 1-table   supplement 1) . Parasite cultures were monitored for recrudescence by microscopy for up six weeks post electroporation. To test for successful editing, the K13 locus was amplified directly from whole blood using the primer pair p26+p27 (Supplementary file 5) and the MyTaq™ Blood-PCR Kit (Bioline). Primer pairs p39+p40 and p50+p51 were used to amplify fd and mdr2, respectively. PCR products were submitted for Sanger sequencing using the PCR primers as well as primer p28 in the case of K13, p41 (fd) or p52 (mdr2). Bulk-transfected cultures showing evidence of editing by Sanger sequencing were cloned by limiting dilution.

Synchronized parasite cultures were obtained by exposing predominantly ring-stage cultures to 5% D-Sorbitol (Sigma) for 15 min at 37°C to remove mature parasites. After 36 h of subsequent culture, multinucleated schizonts were either purified over a density gradient consisting of 75% Percoll (Sigma). Purified schizonts were incubated with fresh RBCs for 3h, and early rings (0-3 hours post invasion; hpi) were treated with 5% D-Sorbitol to remove remaining schizonts.

In vitro RSAs were conducted as previously described, with minor adaptations (Straimer et al., 2015) . Briefly, tightly synchronized 0-3 hpi rings were exposed to a pharmacologicallyrelevant dose of 700 nM DHA or 0.1% dimethyl sulfoxide (DMSO; vehicle control) for 6 h at 1% parasitemia and 2% hematocrit, washed three times with RPMI medium to remove drug, transferred, and cultured for an additional 66 h in drug-free medium. Removal of media and resuspension of parasite cultures was performed on a Freedom Evo 100 liquid-handling instrument (Tecan). Parasitemias were measured at 72 h by flow cytometry (see below) with at least 50,000 events captured per sample. Parasite survival was expressed as the percentage value of the parasitemia in DHA-treated samples divided by the parasitemia in DMSO-treated samples processed in parallel. We considered any RSA mean survival rates <2% to be ART sensitive.

Flow cytometry was performed on an BD Accuri TM C6 Plus cytometer with a HyperCyt plate sampling attachment (IntelliCyt), or on an iQue3® Screener Plus cytometer (Sartorius). Cells were stained with 1´SYBR Green (Invitrogen) and 100 nM MitoTracker DeepRed (Invitrogen) for 30 min and diluted in 1´PBS prior to sampling. Percent parasitemia was determined as the percentage of MitoTrackerpositive and SYBR Green-positive cells. For RSAs, >50,000 events were captured per well.

Fitness assays with African K13-edited parasite lines were performed by co-culturing isogenic wild-type unedited and mutant edited parasites in 1:1 ratios. Assays were initiated with tightly synchronized trophozoites. Final culture volumes were 3 mL. Cultures were maintained in 12well plates and monitored every four days over a period of 40 days (20 generations) by 25 harvesting at each time point a fraction of each co-culture for saponin lysis. gDNA was then extracted using the QIAamp DNA Blood Mini Kit (Qiagen). The percentage of the WT or mutant allele in each sample was determined in TaqMan allelic discrimination real-time PCR assays. TaqMan primers (forward and reverse) and TaqMan fluorescence-labeled minor groove binder probes (FAM or HEX, Eurofins) are described in Supplementary file 7. Probes were designed to specifically detect the K13 M579I or C580Y propeller mutations. The efficiency and sensitivity of the TaqMan primers was assessed using standard curves comprising 10-fold serially diluted templates ranging from 10 ng to 0.001 ng. Robustness was demonstrated by high efficiency (88-95%) and R 2 values (0.98-1.00). The quantitative accuracy in genotype calling was assessed by performing multiplex qPCR assays using mixtures of WT and mutant plasmids in fixed ratios (0:100, 20:80, 40:60, 50:50, 60:40, 80:20, 100:0). Triplicate data points clustered tightly, indicating high reproducibility in the data across the fitted curve (R 2 = 0.89 to 0.91).

Purified gDNA from fitness co-cultures was subsequently amplified and labeled using the primers and probes described in Supplementary file 7. qPCR reactions for each sample were run in triplicate. 20 μL reactions consisted of 1´QuantiFAST reaction mix containing ROX reference dye (Qiagen, Germany), 0.66 µM of forward and reverse primers, 0.16 µM FAM-MGB and HEX-MGB TaqMan probes, and 10 ng genomic DNA. Amplification and detection of fluorescence were carried out on the QuantStudio 3 (Applied Biosystems) using the genotyping assay mode. Cycling conditions were as follows: 30s at 60°C; 5 min at 95°C; and 40 cycles of 30s at 95°C and 1 min at 60°C for primer annealing and extension. Every assay was run with positive controls (WT or mutant plasmids at different fixed ratios). Notemplate negative controls (water) in triplicates were processed in parallel. Rn, the 26 fluorescence of the FAM or HEX probe, was normalized to the fluorescence signal of the ROX reporter dye. Background-normalized fluorescence (Rn minus baseline, or ΔRn) was calculated as a function of cycle number.

To determine the WT or mutant allele frequency in each sample, we first confirmed the presence of the allele by only retaining values where the threshold cycle (Ct) of the sample was less than the no-template control by at least three cycles. Next, we subtracted the ΔRn of the samples from the background ΔRn of the no-template negative control. We subsequently normalized the fluorescence to 100% using the positive control plasmids to obtain the percentage of the WT and mutant alleles for each sample. The final percentage of the mutant allele was defined as the average of these two values: the normalized percentage of the mutant allele, and 100% minus the normalized percentage of the wild-type allele.

Fitness assays with Dd2 parasite lines were performed as previously described (Ross et al., 2018) . Briefly, K13-edited parasite lines were co-cultured in 1:1 ratios with an eGFP-positive (eGFP + ) Dd2 reporter line. Fitness assays were initiated with tightly synchronized trophozoites in 96-well plates, with 200 μL culture volumes. Percentages of eGFP + parasites were monitored by flow cytometry every two days over a period of 20 days (10 generations).

Flow cytometry was performed as written above, except that only 100 nM MitoTracker DeepRed staining was used to detect total parasitemias, since SYBR Green and eGFP fluoresce in the same channel.

The fitness cost associated with a line expressing a given K13 mutation was calculated was calculated relative to its isogenic WT counterpart using the following equation:

where P' is equal to the parasitemia at the assay endpoint, P is equal to the parasitemia on day 0, n is equal to the number of generations from the assay start to finish, and x is equal to the fitness cost. This equation assumes 100% growth for the WT comparator line. For qPCR and GFP-based fitness assays, days 32 and 20 were set as the assay endpoints, resulting in the number of parasite generations (n) being set to 16 and 10, respectively.

Health care facilities were in charge of collecting anonymized P. falciparum positive cases.

Identification of individuals cannot be established. The studies were approved by ethics committees listed in Supplementary file 4. We note that the sponsors had no role in the study design or in the collection or analysis of the data. There was no confidentiality agreement between the sponsors and the investigators.

We thank Dr. Pascal Ringwald (World Health Organization) for his support and feedback. Table 1 and Table 1-table   supplement (A, B) RSA survival rates for Dd2 (Indochina) and Cam3.II (Cambodia) P. falciparum parasites expressing wild-type or mutant K13. Gene-edited parasites were generated using CRISPR/Cas9 or zinc-finger nucleases (ZFNs). Control (ctrl) lines express silent-binding site mutations at the K13 gRNA cut site. Unedited parental lines are described in Table 1 and Table 1-table supplement 

shown as means ± SEM (detailed in Figure 6- Total   WT  308  270  352  30  183  316  96  382  220  871  137  3165   SYN  3  2  8  1  3  12  0  7  8  11 0 55 K13 , Kelch13, PF3D7_1343700; pfcrt , Plasmodium falciparum chloroquine resistance transporter, PF3D7_0709000; mdr1 , multi-drug resistance protein 1, PF3D7_0523000; dhfr, dihydrofolate reductase, PF3D7_0417200; dhps, dihydropteroate synthase, PF3D7_0810800; fd , ferredoxin, PF3D7_1318100; mdr2 , multi-drug resistance protein 2, PF3D7_1447900; arps10 , apicoplast ribosomal protein S10, PF3D7_1460900.1; ap-2 µ , AP-2 complex subunit mu, PF3D7_1218300; ubp1 , ubiquitin binding protein 1, PF3D7_0104300. n/a 19 n/a n/a n/ a  46  110  104  4  12  59  225  13  18  31  74  2  0  68  93  1  1194 G449A 0 0 0 n/a 0 0 0 n/a 0 0 0 n/a 2 n/a n/a n/a N, number of independent experiments, each with technical duplicates. c P value calculated for relative to the respective wild-type parental line. P value determined by unpaired t test. ns, not significant. ** P <0.01, *** P <0.001. SEM, standard error of the mean; WT, wild-type; ZFN, zinc-finger nuclease. Single plasmid all-in-one approach Supplementary file 1. CRISPR/Cas9 strategy for editing the K13 locus. All-in-one plasmid approach used for CRISPR/Cas9-mediated K13 gene editing, consisting of a K13-specific donor template for homology-directed repair, a K13-specific gRNA expressed from the U6 promoter, a Cas9 cassette with expression driven by the calmodulin (cam) promoter, and a selectable marker (human dhfr, conferring resistance to the antimalarial WR99210 that inhibits P. falciparum dhfr). The Cas9 sequence was codon-optimized for improved expression in P. falciparum. Donors coding for specific mutations of interest (e.g. K13 C580Y, red star) were generated by site-directed mutagenesis (SDM) of the K13 wild-type donor sequence. Green bars indicate the presence of silent binding-site mutations that were introduced by SDM to protect the edited locus from further cleavage. The lightning bolt indicates the location of the cut site in the genomic target locus. Primers used for SDM and cloning and transfection plasmids are described in Supplementary files 5 and 6, respectively. 14 Supplementary file 3. CRISPR/Cas9 strategy for editing the ferrodoxin (fd) and multidrug resistance protein 2 (mdr2) loci.

All-in-one plasmid approaches used for CRISPR/Cas9-mediated editing of (A) the ferredoxin (fd) locus or (B) the multidrug resistance protein 2 (mdr2) locus. Plasmids consisted of a fd (A) or mdr2 (B) specific donor template for homology-directed repair, a locus-specific gRNA expressed from the U6 promoter, a Cas9 cassette with expression driven by the cam promoter, and a selectable marker (hdhfr, conferring resistance to WR99210). Donors coding for specific mutations of interest (i.e. fd D193Y or mdr2 T484I) were generated by site-directed mutagenesis (SDM) of the WT donor sequences. Red stars indicate the presence of silent binding-site mutations (also introduced by SDM) used to protect edited loci from further cleavage. Primers used for SDM and cloning and transfection plasmids are described in Supplementary files 5 and 6, respectively.

Country Year(s) Sample collection and K13 genotyping Ethics -Study approved by The Gambia

Trial registration number ACTRN12618000517279)

Institutional Ethics Committee of CNRFP and the Health Research Ethics Committee of Burkina Faso (PACTR201708002499311)

The National Malaria Control Programme in CAR Republic of the Congo

Equatorial Guinea

the Ifakara Health Institute (IHI) ethics review committee (ERC), the National Medical Research Coordinating Committee (MRCC) of the National Institute for Medical Research (NIMR) and by the WHO research ethics review committee Rwanda

the Rwanda National Ethics Committee on 16

The Ministry of Health of the Federal Government of Somalia, the Ministry of Health of Puntland, and the WHO Research Ethics Review Committee ACTRN12616001005448

From the Cambodian National Ethics Committee for Health Research, the Institutional Review Board of the Naval Medical Research Center, the Technical Review Group of the WHO Regional Office for the Western Pacific, and the Institutional Review Board of the National Institute of Allergy and Infectious Diseases For the following countries, K13 genotyping was performed at the Pasteur Institute in Paris: The Gambia

A worldwide map of Plasmodium falciparum K13-propeller polymorphisms

Limited artemisinin resistance-associated polymorphisms in Plasmodium falciparum K13 -propeller and PfATPase6 gene isolated from Bioko Island, Equatorial Guinea

High therapeutic efficacy of artemether-lumefantrine and dihydroartemisinin-piperaquine for the treatment of uncomplicated falciparum malaria in Somalia

fwd, forward; rev, reverse; WT, wild-type.