Impact of different mutations on Kelch13 protein levels, ART resistance, and fitness cost in Plasmodium falciparum parasites

ABSTRACT Reduced susceptibility to ART, the first-line treatment against malaria, is common in South East Asia (SEA). It is associated with point mutations, mostly in kelch13 (k13) but also in other genes, like ubp1. K13 and its compartment neighbors (KICs), including UBP1, are involved in endocytosis of host cell cytosol. We tested 135 mutations in KICs but none conferred ART resistance. Double mutations of k13C580Y with k13R539T or k13C580Y with ubp1R3138H, did also not increase resistance. In contrast, k13C580Y parasites subjected to consecutive RSAs did, but the k13 sequence was not altered. Using isogenic parasites with different k13 mutations, we found correlations between K13 protein amount, resistance, and fitness cost. Titration of K13 and KIC7 indicated that the cellular levels of these proteins determined resistance through the rate of endocytosis. While fitness cost of k13 mutations correlated with ART resistance, ubp1R3138H caused a disproportionately higher fitness cost. IMPORTANCE Parasites with lowered sensitivity to artemisinin-based drugs are becoming widespread. However, even in these “resistant” parasites not all parasites survive treatment. We found that the proportion of surviving parasites correlates with the fitness cost of resistance-inducing mutations which might indicate that the growth disadvantages prevents resistance levels where all parasites survive treatment. We also found that combining two common resistance mutations did not increase resistance levels. However, selection through repeated ART-exposure did, even-though the known resistance genes, including k13, were not further altered, suggesting other causes of increased resistance. We also observed a disproportionally high fitness cost of a resistance mutation in resistance gene ubp1. Such high fitness costs may explain why mutations in ubp1 and other genes functioning in the same pathway as k13 are rare. This highlights that k13 mutations are unique in their ability to cause resistance at a comparably low fitness cost.

Mutations in the Pfkelch13 gene (k13) encoding the K13 protein are considered a major contributor to reduced ART susceptibility (9,10) and this protein was recently found to be involved in endocytosis of host cell cytosol in the ring stage of the parasite (11).Thirteen mutations affecting the C-terminal propeller domain of K13 have been observed to reduce ART susceptibility as evidenced by delayed clearance in patients and in vitro (F446I, C469Y, N458Y, M476I, Y493H, R539T, I543T, P553L, R561H, P574L, C580Y, R622I, and A675V) and others have been associated with reduced susceptibility (3,4).Several other proteins, mostly from the K13-compartment, confer in vitro ART resistance in RSA either when downregulated or when mutated (11)(12)(13)(14)(15).One of them is ubp1, for which the R3138H was shown to result in moderate resistance in RSA (11) and which was associated with reduced susceptibility in patient samples from Asia, although it occurred only rarely (16).The contribution of mutations affecting K13 compartment proteins (other than K13) to reduced ART susceptibility in endemic countries is largely unknown.
While ART resistance mutations are frequent in South East Asia, up to 100% in some locations (21)(22)(23), they are rarer in Africa (22,24,25).Six validated resistance-conferring k13 mutations (M476I, P553L, R561H, P574L, C580Y, and A675V) have occasionally been reported in African countries at moderate to low frequency of 4.1% or less (26)(27)(28).Three exceptions are the R561H mutation in Rwanda, which was reported more frequently since 2019, most recently up to 16% (29)(30)(31), 11% of A675V mutation in Uganda in 2019 (28) and R622I in Eritrea at 14% (4).It is not clear whether further resistance mutations exist in Africa.Most studies focused on mutations occurring in Asia and there is no systematic k13 surveillance in Africa.In agreement with the rare occurrence of resistance mutations, delayed parasite clearance occurs at less than 2% in Africa (32), with the exception of the Masaka region in Rwanda at 16%, which is where k13 R561H occurs (31).It was postulated that resistance would emerge later in Africa than in Asia because of the higher transmission rate and higher level of immunity (33,34).
The reduced endocytosis in ART-resistant parasites deprives the parasite of amino acids (35) and there is clear evidence for a fitness cost of ART resistance (10,36).Usually, considerably less than 50% of parasites from resistant patient isolates survive in the RSA (9) and the delayed clearance rather than full loss of susceptibility in patients could indicate that it is possible for parasites with higher levels of ART resistance to arise.However, it is at present unclear if and to what extent the fitness cost impedes higher levels of resistance.It is also unclear if combinations of mutations could lead to hyper-resistant parasites.
Here, we tested 135 mutations in proteins from the K13-compartment including 4 mutations in K13 itself and found that none of them confers significant levels of resistance over GFP-tagged wild-type (WT) K13.However, some of the k13 mutations resulted in a significantly lower susceptibility to ART than the 3D7 background.Together with known high resistance mutations, this provided us with 3D7-based laboratory lines with different resistance levels.A comparison of parasites with different resistance levels showed a correlation between resistance, fitness cost, and K13 protein abundance of a given variant, suggesting a constraint of resistance levels by the fitness cost.This was confirmed using parasites artificially selected for higher resistance.We also assessed the threat of hyper-resistance, the impact on fitness cost on this and the potential impact of k13 double mutations and combinations with resistance mutations outside k13.

Tested mutations in K13-compartment proteins do not cause in vitro ART resistance
We aimed to assess whether further mutations in addition to the already identified resistance mutations could cause reduced ART susceptibility.As not only k13, but also genes of K13 compartment proteins such as ubp1 can cause reduced susceptibility, we analyzed non-synonymous mutations in k13 and genes of other K13 compartment proteins from African patient samples available from WWARN (37), the Ghanaian Fever Without Source study (38) and MalariaGen (39).These included mutations in the genes encoding K13, KIC1, KIC2, KIC4, KIC5, KIC7, KIC9, UBP1, EPS15, and MyoF (previously annotated as MyoC) (list of mutations in Table S1).In addition, we selected one untested mutation in AP2α, which was reported to be present in in vitro ART-selected parasites that displayed a K13-independent reduced ART-susceptibility (40).Due to the large number of mutations in the K13 compartment proteins (a total of 131 mutations in 10 genes), we created parasites carrying multiple mutations (multi-mutants) by introducing a recodonized sequence of the target gene that encodes several mutations in the same gene into 3D7 parasites using selection-linked integration (SLI) (41) (Table S1; Fig. S7C  and D).In total, 125 mutations were included in eight multi-mutants and six mutations were tested individually (Fig. 1A and B).RSAs [using dihydroartemisinin (DHA)] with multi-mutants of compartment members of K13 showed no change in susceptibility to ART, suggesting that none of the mutations in the genes encoding KIC1, KIC2, KIC4, KIC5, KIC7, KIC9, UBP1, EPS15, and MyoF resulted in ART resistance (Fig. 1A and B).One exception was AP2α H817P which significantly increased RSA survival, although this was not above 1% which is considered the cutoff for clinically relevant resistance (8).Yet this increase might explain the occurrence of this mutation in parasites that had been selected for ART resistance through ART exposure in vitro but that lacked K13 mutations (40).
For k13 itself we investigated the mutations V520A, V589I, and E612K that have been found with low to moderate prevalence (0.8-5.0%) in different malaria-endemic regions in Africa and were not previously tested for ART resistance in vitro (Fig. 1C).All mutations originated from studies where the samples had been taken after ACT started to be used in the respective country, with exception of one study in Kenya in 2002 (Data Set S1).We noticed that k13 V520A and V589I occurred in regions with above average and average malaria incidence, respectively, while k13 E612K, C580Y, R539T, and k13 R561H, which was recently detected in Rwanda, occurred in regions with below average malaria incidences (Fig. S1; Data Set S2).E252Q occurred in Thailand and Myanmar between 2005 and 2013 where it was previously associated with slow parasite clearance (37) but not tested in vitro.Although only detected in Asia, we included this mutation because it might have been missed in Africa as typically only the propeller domain (amino acid residues 444-721) is sequenced when screening for k13 mutations.The respective mutations were introduced into the k13 locus of P. falciparum 3D7 parasites together with fluorescent protein GFP to result in a GFP-K13 fusion as done previously (11).The susceptibility of these parasites to ART was tested by RSA.Introduction of the GFP to the k13 locus already led to small but significant reduction in ART-susceptibility (0.2 ± 0.2% vs 0.6 ± 0.3%) (K13-WT, Fig. 1D), suggesting that the addition of the GFP to the protein's N-terminus resulted in a very mild reduction in K13 activity.All three mutations caused a small further decrease in ART-susceptibility although this was not significantly different over k13 WT but for k13 V520A and E252Q was significant over the 3D7 parent (Fig. 1D).We conclude that these mutations likely have no relevance for ART resistance in endemic settings but provide us with parasites with a mildly reduced ART-susceptibility compared with 3D7.

Double mutations in k13 do not significantly increase ART resistance
To anticipate how k13 resistance phenotypes could further develop in endemic areas, we investigated whether the combination of the k13 C580Y mutation (which is highly prevalent in SE Asia) with the V520A mutation might potentiate resistance, e.g., through a synergetic effect.Further, we investigated whether the two known resistance mutations k13 C580Y and R539T could combine to result in hyper-resistance.To obtain the corresponding cell lines, we again modified the endogenous kelch13 locus of 3D7 parasites using SLI so that the double-mutated K13 was fused to GFP.The K13 V520A+C580Y parasites did not display an increased resistance level in RSAs (11.5% survival) compared with parasites with only the C580Y mutation (16.3% survival) (Fig. 2A), indicating that there is no synergistic effect.Similarly, K13 R539T+C580Y parasites showed no significant change in RSA survival (33.4% survival) compared with parasites harboring R539T alone (38.4% survival) but significantly more survival than the C580Y alone, suggesting no additive or synergistic effect.In conclusion,-at least with the tested mutations-there is no indication for additive or synergistic effects.

Selection by consecutive RSAs significantly increases ART resistance in k13 C580Y parasites
In a further attempt to anticipate increased resistance, we tested whether parasites with a common resistance mutation can be selected for increased resistance by performing consecutive RSAs.For this, an RSA was carried out with K13 C580Y parasites and surviv ing parasites were subjected to a consecutive RSA as soon as they reached sufficient parasitemia following the previous RSA (Fig. 2B).After 29 iterations of RSAs, we obtained the parasite line K13 C580Y -29 th .These parasites displayed significantly increased survival in RSA (39.2%) compared with the original K13 C580Y parasites at the start of this experiment (12.4%) (Fig. 2C), showing that repeated ART treatment of k13 C580Y-harboring parasites can increase ART resistance.Sequencing of k13 in these parasites showed that this was not due to further mutations in k13 in addition to the C580Y already present at the beginning of the experiment (Fig. S2).We thus performed whole genome sequencing and found 930 SNPs, 1929 InDels, 9 copy number variants and 13 structural variants to be present in the K13 C580Y -29 th parasites but not in the parental K13 C580Y parasites (Data Set S3).Of the SNPs, 56 caused non-synonymous mutations in coding regions.Of the InDels, 89 affected coding regions by shifting the reading frame, affecting splice regions or inserting bases in frame.All copy number variants affected regions containing var, rifin, or stevor genes and three of the structural variants affected coding regions, namely two rifin genes and one var gene (Data Set S3).Compared to the parental genome of K13 C580Y parasites, no changes in any genes of Kelch13-compartment proteins, including all KICs, nor in known resistance-associated genes were detected, with the exception of a one adenosine insertion in a stretch of 29 adenosines in pi3k (PF3D7_0515300) (Data Set S3).As this insertion had a low-quality score and allele frequency and would disrupt 97% of this essential gene, we suspected this was an amplification artifact.Repeated Sanger sequencing with different primers resulted in ambiguous results in regard to the number of As in the said stretch.To resolve this, we sequenced the parental K13 C580Y parasites which resulted in highly similar sequencing electropherogram compared to those obtained with the selected resistance line, indicating that parent and serially selected line are identical in that region (Fig. S3A).In addition, we detected a mutation (R277G) in the non-functional stump of K13 which results from introducing the GFP-fused K13 copy by the SLI method (Fig. S3B).However, overexpression of this N-terminal part (NTP) with or without the mutation or recreating the K13 C580Y parasites with the change in the NTP did not increase ART resistance (Fig. S3C), indicating that this change is not the reason for the altered susceptibility in the serially RSA-selected parasites.
To test whether changes in mRNA levels are responsible for the increased resistance in K13 C580Y -29 th parasites, we determined k13 transcript levels in ring-stage parasites 8-12 h post-invasion by qPCR.To exclude contamination from older parasites we also determined transcript levels of etramp5 and exp2, which peak in the early trophozoite stage (42)(43)(44).We found no significant differences between k13 transcript levels between the K13 WT , K13 C580Y , and K13 C580Y -29 th parasites (Fig. S4).

Cellular levels of K13 and KIC7 proteins determine ART resistance
Previously, we and others showed that the level of K13 in the K13 C580Y and K13 R539T parasites is lower than in K13 wt parasites and that reducing K13 abundance resulted in resistance (11,17,19).We assessed the cellular K13 levels in the parasites for all k13 mutations in this study, to find whether a change in K13 level generally accompa nies resistance mutations.We previously used either Western blots or direct microscopybased measurements of GFP fluorescence in the cells to determine K13 levels (11).Here, we used direct microscopy-based measurement of K13 levels because in comparison to Western blots, which are only semi-quantitative, this procedure is not susceptible to the influence of cell lysis and the multiple steps required for blotting and detection.These experiments showed no significant reduction in the amount of K13 with the V520A (94% of WT) and the V589I (91% of WT) mutations (Fig. 3A).The E252Q mutation showed 105% of the amount (non-significant) of WT (Fig. 3A).The resistance mutations k13 C580Y, k13 R561H and k13 R539T had 52%, 51%, and 51% of the K13 amount of WT, respectively (Fig. 3A).As expected from their RSA resistance behavior, parasites with the k13 V520A C580Y double mutation harbored similar levels of K13 (48% of WT) as parasites with the k13 C580Y alone, and parasites with k13 R539T and C580Y (57% of WT) had similar levels as both k13 C580Y and k13 R539T individually.The K13 levels in K13 C580Y -29 th parasites showed a further reduction (32% of WT) when compared with K13 C580Y parasites, even though this reduction was not significant when comparing means (Fig. 3A), only when comparing all data points (Fig. S5A).This difference could be due to difference in parasite age if K13 levels increase during ring-stage development.If this were the case, K13 C580Y parasites would need to have K13 levels in young rings that are considerably lower than the 52% observed in this assay.To test this, we compared K13 levels in young (0-3 h) and late (12-15 h) rings (Fig. S5B).In K13 WT parasites, we observed no change.In K13 C580Y parasites there was a small increase from 51% to 68% of K13 WT , showing that in principle the age of ring-stage parasites can influence the amount of K13.Yet already in young rings the K13 levels in K13 C580Y were considerably higher than the average in the K13 C580Y -29 th parasites.Thus, parasite age does not explain the lower K13 levels in K13 C580Y -29 th parasites compared to the parental K13 C580Y .
We plotted the cellular K13 levels against RSA survival (in vitro resistance).Cellular K13 levels correlated with in vitro resistance (Pearson's r = −0.81,P = 0.005) (Fig. 3B).We note that this was largely because the mutations with low K13 levels clustered into a group conferring significant in vitro resistance over GFP-K13-WT and the other mutations into a group with high K13 levels (Fig. 3B).
To test whether RSA-resistance not just correlates with but is determined by the amount of K13 in its cellular location, we used K13 WT 1xNLS parasites (11).In this line, K13 protein is removed into the nucleus by a nuclear localization signal (NLS) upon the addition of rapalog (termed knock sideways), resulting in a conditional reduction of K13 at its cellular location.Previously, 250 nM rapalog was used to achieve a knock sideway of K13 (41).Here, we used different concentrations of rapalog to result in different levels of K13 protein remaining at its site of action.ART resistance, measured by RSA, decreased as the rapalog concentration and hence K13 knock sideways decreased, confirming that RSA-survival is a function of K13 levels in the parasites (Fig. 3C).These results confirm observations made on parasites in which the K13 abundance was titrated on mRNA level through the glmS system (20).To confirm that this was due to a reduction in endocytosis, we performed a second titration using KIC7, a K13-compartment protein, which was also knocked aside using the same decreased rapalog concentrations as for K13 (Fig. 3D).Again, RSA survival was dependent on the level of knock sideways of KIC7 (i.e., amount of KIC7 at the K13 compartment), indicating that the activity of the endocytic process determines resistance and that this occurs through the available amount of some of the proteins involved in this process.

Fitness cost correlates with ART resistance
Several studies modeling fitness and within-host competition have highlighted the importance of parasite fitness and propose that resistant parasites with reduced fitness are less likely to establish themselves in high transmission areas like most of Africa (34,45,46).Some K13 resistance mutations, including C580Y and R561H, have been shown to have fitness costs (23,36).To measure the impact of k13 mutations on fitness of our parasites with isogenic backgrounds in this work, we first monitored in vitro growth over 96 h.This confirmed the fitness cost of k13 C580Y and suggested that generally, fitness cost might correlate with resistance but the variation between experiments was too high for reliable conclusions (Fig. 4A and B).
For better fitness cost measurements, parasite proportions were determined in a mixed culture of a mutant and the 3D7 parent parasite line, as previously described (23,47).Parasites were mixed 1:1 with the 3D7 parasites and the proportion of mutated (fluorescent) parasites was tracked until it made up less than 5% of the parasite popula tion (Fig. 4C; Fig. S6).The parasites with K13 mutations (V520A, V589I, E252Q, C580Y, and C580Y-29th) declined in proportion compared to 3D7 in all six experiments whereas GFP-K13 WT did not or only to a small extent (Fig. S6).Similar to what was observed when parasite growth was tracked in individual cultures, there was a trend that the RSAsurvival caused by a mutation correlated with the fitness cost inflicted by the mutation (Fig. 4D).However, this trend was not significant (Pearson's r = 0.69, P = 0.13) because k13 V520A behaved as an outlier in half of the experimental repeats, in which its growth slowed severely after the start of the experiment (Fig. S6), while in the other half it grew in line with the trend of the other mutations (Pearson's r = 0.90, P = 0.04 when V520A was entirely excluded) (Fig. 4C and D).Together with the findings on K13 abundance (Fig. 3), it is therefore likely that both fitness cost and ART resistance are caused by the reduced amount of K13 in the cell as a result of the destabilization of the K13 protein through the given mutation.

ubp1 R3138H has a disproportional fitness cost that is further aggravated in K13 C580Y parasites
After assessing the effect of the k13 double mutations V520A + C580Y and R539T + C580Y, we wondered whether there are combinations of mutations or gene disruptions beyond k13 that harbor the risk of very high resistance levels.We tested this by combin ing k13 C580Y with ubp1 R3138H and the resulting K13 C580Y + UBP1 R3138H parasites did not display a significantly increased resistance level (17.3% mean survival) compared with parasites with the C580Y mutation alone (Fig. 5A).We also attempted to disrupt both kic4 and kic5, two genes we were previously able to disrupt individually (11), or combine their disruption with UBP1 R3138H but were unsuccessful despite six attempts per combination, which may indicate that the fitness cost of combining these prevented us from obtaining these parasite lines.
We further characterized the K13 C580Y + UBP1 R3138H parasites and found that they displayed a similar K13 abundance to resistance level ratio (Fig. 5B and C) as the k13 mutation (Fig. 3B).The growth of K13 C580Y + UBP1 R3138H parasites was slower than that of K13 C580Y parasites and UBP1 R3138H parasites, which each confer a fitness cost, both when measured in individual cultures and direct competition assays (Fig. 5D through G).The ratio of fitness cost to resistance level was higher than for k13 mutations alone (Fig. 5E  and G).To determine whether just the combination of the two mutations led to such a reduced growth rate or whether it was due to ubp1 R3138H by itself, we determined the growth rate of UBP1 R3138H parasites.Interestingly, already the ubp1 R3138H mutation alone caused a disproportionally low growth compared with the level of ART resistance it affords when compared to all tested k13 mutations (Fig. 5E).This is congruent with our previous finding that in contrast to K13 which is only required in rings, UBP1 is also important for hemoglobin endocytosis in trophozoites (11) and could explain why we were not able to combine this mutation with kic4 and kic5 TGDs and why this mutation has rarely been observed in patient samples.Overall, due to its disproportionally high fitness cost and lacking synergism with k13 mutations, ubp1 R3138H and possible mutations in other K13 compartment proteins important for endocytosis in trophozoites, are therefore unlikely to form combinations with k13 mutations in endemic countries.

DISCUSSION
In this study, we tested 131 mutations in ten genes using a genetic system that allowed us to introduce several mutations into the same gene simultaneously.We found that none of the tested mutations in the genes encoding KIC1, KIC2, KIC4, KIC5, KIC7, KIC9, UBP1, EPS15, and MyoF resulted in significantly decreased susceptibility to ART, yet we cannot fully exclude that epistatic interactions masked effects of individual mutations included in the multi-mutants.We also tested K13 mutations.We found that N-termi nally fusing a tag to K13, even without any mutations, very mildly reduced the suscept ibility to ART.This effect is negligible compared with the larger effects of common resistance mutations such as C580Y, R561H, and R539T and likely does not generally impact experiments with these parasites but needs to be considered when assessing mutations with small effect sizes.Thus, we found that the four previously unstudied k13 mutations did not cause significant resistance in the RSA compared with K13-WT.The small non-significant decrease in susceptibility that was observed for k13 V520A, V589I, and E252Q likely has no clinical impact but we cannot exclude that these variants have some small advantage in the population that is difficult to assess.However, all of these mutations that were tested in competition assays were consistently outcompeted by 3D7 whereas GFP-K13 WT was not, indicating that they have a small disadvantage over WT.Despite likely not being clinically relevant, parasites with a GFP-fused K13 with or without the V520A, E252Q, and E612K were significantly less susceptible to ART in RSA compared to 3D7 and together with the high resistance parasites offered us a tool to explore the properties of parasites with different levels of in vitro ART resistance.
We also found that the combination of k13 C580Y with k13 V520A, or of two resistance mutations with each other, like k13 C580Y and k13 R539T, and k13 C580Y with ubp1 R3138H did not increase ART resistance.These are the first combinations of C580Y with other naturally occurring mutations generated in isogenic backgrounds.A similar non-additive result was obtained in a recent study for coronin R100K and E107V, two in vitro selected mutations, combined with k13 C580Y (15).Previous work has shown that the genetic background can influence the resistance level of parasites with identical k13 resistance alleles (10,23), indicating that mutations outside the K13 compartment exist that modulate the ART susceptibility of parasites with k13 resistance mutations and remain to be determined.Repeated ART treatment in consecutive RSAs of k13 C580Y-harboring parasites increased ART resistance, but this was not due to additional k13 mutations apart from C580Y or mutations in any other gene known to be involved in drug resistance.Parasites from such selections might thus be a promising tool to study factors outside of k13 that influence resistance levels in parasites with a resistance-conferring allele of k13.
We found that the amount of K13 and KIC7 at the K13 compartment in the cell determines the level of in vitro resistance, confirming that the endocytic process is important for ART resistance and suggesting that the amount of protein directly influences the rate of endocytosis and level of ART susceptibility.The same effect on resistance was observed in a k13 mRNA titration recently (20) and a role of K13 protein abundance rather than a specific functional change in resistance is also supported by the finding that increasing levels of K13 C580Y reverts resistant parasite back to sensitivity (11).The finding that KIC7 titration had a similar effect strengthens this conclusion and further supports that resistance is due to altered endocytosis levels.We found that all tested resistance mutations had low K13 levels and all other tested mutations had high K13 levels.The protein levels in the K13 double mutants were comparable to that of the single resistance mutation, indicating that at least in these combinations, multiple mutations do not further destabilize K13.In conclusion, for all tested k13 resistance mutations, the amount of K13 in the parasite, likely through destabilization of K13, was a factor contributing to or causing the resulting resistance.
We further showed that the fitness cost of k13 mutations generally correlated with resistance.This agrees with a recent study looking at k13 M579I, C580Y, and R561H in several African strains including 3D7 although no such correlation was observed in the Asian strain Dd2 (23).It is unclear what caused these differences between African strains and Dd2.Nonetheless, it showed that the genetic background in which the mutations occur can influence the fitness cost and the resistance level (23,36,48).Many differences have been observed in ART-resistant parasites compared with sensitive parasites (18,21,(49)(50)(51)(52)(53)(54) and some of these might correspond to such changes in the background that alter the effect of K13 mutations, although some of these changes likely also are downstream effects of the resistance mechanism (35).The data presented here originates from a set of isogenic 3D7-derived parasites which differ from each other only in the studied mutations.While this limits the extrapolation of the conclusions to other parasite strains, it presents the advantage of separating effects caused by the studied mutations from effects of the genetic background.Due to its long presence in culture, 3D7 also avoids variations that may arise in more recently culture-adapted lines that often show reduced growth levels and may still acquire further changes during continued culture to adapt for better growth in culture.
k13 E252Q was previously observed to inflict a fitness cost that is smaller than that of k13 C580Y; however, these experiments were not performed using parasites with isogenic backgrounds (47).Our results confirm this finding in an isogenic background.The E252Q mutation caused low but significant in vitro resistance over 3D7 but not over WT in our study.It was previously associated with delayed parasite clearance in patients, suggesting that either E252Q can only confer resistance in a specific parasite background or that it was not the cause of the delayed clearance.Regarding K13 C580Y -29 th , it should also be noted that the long duration of the competition experiments may permit the K13 C580Y -29 th to revert back to a state similar to the K13 C580Y parasites, as the continuous RSA selection pressure had to be lifted in the competition assay.
We observed a disproportionally high fitness cost in parasites with the ubp1 mutation conferring ART resistance.The high fitness cost is likely due to its importance for hemoglobin endocytosis in trophozoites.This is in agreement with the finding that lowering endocytosis by conditional inactivation or disruption of K13 compartment proteins impairs parasite growth (11,55) and that parasites with a k13 ART resist ance mutation are hypersensitive to low amino acids (35).The small fitness cost of k13 mutations compared to other K13-compartment proteins is likely due to its role in endocytosis exclusively in rings (11).Our findings therefore highlight the unique property of K13 which stands in contrast to resistance-conferring changes in KICs that also affect endocytosis in later stage parasites, thus incurring a higher fitness cost.This is a likely reason why k13 is the predominant gene mutated to cause delayed clearance in patients.Due to their high fitness cost, changes outside k13 are less likely to arise and the resistance level, as observed with the few found in the field (16,56,57), is low.The high fitness cost of mutations of proteins involved in endocytosis outside k13 also indicates that combinations of them with k13 resistance mutations are unlikely to lead to hyper-resistant parasites in the field.We also did not detect any indication for additive or synergistic effects on in vitro ART resistance when the ubp1 resistance mutation and k13 C580Y were both present in the parasite or if two k13 mutations were combined.
We also found that a combination of the two most common resistance mutations observed in SEA, k13 C580Y and k13 R539T, did not lead to hyper-resistance in laboratory strain 3D7.While this shows that the combination of these two prominent resistance mutations together does not reduce the resistance levels in these laboratory parasites, it is important to note that we cannot exclude that other combinations of k13 mutations or combinations in other genetic backgrounds lead to increased resistance.Based on the one tested combination, our results also indicate that mutations with minimal increase in ART resistance such as V520A are not a stepping stone for potentiated resistance when combined with current resistance mutations.The only means by which we were able to increase in vitro ART resistance was a constant RSA selection regime.As in our isogenic parasites, higher resistance was accompanied by higher fitness costs, it is possible that fitness cost limits the level of ART resistance parasites can reach.

Plasmid construction
All primers used in this study are listed in Data Set S4.The mutations tested in k13 and k13 compartment proteins were chosen to include only mutations that were present in a maximum of 5% of samples taken in the sampling study.For the multimutants, the pSLI-TGD plasmid (41) was used (Fig. S7A, C and D).The homology region was directly linked to a synthesized functional codon-changed version of the corresponding candidate containing all selected mutations (Genscript).The fragments of the homology region and the mutated recodonized sequence of all candidates were cloned by Gibson assembly into the pSLI-TGD vector via NotI/MluI, resulting in the vectors pSLI-KIC1mutpool, pSLI-KIC2mutpool, pSLI-KIC4mutpool, pSLI-KIC5mutpool, pSLI-KIC7mutpool, pSLI-KIC9mutpool, pSLI-UBP1mutpool, and pSLI-MyosinFmutpool.Sequencing was performed to confirm the absence of undesired mutations.
The N-terminal SLI plasmid of K13, pSLI-N-GFP-2xFKBP-K13-loxP (41) was modified for the Kelch13 mutant parasites as follows: the mutation was obtained by amplifying the codon-changed synthesized version of k13 using primers encoding the mutation.The resulting two fragments were cloned into the pSLI-N-GFP-2xFKBP-K13-loxP via AvrII/StuI.
To generate K13 V520A+C580Y and K13 R539T+C580Y , pSLI-N-GFP-2xFKBP-K13_C580Y-loxP (41) was modified by amplifying k13 from this plasmid using primers to introduce the mutation and ligating the resulting fragments into the same vector at the AvrII/XhoI sites.
The k13 C580Y mutation was introduced into UBP1 R3138H parasites using an SLI2a plasmid based on the pSLI-N plasmid (41) but containing the BSD resistance gene instead of the hDHFR gene (58).The hDHFR was excised using BamHI and HindIII.KIC4-TGD and KIC5-TGD plasmids were created by inserting the respective homology regions (11) into SLI2a plasmids based on the pSLI-TGD plasmid (58).
To recapitulate the mutation in the NTP of the K13 C580Y -29 th parasites, a small fragment of the original NTP that encoded R277 was recodonized and the R277G mutation was introduced using long primers.A second PCR fragment was then generated to provide a new homology region for integration.The two PCR fragments were joined by overlap PCR and ligated into pSLI-N-GFP-2xFKBP-K13 C580Y-loxP (41) via Not1/Pme1.To episomally express the NTP with and without the R277G mutation it was amplified from either pSLI-N-GFP-2xFKBP-K13_C580Y-loxP or pSLI-N-GFP-2xFKBP-mutationHR-K13_C580Y-loxP (41) and ligated by Gibson assembly into pmCherry-Kelch13_nmd3_BSD via Xho1/SmaI replacing k13 (11).

In-vitro ring-stage survival assay 0-3h (RSA) and consecutive RSAs
All RSAs were performed according to the standard procedure described previously (8).Zero-to three-hour-old rings were treated with 700 nM DHA (AdipoGen) for 6 h and cultivated for another 66 h at 37°C.Giemsa smears were taken and parasite survival rate was determined by comparing the parasitemia of viable parasites after DHA against the parasitemia of the untreated control.Parasites were defined as resistant when the mean survival rate exceeded the cutoff value of 1% (8).
For the consecutive RSA, K13 C580Y parasites (11) were used.At 66 h after the DHA pulse, Giemsa smears were taken and the surviving parasites of the DHA-treated sample were re-cultivated in a new petri dish.After sufficient parasitemia was reached, the parasites originating from the RSA survivors were subjected to a new RSA.This proce dure was continuously repeated for 30 rounds.The k13 gene of the resulting parasites K13 C580Y -29 th was sequenced and showed no change compared to the starting cell line which harbored a recodonized k13 with the C580Y mutation (11).

Fluorescence microscopy
Microscopy was performed as described earlier (62).A Zeiss Axio Imager M1 or M2 provided with a Hamamatsu Orca C4742-95 camera was used for imaging.Zeiss Plan-apochromat 63× or 100× oil immersion objectives with 1.4 numerical aperture were used.Images were edited using Corel Photo Paint X8 and brightness and intensity were adjusted.Images that were used for quantification were not adjusted for brightness and intensity.

Measurement of protein amount by fluorescence intensity
GFP-K13 parasites were synchronized two times using 5% sorbitol at intervals of 2 days.After the second sorbitol synchronization, the cell lines were cultivated at 37°C for two more hours and then GFP signal of the ring-stage parasites was detected by fluorescence microscopy using the 63× oil immersion objective.GFP-K13 WT parasites were always imaged alongside parasites carrying mutations and were used to normalize the signal of mutation-harboring parasites.Parasites were selected based on DIC and then exposed for 200 ms to image green fluorescence.Total intensity of the GFP signal in foci was measured and background signal subtracted using ImageJ [ImageJ2 2018 (63)].

Growth assessment
The parasitemia of a mixed-stage parasite culture was measured by flow cytometry (41) and based on this the parasitemia was adjusted to 0.05-0.1% parasitemia in 2.5% hematocrit.The parasitemia was then measured again by flow cytometry to determine the start parasitemia.These parasites were cultivated for 96 h and the medium was changed every 24 h.After 96 h, parasitemia was measured again by flow cytometry and divided by the starting parasitemia to obtain the fold change in parasitemia.

Competition assay
Schizonts were isolated using 60% percoll purification, washed once with medium and cultured for 6 h to allow invasion of merozoites into new red blood cells.Remaining schizonts were removed by a 10-min incubation in 5% sorbitol solution.The resulting 0-to 6-h-old parasites were cultured at 37°C for 20-24 h after which the parasitemia was measured by flow cytometry (41).Based on the determined parasitemia, the mutant K13 cell lines were co-cultivated in a 1:1 ratio with 3D7 control in a 5-mL petri dish.The proportion of GFP-positive parasites was assessed by fluorescence microscopy until one parasite line reached 95% of parasite proportion.

Measurement of relative Kelch13 mRNA abundance
Tightly synchronized rings (8-12 h post-invasion) were pelleted, dissolved in 10 pellet volumes TRIzol (Life Technologies), and immediately incubated at 37°C for 5 min.Thereafter the samples were transferred to −80°C and stored until RNA extraction.To extract the RNA, the TRIzol samples were thawed, then mixed with 1/5th chloroform prior to centrifugation for 30 min at 4°C.The top phase, containing the RNA, was then purified using the RNeasy Mini Kit (QIAGEN) with an additional in-solution DNase I (QIAGEN) treatment.gDNA contamination of the RNA was tested by qPCR using sbp1 (PF3D7_0501300) primers across an intron (Data Set S4) and DNase I treatment was repeated if necessary.For synthesis of cDNA, random primer (Invitrogen) and dNTP's (Invitrogen) were incubated with the RNA for 5 min at 65°C, before addition of Super script III Reverse Transcriptase (Invitrogen) and RNAseout (Invitrogen) and the synthesis program run (5 min 25°C, 1 h 50°C, and 15 min 70°C).
For quantitative reverse transcription real-time PCR of the cDNA, the desired primer pair (Data Set S4) and QuantiTect SYBR Green PCR master mix (QIAGEN) were prepared in technical duplicates for each sample.Analysis was performed using the LightCycler 480 (Roche) with cycling parameters: 95°C 15 min initial denaturing, 40 cycles of 95°C 15 s, 60°C 1 min, a final denature and annealing step of 95°C 1 min, 60°C 1 min and the generation of a melt curve (60-95°C).Ct calculation was performed using LightCycler 480 software using the absolute quantification fit points method.Primer amplification efficiency was calculated for each primer pair over a 6-point dilution series.Relative abundance of genes of interest was calculated for each sample relative to arginyl tRNA synthetase (PF3D7_1218600; Data Set S4) and to the primer pair amplification efficiency.

Statistical analysis and malaria incidence data
Unpaired or paired two-tailed t tests were performed as indicated and Pearson's r was calculated using GraphPad Prism 9.0.2.Linear regressions were fit to log-transformed data (GraphPad Prism).All error bars shown are standard deviations.Malaria incidence data were downloaded from https://ghdx.healthdata.org/gbd-results-tool.

FIG 1
FIG 1 135 tested mutations in K13-compartment proteins do not cause ART resistance.(A) Mutations introduced in the multi-mutant of KIC1 are exemplarily shown in a schematic and fluorescence images of the resulting parasites harboring multi-mutant-KIC1 are shown.Scale bar: 5 µm.RSA survival for all multi-mutants is displayed (% survival compared to control without DHA 66 h after 6 h DHA treatment in RSA).MyoF multi-mutant was compared with 3D7 as matching WT was not available.(B) RSA survival for single mutations in K13 compartment proteins.Eps15 K447R was compared with 3D7 as matching WT was not available.(C) Position of k13 mutations as well as BTB domain and propeller domain are shown on K13.(D) RSA of different k13 mutant cell lines.Error bars show standard deviations.WT, wild type; red bars indicate the mean parasite survival rate of the respective cell line, each dot represents an independent experiment, green dashed line represents the 1% cutoff value defining clinically relevant ART resistance.Error bars show standard deviations.P values derived from two-tailed unpaired t tests.

FIG 2
FIG 2 DHA-selection but not combination of two mutations increases resistance.(A) RSA survival of the different k13 mutant cell lines shown.WT, wild type; each point represents an independent RSA.Position of k13 mutations is shown on K13.Domains are colored as in Fig. 1.Red bars show mean; error bars show standard deviations.P values derived from two-tailed unpaired t tests.(B) Scheme of experimental procedure of consecutive RSA cycles performed with DHA survivors of the respective prior cycle.(C) Parasite survival rate of K13 C580Y (% survival compared to control without DHA) 66 h after 6 h DHA treatment in RSA.Six experiments per cycle were performed, consisting of six replicates (see color code).Replicates 1, 2, 3 (orange) and 4, 5, 6 (green) were each started on the same days.Black bars show mean.P value is indicated, two-tailed paired t test.

FIG 3
FIG 3 Resistance is inversely correlated with K13-and KIC7-abundance.(A) K13-abundance measured by the GFP fluorescence intensity of the single focus observed in ring-stage parasites with the indicated K13 expressed from the endogenous locus and fused to GFP, normalized to the fluorescence in parasites with the identically modified endogenous locus but with a WT GFP-K13.Each small dot represents the measured value from one focus in one parasite.Large dots of the same color as small dots represent mean of the respective measured values and each large dot represents one biological replicate.Each biological replicate (large dot) consists of 20 individual measurements (small dots).P values derive from comparing means (large dots) by one-way ANOVA.Example fluorescence microscopy images are shown.Scale bar: 5 µm.(B) Mean of parasite survival in RSA plotted against mean of K13-abundance.Error bars show standard deviations.(C and D) RSA survival of (C) K13 1xNLS parasites or (D) KIC7 1xNLS parasites grown in the presence of 12.5, 25, 125, or 250 nM rapalog for 3 h before and 6 h during the DHA exposure of the RSA.Red bar, mean; standard deviation indicated; each dot derives from an independent experiment.

FIG 4 FIG 5
FIG 4 Fitness cost correlates with resistance: (A) Parasite growth, calculated as fold-change of parasitemia after 96 h compared to parasitemia at 0 h, normalized to WT growth for different cell lines with indicated k13 mutations.Error bars show standard deviation.P values derived from two-tailed unpaired t tests.(B) Mean of growth of strains harboring different k13 mutations plotted against mean of parasite survival in RSA.Error bars show standard deviation.(C) Fitness cost of parasites harboring different k13 mutations grown in competition with 3D7 parasites given as loss of parasite proportion/generation.Error bars show standard deviation.P values derived from two-tailed unpaired t tests.(D) Mean of fitness cost plotted against mean of parasite survival in RSA.All error bars show standard deviations.

FIG 5 (
FIG 5 (Continued) t tests.(B) K13-abundance measured GFP fluorescence intensity of the single focus observed in ring-stage parasites of the indicated K13 expressed from the endogenous locus and fused to GFP, normalized to the fluorescence in parasites with the identically modified endogenous locus but with a WT GFP-K13.Small dots represent measured values for individual parasites.Large dots of the same color as small dots represent mean of the respective measured values and each large dot represents one biological replicate.Dots of the same color were obtained as part of the same experimental repeat.P values derive from two-tailed unpaired t tests.(C) K13 abundance plotted against parasite survival in RSA for K13 C580Y + UBP1 R3138H parasites (yellow).Gray dots show data of k13 mutations from Fig. 3B for comparison.Error bars show standard deviation.(D) Growth, calculated as fold-change of parasitemia after 96 h normalized compared to parasitemia at 0 h for different cell lines with indicated k13 or ubp1 mutations.Error bars show standard deviation.P values derived from two-tailed unpaired t tests.(E) Growth of strains harboring ubp1 R3138H with (yellow) and without (orange) k13 C580Y plotted against parasite survival in RSA.Gray dots show data of k13 mutations from Fig. 4B for comparison.Error bars show standard deviation.(F) Fitness cost for different cell lines with indicated k13 or ubp1 mutations grown in mixed cultures with 3D7.Error bars show standard deviation.P values derived from two-tailed unpaired t tests.(G) Fitness cost K13 C580Y + UBP1 R3138H parasites (yellow) plotted against parasite survival in RSA.Gray dots show data of k13 mutations from Fig. 4D for comparison.All error bars show standard deviations.