Evidence of a Mild Mutator Phenotype in Cambodian Plasmodium falciparum Malaria Parasites

Malaria control efforts have been continuously stymied by drug-resistant strains of Plasmodium falciparum, which typically originate in Southeast Asia prior to spreading into high-transmission settings in Africa. One earlier proposed explanation for Southeast Asia being a hotbed of resistance has been the hypermutability or “Accelerated Resistance to Multiple Drugs” (ARMD) phenotype, whereby multidrug-resistant Southeast Asian parasites were reported to exhibit 1,000-fold higher rates of resistance to unrelated antimalarial agents when compared to drug-sensitive parasites. However, three recent studies do not recapitulate this hypermutability phenotype. Intriguingly, genome sequencing of recently derived multidrug-resistant Cambodian isolates has identified a high proportion of DNA repair gene mutations in multidrug-resistant parasites, suggesting their potential role in shaping local parasite evolution. By adapting fluctuation assays for use in P. falciparum, we have examined the in vitro mutation rates of five recent Cambodian isolates and three reference laboratory strains. For these studies we also generated a knockout parasite line lacking the DNA repair factor Exonuclease I. In these assays, parasites were typed for their ability to acquire resistance to KAE609, currently in advanced clinical trials, yielding 13 novel mutations in the Na+/H+-ATPase PfATP4, the primary resistance determinant. We observed no evidence of hypermutability. Instead, we found evidence of a mild mutator (up to a 3.4-fold increase in mutation rate) phenotype in two artemisinin-resistant Cambodian isolates, which carry DNA repair gene mutations. We observed that one such mutation in the Mismatch Repair protein Mlh1 contributes to the mild mutator phenotype when modeled in yeast (scmlh1-P157S). Compared to basal rates of mutation, a mild mutator phenotype may provide a greater overall benefit for parasites in Southeast Asia in terms of generating drug resistance without incurring detrimental fitness costs.


Construction of the P. falciparum exo1Δ mutant
To generate a P. falciparum exo1Δ knockout strain, we generated the pcamBSD-exo1 plasmid that was used to genetically disrupt the pfexo1 (PF3D7_0725000) open reading frame. First, a 1187 bp fragment of pfexo1 encoding NotI and PstI restriction sites was PCR amplified from Dd2 gDNA using primers 5'-CTTGCGGCCGCGTAATGATTTTCATATAACTGGTATGGG-3' and 5'-CTTCTGCAGCATCGAAAGTATGTTCACACGTTCCG-3'. This NotI and PstI digested fragment was cloned into pcamBSD [23], which encodes the blasticidin-S-deaminase (bsd) selectable marker, yielding pcamBSD-exo1. Dd2 parasites were transfected with purified circular pcamBSD-exo1 plasmid as described [22]. One day post-electroporation, parasites were exposed to 2 μg/ml of blasticidin (Invitrogen). Blasticidin media was replaced daily until day 7 post-electroporation and every 2 days thereafter. Transformed parasites were detected by light microscopy 11 days post-electroporation, maintained in blasticidin-containing media, and periodically checked by PCR for single site crossover-based plasmid integration into the pfexo1 locus. This PCR screen used the primers 5'-TCACTAAAGGGAACAAAAGCTGG-3' and 5'-AAGGTCATCCTTCTTTTCCCAC-3'. Plasmid integration was detected on day 31 postelectroporation. The mixed culture was then cloned by limiting dilution in a 96-well plate [24]. To detect clones, plates were screened by flow cytometry on an Accuri C6 flow cytometer coupled to a HyperCyt autosampler, with infected red blood cells (iRBCs) labeled using 1.6 μM Mito Tracker Deep Red and 2× SYBR Green (Invitrogen) [25]. Integration-positive clones were confirmed by PCR.

KAE609 fluctuation assays
For a schematic representation of the fluctuation assay, see Fig 1A. On day 0, each parental parasite strain was first cloned by limiting dilution. Single clones were expanded to~5×10 8 total iRBCs and then seeded into two 96-well plates. Stocks were also cryopreserved at this time. The first plate was inoculated with 2.5×10 8 total iRBCs of a parental clone at~3% hematocrit in selective media containing 2.5 nM KAE609 (kindly provided by Dr. Thierry Diagana, Novartis Institute of Tropical Diseases) [19]. The second plate (0.8 parasites/well) was a control in which a parental clone was cloned again by limiting dilution in a 96-well plate (without drug pressure) to calculate an accurate starting culture density (number of iRBCs/ml). KAE609-containing selective media (with 0.9% hematocrit) was added to drug-pressured 96-well plates daily for 4 days, followed by the addition of fresh selective media (without RBCs) every 48 hours. Fresh RBCs were added on days 14, 17, and 20. Drug-resistant clones were detectable by flow cytometry generally around day 20 post-inoculation (matching the growth rate for unpressured clones).

Fluctuation assay mutation rate calculation
Mutation rates were calculated for both parasite and yeast fluctuation assays using the Ma-Sandri-Sarkar Maximum Likelihood Estimator (MSS-MLE) Method via the online Fluctuation Analysis Calculator (FALCOR) [26]. In this analysis, r is the total count of KAE609-resistant clones per parental clone tested, and N t is the corrected starting number of parasites determined from each clone's control plate. MSS-MLE is considered to be a more accurate method of calculating mutation rates as compared to the Lea-Coulson or other methods [27]. Statistical analysis was performed using Prism 6 (GraphPad Software, Inc.).

In vitro susceptibility assays
In vitro IC 50 values were measured by incubating parasites for 72 hours across a range of KAE609 concentrations (0.2-100 nM). Final parasitemia were measured by flow cytometry and IC 50 values were calculated by non-linear regression analysis [25]. clones. Two-tailed Student's t-tests were performed for each strain relative to 3D7 with the exception of Dd2 exo1Δ, which was compared to Dd2, its isogenic background parent (#). 95% confidence intervals (lower and upper limits) are denoted in brackets. Fold changes (fc) are listed relative to 3D7. Dd2 exo1Δ is compared to Dd2. * p = 0.02, ** p < 0.01.

lys2ΔBgl fluctuation assays
All yeast strains were propagated for three days on YEPD+G418 plates. Single clones were inoculated in 10 ml of YEPGE+G418 liquid medium and propagated for 2 days at 30°C with constant shaking. Cells were harvested by centrifugation, washed once with ddH 2 O, and resuspended in 1 ml of ddH 2 O. Dilutions were plated onto SC-Lys plates for 3 days at 30°C to select for Lys + revertants, or control YEPD plates to calculate an accurate starting inoculum. Nine clones were tested for SJR0538-WT and SJR0538-P157S, and eleven for SJR0538-K30A-E31A.

Some Cambodian isolates exhibit a mild mutator phenotype
To determine mutation rates for a diverse selection of P. falciparum strains, we adapted the fluctuation assay, a common microbiological method to measure and compare mutation rates [30][31][32], to P. falciparum. Fluctuation assays estimate mutation rates based on the distribution of mutant clones that arise in parallel cultures subjected to selective pressure.
In our P. falciparum-adapted fluctuation assay (Fig 1), we pressured parasites with the spiroindolone compound KAE609 (also known as NITD609 or cipargamin) [33]. We chose KAE609 for three major reasons: (1) Resistance to low-level (2.5 nM) KAE609 can be readily conferred by a number of unique SNPs in pfatp4 [19,[34][35][36]. Given more than one independent SNP in pfatp4 can yield resistant clones, we were therefore able to capture a broad range of mutation rates, including for low or basal rates (e.g. that of the drug-sensitive reference 3D7 strain). In the original ARMD report, only one of five strains pressured with 5-FOA yielded a frequency of resistance measurement [8]. (2) Using KAE609 is much less time and resource intensive than using drugs that select for resistance at much lower frequencies (3). KAE609 is also fast acting [37]. Mutations that impart resistance would therefore have to be present at the time of drug exposure in order for parasites to survive. This speed of action was preferred to slower acting agents that might allow resistance-conferring SNPs to be encoded beyond the first generation of selective pressure, which would yield inflated mutation rates.
In our KAE609 fluctuation assays (Fig 1), we assayed 3D7, W2, and Dd2, which have been shown to have similar basal mutation rates [10,11], and therefore served as wild-type controls. As a benchmark for a mild mutator phenotype, we generated and assayed a P. falciparum Exonuclease I knockout strain (Dd2 exo1Δ). exo1Δ strains exhibit mild mutator phenotypes in other organisms, as Exonuclease I often mediates the fastest, although non-essential, mechanism for many DNA repair pathways [30,38]. We also tested V1/S, a multidrug-resistant Vietnamese strain [39,40] that carries two DNA repair gene mutations (Table 1) found in contemporary artemisinin-resistant Cambodian isolates, suggesting that it too may exhibit a mild mutator phenotype. Finally, we tested five recently derived Cambodian isolates that encode various DNA repair mutations [9]. These isolates were chosen to encompass the potential contributions of DNA repair mutations in Cambodian isolates. For instance, PH0167-C is artemisinin-sensitive and encodes no mutations overrepresented in artemisinin-resistant isolates (Table 1). PH0167-C nonetheless harbors two unique DNA repair gene mutations, not associated with artemisinin resistance. In contrast, PH0212-C encodes four mutations associated with artemisinin resistance, in particular for the two factors Mlh1 and Pms1 that are both essential for MMR.
Parallel cultures of each strain were pressured and maintained in 2.5 nM KAE609 in 96-well plates until resistant clones were detectable by flow cytometry (See Materials and Methods). To confirm that our fluctuation assays yielded KAE609-resistant clones, we determined halfmaximal inhibitory concentration (IC 50 ) values using 72 hour in vitro susceptibility assays for each parent and two randomly selected drug-pressured clones. Sanger sequencing of the pfatp4 gene from resistant clones yielded 16 total mutations, 13 of which have not been previously reported ( Table 2).
As mentioned above, the polymorphic nature of pfatp4 in response to KAE609 allowed us to calculate mutation rates for all strains, including wild-type strains, which was previously not possible for 5-FOA [8]. It is important to note that these mutation rates are specific to KAE609, for which resistance in vitro can be mediated by single SNPs, and are useful for calculating relative differences between strains. However, the absolute KAE609 mutation rates should not be directly compared with those calculated using other drugs or methods of calculation unless the studies are conducted in parallel. As examples, selection for mefloquine resistance typically results in pfmdr1 gene amplification [41], for chloroquine multiple mutations in pfcrt are required [42], and for atovaquone mutations involve single SNPs in the genome of mitochondria, which are present as multiple copies per parasite [43].
With all strains, mutation rates were compared to 3D7, except for Dd2 exo1Δ that was compared to its parental Dd2 strain. These data revealed no significant differences between 3D7, W2, and Dd2 in their KAE609 mutation rates. This result was not consistent with the original ARMD study, which reported that W2 was hypermutable relative to 3D7 [8]. We also observed no significant differences in mutation rates amongst the three Cambodian isolates PH0167-C, PH0306-C, and PH0164-C that carried fewer DNA repair gene mutations compared to the two other isolates (Table 1, Fig 1B). These observations provide additional evidence against an ARMD phenotype being present in the W2 strain.
Results with our Dd2 exo1Δ line showed a statistically significant 2.5-fold increase in mutation rate relative to its isogenic background parent Dd2. This is similar to yeast exo1Δ strains, which exhibit a 3-fold increase in mutation rate in lys2ΔBgl fluctuation assays [44], and confirms the usefulness of Dd2 exo1Δ as a benchmark mild mutator in P. falciparum fluctuation assays.
We observed mild mutator phenotypes in three additional strains. V1/S exhibited a nonstatistically significant 1.7-fold increase in mutation rate relative to 3D7. However, its mutation rate was very similar to Dd2 exo1Δ. PH0482-C and PH0212-C exhibited statistically significant 2.3-and 3.4-fold increases relative to 3D7, respectively ( Fig 1B). As stated above, PH0482-C and PH0212-C encode the most DNA repair gene mutations that cluster with artemisinin resistance. These data suggest some correlation between DNA repair gene mutations and the mild mutator phenotype, which merits further investigation. Table 1. List of strains used in this study and their DNA repair gene haplotypes associated with artemisinin-resistance. Haplotypes determined using SNP dataset available on PlasmoDB. c Reported in [20]. d Reported in [21]. e Haplotype is associated with artemisinin resistance in Cambodia, reported in [9]. f Variant residue reported in [9]. WT = wild-type. KO  Our data also yielded useful insights into KAE609. As stated above, we observed 13 novel pfatp4 mutations in KAE609-resistant clones (S1 Fig). Prior studies using KAE609 or other PfATP4 inhibitors have to date reported 24 mutations in this gene [19,35,36,45,46]. The polymorphic nature of pfatp4 in response to low drug concentrations exemplifies its ideal use for fluctuation analyses in P. falciparum. In vitro KAE609 susceptibility experiments, however, cannot recapitulate the pharmacokinetic and pharmacodynamic properties of this compound in vivo. Indeed, efficacy studies in humans will be required to assess the parasitological and clinical risk associated with the polymorphic nature of pfatp4. Pharmacokinetic studies in humans show that a 300 mg dose of KAE609 can produce a maximum plasma concentration of around 2 μg/ml in 3 to 8 hours [47]. This equates to roughly 5 μM KAE609, which is 2,000-fold higher than the 2.5 nM concentration used in our drug pressure experiments. Also, the largest changes in IC 50 values attributed to single or double mutations in response to 2.5 nM KAE609 so far have been only 20 to 30-fold ( [19] and Table 2).

The Cambodian mutant Mlh1 allele exhibits a minor mutator phenotype in yeast
We sought to model the DNA repair gene mutations found in artemisinin-resistant parasites in a yeast system to test for a possible association between these mutations and mutator phenotypes. Among these mutations, only one gene, the MMR factor Mlh1, had a mutation (pfmlh1-P203S) that could be mapped to a yeast ortholog [48], in this case Mlh1 (scmlh1-P157S) (Fig  2A). The ability to map the Mlh1 mutation is serendipitous for a number of reasons. First, Mlh1 is an essential MMR protein that plays a major role in the generation of SNPs [18,49]. Second, the pfmlh1-P203S mutation is found in the mild mutator P. falciparum strains PH0212-C and PH0482-C. Third, the homologous residue in yeast, scmlh1-P157, is located in a conserved region near motif IV of the essential N-terminal ATPase domain (Fig 2A). Fourth, in yeast the mutation to lysine at residue P157 produces a strong mutator phenotype in an exo1Δ background, which amplifies the mutator effect [30]. A mutation at the homologous residue in human MLH1 has also been associated with hereditary nonpolyposis colorectal cancer [50], a cancer mediated by mutations in MMR genes. Finally, among the 3,000 P. falciparum isolates sequenced in the Pf3k database, the pfmlh1-P203S mutation has only been found in Southeast Asia (in Cambodia, Myanmar, Thailand, and Vietnam), and never in Africa or Bangladesh [51].
To test the effect of scmlh1-P157S, we transformed a Mlh1 knockout yeast strain (SJR0538: lys2ΔBgl mlh1Δ) with plasmids encoding either wild-type scmlh1-WT, or scmlh1-P157S, or scmlh1-K30A E31A. In yeast, scmlh1-WT was shown to fully rescue MMR functionality, whereas the ATPase-dead scmlh1-K30A E31A plasmid phenocopied a full Mlh1 knockout [52]. To test for mutability differences, each yeast strain also encoded an internal frameshift in LYS2 (lys2ΔBgl), which rendered the strain auxotrophic for lysine. Reversion to lysine prototrophy requires a return to wild-type functionality, but not necessarily wild-type sequence, as a number of different reversion events in lys2ΔBgl can restore function. Therefore, the lys2ΔBgl fluctuation assay measures an artificially increased mutation rate [28], similar in concept to our KAE609 assay, as in both assays several distinct genetic changes can produce the selected phenotype. For each of the three scmlh1 strains, we measured mutation rates by measuring the return to growth on media lacking lysine (Fig 2B).
Given the conservation of the region in which it lies, the mlh1-P203S mutation in P. falciparum likely also mediates a mild mutator phenotype, particularly if compounded with additional DNA repair mutations (e.g. Pms1). Gene editing experiments are currently ongoing to examine the contributions of MMR mutations to mild mutator phenotypes in P. falciparum.
In our study, we have focused on the effects of DNA repair gene mutations in the intraerythrocytic, asexual cycle. It is tempting to consider that the effects of these mutations might also extend to the sexual stages that genetically recombine in the Anopheles mosquito host. In meiosis, for instance, the complex of Mlh1-Mlh3 (MutLγ) and Exo1 has been shown to resolve meiotic crossovers [53]. Additionally, the Mlh1-Pms1 complex (MutLα) prevents post-meiotic segregation and divergence from Mendelian inheritance [54]. We posit that P. falciparum DNA repair mutations might reduce meiotic recombination and thereby contribute to the maintenance of multigenic drug-resistant traits in Southeast Asian strains. Such a mechanism would be consistent with the over-representation of these DNA repair genes in artemisininresistant parasites in Cambodia, which comprise genetically distinct parasite subpopulations whose primary resistance determinant K13 appears to be in linkage disequilibrium with multiple genetic traits on distinct chromosomes [9]. Dissecting the role of these DNA repair genes in meiosis will require the investigation of rates of chromosomal crossovers in gene-edited P. falciparum parasites examined in Anopheles mosquitoes, a complex but exciting undertaking.
Since the ARMD phenotype was initially described [8], examining whether this is a consistent characteristic of multidrug-resistant parasites has focused on finding hypermutators with very large variations in mutability rates compared to drug-sensitive parasites. Studies in natural bacterial populations present an alternative perspective. Hypermutator strains can and do acquire resistance to drugs quickly in vitro, but when propagated longer become outcompeted by mild mutators due to an accumulation of additional detrimental mutations [17,18]. Mild mutators can more readily provide an appropriate balance between the ability to adapt to drug pressure via the acquisition of resistance-conferring mutations and the need to maintain overall fitness [14,16,18]. This may have been important for the acquisition of resistance to earlier first-line antimalarial therapies such as CQ or pyrimethamine+sulfadoxine, or current firstline artemisinin-based therapies, which require multiple mutations in one or more genes to build full resistance [6,42,55]. Further research will be necessary to determine to what extent a mild mutator phenotype in P. falciparum facilitates the acquisition of multidrug resistance in parasite populations, as has been observed in bacteria.

Conclusion
We provide evidence of a mild mutator phenotype in some contemporary Cambodian isolates, using a fluctuation assay adapted for use in P. falciparum. These strains are artemisinin-resistant and carry multiple DNA repair gene mutations. One such mutation, pfmlh1-P203S, drives a mild mutator phenotype when modeled in yeast. In describing this phenotype, we provide further evidence that the ARMD phenotype cannot be recapitulated, under the drug selection conditions adopted herein. As such, we posit that the focus in P. falciparum should be shifted from the ARMD phenotype to a more detailed examination of mild mutator phenotypes, which have been shown to be more relevant than hypermutators for antibiotic resistance and evolution of natural bacterial populations [13][14][15][16].