Disparate selection of mutations in the dihydrofolate reductase gene (dhfr) of Plasmodium ovale curtisi and P. o. wallikeri in Africa

Plasmodium ovale curtisi and P. ovale wallikeri are both endemic in sub-Saharan Africa, the Middle East and Southeast Asia. Molecular surveillance data for drug resistance in P. ovale spp. is limited at present. We analysed polymorphisms in the podhfr, pocrt and pocytb genes of P. ovale spp. in 147 samples collected from travelers returning to China from Africa. Two podhfr mutations, S58R and S113N/T were detected in P. ovale curtisi with high/moderate frequencies of 52.17% and 17.39%, respectively. Evidence of positive selection (dN/dS = 2.41) was found for podhfr in P. ovale curtisi and decreased diversity (He) of microsatellite markers flanking the mutant alleles suggests that selective sweeps have occurred for both. Mutations E34G (1.50%) and L43V (1.50%) in pocrt of P. ovale curtisi, and E34G (3.70%), I102M (1.80%) and V111F (1.80%) of P. ovale wallikeri were found at low frequencies. Mutations R66K (6.20%), R75K (11.63%) and R95K (3.88%) of pocytb were found in both P. ovale curtisi and P. ovale wallikeri. These results suggest that the podhfr gene of P. ovale curtisi may be subject to drug selection in Africa, warranting further attention. We observed significant differences in the prevalence and distribution of podhfr mutations between the two P. ovale species, suggestive of fundamental biological differences between them.


Introduction
Malaria remains a major global public health threat with an estimated 241 million cases and 627,000 deaths reported worldwide in 2020, with more than 90% of these cases occurring in Africa [1]. There are six species of malaria parasite that commonly cause malaria in humans: Plasmodium falciparum, Plasmodium vivax, Plasmodium malariae, Plasmodium knowlesi, Plasmodium ovale curtisi and Plasmodium ovale wallikeri. The latter two parasites were previously considered to be a single species, P. ovale, but are now known to be separate species. The prevalence of P. ovale spp. is probably underestimated due to low parasite densities and difficulties in detection by microscopy [2,3].
Global malaria control and elimination has made great progress. There remain, however, many challenges among which antimalarial drug resistance is of major concern [4]. Malaria parasite resistance to antimalarial drugs manifests in delayed or incomplete clearance of parasites during their proliferation in the blood of the host. This reduction in susceptibility allows drug tolerant and resistant parasites to persist in the blood post-treatment, where they may produce gametocytes capable of infecting mosquitoes.
Multi-drug resistant strains of both P. falciparum and P. vivax have been described. Following the emergence of Chloroquine (CQ) resistant P. falciparum, sulfadoxine-pyrimethamine (SP) was used extensively as a safe and inexpensive alternative first-line treatment in many malaria-endemic areas [5]. However, use of this drug has been hampered by the development and spread of drug resistance [6], and it has been replaced by artemisinin combination therapies (ACTs) in most countries. Despite this, the WHO still recommends its use as an intermittent preventive treatment during pregnancy (IPTp) and in infants (IPTi) in sub-Saharan Africa [7]. SP belongs to the antifolate drug class which target the folate-synthesis pathway of Plasmodium parasites. Dihydrofolate reductase (DHFR) plays an important role in the synthesis of folic acid by malaria parasite [8]. Pyrimethamine binds to DHFR, competitively excluding its natural substrate, dihydrofolate [8]. Point mutations in the P. falciparum DHFR gene (dhfr) reduce the affinity between DHFR and pyrimethamine, reducing the latter's ability to inhibit folic acid synthesis, leading to parasite resistance to the drug [8]. Three mutations N51I, C59R, and S108N in pfdhfr are now widespread in Africa [9].

PLOS NEGLECTED TROPICAL DISEASES
Although SP is not and never has been recommended to treat patients with P. ovale spp. malaria, it is possible that the presence of the drug in populations in which both P. falciparum and P. ovale spp. are prevalent may select for resistance in the latter parasite by proxy [10].
At present, in vitro culture of P. ovale spp. is not possible, and in vivo drug efficacy testing is very difficult. The only viable approach to assaying whether P. ovale spp. parasites may have developed resistance against anti-malarial drugs is to detect the prevalence of mutations in orthologs of genes associated with drug resistance in other Plasmodium species. Currently, molecular investigations of drug resistance markers in P. ovale spp. are limited. We analyzed the prevalence of genetic polymorphisms in podhfr,pocrt and pocytb, the orthologues of which are known to be associated with drug resistance in P. falciparum, in P. ovale spp. samples collected from imported malaria cases returning from Africa to China.

Ethics statement
This study was reviewed and approved by the Institutional Ethics Committee of Jiangsu Institute of Parasitic Diseases (JIPD). The written informed consent was obtained from all participants.

Sample and data collection
Whole blood samples were collected from P. ovale spp. positive migrant workers returning to Jiangsu province of China from Africa between 2012 and 2016. Jiangsu Province is located in east China with only imported malaria cases reported since 2012 [11]. Case epidemiological information was collected from the web-based China Routine Disease Surveillance Information System (CRDSIS) and Jiangsu Provincial Malaria Weekly [12]. Plasmodium ovale spp. genomic DNA was extracted from 200μl whole blood using the QIAamp DNA Mini Kit (Qiagen Valencia, CA, USA) according to the manufacturer's protocol. Plasmodium infection was initially diagnosed by microscopy and confirmed by species-specific nested PCR [13] in Jiangsu Provincial Reference Laboratory of Malaria Diagnosis. Only P. ovale spp. single species infections were included in this study. Plasmodium ovale spp. samples were genotyped to discriminate between of P. ovale curtisi and P. ovale wallikeri using Real-time PCR as previously described [14].

Podhfr-flanking microsatellite analysis
Five microsatellite loci flanking the podhfr gene (upstream: -47kb, -33.5kb; downstream: 0.3kb, 33.1kb and 39.2kb) were used to identify potential selective sweeps associated with the spread of drug resistant alleles of podhfr. Microsatellite loci within a flanking region of 100kb around the podhfr gene of P. ovale curtisi (pocdhfr) were screened using Tandem Repeats Finder [16]. Fluorescently labeled primers and PCR conditions for microsatellite genotyping are provided in S2 Table. Products labelled with Hex fluorophore were separated and detected on an ABI 3730 capillary sequencer (Thermo Fisher Scientific, US) by Sangon Biotech (Shanghai, China). DNA fragment sizes were determined and visualized using GeneMapper software Version 3.2 (Applied Biosystems, Foster City, CA, USA). The dominant allele (highest intensity peak) at each locus, and any additional alleles with peak height � 33% of the dominant allele peak heights of > 200 rfu (fluorescence units) were scored [17]. Genotyping success was defined as the presence of at least one allele at a given locus in a given sample.

Population genetics analysis
The number of single nucleotide polymorphisms (SNPs), average pairwise nucleotide diversity (π), haplotype diversity (Hd) and number of haplotypes (H) were calculated using DnaSP Version5.10.01 [18]. The presence of selection was analyzed by comparing the ratio of non-synonymous to synonymous mutations. Tajima's D, Fu and Li's D � , and Fu and Li's F � tests were analyzed to measure the degree of deviation from neutrality using DnaSP Ver-sion5. 10.01 [18]. The rates of non-synonymous mutations (d N ) and synonymous mutations sites (d S ) were computed by Z-tests based on the Nei & Gojobori method with the Jukes and Cantor correction and 1000 bootstrap replications using MEGA Version6.06 [19]. The diversity of microsatellites flanking podhfr was estimated by the expected heterozygosity (H e ) using GenAlEx Version6.5 [20], where H e = [n/(n-1)](1-Sp i 2 ) and n represents the number of isolates and P i is the frequency of the ith allele. Microsatellite polymorphism parameters including allele number and expected heterozygosity (H e ) were calculated for different geographical regions and mutant strains. Mann-Whitney U and Wilcoxon rank sum tests were used to determine the significance of differences in H e (P<0.05). Pair-wise F st was employed as a proxy of the genetic distance between pairs of populations and was calculated using Arlequin Version 3.5 [21].

Polymorphisms in podhfr, pocrt and pocytb
Nine distinct mutations were found in podhfr, among which the most common were S58R (36/69, 52.1%) and S113N (10/69, 14.5%) with higher frequencies observed in P. ovale curtisi than P. ovale wallikeri ( Table 2 and Fig 1A). Most of the S58R mutations (32/36, 88.9%) in P. ovale curtisi were from Central Africa, among which 20 were from Equatorial Guinea, five Most of the S113N mutations (8/10, 80%) in P. ovale curtisi were also from Central Africa with five samples from the Democratic Republic of the Congo and three samples from the Republic of Congo. The frequency of S113N mutation in Central Africa (18.6%, 8/43) was slightly higher than Southern Africa (16.7%, 2/12). All 10 samples of P. ovale curtisi with S113N mutations were S58R/S113N double mutants. Seven and five podhfr haplotypes were found with haplotype diversities (Hd) of 0.681 and 0.192 in P. ovale curtisi and P. ovale wallikeri, respectively (Table 3).
For pocrt, four mutations were found, all at low frequencies. E34G and L43V were observed in only one sample each, and only in P. ovale curtisi.E34G was observed in two samples of P. PLOS NEGLECTED TROPICAL DISEASES ovale wallikeri. I102M and V111F were observed in only one sample each and only in P. ovale wallikeri ( Table 2 and Fig 1B). For pocytb, four mutations were found, with R75K present at the highest frequency of 12.73% in P. ovale curtisi and 10.81% in P. ovale wallikeri ( Table 2 and Fig 1C). More details regarding the population genetics of pocrt and pocytb are given in Table 3. Fig 2 shows the proportion of mutant alleles of P. ovale curtisi and P. ovale wallikeri in isolates from West, Central and Southern Africa. The individual mutations S58R and S113N were  observed in both species. In P. ovale curtisi, S58R was observed both as a single mutation and in combination with S113N, whereas in P. ovale wallikeri, S58R only appeared in combination with F57L in a double mutant allele, and with F57L and S113T as part of a triple mutant allele. All other mutations were unique to one or other of the species. The P. ovale curtisi population from Central Africa was the only one in which mutant alleles were more prevalent than wild types, with 74% (32/43) of isolates carrying either S58R single mutations or S58R + S113N double mutations. The former allele was also observed (at lower prevalence) in both West and Southern Africa, making it the most geographically widespread mutant allele. Mutant alleles were present at a prevalence of 59% (41/69) in P. ovale curtisi, and 10% (7/69) of P. ovale wallikeri isolates.

Selective sweeps involving podhfr mutations in P. ovale curtisi
The d N /d S ratio for P. ovale curtisi podhfr was 2.41, indicating the presence of positive selection pressure. pocrt and pocytb did not appear to be under the influence of positive selection, and there were no significant differences in neutral evolutionary statistics (Table 3). Based on analysis of five microsatellite loci, the genetic diversity of podhfr S58R mutant strains (H e = 0.575 ±0.069) was significantly lower than that of wild-type strains (H e = 0.682±0.045, P = 0.011) in P. ovale curtisi (Fig 3A and S4 Table). The genetic diversity of S113N/T mutant strains (H e = 0.511±0.074,) was significantly lower than that of wild-type strains (H e = 0.661±0.060, P = 0.001) in P. ovale curtisi (Fig 3B and S4 Table). This decreased genetic diversity in loci surrounding mutant alleles suggests a selective sweep has occurred involving podhfr gene mutations in P. ovale curtisi. The number of alleles for each microsatellite marker in P. ovale spp. isolates of each population is shown in S3 Table. Microsatellite polymorphism was relatively similar in different regions / countries in Africa (Fig 3C and 3D and S4 Table). The genetic differentiation coefficients of P. ovale spp. strains among Southern Africa, West Africa and Central Africa differed. At the regional level, the genetic differentiation between West Africa and Central Africa (F st = 0.038, P >0.05) was the highest (S5 Table). At the country level, the genetic differentiation between Nigeria and Equatorial Guinea (F st = 0.041, P >0.05) was the highest (S6 Table). There was significant genetic differentiation between S58R mutant strains and the wild-type strains (F st = 0.103, P<0.05) (S7 Table), and between the S113N/T mutant strains and the wild-type strains (F st = 0. 087, P<0.05) (S8 Table).

Discussion
We analyzed genetic polymorphisms in P. ovale dhfr, crt and cytb in 147 samples collected from imported cases of malaria in migrant workers returning to China from Africa. These imported cases were of non-immune migrant workers (and not short-term travelers), who had resided in African countries for a relatively long periods of time, which enabled us to sample a large number of parasites from a widespread geographical area. Furthermore, these workers spent extended time periods in Africa, and would have experienced correspondingly longer exposure times than short-term travelers.
Overall, a low level of genetic polymorphism in these drug resistance-related genes were observed. We observed the highest level of polymorphism in the dhfr gene (π = 0.00118), and this polymorphism was significantly higher in P. ovale curtisi (π = 0.00173) compared to P. ovale wallikeri (π = 0.00062). Non-synonymous mutations at the 58th and 113th amino acids positions of dhfr were observed in both P. ovale spp. sub-species. Furthermore, S58R and S113N were detected in P. ovale curtisi with high/moderate frequencies of 52.17% and 17.39%, respectively, but were rare in P. ovale wallikerii which may be related to biological differences between the species that lead to differing susceptibilities to drugs. These differences may inform different public health policies targeting these two P. ovale spp. species. The S58R mutation has also been reported in other Plasmodium species. For example, three mutations (F57L, S58R, S117N) in P. vivax were found with high frequency in cases from Thailand and were considered to be associated with high levels of SP resistance through patient follow-up [22]. Seven mutations in dhfr (K55E, S58R, S59A, F168S, N194S, D207G and T221A) were also identified in 123 samples of P. malariae from Asia and Africa [23].
It is now established that the parasite species previously known as 'Plasmodium ovale' is in fact two separate species, which, despite their clear separate species status are still referred to with the sub-species names P. ovale wallikeri and P. ovale curtisi. They are indistinguishable by microscopy, but have been shown to constitute sympatric and non-recombining species. Their ranges overlap, and they are sometimes found in co-infections of the same individual. To date, there is little evidence of major differences between the species phenotypically [24], but genotypic studies indicate distinct genetic lineages with little evidence of introgression [25]. In this context, our findings of distinct mutation profiles in the dhfr gene of both species, and of significantly greater prevalence of mutations in P. ovale curtisi compared to P. ovale wallikeri, perhaps constitute the strongest evidence to date of an obvious biological difference between the species. Striking differences were observed between the species at the dhfr locus, with no allelic types in common between the species, and with a much larger proportion of P. ovale curtisi isolates carrying mutations compared to those of P. ovale wallikeri. These differences were particularly obvious when considering isolates from Central Africa, where the majority of P. ovale curtisi isolates carried mutated dhfr alleles, whereas in P. ovale wallikeri isolated from the same region, parasites carrying wild-type isolates were dominant.
As the two species have previously been shown to be sympatric in this region, we assume that both are likely to have been exposed to antifolate drugs to the same degree. Given this, several possible scenarios for the observed disparity between the species at the dhfr locus present themselves. Firstly, P. ovale wallikeri may develop resistance to antifolates at loci other than dhfr. Some drug resistance phenotypes in malaria parasites involve mutations in multiple genes, and some species develop resistance to the same drug through mutations at different loci (for example, resistance to CQ is mediated through mutations in the chloroquine transporter gene, crt, and the multidrug resistance gene mdr1 in P. falciparum, but not in P. vivax or P. chabaudi [26]. However, a low prevalence of mutations was observed in the dhfr gene of P. ovale wallikeri, including those, such as S58R and S113N, the orthologues of which are known to confer antifolate resistance in other species, suggesting some conservation of resistance mechanisms in this species. A second possible explanation for the relatively low prevalence of mutations in dhfr in P. ovale wallikeri compared to P. ovale curtisi, is that antifolate use in the host population results in less selective pressure for the acquisition of resistance in the former species compared to the latter. This may be due to inherent susceptibility differences between the species, for instance with P. ovale wallikeri being naturally more tolerant to pyrimethamine than P. ovale curtisi. This situation is not without precedent, as Plasmodium yoelii, the malaria parasite of African thicket rats, is known to be far more naturally resistant to chloroquine than its rodent malaria parasite cousin P. chabaudi [27] (both these species, like the P. ovale spp., naturally infect the same mammalian hosts).
Further explanations may include reduced exposure of P. ovale wallikeri to antifolates compared to P. ovale curtisii, perhaps due to range differences or to niche differences within the blood. Furthermore, given that antifolate treatment of P. falciparum is probably the main driver of resistance in P. ovale spp., it is possible that differences in the P. ovale species propensity to co-infect with P. falciparum may influence the acquisition of resistance mutations. Additionally, it is possible that P. ovale wallikeri may suffer from a greater fitness cost of dhfr mutations than P. ovale curtisi. Further studies are required to shed light on this phenomenon.
While the S58R mutation was most commonly observed in single mutation pocdhfr alleles, the S113N mutation was always found in combination with S58R, suggesting that the acquisition of pyrimethamine resistance in P. ovale curtisi may occur in a step-wise manner, with the S58R mutation necessary before S113N is selected. This mirrors the situation in P. falciparum, in which the pfdhfr S108N mutation is thought to be an essential first step prior to the acquisition of further mutations such as pfdhfr C59R. Of note, the order of selection is reversed in P. ovale curtisi compared to P. falciparum, with the pocdhfr S113N mutation (perhaps orthologous to pfdhfr S108N) following after the podhfr S58R mutation (orthologous to pfdhfr C59R). In P. ovale curtisi, the picture is somewhat different, with the S58R and F57L mutations only ever appearing together, and the S113N and S113T alleles appearing exclusively in combination with T62R and F57L + S58R, respectively. It should be stressed, however, that there is no experimental evidence linking the mutations described in podhfr to pyrimethamine resistance. We found regional differences in the distribution of putative pyrimethamine resistance associated mutations in P. ovale curtisi podhfr. Both S58R single mutants and S58R + S113N mutations were more prevalent in Central Africa compared to west and southern Africa. Given the lack of polymorphism found in microsatellite markers linked to these alleles suggests a single origin and a subsequent selective sweep, it seems likely that these mutant alleles originated in Central Africa from where they are spreading to the rest of the continent. Further surveys of P. ovale spp. populations in Africa are required to test this hypothesis.
It should be noted that antifolates are no longer recommended as treatment for malaria in Africa, so it is possible that rather than observing an ongoing selection of resistance-associated mutations, we may actually be witnessing the reverse; the re-emergence and spread of wildtype parasites in the absence of antifolate drug pressure. However, it is impossible to assess the impact of ongoing antifolate use in the populations sampled here, either for non-malaria related reasons such as prophylaxis among HIV infected people, or as non-officially prescribed treatments for malaria. Clarifying this situation will require further sampling of the P. ovale spp. populations in the future, combined, perhaps, with retrospective analysis of archived samples.
The d N /d S ratio of P. ovale curtisi podhfr in was 2.41, indicative of the action of directional selection on this gene. Previously, microsatellites markers have been used to measure the degree of genetic polymorphism at loci around genes linked to drug resistance, in order to identify potential 'selective sweeps', in which a particular allele of a gene becomes dominant in a population following its emergence in a single, or limited number of parasite(s). For example, a study involving the analysis of four microsatellite loci flanking the dhfr gene from 50 P. falciparum strains in Ghana revealed a selective sweep of the mutant allele probably driven by the use of SP [28]. In another study, twenty-six microsatellite markers flanking the dhfr gene were used to study P. falciparum strains collected in Southeastern Africa [29]. It was found that genetic diversity within 70 KB of the allele associated with the highest degree of resistance was significantly reduced; evidence of a selective sweep following the widespread use of SP.
In order to further explore whether the podhfr gene of P. ovale curtisi has been subjected to drug selection pressure, we analyzed the diversity of five microsatellite loci flanking the gene. Samples of P. ovale curtisi were divided into different populations based on their geographical origin and mutation sites, and their diversity analysed. We found that allele diversity did not differ significantly between populations, but there was a greater degree of diversity in microsatellite markers flanking wild-type alleles than those flanking mutant alleles.
There were no statistical differences in expected heterozygosity at each microsatellite locus between the three geographic regions. The expected heterozygosity of microsatellite loci flanking the S58R mutant allele was significantly lower than those flanking the wild-type allele. Similarly, the expected heterozygosity of markers flanking the S113N/T mutant was lower than those of the wild-type allele. This reduced heterozygosity around the mutant alleles is evidence selective sweeps, probably driven by exposure to antifolate drugs. Both the S58R and S113N/T mutations were most commonly found in Central and West Africa, which may be related to the drug policy in these regions (CQ was replaced by SP for intermittent preventive treatment in sub-Saharan African countries).
Point mutations at codons 72, 74, 75, and 76 in the pfcrt gene confer resistance to CQ in P. falciparum [30], and the mutations C101F, H97Y, F145I, M343L, G353V are associated with Piperaquine (PQ) and Amodiaquine (AMQ) resistance [31,32]. At present, there are no reports on polymorphisms of the pocrt gene in P. ovale spp. We found four non-synonymous mutations, E34G, L43V, I102M and V111F in pocrt, all at low frequencies. Orthologs of these mutations have not been described in P. falciparum. Mutations in the pfcytb gene confer drug resistance to atovaquone-proguanil in P. falciparum [33]. In a previous study only one nonsynonymous cytb mutation (M248I) was found in 19 P. ovale spp. samples from Africa [34]. We found three non-synonymous mutations, R66K, R75K and R95K, which were present at relatively high frequencies. The orthologue of the P. falciparum atravaquinone resistance associated mutation at position 268 was not found. Selection pressure analysis showed that the pocytb gene was under purifying selection (d N /d S = 0.276) in P. ovale curtisi.
Although a relatively small number of samples were analysed in this study and sample sources are unevenly distributed throughout Africa, we found evidence of positive selection on podhfr, and of the signature of selective sweeps around the mutant alleles. Despite the fact that the mutations observed have orthologues in P. falciparum that are known to be associated with resistance against anti-folate drugs, there is currently no evidence to link the P. ovale spp. mutations to drug resistance. Further in vitro or in vivo drug assays are required to verify the involvement of these mutations in P. ovale spp. response to drug treatment.

Conclusion
High frequencies of S58R and S113 N/T mutations were observed in podhfr in P. ovale curtisi isolates from Africa, and evidence of selective sweeps around these mutant loci were recorded. The podhfr gene in P. ovale curtisi may, therefore, have undergone selection pressure due to exposure to antifolate drugs. Further P. ovale spp. resistance surveillance in Africa is recommended.
Supporting information S1