Unexpected selections of Plasmodium falciparum polymorphisms in previously treatment-naïve areas after monthly presumptive administration of three different anti-malarial drugs in Liberia 1976–78

Background To assess the effect on malaria prevalence, village specific monthly administrations of pyrimethamine, chlorproguanil, chloroquine or placebo were given to children in four previously treatment-naïve Liberian villages, 1976–78. Plasmodium falciparum in vivo resistance developed to pyrimethamine only. Selection of molecular markers of P. falciparum resistance after 2 years of treatment are reported. Methods Blood samples were collected from 191 study children in a survey in 1978. Polymorphisms in pfcrt, pfmdr1, pfdhfr, pfdhps, pfmrp1 and pfnhe1 genes were determined using PCR-based methods. Results Pfcrt 72–76 CVIET was found in one chloroquine village sample, all remaining samples had pfcrt CVMNK. Pfmdr1 N86 prevalence was 100%. A pfmdr1 T1069ACT→ACG synonymous polymorphism was found in 30% of chloroquine village samples and 3% of other samples (P = 0.008). Variations in pfnhe1 block I were found in all except the chloroquine treated village (P < 0.001). Resistance associated pfdhfr 108N prevalence was 2% in the pyrimethamine village compared to 45–65% elsewhere, including the placebo village (P = 0.001). Conclusions Chloroquine treatment possibly resulted in the development of pfcrt 72–76 CVIET. Selection of pfmdr1 T1069ACG and a pfnhe1 block 1 genotypes indicates that chloroquine treatment exerted a selective pressure on P. falciparum. Pyrimethamine resistance associated pfdhfr 108N was present prior to the introduction of any drug. Decreased pfdhfr 108N frequency concurrent with development of pyrimethamine resistance suggests a non-pfdhfr polymorphisms mediated resistance mechanism. Electronic supplementary material The online version of this article (doi:10.1186/s12936-017-1747-6) contains supplementary material, which is available to authorized users.


Background
Plasmodium falciparum has historically shown considerable ability to develop resistance to antimalarial drugs. Increased malaria specific morbidity and mortality have been associated with development and spread of quinine (QN), chloroquine (CQ) and sulfadoxine-pyrimethamine (SP) resistance. Reports of resistance to the currently recommended artemisinin-based combination therapy are consequently of major concern and highlight the importance of understanding the evolution of drug resistance [1][2][3][4][5][6].
At least fifteen single nucleotide polymorphisms (SNP) in the P. falciparum chloroquine resistance transporter (pfcrt) gene have been associated with CQ resistance in field isolates [7][8][9]. Specific haplotypes at positions 72-76 have been linked to regional evolution of CQ resistance and the 76T allele appears essential for resistance [10][11][12]. Drug resistance associated SNPs have also been identified in the multidrug resistance gene 1 (pfmdr1) including N86Y, Y184F, S1034C, N1042D, F1226Y and D1246Y. Various combinations of these SNPs have been shown to modulate levels of drug resistance/tolerance to QN, CQ, amodiaquine, mefloquine and lumefantrine [13].
The P. falciparum Na+/H+ exchanger 1 gene (pfnhe1) has been associated with QN resistance [14,15]. Specifically, >1 DNNND repeat and one DDNHNDNHNND repeat in the coding microsatellite ms4760 have been associated with reduced in vivo QN sensitivity [14,16,17]. The function of pfnhe1 has not been fully determined, but it has been proposed that it actively effluxes protons to maintain a pH of 7.4 within the parasite [18]. Pfnhe1 has also been suggested to play a role in CQ resistance [19,20], but this was subsequently disputed [21,22].
The aim of this study was to determine possible early P. falciparum in vivo selection of genetic markers associated with antimalarial drug resistance after repeated exposure to CQ, PYR or chlorproguanil (CPGN). A large field study conducted between 1976 and 1978 in four villages in northern Liberia where virtually no anti-malarials had been previously used was therefore revisited [29]. Children aged 2-9 years received monthly administrations of CQ, PYR, CPGN or placebo depending upon which village they lived in. One drug was consistently used in each village. During the study period P. falciparum developed in vivo resistance to PYR and partially reduced susceptibility to CPGN whilst CQ remained effective [29][30][31]. The in vivo drug susceptibilities were assessed as Day 7 clearance of parasitaemia by microscopy after single doses of 10 mg/kg CQ, 2 mg/kg PYR or 1.5 mg/kg CPGN respectively. For CQ the clearance rates were 100% among children in CQ village and the control village. For PYR the clearance rates were 7% among children in PYR village and 96% in the control village. For CPGN the clearance rates were 53% among children in CPGN village and 96% in the control village. Sequence variation in pfcrt, pfmdr1, pfnhe1, pfmrp1, pfdhfr and pfdhps in samples collected after 2 years of monthly presumptive treatment with either CQ, PYR, CPGN or placebo were determined.

Study area
The clinical study from which samples were used was conducted in a rural area of northern Liberia 25-35 km from Yekepa. Four villages, about 5 km from each other, were included in the study: Bondi, Baytonwee, Bonah and Kinon. Each village had around 500 inhabitants. The villages were separated by dense forest, which made access to and between the villages difficult. Access to healthcare and specifically anti-malarial treatment was very limited when the study started in 1976. The climate is tropical with a dry season from November to April and a rainy season from May to October. The temperature is generally between 21 and 32 °C. Malaria was holoendemic when the clinical study was conducted [29,30]. In 1976 (prior to deployment of monthly presumptive treatment) the overall parasite prevalence in the study area assessed by blood slide microscopy in 2-9 years old children was 82% for P. falciparum, 39% for Plasmodium malariae and 9% for Plasmodium ovale [32].

Study group, consent, drug administration, sample collection and storage
Children 2-9 years old, present in the villages in November 1976 were invited to participate in the study. All (n = 282) agreed to participate after verbal informed consent. The four villages were visited every 14 days for 3 years. Children were given single dose anti-malarial drugs every 4 weeks for 2 years as follows. Bondi: 8-15 mg base/kg dose of CQ. Baytonwee: 1.0-2.0 mg base/kg of CPGN. Bonah: 1.3-2.5 mg base/kg of PYR. Kinon: 1 or 2 tablets of Vitamin B (placebo). Drug intake was supervised by the investigators [29].
Children in all villages were also treated with 8-15 mg/ kg base of single dose CQ in case of fever and confirmed malaria infection throughout the two-year study period. At the time this was considered a therapeutic dose of CQ [29]. After 2 years of monthly presumptive treatment, capillary whole blood samples were collected 4-6 weeks after last intake of study drug (i.e. from 28th of February to 17th of March, 1978) from all children (n = 191) present and still included in the study. The children were now 4-11 years old. Samples were coded and stored at −20 °C. The prevalence of P. falciparum in each treated village, as assessed by microscopy, was 23% (CQ), 94% (PYR), 73% (CPGN) and 88% (placebo) [29], whereas P. malariae and P. ovale were rarely detected.
The study was approved by the Liberian Institute of Biomedical Research [29]. Molecular analyses were approved by the Stockholm regional ethics board (reference number: 2013/836-32).

Molecular analysis
DNA was extracted using QIAamp Blood Mini Kit (QIAgen Biosciences, Germantown, MD, USA). The samples were extracted separately from other samples in our research laboratory to minimise the risk of contamination. Extracted DNA was stored at −20 °C until use.

Statistics
Data were entered, validated and analysed on Microsoft Excel 2003 and StataCorp 12. Allele proportions were calculated by dividing the number of samples with a certain allele by the number of samples with an identifiable allele at that position. Thus mixed infections contributed to the proportion of both alleles. Associations were determined using Fishers Exact test using StataCorp 12.

Results
A total of 191 samples were included in the study, of which 50, 48, 48 and 45 were from the CQ, PYR, CPGN and placebo village, respectively. At least one SNP was identified in all samples (not only microscopy positive) and each SNP was identified in 81-98% of samples. Negative controls were used throughout without any signs of contamination.

Pfcrt
Pfcrt K76 was found in 185/186 (99%) samples. A single sample carrying pfcrt 76T was detected in the village using CQ. The CVMNK haplotype was found in 177/178 (99%) successfully sequenced samples and the CVIET haplotype was found in the sample that had pfcrt 76T by RFLP. The pfcrt CVIET result was confirmed by PCR-RFLP and sequencing after re-extracting the sample alone. The pfcrt haplotype at codons 220, 271, 326, 356 and 371 was determined in the CQ and placebo villages only. The wild type AQNIR haplotype was found in all successfully amplified samples (88/95 [93%]) including the sample with the pfcrt CVIET haplotype.

Pfdhps
Allele proportions are presented in Table 3. The sulphadoxine resistance associated pfdhps 613S was found in 10%   the findings were significantly different compared with the PYR village (P = 0.01).

Discussion
This study provides a unique insight into the early appearance of genetic markers associated with antimalarial drug resistance and their selection 2 years after monthly presumptive treatment with CQ, PYR and CPG or placebo in a previously virtually anti-malarial treatment-naïve area of Liberia. One sample had the pfcrt 76T SNP in the 72-76 CVIET haplotype not previously recorded in Africa as early as 1978. Pfcrt CVIET was confirmed by repeated PCR analyses on re-extracted DNA several months after the first analyses. Contamination is, therefore, unlikely to explain this remarkable finding. Throughout the world pfcrt 76T has been linked to CQ resistance within various pfcrt 72-76 haplotypes. By 1978, pfcrt CVIET had existed in Southeast Asia for many years but not in Africa. Molecular analysis has shown that the CVIET later found throughout Africa was derived from this Southeast Asian lineage [8]. Southeast Asian pfcrt typically have additional downstream SNPs at codons 220, 271, 326, 356 and 371. Our sample had those typical among African origin CQ sensitive genotypes (and 3D7) [39]. Furthermore, the study area was very isolated making importation unlikely. The most likely explanation is thus that pfcrt 72-76 CVIET arose in the CQ study village de-novo following a period of intermittent presumptive treatment with CQ.
Irrespective of how this isolated CVIET haplotype came to be in the study area it did not seem to have become widespread as only pfcrt CVMNK parasites were found in 50 samples collected 3 years later in the same area [40]. This observation is consistent with longitudinal surveillance of high in vivo efficacy and in vitro susceptibility to CQ in Bondi (CQ village) between 1976 and 1978 [29,41]. In the ordered evolution of pfcrt SNPs that eventually give rise to CQ resistance described by Summers et al. [42], having CVIET alone with no other pfcrt SNPs only resulted in a small increase in Pfcrt's ability to transport CQ and hence to confer resistance [42]. Furthermore, these mutations interfere with haemoglobin digestion resulting in a loss of fitness [43]. Thus CVIET might have developed in Liberia but did not become widespread as it provided only weak selective advantage in this holoendemic area.
The increased frequency of the synonymous SNP pfmdr1 T1069 ACG and decreased diversity in pfnhe1 block I in the CQ treated village indicates that CQ exerted a selective pressure on the drug-naïve parasite population. In line with this both genes have previously been linked to CQ resistance [19,20] and previously described block I ms4760 haplotypes in samples collected after the spread of CQ resistance had the same block I type as that selected for in the CQ village [36]. However, as pfmdr1 T1069 ACG is a synonymous SNP it was not selected itself but is probably a marker for selection of a parasite strain or a pfmdr1 genotype that had a survival advantage when CQ was used.
Decreased in vitro and in vivo QN susceptibility has been associated with the pfnhe1 ms4760-1 haplotype that has two DNNND and one DDNHNDNHNND and with having more than one DNNND repeat and one DDNHNDNHNND repeat in some studies but not in others [14,16,36,44]. Though QN was not used in the Yekepa area at the time of this study the frequency of pfnhe1 ms4760-1 (26%) was similar to the frequency found at day 0 in a recent study in Mali where ms4760-1 was associated with decreased QN susceptibility [16]. Considering the geographic separation of Liberia and Mali results should be compared with considerable caution. However, similar frequencies found in Liberia in 1978, prior to a drug selective pressure and in Mali and Senegal several decades later suggest that ms4760-1 has not been under a strong selective pressure.
The PYR resistance associated pfdhfr 108N was surprisingly not selected for in the PYR village despite development of in vivo resistance to PYR within a year Table 3 Allele frequencies at resistance associated pfdhps codons 436, 437, 514, 540, 581 and 613 Resistance associated alleles are shown in italics a The frequency of samples with the resistance associated pfdhps 613S was lower in the pyrimethamine treated village compared to pooled frequencies from the other villages (P < 0.01) and the village using CQ (P = 0.001) of initiating monthly intermittent presumptive therapy [29]. In contrast it was found at significant frequencies in the placebo and CQ villages where no antifolate drugs had been used, suggesting that it was a wild type in Liberia in 1978. Similarly, resistance associated pfdhfr 51I, 59R, 108N were found in 9/66 samples collected in The Gambia in 1984 prior to widespread use of SP [45]. These intriguing results suggest that there may have been an alternate mechanism of PYR resistance in our study village as has previously been suggested [13,46]. Furthermore, the 108N did not provide a selective advantage in the presence of this alternate mechanism but rather the opposite. Amplifications of the pfdhfr or GTP-cyclohydrolase (gch1) genes have been linked to PYR resistance in the past [47,48]. Gene amplifications could have occurred in Liberia and might thus constitute an early method of PYR resistance in line with the correlation between pfmdr1 amplifications and drug resistance to mefloquine and artemisinin derivatives [49]. An alternative mechanism of PYR resistance might be a greater influx of folate into the parasite as has been suggested to be mediated by pfmrp1 1466K [28]. However, in Liberia only pfmrp1 R1466 was found. Concerning pfdhps, PYR is not likely to directly exert a selective effect, yet the frequency of pfdhps 613S (that is related to sulphadoxine resistance) was lower in the PYR treated village. Perhaps this represents an indirect effect of selection of other genes. Chlorproguanil (CPGN) is believed to be a DHFR inhibitor that is partly metabolized to the probably more P. falciparum active chlorcycloguanil in vivo. In line with this parasites with pfdhfr 16V and 108T and 108N have been associated with decreased susceptibility to the structurally similar cycloguanil in vitro and ex vivo, [26,[50][51][52]. Finding no difference in pfdhfr 108N frequency in the CPGN village compared with the placebo village indicates that 108 N was not the cause of the partly reduced CPGN susceptibility seen in the study area [29].

Conclusion
Unique data on the effect of monthly intermittent presumptive CQ, PYR and CPGN therapy on selection of known resistance genes in previously treatment-naïve P. falciparum populations is presented. The pfcrt 72-76 haplotype CVIET was detected in the CQ village and possibly developed de-novo. However, none of the downstream pfcrt SNPs commonly found throughout CQR genotypes in Africa today were detected. Decreased variation of pfnhe1 was seen and a synonymous pfmdr1 1069T SNP was selected detected in the CQ village suggesting that CQ exerted a selective pressure on the drug naïve parasite population. The PYR resistance associated SNP pfdhfr 108N was highly prevalent in this antifolate naïve parasite population and intriguingly deselected concomitant with emergence of PYR resistance suggesting alternative mechanisms of PYR resistance. The results provide new insights into the evolution of anti-malarial drug resistance in Africa.