Prevalence of chloroquine and antifolate drug resistance alleles in Plasmodium falciparum clinical isolates from three areas in Ghana

Background: The emergence and spread of resistance in Plasmodium falciparum to chloroquine (CQ) necessitated the change from CQ to artemisinin-based combination therapies (ACTs) as first-line drug for the management of uncomplicated malaria in Ghana in 2005. Sulphadoxine-pyrimethamine (SP) which was the second line antimalarial drug in Ghana, was now adopted for intermittent preventive treatment of malaria in pregnancy (IPTp). Methods: To examine the prevalence of molecular markers associated with CQ and antifolate drug resistance in Ghana, we employed restriction fragment length polymorphism polymerase chain reaction to genotype and compare single nucleotide polymorphisms (SNPs) in the P. falciparum chloroquine resistance transporter ( pfcrt, PF3D7_0709000), multidrug resistance ( pfmdr1, PF3D7_0523000), bifunctional dihydrofolate reductase-thymidylate synthase ( pfdhfr, PF3D7_0417200) and dihydropteroate synthase ( pfdhps, PF3D7_0810800) genes. Parasites were collected from children with malaria reporting to hospitals in three different epidemiological areas of Ghana (Accra, Kintampo and Navrongo) in 2012-2013 and 2016-2017. Results: The overall prevalence of the CQ resistance-associated pfcrt 76T allele was 8%, whereas pfmdr1 86Y and 184F alleles were present in 10.2% and 65.1% of infections, respectively. The majority of the isolates harboured the antifolate resistance-associated pfdhfr alleles 51I (83.4%), 59R (85.9 %) and 108N (90.5%). Pfdhps 437G and 540E were detected in 90.6% and 0.7% of infections, respectively. We observed no significant difference across the three study sites for all the polymorphisms except for pfdhps 437G , which was more common in Accra compared to Kintampo for the 2016-2017 isolates. Across both pfdhfr and pfdhps genes, a large proportion (61%) of the isolates harboured the quadruple mutant combination ( I 51 R 59 N 108/ G 437). CQ resistance alleles decreased during the 12 years after CQ withdrawal, but an mediate SP resistance alleles increased. Conclusion: Surveillance of the prevalence of resistance alleles is necessary in monitoring the efficacy of antimalarial drugs.


Abstract
The emergence and spread of resistance in Background: Plasmodium to chloroquine (CQ) necessitated the change from CQ to falciparum artemisinin-based combination therapies (ACTs) as first-line drug for the management of uncomplicated malaria in Ghana in 2005. Sulphadoxine-pyrimethamine (SP) which was the second line antimalarial drug in Ghana, was now adopted for intermittent preventive treatment of malaria in pregnancy (IPTp).
To examine the prevalence of molecular markers associated Methods: with CQ and antifolate drug resistance in Ghana, we employed restriction fragment length polymorphism polymerase chain reaction to genotype and compare single nucleotide polymorphisms (SNPs) in the P. falciparum chloroquine resistance transporter ( PF3D7_0709000), multidrug pfcrt, resistance ( PF3D7_0523000), bifunctional dihydrofolate pfmdr1, reductase-thymidylate synthase ( PF3D7_0417200) and pfdhfr, dihydropteroate synthase ( PF3D7_0810800) genes. Parasites pfdhps, were collected from children with malaria reporting to hospitals in three different epidemiological areas of Ghana (Accra, Kintampo and Navrongo) in 2012-2013 and 2016-2017.
437G and 540E were detected in Pfdhps 90.6% and 0.7% of infections, respectively. We observed no significant difference across the three study sites for all the polymorphisms except for 437G which was more common in Accra compared to Kintampo for pfdhps , the 2016-2017 isolates. Across both and genes, a large pfdhfr pfdhps proportion (61%) of the isolates harboured the quadruple mutant

Introduction
Malaria remains a major global health concern especially in sub Saharan Africa. P. falciparum malaria is considered the most severe and also the leading cause of morbidity and mortality, especially among children under five years (Schumacher & Spinelli, 2012). In 2016 a global estimate of 216 million malaria cases was reported, which led to about 445,000 deaths (WHO, 2017). The global malaria mortality rate, however, has reduced by 29% since the year 2010, as a result of increased preventive and control measures (WHO, 2016).
The use of antimalarial drugs for malaria treatment and prevention has played an integral role in the control of the disease over the decades (Cui et al., 2015;Gosling et al., 2011;Greenwood, 2004;Schlitzer, 2007). Unfortunately, the emergence and the spread of drug resistant P. falciparum strains militated against the use of antimalarial drugs for the containment of the disease (Lin et al., 2010). P. falciparum chloroquine (CQ) resistant strains were first reported in the 1950s in Southeast Asia along the Cambodia-Thailand border (Young et al., 1963) and subsequently reported in other countries globally. Currently, the parasite has been reported to have developed resistance to most available artemisinin monotherapies and this is exhibited by reduced parasite clearance rates and/or treatment failures (Dondorp et al., 2009). ACTs are now the frontline drugs for treating uncomplicated P. falciparum malaria in almost all countries that are endemic with malaria, including Ghana (WHO, 2016).
Point mutations in specific genes in the parasite genome are implicated in resistance to specific antimalarial drugs (Cui et al., 2015;Fidock et al., 2000;Sidhu et al., 2002). A point mutation in the P. falciparum chloroquine resistance transporter gene (pfcrt, PF3D7_0709000) that replaces lysine with threonine at codon 76 had become a common single nucleotide polymorphism (SNP) in parasite populations as it is a critical mediator of resistance to CQ (Babiker et al., 2001). In addition, mutations in the P. falciparum multidrug resistance gene 1 (pfmdr1, PF3D7_0523000) that result in amino acid substitutions at positions N86Y and Y184F have been reported to confer parasite resistance to CQ, amodiaquine (AQ) and lumefantrine (L) (Duraisingh & Cowman, 2005 In Ghana, prior to the withdrawal of CQ a prevalence range of between 46%-98% of the mutant pfcrt 76T was reported across five sentinel sites (Duah et al., 2007). Interestingly, studies in other settings have shown that the replacement of CQ with ACTs resulted in a decline in the frequency of the mutant alleles and concomitant restoration of CQ susceptibility (Laufer et al., 2006;Mwai et al., 2009). In a study that was conducted in Tanzania, more than 90% recovery of the sensitive pfcrt K76 allele was reported after 10 years of CQ use being officially discontinued (Mohammed et al., 2013). Follow -up studies in Ghana have reported a decline in the prevalence of pfcrt 76T and pfmdr1 86Y but an increasing prevalence pfdhfr I 51, R 59, N 108 and 437G resistant alleles from 2003 to 2010 (Duah et al., 2013;Duah et al., 2012) This study sought to ascertain the population trends in the prevalence of known drug-resistance-related point mutations in pfcrt, pfmdr1, pfdhfr and pfdhps in clinical isolates from three different malaria-endemic areas in Ghana a decade following the introduction of ACTs.

Amendments from Version 1
The majority of reviewers' recommendations have been effected, and for the others, appropriate explanations/rebuttals have been provided. Details of the cross-sectional sampling in 2012-2013 and 2016-2017 and the genotyping methods used have been included. In addition, an explanation has been provided for the choice of the 2-14 year age group in the study. The data have been disaggregated and reanalyzed separately for the 2012-2013 and 2016-2017 time points. This study depended on archived samples from an erythrocyte invasion study so only samples that were successfully genotyped for any number of mutations were presented in Table 1. Table 1 now compares data from the same time points across study sites while Table 2 now captures frequencies of allele combinations in both pfdhfr and pfdhps. In this study mixed infections were not examined and we have provided an explanation for that in the revised manuscript. A map of Ghana has been included as Figure 1 to show the distance between Sunyani and Kintampo and the use of data for Sunyani to represent Kintampo has been corroborated with relevant literature. The graph on trend analyses is now designated Figure 2. References have been updated to capture extant literature in Ghana and other West and Eastern African countries. The discussion on IPTp and SMC uptake in Ghana and Burkina Faso has been expanded. Further details on the role of ACTs in the selection of mdr1 mutations has been included and relevant references provided therein. The major findings of the study have now been captured in the discussion as recommended. In addition, the use of antifolate drugs for the management of opportunistic infections in HIV has been discussed. The conclusion has been revised to align with the study objectives.

Ethical consideration
This study was approved by the Ethics Committees of the Ghana Health Service (GHS-ERC:12/05/12), the Kintampo Health Research Centre (KHRCIEC/FEA/2011-13), the Navrongo Health Research Centre (NHRC-IRB135/08/2012) and the Noguchi Memorial Institute for Medical Research (NMIMR) (NMIMR-IRB CPN 004/11-12). Informed consent of parents or guardians for all participants was obtained. An additional assent was also obtained from children aged 10-14 years prior to recruitment.

Study sites and sample collection
This study leveraged the availability of samples from a concurrent study at the time on erythrocyte invasion mechanisms and whole genome sequencing of the malaria parasites. The appropriate sample collection at the time to meet the erythrocyte invasion studies, the whole genome sequencing of the malaria parasites was adopted. We used the samples so gotten to carry out the drug resistance study. The choice of 2-14 years was premised on development of immunity that was key in the erythrocyte invasion study. . Samples were obtained from participants during the rainy seasons at the respective study sites. P. falciparum genomic DNA was analyzed for the prevalence of known antimalarial drug resistance SNPs in pfcrt (K76T), pfmdr1 (N86Y and Y184F), pfdhfr (N51I, C59R and S108N) and pfdhps (A437G and K540E) across the three study sites. Malaria was diagnosed using the first response ®malaria Ag. (HRP2) card test (Premier Medical Corporation, Ltd., Mumbai, India) and confirmed by microscopy. Venous blood samples were obtained and depleted of leucocytes using lymphoprep gradient centrifugation, followed by passage through Plasmodipur filters (EuroProxima, Arnhem, Netherlands), and the resulting infected red blood cells were stored at -20°C until DNA extraction.
Extraction of genomic DNA and nested PCR Plasmodium gDNA was extracted from the samples using the QIAamp Blood Midi Kit (Qiagen, Manchester, UK) as per manufacturer's instructions and stored at -20°C. Both outer and nested PCRs were carried out to amplify regions flanking known point mutations in pfcrt (K76T), pfmdr1 (N86Y and Y184F), pfdhfr (N51I, C59R, and S108N) and pfdhps (A437G and K540E) that mediate antimalarial drug resistance. All PCRs were carried out at final volume of 25 µL containing 1X of Maxima Hot Start Green PCR master mix (Thermo Scientific, Waltham, MA, USA) and 250 nM of each of the forward and the reverse primers. Five microlitres of the purified P. falciparum gDNA was used as template in the outer PCR and 1 µL of the resulting products was used as template DNA in the nested PCR. Previously reported primer sets and cycling conditions for both the outer and the nested PCRs were used (Djimdé et al., 2001a;Duraisingh et al., 1998). Prior to the restriction digest, 5 µL of the nested PCR products were resolved on 2% agarose gel stained with ethidium bromide and images were resolved using the Amersham Imager 600 (General Electric Healthcare Life Sciences, Chicago, IL, USA).

Restriction digestion of nested PCR amplicons
The resulting nested PCR products for each of the four genes containing the SNP alleles of interest were analyzed by restriction fragment length polymorphism (RFLP). Each of the restriction digestion reactions was set at a final volume of 15 µL containing 5 µL of the nested PCR product, 1X FastDigest Green buffer and 0.3 µL of the appropriate restriction enzyme (Thermo Scientific). The restriction enzymes used, incubation temperature, incubation time as well as the expected band sizes for the wild-type and the mutant alleles of the point mutations were as reported in previous studies (Djimdé et al., 2001a;Duraisingh et al., 1998). Ten microlitres of the restriction digestion fragments were resolved on 2 % agarose gel stained with ethidium bromide and the resulting image resolved with the Amersham Imager 600 (GE, USA). Purified DNA obtained from laboratory strains of P. falciparum (Dd2, 3D7, FCR3, K1, 7G8 and W2) were used as controls for the sensitive and resistant alleles for each gene.

Data analysis
Data was analyzed using the Stata version 14.2 (Texas, USA), and the GraphPad Prism (Version 6.01). Analysis of contingency tables of frequency distribution of the point mutations between the study sites were analyzed by chi-square test. In addition, allele combination frequency distribution of the 2012-2013 isolates were compared to the 2016-2017 isolates utilizing the Fisher exact test for expected lower cell counts taking each marker as independent. All statistical tests were two-tailed and statistical significance was defined at P < 0.05.

Results
Prevalence of alleles in P. falciparum genes that mediate chloroquine and antifolate drug resistance The prevalence of antimalarial drug resistance alleles in three different transmission zones were determined and compared across sites and sampling time points.
. We did not observe any significant differences in the distribution of isolates harbouring pfcrt K76T, pfmdr1 N86Y or pfmdr1 Y184F point mutations across the three transmission zones (P > 0.05 for all three SNPs), although all the three mutant alleles were found at a higher prevalence in Navrongo (2012-2013) compared to Kintampo (2012-2013 and 2017) and Accra (2016-2017) ( Table 1). The total prevalence of pfcrt 76T (8%) and pfmdr1 86Y (10.2%) mutant alleles were comparable (P = 0.39). Compared to CQ resistance-associated alleles, higher frequencies were observed in the three study sites for all the antifolate drug resistance-associated alleles, except pfdhps K540E (Table 1 and Dataset 1). The frequency distribution of isolates harbouring the pfdhfr 51I, 59R and 108N mutant alleles were also comparable across the study sites (P > 0.05 for all the three loci). However, the distribution of pfdhps 437 G mutant allele, was significantly different across the study sites (P = 0.01). Pfdhps 437G was significantly higher in Accra Trends in the prevalence of antimalarial drug resistance markers in the study populations To investigate the dynamics of the drug resistance alleles in the selected areas, we compiled data from previous studies that reported the frequencies of the various mutations in the same or near-by communities. Thus, the current data from Navrongo were compared to previous data from the same area, while data from Kintampo were compared to published data from Sunyani. Kintampo and Sunyani are located in the same region (Brong Ahafo) but approximately 122 Km apart (Figure 1) Sunyani lies in the Forest Zone whilst Kintampo lies within the Forest Savanah transition zone, however, both sites have similar   Table 2). The frequency of I 51 R 59 N 108/G 437 was comparable across the study sites (P > 0.05 for all comparisons). Low prevalence (<10%) allele combinations in both pfdhfr and pfdhps were observed for the triple mutant allele combination (R 59 N 108 /G 437 , I 51 R 59 /G 437 and I 51 N 108 /G 437 ) and these were also comparable across the study sites (P > 0.05 for all haplotypes) ( Table 2).

Discussion
P. falciparum resistance to antimalarial drugs remains one of the biggest threats to the control and elimination of malaria globally. In Ghana, a change in the use of CQ to ACTs was implemented in 2005 as a result of high rate of malaria treatment failure (Duah et al., 2007). In this study, we determined the prevalence of alleles associated with CQ and antifolate resistance using clinical isolates from three malaria endemic regions with varying transmission intensities in Ghana. We observed a decreasing prevalence of CQ resistance-associated alleles but an increasing prevalence of SP resistance-associated alleles. The distribution of the alleles across the three study sites were not significant, except for pfdhps 437G which was significantly higher in Accra compared to Navrongo and Kintampo. The frequency of pfdhfr/pfdhps haplotypes in 2012-2013 and 2016-2017 were not significantly different across the three study sites. Both in vitro and molecular surveillance studies have associated CQ resistance mainly with the pfcrt 76T allele, but also with pfmdr1 86Y and 184F alleles. Pfcrt 76T and pfmdr1 86Y mutant alleles have also been reported to decrease P. falciparum susceptibility to amodiaquine  Ghana and this may explain the low resistance in the former. Furthermore, since drug resistance evolution is spatiotemporal the differences in periods of sampling could also account for the differences observed. The high prevalence may be due to SP intervention in groups such as pregnant women and young children acting as reservoirs of infections with resistance alleles as a direct consequence of continuous use of SP in IPTp and SMC campaigns that fuel transmission of these alleles in the general population. Another important factor may be the unauthorized use of SP for self-medication as it is readily available at health centres and pharmacy shops in the study areas (Abuaku et al., 2004), particularly because it is a single dose drug with very minimal to no adverse reactions. Co-trimoxazole is used in Ghana (Fadeyi  et al., 2015), however, there is limited data on its usage in the three study sites. Besides the prevalence of HIV in Ghana is only 3 % (Ghana AIDs Commission, 2017) and therefore the use of antifolate drugs such as cotrimazole for the management of opportunistic infections is not as widespread as the use of antifolate antimalarial drugs. The study indicates that CQ sensitive parasites have again become more common since the replacement of CQ with a variety of ACTs as first-line treatments of uncomplicated malaria in Ghana. This notwithstanding, our findings also show that between 5% to 14% of clinical infections may still carry CQ resistant parasites, which suggest that ACT partner drugs such as AQ that are widely used in Ghana may still be maintaining significant selection pressure on the pfcrt locus. In addition, the increasing prevalence of the pfdhfr/pfdhps partial SP resistance haplotypes could result in the fixation of these alleles within the parasite population. The continuous use of SP for IPTp and SMC may result in emergence of the "full" SP resistance haplotype and compromise the use of SP IPTp and SMC are the two significant sources of SP drug pressure on the parasite population in pregnant women and young children, respectively. Recent studies have shown that these interventions have contributed to reduction in maternal and child morbidity and mortality (Coldiron et al., 2017). Undoubtedly, these interventions are critical (York, 2017), but could easily be undermined by rising resistance in these populations. Therefore, it is very important to closely monitor the prevalence of molecular markers of resistance associated with antifolate antimalarial drugs to guide policies on the continuous use of these drugs in Ghana and other African countries. There are probably other factors that contribute to the evolution of resistance markers as SP has been shown to be efficacious even in the face of fixation of SP resistant alleles (Iriemenam et al., 2012).

Conclusion
This study reports an increasing prevalence of CQ sensitive clinical isolates after 12 years of CQ withdrawal at three different study sites that capture the eco-epidemiology of malaria in Ghana. The prevalence of the antifolate drug resistant alleles remain relatively high across the study sites. Besides, there is an increasing trend in the frequency of SP-resistance associated alleles at all sites. Taken together, these observations point to the need for a robust antimalarial drug discovery strategy to provide a vast array of alternatives for chemotherapy in readiness for the likelihood of future poor parasite response to the use of SP for prevention of malaria in pregnant women and for SMC in children. However, it is pre-mature to recommend the discontinuation of SP use due to the high prevalence of antifolate drug resistance alleles since the drug can be efficacious where there is fixation of these alleles.

Data availability
The data supporting this article is available online at Open

Edwin Kamau
Walter Reed Army Institute of Research, Silver Spring, MD, USA Abugri analyzed the prevalence of chloroquine (CQ) and sulphadoxine-pyrimethamine (SP) et al. resistance in clinical isolates in Ghana. The authors used known molecular markers of resistance in three genes that modulate sensitivities to these two different antimalarial drugs; pfcrt and pfmdr1 for CQ resistance, and pfdhfr and pfdhps for SP resistance. Samples were collected from three different locations with different transmission intensities in two different time periods, 2012/2013 and 2016/2017. It is a well written manuscript but I have several comments. the same location, this can be extremely problematic and misleading. The authors must make strong arguments and justify their point of view much more strongly than currently presented. See references below. In the analysis of pfdhfr and pfdhps haplotypes, only Kintampo has samples collected in 2012/2013 and 2016/2017, comparing three different locations which Navrongo only has 2012/2013 samples/data and Accra has only 2016/2017 samples/data is not completely accurate (as presented 2012/2013 vs. 2016/2017). This analysis must be revisited. The difference might be just because the samples were collected in different locations; it is important this is possibility is ruled out. Discussion: The first paragraph of the discussion needs to capture and highlight the key findings of the study, consider re-writing the first paragraph of the discussion to capture the key findings. table to give an idea of the amplification success (proportion of samples with genotypes) for each of the markers analysed. This will be helpful in interpreting the results in subsequent sections.
In the introduction, it was stated that, "This study sought to ascertain the population trends in the prevalence of known drug-resistance-related point mutations in and pfcrt, pfmdr1, pfdhfr in clinical isolates from three different malaria-endemic areas in Ghana a decade pfdhps following the introduction of ACTs" and it was also mentioned in the data analysis section that the data was pooled for 2002-2003 and 2016-2017. However, the section on the prevalence of alleles of CQ and antifolates as well as table 1 don't show any data on the prevalence of the alleles at different time points (trends). Such data should be incorporated in the table and summarized in the text as well. While in the text authors refer to "prevalence of alleles", the heading of table 1 talks about the "frequency of SNPs". These are two different things and should be reconciled. The authors should also explain how they dealt with samples which had mixed infections and therefore mixed alleles.  Since table 2  presents combined and mutations, all the reported haplotypes should cover both dhfr dhps genes because the reported number and prevalence are basically considered to have combined parasites with/without mutations in the two genes (as reported in table 2). Thus, the following statement, "Double mutant allele combinations of ( , and pfdhfr I N R N ) had comparably low frequencies" cannot be shown in the same table (as for the I R triple mutations) without including the allele on . The double mutation dhfr dhps I G reported in text is not shown in table 2.

Discussion:
The authors mentioned that, "The percentages of the 51I (81%), 59R (82%), 108N pfdhfr (88%) and 437G (88%) mutant alleles reported in this study are relatively higher pfdhps when compared to the 71%, 42%, 64% and 80% prevalence reported in a recent study in a neighbouring country, Burkina Faso (Cisse ., 2017), this could be due to differences in et al the uptake of IPT in both countries." However, they do not provide any data or evidence to show the differences in uptake of IPTp in the two countries. At the end of the second paragraph, authors state that, "Unlike 76T and 86Y, pfcrt pfmdr1 the prevalence of 184F mutant allele (65%) appears to have not varied so much pfmdr1 from 2005 to 2017 when compared to the 43% to 69% prevalence reported from 2005 to 2010". The role of ACTs in the selection of mutations should be discussed, due to mdr1 available evidence of increasing N86 and 184 after introduction of ACTs as reported in F some countries. This study covers molecular markers which are important for drugs used in both IPTp and SMC, the discussion should provide prominence on these interventions, in order to make a strong case for future surveillance and monitoring of these markers. This will provide an opportunity to monitor the impact of IPTp and SMC on the molecular markers of amodiaquine and SP resistance.

Conclusion:
The conclusion should be revised to align it to the objectives of the study which was stated at the end of the introduction; "This study sought to ascertain the population trends in the prevalence of known