Molecular epidemiology of drug-resistant malaria in western Kenya highlands

Malaria is a major public health problem in sub-Saharan Africa and Plasmodium falciparum infection is a leading cause of morbidity and mortality inflicting a huge economic burden in countries where the disease is endemic [1]. It is estimated that death toll of malaria exceeds one million people each year, and the victims are primarily children under the age of five [2]. Until the early 1980s, the African highlands (generally referred to areas of >1,500 m above sea level) were either free of malaria or had very low incidences of the disease; however, since the late 1980s a series of malaria epidemics has occurred [3–9]. Among the many factors that may contribute to the highland malaria epidemics, resistance of the parasites to multiple antimalarials has not been extensively investigated. Resistance to antimalarial drugs is one of the major obstacles for effective malaria control. The first case of chloroquine (CQ) resistance in Kenya was reported in 1977 [10]. In 1993, resistance levels had reached 70% [11]. In 1998, the Ministry of Health of Kenya changed the first line of treatment from chloroquine to sulfadoxine – pyrimethamine (SP; Fansidar^®) [12]. In 2004, the Ministry of Health of Kenya officially changed the first-line drug to artemether/lumefantrin (Coartem™) [13].

Drug resistance in malaria parasites is associated with genetic mutations in target genes and can be monitored using molecular methods. CQ resistance is determined by the major point mutation at codon 76 of the P. falciparum CQ resistance transporter (pfcrt) gene [14], which is highly correlated with increased clinical CQ tolerance and treatment failure [14–16]. In addition, point mutations in P. falciparum multi-drug resistance gene 1 (pfmdr1) (e.g., N86Y, Y183F, S1034C, N1042D, and D1246Y) have been shown to modulate CQ resistance [17] and possibly lumefantrine resistance [18]. Resistance to antifolates is associated with point mutations in the dihydrofolate reductase (pfdhfr) and dihydropteroate synthetase (pfdhps) genes [19]. Pyrimethamine (PY) resistance is conferred by the key mutation at codon 108 in the pfdhfr gene, while additional mutations at positions 51 and 59 increase the levels of resistance [20]. The 164L mutation common in Southeast Asia has been shown to confer PY resistance. Although this mutation is not common in Africa, a recent study by McCollin et al detected the 164L mutation in western Kenya [21]. Similarly, sulfadoxine (SD) resistance is conferred by a key mutation at codon 437 in the pfdhps gene and modulated by additional mutations at codons 436, 540, 581, and 613 [22]. Multiple mutations in these two genes result in antifolate treatment failure [23]. Molecular methods have been developed to detect point mutations in genes responsible for drug resistance [24–26], and have been widely used to monitor drug resistance in epidemiological surveys [16]. In this study, we have analyzed the mutations of antimalarial drug resistance genes for CQ (pfcrt and pfmdr1) and antifolates (pfdhps and pfdhfr) in samples from patients with acute malaria infections and school children with asymptomatic infections at one lowland and two highland locations in western Kenya.

In the present study, we analyzed the mutations of four known drug resistance genes in both highland and lowland parasite populations of western Kenya. We showed that the frequencies of key mutations in the pfcrt, pfmdr1, pfdhps, and pfdhfr genes that were implicated in resistance to CQ and SP were very high. We further demonstrated that there was no difference in the frequencies of key mutations between symptomatic and asymptomatic malaria volunteers. Our experiment design did not allow us to test clinical or parasitological efficacy in symptomatic infections after treatment; nor was the mutation prevalence/in vivo resistance before 1998 when the national policies of treating uncomplicated malaria were changed to SP was examined. Nevertheless, the big sample size in this study coupled to lack of differences in the frequency of resistant parasite genotypes in the two highland areas and the low land area clearly demonstrates a high frequency of drug resistant mutants circulating in the study areas.

Resistance to CQ is largely determined by the K76T mutation in the pfcrt gene, and enhanced by mutations at other sites of this gene and mutations in the pfmdr1 gene. Since resistant phenotypes often have fitness costs [37], their prevalence is expected to decline after removal of the selective pressure. For example, the prevalence of mutant alleles of pfcrt 76T decreased from 64.5% in 2002 to 16% in 2004 and that of mutant pfmdr1 86Y alleles decreased from 46.6% to 2.7% two and half year after CQ withdrawal in coastal Tanzania [38]. Kublin et al. reported that the prevalence of the chloroquine-resistant pfcrt 76T genotype decreased from 85% in 1992 to 13% in 2000. In 2001, chloroquine cleared 100% of 63 asymptomatic P. falciparum infections, no isolates were resistant to chloroquine in vitro, and no infections with the chloroquine-resistant pfcrt 76T genotype were detected [39]. Similarly, Laufer et al. demonstrated that chloroquine was again an efficacious treatment for malaria, 12 years after it was withdrawn from use in Malawi [40]. Although Kenya had a similar drug policy, the pfcrt K76T mutation in western Kenya was still predominant. Consistent with the observation of high pfcrt K76T mutation frequency in the study sites, pfmdr1 mutations that enhance CQ resistance (e.g., 86Y) were also present at relatively high frequencies (37.2 – 45.2%) in western Kenya. This may be partly caused by wide availability and consumption of CQ after it was replaced as first-line antimalarial drug by antifolate drug (SP) in 1998 [12, 13]. Our results indicated that parasite populations in western Kenya were still highly resistant to 4-aminoquinoline drugs.

Several years after introduction of SP as the first-line antimalarial drug in Kenya, mutant genotypes have already become common. We found that mutations at codon 108 of the pfdhfr gene and codon 437 of the pfdhps gene, which are the major determinants for sulfadoxine and pyrimethamine resistance, respectively, were highly prevalent in our study sites. Furthermore, increased resistance to antifolate drugs was correlated with mutations at additional sites, which were also detected in our present study. In Kenya, Nzila et al. [41] reported that the pfdhfr triple mutant (codons 108, 51 and 59) was associated with seven-day treatment failure using Fansidar, and this association was strengthened by the presence of double mutations of the pfdhps gene at codons 437 and 540. Such "quintuple" mutants were also strongly correlated with Fansidar treatment failure in Malawi [26]. These quintuple mutants have already reached 80–87% prevalence in the parasite populations in western Kenya. This is consistent with the observation in another highland site in Chogoria, near Mount Kenya [42]. Since mutations at different sites of the pfdhfr and pfdhps genes might be a stepwise process, two mutations at codon 59 of pfdhfr and codon 540 of pfdhps can accurately predict the presence of this quintuple mutant [26]. The molecular method for monitoring antifolate resistance in P. falciparum can thus be simplified to detect the presence of these two mutations. This method has been shown to provide the best means of predicting clinical treatment outcomes in the patient population, which consists primarily of children from endemic areas of Africa [36]. In our study, either of these two mutations has exceeded 75% prevalence in the parasite populations, suggesting that at least 50% of the parasites carry these double mutations. Consistent with other genotype studies of pfdhfr from Kenyan isolates, we did not observe mutations at codon 164 [23, 41, 43], although McCollum et al. identified 164L mutation in several samples in western Kenya [21].

The causes of the malaria epidemics in the East African highlands are complex and could be the result of interactions between environmental changes, vectors and parasites. Antimalarial drug resistance has been invoked as a major factor [9], however the molecular epidemiology of drug resistance in these highlands is unknown. Mbaisi et al. [43] reported a significantly lower prevalence of mutations at codon 437 of the pfdhps gene and codons 86 and 1246 of the pfmdr gene in lowland (Kisumu) samples than in highland (Kericho, Magadi, and Entosopia) samples in western Kenya, whereas this trend was the opposite for the mutation at codon 436 of the pfdhps gene. In contrast, our study did not detect significant differences in frequency of mutations at 14 codons in four drug resistant genes between samples from lowland and highland areas in western Kenya, although the prevalence of mutations at codons 86 and 1246 of the pfmdr gene was higher in samples from the lowlands (Kombewa) than in those from the highlands (Kakamega and Kisii). This result suggests either a similar drug selection pressure or significant gene flow between lowland and highland parasite populations aided by human travel [44], as detected by microsatellite and merozoite surface protein 1 and 2 (msp-1 and msp-2) [45].

Together with the evidence that there was no significance difference in the frequencies of key resistance-conferring mutations in pfcrt, pfmdr1, pfdhps, and pfdhfr genes between symptomatic and asymptomatic malaria infections in western Kenya highland, high frequencies of these mutations in symptomatic and asymptomatic infections suggest that drug resistance of malaria parasites may be an important contributor to malaria-induced morbidity and mortality. However, the role of drug resistance as a driving force for malaria outbreaks in the highlands has not been established. If drug resistance were the main driving force, the lowland and highland sites should expect similar epidemic pattern of malaria incidence. Moreover, drug resistance level should increase over time until a new antimalarial drug is adopted. Thus, the number of malaria patients would be expected to increase gradually over time, but malaria incidence should not exhibit a large inter-annual variation. In contrast, dramatic fluctuations in malaria incidence in the African highlands were observed [9, 46, 47]. Significant association between climate variability and malaria incidence suggests that climate factors may play an important role in the East African highlands [47]. Further studies are required to examine the interactions between climate factors and drug resistance on malaria incidence in African highlands.