The emergence and spread of drug resistant malaria poses the greatest challenge to the control and elimination of malaria in Swaziland. Despite reports of chloroquine failure suspected to be a result of drug resistance in the 1980s 1 , CQ has remained the first-line drug for treatment of uncomplicated Plasmodium falciparum malaria in Swaziland. The combination of sulphadoxine and pyrimethamine (SP) has been an effective and affordable alternative single dose treatment for CQ-resistant parasites since its first introduction to Africa in the 1960s. Use of SP increased rapidly in countries with CQ resistance over the next 30 years. As a result, resistance to SP is now globally widespread 2 , leading most malaria endemic countries in Africa to abandon SP alone as first-line treatment. In Swaziland, SP was introduced in 1993 as a second-line treatment for uncomplicated malaria following reports of CQ failure in Mozambique 3 4 5 , South Africa (KwaZulu-Natal) 6 and Swaziland 1 . To this day, SP remains a second-line anti-malarial in Swaziland despite its withdrawal as a first-line anti-malarial in KwaZulu-Natal (2001) and Mozambique (2002). Resistance to CQ is associated with polymorphisms in the pfcrt and pfmdr1 genes, and resistance to SP with pfdhfr and pfdhps polymorphisms. No data are available concerning the diversity of these loci in Swaziland.

The mutation at codon 76 of the pfcrt gene, where lysine is replaced by threonine (K76T) has been strongly associated with P. falciparum resistance to CQ and subsequent treatment failure 7 . Mutations in the pfcrt gene at codons other than K76T are said to modulate CQ resistance and/or compensate deleterious effects of K76T mutations 8 . Mutations in the pfmdr1 gene (N86Y, Y184F, S1034C, N1042D and D1246Y) have been reported to modulate the level of CQ resistance 9 10 11 . Recent studies have further associated some alleles of the pfmdr1 gene with decreased susceptibility to artemisinin-containing drugs. Artemether-lumefantrine (AL) is currently the most favoured combination therapy against uncomplicated Plasmodium falciparum malaria in southern Africa. However, in vivo studies of AL-treated patients have shown enhanced survival of parasites carrying the pfmdr1 polymorphisms (86N, 184F and 1246D) and selection against the YYY haplotype 12 13 . Studies have also reported an increased pfmdr1 copy number in parasites of Asian origin following AL treatment 14 . Selection of the NFD haplotype and increased pfmdr1 copy number could result in reduced susceptibility of parasites to AL and further investigation of pfmdr1 in areas implementing AL treatment policy are clearly needed.

Resistance to SP has been shown to occur following changes in the amino acid sequences of the dihydrofolate reductase (DHFR) 15 and dihydropteroate synthase (DHPS) 16 enzymes, with both in vitro and in vivo studies finding associations between mutations in the dhfr gene of P. falciparum and resistance to pyrimethamine 17 18 and between mutations in the pfdhps gene and resistance to sulphadoxine 19 20 . The level of resistance to SP has been shown to increase with accumulation of mutations in codons 51, 59, 108 and 164 in pfdhfr and in codons 437 and 540 in pfdhps 21 22 ; the contribution to resistance of mutations at positions 431, 436, 581 and 613 of pfdhps remains unclear 23 .

This report describes molecular screening of 79 isolates from Swaziland collected in 1999 and 2007. Resistance-associated polymorphisms in the pfcrt, pfmdr1, pfdhfr and pfdhps genes were determined by multiplex qPCR and direct DNA sequencing of conventional PCR products.

Malaria is still a major concern in Swaziland despite recent reports of significant reduction of disease incidence by more than 95% 29 . Selection of effective and affordable anti-malarials for treatment and control remain extremely important and complicated in this era of emergence of P. falciparum parasites resistant to multiple anti-malarials. CQ failure in Swaziland was first reported in 1987 1 . More than twenty years later, Swaziland retained CQ as a first-line anti-malarial for treating uncomplicated malaria. Although SP has not been used as a first-line anti-malarial in Swaziland, it is still of interest to determine whether parasites remain susceptible to the drug especially following reports of resistance from neighbouring countries. SP is used as a second-line anti-malarial for CQ resistant (CQR) parasites.

The high prevalence of the pfcrt CVIET haplotype, which originated in CQR parasites from Southeast Asia 30 31 , and the absence of CQR haplotype SVMNT (South American) among isolates examined in this study is consistent with the earlier presumed spread of CQR parasites from Southeast Asia to Africa via the Indian subcontinent. Reports of SVMNT occurrence described in isolated areas of East Africa are probably a result of resistance to AQ or its metabolite desethyl-amodiaquine (DEAQ) following use of AQ as monotherapy 32 . High prevalence of 76T mutations both in 1999 and 2007 suggests that CQ resistance is widespread in Swaziland. Studies have also shown that acquisition of 76T mutations favours severity and multiplicity of malaria infection 33 and, together with pfmdr1 mutations, is associated with life-threatening complications such as severe anaemia (Hb<5 g/dl) in young children 34 .

No published data are available on the in vivo efficacy of CQ in Swaziland. Evidence of heavy over-prescription is described here, in which only six of 252 patients treated for malaria in 2007 proved to be parasite positive by both retrospective slide-reading and PCR. Previous studies elsewhere have shown that the 76T and 86Y mutations occur among parasites surviving CQ treatment, including those responsible for production of transmissible gametocytes 35 36 . In CQ-treated children, parasites with pfmdr1 86Y and pfcrt 76T are transmitted more frequently than other genotypes and this mechanism probably supports continued circulation of these genotypes 37 . Further studies among malaria patients in Swaziland are required to estimate any impact of these mutations on treatment outcome. The pfmdr1-1246Y allele, found at 49% prevalence among 1999 isolates, was absent among 2007 isolates, suggesting withdrawal of CQ use may lead to further beneficial changes in the parasite population in Swaziland.

The novel 86F mutation in pfmdr1 identified in two samples from this study has not previously been described in field-collected parasites. In the isolates described here, the phenylalanine at position 86 is encoded by the codon UUU, which has two base changes from the wild-type asparagine codon AAU, but only a single base-change from the common CQ-resistance-associated tyrosine codon, UAU. The 86F allele is therefore likely to have arisen from the 86Y form of pfmdr1 by a single point mutation, rather than directly from the wild-type 86N form, which would require a minimum of two changes. This is consistent with in vitro observations 27 .

The absence of the wild-type pfmdr1-NYD haplotype in 2007 and its low prevalence among 1999 isolates suggest that CQ resistance had developed over a long time. The high prevalence of the pfmdr1-NFD haplotype in 2007 and absence of the YYY haplotype is of interest, as studies elsewhere in Africa have reported selection for the NFD allele of pfmdr1 and loss of the YYY allele following treatment with artemether-lumefantrine (AL) 12 13 . Since there is no history of AL use in Swaziland, this pattern may be a result of the impact of drug policy changes in neighbouring South Africa, KwaZulu-Natal (KZN) and Mozambique. Swaziland is a very small country and the influence of drug policies in Mozambique and South Africa (KZN) cannot be ruled out. KZN introduced AL in 2001 while Mozambique introduced artesunate-sulphadoxine/pyrimethamine (AS-SP) combination therapy in 2002 and artesunate/AQ in 2004. Thus, deployment of AL in KZN and the introduction of AS-SP treatment policy in Mozambique could have influenced selection of 86N parasites through successful transmission compared to their mutant counter-parts. However, this phenomenon can only be confirmed through analysis of a larger number of samples from Swaziland. Transmission studies reported a four-fold reduction in transmissibility to mosquitoes of the 86Y allele compared to the 86N allele following addition of artesunate to CQ 35 . The same study reported an association between CQ monotherapy and increased 86Y transmissibility. Other studies have shown increased tolerance of laboratory parasite clones to mefloquine and halofantrine, (which are related to lumefantrine) if they carried the 86N allele 10 or the 1034S/1042N/1246D haplotype 38 . Therefore, the prevalence of NFD haplotypes in Swaziland has to be closely monitored as it could compromise the effectiveness of combinations involving lumefantrine in the future. Unfortunately, data on the prevalence of NFD and YYY haplotypes in KZN following AL implementation in 2001 are not available.

Regional drug pressure, as well as prescription of SP to CQ-resistant cases probably partly explains the high prevalence of dhfr triple-mutant haplotypes (82%) observed in 1999 in Swaziland. The prevalence of the dhfr-CIRNI triple-mutant haplotype in Zone 1 in Southern Mozambique, the zone closest to Swaziland, was 47.5% in 1999 39 which is lower than that observed in Swaziland, suggesting only partial influence from Mozambique, possibly because SP was in use in Swaziland as a second-line drug. In contrast, there was a high prevalence of the triple-mutant haplotype at Masvold Hospital in northern KZN 40 in 1999. The results given here suggest that the prevalence of dhfr mutations in Swaziland was high compared to that in neighbouring countries. Thus, regional drug pressure alone may not explain allele prevalence in Swaziland. SP drug pressure in Swaziland may also have been further enhanced by the use of other anti-folate compounds, such as the combination antibiotic cotrimoxazole, which have cross-resistance with SP. As HIV infection rates are high in Swaziland, cotrimoxazole is commonly prescribed as prophylaxis against opportunistic infections to HIV patients 2 , and may have contributed to persistence of CIRNI triple-mutant malaria parasites in Swaziland.

The results presented here strongly support the recent move to replace CQ with the highly efficacious artemisinin-combination therapies (ACTs) in Swaziland. The replacement of presumptive treatment with a quick, simple and cheaper diagnostic method such as rapid diagnostic tests (RDTs) to reduce inappropriate prescription of drugs should reduce the cost of ACT drug and further delay development of resistance. Deployment of AL in Swaziland is likely to further reduce prevalence of YYY haplotype but could increase the NFD haplotype, and this may help preserve good AQ efficacy 12 13 41 , as will the recent withdrawal of AQ use in neighbouring Mozambique. In the absence of the pfmdr1 YYY haplotype, AQ shows in vivo and vitro efficacy against parasite clones with the CVIET pfcrt haplotype 13 42 ; parasites of this genotype are shown here to be abundant in Swaziland.

This study is the first to document the presence and circulation of CQR parasites in Swaziland. The high prevalence of pfcrt-76T, pfmdr1-86Y, and pfmdr1-1246Y among 1999 isolates suggests that the population of isolates circulating in Swaziland has been selected by the long use of CQ. These findings thus strongly support the withdrawal of CQ as a first-line anti-malarial drug for uncomplicated P. falciparum malaria, and replacement with ACT. SP remains available for use in pregnant women, but the occurrence of the SGEAA haplotype in Swaziland, previously associated with resistance to sulphadoxine, suggests that SP use for treatment of uncomplicated malaria might rapidly select for these SP-resistant parasites. A weakness of this study is the small number of isolates obtained during the health facility survey in 2007. Genetic analysis of parasites from larger surveys in Swaziland and neighbouring countries is required for better understanding of the development of anti-malarial drug resistance in this region.

The authors declare that they have no competing interests.

SD and CJS conceived the study. SD performed fieldwork. SD and KB performed laboratory work. SD and CJS analysed data. All authors contributed to the paper, and approved the final manuscript.