Large scale prevalence of Plasmodium falciparum chloroquine (CQ), amodiaquine (AQ) and sulphadoxine/pyrimethamine (SP) resistance has forced countries with endemic malaria to change to the artemisinin combination therapies (ACTs) as first-line treatment against uncomplicated malaria. The ACT drugs, currently being implemented in sub-Saharan Africa combine various artemisinin analogues with novel drugs (lumefantrine, i.e. Coartem^®) or with already widely used drugs such as AQ and SP. ACTs are highly efficacious and are likely to decrease the rate of development of resistance, and thus expand the useful lifetime of the drugs 1. Artemisinin derivates are acting and are eliminated rapidly after treatment and the development of resistance to ACTs will probably depend on the already existing background level of resistance to the partner drugs in the parasite population, as it has been observed in areas with high levels of resistance to SP and AQ 23. Currently, SP combined with artesunate (ART) is the first-line treatment policy in few African countries, such as Mozambique (from 2004) and Northern Sudan 45. In Mozambique, prior to adoption of the ACT strategy, the combination of AQ and SP as the first-line treatment was implemented in late 2002 following reports of high levels of chloroquine (CQ) resistance 6. However, one year prior to this change, SP alone was used to treat uncomplicated malaria cases in the study area. Additionally, SP has recently been used in programmes examining the efficacy of intermittent presumptive treatment (IPT) intervention in pregnant women and infants in Mozambique 78.

The prevalence of combinations of single nucleotide polymorphisms (SNP) in the P. falciparum dihydrofolate reductase (Pfdhfr) and dihydropteroate synthetase (Pfdhps) genes have been correlated with resistance to SP in vivo 910. In Africa, the Pfdhfr triple mutation, N51I+C59R+S108N, with wildtype at codons 50 (C50) and 164 (I164), is combined into a highly prevalent and resistant haplotype, CIRNI. For Pfdhps the double mutation, A437G + K540E, with wildtype at codons 436 (S436) 581 (A581) and 613 (A613) are combined into the resistant haplotype SGEAA. The combination of the two; CIRNI-SGEAA, have been suggested as a molecular marker of in vivo resistance 10. However, recent studies have questioned the association between this combination of molecular markers and drug resistance in vivo by showing that differences in responses to SP treatment was a consequence of acquired immunity rather than parasite factors 1112.

Resistance to both CQ and AQ is mainly associated with a single K76T mutation in the P. falciparum chloroquine resistance transporter gene (Pfcrt), but are probably involving mutations in the P. falciparum multi-drug resistance (Pfmdr1) gene as well 1314. Three main haplotypes in codons 72–76 of the Pfcrt gene exist, namely the wild type CVMNK and the CQ resistant haplotypes, CVIET and SVMNT 15, where CVIET is by far the most dominant mutant haplotype in Africa while SVMNT haplotypes has only been observed sporadically 16.

The objective of this study was to determine the frequency of Pfdhfr, Pfdhps and Pfcrt resistance related haplotypes in P. falciparum isolates collected before and a year after SP was introduced as first-line treatment, as replacement for CQ in the study area. Moreover, since Mozambique was implementing a treatment policy with SP in combination with artesunate as first line treatment of uncomplicated malaria, the findings of this study in addition to recent studies 1718 is providing crucial baseline information of the level of drug resistance related haplotypes present in the parasite population prior to the adoption of the ACT strategy in the country.

A total of 244 and 192 samples were P. falciparum positive by microscopy in cross-sectional survey 1 and 2, respectively out of which 171 and 173 samples were chosen randomly for the genotyping analysis. The populations in the two surveys differed in respect to age and parasite density: In survey 2, the median age was significantly lower (P = 0.013) while the parasite density was significantly higher (P < 0.001) compared to survey 1 (Table 1). Only Pfdhfr wildtypes at c164 was found, while for Pfdhps, only wildtypes were observed at c518 and c613.

A significant increase in the frequency of Pfdhps double mutant haplotype, SGEAA from 13.5% to 34.8% over a year was found (χ² = 12.24, P < 0.001), mainly decreasing the frequency of Pfdhps wildtypes SAKAA/AAKAA from 80.8% to 64.3% (χ² = 5.893, P = 0.015; Figure 1A). For Pfdhfr, there was only a marginal difference between the two surveys with a frequency of the triple mutant haplotype, CIRNI at 46.5% and 53.2% in survey 1 and 2, respectively (χ² = 0.70, P = 0.405; Figure 1B). When combining the two loci, the frequency of parasites carrying the highly resistant Pfdhfr/Pfdhps quintuple haplotype increased from 8.0% to 25.8% (χ² = 7.78, P = 0.005) between survey 1 and 2. The change in frequency was not at the expense of the wildtype Pfdhfr/Pfdhps haplotype, CNCSI-SAKAA that remained stable at 11%, but rather on the double mutant haplotypes, where the CNRNI-SAKAA accounted for a decrease from 26.7% to 18.0% (χ² = 1.32, P = 0.250) and the triple mutant haplotype, mainly the CIRNI-SAKAA that decreased from 37.3% to 27.0% (χ² = 1.57, P = 0.210) (Figure 2). Logistic regression analysis revealed that all of the above mentioned statistical results (with regards to significance) persisted after appropriate adjustment for P. falciparum density and age.

The prevalence of single nucleotide polymorphisms in Pfdhfr and Pfdhps is shown in Table 2. In Pfdhfr, only minor differences can be seen between the two cross sectional surveys where a strong trend for a decrease in the prevalence of the 108S wildtype is apparent (from 49.6% to 38.0%, χ² = 3.06, P = 0.080). For Pfdhps, as expected from the combined haplotype data, both a decrease of the 436/437SA wildtype (from 85.7% to 71.0%, χ² = 7.18, P = 0.007) and an increase in the mutant 436/437SG mutant type (from 56.4% to 68.5%, χ² = 3.46, P = 0.063) is evident. Furthermore, the prevalence of the 540 K wildtype is decreasing significantly from 92.1% to 71.6% (χ² = 16.66, P < 0.001).

Major differences in prevalence of infections with Pfdhps and combined Pfdhfr/Pfdhps mutations were found when comparing different age groups (i.e. 1–5, 6–14 and > 14 years), however only in the second survey (Table 3): The prevalence of quintuple mutant haplotype was significantly higher in children under fives compared to the older age groups (χ² = 6.04, P = 0.049) and a strong tendency for a higher frequency of Pfdhps double mutant haplotypes in children under fives compared to the older groups was seen (χ² = 5.33, P = 0.069).

The samples analysis of the Pfcrt gene showed almost exclusively the occurrence of the mutant haplotype CVIET in both years (92.8% in survey 1 and 95.3% in survey 2; χ² = 0.74, P = 0.389), while the remaining samples were the wild type CVMNK haplotype.

In late 2002, Mozambique changed the treatment policy from CQ to SP/AQ against uncomplicated malaria. Prior to this change, the present study implemented SP alone as first line treatment in the study area and as a possible consequence, the study observed a rapid increase in frequency of double SGEAA-Pfdhps and quintuple CIRNI-SGEAA-Pfdhfr/Pfdhps mutant haplotypes at the community level just a year after. Prior to the introduction of SP the consumption of this drug was very low in the study area 19. However, since the study did not measure the actual use of SP outside the health services, the study cannot rule out alternative explanations such as consumption of other antifolate drugs and for instance changes in migration. The triple mutant CIRNI-Pfdhfr haplotype, was highly prevalent prior to the change (46.5%), also observed in 2000 in Matola (50.9%), a peri-urban area within Maputo province 21. Similarly, Raman et al observed a remarkable increase of the triple mutant haplotype in 13 sentinel sites stratified into three zones in Maputo province between 1999 and 2003, almost reaching 100% in 2003 18, while Fernandes et al found 93% of the triple mutant haplotype in samples from children with uncomplicated malaria attending a Health Centre in Maputo city of Mozambique in 2004 17.

This high frequency of Pfdhfr mutations before the actual implementation of SP/AQ as first-line drug in Mozambique may be due to the use of other chemically related drugs such as co-trimoxazole, which is commonly used in the study area for treating bacterial infections and as a preventive measure in HIV/AIDS patients (Dr Yunuss Assane, MoH, personal communication). While not shown in vivo, in vitro cross-resistance has been reported between co-trimoxazole and SP components 22. More importantly, as suggested by Raman et al, since neighbouring KwaZulu-Natal province in South Africa implemented SP as early as 1988, the high frequency before SP implementation in Mozambique may be caused by SP drug pressure from the border region 18.

In the present study, the most notable change was the increase of the double mutant Pfdhps SGEAA haplotype from 13.5% to 34.8% over only one year. In the study by Raman et al, the frequency of the double mutant haplotype peaked in 2001 at 19–25% in all three zones followed by a remarkable and significant decrease in 2003 and 2004 to less than 10% 18, in sharp contrast to the findings by Fernandes et al, where the frequency was 51% in their 2004 study 17, presumably due to differences in sampling where the latter study was done on samples from symptomatic children attending a health centre. Such a difference in frequency in symptomatic cases versus asymptomatic carriers has likewise been shown in a study performed in Tanzania 23.

In the second survey of this study, the quintuple Pfdhfr/Pfdhps haplotype was more frequent in children under fives compared to the older age groups. Similar findings, but rather for Pfdhfr haplotypes alone was seen by Raman et al 18 and is most likely due to a stronger survival advantage of the mutant parasites in the younger age groups due to the more frequent treatments this group receive. It also pinpoints a potential pitfall when measuring prevalence of molecular markers of drug resistance in under-fives when these data are extrapolated to cover the population as a whole.

The high prevalence of these mutant haplotypes in the area as early as 2001 along with the findings by Fernandes et al 17 and Raman et al 18, questions the current efficacy of the first-line antimalarial drug in Mozambique, a combination of ART and SP. Results from in vivo study carried out in the same area in 2001 have described a relatively high adequate clinical response to SP after 14 days of follow-up at 82.7% 6. However, the levels of treatment failures and RI and RII parasitological resistance increased when follow-up was extended to 28 days, indicating a decline in the efficacy of SP as early as 2001. Most likely, initially ART/SP was an efficient treatment against uncomplicated malaria in Mozambique, however the risk of clinical treatment failures could be expected to appear within a short span of years, since the benefits of ACT combinations are diminished by the inclusion of an ineffective partner drug, such as SP, where ART/SP may be equivalent to ART monotherapy. Despite that mutations in Pfdhfr/Pfdhps may not directly cause clinical ART/SP treatment failures, a high prevalence of SP resistance-related mutations may rather indicate a risk of emerging tolerance to ART endangering the whole ACT strategy in the region. Close and frequent monitoring of the efficacy of the drug combination should be performed as well as formulation of a strategy for alternative first line treatment.

The high frequency of quintuple Pfdhfr/Pfdhps haplotypes found here as early as 2002 most likely cause ideal conditions for the development of ART tolerance in the P. falciparum populations and may even endanger the sensitivity to the second-line drug, Coartem.

SE, PM, AMR, RT and MA conceived and designed the study. SE, FA, FXG-O and RT was responsible for the cross sectional sampling. SE performed the experiments and analyses with MA. SE, FXG-O, PM, FA, RT and MA participated in manuscript preparation. All authors read and approved the final manuscript.