As elsewhere in the world, a very rapid development of resistance to anti-malarial drugs in Africa requires a regular monitoring in multiple and strategic points. With 53% of the population living in cities against 38% WHO African region, Cameroon is a highly urbanized African country . This demonstrates the importance of epidemiological studies in large cities such as Yaoundé, which currently has a population of over 1,800,000 inhabitants.
In the present study, a high prevalence of mutations associated with drug resistance was found in Yaoundé and its suburbs in both codon 76 of the pfcrt gene (83%) and codon 86 of the pfmdr1 gene (93%) when all samples with mixed genotype were classified as mutant (Table 1). These results are in agreement with several other studies. Previous works of Basco et al carried out in Yaoundé showed that 70% of 157 P. falciparum clinical isolates had the mutant pfcrt 76T codon in 2001 , and a large majority of isolates (88% of 64) carried the pfmdr1 86Y mutant allele between 1997 and 2000 . Similarly, Mbacham et al reported 77% and 76% prevalence of mutant pfcrt 76T and pfmdr1 86Y codons, respectively, in samples collected during the period 2004-2006 in Yaoundé . Despite different classification of double populations and techniques with different sensitivity, the prevalence of mutations appears to increase (pfcrt) or remains at a high and relatively stable level (pfmdr1) until 2009 in spite of the official withdrawal of CQ from Cameroon in 2002. In some endemic areas, stopping the widespread use of CQ resulted in a return of chloroquine-sensitivity associated with the reappearance of wild-type genotypes. In the absence of drug pressure, P. falciparum wild-type haplotypes have a selective advantage over mutants. For example, in 1993, Malawi was the first sub-Saharan African country to replace CQ with SP nationwide in response to the high rates of CQ-resistant falciparum malaria. This change resulted in a decrease in the prevalence of the mutant pfcrt haplotype associated with CQ resistance from 85% in 1992 to 13% in 2000. For pfmdr1 86Y mutant codon, the same study showed similar results but with lower amplitude (from about 60% in 1993 to around 20% in 2000) . The recovery of CQ-sensitivity phenotype and genotype was also observed elsewhere in Malawi , Kenya  and China .
This stability of mutant pfcrt 76T and pfmdr1 86Y genotypes observed in Yaoundé and suburb may be the result of many factors. First of all, the choice of the replacement treatment logically influences the type of selected isolates. The use of SP, which has no influence on the selection of mutant pfcrt and pfmdr1 genotypes, has been shown to favour, by a phenomenon of selective advantage, the reappearance of CQ-sensitive isolates harbouring wild-type pfcrt K76 and pfmdr1 N86 genotypes [39–41]. The use of AL or artesunate-mefloquine (AS-MQ) seems to favour the return to the predominance of wild-type pfmdr1 N86 genotype and, to a lesser extent, to the wild-type pfcrt K76 genotype by an active selection [14, 43–45]. Inversely, AQ, a close Mannich base analogue of CQ, or AS-AQ promotes the maintenance of CQ-resistant isolates with the mutant pfcrt and pfmdr1 genotypes by an active selective pressure [20, 46], as observed in the present study. Whereas in East African countries like Malawi or Kenya, SP or AL had largely replaced CQ , in Yaoundé, in 2005, AQ was still prescribed as a first-line anti-malarial drug in 20% and 63% of adults and children under five years old, respectively , and AS-AQ in 4.5% and 1.5%. AL was used only in 8.3% and 2.4%, AS-MQ in 1.5% and 0.8%, and SP in 5.8% and 0% of adults and children less than five years old, respectively .
Secondly, the changes of P. falciparum resistance phenotype and genotype after the withdrawal of CQ depend on the rapidity of drug replacement. For example, in Malawi where a profound and rapid return to CQ sensibility was observed, the change in drug policy from CQ to SP was swift and efficient, so that SP became the only available anti-malarial drug in less than one year after the implementation of the new drug policy. In contrast, these changes were progressive and lasted several years in many areas as in Cameroon. In fact, in Yaoundé, although the National Malaria Control Programme of Cameroon had replaced CQ by AQ in 2002 and then AQ monotherapy by AS-AQ since January 2004, CQ was still largely accessible through the informal outlets (e.g. food market) in August 2005 .
Finally, in a more general way, fitness loss of mutant P. falciparum might be associated with the development of compensatory mechanisms able to maintain mutant parasites even in the absence of drug pressure . This feature might explain, at least in part, the persistence of mutant pfcrt codon in Southeast Asia and South America [49–51] and also in Cameroon, as described here.
In Mfou, a higher frequency of mixed pfcrt haplotypes was observed at the expense of mutant pfcrt population. This observation was not done for pfmdr1 haplotypes. A possible reason for this observation is a drug pressure selection different from that existing in Yaoundé.
Since the probes used to detect the mutation in codon 76 of pfcrt gene were not able to detect that of codon 72, a new technique using LNA probes was developed in the present study to discriminate the mutant SVMNT haplotype (72S mutation) from the wild-type CVIET haplotype (C72 wild-type). Previous studies and data collected from countries like Bolivia or India suggested that AQ has an early and prominent role in the selection of parasites carrying SVMNT haplotype associated with drug resistance . These parasites are highly resistant to AQ, but only moderately resistant to CQ. Contrary to CVIET haplotype, once the SVMNT haplotype emerges in a given parasite population and CQ and AQ are removed, the repopulation of sensitive strains may be very slow to occur . As the SVMNT haplotype was recently described in Tanzania and Angola [22, 23], it was important to verify whether this haplotype existed in Yaoundé. None of the samples tested for the codon 72 of pfcrt was found to carry the SVMNT haplotype. These results are contrary to what was observed in nearby African countries, such as in Ghana , Tanzania  and Angola  where the prevalence of this haplotype was between 3.9% and over 50%. It is possible that the observed predominance of SVMNT haplotype in Angola is the result of frequent travels of Brazilian and Angolan citizens between the two countries , which is not the case in Cameroon. However, the monitoring of pfcrt codons 72-76 should be pursued because AQ has long been prescribed in Cameroon before and since the cessation of the use of CQ (2002) and until 2005  and seems to have an important role in the selection of the SVMNT haplotype .
The amplification of pfmdr1 gene has been more closely linked to MQ and halofantrine (HAL) resistance [53–55]. In this study, pfmdr1 amplification was not observed in Yaoundé between 2005 and 2009. Elsewhere in Africa, the situation seems to be contrasted. In various studies conducted in East Africa, only four samples were found with pfmdr1 gene duplication, one in Kenya and three in Sudan (near the Ethiopian border) among 475 isolates tested (57 in Sudan , 72 in Kenya , 186 in Zanzibar  and 160 in Malawi ). In West Africa, on the one hand, none of 580 samples tested in Liberia and Guinea-Bissau between 1981 and 2005 was found to be duplicated ; on the other hand, two studies had identified in Burkina Faso, Ivory Coast, Togo and Madagascar, six pfmdr1 duplications among 112 samples tested [27, 56]. In Central Africa, data are limited since only 32 samples were screened and all of them had a single copy of pfmdr1 gene . In this region, an exception is the study of Uhlemann et al who found the duplication of pfmdr1 gene in five of 62 clinical isolates tested (8%) in 1995 in Lambaréné, Gabon . Four of these five patients harboured the wild-type N86 pfmdr1 codon even though during this period 90% of isolates carried the mutant pfmdr1 codon 86 around Lambaréné . However, in 2002 at the same study site, none of 37 samples tested had pfmdr1 gene duplication. These observations on pfmdr1 gene amplification in Lambaréné are difficult to explain outside of the possible selection of such a clone by previous clinical trials on the same site, using low dose of mefloquine [58–60]. Nevertheless, these data showed that P. falciparum isolates from Central Africa can have pfmdr1 gene duplication.
The lack of pfmdr1 gene duplication in Yaoundé may possibly be due to the very low use of MQ or HAL, which represented only 1.5% of first-line treatments against malaria in 2005 , but also partly to the high prevalence of the pfmdr1 Y86 mutant allele. Indeed, in Southeast Asia, pfmdr1 amplification has been suggested to be incompatible in the presence of the mutant pfmdr1 86Y allele . However, the conclusion of that Asian study has not been confirmed in Africa, where the existence of parasites harbouring a duplicated pfmdr1 gene with mutant 86Y codon has been reported from Sudan , Gabon  and Ivory Coast [27, 56].
The molecular analysis performed in the present study did not find any pfcrt 72S mutation, which may be a good sign for the continued use of AQ in combination with AS. A regular evaluation of AS-AQ efficacy, in parallel with molecular surveillance, is required to ensure the utility of AS-AQ in Cameroon. This ACT contributes to the maintenance of a high prevalence of mutant pfcrt 76T and pfmdr1 86Y alleles. The pressing question is to predict how these parasites will evolve in the presence of AL pressure. Several scenarios can be envisioned. Firstly, they could behave like Southeast Asian isolates and will not progress to the duplication of pfmdr1 gene in the absence of wild-type pfmdr1 N86 allele. Secondly, as already observed in some cases in Africa [9, 27, 56, 57], the parasites may acquire multicopies of pfmdr1 despite the pfmdr1 86Y mutation. Only regular and exhaustive molecular monitoring of P. falciparum clinical isolates can provide the answer. However, the relevance of these results would be improved if they were associated with information on different anti-malarial drugs that are really taken by the patients because these data often differ from the current national recommendation.