pfmdr1 mutations associated with chloroquine resistance incur a fitness cost in Plasmodium falciparum

Authors


Summary

Efforts to control malaria worldwide have been hindered by the development and expansion of parasite populations resistant to many first-line antimalarial compounds. Two of the best-characterized determinants of drug resistance in the human malaria parasite Plasmodium falciparum are pfmdr1 and pfcrt, although the mechanisms by which resistance is mediated by these genes is still not clear. In order to determine whether mutations in pfmdr1 associated with chloroquine resistance affect the capacity of the parasite to persist when drug pressure is removed, we conducted competition experiments between P. falciparum strains in which the endogenous pfmdr1 locus was modified by allelic exchange. In the absence of selective pressure, the component of chloroquine resistance attributable to mutations at codons 1034, 1042 and 1246 in the pfmdr1 gene also gave rise to a substantial fitness cost in the intraerythrocytic asexual stage of the parasite. The loss of fitness incurred by these mutations was calculated to be 25% with respect to an otherwise genetically identical strain in which wild-type polymorphisms had been substituted at these three codons. At least part of the fitness loss may be attributed to a diminished merozoite viability. These in vitro results support recent in vivo observations that in several countries where chloroquine use has been suspended because of widespread resistance, sensitive strains are re-emerging.

Introduction

The emergence and spread of multidrug-resistant Plasmodium falciparum constitutes a considerable obstacle to efforts to control malaria, particularly in sub-Saharan Africa where the bulk of clinical cases occur. The repertoire of antimalarial drugs is limited, and most are far more expensive than conventional first-line treatments like chloroquine or sulfadoxine-pyrimethamine. Rational disease management strategies, which make maximum utility of the available drugs, are therefore required (WHO, 2001).

Although the molecular basis for the resistance of P. falciparum to many drugs has not yet been fully characterized, there is strong in vitro evidence that P. falciparum chloroquine resistance transporter (PfCRT) and P-glycoprotein homologue 1 (Pgh1), two proteins localized to the parasite's digestive vacuole membrane (Cowman et al., 1991; Fidock et al., 2000), can play important roles in determining sensitivity to multiple antimalarials including chloroquine (CQ), the most widely used and extensively studied antimalarial. Laboratory studies have shown that transfection of chloroquine-sensitive (CQS) parasites with the pfcrt gene from several different chloroquine-resistant (CQR) strains can confer resistance to CQ, and increase susceptibility to quinine, mefloquine and artemisinin (Fidock et al., 2000; Sidhu et al., 2002). Similar allelic exchange experiments have demonstrated that mutations in Pgh1, encoded by the pfmdr1 gene, can modulate the degree of sensitivity to CQ and artemisinin, and confer resistance to mefloquine, quinine and halofantrine (Reed et al., 2000).

Field surveys confirm that pfmdr1 and pfcrt are probably the primary determinants of parasite resistance to CQ and other quinoline drugs. In isolates from most areas of the world where falciparum malaria is endemic, there is typically a positive correlation between mutations in either or both these genes, and in vitro CQ resistance (Foote et al., 1990; Zalis et al., 1998; Duraisingh et al., 2000a; Babiker et al., 2001; Durand et al., 2001; Vieira et al., 2001; Chen et al., 2003; Lim et al., 2003; Pickard et al., 2003). However, with the identification of at least four different foci of emergence of CQ resistance (Wootton et al., 2002), the genetic basis of resistance to CQ may well vary in different geographical regions, albeit with strong tendencies for convergence at the pfmdr1 and pfcrt loci. A recent theoretical analysis has indicated that as many as nine putative transporter genes (including pfmdr1 and pfcrt) show significant associations with decreased CQ sensitivity, with linkage disequilibria suggesting there may be multiple interactions between these genes (Mu et al., 2003).

It has been shown in a wide variety of organisms that acquisition of drug resistance can incur a biological cost (Bjorkman et al., 2000; Nagaev et al., 2001; Fohl and Roos, 2003; Lu et al., 2004; Mariam et al., 2004). Genetic mutations that alter the structure or function of key proteins in order to evade the actions of deleterious compounds may also compromise the normal function of those proteins. In the absence of the selective pressure favouring the mutant form, the drug-resistant organism is often less fit and can be displaced in the population by the sensitive wild-type, provided the latter has not itself been eliminated.

There have not been many studies of the relative ability of drug-resistant forms of Plasmodium species to persist when drug pressure is removed. Shinondo et al. (1994) showed that in the absence of selection, pyrimethamine-resistant P. berghei parasites underwent sporogony in the mosquito vector more slowly than the pyrimethamine-sensitive strain from which they were derived, consistent with the sensitive parasites having a transmission advantage. In a more direct comparison, Rosario et al. (1978) showed that under drug-free conditions, a pyrimethamine-sensitive strain of P. chabaudi outgrew a pyrimethamine-resistant strain in two out of three experiments in which mice were inoculated with equal proportions of the two strains. However, in the same study a CQS strain was outgrown by a CQR strain in all four experiments. In another direct comparison, Peters et al. (2002) showed that in mixed cultures of erythrocytic stages of P. falciparum, atovaquone-resistant lines were 5–9% less fit than the atovaquone-sensitive strain from which they had been derived, and were subsequently outcompeted in the absence of selection.

Several recent epidemiological studies also support the hypothesis that drug-resistant P. falciparum parasites may be at a competitive disadvantage when drug pressure is removed. In vitro CQ resistance in Hainan, China, gradually decreased from 98% in 1981 to 61% in 1991 following use of the drug being officially discontinued in 1979 (Liu et al., 1995). In Vietnam, where artemisinin replaced CQ as the first-line treatment for falciparum malaria in 1992, comparative studies have shown that in vitro and in vivo parasite sensitivity to CQ has increased significantly over time (Nguyen et al., 2001; Thanh et al., 2001; Nguyen et al., 2003). A field survey in Malawi, where sulfadoxine-pyrimethamine replaced CQ as the first-line antimalarial in 1993, showed that in vitro chloroquine resistance in the local area had decreased from 47% in 1988 to 3% in 1998, noting this was also accompanied by a significant reduction in the prevalence of one of the key pfcrt mutations associated with resistance (Mita et al., 2003). This was subsequently shown to be attributed to expansion of the wild-type pfcrt allele in the parasite population, rather than a back mutation in the mutant allele (Mita et al., 2004). In a similar survey, Kublin et al. (2003) found that the increases in in vitro and in vivo parasite sensitivity to chloroquine observed in Malawi were accompanied by significant decreases in the frequency of mutations in pfcrt and pfmdr1 associated with resistance.

In this study we have investigated whether, in the absence of selective pressure, the parasite's acquisition of particular CQ resistance-enhancing mutations in Pgh1 incurs a cost in terms of parasite fitness. Using transfected P. falciparum strains described previously (Reed et al., 2000), we have demonstrated that pfmdr1 mutations associated with enhanced CQ resistance incur a ∼25% fitness cost relative to the wild-type. Analysis of the growth cycle of the parasites suggests that this biological cost may be attributed, at least in part, to a reduction in the viability of developing merozoites.

Results

pfmdr1 mutations, which modulate CQ resistance, are associated with a fitness cost

The three strains used in these experiments all had their endogenous pfmdr1 locus modified by allelic replacement, as described previously (Reed et al., 2000). Briefly, D10-mdrD10 is a transfection control derived from the CQS strain D10, and expresses wild-type Pgh1. 7G8-mdr7G8 is a transfection control derived from the CQR strain 7G8, in which four single-nucleotide polymorphisms in pfmdr1 result in a mutant form of Pgh1 containing four amino acid changes relative to the wild-type: Y184F, S1034C, N1042D and D1246Y. 7G8-mdrD10 is also derived from the CQR strain 7G8, but has had its pfmdr1 gene partially replaced by that of D10, so that it retains the Y184F mutation of its 7G8 parent, but has the wild-type amino acids at the other three positions. The CQ sensitivity of 7G8-mdrD10 was previously shown to be intermediate between that of D10-mdrD10 and 7G8-mdr7G8, and the level of expression of Pgh1 was found to be the same in all three strains (Reed et al., 2000). DNA fingerprinting by Southern blot analysis of genomic DNA confirmed that 7G8-mdr7G8 and 7G8-mdrD10 had identical band patterns (data not shown), indicating that no gross genomic changes or cross-contamination with other laboratory cultures had occurred. We also found (data not shown) by polymerase chain reaction (PCR) that the strains retained their respective transfection plasmids (integrated into the parasite genome), as described previously (Reed et al., 2000). Further, assays of drug sensitivity including quinine, halofantrine, mefloquine and artemisinin reproduced the spectrum of sensitivity profiles of these strains obtained in the initial study (data not shown).

Cultures containing 1:1 mixtures of different pairings of these three strains were grown in the absence of any selective drug pressure. Comparisons between 7G8-mdr7G8 and 7G8-mdrD10 permit direct analysis of the effects of polymorphisms in pfmdr1. Comparisons involving D10-mdrD10 do not invite such conclusions; they simply provide an indication of the magnitude of fitness differences that may exist between parasites of different lineages.

The composition of the cultures was monitored by periodically testing the mixture for sensitivity to CQ. All mixed cultures initially displayed a biphasic CQ dose–response curve, indicative of the characteristics of both parent strains (Fig. 1), with an inflexion point at around 50% growth confirming that the cultures started as equal proportions. It should be noted that CQS strains accumulate much higher quantities of CQ than CQR strains; an increase in the number of CQS parasites results in increased depletion of CQ from the medium (as a result of uptake of the drug by the CQS parasites) and a consequent rightward shift of the CQ dose–response curve. This ‘inoculum effect’ (Gluzman et al., 1987; Geary et al., 1990) is the reason for the CQS component of the biphasic curve of mixtures containing D10-mdrD10 showing a leftward shift relative to the dose–response curve for the parental D10-mdrD10 strain (Fig. 1A and C).

Figure 1.

CQ dose–response curves showing the change in drug-sensitivity of representative mixed cultures over time. (A) D10-mdrD10 (CQS) and 7G8-mdr7G8 (CQR); (B) 7G8-mdr7G8 (CQR) and 7G8-mdrD10 (1/2 CQR); and (C) D10-mdrD10 (CQS) and 7G8-mdrD10 (1/2 CQR). These were cultured in the absence of selective pressure until the biphasic curve characteristic of the mixed population resolved to and could not be distinguished from the curve representing one of the two parent strains (broken lines). In mixed cultures where the CQR strain 7G8-mdr7G8 was outcompeted (A and B), 75 nM CQ was added to the culture medium and CQ sensitivity was again measured once sufficient parasites were growing in the presence of the drug.

The CQR strain 7G8-mdr7G8 was progressively outcompeted by both the CQS strain D10-mdrD10 and the partially resistant (1/2 CQR) strain 7G8-mdrD10 (Fig. 1A and B). After 34 days, the dose–response curve of the D10-mdrD10/7G8-mdr7G8 mixture could not be distinguished from that of D10-mdrD10 (Fig. 1A). Similarly, the dose–response curve of the 7G8-mdrD10/7G8-mdr7G8 mixture could not be distinguished from that of 7G8-mdrD10 after 34 days (Fig. 1B), at which point the first part of the experiment was deemed to have reached completion.

Subsequently, 75 nM CQ was added to these two cultures to test whether the CQR strain 7G8-mdr7G8 was still present in the culture, albeit at levels not measurable by this assay. This caused the amount of parasites in both cultures to drop to levels below those detectable by routine microscopic examination of thin-film slides, after 9 days in the case of D10-mdrD10/7G8-mdr7G8, and after 22 days in the case of the more resistant mixture of 7G8-mdrD10 and 7G8-mdr7G8. In both cases, approximately 14 days after the parasitaemia had decreased to below detectable levels, the parasitaemia recovered to above 10%, at which point CQ sensitivity tests were again performed on the D10-mdrD10/7G8-mdr7G8 and 7G8-mdrD10/7G8-mdr7G8 mixed cultures. These clearly showed that the CQR 7G8-mdr7G8 was now dominating both cultures, as expected by the selection regime (Fig. 1A and B).

In contrast to the situation with the D10-mdrD10/7G8-mdr7G8 and 7G8-mdrD10/7G8-mdr7G8 mixtures, it appeared that there was very little difference between the growth rates of 7G8-mdrD10 and D10-mdrD10; after 62 days the biphasic curve was still very close to that at the start of the competition, with periodic fluctuations to either side of the level of the 1:1 mixed culture (Fig. 1C).

Allele-specific real-time PCR allows quantitation of the loss of fitness

While monitoring for changes in CQ sensitivity, genomic DNA from each mixed culture was periodically isolated, and the concentration of the specific pfmdr1 alleles present was measured by real-time PCR as described in Experimental procedures. This strategy had been verified by the successful identification of the proportions of 7G8-mdr7G8 and 7G8-mdrD10 in unknown mixed samples in blind tests (data not shown).

The real-time PCR assay confirmed that 7G8-mdr7G8 parasites were outgrown by both D10-mdrD10 and 7G8-mdrD10 parasites (Fig. 2A and B), and confirmed that there was no significant fitness difference between 7G8-mdrD10 and D10-mdrD10 (P = 0.88, Student's t-test), as there were periodic fluctuations in the levels of these two strains over the course of 62 days but no overall trend in favour of either (Fig. 2C).

Figure 2.

Results of competition within representative mixed cultures as determined by allele-specific real-time PCR. (A) D10-mdrD10 and 7G8-mdr7G8; (B) 7G8-mdr7G8 and 7G8-mdrD10 and (C) D10-mdrD10 and 7G8-mdrD10. The ordinate shows the log of the ratio of the sizes of the two strain populations present in the culture. The ‘fitness’ of the CQR strain 7G8-mdr7G8 was ∼70% that of the CQS strain D10-mdrD10 (A). Similarly, 7G8-mdr7G8 had ∼75% of the fitness of 7G8-mdrD10 (B), a difference attributable to the three single-nucleotide polymorphisms in pfmdr1 by which these two strains differ. After day 34, 75 nM CQ was added to the culture medium to reselect the outcompeted CQR strain 7G8-mdr7G8 (A and B). There was no significant difference in the fitness of D10-mdrD10 and 7G8-mdrD10 (C).

A plot of the logarithm of the ratio of the two strains in a given mixture against time produced a linear relationship, from which the fitness differential could be derived (Hartl and Clark, 1997). Each competition experiment was performed twice in succession. The fitness defect of 7G8-mdr7G8 with respect to the wild-type D10-mdrD10 was ∼30% per generation, while the fitness defect in 7G8-mdr7G8 relative to 7G8-mdrD10 was ∼25% per generation. The fitness defects derived in these two mixtures were not significantly different from one another (P = 0.09; NB  there is increasing error as ratios get larger), an observation consistent with the lack of significant fitness difference in the D10-mdrD10/7G8-mdrD10 mixture (Fig. 2C).

The pfmdr1-associated fitness cost is attributed in part to a decrease in merozoite viability

In all organisms, the fitness of an individual of any particular phenotype is generally defined in terms of two factors, survival and reproduction (Hartl and Clark, 1997). Having established that 7G8-mdr7G8 was less fit than both 7G8-mdrD10 and D10-mdrD10 in the absence of selection, a preliminary investigation of the viability and fecundity of the three strains was performed in order to try and determine the cause.

Approximately 40–42 h after invading the erythrocyte, the now mature trophozoite-stage parasite subdivides to produce new merozoites enclosed within the parasitophorous vacuole membrane, a process known as schizogony. At around 48 h post invasion the infected erythrocyte lyses, allowing the merozoites to escape into the extracellular medium where they invade new, uninfected erythrocytes and rapidly develop into the characteristic ‘ring’ stage. Thin-film slides collected during the course of three successive generations of unmixed cultures of the three strains were analysed. It was found that there was no significant difference between the average number of merozoites (∼20) produced by each strain at schizogony (P > 0.1) (Fig. 3A). However, the number of ring-stage parasites, which were produced in the subsequent intraerythrocytic cycle, did differ significantly between the three strains (Fig. 3B). For each D10-mdrD10 trophozoite, there were 14.1 ± 0.9 (±SEM, n = 3) rings in the succeeding generation, compared with 11.4 ± 0.3 for 7G8-mdr7G8. With its altered pfmdr1 locus, 13.2 ± 0.7 7G8-mdrD10 rings were produced per trophozoite; this was significantly higher than 7G8-mdr7G8 (P = 0.016), but not significantly different from D10-mdrD10 (P = 0.23).

Figure 3.

Replicative success of the individual strains in standard in vitro culture conditions. Although there was no significant difference in the average number of merozoites produced per mature schizont amongst all three strains (A), there were significantly more rings produced per D10-mdrD10 (14.1 ± 0.9) and 7G8-mdrD10 (13.2 ± 0.7) trophozoite than was produced by 7G8-mdr7G8 trophozoites (11.4 ± 0.3).

These figures indicate an ∼19% deficit in the ability of 7G8-mdr7G8 merozoites to invade new erythrocytes and develop to the ring stage when compared with D10-mdrD10, and an ∼14% deficit with respect to 7G8-mdrD10. Highly consistent parasitaemias at different stages within the same generation indicated that there was no significant parasite mortality associated with the development from ring stage to trophozoite stage in any generation of any strain (data not shown).

Discussion

The replacement of the mutant pfmdr1 codons 1034C, 1042D and 1246Y with the wild-type codons S1034, N1042 and D1246 in the CQR P. falciparum strain 7G8 had previously been shown to reduce the level of CQ resistance in the parasite (Reed et al., 2000). Using the same transfected strains in the present study, we have shown that these resistance-modulating pfmdr1 mutations impose a significant in vitro fitness cost on the intraerythrocytic stage of the parasite. When mixed in equal proportions and cultured together under standard conditions, the CQR transfectant 7G8-mdr7G8 was outgrown by 7G8-mdrD10 (Figs 1B and 2B), a transfectant with reduced CQ resistance and differing from 7G8-mdr7G8 by only three single-nucleotide polymorphisms at the pfmdr1 locus. In contrast, no significant fitness difference was observed between 7G8-mdrD10 and D10-mdrD10, a CQS strain that also outgrew 7G8-mdr7G8. The same results were obtained in two independent competition experiments. Although 7G8-mdrD10 was considerably more fit in the absence of CQ selection than its parent 7G8-mdr7G8, and not significantly different in fitness from D10-mdrD10 (Figs 1C and 2C), we do not suggest that pfmdr1 is the only factor that determines the levels of fitness in the D10 and 7G8 lineages. Rather, we simply conclude that alterations in pfmdr1 were sufficient to overcome the in vitro fitness differences observed between D10-mdrD10 and 7G8-mdr7G8 (Figs 1A and 2A).

The difference between 7G8-mdr7G8 and 7G8-mdrD10 was principally attributable to the enhanced viability of 7G8-mdrD10 merozoites, which produced more ring-stage parasites in each subsequent generation than were produced by an equivalent number of 7G8-mdr7G8 merozoites. Whether this is attributed to the formation of merozoites with a reduced ability to invade new erythrocytes, or to survive the first 1–4 h post invasion, requires further investigation. At this stage it is not clear how mutations in Pgh1, the protein encoded by pfmdr1, might cause such a pronounced effect on parasite development or overall fitness, as the normal physiological functions of Pgh1 remain to be established. It has been postulated that CQ resistance is mediated by the active efflux of the drug from the parasite's digestive vacuole (Krogstad et al., 1987; Sanchez et al., 2003), where it has its toxic effect (Slater, 1993; Sullivan et al., 1996). If Pgh1, a member of the ABC transporter superfamily and a homologue of the human P-glycoprotein (Foote et al., 1989; Wilson et al., 1989), is one of the molecules diverted to the process of drug efflux via an altered activity or substrate specificity, it could be at the expense of any number of processes responsible for normal parasite development and maturation. There might also be differences in generation time for these different lines. If so, however, they appear to be quite subtle, as the cultures used to generate the merozoite viability data developed uniformly during the three consecutive cycles of parasite growth studied, so that there was no discernible lag at the end of this experiment.

It should be noted that the in vitro cost imposed by these particular pfmdr1 mutations in the absence of drug pressure may not necessarily translate to an in vivo cost. Several studies have shown that compensatory mutations elsewhere in an organism can ameliorate the costs of acquisition of resistance (Sherman et al., 1996; Bjorkman et al., 1998; Nijhuis et al., 1999; Levin et al., 2000; Bjorkholm et al., 2001). The fitness disadvantage incurred by the mutant form of pfmdr1 during the erythrocytic, asexual stage may not necessarily result in a net disadvantage in the context of the complete parasite life cycle. Further investigation is required into whether the mutant form of pfmdr1 may, for example, increase gametocytogenesis and thus enhance transmission to the mosquito vector (Tchuinkam et al., 1993; Robert et al., 1996; Mendez et al., 2002), alter sporogony in the mosquite vector (Shinondo et al., 1994), or increase parasite viability/efficiency in the human host (Wernsdorfer et al., 1995).

Even if certain mutations in pfmdr1 (and pfcrt) do result in fitness differences in vivo, CQR strains bearing such mutations will not necessarily be displaced from the parasite population in the absence of drug selection. In some parts of south-east Asia and most of South America, polymorphisms associated with resistance in one or both these genes are near-ubiquitous (Povoa et al., 1998; Zalis et al., 1998; Labbe et al., 2001; Vieira et al., 2001; Contreras et al., 2002; Cortese et al., 2002; Lopes et al., 2002; Vinayak et al., 2003; Casey et al., 2004; Huaman et al., 2004; Vathsala et al., 2004), a situation attributed to relatively higher levels of symptomatic infections and consequently increased levels of drug usage (Warhurst and Duraisingh, 2001; Hastings, 2003). In such circumstances, cessation of CQ use would not automatically lead to an expanded population of parasites carrying the wild-type alleles as the possibility of mixed infections, and thus competition between resistant and sensitive parasites, would be minimal.

Nevertheless, our observations are consistent with recent findings in parts of China (Liu et al., 1995), Vietnam (Nguyen et al., 2001; Thanh et al., 2001; Nguyen et al., 2003) and Malawi (Kublin et al., 2003; Mita et al., 2003; Mita et al., 2004), where concerted efforts to halt CQ use have resulted in an increase in the in vitro and in vivo sensitivity of field isolates to the drug, and a concomitant decrease in the prevalence of certain mutations in pfmdr1 and pfcrt associated with resistance. It was shown in Malawi (Kublin et al., 2003) that the frequency of the K→T mutation at codon 76 of pfcrt found in most [but not all (Ariey et al., 2002; Rason et al., 2002; Thomas et al., 2002; Vinayak et al., 2003)] strains resistant to CQ in vitro declined sixfold from 1992 to 2000. The frequency of the D→Y mutation at codon 1246 of pfmdr1 also decreased significantly, although only halving in the same period. The decrease in the frequency of K76T in this region was shown to be attributed to the expansion of the wild-type population rather than a reversion event at this locus, a finding corroborated by another study in this area (Mita et al., 2004). These results suggest that there may be a greater selective pressure on the resistant form of pfcrt in this area than there is on the resistant form of pfmdr1, so that in the absence of drug the fitness deficit incurred by mutations in pfcrt is greater than the fitness deficit incurred by mutations in pfmdr1.

A similar inference can be drawn from a recent study in eastern Sudan (Abdel-Muhsin et al., 2004), where CQ is still used as the first-line drug. Between 1990 and 2001 the frequencies of mutations in pfmdr1 (N→Y at codon 86) and pfcrt (K76T) strongly associated with an increase in in vitro CQ resistance (Babiker et al., 2001) increased steadily and significantly by 2–3% annually. However, malaria in this area is a highly seasonal disease, and antimalarial use by the inhabitants is generally limited to the epidemic period in the few weeks following the annual rains. In the two successive years where interseasonal variation in allelic frequencies was monitored, it was found that there was a significant decrease in the frequency of K76T in the period following cessation of antimalarial use in the non-epidemic period. This relatively rapid change in allele frequency in vivo suggests a considerable fitness deficit may be associated with the mutant form of pfcrt in this region. There was no significant decrease in the frequency of the N86Y pfmdr1 mutation over the same seasonal intervals, supporting the hypothesis that mutations in pfcrt associated with CQ resistance may be more strongly affected by the presence or absence of drug pressure in vivo than the N86Y mutation in pfmdr1[a polymorphism not investigated in this study, but one that several studies have shown to be linked to CQ resistance in Africa (Sutherland et al., 2002; Abdel-Muhsin et al., 2004)].

The fitness differences shown here to result from mutations in pfmdr1 in the absence of CQ selection may be exacerbated in the presence of mefloquine (MQ). There is predominantly an inverse correlation between CQ and MQ resistance (Cowman et al., 1994; Peel et al., 1994; Duraisingh et al., 2000b), such that D10 is MQ-resistant, and 7G8 MQ-sensitive. However, Reed et al. (2000) showed that 7G8-mdrD10 is nearly as MQ-resistant as D10-mdrD10, with additional transfectants indicating that polymorphisms in pfmdr1 codons 1034, 1042 and 1246 are capable of determining levels of MQ sensitivity. Therefore in areas where the treatment for malaria includes MQ but not CQ, parasites with mutations at pfmdr1 codons 1034, 1042 and 1246 would be doubly disadvantaged with respect to parasites with the wild-type pfmdr1 allele, in terms of both decreased fitness and heightened drug sensitivity. This is borne out by recent findings in Thailand, where use of CQ as the first-line treatment for falciparum malaria was discontinued in 1972 in favour of Fansidar (a combination of the antifolate drugs sulfadoxine and pyrimethamine), which in turn was replaced by Fansimef (MQ with sulfadoxine/pyrimethamine) in 1985 (Thaithong et al., 1988; Labbe et al., 2001). Several studies have shown that more than 50% of Thai isolates are resistant to MQ in vitro, and that the prevalence of the D10-like N1042 and D1246 pfmdr1 polymorphisms is greater than 90% (Price et al., 1999; Lopes et al., 2002; Price et al., 2004), consistent with both a fitness advantage in the absence of CQ and previous observations associating MQ resistance with these polymorphisms (Duraisingh et al., 2000b; Reed et al., 2000). Despite the prevalence of the wild-type (CQS) pfmdr1 allele, in vivo CQ resistance in Thailand remains high, probably as a consequence of the CQR-conferring K76T mutation in PfCRT having reached near-fixation throughout the country (Price et al., 1999; 2004; Labbe et al., 2001; Lopes et al., 2002).

In conclusion, our observations of an in vitro fitness cost associated with mutations in pfmdr1 suggest that it may be possible to reintroduce CQ to countries where its use has been suspended and the reservoir of CQS strains has subsequently expanded. However, the relative ease with which CQ selection retrieved 7G8-mdr7G8 parasites from mixed cultures where they had been massively outcompeted (Figs 1A and B, 2A and B) highlights the potential dangers of reselection of residual CQR parasites by CQ monotherapy. This could be avoided by deploying CQ to appropriate areas as part of a combination regimen, in accordance with current WHO recommendations for improving the therapeutic efficacy of antimalarial treatments (WHO, 2001).

Experimental procedures

pfmdr1 transfected strains

Plasmodium falciparum strains in which the pfmdr1 locus had been manipulated by transfection (Reed et al., 2000) were obtained from Professor Alan Cowman (Walter and Eliza Hall Institute, Melbourne). Three of these strains were examined in this study: D10-mdrD10, a transfection control retaining the wild-type D10 pfmdr1 sequence associated with chloroquine sensitivity (key codons: N86, Y184, S1034, N1042, D1246); 7G8-mdr7G8, a transfection control retaining the 7G8 pfmdr1 allele associated with chloroquine resistance (N86, 184F, 1034C, 1042D, 1246Y, mutated residues are underlined) and 7G8-mdrD10, parasites with a 7G8 background but in which three of the four 7G8 pfmdr1 mutations had been replaced with the wild-type codons from D10 (N86, 184F, S1034, N1042, D1246). These strains were thawed from cryopreserved stocks 5 weeks prior to commencement of the experiments, reducing the possibility that differences between them were attributed to spontaneous mutations accumulated during continuous culture. Although these strains retain selectable markers as a consequence of the transfection process (Reed et al., 2000), they were not exposed to selection during this 5 week period, as we have found the integration of the transfection plasmids to be very stable, and exposure to selection drugs to be unnecessary for retention of the transfection plasmids (and the corresponding drug-resistance phenotype) over periods of many months. PCR tests confirmed the presence of the selectable markers and appropriate pfmdr1 mutations in the freshly thawed stocks (data not shown).

Culture conditions

Plasmodium falciparum parasites were cultured under standard conditions as described previously (Allen and Kirk, 2004). The cultures were tightly synchronized by a double exposure to sorbitol (Lambros and Vanderberg, 1979) during the juvenile (‘ring’) stage of their life cycle, ∼10 h apart, to eliminate mature trophozoites from both sides of the window of sorbitol insensitivity. This was repeated for three successive generations prior to commencement of the growth competition experiments.

Competition experiments

Thin-film slides of highly synchronized cultures of D10-mdrD10, 7G8-mdr7G8 and 7G8-mdrD10 were prepared and treated with Giemsa's stain in order to determine parasite density. These cultures were then set to a haematocrit of 1.8% and parasitaemia of 8%, and used to prepare mixed cultures containing equal proportions of (i) D10-mdrD10 and 7G8-mdr7G8, (ii) D10-mdrD10 and 7G8-mdrD10, and (iii) 7G8-mdr7G8 and 7G8-mdrD10. These mixtures were cultured continuously until measurement of in vitro CQ susceptibility showed that one strain was clearly dominant. To ensure that access to media nutrients and uninfected erythrocytes was not limiting, 75–90% of each culture was removed 8–12 h after the start of each new cycle, and fresh erythrocytes added to a final culture haematocrit of 2%. In this manner the parasitaemia was routinely reset to ∼3–5%. The ring-stage parasites removed from the flasks in each cycle were used as required to determine the CQ sensitivity of the mixture, and to provide genomic DNA for real-time PCR analysis of the relative proportions of each strain at that time point. Where it was determined that the CQR 7G8-mdr7G8 had been outcompeted in a mixture, 75 nM chloroquine was added to the medium on each media change, and once the culture had attained a sufficient parasitaemia, the CQ sensitivity of the mixture, and relative proportions of each strain, were again measured. No other selection agents were used during the competition experiments. At the end of the competition experiments we used PCR to confirm that the recombinant loci of the strains in each mixture were intact (data not shown).

In vitro testing of CQ sensitivity

Every 6–8 days a proportion of the ring-stage parasites not required to propagate the culture was used to determine the CQ sensitivity of the mixture by a modification of the standard microdilution technique described previously (Desjardins et al., 1979). Briefly, thin-film slides of each mixture – as well as the parental D10-mdrD10, 7G8-mdr7G8 and 7G8-mdrD10 strains – were prepared, and the parasite density of each determined. The parasitaemia and haematocrit of each was then adjusted to 2% and 1%, respectively, in RPMI media as described in Culture conditions but with a final hypoxanthine concentration of 1 µM. These cultures were incubated in 96-well plates containing 18 different CQ concentrations, from 3 to 400 nM, for 24 h at 37°C. [3H]-Hypoxanthine (Amersham) was then added to each well to a final concentration of 2 µCi ml−1, and the plates incubated at 37°C for a further 18–20 h. The plates were frozen and thawed, and parasite DNA and RNA was harvested onto glass fibre filter paper (Packard) using a Filtermate 196 Harvester (Packard). Incorporation of [3H]-hypoxanthine was measured using a Topcount Scintillation Counter (Packard). Sigmoidal curves (described by a four-parameter equation) were fitted to the data by least-squares regression with SigmaPlot 2001 (SPSS), and the dose–response curves for each mixture were compared with those of its parent strains to determine how the competition was proceeding.

Genomic DNA extraction

Every 4–6 days, a proportion of the ring-stage parasites not required to propagate the culture were treated with saponin essentially as described previously (except that centrifugation was performed at 12 000 g) (Saliba and Kirk, 1999) in order to remove erythrocyte haemoglobin. Genomic DNA was then extracted from the isolated parasites using a DNeasy Plant Mini Kit (Qiagen), and stored at −20°C. The concentration of each DNA extract was established using a Cary spectrophotometer, and diluted to 4 ng µl−1 before use in real-time PCR assays. Genomic DNA harvested before and after the experiments was found to be negative for mycoplasma contamination using a PCR detection method (Uphoff and Drexler, 2002).

Allele-specific real-time PCR quantitation of strains in mixtures

Allele-specific real-time PCR was performed to measure the relative proportions of each strain in the mixed cultures. Primers were designed using Primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). Two alternate forward primers were designed to detect specific single-nucleotide polymorphisms in the wild-type and mutant pfmdr1 alleles. PfD10 (5′-GCA GCT TTA TGG GGA TTC A-3′) was designed to anneal preferentially to the wild-type form of pfmdr1, its 3′ end matching the start of the AGT codon encoding the amino acid serine at position 1034. Pf7G8 (5′-TGC AGC TTT ATG GGG ATT CT-3′) was similarly designed to detect the mutant pfmdr1 form, its 3′ end preferentially annealing to the start of the TGT cysteine codon at the same position. The antisense primer for both reactions was Pfanti (5′-TCC ACC ATC ATC TCT TAC ATC A-3′). In each real-time PCR the primers were present at a final concentration of 300 nM. Every genomic DNA sample was tested with both primer pairs in order to determine the concentrations of each strain in the sample.

Real-time PCR amplification was carried out in a Rotor-Gene 2000 Real-Time Cycler (Corbett Research). To detect the D10 form of pfmdr1, reaction tubes contained 10 µl QuantiTect 2× SYBR Green PCR Master Mix (Qiagen), PfD10, Pfanti, 5 µl of 4 ng µl−1 genomic DNA, and sterile water to a final volume of 20 µl. The PCR programme was as follows: 15 min at 95°C to activate the HotStarTaq in the SYBR Green Master Mix; 40 cycles of denaturation (94°C for 15 s), annealing (57°C for 40 s) and extension (72°C for 40 s); followed by a melting curve analysis from 60°C to 95°C at 1°C s−1. To detect the 7G8 form of pfmdr1, the same protocol was followed, except that the primer pair used was Pf7G8 and Pfanti. The concentration of strain-specific DNA in each mixture was determined by comparing the threshold cycle with that of control reactions containing known proportions of each strain, also at a final DNA concentration of 1 ng µl−1. All reactions were performed in triplicate.

Quantifying the fitness loss

The relative fitness (w) of the less fit strain in each competition was derived from the equation log(pt/qt) = log(p0/q0) + t·log(w), where p is the proportion of the fitter strain, q the proportion of the less fit strain, and t the number of asexual generations that had occurred over the course of the experiment (Hartl and Clark, 1997). A least-squares regression plot of log(pt/qt) against t produces a linear relationship in the form y = mx + b, where m = log(w) and b = log(p0/q0).

Microscopic examination of strain reproductive success

Thin-film slides of highly synchronized ring-stage D10-mdrD10, 7G8-mdr7G8 and 7G8-mdrD10 strains were prepared, and the parasite density of each was determined. The parasitaemia and haematocrit was, in each case, adjusted to 2% and 5%, respectively, and the cultures were allowed to mature. Further slides were prepared 24 h later, at the trophozoite stage, and then again every hour once schizogony had commenced (∼44–48 h post invasion). Once the cultures consisted uniformly of ring-stage parasites again (∼6–12 h post invasion), the parasitaemia and haematocrit of each was again readjusted to 2% and 5%, and the entire process repeated twice more. In this manner slides representing three successive generations of each strain were collected for examination. Parasitaemias were established by counting the number of rings per 1000 erythrocytes, and confirmed by establishing the number of trophozoites per 5000 erythrocytes in each culture following parasitaemia and haematocrit adjustment. The reproductive success of each strain was then quantified in terms of the number of ring-stage parasites produced per trophozoite in the preceding generation. The slides collected at schizogony were used to calculate the number of merozoites produced per schizont. Only intact schizonts in which distinct, mature merozoites could be distinguished unambiguously were counted. At least 30 such schizonts were examined per strain per generation.

Acknowledgements

This work was supported by Australian National Health and Medical Research Council Grant 179804. We thank Professor Alan Cowman (WEHI) for providing the transfected parasite strains examined in this study, and the ACT Red Cross Blood Transfusion Service for the provision of blood.

Ancillary