The interplay between drug resistance and fitness in malaria parasites
For correspondence. E-mail firstname.lastname@example.org; Tel. (+1) 415 206 8845; Fax (+1) 415 648 8425.
Controlling the spread of antimalarial drug resistance, especially resistance of Plasmodium falciparum to artemisinin-based combination therapies, is a high priority. Available data indicate that, as with other microorganisms, the spread of drug-resistant malaria parasites is limited by fitness costs that frequently accompany resistance. Resistance-mediating polymorphisms in malaria parasites have been identified in putative drug transporters and in target enzymes. The impacts of these polymorphisms on parasite fitness have been characterized in vitro and in animal models. Additional insights have come from analyses of samples from clinical studies, both evaluating parasites under different selective pressures and determining the clinical consequences of infection with different parasites. With some exceptions, resistance-mediating polymorphisms lead to malaria parasites that, compared with wild type, grow less well in culture and in animals, and are replaced by wild type when drug pressure diminishes in the clinical setting. In some cases, the fitness costs of resistance may be offset by compensatory mutations that increase virulence or changes that enhance malaria transmission. However, not enough is known about effects of resistance mediators on parasite fitness. A better appreciation of the costs of fitness-mediating mutations will facilitate the development of optimal guidelines for the treatment and prevention of malaria.
Malaria is one of the most important infectious diseases in the world. Recently, important gains in the control of malaria have been reported in some areas, and there is increasing optimism regarding the potential for elimination of malaria from many regions (Feachem et al., 2010; Tatem et al., 2010). However, despite recent gains, malaria remains an overwhelming problem in much of the tropical world, and it continues to cause hundreds of millions of illnesses and up to about one million deaths each year (Snow et al., 2005; Murray et al., 2012). Most serious illnesses and deaths from malaria and also most drug-resistant infections are due to infection with Plasmodium falciparum, the most pathogenic human malaria parasite, and this review will focus principally on studies with that organism.
The control and eventual eradication of malaria depend on a rather small set of tools. For control of anopheline mosquito vectors insecticide impregnated bednets and indoor residual spraying of insecticides are increasingly used, and their utility has been clearly demonstrated (Okumu and Moore, 2011), but their efficacy will be limited without coincident efforts directed against malaria parasites. An effective vaccine against malaria would be extremely valuable. Unfortunately, although the RTS,S vaccine, which has offered modest protection against malaria in African children (Olotu et al., 2013), may be available in a few years, no highly effective vaccine is on the horizon. Considering the limitations of vector control and vaccines, appropriate use of antimalarial drugs remains a cornerstone of malaria control.
Drugs have two key roles for malaria control. First, in addition to offering obvious benefit to an ill individual, prompt and effective treatment for malaria limits the development of parasites into gametocytes, thus blocking transmission to mosquitoes and subsequently to other individuals (Gosling et al., 2011). Available drugs have varied activity against gametocytes that are present at the time of treatment, however. Second, there is increasing consideration of the use of drugs to prevent malaria in endemic populations, either as intermittent therapy or as low-dose chemoprophylaxis (Greenwood, 2010). For all indications, we are dependent on a rather small armamentarium of antimalarial drugs. The efficacies of many of these drugs are limited by resistance, and recent evidence suggests that parasites are becoming resistant to our newest agents. However, the extent of resistance varies, such that in many cases drugs with resistance concerns are nonetheless offering good effectiveness. In the setting of widespread but varied levels of drug resistance, the impact of resistance on parasite fitness is of great importance. Choices of optimal regimens for treatment and chemoprevention will be facilitated by an understanding of the impacts of resistance selection on the abilities of parasites to cause disease and to be transmitted.
The interplay between drug resistance and fitness in bacteria and viruses
With many microbial pathogens, drug resistance comes with a fitness cost. Considering bacterial infections, antibiotic use selects for resistant bacteria via multiple mechanisms, including alterations in antibiotic target genes and increases in drug efflux, but resistant bacteria are typically less fit (Andersson and Hughes, 2010). Fitness can be measured in vitro by comparing growth rates and in competitive growth experiments. Importantly, fitness can be mediated by impacts of a trait on other organisms (e.g. production of a factor toxic to sensitive bacteria), such that a fitness advantage will only be recognized in a competitive growth experiment (Gordo et al., 2012). Fitness can also be assessed in animal model studies and by considering the clinical consequences of infection with drug-sensitive and drug-resistant organisms. Over time, resistant bacteria can evolve into organisms with improved fitness due to the acquisition of compensatory mutations. Non-lethal selective pressure from low levels of antibiotics may enhance the likelihood of resistance selection (Andersson and Hughes, 2010). Removal of antibiotic pressure can allow reversion to drug-sensitive organisms, but resistant bacteria can be quite stable, in part due to compensatory mutations that improve fitness.
Studies with viruses have also demonstrated ready selection of resistance (Götte, 2012). In general mutant variants with a low genetic barrier (relatively few genetic changes required) are selected most rapidly by antiviral selective pressure. Subsequently, more fit variants are selected more slowly. As with bacteria, measures of viral fitness can include relative rates of replication of different strains in cell lines, comparing replication either in parallel assays or with competition assays (Wargo and Kurath, 2012).
Overall, characterization of the impacts of antimicrobial drug resistance on bacterial and viral fitness has been complex. Measures of fitness may vary depending on experimental methodology. Resistance to one agent may have important impacts on resistance to other drugs, and thereby impact upon fitness. Fitness also must be considered in the context of both the replication and transmissibility of microorganisms, and these two features are often not linked. It is important to take these factors into account when considering assessments of the interplay of drug resistance and fitness in malaria parasites. However, as eukaryotes with complex asexual and sexual life cycles, malaria parasites differ importantly from the prokaryotic and viral model systems.
Antimalarial drugs act principally to eliminate the erythrocytic stages of malaria parasites that are responsible for human illness. The standard treatment for falciparum malaria has changed in recent years. With frequent resistance to older drugs, artemisinin-based combination therapy (ACT) is recommended for the treatment of uncomplicated falciparum malaria in nearly all areas (World Health Organization, 2010). ACT consists of a potent artemisinin component, which rapidly clears most parasites, plus a longer acting partner drug, which eliminates remaining parasites after the artemisinin is cleared (Nosten and White, 2007). The most important ACTs that are now available are artemether/lumefantrine, artesunate/amodiaquine, artesunate/mefloquine, and dihydroartemisinin/piperaquine. ACTs are also effective against non-falciparum malaria. Multiple drugs are used to prevent malaria. Recommendations for travellers from non-endemic to endemic areas generally advocate use of atovaquone/proguanil, mefloquine or doxycycline in low-dose chemoprophylactic regimens (Schlagenhauf and Petersen, 2008). In Africa intermittent preventive therapy is advocated in some high risk populations, including sulphadoxine/pyrimethamine during pregnancy and amodiaquine/sulphadoxine/pyrimethamine as seasonal malaria chemoprophylaxis in areas with relatively little drug resistance (Greenwood, 2010).
Available antimalarial drugs can be divided into seven classes (Table 1). The 4-aminoquinoline chloroquine was the gold standard for the treatment of uncomplicated malaria for many years, but it is no longer appropriate for the treatment of falciparum malaria in nearly all areas due to drug resistance. Amodiaquine is subject to similar resistance mechanisms, but it often provides adequate efficacy against parasites with the genetic changes that mediate chloroquine resistance. The main current use of amodiaquine is as a component of the ACT artesunate/amodiaquine. A third 4-aminoquinoline, piperaquine, was widely used to treat and prevent malaria in China a few decades ago, but it then fell into disfavour due to increasing drug resistance (Davis et al., 2005). More recently piperaquine has become a component of another ACT, dihydroartemisinin/piperaquine. The 8-aminoquinoline primaquine has some activity against erythrocytic parasites, but it is used principally to eliminate parasite liver stages, including the exoerythrocytic forms that precede erythrocytic infection in all species and the hypnozoites that cause latent infections with Plasmodium vivax and P. ovale. Primaquine also acts against gametocytes, thereby lowering transmission of parasites to mosquito vectors. Quinine is an aryl-amino alcohol that is our oldest antimalarial drug, used as cinchona bark since the 1600s and in its pure form since 1820 (Meshnick and Dobson, 2001). Quinine is quite hard to tolerate, and its use is best limited to the treatment of severe malaria. Important related drugs are mefloquine and lumefantrine, both of which are components of ACTs.
Table 1. Antimalarial drugs: mechanisms of action and resistance
|4-Aminoquinoline||Chloroquine||Inhibition haemozoin formation → toxicity from free haem||Tx and CPx, but limited by R||Yes||High level R → tx failures||SNPs in pfcrt (76T primary mediator) and pfmdr1|
|Amodiaquine||Tx falciparum malaria in combination with AS||Yes||High level R → tx failures; Efficacy good combined with AS in Africa||SNPs in pfcrt and pfmdr1|
|Piperaquine||Tx falciparum malaria in combination with DHA||Yes||R described in China in 1980s, but not recently||Unknown|
|8-Aminoquinoline||Primaquine||Unknown||Elimination liver hypnozoites of P. vivax and P. ovale||Yes||Failure of radical cure of P. vivax and P. ovale||Unknown|
|Aryl-amino alcohol||Quinine||Unknown||Tx severe falciparum malaria||Yes||Mostly low-level R in SE Asia||Unknown|
|Mefloquine||Unknown||Tx falciparum malaria in combination with AS; CPx||Yes||Failures of MQ and AS/MQ seen in SE Asia||↓ S with pfmdr1 WT sequences and ↑ copy number|
|Lumefantrine||Unknown||Tx falciparum malaria in combination with AM||No||Some variation in activity, but no definitive clinical R||↓ S with pfmdr1 WT sequences and ↑ copy number|
|Antifolates||Pyrimethamine||Inhibition DHFR||Combined with sulphadoxine for Tx and IPT||Yes||R in many areas||Multiple SNPs in pfdhfr mediate step-wise ↑ in R|
|Trimethoprim||Daily with sulphamethoxazole protects against malaria||Yes|
|Proguanil||Combined with ATV (Malarone) for Tx and CPx||Yes||R in many areas; Malarone active even with proguanil R|
|Sulphonamides||Inhibition DHPS||Used in combination with DHFR inhibitors||Yes||R in many areas; SP retains good efficacy in W. Africa||Multiple SNPs in pfdhps mediate step-wise ↑ R|
|Naphthoquinone||Atovaquone||Inhibition cytochrome bc1 complex||CPx and Tx, in both cases in combination with proguanil||Yes||R selected rapidly with monotherapy||SNPs in pfcytb|
|Antibiotic||Doxycycline, clindamycin||Inhibition apicoplast protein synthesis||CPx; Tx in combination with quinine||No||Some variation in activity, but no definitive clinical R||Polymorphisms in homologues of bacterial R mediators associated with ↓ S|
|Artemisinin||Artesunate, artemether, dihydroartemisinin||Uncertain||Cornerstone of ACT regimens for Tx falciparum malaria; intravenous AS gold standard for Tx severe malaria||Yes||Delayed parasite clearance in clinical trials||↓ S with pfmdr1 WT sequences and ↑ copy number; Mechanism delayed clearance unknown|
Antifolates, which were developed to treat bacterial infections, target parasite dihydrofolate reductase (DHFR) and dihydropteroate synthase (DHPS), offering synergistic antimalarial activity. Sulphadoxine/pyrimethamine has the distinct advantage of single-dose therapy, but its treatment efficacy is seriously limited by drug resistance. Trimethoprim/sulphamethoxazole is not routinely used to treat malaria, but it offers fairly effective protection against malaria when provided as regular prophylaxis against multiple infections in those with HIV infection. The naphthoquinone atovaquone acts against the mitochondrial cytochrome bc1 complex. Combined with the DHFR inhibitor proguanil it offers effective therapy and chemoprophylaxis for falciparum malaria. A number of antibiotics that are prokaryotic protein synthesis inhibitors have antimalarial activity due to action against the protein synthesis machinery of the apicoplast organelle (Dahl and Rosenthal, 2008). Doxycycline is used for chemoprophylaxis against malaria, and doxycycline or clindamycin are combined with quinine to treat falciparum malaria.
The most important new class of antimalarials is the artemisinins, which were developed from a natural product remedy in China. Artemisinin is a potent antimalarial, but the derivatives artesunate, artemether and dihydroartemisinin are most widely used, all as components of ACT regimens. Indeed, the use of artemisinins outside of combination regimens is strongly discouraged by the World Health Organization due to fear of selecting for resistance to this important class of drugs. Artemisinins are highly effective against acute malaria, but short acting, so combination with longer-acting drugs in ACTs allows short (3-day) courses of treatment that protect against the selection of resistance to the artemisinin component (Nosten and White, 2007). Due to its rapid action intravenous artesunate is also the new gold standard for the treatment of severe falciparum malaria, with documented survival advantages compared with intravenous quinine (Dondorp et al., 2005; 2010).
Antimalarial drug resistance
Resistance has been described for nearly all available drugs (Table 1). Established mediators of resistance are single-nucleotide polymorphisms (SNPs) and changes in copy number in genes encoding putative drug transporters and some enzyme targets. For many drugs the extent of resistance is uncertain and mechanisms of resistance are unknown. Resistance can be assessed by in vitro assessment of sensitivities of cultured P. falciparum, by evaluation of genetic polymorphisms associated with resistance, by consideration of the clinical consequences of polymorphisms present at the time of treatment, or by assessing the selective pressure of antimalarial treatment on subsequent infections. Studies considering all of these factors have shed light on the extent of resistance and on mechanisms of resistance.
Informatic studies have identified multiple predicted transporter genes in P. falciparum (Table 1); SNPs in 11 of these genes were associated with decreased sensitivity to chloroquine or quinine (Mu et al., 2003), although a subsequent study could not confirm associations between most of the identified polymorphisms and drug sensitivity in clinical isolates (Anderson et al., 2005). Many predicted P. falciparum transporters are members of the ATP-binding cassette (ABC) transporter superfamily. ABC transporters are responsible for the transfer of a range of substances across concentration gradients in an energy-dependent manner (Koenderink et al., 2010). Polymorphisms in transport proteins can mediate resistance to many agents active against cancer and infectious diseases via enhancing efflux of the drugs from cells (Borges-Walmsley et al., 2003). It appears that a number of plasmodial proteins transport different drugs and that polymorphisms in these proteins may impact upon drug sensitivity (Picot et al., 2009).
Polymorphisms in the P. falciparum multidrug resistance-1 (pfmdr1) gene, which encodes the P-glycoprotein homologue, impact on sensitivity to multiple antimalarial drugs (Foote et al., 1990; Valderramos and Fidock, 2006; Sanchez et al., 2010). In humans, P-glycoprotein polymorphisms are associated with resistance to cancer drugs (Sharom, 2011). In P. falciparum, the function of the pfmdr1 product is unknown, but the protein localizes to the membrane of the food vacuole, the site of action of a number of drugs, suggesting that it is a drug transporter (Cowman et al., 1991). Data on associations between pfmdr1 polymorphisms and drug sensitivity are complex, but overall suggest that changes in pfmdr1 sequence or copy number alter transport of multiple drugs in or out of the parasite food vacuole, with individual polymorphisms leading to opposite effects on different drugs (Koenderink et al., 2010). Mutations at pfmdr1 N86Y and D1246Y (for this and other P. falciparum genes wild-type sequence is based on the 3D7 reference strain), which are common in Africa, have been linked to decreased sensitivity to chloroquine and amodiaquine, but increased sensitivity to lumefantine, mefloquine and artemisinins (Reed et al., 2000; Duraisingh et al., 2000a,b; Mwai et al., 2009). Other polymorphisms primarily seen outside Africa (including 1034C, 1042D, and increased gene copy number) are associated with altered sensitivity to lumefantrine, mefloquine and artemisinins (Reed et al., 2000; Pickard et al., 2003; Sidhu et al., 2005; 2006; Veiga et al., 2011). Considering infections that emerge soon after prior therapy, amodiaquine-containing regimens selected for the 86Y and 1246Y mutant alleles (Humphreys et al., 2007; Nsobya et al., 2007; Zongo et al., 2007) and for parasites with decreased in vitro sensitivity to the active metabolite monodesethylamodiaquine (Nawaz et al., 2009) in subsequent infections. In contrast, therapy with artemether-lumefantrine selected for the N86 and D1246 wild-type alleles in subsequent infections within 60 days of prior therapy (Sisowath et al., 2005; Humphreys et al., 2007; Zongo et al., 2007; Happi et al., 2009; Some et al., 2010; Baliraine and Rosenthal, 2011). Importantly, impacts of pfmdr1 polymorphisms on drug sensitivity are modest, correlations between particular polymorphisms and treatment efficacy have not been seen, and the ACTs artesunate-amodiaquine and artemether-lumefantrine remain highly efficacious for the treatment of uncomplicated falciparum malaria in Africa (Dorsey et al., 2007; Four Artemisinin-Based Combinations Study Group, 2011). However, as seen for chloroquine and amodiaquine, pfmdr1 polymorphisms may contribute, with additional polymorphisms, to higher level resistance to increasingly used components of ACTs.
Soon after the identification of pfmdr1 it became clear that polymorphisms in this gene are not the primary mediators of chloroquine resistance. Subsequently, analysis of progeny of a genetic cross between chloroquine-sensitive and -resistant strains led to the identification of pfcrt (Fidock et al., 2000), which encodes a food vacuole membrane protein that is predicted to be a member of the drug/metabolite transporter superfamily (Martin and Kirk, 2004; Tran and Saier, 2004). The function of pfcrt is unknown, but apparently essential, as disruption of the gene has not been possible (Ecker et al., 2012). Pfcrt is highly polymorphic, but one SNP, K76T, is the primary mediator of chloroquine resistance (Lakshmanan et al., 2005; Ecker et al., 2012). The 76T mutation appears to act principally by increasing the export of chloroquine from the food vacuole, but the mechanism of pfcrt 76T-mediated chloroquine resistance is incompletely understood (Ecker et al., 2012). Other pfcrt SNPs always accompany 76T in field isolates, and these likely encode compensatory mutations that allow parasites containing 76T to maintain adequate fitness; some other SNPs may also contribute directly to the drug-resistant phenotype. The 76T mutation also mediates decreased sensitivity to monodesethylamodiaquine, and studies with genetically modified parasites have shown it to mediate increased susceptibility to mefloquine and artemisinins (Sidhu et al., 2002; Lakshmanan et al., 2005), suggesting the same reciprocal relationship between sensitivities to aminoquinolines and other drugs as described for certain pfmdr1 polymorphisms.
Plasmodium falciparum multidrug resistance protein-1 (Pfmrp1) is a member of the ABC transporter superfamily (Koenderink et al., 2010). Unlike the products of pfmdr1 and pfcrt, Pfmrp1 localizes principally to the parasite plasma membrane (Kavishe et al., 2009). In studies of culture adapted P. falciparum, SNPs in pfmrp1 were linked to decreased sensitivity to chloroquine and quinine (Mu et al., 2003). Two SNPs that appear to be common in African parasites, I876V and K1466R, were selected by prior treatment with artemether/lumefantrine (Dahlstrom et al., 2009a) and sulphadoxine pyrimethamine (Dahlstrom et al., 2009b) respectively. Disruption of the pfmrp1 gene yielded parasites with diminished growth and increased sensitivity to chloroquine and other drugs, suggesting a role for this protein in the efflux of antimalarial drugs from the parasite and in parasite fitness (Raj et al., 2009).
Quantitative trait locus analysis identified three genes predicted to play roles in the responsiveness of P. falciparum to quinine, pfcrt, pfmdr1 and pfnhe1, which encodes a putative sodium-hydrogen exchanger and is highly polymorphic (Ferdig et al., 2004). Reducing the expression of pfnhe1 by ∼ 50% using allelic exchange led to a 30% increase in quinine sensitivity in some but not other parasite strains (Nkrumah et al., 2009). Recent studies evaluating associations between polymorphisms in a pfnhe1 microsatellite, in vitro parasite sensitivity, and clinical responses to various drugs have been inconsistent, but these polymorphisms appear to have a modest impact on sensitivity of parasites to quinine, and possibly other drugs (Henry et al., 2009; Andriantsoanirina et al., 2010; Meng et al., 2010; Okombo et al., 2010; Baliraine et al., 2011; Sinou et al., 2011). Pfatp4 encodes a P. falciparum plasma membrane protein that appears to be a sodium efflux pump (Spillman et al., 2013). Mutations in pfatp4 have been linked to altered sensitivity to a number of candidate antimalarials, with good evidence that the transporter is the target of highly active spiroindolones (Rottmann et al., 2010).
Resistance to antifolates and atovaquone
The best-characterized mediators of drug resistance in P. falciparum are mutations in the pfdhfr and pfdhps genes, which encode sequential enzymes in the folate pathway common to a wide range of eukaryotic and prokaryotic organisms. A series of mutations in pfdhfr and pfdhps mediate increasing resistance to antifolate combinations (Gregson and Plowe, 2005). In Africa, pfdhfr S108N, N51I and C59R and pfdhps A437G are now very common and mediate low-level resistance to sulphadoxine-pyrimethamine. A fifth mutation, pfdhps K540E, is common in eastern and southern Africa, and mediates a higher level of resistance (Pearce et al., 2009). Additional mutations seen most commonly outside Africa, include pfdhfr I164L, pfdhps A581G and pfdhps A613S, appear to mediate high level of resistance. Atovaquone leads to collapse of mitochondrial membrane potential via inhibition of the cytochrome bc1 complex. Resistance to atovaquone develops rapidly, and is mediated by a number of mutations in the cytochrome b (pfcytb) gene (Vaidya and Mather, 2000; Musset et al., 2007).
Resistance to antibiotics
Resistance of P. falciparum to antibiotics has not been well studied clinically, but varied sensitivities to tetracyclines of unknown clinical significance have been seen in vitro (Briolant et al., 2009). Sequence polymorphisms and variation in copy number in P. falciparum homologues of bacterial mediators of tetracycline resistance (pfmdt and pftetQ) were associated with decreased drug sensitivity (Briolant et al., 2010). A survey of parasites from the Amazon identified a SNP in an apicoplast gene encoding a homologue of a ribosomal protein in which mutations mediate clindamycin resistance in bacteria; this SNP was associated with in vitro clindamycin resistance in three clinical isolates (Dharia et al., 2010). Resistance to azithromycin has also been linked to mutations in an apicolast-encoded ribosomal protein (Sidhu et al., 2007).
Resistance to artemisinins
A clinical study identified P. falciparum with decreased in vitro sensitivity to artemether from French Guiana and associations between the S769N SNP in the pfatp6 gene, which encodes a SERCA-type Ca++ ATPase, and decreased artemether sensitivity (Jambou et al., 2005). This finding was of great interest due to the vital importance of artemisinin antimalarials and a prior report suggesting that artemisinins exert antimalarial activity by inhibiting PfATP6 (Eckstein-Ludwig et al., 2003). However, while in vitro inhibition of PfATP6 by artemisinins has been demonstrated in Xenopus oocytes (Uhlemann et al., 2005; Pulcini et al., 2013) and pfatp6 has been shown to be highly polymorphic (Dahlstrom et al., 2008; Tanabe et al., 2011), with one SNP (L263E) ablating artemisinin sensitivity (Uhlemann et al., 2005), mutations in field isolates differed from those shown to alter artemisinin sensitivity, the S769N mutation was not seen in isolates from other areas, and introduction of the L263E (Valderramos et al., 2010) or S769N (Cui et al., 2012) mutations did not alter sensitivity of malaria parasites to artemisinins. Thus, the mechanism of resistance to artemisinins remains uncertain. Evaluations of resistance mechanisms should be viewed in light of recent observations of delayed clearance of parasites after treatment of malaria patients in South-east Asia with artemisinins (Noedl et al., 2008; Dondorp et al., 2009). Despite extensive effort, this reproducible clinical phenotype has not been clearly linked to an in vitro phenotype, with parasites that cleared slowly showing decreased drug sensitivity in some (Noedl et al., 2008; Lim et al., 2010), but not other (Dondorp et al., 2009) studies. Recent studies using novel assays have shown association between parasites with delayed clearance and in vitro survival after exposure to short pulses of high concentrations of artemisinins during the ring stage (Witkowski et al., 2013), consistent with observed marked sensitivity of early ring-stage parasites to short pulses of high concentrations of the drugs (Klonis et al., 2011; 2013). In addition, in vitro selection of parasites with decreased artemisinin sensitivity has typically led to no change in standard sensitivity assays, but increased recrudescence after treatment, suggesting that decreased sensitivity may be due to changes in recovery from drug-induced parasite dormancy (Witkowski et al., 2010; Tucker et al., 2012). As noted above, polymorphisms in pfmdr1 are associated with modest changes in artemisinin sensitivity, but polymorphisms in this gene or in pfatp6 have not been clearly linked to delayed clearance in clinical studies. Ongoing studies are working to identify genetic determinants of delayed clearance of parasites after artemisinin treatment, and these appear to be complex (Cheeseman et al., 2012; Miotto et al., 2013; Takala-Harrison et al., 2013). Changes in artemisinin sensitivity will likely impact upon parasite fitness, but information on such associations are not yet available.
Drug resistance and fitness in malaria parasites
Means of assessing the fitness of malaria parasites
We lack a specific measure of fitness in malaria parasites. Culture systems for P. falciparum allow careful comparison of growth rates, but distinguishing strains with only modest differences can be challenging. A method that is probably superior is the direct comparison of growth by co-culture of competing strains, with genetic determinants allowing one to distinguish the strains and identify overgrowth by the more fit strain. Comparisons of clinical isolates are arguably more relevant than those of engineered laboratory strains, but identifying meaningful differences between unrelated strains is challenging. Animal model studies have also provided insights, but these are limited by analysis of non-human parasites. The most convincing demonstrations of the fitness costs of antimalarial drug resistance have come from clinical studies. Most notably, loss of chloroquine selective pressure has been followed by rapid reemergence of chloroquine-sensitive parasites, as will be discussed below.
Fitness consequences of antimalarial drug resistance: in vitro studies
The first direct comparison of the fitness of drug-sensitive and -resistant malaria parasites used competition experiments to compare the growth of P. falciparum selected in vitro for resistance to atovaquone with that of the parent strain. Parasites selected for resistance had multiple mutations in pfcytb. Those with a single M133I mutation had 25-fold decreased atovaquone sensitivity and those with the M133I and G280D mutations had 1000-fold decreased sensitivity (Korsinczky et al., 2000). In competition growth experiments, the single mutant line grew as well as the parent parasite, but the more resistant double mutant was out-competed by the parent line, with a 5–9% loss of fitness calculated based on relative growth rates (Peters et al., 2002). Molecular modelling predicted that the G280D mutation altered the orientation of a putative ubiquinone-binding residue, likely altering enzyme function, while the M133I mutation would have minimal effect on enzyme conformation. Another pfcytb mutation, Y268S, was seen in a P. falciparum isolate from Thailand, and led to marked alterations in enzyme function (Fisher et al., 2012). Decreased enzyme activity was associated with increased expression of mitochondrial cytochrome bc1 complex genes, apparently compensating for detrimental effects of the mutation, and offering an example of parasites reaching a balance to maintain fitness in the setting of drug-resistance.
A similar approach was used to evaluate the relative fitness of P. falciparum engineered to contain different pfmdr1 haplotypes. Competitive growth was compared between the chloroquine-sensitive D10 strain, the chloroquine-resistant 7G8 strain, and a strain in which most of the pfmdr1 gene of D10 was replaced by that of 7G8, introducing the S1034C, N1042D and D1246Y mutations (Hayward et al., 2005). Introduction of the mutant pfmdr1 in D10 partially reversed chloroquine resistance and mediated decreased sensitivity to mefloquine, halofantrine and artemisinin (Reed et al., 2000). These results demonstrate the reciprocal impacts of the pfmdr1 polymorphisms on sensitivity to chloroquine (which is mediated principally by a different mutation, pfcrt K76T) and to mefloquine, halofantrine and artemisinin. Considering fitness, the chloroquine-resistant 7G8 strain was outcompeted by the chloroquine-sensitive D10 strain and by the intermediate sensitivity 7G8 strain containing D10 pfmdr1 sequence. Improved fitness of the more chloroquine-sensitive strains was evidenced by gradual increases in chloroquine sensitivity of mixed cultures until, after 34 days, sensitivities were identical to that of a pure D10 culture (Hayward et al., 2005). Comparison of growth rates led to an estimate of ∼ 25% loss of fitness mediated by introduction of the three pfmdr1 mutations. However, the relative importance of the individual SNPs in mediating resistance is unclear, and the N86Y mutation, which appears to be the principle mediator of altered drug sensitivity in Africa, was not studied.
Increases in the copy number of pfmdr1 have an important impact on sensitivity to numerous antimalarials, with increased copy number increasing sensitivity to chloroquine, but decreasing sensitivity to mefloquine, quinine and artemisinins (Pickard et al., 2003; Phompradit et al., 2011; Veiga et al., 2011). Increased pfmdr1 copy number is common in Asia, but not Africa. As is the case with resistance-mediating SNPs, copy number variation is likely to impact upon parasite fitness. This factor was studied by selecting in vitro for resistance to mefloquine in a Thai strain of P. falciparum (Preechapornkul et al., 2009). Decreasing susceptibility to mefloquine was associated with increasing copy number, but pfmdr1 SNPs were not selected. Mefloquine-resistant clones with average copy numbers of 2.3 (approximately fivefold less sensitive to mefloquine than the parental strain) and 3.1 (approximately sixfold less sensitive) were co-cultured with the parent strain (1 copy number), and the selected strains showed a fitness disadvantage, with overgrowth of single copy number strains over 3–4 weeks. Modelling predicted a loss of fitness, compared with the parental strain, of 6.3% and 8.7% for parasites with 2.3 and 3.1 copies of pfmdr1 respectively. Modelling further suggested that under drug pressure pfmdr1 amplification is a common event, with increases from one to two copies occurring once in every 108 parasites and from two to three copies once in every 5000 parasites. Thus, the selective pressure for increased pfmdr1 copy number is great, but gene amplification is apparently kept in check by the fitness disadvantage of increased copy number.
Considering impacts of pfmdr1 polymorphisms on the fitness of clinical isolates, P. falciparum was cultured from children with malaria in a region of Uganda with high multiplicity of infection (and thus high likelihood of mixed infections), and changes in the prevalence of the pfmdr1 N86Y and D1246Y alleles were followed over time (Ochong et al., 2013). Most cultures did not undergo changes in culture, but for those that did show selection 8/11 selected towards mutant 86Y, 9/14 selected towards wild-type D1246, and 5/7 with selection at both alleles selected towards 86Y and D1246. Surprisingly, the results suggest a mixed picture, with fitness advantages for parasites with pfmdr1 mutant 86Y and wild-type D1246 alleles.
The gch1 gene encodes GTP cyclohydrolase, which acts upstream of DHPS and DHFR in the folate synthesis pathway. This gene had increased copy number in Asian P. falciparum with folate resistance mutations, suggesting a compensatory mechanism to increase fitness in these drug-resistant parasites (Nair et al., 2008). Genetic manipulation of P. falciparum to increase gch1 copy number led to modest decreases in pyrimethamine sensitivity in most tested parasite lines (Heinberg et al., 2013). However, results were complex, with varied effects in different strains, and with evidence for detrimental effects of marked overexpression of GTP cyclohydrolase. These results offer another example of how selection of parasite alterations by drug pressure is balanced by effects on parasite fitness.
Fitness consequences of antimalarial drug resistance: animal model studies
Studies using murine malaria models have also attempted to assess the fitness costs of drug resistance. Growth of a pyrimethamine-resistant clone of the rodent parasite Plasmodium chabaudi was compared with that of its drug-sensitive parent strain by co-infection of mice. The sensitive strain appeared to outgrow the resistant strain in two of three experiments (Rosario et al., 1978). In the same studies, a chloroquine-resistant strain of P. chabaudi outgrew a chloroquine-sensitive strain in four of four experiments, in one case even when the inoculum was 90% sensitive, and only 10% resistant parasites. This experiment suggests a surprising fitness advantage for chloroquine-resistant parasites that is difficult to reconcile with other reports. In another study P. chabaudi selected for resistance to pyrimethamine were outcompeted by sensitive parasites, but after many passages the resistant strain gained a fitness advantage over both the parent strain and the original resistant strain (Walliker et al., 2005). Thus, determinants of fitness are complex, and compensatory mutations gained over time may mediate fitness improved even over that of the parental drug-sensitive strain.
In another strategy to study fitness, P. chabaudi was passaged in mice to increase virulence (Mackinnon et al., 2002). In this case the features described as virulence (increased parasite growth rate and mouse mortality) can also be seen to represent fitness. The parasites selected for virulence were consistently less susceptible than parent strains to treatment with pyrimethamine or artemisinin, and differences were not explained by differences in known resistance-mediating polymorphisms (Schneider et al., 2008; 2012). Thus, independent of specific resistance mediators, parasites selected for increased fitness were more resistant to drug treatment. In another study P. chabaudi were selected by passages in mice for resistance to artesunate-mefloquine. The parasites selected for resistance grew as well as parent drug-sensitive parasites, in this case showing no apparent fitness cost of drug resistance (Rodrigues et al., 2013).
To study impacts of resistance on fitness in mosquito stages of malaria parasites, mice infected with Plasmodium berghei were treated with pyrimethamine to select for resistance. Selected parasites were similar to sensitive parasites in the mouse, but they proceeded more slowly through mosquito development (Shinondo et al., 1994). Thus, drug-resistant strains might be less capable than sensitive parasites to transmit malaria. In contrast, when mosquitoes were fed on mice infected with a mixture of parasites differing in virulence as described above, the more virulent parasites were more readily transmitted (Schneider et al., 2012). In other studies, when mice were infected with P. berghei encoding wild-type or resistance-mediating pfcrt haplotypes, blood-stage parasites were equally sensitive to chloroquine, but mutant parasites had enhanced mosquito infectivity in the presence of chloroquine, suggesting that drug resistance enhanced transmission (Ecker et al., 2011).
Fitness consequences of antimalarial drug resistance: insights based on clinical trials
The most valuable insights available to date on the interplay between drug resistance and fitness in malaria parasites probably come from clinical observations. Of interest is the geography of selection of drug resistance. Interestingly, resistance to three major classes of antimalarials, aminoquinolines, aryl amino-alcohols and antifolates, has appeared to originate in South-east Asia or South America, despite the fact that the large majority of episodes of falciparum malaria occurs in Africa (Ekland and Fidock, 2007; Mita and Tanabe, 2012). This pattern might be explained by many factors, but a compelling explanation is that resistance develops most easily in areas of low malaria transmission intensity. These areas will have relatively non-immune human populations that are more likely than highly immune Africans to harbour resistant parasites with diminished fitness. They are also less likely to harbour polyclonal infections in which less fit resistant parasites may be outcompeted by sensitive strains. With time, as seen in animal studies, resistant parasites may develop compensatory mutations that increase fitness and enable spread even in high transmission areas. This scenario can explain the spread of resistance to chloroquine, antifolates and other drugs from Asia to Africa.
A look at clinical chloroquine resistance provides our best example of the interplay between resistance and fitness. Chloroquine-resistant P. falciparum was highly prevalent worldwide by the 1990s, but chloroquine use remained common, providing continued selection for the resistant phenotype. Then, in some well-studied areas, chloroquine was discontinued as the standard treatment for falciparum malaria. In Malawi, chloroquine use was eliminated and sulphadoxine-pyrimethamine made the national treatment regimen in 1993. Remarkably, dramatic shifts in parasite populations were seen. In samples collected from children with malaria between 1992 and 2000 the prevalence of the resistance-mediating pfcrt 76T mutation decreased steadily from 85% to 13% (with resistance to antifolates increasing concurrently) (Kublin et al., 2003). Following these changes, 99% efficacy of chloroquine for the treatment of falciparum malaria was demonstrated (Laufer et al., 2006). A similar, but less dramatic pattern was seen on Hainan Island, China, where chloroquine was discontinued as the standard treatment for malaria in 1979, and the prevalence of pfcrt 76T decreased from 90% in 1978–1981 to 54% in 2001 (Wang et al., 2005). The obvious explanation for dramatic changes after withdrawal of chloroquine is that discontinuation of drug pressure allowed reemergence of minority populations of chloroquine-sensitive parasites. Despite the obvious virulence of chloroquine-resistant P. falciparum, chloroquine-sensitive parasites have a clear fitness advantage.
Another means of assessing fitness costs is to compare parasites infecting individuals at different times of the year in areas with highly seasonal malaria (Babiker et al., 2013). During the dry season in these areas symptomatic malaria is uncommon and selective pressure from antimalarial drug use is low, but it is not uncommon for individuals to maintain low-level parasitaemia. A number of studies have shown increased prevalence of drug-sensitive parasites during the dry season compared with that during the transmission season. In samples from Sudan, the prevalence of the pfcrt 76T mutation was described as greater in samples from the dry season (Abdel-Muhsin et al., 2004), but these samples were collected just after the transmission season, so actually demonstrated selection of mutant parasites by recent drug pressure. This explanation was supported by a study from The Gambia in which the prevalence of the mutant alleles pfcrt 76T and pfmdr1 86Y decreased with increasing time since the transmission season (Ord et al., 2007). A study from Indonesia showed a similar, albeit less dramatic pattern, with decreased prevalence of the pfcrt 76T, pfmdr1 86Y, pfdhfr 108N and pfdhfr 59R mutations during the dry season (Asih et al., 2009). Taken together, available studies suggest a consistent pattern. During the dry season, with decreased use of antimalarial drugs, fitness advantages lead to replacement of mutant by wild-type parasites.
Asymptomatic plasmodial infections are common in high transmission areas in which human immunity is strong. Insight into fitness determinants may be gleaned by comparing parasites causing symptomatic and asymptomatic infections. A study in Kenya found no difference in the prevalence of pfcrt 76T or antifolate mutations between parasites causing symptomatic and asymptomatic infection (Zhong et al., 2008). In contrast, a recent cross-sectional study by our group in Uganda showed that children with asymptomatic infections had higher prevalence of the mutant pfcrt 76T, pfmdr1 86Y and pfmdr1 1246Y alleles compared with those with symptomatic infection, suggesting greater virulence for wild-type parasites (S. Tukwasibwe, unpubl. obs.). The impact of resistance-mediating polymorphisms on parasite densities may offer additional insight into parasite fitness. In Sudan, wild-type parasites achieved modestly higher parasite densities than parasites with the pfcrt 76T polymorphism; impacts of other resistance-mediating SNPs on parasite density were varied (Osman et al., 2007).
In a novel approach to assessing fitness determinants, the genotypes of parasites isolated from humans and from anopheline mosquitoes in Zambia were compared (Mharakurwa et al., 2011). In this region in which sulphadoxine-pyrimethamine had been heavily used, the prevalence of mutations that mediate resistance to pyrimethamine (pfdhfr 108N, 51I, 59R) was much higher in parasites infecting humans than those infecting mosquitoes. Other polymorphisms associated with resistance to another antifolate, cycloguanil, but rare in human infections in Africa, were much more common in mosquito infections. Thus, human infections under the selective pressure of frequent drug use may differ from those in mosquitoes, in which drug pressure is absent and other selective pressures may be in play.
Considering impacts of resistance-mediating polymorphisms on transmission, mosquitoes were fed on blood from Gambian children with malaria, and the transmissibility of drug-sensitive and -resistant parasites was compared (Hallett et al., 2004). Mosquito infection was much greater after feeding on blood containing gametocytes with the pfcrt 76T and pfmdr1 86Y mutations compared with blood with wild-type gametocytes. In another study in the Gambia, compared with those with wild-type infections, children with parasites containing pfcrt 76T had higher gametocyte densities, but those with pfmdr1 86Y lower gametocyte densities (Ord et al., 2007). In samples from Sudan, gametocyte production was greater in infections with the pfcrt 76T or pfmdr1 86Y polymorphisms (Osman et al., 2007). Thus, for these polymorphisms increased transmission may have in part circumvented fitness costs incurred in erythrocytic parasites, contributing to the spread of chloroquine resistance.
Conclusions and future perspectives
Studies utilizing cultured malaria parasites, animal models and samples collected from infected individuals have generally shown that resistance-mediating polymorphisms lead to malaria parasites that are out-competed by wild type in culture and in animals and that, in human infections, are replaced by wild type when drug pressure diminishes. However, results have been complex. Fitness costs of resistance may be offset by compensatory mutations that increase parasite virulence, and some polymorphisms associated with decreased fitness in erythrocytic parasites may improve transmission to mosquitoes.
We are at a critical juncture in man's battle with malaria. On one hand, we have a great opportunity for improved control leading to elimination, with improved tools to control mosquito vectors, a vaccine on the horizon, and highly effective ACT treatment regimens. These tools have led to important decreases in malaria in some areas in recent years. On the other hand, malaria remains an overwhelming problem in many areas, especially in Africa, and control measures are threatened by drug resistance, insecticide resistance, and parasite diversity and immune evasion mechanisms that challenge the development of a highly effective vaccine. In this context it is very important that we do not lose our best drugs, the ACTs, which have only recently reached widespread use around the world, and are likely the main contributor to recent advances in malaria control. But, all ACTs are already in serious jeopardy, with early signs of selection of resistance to artemisinins in South-east Asia and resistance concerns for all partner drugs. Further, we do not yet have any solid replacements for ACTs for the treatment of drug-resistant falciparum malaria. To maintain long effective lifespans for the ACTs we need to know how best to use them and other drugs for the treatment and prevention of malaria. More work is needed, building on recent in vitro, in vivo, and clinical studies, to understand the specific effects of different resistance-mediating polymorphisms on the ability of malaria parasites to cause serious infections and to be transmitted.
Relevant recent work in the author's laboratory has been supported by the National Institutes of Health and the Doris Duke Charitable Foundation.