Keywords

Introduction

Five Plasmodia spp. infect humans, and these cause distinct malarias that are distinguished by different pathophysiology and rates of mortality. These unicellular eukaryotic parasites belonging to the phylum Apicomplexa exist in the body as multiple highly differentiated forms. Mixed infections with multiple strains and multiple species can occur, and the pathophysiology of malaria in pre-immune versus naïve children, adults, and pregnant women differs. The overall point is that malaria is actually a spectrum of diseases with a variety of effects on different human populations, which presents many unique challenges in controlling the disease. The emergence and spread of drug-resistant strains of P. falciparum and P. vivax have further complicated treatment and threaten the lives of millions annually.

The life cycle of malarial parasites is complex, involving two hosts (Anopheles mosquitoes and humans) or, for P. knowlesi, three (transmission of this species is zoonotic via macaques). P. falciparum and P. vivax infections are the most common, with the former causing most mortality. P. falciparum sporozoites injected into the skin during an Anopheles blood meal quickly migrate to the liver, invade hepatocytes, and are then released back into the blood stream approximately 2 weeks later as large clusters of merozoites called merosomes. The individual merozoites then rapidly invade red blood cells (RBC). Once in the erythrocyte, the parasite proceeds through ring, trophozoite, and schizont stages of development before lysing the RBC within 48 h and emerging as ≥ 8 new merozoites. These then reinvade fresh RBC. Most clinical symptoms of malaria are a consequence of the RBC cycle, and most antimalarial drugs act against the RBC stages. See Bogitsch et al. (2005) for a detailed discussion of the parasite life cycle.

Quinoline Drugs, Emergence of Resistance, and Drug–Heme Interactions

Multiple effective classes of antimalarial drugs exist including the quinolines (4-amino-quinoline, 8-amino-quinoline, and quinoline methanols), the reactive endoperoxides (artemisinins), and antifolates such as pyrimethamine (typically administered in combination with sulfadoxine) that poison pyrimidine biosynthesis or utilization. Quinoline-based drugs (Fig. 1) have long been used in the battle against malaria, beginning with quinine [QN], originally extracted from the leaves of the cinchona tree. Upon the synthesis of chloroquine (CQ) and other 4-aminoquinolines during World War II (see Kauer et al. (2010) for a recent review), efficacious, cost-effective antimalarial drug therapy became readily available worldwide. Resulting widespread use of CQ led to the emergence of CQ-resistant (CQR) P. falciparum parasites, which was aided by prophylactic use and population-based dosing directed towards global eradication (Foley and Tilley 1998). Today, the majority of P. falciparum infections in S.E. Asia are CQR, and >50 % are CQR in many African countries, but pockets of CQ-sensitive (CQS) P. falciparum malaria still exist in South America and elsewhere. Importantly, quinoline antimalarials (QN), amodiaquine (AQ) (first synthesized at Parke-Davis in the late 1940s Burckhalter et al. 1948), and mefloquine (MQ), despite prolonged use, are still effective against most CQR parasites, and AQ and MQ are also important partners in approved artemisinin-based combination therapies (ACTs) (Schlagenhauf et al. 2010). Quinoline-quinoline and quinoline-non-quinoline combinations (Bell 2005), as well as quinoline-resistance reversal strategies (Peyton 2012), are additional components of ongoing development of new therapy.

Fig. 1
figure 1

Common antimalarial drugs

Full size image

In recent decades, as CQR strains have both spread around the globe and continued to evolve, new geographically distinct CQR strains with unique genotypes and phenotypes have appeared. Resistance to the antifolate drugs is also now widespread, and delayed clearance of parasites is now seen in some patients treated with artemisinin-based drugs (Cheeseman et al. 2012; Takala-Harrison et al. 2013) which may be an early sign of emerging resistance to this class of compounds as well. To use current drugs more effectively, and to develop new therapies, molecular details of antimalarial drug resistance phenomena must be elucidated.

Detailed biochemical and molecular analyses of drug resistance is complicated, but conceptually, drug resistance phenomena are quite simple. A drug must interact with one or more molecular targets to exert its effect, and so resistance to that drug is due to either (or both) disruption of that interaction or to altered signal transduction propagated from the drug–target interaction that would normally promote growth arrest or cell death. Disruption of drug–target interactions seen in drug resistance phenomena typically fall into one of three categories: increased enzymatic degradation of the drug, mutation or altered expression of the drug target, or altered cellular transport of the drug. In some cases, particularly in examples of “multidrug” resistance, more than one mechanism may be relevant. There are many examples of altered signal transduction related to drug resistance, with the best understood being disrupted apoptotic signal transduction (disrupted induction of programmed cell death) in multidrug-resistant tumor cells.

In the case of P. falciparum resistance to CQ and related quinoline antimalarial drugs, there is no known enzymatic degradation of drug that reduces drug–target interactions. As described in more detail below, the principal target for quinoline drugs is believed to be ferriprotoporphyrin IX (FPIX) heme released upon the host red blood cell hemoglobin (Hb) catabolism. Thus, the drug target cannot be mutated by the drug-resistant parasite because it is synthesized by the host. This leaves altered cellular transport of quinoline drugs as the likely pathway to disrupting drug–target interactions. However, as more has been learned about how free FPIX is processed by the parasite, it may also be possible that target accessibility is reduced in interesting novel ways, as briefly described below, which would then also disrupt drug–target interactions.

Regarding signal transduction related to parasite growth or death, very little is currently known about how quinoline drugs might affect that signaling, or how that signaling might be altered in drug-resistant parasites. Being a single-celled microorganism without clear caspases and other key apoptosis effectors encoded within its genome, P. falciparum does not appear to express a typical apoptosis pathway (Sinai and Roepe 2012), and our understanding of cell cycle regulation for the parasite is limited (Doerig et al. 2002; Halbert et al. 2010). Thus, changes in the signal transduction relevant for cell cycle regulation or cell death have not yet been inspected in any detail for drug-resistant P. falciparum malaria, although one very recent paper suggests that autophagy signaling may be related to parasite cell death (Gaviria et al. 2013). Progress in understanding signal transduction and other biology relevant for drug resistance in the related pathogen P. vivax is even more limited, but just as crucial (Shanks 2012; Douglas et al. 2012).

To analyze how drug–target interactions might be perturbed in drug resistance, the drug target must be understood in molecular terms. Quinoline antimalarials have long been thought to target FPIX heme within the digestive vacuole (DV) of the parasite, which is released upon Hb digestion during the trophozoite stage of the intraerythrocytic cycle (Banerjee et al. 2002; Gamboa de Domínguez and Rosenthal 1996; Elliot et al. 2008). The parasite must digest most Hb found in the RBC cytosol, both to provide room for very rapid trophozoite growth and to obtain necessary amino acids. FPIX is toxic in its free state (Fitch et al. 1983), and due to the lack of a heme oxygenase pathway, the malarial parasite must sequester FPIX as nontoxic crystalline hemozoin (Hz). At cytostatic dosages, quinoline drugs slow the production of Hz, presumably by interacting with uncrystallized Hz precursors, growing Hz crystal faces, or both. This inhibition of Hz presumably leads to the buildup of free heme which is then believed to inhibit Hb-degrading proteases (Vander Jagt et al. 1987) leading to growth arrest. Precisely how quinoline drugs target heme to inhibit Hz formation in vivo is not fully known nor is it understood whether different quinoline antimalarials inhibit Hz via similar or different pathways. Different pathways are likely since some CQR parasites remain sensitive to related quinoline drugs such as QN, AQ, and MQ (Fig. 1), whereas others do not.

Hz is a crystal of heme dimers, and the unit cell is a unique heme structure, with the ferric iron of each FPIX coordinated to a carboxyl side chain of an adjacent moiety. These “head-to-tail” dimers are stabilized in the crystal lattice via hydrogen bonding (Pagola et al. 2000; Bohle et al. 2012). Current evidence strongly supports catalysis of Hz formation by lipid in vivo (Jackson et al. 2004; Pisciotta et al. 2007; Gorka et al. 2013b), and certain lipids are known to efficiently catalyze Hz crystal growth in vitro (Jackson et al. 2004; Pisciotta et al. 2007; Egan et al. 2006). Structures for a number of quinoline drug-FPIX heme structures have recently been solved, and these are reviewed elsewhere (Gorka et al. 2013a).

As a diprotic weak base with pKa of 8.4 and 10.2, CQ exists as neutral, singly, or doubly charged compound under biological conditions, and these different drug species have different reactivity towards multiple chemical forms of free heme (i.e., monomers vs dimers in either aqueous or lipid phase, see Gorka et al. 2013a). In vitro, some drug–heme species aggregate and fall out of solution, generating amorphous drug–heme aggregates that then reestablish aqueous equilibria between heme species not complexed with drug. Others prefer to partition into lipid as 1:1 drug–heme complexes (Alumasa et al. 2010; Casabianca et al. 2008). This drug–heme chemistry likely competes with heme-to-hemozoin conversion; however, quantification of drug–heme aggregation or lipid partitioning phenomena within the parasite has not yet been done. Factors that reduce efficiency of quinoline drug–FPIX heme binding will alter DV retention of drug as well as the rate of Hz formation and could therefore contribute to resistance in multiple ways.

Genetic Basis of CQR

CQR is both spreading and continuing to genetically evolve via ongoing selective pressure. As briefly summarized in the next section, it was initially thought that ATP-binding cassette (ABC) protein drug pumps for CQ and other antimalarials must exist in drug-resistant P. falciparum and that these would be similar to the drug pump proposed for tumor cells (HsMDR1, or Pgp) believed by many investigators to directly translocate vinblastine, doxorubicin, and other antitumor drugs out of drug-resistant tumor cells. However, subsequent work showed that genotypes of some drug-resistant P. falciparum do not necessarily include mutation or increased expression of pfmdr genes (Wellems et al. 1990; Barnes et al. 1992) and that other genetic events must therefore be important. Similar to P. falciparum multidrug resistance (PfMDR) protein versus malarial MDR, the precise role of HsMDR1 protein in tumor MDR has been questioned (e.g., Roepe et al. 1996) as clinical pathology data have not correlated HsMDR1 overexpression with clinically relevant drug resistance as strongly as initially suspected.

Initial Genetic Studies of CQR: CG2 Versus Na+/H+ Exchange

More detailed genetic definition of CQR begins with cloning of pfcg2, which was suggested to be a resistance determinant based on quantitative trait loci (QTL) analysis of the progeny of a CQS × CQR parasite cross (Su et al. 1997). Subsequently, Lanzer and coworkers concluded that mutated PfCG2 protein was a dysregulated Na+/H+ exchanger that also pumped CQ out of parasites (Sanchez et al. 1998; Wünsch et al. 1998). Wellems and colleagues questioned this (Wellems et al. 1998). Subcellular localization of PfCG2 reveals it resides in vesicle-like structures near the parasitophorous vacuolar space (pvs) and the DV (Cooper et al. 2005) but is not localized within the plasma or pv membranes of the parasite as originally envisioned (Sanchez et al. 1998). A follow-up study (Bray et al. 1999) questioned Na+ dependency for CQR phenomena, arguing against a strong role for Na+/H+ exchange in CQ transport or CQR. More recently, QTL analysis, availability of the P. falciparum genome, and novel single-cell imaging of Na+/H+ exchange in a series of drug resistant progeny suggested that altered Na+/H+ might be related to QN resistance (QNR), but not to CQR, and that the relevant exchanger is not PfCG2, but Plasmodium falciparum Na+/H+ exchanger (PfNHE) (Ferdig et al. 2004; Bennett et al. 2007).

Wellems and colleagues found that mutant PfCG2 did not confer CQR in and of itself (Fidock et al. 2000a). Attention thus focused on another gene found within the same 36 kbp fragment that harbored pfcg2, namely, pfcrt (Fidock et al. 2000b). Results described in this and additional papers (Sidhu et al. 2002; Cooper et al. 2002) show that mutations in the PfCRT protein are the ultimate determinant of CQR (and of some degree of resistance to other drugs) in P. falciparum malaria (see “PfCRT,” below). While PfCRT mutations play a do minant role, importantly, PfMDR1 protein appears to modulate cross-resistance patterns in interesting ways.

The Elusive Role of PfMDR1

Early studies of CQR showed that drug resistance was associated with decreased drug accumulation (Krogstad et al. 1987) that was reversed by the ion channel blocker verapamil (VPL). Similar phenomena had been seen in drug-resistant tumor cells; thus, early on Wirth and colleagues screened P. falciparum for Hsmdr1 homologues and identified Pfmdr1 and Pfmdr2 (Wilson et al. 1989). Another group found Pfmdr1 to be upregulated in some CQR P. falciparum (Foote et al. 1989). But subsequent experiments (Barnes et al. 1992) showed that Pfmdr1 overexpression did not correlate with CQR. This was not entirely surprising since Wellems et al. had earlier shown that CQR did not segregate with the Pfmdr1 chr 5 locus in progeny from a CQS × CQR genetic cross (Wellems et al. 1990).

On the other hand, polymorphisms in pfmdr1 were also associated with CQR early on (Foote et al. 1990). While CQS isolates had identical pfmdr1 sequences, there were five changes in CQR isolates. In strains K1 and ITG2, N86Y was the only change. CQR strain 7G8 had four mutations: Y184F, S1034C, N1042D, and D1246Y. The 184F mutation was postulated as not likely involved in CQR since it was also found in CQS strains. Thus, the pfmdr1 overexpression hypothesis was revised to suggest that CQR strains expressed mutant pfmdr1 but did not necessarily overexpress wild-type pfmdr (Foote et al. 1990).

Subsequently, when MQR P. falciparum were selected to higher levels of MQR, pfmdr1 was found to be amplified (Cowman et al. 1994), and a very interesting inverse relationship between resistance to MQ and CQ was observed in the series of strains. Also, halofantrine (HF) and QN resistance increased with increasing pfmdr1, whereas AQ resistance did not (Cowman et al. 1994). However, when CQR strain K1 was selected versus HF, it did not result in MQR or amplification of pfmdr1 (Ritchie et al. 1996). In another study, which used allelic exchange of pfmdr to probe these questions, incorporation of pfmdr1 7G8 polymorphisms into a CQS strain not previously exposed to drug had no effect on CQR, but incorporating wild-type pfmdr1 into a CQR strain expressing mutant pfmdr1 did decrease the level of resistance by half (Reed et al. 2000). Also, the CQS strains expressing mutant pfmdr1 alleles showed some mild QNR and altered sensitivity to MQ and HF. Variations on this theme have also been described by Fidock and colleagues (Sidhu et al. 2005). Essentially, these data suggest that the pfmdr1 effects measured by Reed et al. may be strain specific, and they bring us to our current understanding (Roepe 2009). It seems unlikely that mutations in pfmdr1 confer CQR in and of themselves, but they can provide an important modulatory effect in some strains and isolates (Price et al. 2004; Dorsey et al. 2001; Patel et al. 2010). Interestingly, a recent report shows that PfMDR1 binds a high-affinity CQ photoaffinity analogue, suggesting that the protein does indeed react with quinoline drugs in some fashion (Pleeter et al. 2010), but the significance of this binding remains to be elucidated.

PfCRT

As mentioned, work by Wellems and colleagues showed that pfmdr1 was unlikely to cause CQR since the relevant region of chr 5 harboring pfmdr1 did not segregate with the CQR phenotype in a genetic cross (Wellems et al. 1990). Another key paper suggested that the CQR locus resided within the cg2 gene on chr 7 (Su et al. 1997), but this paper also showed that one CQS strain (Sudan 106/1) carried CQR-associated cg2 yet was nonetheless CQS. The 36 kbp chr 7 locus that segregated with CQR was thus reexamined, and a previously unrecognized gene, now known as pfcrt, was found (Fidock et al. 2000b). Mutations in pfcrt are the central determinant for P. falciparum CQR. The 13 exons of pfcrt span 3.1 kbp and encode a 424 amino acid, 48.6 kDa protein. Mutant-crt alleles found in CQR parasites contain a number of point mutations that confer multiple amino acid substitutions, with the pattern of mutations depending on the region of the globe from which the CQR parasite originates (Table 1). CQR arose (and continues to evolve) independently in at least five locations – S.E. Asia (which then spread to Africa), Papua New Guinea, Peru, Colombia, and the Philippines (Wooton et al. 2002; Chen et al. 2003). CQR parasites from S.E. Asia and Africa carry 7–8 point mutations, whereas South American CQR strains carry 5. Novel patterns continue to be discovered, including new alleles recently identified in the Philippines (Chen et al. 2003), Cambodia (Durrand et al. 2004), Columbia (Echeverry et al. 2007) China (Yang et al. 2007), and Thailand (Chaijarkoenkul et al. 2011). Based on these mutations, it appears that at least four amino acid substitutions are required for conversion to CQR, with a change at codon 76 always required. It is not completely understood why South American CQR strains segregate into two groups with distinct mutations, but a likely explanation is variable AQ selective pressure (Sá et al. 2009). The pattern of PfCRT mutations thus provides identification of the likely geographic origin of a CQR isolate. The number of mutations apparently required for conversion to CQR explains two riddles, namely, why CQR took so long to appear on a large scale and why it had historically been impossible to create CQR strains from CQS in the laboratory via drug selection pressure.

Table 1 All known PfCRT isoforms. Amino acid substitutions relative to wild type are shown in green. Where literature shows more than one measured IC50 for a strain harboring the PfCRT isoform, the high and low values are reported
Full size table

Thus, over the past 10 years, it has become clear that a number of distinct pfcrt alleles encoding unique PfCRT isoforms exist (see Table 1). These have presumably arisen for two reasons: (1) different antimalarial drug use in various regions of the globe has provided different selective pressure for the persistence of PfCRT mutations in these regions and (2) the patterns of mutations may provide different “fitness” advantages, some of which could be more specific to one region versus another. Ongoing efforts to sequence entire genomes of multiple P. falciparum strains and isolates will help further explain this ongoing parasite evolution (e.g., Wooton et al. 2002; Volkman et al. 2007). To date, a full molecular understanding of the relative resistance-conferring function of the different PfCRT isoforms known to exist is yet to be elucidated.

Similar to PfMDR1, PfCRT protein is localized to the DV membrane (Cooper et al. 2002) and is a polytopic integral membrane protein that performs some type of transport function (see Roepe 2011; Ecker et al. 2012 for recent reviews). Most functional hypotheses for PfCRT involve either ion or drug transport or both, since CQR parasites accumulate less antimalarial drug versus time relative to CQS (see below) and quinoline antimalarial drugs are hydrophobic weak bases. In fact, CQ and related drugs are dibasic, and the DV is known to be quite acidic. So, passive concentration of CQ within the DV (where FPIX heme CQ target is found) is dependent upon the square of the net pH gradient and will be 105–106-fold by the predictions of weak-base partitioning theory. A repercussion is that very subtle changes in DV pH will have quite significant consequences for drug sequestration. Regulation of DV pH is not fully understood, but it includes a V type H+ ATPase that hydrolyzes cytosolic ATP to pump H+ into the DV. Interestingly, small changes in DV pH and volume caused by mutation of PfCRT have been measured in some studies (Roepe 2011; Gligorijevic et al. 2006). These can affect drug partitioning, FPIX heme to Hz biomineralization, and the chemistry of Hz inhibition by drug (see Gorka et al. (2013a) for a more extensive discussion).

Altered Drug Transport Observed in CQR P. falciparum

Coy Fitch first observed reduced accumulation of CQ into iRBC infected with drug-resistant parasites (Fitch 1969), and decreased retention of CQ for CQR parasites was subsequently reported in zero-trans efflux experiments (Krogstad et al. 1987). This was termed “increased efflux” and quantified as the percent preloaded 3H-CQ remaining versus time after dilution into a drug-free medium. Efflux was hypothesized to be 40–50 times faster for CQR parasites (Krogstad et al. 1987); however, no molar quantification of transport (moles-free CQ per parasite per unit time) was possible. Soon thereafter, as described above several papers suggested that CQR was due to outward pumping of CQ by a P. falciparum homologue of human P-glycoprotein (HsMDR1 or Pgp), named PfMDR1, and encoded by the pfmdr1 gene on Pf chr 5 (Wilson et al. 1989; Foote et al. 1990). However, at about the same time, Wellems and colleagues reported that the determinant for CQR identified in a Mendelian cross resided on Pf chr 7, not chr 5 (Wellems et al. 1990), which led to the search for other genes involved in CQR and the subsequent identification of pfcrt 10 years later (Fidock et al. 2000b) (see “Genetic Basis of CQR” above).

Many cell-based drug influx and/or efflux studies have been reported for CQS versus CQR parasites (e.g., Geary et al. 1986; Bray et al. 1992; Roepe 2011). These were performed using iRBC populations or detergent-extracted parasites and various filtration or oil layer centrifugation approaches. Non-saturable drug accumulation was often calculated (e.g., see Hawley et al. 1998), and the different protocols quickly generated a variety of data. Consensus was nonetheless eventually reached, namely, at low external [CQ] (1–50 nM) CQR parasites typically accumulate 2–10-fold less CQ in similar time relative to CQS. Depending on calculated non-saturable accumulation subtracted from these data, some studies (Bray et al. 1992) hypothesize that saturable uptake differs by as much as 100–1,000 fold for CQR versus CQS parasites; however, this conclusion rests on mathematical modeling assumptions. Measured differences in net CQ accumulation are typically 2–10 fold (see Roepe 2011 for more detailed review).

A proposed faster rate of drug efflux back out of the iRBC was one popular explanation for reduced iRBC drug accumulation early on (Krogstad et al. 1987; Martin et al. 1987), and initially the PfMDR1 protein (see section “The Elusive Role of PfMDR1”) was thought to mediate the hypothesized increased efflux. After identification of PfCRT (see section “The Elusive Role of PfMDR1”), it was proposed that PfCRT, not PfMDR1, was responsible for the putative increased cellular drug efflux. Several interpretations were offered: (1) PfCRT-mediated outward pumping of drug from CQR parasites, (2) PfCRT-mediated drug counterflow (exchange) in the presence of appropriate drug gradients, (3) altered binding to intracellular targets caused by PfCRT mutations promoted decreased drug retention (increased efflux) in zero-trans efflux experiments.

However, both PfCRT and PfMDR1 proteins reside in a subcellular organellar membrane (the DV membrane), with three additional membranes between it and the outside of the iRBC. How one transporter at the DV membrane (via whatever thermodynamic mechanism one chooses to invoke) could kinetically compete with the fast passive influx of drug across these other membranes to result in net movement of drug out of the entire iRBC remains unclear. This leads to the proposal that PfCRT facilitates downhill leak of charged CQ from the DV to the cytosol (Zhang et al. 2004). Thus, PfCRT would not “pump” CQ from the iRBC per se but would promote decreased accumulation over time by lowering time-dependent CQ binding to intra-DV targets (e.g., FPIX heme, discussed above).

Interestingly, with one noted exception (Geary et al. 1986), all cell-based transport during this period was assayed at 1–10 nM levels of CQ. However, these concentrations are approximately 100–1,000 times below peak plasma [CQ] in malaria patients. In situ autoradiography of trophozoites showed that accumulated 3H-CQ localized nearly exclusively to the DV of the mid-stage trophozoite (Sullivan et al. 1996), but these experiments were done at even lower external levels of drug (pM). Nonetheless, when abundant PfCRT was found expressed in the DV membrane, nearly all evidence seemed to point in favor of a straightforward interpretation of the Fitch hypothesis (Fitch 1969) and a DV membrane drug pump or drug channel explanation for CQR. Only recently have resistance phenomena related to higher (cytocidal) levels of CQ been investigated and evaluated versus drug transport phenomena (see below “New Insights: Cytostatic Versus Cytocidal Resistance”).

An additional caveat for interpreting altered drug accumulation in iRBC is that some studies have reported altered DV pH and volume for CQR parasites (reviewed in Roepe 2011). It seems likely that traffic of endogenous DV osmolytes (ions and/or small molecule metabolites) is perturbed upon mutation of wild-type PfCRT to CQR isoforms found in CQR parasites (Roepe 2011). Altered osmolyte traffic perturbs regulation of important biochemical characteristics of the DV (e.g., pH, volume, ionic composition) that have direct effects on the efficiency of quinoline drug–FPIX heme interactions, and hence on net drug accumulation versus time (see Gorka et al. 2013a).

Vesicle Studies

From studies with whole cells, it was not entirely clear whether reduced accumulation of drug versus time for iRBC harboring CQR parasites was due to transporter-mediated efflux from the DV, altered binding of drug to FPIX heme targets caused by perturbations in DV physiology, or some combination. In most cases, analysis of drug transport with vesicles or proteoliposomes reduces complexity in interpretation. Initial vesicle-based studies of hypothesized PfCRT drug transport function used plasma membranes from yeast and first tested for direct binding of 3H-CQ to PfCRT (Zhang et al. 2004). Scatchard analysis indicated a single-drug binding site in PfCRT, and, surprisingly, that CQS and CQR isoforms of PfCRT have similar affinity for CQ (Kd = 435 and 385 nM, respectively). A recent follow-up study of CQ binding used covalent attachment of a perfluoroazido-tagged CQ probe to quantify CQ probe binding versus other quinoline antimalarials and to further define the drug binding site in PfCRT, which is predicted to be disposed towards the DV side of the DV membrane (Lekostaj et al. 2008). Satisfyingly, this binding site can also easily place the quinoline ring of CQ near mutations in PfCRT isoforms that are known to modulate response to drugs (Fidock et al. 2000b; Cooper et al. 2002). It is clear at this point that both wild-type (CQS) and mutant (CQR) isoforms of PfCRT bind CQ at a single-drug binding site, that related quinoline drugs such as MQ and QN compete with CQ for binding to this site, and that chemoreversal agents such as VPL inhibit CQ binding only for some PfCRT isoforms (Lekostaj et al. 2008).

The paper showing direct binding of 3H-CQ (Zhang et al. 2004) also examined 3H-CQ efflux from inside–out plasma membrane vesicles (ISOV) via flow dialysis techniques and concluded that one CQR isoform of PfCRT mediated downhill passive efflux of 3H-CQ faster than that observed for control ISOV or ISOV harboring CQS PfCRT. This was the first direct biochemical evidence in support of CQ transport by PfCRT. Subsequently, another paper applying a similar approach with vesicles made from D. discoideum reached similar conclusions after observing that vesicles harboring mutant PfCRT accumulated less CQ than those harboring wild-type PfCRT (Naude et al. 2005). In this study as well as an earlier yeast vesicle study (Zhang et al. 2002), additional evidence for ion or osmolyte transport via PfCRT was also obtained. A more recent D. discoideum vesicle study supports the notion that CQ transport is likely driven by electrochemical potential (Papakrivos et al. 2012).

Analysis of CQ Transport Using Proteoliposomes and Oocytes

Injection of oocytes with modified pfcrt mRNA followed by measuring 3H-CQ accumulation into individual eggs (Martin et al. 2009) and purification of recombinant PfCRT from yeast followed by reconstitution into proteoliposomes (PLs) and analysis of fluorescently tagged CQ efflux from these PLs (Paguio et al. 2009) have both recently been pursued to test conclusions regarding PfCRT-mediated drug transport (see Roepe 2011 for more detailed review). Both the PL and oocyte approaches provide the best evidence for direct CQ transport by PfCRT; however, there are important differences in interpretation between the two studies. One is quantification of apparent turnover (mole drug/mol transporter(s)), and the other is whether both CQS and CQR isoforms are capable of drug transport. The paper reporting CQ transport in oocytes (Martin et al. 2009) does not calculate explicit turnover, presumably because the transport measured in this study is quite slow and does not plateau. However, initial rates are calculated and expressed as pmol CQ/oocyte/h. Assuming site density of PfCRT is within the range reported for many other transporters and channels expressed in oocytes (68), then these data convert to 0.002–0.02 CQ molecules/PfCRT(s) (CQS isoform) and 0.009–0.09 CQ molecules/PfCRT(s) (CQR isoform) at 300 nM external 3H-CQ. This estimated turnover is 1–2 orders of magnitude lower than that measured with fluorescent drug probe and purified protein reconstituted into PLs (Paguio et al. 2009) and does not appear sufficient to account for reduced CQ accumulation in the parasite DV for CQR parasites (Cabrera et al. 2009; Roepe 2011).

The PL experiments are done with preparations wherein efflux of free CQ probe from an acidified PL interior can be measured instead of influx into the oocyte from a neutral egg perfusate. Experiments with these PLs and the fluorescent CQ analogue yield turnovers that are much higher than those computed from the oocyte data and that are also found to be highly dependent on the magnitude of ΔpH and ΔΨ (Paguio et al. 2009). This is expected for a DV transporter since high electrochemical driving forces exist across the DV membrane. At 5 μM NBD-CQ, turnover numbers were determined to be 0.8 NBD-CQ molecules/PfCRT(s) in the presence of a 1 unit ΔpH and 0 mV ΔΨ, 1.6 NBD-CQ molecules/PfCRT(s) in the presence of a 2 unit ΔpH and 0 mV ΔΨ, and 3.4 NBD-CQ molecules/PfCRT(s) in the presence of a 2 unit ΔpH and ~120 mV ΔΨ+ (Paguio et al. 2009).

Data from the two approaches probably differs for several reasons. First, oocyte plasma membranes have low electrochemical driving force that cannot be conveniently manipulated, whereas driving force can be both increased significantly and conveniently manipulated for the PLs. Also in the PL experiments, the only modification to the amino acid sequence of PfCRT is a hexa-His tag added at the C-terminus, but in oocytes, four putative lysosomal-/endosomal-targeting motifs were removed in PfCRT by replacing 15 residues (a.a. # 17, 20, 22, 23, 26, 27, 47, 48, 50, 51, 409, 412, 414, 421, 422) with alanine (Martin et al. 2009). It is possible that, along with very different electrochemical driving force, these extensive modifications to the PfCRT primary sequence affect catalytic efficiency of CQ transport.

A second key difference in comparing these studies is the relative transport measured for CQS versus CQR isoforms of PfCRT. In the PL study, small differences in CQ transport by the two isoforms are noted when transport is measured at the same ΔpH and ΔΨ. In contrast, the oocyte system does not show statistically significant CQ transport above background for the CQS isoform of PfCRT. Levels of transporter are difficult to quantify for the oocyte system, and western blot data that directly compares CQS to CQR PfCRT isoform expression in the oocytes is not shown in this study (Martin et al. 2009). Also, as mentioned, the extensive N-terminal modifications that are necessary for effective oocyte expression of PfCRT could be compromising activity, as could different lipid composition in the different membrane systems.

Most recently, Baro et al. (2011) engineered galactose-inducible expression of PfCRT in metabolically active growing yeast. Since the majority of PfCRT protein in this system is expressed at the plasma membrane and since the topology of the protein (cytosolic domains remain cytosolic, intra-DV domains are disposed outside the cell) as well as DV membrane bioenergetics (high delta pH acid outside) are preserved relative to DV-disposed PfCRT, function of PfCRT can be analyzed by plating or growing the live yeast versus various external [CQ]. In these experiments, both CQS and CQR isoforms of PfCRT are found to transport CQ, but the CQR isoforms transport at higher efficiency and with a stronger dependence on membrane potential (Baro et al. 2011), consistent with observations made with purified PfCRT reconstituted into proteoliposomes (Paguio et al. 2009). A recent review (Roepe 2011) discusses other factors that could explain the disagreements between oocyte versus yeast and proteoliposome data; regardless, the important point is that considerable evidence now exists in support of direct transport of CQ by PfCRT driven by membrane potential. There is not yet complete agreement on whether this transport is more “channel-like” versus “carrier-like” (see also Papakrivos et al. 2012), thus additional studies with purified protein will be needed to refine the remaining hypotheses.

Most recently, (Baro et al. 2013) the inducible yeast expression system has been used to quantify differences in CQ transport by multiple CQR-associated PfCRT isoforms. Surprisingly, plots of CQ transport efficiency versus CQ IC50 for the strains in which the PfCRT isoforms are found reveal very poor correlation. A simple interpretation is that mutant PfCRT alone does not explain the degree of CQR.

New Insights: Cytostatic Versus Cytocidal Resistance

The above summary presents a satisfying model for PfCRT-mediated CQR further modulated by PfMDR1 and perhaps other factors. However, this assessment of CQR phenomena is based upon an incomplete definition of CQ pharmacology. With one exception (Paguio et al. 2011), for decades, all quantification of CQR has been via computing a ratio in CQ IC50 for CQR versus CQS strains or isolates. IC50 are determined from long-term growth inhibition assays wherein live parasites are grown for 1–3 iRBC cycles in the constant presence of CQ. These IC50 are in the 101–102 nM range (depending on the strain) and are relatively easy to obtain, including in high-throughput fashion with live cells (Smilkstein et al. 2004; Bennett et al. 2004a). Growth inhibition of parasites is highly relevant to the development of antimalarial therapy, because a good antimalarial drug should prevent increases in parasitemia and recrudescence. But it is also true that when CQ is administered to a malaria patient, the plasma concentration of the drug is typically >1 μM (not 10–100 nM), for at least the first 6–12 h. The most important initial effect of CQ therapy is significant reduction of parasitemia from 1012 to 1011 parasites to ≤109, within hours. Meaning, successful clinical administration of CQ kills many parasites, it does not merely prevent their growth. A patient infected with CQR P. falciparum does not show this dramatic drop in parasitemia due to parasite death from micromolar CQ dose. Meaning, clinically relevant CQR can also be defined via an elevated LD50 (“Lethal Dose”). LD50 (defined as survival after bolus dose of CQ, see Paguio et al. 2011; Cabrera et al. 2009) have only been reported for two laboratory strains of P. falciparum. Many more such studies obviously need to be done. In one recent study, when drug accumulation is analyzed for intact iRBC using LD50 levels of drug (not IC50 levels), reduction in DV accumulation is not found for the CQR parasites relative to CQS (Cabrera et al. 2009). In fact, this study reports that CQR parasites can accumulate more toxic CQ relative to CQS and still exhibit resistance to drug-induced cell death. It is of course not uncommon for antimicrobial or anticancer drugs to show both growth inhibitory (cytostatic) and cell killing (cytocidal) effects. When they do, cytocidal activity often (but not always) requires higher levels of drug or longer exposure to drug. It is also not uncommon for targets that are relevant for cytostatic functions of a drug to differ from those that are relevant for cytocidal. It is critical then to point out that, with two exceptions (Geary et al. 1986; Cabrera et al. 2009), nearly all detailed CQ transport analyses for CQR versus CQS parasites, vesicles, or oocytes that have been used to develop models for CQR and PfCRT function have been done at sub IC50 levels of drug (typically, 1–50 nM). This has led to logical explanations for CQR that are relevant for resistance to the cytostatic functions of CQ (“CQRCS”) but that, at least initially, do not appear to be completely relevant for resistance to the cytocidal functions of CQ (“CQRCC,” Cabrera et al. 2009). To fully elucidate CQR, and in the design of additional antimalarial chemotherapy, both facets are critical, and more work is required to understand the latter.

At LD50 doses of the drug, [CQ] in the DV will be very far above Kd for PfCRT, suggesting any direct PfCRT-mediated CQ transport would be overwhelmed by increased passive diffusion of the drug. Assuming therefore that targets for CQ static versus cidal effects differ (see also Gorka et al. 2013b), then those relevant for the latter might exist outside the DV, in the cytosol, in the nucleus, or in some other organelle. A search for these possible targets, better definition of LD50 versus IC50 for a number of drugs, and continued analysis of the differences between strains harboring various PfCRT and PfMDR1 isoforms are important topics for future research. These points are likely also relevant for defining what is apparently a non-mutated CRT mechanism for CQR in P. vivax malaria (Nomura et al. 2001; Baro et al. 2011).

Most recently, at least some of the additional biochemistry relevant for CQRCC has been revealed (Gaviria et al. 2013). By performing LD50-directed QTL for progeny of the HB3 (CQS) × Dd2 (CQR) cross (instead of IC50-directed; Fidock et al. 2000b), it was found that CQSCS and CQRCC are genetically distinct, and additional chromosomal loci were found to be inherited in an LD50-dependent fashion. Analysis of these loci suggests biological processes that could be altered in the development of resistance to the cytocidal effects of CQ. One process in particular, similar to autophagy, was found to be dysregulated in CQR parasites via analysis of PfATG8 protein distribution upon exposure to LD50 levels of CQ (Gaviria et al. 2013).

Conclusion

Resistance to CQ and other common antimalarials has historically been quantified by ratioing IC50 that quantify drug cytostatic activity. Drug transport experiments at IC50 dosages, genetics, and molecular pharmacology of drug–heme interactions have generated a molecular model for cytostatic CQR wherein mutations in PfCRT cause increased, electrochemically downhill, leak of CQ (and possibly other quinoline drugs) from the DV. This model is strongly supported by recent experiments with purified PfCRT protein that both define a single CQ-binding site and that show membrane potential-driven transport by PfCRT. Some conflicting interpretations regarding the efficiency of drug transport by different PfCRT isoforms exist based on other recent experiments with oocytes, but overall these experiments also support PfCRT-mediated CQ transport. Questions that remain to be elucidated include defining relative affinities of related quinoline drugs for PfCRT isoforms and determining the efficiency with which they might be transported. Also, resistance to the cytocidal action of CQ does not appear to require decreased cellular accumulation of CQ. The precise role that PfCRT and other transport proteins (PfMDR1, PfNHE) play in this phenomenon remains to be explored.

The Fitch/Macomber/Sprinz hypothesis (Fitch 1969; Macomber and Spintz 1967) states that the DV is the principal site of CQ accumulation because heme released from Hb catabolism within the DV is its principal molecular target. Simply moving CQ from the DV faster than it passively diffuses back inward from the cytosol would then decrease binding to heme target that is continuously delivered as Hb is digested, and thereby cause CQR. In addition, the dynamics of CQ2+ versus CQ+ versus CQ binding to different chemical forms of heme (monomer vs μ-oxo dimer vs head-to-tail dimer) might be different in CQS versus CQR parasites, further contributing to decreased retention of CQ. In support of both models, several studies have defined both covalent and non-covalent heme–drug complexes for CQ and related drugs that are influenced by CQ protonation (Gorka et al. 2013a). Decreased pH and volume-dependent CQ-heme adduct accumulation within the DV of CQR parasites may be part of the explanation for elevated IC50 in these CQR parasites. However, the early observations of Geary and Ginsburg (Geary et al. 1986), along with recent distinction between IC50 versus LD50 phenomena (Paguio et al. 2011; Gaviria et al. 2013), continue to hint that the mechanism of clinically relevant cytocidal CQR is likely multifaceted. Perhaps related to this, interestingly, mutations in the P. vivax orthologue PvCRT apparently do not cause CQR in P. vivax malaria (Nomura et al. 2001). We suggest the additional layers to CQR in P. falciparum, which likely represent biochemistry relevant to raising CQ LD50 (but not necessarily IC50), will provide important clues for the mechanism of CQR in P. vivax malaria.

Lastly, for both P. falciparum and P. vivax malaria , parasite CQR can be overcome with new chemotherapy that can be perfected by screening versus CQR strains (e.g., Peyton 2012). It has been hoped that detailed knowledge of CQR would expedite second-tier drug development versus malaria, and this hope is now beginning to be realized. A key component to new antimalarial chemotherapy is (and will continue to be) the additive or synergistic effects of two or more drugs given in combination. We suggest more detailed quantification of IC50 versus LD50 synergies is an important new avenue to explore in antimalarial drug discovery (Gorka et al. 2013c).