Early support for the involvement of P. falciparum MRPs in malaria drug resistance came from one of the first re-sequencing projects after unveiling P. falciparum’s genome (Mu et al., 2003). Upon the ORF sequencing of 49 putative transporter genes in 97 culture-adapted parasites from widely diverse geographical origins, associations with chloroquine (and quinine, see below) IC₅₀ levels were searched. Positive associations were found between pfmrp1 (originally termed “G2”) H191Y and S437A SNPs and increased CQ IC₅₀’s [(³H) hypoxanthine uptake methods, 48 h (Desjardin et al., 1979)] Such results were consistent with the fact that chloroquine (CQ) is an MRP1 substrate in human cells (Vezmar and Georges, 1998), and a potential factor in modulating the use of this drug for arthritis management (Oerlemans et al., 2006).

A re-evaluation of these pfmrp1 SNPs was subsequently performed in a set of 107 ex-vivo characterized infections from a single hospital in the Thai-Myanmar border during a short period of time, avoiding biasing from comparing parasites derived from populations with different genetic structures (Anderson et al., 2005). The analysis did not reproduce the CQ vs. H191Y/S437A associations, albeit it has to be noted that both pfmrp1 polymorphisms were fixed (Y191: 100%) or near fixed (A437: 86%) in the region, severely limiting the power of the study. Also, the remaining of the gene ORF was not investigated, not ruling out the potential involvement of other SNPs, as later investigated (e.g. Veiga et al., 2011; Bai et al., 2018). Such SNP associations with CQ IC₅₀s [HRP2 double-site Sandwich linked immunosorbent assay, 72 h (Noedl et al., 2005)] were not observed in the analysis of a set of 48 SE Asian parasites from the same Thai-Myanmar Western border region, where the full pfmrp1 ORF was characterized (Veiga et al., 2011). Differently from the (Mu et al., 2003) collection of analyzed strains, all these SE Asian isolates were highly resistant to CQ. This might have overshadowed a potential secondary pfMRP1 contribution, likely to be only detectable through a much larger trial. A more recent study in the Mae Sot region found a higher prevalence of the resistance associated 191Y allele (13%) among 119 adapted parasites, having actually reported a significant association with CQ IC₅₀s Phompradit et al., 2014). In another study performed with culture adapted parasites now from the China-Myanar border during the 2007–2012, the T1007M polymorphism was associated with higher CQ IC₅₀s (Bai et al., 2018).

By mainly using parasites adapted from returning travellers to France, Pradines and collaborators published a series of reports analyzing relatively small sets of strains (n = ca. 20–23) from different geographical origin and range of CQ sensitivities (Henry et al., 2008; Henry et al., 2009) (Table 1). Contrarily to the reports focused on SE Asia, the authors repeatidly observed associations of the H191Y and S437A polymorphisms and increased CQ IC₅₀ values in these sets of parasites of heterologous origin (it is important to note that between reports a significant degree of overlap exists between the sets of strains used). These studies, supportive of the aforementioned initial observations (Mu et al., 2003) were limited to the analysis of these two SNPs, opening the possibility of the latter to be in linkage with yet other participating polymorphisms in the pfmrp1 ORF. Consistent with this view, the I876V SNP—previously connected with in vivo artemether-lumefantrine (AL) response (Dahlstrom et al., 2009) (see below)—was seen associated with increased CQ IC₅₀s (SYBR green based fluorescence assay, 72 h) (Meng et al., 2010) in 63 P. falciparum adapted parasites from the Northern Thai-Chinese border (Hao et al., 2013). The fully sequencing of the pfmrp1 coding region did not unveil other associations, including with the H191Y and S437A SNPs, which were as prevalent as I876V. The global prevalence of the latter SNP was speculated to be a remnant of the worldwide spread of CQ resistance during the XX Century (Hao et al., 2013). A later, smaller study, found an additional association with the I1390 allele (Gupta et al., 2014), which has not been previously detected (Veiga et al., 2011).

These associations between CQ response and pfMRP1 are supportive of previous genetic modification works in W2, reporting a significant decrease in IC₅₀’s (72-h incubation essays) of ca. 1.6 fold (∼160–∼100 nM) pfmrp1 excision (Raj et al., 2009). The authors further explored the possibility of these differences to be related with the capacity of pfMRP1 to transport CQ. This was performed through 6-point drug accumulation essays, using radioactive labeled (³H) chloroquine for a period of 1 hour. Scintilation rate measures of the intact parasites showed a significant increase in CQ accumulation in the pfmrp1 knocked-out parasite. At the final time point of the experiment (60 min), the difference in accumulation was ca. 2 fold.

Somewhat surprisingly, this sensitization effect was not registered by Rijpma et al. (2016) when deleting pfmrp1 (and pfmrp2) in the NF54 strain. In fact, and as referred in more detail below, the ablation of these genes essentially did not significantly influence the NF54 in vitro IC₅₀s against quinolone drugs—with the possible exception of MQ—and artemisinins. The reasons behind these marked differences are not clear. As previously referred in terms of fitness effects, they might be related to the different genomic context between the South East Asian W2 and the NF54 East African strains.

CQ was recently demonstrated in vitro to directly interact with pfMRP1. Using a CQ 7-nitrobenz-2-oxa-1,3-diazole (NBD)-labeled fluorescent derivative, Woodland and collaborators detected significant fluorescence in the parasite plasma membrane. A drug-labeled affinity matrix based approach subsequently identified a 200–250 Kd plasma membrane associated protein. Through mass spectrometry-based proteomics, this protein was further shown to be pfMRP1 (Woodland et al., 2018), supporting the concept of CQ as a substrate for this ABC transporter, and consequently, a factor in the parasite response.

As for pfMRP2, a suggested trend for association between the synonymous SNP t3414c and chloroquine IC₅₀s (660 vs. 564 nM) has been reported (Veiga et al., 2014). Any contribution of ORF polymorphism of this transporter in CQ response is likely to be minor. The question remains on the importance of non-coding SNPs located in the putative 5′regulatory regions of the gene, taking in consideration the implications of transcriptional control in this gene and drug response (Mok et al., 2014).

As for amodiaquine’s clinically relevant metabolite desethylamodiaquine (DEAQ), less data is available (Table 1). The pfMRP1 H191Y/S437A SNPs have been reported to be significantly linked with increased DEAQ IC₅₀s in 15 adapted parasites (Henry et al., 2008), a result subsequently supported through the same approach (Parquet et al., 2009; Briolant et al., 2010; Pradines et al., 2010; Pascual et al., 2011). Such an association was not previously observed in a set of 15 South American parasite strains (Western Colombia) (Echeverry et al., 2007) where, albeit a large range of DEAQ (and CQ) responses (DEAQ IC₅₀’s ca. 40–350 nM, CQ IC₅₀’s ca. 115–415 nM), all pfmrp1 ORFs were 3D7-like (“wild type”) (Dahlstrom et al., 2009).

A highly sensitive approach for detecting the involvement of a particular gene in antimalarial drug response is the monitoring of gene polymorphism selection upon antimalarial therapy. In general there are very few in vivo studies investigating this treatment driven selection of pfmrp SNPs among recurrent infections. To our knowledge, no such analysis has been performed in CQ or DHA-PPQ efficacy trials. As for AQ/DEAQ, an early report studying the H191Y/S437A SNPs in an AQ monotherapy study in Kenya did not show signs of selection (Holmgren et al., 2006). It must be anyway stressed that the low prevalence of both mutant alleles (<10%) and the small size of the study (81 patients), severely limited its power. Nevertheless, a larger follow-up analysis of two ASAQ clinical trials in Zanzibar and mainland Tanzania did not show any signs of H191Y/S437A selection (Dahlstrom et al., 2009), supporting the previous observations in Kenya. The subsequent analysis of an effectiveness study in Benin could not reach conclusions concerning the I876V SNPs due to rarity of the mutant alleles, while K1466R did not show trends of selection, or associations with DEAQ ex-vivo IC₅₀s (Dahlstrom et al., 2014). Another study, now an efficacy trial in Liberia confirmed the low prevalence of the 876V (<1%) and 1466R (ca. 2.5%) alleles, not allowing the drawing of conclusions (Otienoburu et al., 2016). The available data prompts the conclusion that pfmrp1 is at most a secondary contributor for AQ/DEAQ resistance.

As with CQ, (Mu et al., 2003) reported significant positive associations between the pfmrp1 (“G2”) 191Y and 437A alleles and increased IC₅₀ values. This result was later repeatedly confirmed in smaller collections of parasite strains using similar isotopic in vitro tests (Henry et al., 2008, 2009; Parquet et al., 2009; Parquet et al., 2010; Pradines et al., 2010; Briolant et al., 2010; Pascual et al., 2011) (Table 1). The importance of the 191 and 437 a.a. status with QN response field samples was not possible to be robustly tested by Anderson et al. (2005) in a large set (>200, 1998–2003 period) of ex-vivo characterized SE Asian parasites, due to the rarity of these polymorphisms found in that study, as previously referred. A subsequent study, also performed in the Mae Sot region (2006–2009 period) in 119 culture-adapted parasites extended the number of a.a. positions under scrutinity, with a significant association found between QN IC₅₀s and F1390I (F1390: 238 nM, 1390I: 190 nM) (Phompradit et al., 2014).

The whole deletion of pfmrp1 in the context of the W2 clone (W2/MRP∆) gave rise to a 1.8 fold (∼140–∼80 nM) decrease in IC₅₀ (72 h exposure) to piperaquine. This is consistent with data where the uricosuric drug probenecid (PBN) has been shown to be able to sensitize P. falciparum to this antimalarial (Masseno et al., 2009), albeit it is important to note the relatively unspecificity of this reversal agent, which might be influencing other transport systems. The pfmrp1 191 and 437 SNPs were not associated with in vitro sensitivity in 23 parasite strains (Briolant et al., 2010), a result supported by Hao et al. (2013) when analyzing 63 strains from Northern Myanmar (Table 1). In the same region, a subsequent study found a significant increase in PPQ IC₅₀s among pfmrp1 1007M and 785N carrying parasites (Bai et al., 2018), reinforcing the view of the need of full ORF sequences in order to more robustely unveil valuable associations with this gene.

Lumefantrine (LUM) is part of the most used antimalarial worldwide, in combination with artemether. Two independent Coartem^® (Artemether-Lumefantrine) clinical comparative efficacy trials in East Africa documented a significant selection of the pfmrp1 876I allele among recurrent parasites (frequencies of 11% vs 2.6%, respectively) (Dahlström et al., 2009). The I876V SNP is immediately adjacent to the Walker A/NBD1 (Figure 1). In silico analysis of the I→V change have suggested that even though the structural differences between these two a.a. are small, it may affect the rate of the ATP hydrolysis cycle of this ABC transporter (Dahlström et al., 2009). The pfmrp1 876I association was specific for Coartem^® - no significant selection was detected concerning the comparator chemotherapies, ASAQ as previously referred, and SP.

Mefloquine sensitivity was not changed by pfmrp1 deletion in the W2 strain (Raj et al., 2009). These results are somewhat consistent with the in vitro data by Wurtz et al. (2010), where no significant associations were found between the 191 and 437 SNPs and MQ IC₅₀s (n = 21 strains). On the other hand, in NF54, The removal of pfMRP1 led to a significant 2-fold increase in IC₅₀ values (121 vs 241 nM) (Rijpma et al., 2016). Intrigingly, the deletion of pfMRP2 in this strain leads to an identical outcome (IC₅₀ = 242 nM), and a similar one when both are missing (IC₅₀ = 272 nM). This lack of effect reinforcement is somewhat unexpected, possibly meaning that whatever influence of the MRP system in MQ response, it reaches its maximum outcome with one of the proteins missing.

In South East Asia, pfMRP1 F1390I has been associated with significant changes in IC₅₀ values (72 h exposure) in a set of 48 culture-adapted parasites from this region (Veiga et al., 2011). F1390 carriers had significantly increased IC₅₀s for both mefloquine (ca. 122 nM vs. 37 nm) and lumefantrine (16 nM vs. 6 nM). This 3’ ORF region polymorphism, frequent in the SE Asian regions (Nyunt et al., 2015, 2017), codes for an amino acid position located at—or at least in close interaction—with TM11. Assuming TM as part of the internal lining of the tranporter interacting with the drugs, it is possible that mutations in these regions (like the 191 and 437 a.a.) will directly participate in the transport process (Table 1). In 43 culture-adapted parasites from Burmanese patients, a significant association between the 191Y, 437A and 876V alleles and increased mefloquine IC50′s was reported, but no link with the F1390I SNP (Phompradit et al., 2014a). In another in vitro study from the northern Miyanmar-China region, the H785N SNP was associated with significant changes in LUM IC₅₀s (Bai et al., 2018).

Considering the documented lower Artemether-Lumefantrine efficacy in these regions (Price et al., 2006), it is conceivable that pfmrp1 genes harbouring this variant might constitute an additional contributing resistance factor, along with 86N carrying pfmdr1 duplications (Calçada et al., 2020).

Concerning pfMRP2, IC₅₀ determination (48 h exposure, [³H]hypoxanthine incorporation) of a selected set of culture adapted parasites (n = 67) pointed for a significant association between the presence of seven copies of ORF-located short repeats [DNNNTS/NNNNTS, indel region III (Veiga et al., 2014) (Figure 1)] and IC₅₀ for lumefantrine (Okombo et al., 2013). Such associations were not seen among SE Asian culture adapted parasites (Veiga et al., 2014).

When considering each of the two MRPs independently, (Rijpma et al., 2016) did not documented changes in di-hydroartemisinin sensitivity in NF54, although a suggestive trend was noticed of less sensitivity to DHA when pfmrp1 and pfmrp2 were both ablated (43 vs. 134 nM). Again this comes in contrast with the reported for W2, where the deletion of pfmrp1 led to a prompt sensitization to artemisinin (from 11 to 5 nM), not observed in NF54. Concerning sequence variation, the status of the previously mentioned pfMRP1 F1390I SNP was shown to influence the parasite in vitro response (72 h exposure), of both artesunate and DHA (Veiga et al., 2011), the F1390 allele being associated with increased IC₅₀s (artemisinin: 9 vs 3 nM; DHA: 1.5 vs. 0.9 nM). This initial observation was supported in a larger set of SE Asian parasites (Bai et al., 2018). The sensitivity pattern associated to the presence of F1390I is operationally worrying, as it was seen to simultaneously increase parasite resistance against the two available amino-alcohol quinolone based ACTs, mefloquine-artesunate and lumefantrine-artemether. In sub-Saharan Africa, where Artemether-Lumefantrine represents the first line treatment in the majority of the national malaria control programs, the F1390 allele is fortunately still unusual (<1%), contrarily to Asia, where it can reach >50% in Indochina (Dahlstrom et al., 2009) ² .

In Nigeria, a field ex-vivo study attempted to find a link between the pfmrp1 S437A polymorphism and response to artemether (Bustamante et al., 2012), an important question considering the importance of AL in the continent. Unfortunately, the rarity of this allele (3% frequency in this study) precluded any meaningful conclusions. In terms of in vivo approach, ACT day-3 surveillance trials performed in the East and Western borders of Myanmar (Chin and Kayen provinces, respectively) in 2013 did not find associations between D3 positivity and the seven pfmrp1 non-synonymous SNPs detected (Nyunt et al., 2015), not supporting an involvement of this gene in the parasite response to artemisinin type compounds (ART). In contrast, a significant link between the 876V allele and delayed parasite clearance was found in a study conducted among 63 patients in Laiza, Myanmar, treated with DHA-PPQ during 2011–2013 (Lo et al., 2016).

Due to the well-known ART pleiotropic effects and the complex architecture of associated resistance phenotypes, it is likely that other factors—including additional drug transporters—might be involved. The latter is to a certain extent supported by the recent demonstration that the ART-resistance kelch13 mutations lead to a rewiring of the parasite physiology (Mok et al., 2021). Finally, it is important to refer that concerning in vitro phenotype characterization, classical IC₅₀s have been considered insufficient for a proper response profiling to this type of antimalarials (Davis et al., 2020), as they do not generally reflect variation in the main characteristic of artemisinin (partial) resistance, parasite clearance dynamics.

P. falciparum resistance to antifolate drugs, as pyrimethamine and sulfadoxine, is centrally associated with mutations in the parasite endogenous folate pathway, the metabolic target for these drugs. Here, mutations in the pfdhfr (dihydrofolate reductase) and pfdhps (dihydropteroate synthase) genes are critical (Warhurst, 2002), as well as increased copy number events in the GTP cyclohydrolase 1 (pfgch1) gene, coding for the first enzyme of the pathway (Heinberg and Kirkman, 2015). Additional factors have nevertheless been long proposed, in particular with the involvement of transport systems (Wang et al., 1997).

The 3D structure of pfMRP1 and 2 have not been reveald, yet it is assumed that they likely resemble other eukaryotic MRPs (Wilkens, 2015). Bozdech and Ginsburg (2004) have made the early suggestion that the significantly earlier pfmrp2 transcriptional peak—before the emergence of full functional Hb digestions and glutathione synthetase expression—could be related with very different functions, namely the possibility of pfMRP2 acting as an importer of reduced glutathione (GSH) from the RBC, compensating the low activity of the parasite glutathione metabolism during that period of the intra-RBC cycle. This is not an outlandish suggestion. Among eukaryotes, albeit ABC transporters are commonly substrate exporters, a number of importing ones have been described, mainly in evolutionarily more primitive plants (Choi and Ford, 2021), but also in higher mammals, with the human ABCA4 being a prime example (Quazi et al., 2012). In parallel, it is to note that the most well studied P. falciparum ABC transporter, the P- glycoprotein homologue (Pgh) is actually considered to act as an importer towards the food vacuole lúmen (Rohrbach et al., 2006), albeit it has the advantage of being inserted in an internal membrane, which permits the direct access to cytoplasmic ATP for the protein NBDs. Considering the present available data, the notorious functional flexibility of the ABCC proteins, and the fact that the NBDs of a pfMRP importer in the parasite plasma membrane is still not actually exposed to extracellular space, but rather to the parasitophorous vacuole lumen, this possibility should not be discarded.

Under the “importer hypothesis”, it is then possible that pfMRP1 might actually work as a secondary folate transporter towards the parasite, complementing the function of the characterized pfFT1/pfFT2 proteins (Salcedo-Sora et al., 2011). Such configuration could explain the selection of the pfMRP1 1466R SNP, now hypothesized to be related with a more efficient folate importer structure, increasing the intra-cellular competition between this compound and pyrimethamine accessing DHFR (Dahlstrom et al., 2009a)—one is anyway left wondering why this allele is not the dominant one in nature. It would also support the strong MTX decrease in sensitivity in the absence of pfMRP1, as this drug—structurally related with folate—would simply be less transported towards the parasite. Also, and as proposed for pfFT1/pfFT2, it would explain the sensitization effect of probenecid, a long known resistance reversal agent that also affects MRPs (Barrand et al., 1997). Intriguingly, the parasite does not become more resistant to the structurally less folate-like pyrimethamine. In fact, pfMRP1 absence slightly decreases PYR IC₅₀s, as well as sensitivity (and intra-cellular accumulation) to quinolone antimalarials in W2 (Raj et al., 2009) (but not 3D7-like NF54), pointing for a drug exporter function. This is further supported by the decrease in GSH accumulation arguing for a typical MRP exporter activity, the MRP absence making the parasite more vulnerable to direct and indirect (REDOX stress) antimalarial action.

Concerning pfMRP2, and loosely following the Ginsburg/Bozdech suggestion, one could also argue that the disruption of the hepatic merozoite development through pfmrp2 deletion (Rijpma et al., 2016) is related with a resulting lack of import of essential substrates (e.g. co-factors, as it happens with some bacterial ABC importers) obtained from the liver cell during that stage.

In conclusion, part of the available data could be explained by both MRPs working as importers and/or exporters, leading to the rather intriguing possibility of P. falciparum MRP system having at least two main alternative conformational states, associated with markedely different insertion structures in the membrane, and altered functional capacities.