Genetic markers of artemisinin resistance in Plasmodium spp. parasites.

The vast majority of malaria patients worldwide are currently treated with combination therapy comprising one of the artemisinin family of drugs, characterised by rapid action and short plasma half-life, co-formulated with a longer-lasting drug from the amino aryl-alcohol or quinoline families. There is now a widely perceived threat to treatment efficacy, as reduced susceptibility to rapid artemisinin clearance in vivo has become prevalent among populations of Plasmodium falciparum in the Greater Mekong subregion since 2008. In vitro and in vivo drug selection studies, heterologous cell expression experiments and genetic epidemiology have identified many candidate markers of reduced ring-stage susceptibility to artemisinin. Certain variants of the P. falciparum pfk13 gene, which encodes a kelch domain protein implicated in the unfolded protein response, are strongly associated with slow parasite clearance by artemisinin in the Mekong subregion. However, anomalies in the epidemiological association of pfk13 variants with true treatment failure in vivo and the curious cell-cycle stage specificity of this phenotype in vitro warrant exploration in some depth. Taken together, available data suggest that the emergence of P. falciparum expressing K13 variants has not yet precipitated a public health emergency. Alternative candidate markers of artemisinin susceptibility are also described, as K13-independent treatment failure has been observed in African P. falciparum and in the rodent malaria parasite Plasmodium chabaudi.

A new artemisinin susceptibility phenotype arises in Asia

Noedl et al. [1] first described microscopically detectable Plasmodium falciparum parasites persisting in the peripheral blood of artesunate-treated Cambodian malaria patients in 2008. This was soon followed by more extensive observations from the randomised study of Dondorp et al. [2]. Importantly, the latter study demonstrated that slow parasite clearance was observed in malaria patients from western Cambodia irrespective of whether they had been treated with monotherapy, or with artemisinin-based combination therapy (ACT), but this was not observed in patients from the Thailand–Myanmar border to the northwest. Extensive multicentre studies led by the UK-funded TRAC consortium then provided important evidence, just a few years later, that the slow-clearance phenotype was spreading through P. falciparum populations in the Greater Mekong subregion (GMS) in association with certain variants of the pfk13 gene, but that in parallel African studies any treatment failures seen were not associated with this genetic marker [3].

Identification of pfk13 was the culmination of a search for a marker or markers to better identify P. falciparum with reduced susceptibility to artemisinin, a search that had been hampered by the lack of an in vitro correlate. Curiously, the in vivo slow-clearance phenotype of P. falciparum, spreading through the GMS, was not consistently reflected in estimates of the 50% effective concentration (EC50) of dihydroartemisinin (DHA, the dominant parasiticidal artemisinin metabolite in vivo) against parasites cultured in vitro [2,4]. Only after the Pasteur Institute teams in Phnom Penh and Paris had devised an in vitro DHA susceptibility test specific for the earliest blood-stage parasites, the ring-stage survival assay (RSA) [5], Ariey et al. [6] were then able to demonstrate that certain mutations in the pfk13 locus correlated with the ex vivo parasite phenotype in the RSA. This key locus was identified as a candidate for testing not by genome-wide association studies (GWASs), an approach that promised much [7] but delivered little [8] in terms of identifying specific markers of artemisinin susceptibility, but through 5 years of painstaking drug selection in the laboratory [6]. Pfk13 was one of a few loci in the genome of the culture-adapted Tanzanian P. falciparum isolate F32 that, during this extensive selection process, had accumulated mutations in concert with decreasing DHA susceptibility as measured by the RSA. Alone among these few loci, mutations in pfk13 were also found to correlate with ex vivo DHA susceptibility of Cambodian parasites in the RSA, although the exact mutation identified in mutant F32 parasites has still not been encountered in any wild P. falciparum isolates to date [9]. The identification of key variants of pfk13 that correlate both with slow clearance in patients receiving artesunate monotherapy [3] and with ex vivo RSA data [6] has enabled subsequent research on pfk13. This includes:

studies of the distribution of polymorphisms in the field, particularly in Africa [9-13];

groundbreaking genome-editing studies in vitro demonstrating that single-nucleotide changes in pfk13, mimicking GMS variants, alter P. falciparum DHA susceptibility in vitro [14,15];

GWAS demonstrating the importance of genetic background, including novel alleles at several loci, in the K13-mediated phenotype [16].

The perceived threat to artemisinin-based malaria therapeutics

There is considerable concern that, faced with the spread of pfk13 variants to the wider GMS, the proportion of slow-clearing phenotypes is increasing, thus posing an immediate threat to ACT efficacy within the region and more globally in the future [17-20]. In fact, some authors have called on the WHO to declare artemisinin ‘resistance’ to be a ‘‘public health emergency of international concern’ [18]. However, a careful overview of data from the GMS and beyond provides several arguments against this emotive response. These include the following: In fact, considering all these factors, and the inability of variant pfk13 parasites to survive 48 h of artemisinin in vitro (see below), it seems prudent to cease the practice of calling pfk13 a marker of artemisinin resistance, but rather see it as an important modulator of susceptibility that represents a parasite adaptation to the short treatment regimens deployed against P. falciparum infections in vivo. In the next section, arguments based on the parasite cell cycle will be used to support this unorthodox suggestion, and lend further support to the strategy of extended artemisinin regimens to radically cure P. falciparum infections [25].

It is postulated that a single lineage of multidrug-resistant variant K13 parasites, dubbed PfPailin, underly this proposed emergency. However, despite the claim that ‘transnational spread of this single fit multidrug-resistant malaria parasite lineage is of international concern’ [20], these authors' own data show this not to be the case. To quote: ‘Parasites with the same C580Y (pfk13) haplotype were not clonal since there was significant diversity in microsatellites elsewhere in the genome and in the polymorphic loci in the msp1, msp2, and glurp genes that showed an overall mean He of 0.529 (SE 0.030) for three loci combined.’ [19]. This indicates a relatively high level of genetic recombination among a variety of parasite genomes, with the pfk13 variant locus (and one other gene of interest, pfplasmepsin2) showing a selective sweep across these various genomes. Therefore, there is no single multidrug-resistant lineage and no ‘super-parasite’ emergency.

The association between pfk13 carriage and either slow-clearance phenotypes measured 3 days after treatment or parasite recrudescence after 28 or 42 days of active follow-up in drug efficacy studies is elusive. Firstly, parasite persistence and slow clearance in African studies are not associated with the pfk13 polymorphism [10,21]. Secondly, K13 variants do not determine parasite recurrence after treatment with the ACT DHA–piperaquine (DP), as recently reported for three sites in Vietnam [17]. The Gia Lai site, in which 67% of P. falciparum malaria patients carried the C580Y variant of pfk13, showed 89% treatment success (‘adequate clinical and parasitological response’, ACPR), whereas the ACPR among patients enrolled at Ninh Thuan was 78% (C580Y prevalence rate 4%), and among patients enrolled at Binh Phuoc, the ACPR was a very disappointing 56% (C580Y prevalence rate 73%). In fact, the best interpretation of this study is that it is falling partner drug efficacy that has led to the low ACPR in some sites [17]. Furthermore, in Myanmar, despite a prevalence rate of 25–65% of the C580Y variant pfk13 in 2012 and 2013 in Kawthaung and Myawaddy, 3 days of standard ACT delivered ≥96% ACPR [22].

P. falciparum is rapidly disappearing from the GMS. Despite an increasing proportion of P. falciparum isolates in the region carrying multiple determinants of reduced susceptibility to ACT, the overall numbers of falciparum malaria cases have fallen dramatically; as an example, between 2007 and 2015 malaria hospital admissions fell from 25 to 2 per 100 000 (92% fall) and deaths from 0.15 to 0.05 per 100 000 (67% fall) in Thailand, and only 42% of infections are P. falciparum (WHO World Malaria Report 2016; http://www.who.int/malaria/publications/world-malaria-report-2016/en/). Thus, the malaria control effort in this region needs to be applauded and continued. If there is any public health malaria emergency of concern, it is surely in sub-Saharan Africa, where, in Burkina Faso in 2015, ∼2000 people were hospitalised for malaria, and 30 malaria deaths were estimated to have occurred, per 100 000 population (ibid.).

P. falciparum harbouring variant K13 alleles, and/or demonstrating slow clearance at day 3, are clearly susceptible to extended-duration artemisinin-based treatment, as an ACPR of more than 90% has been achieved in the GMS with 6 days of artesunate monotherapy [23,24] or 3 days of artesunate monotherapy followed by 3 days of ACT [3]. Thus, regimens such as two sequential ACTs with different partner drugs have been postulated as a rational response to signs of falling 3-day ACT efficacy, and a protocol has been developed to evaluate such regimes before ACT failure becomes a serious problem [25].

Variants of pfk13 and the parasite cell cycle in vitro

There is very good evidence from laboratory studies of P. falciparum isolates grown in culture that artemisinin susceptibility varies greatly across the cell cycle [26,27]. The patterns as currently understood for wild-type parasites are represented in cartoon form in Figure 1A. Although all stages are susceptible to artemisinin, a small proportion of the least susceptible stages may survive a single pulse of artemisinin in vitro. Figure 1B illustrates some of the findings for Cambodian-origin parasite clones harbouring K13 variants [27]. Most notably, the exquisite susceptibility of wild-type ring-stage trophozoites is ablated by the pfk13 variant, and this fully resistant stage is extended in duration, thus minimising the proportion of the life cycle spent at the more susceptible stages [mature trophozoite (MT) and early schizont (ES)]. This simplified conceptualisation of the impact of artemisinin on the parasite at different points in the life cycle can be extrapolated, with some caution, to consider the impact of a standard artesunate monotherapy regimen (3 days) on a typical asynchronous natural infection of P. falciparum. Presented as a pictorial model in Figure 2, this provides interesting inferences of what may happen in a treated malaria patient with all stages of the parasite life cycle present (although some stages will be sequestered in the tissues, these are drug-exposed).

Figure 1.

Asexual life-cycle stage specificity of in vitro artemisinin susceptibility of P. falciparum parasites with or without variant pfk13 alleles.

Cartoon figures depict in a stylised manner 5 of the morphological stages recognised in fixed thin films of cultured parasites as labelled [ERT, LRT, MT, ES, and mature schizont (MS)], with the number of repeat cartoons approximating the observed duration of that stage during the 48-h life cycle. Y-axis reflects estimated stage-specific inhibition/killing of a short (3–4 h) pulse of 700 nM DHA from published sources. All wild-type parasite stages are killed by artemisinin, but the short exposure to drug in vivo due to rapid metabolism and extinction of DHA may allow a small proportion of the least susceptible stages to survive. (A) Standard DHA-sensitive cultured laboratory parasite line, as shown for 3D7 by Klonis et al. [26]. All stages are susceptible to artemisinin killing; ERT briefly exhibits exquisite susceptibility, while LRT is the least susceptible stage. (B) Slow-clearing parasite line of Cambodian origin, exemplified by ARC08-88 (clone 4G) described by Hott et al. [27]. It is notable that other artemisinin-tolerant clones described by these authors displayed a shortened in vitro life cycle of under 40 h. Abbreviations: DHA: dihydroartemisinin; h: hours post-invasion in a synchronised ring-stage culture; full susc: development stage is fully susceptible to a 3–4 h pulse of 700 nM DHA, and no further growth is detected after the drug pulse; part susc: partly susceptible to the 700 nM DHA pulse, such that some continued growth occurs after the drug is washed away.

Figure 2.

Pictorial model describing the likely impact of non-synchrony in natural infections: three doses of daily artemisinin leave survivors in vivo.

Cartoon figures representing ERT, LRT, MT, ES and MS are shown in Figure 1. Two hypothetical malaria patients infected with a single, but asynchronous, clone of P. falciparum are depicted under 3 days of artesunate (ART) monotherapy [1,2]. All five developmental periods (represented by the five different cartoon shapes from Figure 1) are assumed to be present at the moment of first treatment. It is not assumed that these are present in equal proportions. Parasites killed or prevented from developing by the current or most recent ART dose are depicted as pale grey ghosts. Viable parasites are presented in full colour. Mid-grey parasites with orange border are drug-affected and significantly reduced in number; these may die or may survive and progress to the next stage. In this figure, altered temporal development as depicted in Figure 1 and seen in vitro has not been depicted, as this is not proved in vivo. Day 3 count: key follow-up time point to identify slow-clearing parasites for in vivo drug trials.

Firstly, in wild-type parasites, a small proportion of parasites at the LRT (late ring-stage trophozoite) stage at the time of the first artesunate dose may survive (dark grey shading), particularly considering that the terminal elimination half-life of DHA (administered as DP) is less than 2 h [28]. These will be schizonts by the time of the next administration, but if this is not exactly spaced by 24 h, some of this cohort of parasites may be mature enough to also have a chance of escaping DHA killing. The third dose will find any remnants of this cohort as LRT again, and hence, some may yet survive. Interestingly, at the day 3 follow-up time point, any survivors will be schizonts and fully sequestered, so peripheral blood microscopy will indicate the patient is fully cured. In this way, the drug treatment is seen to effectively synchronise any survivors into a co-ordinated cohort. Is this an argument for a day 4 time point?

Secondly, in the K13 variant parasites, the benefit of the mutation is entirely due to survival of that proportion of parasites at the most resistant stage — ERT (early ring-stage trophozoite). The pictorial model suggests that, again, day 3 is not the best time point for peripheral microscopy, as survivors will be late trophozoites or schizonts sequestered in the periphery. The second artesunate dose will find relatively vulnerable MT and ES among the resistant survivors, but the third dose again would encounter highly resistant ERT in the resistant cohort (Figure 2, bottom panel, top row). If the perturbations of cell cycle dynamics described by Hott et al. [27] hold in vivo, it may be that resistant rings are present for longer periods during the cell cycle, which may increase the overall survival.

The fact that pfk13-mediated phenotypes are advantageous solely at the early ring stage of the cell cycle (Figure 2) draws attention to the remarkable spike in susceptibility at this stage in wild-type parasites (Figure 1A). It is widely considered that the endoperoxide bridge in artemisinin is strongly activated by haem metabolism in the food vacuole, leading to parasite killing [29], a process that barely occurs during the first few hours of the life cycle, so it is unclear how activation of artemisinin, so lethal to parasites in these first few hours, occurs [18,26]. One plausible hypothesis is that a specific parasite cellular process in the first few hours post-invasion provides a finite burst of activated artemisinin; variant K13 renders ERT resistant to artemisinin simply by preventing this specific event. As haemozoin formation in the growing food vacuole proceeds, artesunate is activated for the rest of intra-erythrocytic development, and thus, K13 variant parasites are fully susceptible to prolonged exposure to artemisinin, as clearly seen in vitro [2,4,5] and in vivo [3,22,23].

K13-independent artemisinin susceptibility phenotypes in Plasmodium spp

As we have seen, painstaking drug selection in the laboratory was the experimental path that led to identification of the pfk13 locus as an important modulator of artemisinin susceptibility. The mode of action of artesunate in killing malaria parasites is almost certainly multifactorial, as the activated endoperoxide moiety causes widespread non-specific damage to intracellular proteins [26,29]. Thus, artemisinin selection is expected to identify other mechanisms that reduce susceptibility, and indeed, this has been the case. The best example is the work of Hunt, Cravo and colleagues in the rodent parasite Plasmodium chabaudi, in which variants in the C-terminus of pcubp1, encoding a ubiquitin carboxyl-terminal hydrolase, were fully validated in vivo by whole-genome sequencing and through reverse genetics in a genetic cross, as a true determinant of artemisinin and chloroquine resistance [30,31]. Subsequent in vitro and in vivo studies have supported, but not yet validated, variants of the orthologous P. falciparum locus pfubp1 as determinants of artemisinin susceptibility [32,33]. An additional locus identified as contributing to high-level artemisinin resistance in the P. chabaudi studies, pcap2mu, encoding a putative adapter protein subunit involved in clathrin-mediated endocytosis [34], has also been partially validated as a key modulator of ACT treatment outcome in P. falciparum in vivo [33] and of DHA susceptibility by generating transgenic parasites in vitro [35]. Further studies of these candidate molecular markers for artemisinin susceptibility, including any yet to be identified, are keenly awaited. Certainly, there is no doubt that K13-independent artemisinin susceptibility phenotypes have arisen [10,36], and may continue to arise in the future, under universal selection pressure from ACT deployment worldwide. This threat makes it imperative that strategies to preserve the efficacy of our current regimens are vigorously pursued in the short term, as we await the next generation of antimalarial drugs [25].

Mutations in pfk13 do not explain all artemisinin treatment failures.

The mechanism of action of artemisinin is highly stage-specific.

K13 mutations reduce parasite susceptibility to artemisinin by rendering a single stage, very young trophozoites, resistant.

Increasing the duration of artemisinin exposure overcomes K13-mediated parasite phenotypes.