A framework for assessing the risk of resistance for anti-malarials in development

Fortunately, development of resistance to artemisinin, on which are based current first-line therapies for uncomplicated and severe P. falciparum malaria, has been relatively slow. This is partly due to recommendations from the World Health Organization (WHO) that only fixed-dose combinations of artemisinin derivatives with other anti-malarials should be used. It is also because of the relative difficulty of generating artemisinin resistance, which has been reassuringly problematic, even in the laboratory[5]. The first signs of a reduction of the anti-parasitic activity of artemisinins are emerging, with a decrease in parasite clearance time being seen in Cambodia and Thailand, and the first hints of a possible molecular mechanism have been suggested[6–8]. Even if the spread of resistance can be controlled, there will be a need for new classes of anti-malarial drugs. This is a race against time, given that the time taken to develop new medicines often exceeds a decade. Developing new medicines that will withstand resistance is a combination of reducing the possibility that resistant parasites will arise and preventing their transmission. First, finding new molecular classes with a low probability that resistance would arise should be a consequence of more holistic cellular based screening, which has the advantage of potentially selecting for polypharmacology; in other words molecules that hit more than one target. Fitness cost is a key factor in this equation as well: if it is not possible to produce a resistant phenotype with a low fitness cost, then these strains are likely to be replaced by sensitive strains when the drug pressure is relaxed. Second, understanding the role of new inhibitors on transmission is also critical at an early stage, since resistant parasites which are inefficiently transmitted are a far lower threat to global public health concern. Anti-folate resistance had a marked differential effect on transmission which may have accelerated its spread[9]: clearly a molecular class with the opposite trend would be preferable. Third, the spread of resistance can be reduced by combining multiple agents with distinct modes of-action; since any organism resistant to one component of the combination should be eliminated by the other. This raises the barrier for resistance, since the parasite would have to acquire both resistances simultaneously, which is far less likely, assuming the mutation events are stochastic[10]. The combination of artemether-lumefantrine has been protected against the emergence of lumefantrine resistance despite being used to treat over 500 million patients, partly because lumefantrine has never been used as a monotherapy. Fixed dose combinations prevent sub-optimal monotherapy treatment, and will restrict the spread of resistance[10, 11], provided the optimal dose ratio can be deployed. Multiple first-line therapies would also help, but the key is to have medicines which are as different as possible and current strategies would only allow use of artemisinin combination therapies[12, 13]. Putting together these three factors in models to simulate drug resistance emergence in the wild is an important task, but is still at an early stage[14–16].

The lifespan of the next generation of anti-malarials in the face of resistance is a critical question in the debate over malaria elimination and eradication. If through judicious selection of the right molecules and good practice, the time interval from introduction to clinically significant resistance for the next generation of therapy can be put at the level of artemisinin, or even better, it is possible that the disease will be eliminated before clinical resistance occurs. This increases the need to have an early assessment of the risk of developing resistance associated with each compound and each combination. This would not be the single decision criterion, but would be part of the overall framework of deciding which compounds to prioritize. The Medicines for Malaria Venture (MMV) is a product development partnership with the mission to discover, develop, and facilitate delivery of new, effective and affordable anti-malarial medicines. Part of the process of developing new medicines is to have a coherent view on the required properties of new medicines, a process often referred to as developing the “target product profile” or TPP. Since one key aspect of this product profile has to be minimal susceptibility to resistance, it is important to have a standardized framework to quantitatively assess the risk of resistance associated with preclinical anti-malarial compounds. Such a framework, based on in vitro P. falciparum experiments, has been developed and is presented. The experimental strategy is discussed together with the benefits and limitations of such a standard approach and its implications for clinical development.

One of the hallmarks of genetically determined drug-resistance is a shift in the 50% inhibitory concentration (IC₅₀) compared to sensitive strains. The first step is establishing that a new compound is fully active on known multi-drug resistant (MDR) P. falciparum strains by measuring its IC₅₀, which is a more robust and reproducible measure than IC₉₀ or IC₉₉. Known anti-malarial resistance is principally determined by only a handful of genes (Figure2). These include specific molecular targets, such as those for antifolates pfdhps (dihydropteroate synthase) and pfdhfr (dihydrofolate reductase) and electron transport inhibitors pfcytb (cytochrome bc1). In addition there are transporters such as pfcrt (chloroquine) and pfmdr1 (multidrug resistance 1).

Since these MDR strains were adapted to culture and cloned many years ago, and have been passaged since that time, they may not represent the wild-type field Plasmodium parasites. Moreover, only P. falciparum can be stably cultured in the laboratory. It is therefore essential to test compounds against contemporary field isolates of all the human-infecting species, including Plasmodium vivax, Plasmodium ovale and Plasmodium malariae. This gives good background information on variability, and allows confirmation that the compound is active on all species. The analysis of at least 10 P. falciparum and 10 P. vivax isolates appears a minimal requirement, if a 20-fold change in IC50 is used as a cut-off[42].

Resistant parasites can be selected by applying drug pressure in vitro and in vivo[22]. These approaches have successfully identified genes, and sometimes codons involved in naturally occurring resistance[22]. The single nucleotide changes behind these mutations are estimated to occur at a rate of approximately 10^-9 per nucleotide site per replication, similar to other eukaryotes[43–45]. Changes in sensitivity are also caused by gene copy number variations, which occur as often as 10^-4 per nucleotide site per replication[45, 46].

The simplest approach is in vitro selection of resistant parasites, which can be applied on a large number of compounds earlier in drug discovery. P. falciparum intraerythrocytic cultures, with starting inocula ranging from 10⁵ to 10⁹ parasites, are exposed to a concentration of the compound nearing IC₉₀ and monitored during 60 days for recrudescent parasites (Figure3a). The minimal inoculum for resistance (MIR) can be determined and is an indirect measure of the likelihood of a resistant genotype to occur and to be selected in vitro. In addition, the IC₅₀ shift of the resistant mutants as compared to the parental sensitive strain is a measure of the resistance intensity[47–49]. Dd2 has an accelerated resistance to multiple drugs (ARMD) phenotype, possibly due to defective DNA repair mechanisms[47, 50], and is therefore an ideal strain to investigate rare mutational events within the practical limits of in vitro culture volumes. Importantly, MIR from ARMD strains are assumed to be lower because of the general higher mutation rate of these parasites and not because they acquire alternative resistance mechanisms. Moreover, ex vivo results seem to be in good agreement with clinical experience. Atovaquone resistance is generated relatively easily in patients, and in laboratory experiments results either in low MIR or large IC₅₀ increases[3, 4, 47]. Chloroquine and artemisinin derivatives, which escaped resistance clinically for a long time despite widespread use as monotherapy, do not readily produce stable resistant mutants in selection experiments[22].

Mutations that give a selective advantage against drugs often cause a reduction in fitness in its absence[53, 54]. A key fitness parameter is the relative asexual growth rate[55], measured in vitro by monitoring cultures seeded with the derived resistant and the original (sensitive) parental strain at a given ratio. The evolution of the strain ratio can then be monitored by DNA sequencing or qPCR[56–58]. The proportional loss of fitness (C) can be calculated from the change in strain ratio ΔR_n, after n generations, where that value is equal to 1/(1-C) ⁿ [58]. If C ≥ 0 the sensitive strain will outgrow the resistant strain, however, if C<0 this would be major potential risk, suggesting that the strain may also out-compete the original wild type in humans. Of note, such a scenario has already been observed with P. chabaudi chloroquine and pyrimethamine resistant parasites displaying a fitness gain even in the absence of drug pressure[53]. A second key parameter of parasite fitness is its transmission potential, which depends not only on its growth rate but also on its ability to produce viable gametocytes. The effect of resistance can also be investigated on the sexual phase of the parasite cycle, using recently developed in vitro methods[59]. A resistant parasite unable to complete a sexual cycle is of no clinical relevance, as it cannot be transmitted.

In vivo resistance selection experiments allow a more physiological variation of the drug concentration and are more representative of the clinical situation[60]. For instance, the pfcytb Y268C mutation conferring atovaquone resistance in the wild could only be reproduced using a rodent in vivo model[61]. However, in vivo experiments do not allow to control the level of drug pressure or the number of parasite subjected to it and are not ideal to profile and cross-compare a large number of early compounds.

One key to understanding the importance of a resistant phenotype is its molecular characterization. Mutations in genes encoding pumps and transporters are concerning, since they could imply parallel resistance generation to a wide variety of different molecules. Point mutations may enable a link with the molecular target to be established. The genetic changes causing resistance can be identified using tiling arrays, full genome sequencing or linkage analysis and candidate genes can be validated using allelic exchange experiments[62–65]. The full genome sequences of a large number of P. falciparum parasites from diverse locations have already been determined[66]. For those molecules like NITD609 for which molecular markers of resistance have already been identified, it will also be important to examine these sequences and determine whether the target gene is highly polymorphic in these natural populations. This straightforward examination will give advance warning of any regions where resistant parasites might already by present, and easily selected by use of the drug.

Additionally, in vitro experiments performed to characterize the compound speed of action and its activity at various stages of the parasite lifecycle can be repeated using resistant parasites in order to better understand the resistance mode-of-action[59, 67]. Recent advances in imaging also allow monitoring the intraerythrocitic life cycle of the parasites in vitro and in vivo and could similarly be used to compare resistant and sensitive parasites[68].

The development of technologies for assessing resistance means that comprehensive resistance testing should take place at all phases of drug development, including post-marketing pharmacovigilance (Figure4). Ideally, the preclinical studies will have produced resistant strains and identified resistance markers, which are especially useful in the post-marketing surveillance phase. Resistance selection could also be performed for both partners of potential therapeutic combination using single-resistant mutants to evaluate the frequency of double-resistance, which should be equal or inferior to the single-resistance selection frequency. The identification of in vitro resistance markers allows careful monitoring for their potential appearance during clinical trials. Analysis of samples during dose ranging studies for resistance selection could have an impact on dose selection, but presumably only if the changes in the IC₅₀ value were relatively minor. Systematic monitoring of resistance selection in Phase III will consolidate the data on frequency and help to consolidate the understanding of the development of resistance in different geographic and endemicity settings. This will be important information for the regulatory filings, for national malaria control programmes, and for coordinated efforts of resistance monitoring[69]. Regulators have universally recognized the importance of a systematic approach to the characterization of new compounds in regard to their resistance generation potential. The US-FDA recommend in their ‘Guidance for Industry for Malaria drug development’ that “the ability of Plasmodium strains to develop resistance when subjected to drug pressure should be examined in appropriate in vitro and/or in vivo models; this examination should include evaluating the potential for cross-resistance to drugs in the same class or in other classes. If resistance is demonstrated, it is important to identify the mechanism of resistance. Attempts should be made to evaluate the clinical significance of any changes in phenotype (e.g., in vitro susceptibility to the drug) or genotype observed in preclinical studies by correlating such changes with clinical outcome”[70].

Characterization of resistance potential has been mentioned in the registration summaries of all anti-malarials submitted stringent regulatory authorities in the past ten years: Coartem® (artemether-lumefantrine, Novartis) to the US-FDA; Eurartesim (dihydroartemisinin-piperaquine, Sigma-Tau) at the EMA; and more recently Pyramax (pyronaridine-artesunate, Shin Poong Pharmaceuticals) which obtained marketing authorization under the EMA’s article 58)[71–73]. The MMV strategy for resistance characterization will provide this information for the regulators and ensure a complete resistance registration package.

The process presented here integrates information about the potential importance of resistance selection and spreading into the preclinical and clinical development programme for new anti-malarial drugs. At the preclinical level, the experimental framework requires a test cascade of six essential steps covered by three specific experimental procedures. The output of this workflow allows to classify the compound risk profiles as low, elevated or major (Figure1 and3b). This is one of the factors that will be discussed as part of the preclinical candidate selection process, a decision matrix which includes safety margin, and severity of safety signals observed, human predicted potency, speed of killing, effect on other lifecycle stages such as gametocytes and hypnozoites[74]. A major risk suggests that the compound will either face pre-existing resistance or will select new clinically significant mechanisms of resistance. This risk level should prevent further development of the compound, in the absence of a clear mitigating factor linked to major synergy in combination studies. An elevated risk is associated where there is a clear signal, and is an obvious concern. Provided that there are a small number of other elevated risks in the candidate selection matrix these compounds can be progressed, but resistance needs to be continually monitored in clinical studies, and will be a key factor in the selection of dose and combination partner. Risk mitigation is essential, making it important to understand the molecular mechanism, and to have genetic markers to survey its occurrence. A low risk classification suggests that resistance is unlikely to arise, although it is still important to continually monitor for changes in IC₅₀ in the clinical program.

The outcome of resistance studies has several implications influencing decision making in drug development. First, the IC₅₀ as part of the calculation for the prediction of human dose should include the range of potencies observed, and focus on the value from the most resistant parasites. This needs to be continually re-evaluated based on emerging data from primary field isolates. Second, potential cross-resistance amongst partners in drug combinations can be checked using laboratory derived resistant strains. Ideal drug combinations are these that regroup compounds driving opposed or even incompatible resistance mechanisms[75]. Third, the knowledge of in vitro identified mutations conferring resistance and susceptible genes allow monitoring during the large scale clinical trials in Phase III and post-launch, to give an early warning signal for the emergence of clinically significant effects.

The experimental strategy presented here also has limitations. It applies to compounds acting against P. falciparum asexual intraerythrocytic stage only. However, these represent the majority of the parasite biomass. Although the next generation of anti-malarials may include increased focus on the hepatic forms, understanding potential resistance selection at these lifecycle stages will be supported by knowledge of the resistance mechanisms in intraerythrocytic stages. Another limitation is the standardized approach itself. Resistance is a complex phenomenon that will require tailor-made studies. For instance, artemisinin resistance as currently observed does not translate into clear IC₅₀ shift, rendering its study in cross-resistance and resistance selection experiments challenging[5]. However a standardized method does at least allow a common framework for comparing compounds and making development decisions.

The workflow presented here allows to measure the risks of resistance generation, the magnitude of the change in IC50, and the effect on transmission at an early stage in drug discovery. By determining the molecular mechanism of resistance, and identifying potential cross-resistance issues, high-risk chemical series can be de-prioritized, in a time and cost effective manner. The generation of a standard resistance profile of new drugs will supply common information to regulators and national control bodies to facilitate decision-making. This information will also allow for the focused monitoring of compounds after registration. This strategy will contribute to developing long-lasting anti-malarial drugs by identifying and monitoring robust compounds during their development.