## Main

Malaria is a human tragedy: by one estimate it may have killed half of all the people who ever lived1 and currently claims the lives of hundreds of thousands of young children every year2. Stark as the numbers are, these figures understate the loss of developmental potential for malaria-endemic countries, the impact on their health-care systems and on their economic growth3,4. However, the past decade has seen a renaissance in the research and development for malaria medicines5,6,7. The critical factors at play here are an increased collaboration between academia and industry and a focus on finding new chemotypes by screening directly against the parasite. A call for the eradication of malaria by the Bill and Melinda Gates Foundation in 2007 (Ref. 8) identified new priorities, such as transmission blocking and chemoprotection in vulnerable populations9, and new target candidate profiles10. The past decade has seen a transformation of the portfolio of malaria medicines, with a dozen or so new chemical entities entering clinical development and breakthroughs in translational research that have enabled human proof-of-concept experiments to be performed rapidly11 (Figs 1,2). Not only have new classes of molecules been discovered, but so have new ways of collaborating and methods to accelerate drug development. However, emerging drug resistance threatens all this success. Artemisinin-based combination therapies (ACTs) are used worldwide as the first line of treatments12, but patients with resistance to artemisinin and its partner drugs are now being routinely identified in Southeast Asia13,14,15,16,17,18. However optimistic the pipeline appears, there is still a need for new chemical agents, especially to prevent relapse of Plasmodium vivax infection and to protect vulnerable populations, especially in early pregnancy. Eradication of malaria will take many decades: the challenge is to continually find innovative ways of targeting the parasite. Paradoxically, the lack of commercial incentives for new malaria medicines has a silver lining: it has created a niche in which academia and industry can continue to work together to pioneer new models of collaboration and openness.

The announcement of the United Nations' Millennium Development Goals19,20 highlighted the eradication of malaria as a key target. The goal outlined in this announcement is to decrease the incidence of malaria by 2015. In 2000, malaria was responsible for 350 million to 500 million infections and more than 1 million deaths, particularly in children under the age of 5 years and in expectant mothers21. Progress in the first decade of the new millennium to reduce these frightening statistics was steady but slow, with 4% year-on-year reductions in deaths driven by the roll out of insecticide-treated bed nets22 and fixed-dose ACTs12. In 2007, the Bill and Melinda Gates Foundation, with the support of the Director General of the WHO (World Health Organization), announced a campaign to drive the eradication of malaria. The advantages of such a call were obvious; the elimination of malaria would have tremendous benefits for public health. However, gains made against malaria are fragile, and whenever a country has decreased its public health focus and funding on malaria control the disease prevalence has climbed dramatically23.

To eradicate malaria, achievable milestones must be set. The publication of the Global Malaria Action Plan (GMAP) by the WHO24 set some of those milestones; for example, a tenfold reduction in malaria incidence and deaths by 2030 (compared with 2015). The first GMAP was published in 2008 and covered the period until 2015. More recently, input and consultations are being sought from experts and regions for GMAP2. Both GMAP2 and the Global Technical Strategy for Malaria (co-ordinated by the WHO) will cover the 10 years between 2016 and 2025. All these plans and proposals call for a transformation of the malaria landscape. That is, even though substantial progress has been made over the past decade with the number of deaths from malaria falling by 4% per year, to achieve a 90% reduction by 2030 we need to accelerate progress to as much as a 10–16% fall each year24. As a public health goal, this is audacious and is clearly not a call to business as usual. Along with the widespread deployment of insecticide-treated bed nets, the discovery of new malaria medicines is a key pillar in reducing malaria-associated deaths by this margin Moreover, in the context of malaria eradication, despite considerable progress towards vaccination, no effective vaccines capable of protecting a sufficient proportion of the population to elicit herd immunity are on the horizon (Box 1).

Since 2000, malaria-associated mortality has been reduced by more than 50%, but emerging drug- and insecticide-resistance continues to pose a major threat2. Insecticides continue to be at the front line, both as larvicides and as part of insecticide-treated bed nets. Although tremendous gains have been made as a result of wider distributions of bed nets22, new classes of insecticides are urgently needed to combat the emergence of resistant mosquito strains. Currently, only pyrethroid is approved for use with bed nets, and resistance among malaria vectors is increasing25.

Medicines have also had a tremendous impact on reducing malaria-associated mortality, and ACTs have been used to treat hundreds of millions of patients, following the isolation of the active ingredient, artemisinin, by Chinese researchers. Here, the landscape has changed dramatically in the past 5 years, and five types of fixed-dose ACTs have been approved by either Stringent regulatory authorities or prequalified for purchase by the WHO (Box 2; Figs 1,2; Table 1).

These successes, however, are no grounds for complacency. We need to protect the medicines we have by ensuring correct deployment and continual vigilance to stop the production and distribution of counterfeit medicines. Even with fixed-dose combinations there remains a risk of resistance emerging, and therefore a need for new medicines. New molecules should shorten the duration of treatment and increase compliance, and also prevent the transmission of the parasite back to the insect vector. If the parasite reservoir is going to be eliminated, medicines have to demonstrate a level of safety such that they can be given to people who are infected with extremely low levels of parasite and are asymptomatic. New molecules that can either prevent relapses from P. vivax or Plasmodium ovale infection or that can provide chemoprotection for vulnerable populations, including expectant mothers, are also needed. This need for protection will only increase as the malaria eradication agenda proceeds. People in areas that are currently disease-endemic will have fewer infections, will lose their partial immune protection and will need chemoprotection when entering the remaining disease-endemic areas. In this article, we summarize the discovery of new classes of medicines, illustrated with several case studies, to underline the progress made and also the challenges that remain in treating and preventing malaria.

## Eradication requires different types of drugs

In 2007, the global agenda for malaria eradication brought together several themes to describe the perfect medicine for treating malaria, termed the SERCaP (single exposure radical cure and prophylaxis)8. However, it soon became clear that no single chemical entity could meet these needs and therefore the ideal medicine would be a combination of two or more ingredients, each of which meets a number of different target candidate profiles10. The first question is how to treat malaria with a single exposure to a drug: the current gold standard for treating malaria is a 3-day course of an ACT. Moving to a single exposure would mean that a health worker could directly observe the administration, and this could provide a critical advantage in a malaria eradication campaign. Currently, one major concern in the field is that the patient frequently does not take the full course of treatment; for example, if the symptoms are resolved rapidly, the danger is that a mother will save the remainder of the treatment for the next time her child is infected. Exposing parasites to subtherapeutic doses leads to more treatment failures and recrudescence of the original malaria parasites, which then encourages the emergence of resistant strains.

A single-exposure cure places a very high barrier for any new molecule: a hyperparasitaemic patient may carry as many as 1 × 1012 parasites in the circulation. Clinically, an acceptable cure rate would result in >95% of treated patients clearing the initial infection, which can be achieved with currently available ACTs. Even though new molecules would be used in combinations (to protect against the emergence of resistance), each active ingredient should still ideally be capable of killing a large proportion, if not all, of the parasites on its own. Medicines such as artesunate are capable of rapidly killing the parasite, reducing the parasite load by as much as 10,000-fold over a single 48-hour life cycle26. Nevertheless, such medicines would need to be active for three or four parasite life cycles to ensure a complete cure27. New medicines active against the blood stages would therefore need both a 'fast kill' profile and be able to maintain a plasma concentration above the minimal inhibitory concentration (MIC) for more than 1 week. Longer duration of action could be also useful in some parts of Africa to provide post-treatment protection, in areas where a child can be bitten almost every night. Five years ago, conventional wisdom dictated that the choice was between short-acting and fast killers, such as artemisinin derivatives, or long-acting drugs with relatively slower parasite reduction rates. It is encouraging to note that many of the new generation of antimalarials are both fast-acting and can maintain active plasma concentrations for more than 1 week. This reflects the increased focus on attempting to increase the in vivo half-life of compounds in the early stages of the drug discovery process.

Another drug class needed in the ongoing battle against malaria is the chemoprotectants. Chemoprotectants should be safe and effective drugs to prevent infection, particularly in vulnerable populations such as expectant mothers and small children, and all residents in areas where continuous, intense transmission is absent and partial malaria immunity does not develop (also known as seasonal chemoprevention). Until recently, the sulfadoxine–pyrimethamine combination has been used for this purpose, but with the emergence of resistance against this combination28,29 chemoprevention is threatened30. New medicines for chemoprevention are therefore needed, and molecules that have activity against the liver schizont stage of the parasite are good chemoprotectants. However, the vulnerability of the target group means that safety and efficacy of such medicines must be first demonstrated in less vulnerable populations. The goal in this strategy is not so much to design new compounds that are safe for chemoprotection in these populations, but instead to accelerate reproductive and juvenile toxicology studies, including for approved drugs, and deprioritize compounds for which significant concerns are raised.

Eradication requires completely different approaches to simple control. One key intervention is to block the transmission cycle; that is, stopping parasite-carrying humans from infecting the mosquitoes. Only the sexual stages of the parasite survive after being taken up in the blood meal of the mosquito when she feeds. Therefore, a medicine that can selectively kill the sexual stages or prevent the completion of their development in infected mosquitoes would halt malarial transmission. Although it is possible to think of new medicines that only target transmission, many of the new compounds currently in development both block transmission and kill the blood stages. In 2012, the WHO recommended that adding a single dose of primaquine (0.25 mg per kg) to current treatment regimens would help block transmission31. The next generation of medicines could have this transmission-blocking component already built in; some of the current candidates do provide partial inhibition of transmission in standard membrane feeding assays. The second requirement is that medicines should be usable in mass screen-and-treat campaigns to remove the parasite even from patients who are asymptomatic. The major difference in such usage of new antimalarials compared with conventional treatment is that these medicines must demonstrate exceptional safety, and therefore larger investments in Phase IV safety studies will be needed.

## Treating relapses from P. vivax and P. ovale

Although 90% of global malaria fatalities are caused by Plasmodium falciparum, up to 2.5 billion people are at risk from P. vivax, which is mostly prevalent in South America and Asia32. P. vivax has an additional complication in its life cycle: dormancy. Forms of P. vivax can remain dormant in the human host from a few weeks to several years, and then re-emerge and cause a new infection32. Radical cure means effectively clearing the body of parasites, including those that are dormant. The target candidate profile for this type of malaria (which we have called TCP3b)10 is based on finding a molecule better than primaquine. Fourteen days of primaquine treatment is sufficient to completely eradicate P. vivax in the body. However, primaquine has an Achilles heel: in patients who are deficient in glucose 6-phosphate 1-dehydrogenase (G6PD), primaquine causes a significant increase in the risk of haemolysis33. G6PD deficiency segregates with protection against malaria. New agents will need to be as effective as primaquine but with a shorter regimen and preferably an improved safety profile.

In 2014, the completion of a clinical study of a new 8-aminoquinoline, called tafenoquine, provided a breakthrough in this regard. Tafenoquine was discovered by researchers at the Walter Reed Army Institute of Research, USA, and developed by GlaxoSmithKline with support from the Medicines for Malaria Venture (MMV). The clinical study showed that tafenoquine could have the same therapeutic effect after a single dose as 14 days of primaquine (but the study was not powered to show this statistically)34,35. The hope is that tafenoquine will be submitted to the regulatory authorities in 2016–2017. However, it still has the haemolytic liabilities of primaquine and will require a point-of-care G6PD diagnostic test; thus, there is room for improvement. New approaches to eliminate the dormant liver stages or hypnozoites require a better understanding of their basic biology. Moreover, the mechanism of action of primaquine is unknown, so the only way to find new drugs is to directly study the effect that compounds have on relapse, as discussed under phenotypic screening below.

It is an interesting paradox that although many individuals use chemoprophylaxis against malaria as a personal safeguard, this is frequently downplayed as an approach in public health. Recent data on the use of sulfadoxine-pyrimethamine plus amodiaquine in the Sahel region helped bring more perspective in this regard. Monthly administration of chemoprotection for the 4 months of the transmission season is extremely inexpensive — less than US$0.60 per child — and is highly effective126. This sets a very high hurdle for a future vaccine in terms of cost effectiveness. To expand chemoprotection there are three main populations that need to be considered. First, new molecules need to be developed that can be used in eastern Africa, where sulfadoxine–pyrimethamine resistance could limit the usefulness of the sulfadoxine–pyrimethamine plus amodiaquine combination36. Second, to protect the entire population, new medicines suitable for use in pregnant women are also a priority, especially because pregnancy increases the risk of malaria infection37. However, African women often do not report pregnancy in the first trimester; therefore, any drug that is given to the general population should ideally be known to be safe in pregnancy. To help address this issue, reproductive toxicology evaluation of new medicines should be accelerated. Typically, reproductive toxicology evaluation is performed during the Phase II studies of a new chemical entity. Currently, at MMV this evaluation is being performed in parallel with Phase I studies, but in the future it should be conducted as part of the preclinical work-up. The third and final group that needs protection is the non-immune travellers. Traditionally, these individuals have not been part of the public health agenda. However, as malaria starts to be eliminated from regions of the world, new non-immune populations will begin to emerge, and mobility becomes a major risk factor. One country with this dichotomy is Zambia, where the disease is beginning to be controlled in many parts in the south of the country, but in the north there are areas where malaria incidence in children aged 5 years and under is greater than 15%38. New chemoprotective regimens are needed that can protect workers from the south of Zambia (or even from other essentially malaria-free zones) travelling to the north. As the pipeline of new medicines to treat malaria grows and strengthens, the pressure to increase the focus on new types of chemoprotection will increase. ## Strategies for developing new medicines Improving on existing success: expanding the medicinal arsenal. The gold standard of malaria treatment is the ACT. This is a fixed-dose combination of an artemisinin derivative (artesunate, artemether or dihydroartemisinin) and a 4-aminoquinoline or aminoalcohol, compounds that trace their ancestry back through chloroquine and eventually to quinine (which was first used in Europe in the seventeenth century) (Fig. 3). Given that artemisinin endoperoxides and their derivatives, such as artesunate, are the mainstay of current therapy, these endoperoxides are logically good starting points for drug discovery. However, one of the biggest concerns with artesunate is that the raw material is derived from plants and there is an 18-month lead time between demand for this material and its supply. This lag in demand and supply has sometimes caused catastrophic changes in price: artesunate prices have varied as much as fourfold in the past decade. In 2002, the MMV in collaboration with the University of Nebraska, USA, Monash University, Australia, and the Swiss Tropical Institute (now known as the Swiss Tropical and Public Health Institute) began a collaboration to attempt to make a fully synthetic endoperoxide molecule that was as effective as artesunate and also as affordable as artesunate at its best price ($400 per kg).

Although there has been much progress in simplifying the total synthesis of endoperoxide antimalarials42,43,44, we are still a long way from achieving a cost-competitive total synthesis. The result of the initial work was OZ277 (Fig. 4), developed through a partnership between the MMV and Ranbaxy, and subsequently called Rbx11160 or arterolane, the synthesis of which was efficient, with a cost of $800 per kg, less than artesunate at its peak price. The activity of arterolane in clinical studies was in the same order of magnitude of activity as the marketed artemether–lumefantrine combination (Coartem; Novartis)45. Ranbaxy continued with the development of arterolane, in partnership with the Department of Science and Technology of the Government of India, combining arterolane with the 4-aminoquinoline piperaquine. A limited Phase III programme has led to the submission and eventual approval of arterolane–piperaquine (under the trade name Synriam) in India in 2013, followed by approval in seven African Nations in 2014, although the medicine has yet to be approved by a stringent regulatory agency or prequalified by the WHO. At 120 Indian Rupees ($2), Synriam is cheaper than many of the marketed ACTs, although the price of these may fall with the anticipated marketing of Sanofi's new, semisynthetic artemisinin. The rapid approval of Synriam based on a single study, and the fact that in the first year the drug was used to treat almost one million patients in India, highlights another issue in malaria drug development: the involvement of local companies can significantly enhance the impact of a medicine.

The original team that developed OZ277 continued with the discovery and development of other new synthetic endoperoxides. At this time the Phase IIa results for OZ277 were available and there were already suggestions that the next-generation compound could solve the issue of artesunate price fluctuations and also have a much longer half-life — with plasma exposures above the MIC for more than 1 week — and therefore could be part of a shorter course of treatment or a single-dose cure. The resultant molecule, OZ439 (also known as artefenomel), has now progressed to Phase IIb combination studies, whereby it is being tested in combination with piperaquine, registered at ClinicalTrials.gov as NCT02083380. The challenges for both OZ439 and OZ277 include the extent to which they can differentiate themselves from their artemisinin heritage. First, artemisinin resistance — or, more correctly, artemisinin insensitivity — is a substantial concern. For example, in patients in western Cambodia there is a clear change in the kinetics of parasite killing but an unclear effect on the half-maximal inhibitory concentration (IC50) for artesunate46,47. Analysis of the parasites has suggested that the cause of this insensitivity is in the early ring stages of the parasite48 and that the kelch13 gene has a role16. The first Phase IIa clinical study of OZ439 showed no strong correlation between kelch genotypes and the speed of parasite killing, and the in vitro data on the early ring stages confirmed that OZ439 is fully active on artemisinin-resistant strains (J. Möhrle and D. Menard, personal communication); more data from ongoing clinical studies will shed further light on this issue. Additionally, one Achilles heel for artemisinin derivatives is that they have been associated with reproductive toxicology: their actions in the embryonic yolk sac in rats have suggested that there might be a risk of spontaneous abortions in the first trimester49. There are no clinical data on the use of OZ277 or OZ439 in pregnancy, but both OZ277 and OZ439 show a safety margin almost a hundred times higher than artemisinin derivatives in preclinical embryo–fetal developmental safety studies.

Improved 4-aminoquinolines are also being pursued. Ferroquine is a good example of such a candidate39,40 (Fig. 3); it is a ferrocene moiety coupled to a 4-aminoquinoline and could theoretically kill the parasite even faster owing to the capacity of ferrocene to generate parasite-killing radicals41. However, improving upon 4-aminoquinolines will be difficult because historical data on 4-aminoquinolines suggest that as monotherapies they kill as fast as artesunate. Ferroquine is being developed by Sanofi in combination with OZ439 (Fig. 4; Table 2). All new compounds must be fully active against the existing collection of primary Plasmodium strains, and, promisingly, ferroquine is active against chloroquine-, amodiaquine- and mefloquine-resistant strains39.

The global portfolio of antimalarials contains two other interesting endoperoxides: artemisone (also known as BAY-44-9585)50, which is a derivative of artesunate at the 10-position and is a potential replacement for artesunate, and the tetraxoane TDD E209, the next generation compound following RKA182, which was discovered by researchers at the University of Liverpool, UK51. The challenge for both of these newcomers is similar to the new endoperoxides currently in clinical trials: to demonstrate a significant advantage in one of the following: safety, compliance, activity against resistant strains or reduced cost compared with other compounds in development.

The 4-aminoquinoline scaffold is relatively old, and few other compounds have been developed based on this scaffold. AQ13 is one example52 (Fig. 3; Table 2), and many compounds have been proposed to have dual activity or to be used as fusion compounds53. Here, caution is advised, as optimizing two activities while retaining safety is not a simple task. The primary issue with all 4-aminoquinolines, even chloroquine, is their potential for prolongation of the cardiac QTc interval, which may be associated with torsades de pointes and cardiac sequelae54. Interestingly, it seems that all the 4-aminoquinolines have the potential to prolong the QTc interval, but in a recent review of 16,382 patients taking piperaquine, no drug-related sequelae have been observed, indicating that these compounds are probably among the exceptions to the strong correlation between QTc and torsades de pointes55.

Better by design: the power of structure to improve a molecule. The classical drug discovery paradigm of the past 20 years, fuelled by the genomic revolution, has been to identify a molecular target, perform experiments that satisfy one's confidence in the role of that target in human disease and then find an inhibitor by high-throughput screening, often combined with three-dimensional structure elucidation56. A good example of this process for malaria is the recent development of DSM265, an inhibitor of Plasmodium dihydroorotate dehydrogenase57. The validation of targets in malaria is always imperfect because it consists of making a gene-deficient parasite and showing a lack of viability. Screening of early-stage compounds against the dihydroorotate dehydrogenase enzyme and the high-resolution crystal structure identified a conformationally induced hydrophobic inhibitor-binding pocket at the end of the nucleotide ring where additional binding interactions could be made58 (Fig. 4).

DSM265 has an interesting parentage because it was suggested at a meeting between erstwhile competitors: the University of Texas, USA, Harvard University, USA, GlaxoSmithKline and Genzyme. DSM265 was also one of the first molecules to take advantage of the early availability of data from the human experimentally induced infection model. Currently, the Queensland Institute of Medical Research, Australia, is studying the effect of medicines on the speed of parasite reduction in human volunteers with very low level, subclinical infections59. In these data and other unpublished observations, parasite reduction rates are similar between volunteers with thousands of parasites per millilitre of blood and patients with thousands of parasites per microlitre of blood for the major classes of antimalarials tested so far. PCR-based parasite detection methods enable monitoring and treatment of the volunteers before they develop malaria symptoms. Another advantage of working with Australian volunteers is that they have no immune protection against the parasite: in that sense they are as naive as children in Africa but are otherwise healthy and robust. We were able to test DSM265 in a challenge cohort only 7 months after the first-in-human (Phase I) study160, which provided an early readout as to the efficacy and pharmacodynamics in 'real' patients and therefore confidence to select a dose for the upcoming patient studies. This compound began Phase IIa monotherapy studies in Peru in early 2015; Peru, as with many South American and Asian countries, has both P. falciparum and P. vivax infections.

A renaissance in phenotypic screening. The major breakthrough in hit discovery in the past 7 years came from an unexpected source: phenotypic screening. A minority of compounds currently in clinical testing had molecularly defined targets from the outset. The majority of compounds currently in development are derived from phenotypic screening; in most cases the targets were subsequently identified (Fig. 2). Advances in automation and in image capture and analysis technology have meant that over the period leading up to 2007, the cost of assaying a compound against primary human erythrocytes infected with P. falciparum fell almost 100-fold to approximately \$1 per data point. Led by collaborations established with the pharmaceutical companies Novartis60 and GlaxoSmithKline61, and the hospital-based drug discovery group at the St Jude Children's Research Hospital, USA62, large compound collections were screened, initially in-house63 and later at The Eskitis Institute, Australia64. A total of ~6 million compounds have been screened to date and, excitingly, more than 25,000 of these have shown half-maximal (IC50) activity at approximately 1 μM or lower against P. falciparum. The advantage of the relatively low price of screening is that it was easier to screen all compounds than to triage first, enabling a wide variety of compounds to be tested with, in some cases, structure–activity relationships determined for positive hits. In addition, it was relatively simple to propose a business model in which compounds would be screened and the details of follow-up collaborations discussed only once the initial hit data were available. This model for precompetitive collaboration helped bypass much of the sterile intellectual property discussions that have plagued neglected disease drug discovery (and drug discovery in general) in the previous decade.

The most advanced compound derived from this approach is the spiroindolone KAE609 (also known as cipargamin)65,66, which was developed in a collaboration between the Novartis Institute for Tropical Diseases in Singapore, its Genome Foundation in San Diego, USA, and the Swiss Tropical and Public Health Institute. The initial singleton hit (Fig. 4) was reasonably active and therefore the main problem in the medicinal chemistry programme was how to improve its in vivo half-life without losing activity against the parasite. Knowledge of the target was not essential in the early stages of identifying a development candidate: this lack of target information was a major change for many in drug discovery and was seen as an inherently high-risk approach. One of the teams described it as “driving in the dark with the headlights off”. Ultimately, the target was discovered to be the Na+-ATPase 4 ion channel PfATP4 (Ref. 67). Interestingly, PfATP4 had been identified as a potential target more than a decade earlier but had not been progressed68; biochemists tend to work on the targets that are easiest to assay. KAE609 rapidly progressed through to Phase I69, and has now completed the first studies in human patients in Thailand70. The results are stunning; KAE609 kills even faster than artesunate — something that was rarely imagined to be possible — and also maintains a plasma concentration above the MIC for several days. In addition, the doses used in the clinic were relatively low: 30–75 mg in total. One of the goals of drug discovery in malaria is to formulate new compounds that are affordable. This does not mean that they have to be cheap on a milligram basis. By weight, KAE609 is 7 times more potent than artesunate, and arguably 40 times more potent than 4-aminoquinolines. As such, KAE609 offers an opportunity not only for a single-dose cure but one that could be even cheaper than current therapy. Novartis hopes that KAE609 can be submitted for approval in 2017 (Ref. 71).

Several other chemical series have emerged that also target PfATP4, including PA21A092 from Drexel University, USA72,73,74, and SJ557733 (also known as SJ733) from a collaboration between St Jude Children's Research Hospital and Rutgers University, USA62,75; both of these series are in preclinical development. Indeed, an analysis of the Malaria Box63, a collection of 400 publicly available compounds distilled from the original 25,000 hits obtained in the high-throughput screening, showed that 7% of these compounds hit PfATP4 (Ref. 76). This phenomenon of frequently hit targets has also been observed in phenotypic screens against tuberculosis, and may reflect the inherent bias of the screening assay format and the chemical libraries tested.

Another candidate that has been developed from a phenotypic hit is MMV390048, which has been progressed by the team led by the University of Cape Town, South Africa (initially identified as compound 15)77. Initial screening identified a cluster of four aminopyridines, which were then optimized to produce an early lead (Fig. 4) and further refined to produce MMV390048. The compound progressed to preclinical development and at the same time its target was identified to be lipid phosphatidylinositol 4-kinase (PfPI4K)78. This target had previously been identified by the Novartis-led group as the target of one of their molecules that had not progressed into preclinical development79. Therefore, PfPI4K is another example of a frequently hit target. MMV390048 has successfully completed the preclinical programme and began Phase I human studies in Cape Town in the first half of 2014. This is another major step in the right direction for antimalarial medicines: it is the first time a new medicine has been taken into human volunteers in Africa, with direct readout of safety in the genetic background of the majority of patients with malaria.

## Future directions

A renaissance of new targets. As mentioned above, the phenotypic screening programmes followed by lead optimization and target identification have resulted in new molecular targets. Thus, target prioritization in this case is driven by real life druggability. Seven new targets and several revitalized targets have been identified by these approaches in the past 5 years (Table 3). These rediscovered targets are all druggable, by definition, and are often in well-characterized target classes (channels or enzymes)80. However, the phenotypic screening data allow priorities to be refined and set by the community, providing new opportunities for structure-based design approaches to groups such as the Structural Genomics Consortium81. They can also form the basis of target-based screening initiatives such as the Innovative Medicines Initiative's European Lead Factory82. In some cases, we are seeing the same target come up from many independent projects, as described above for PfATP4. However, prioritizing the less commonly hit new targets for target-based screening in the future will be an excellent strategy for improving the chemical diversity of our lead optimization projects.

New players and new diversity. Despite the interest in identifying targets, it can be argued that phenotypic screens of diverse chemotypes will continue to be the principal source of new candidate molecules in malaria. The search for radically new chemotypes is always going to be a high priority but is challenging, as illustrated by the disappointing harvest from large combinatorial libraries used in screening in other therapeutic areas83,84. The majority of novel chemotypes will come from two areas. First are natural products, from which many highly innovative but somewhat toxic scaffolds have been identified85. As malaria is a disease that has historically been treated by herbal medicinal products, selecting extracts with demonstrable clinical activity could clearly lead to new drugs86. Diversity in the natural product space has been facilitated by recent alliances with Japanese pharmaceutical companies, which have been aided by the Global Health Innovative Technology Fund, a product development partnership that was established in 2012 (Refs 87,88). Japanese companies have maintained more of a focus on fungal-derived natural products in recent decades. Second, the use of novel heteroatoms has also contributed to chemical diversity. For example, (boron-containing) oxaboroles have provided clinical candidates for bacterial and trypanosome infections; this chemical family also has activity against malaria89, although the compounds for this disease are still in the lead optimization stage.

Improved models for early-stage drug development. Another pathway to discover new chemotypes is to change the design of the biological screen. Two of the biological activities needed for the malaria eradication agenda are the ability to block transmission and to kill the dormant liver stages. In neither case can we perfectly reproduce the biology in high-throughput screening formats; however, as a surrogate for blocking transmission the ability of compounds to kill the late-stage gametocytes (stage V) can be tested90. For the dormant liver stages, we would need to screen surrogates of surrogates — a mouse parasite (Plasmodium berghei) in a non-dormant stage — but at least this means that the screening process can start. What is important is that the investment in the primary biology continues so that screening against primate dormant forms91, or even human cells infected with P. vivax92, can eventually be performed. Again, new classes of drugs for such stages of malaria parasites would truly be transformative and accelerate the decline of malaria — a change that is urgently needed over the next 20 years.

Ultimately, the challenge in drug discovery and development is to show that a medicine is effective in the human clinical situation. Historical development of anti-infective agents has often relied on the use of experimentally induced human infections93,94,95. In recent years, with the development of highly sensitive PCR-based techniques, volunteers can be infected with P. falciparum, and the effect of treatment can be studied on the resultant subclinical infection. Volunteers can be infected using biting mosquitoes96, injection with sporozoites97 or infected erythrocytes11,59. Clearly, sporozoite infection enables the study of the hepatic stages of infection, whereas erythrocytic infection focuses on the blood stages. Studies comparing the kinetics of the response to medicines in these subclinical infections show that the parameters obtained (such as the speed of parasite decline and the MIC) are similar to those obtained clinically. Use of the human experimentally induced infection model provides an early readout on the speed of parasite killing in non-immune individuals and enables an early estimation and assessment of whether the compound could be part of a drug combination that produces sterile cures. This trend will continue: new models are now being developed to study transmission to the mosquito98 and relapse of dormant forms99.

Subsequent clinical development of antimalarials must establish efficacy and safety in patients. A survey of recent Phase II–IV clinical trial locations (as listed in ClinicalTrials.gov) shows that most trials take place in Africa (n = 99) and Southeast Asia (n = 30) (Fig. 5a). Trials in patients from sub-Saharan Africa are important because of their wide genetic diversity100 and enhanced risk of the occurrence of pharmacogenomic complications101,102. However, the distribution of trials within Africa is unbalanced, with virtually all trials taking place in Kenya, Malawi and Burkina Faso (Fig. 5b). This unfortunate situation is due to a combination of regulatory, practical and political barriers. Recently, trial sites have opened up in Uganda, Gabon, Mozambique and the Democratic Republic of Congo, but there is still room for improvement.

Pioneering new ways of working together. It could be said that one of the difficulties of drug discovery in the past 20 years has been the lack of peer review: when projects are solely controlled by one company or university they lack the somewhat painful advantages of the peer-review process. In such an environment of “introvert innovation” (Ref. 103), key decisions on moving projects forward are typically taken by small committees that lack insider access to data for the global pipeline, instead being guided by internal drivers such as the company's growth expectations, individual personalities and internal competition with other therapeutic areas. To address this issue, we have instituted rigorous peer review for malaria drug development by having every step of the process reviewed by external experts from academic institutions, pharmaceutical companies, funding agencies, governments and malaria clinicians in the form of the Expert Scientific Advisory Committee. One question going forward is whether such peer review can be realistically achieved for other, more commercially important diseases.

Malaria is a disease with market failure, whereby the cost of developing new medicines, and the relatively low price that can be charged to the patient or health care provider, offers no simple commercial incentive to invest. Many of the world's major pharmaceutical companies have invested in neglected disease research and development. Indeed, some, such as Novartis and GlaxoSmithKline, even have centres of excellence dedicated to them. Historically, however, this work has been justified as part of corporate social responsibility. Nevertheless, there is the other side of the coin. The lack of a large commercial market has meant that malaria and other diseases of neglected populations have been at the forefront of exploring new modes of collaboration between academia and industry or even within the industry. The initial high-throughput screening of compound libraries against the malaria parasite is the first example, with the overwhelming majority of the top 20 pharmaceutical companies now having agreed not just to screen but also to share the results, and many of these are now in the public domain.

Second, the availability of large volumes of screening data has driven biological research. For example, the Malaria Box63, a collection of 400 compounds that was selected from the 25,000-compound malaria phenotypic screens, has been made freely available and without restriction to 200 groups worldwide. It has been tested on a wide variety of different parasites, including Cryptosporidium parvum104,105, Toxoplasma gondii, Entamoeba histolytica106, Trypanosoma cruzi, Trypanosoma brucei, Leishmania donovani, Perkinsus marinus107 and Schistosoma mansoni108. The data from these assays can then be put into the public domain, along with key drug-related parameters such as bioavailability, metabolism and half-lives. What becomes obvious from these types of collaboration networks is the high demand for collections of quality-controlled, validated compounds for biologists to test in their laboratories.

Third, some companies have adopted a different approach to classical partnerships. Both AbbVie and Hoffman-La Roche have allowed their own in-house experts to provide advice on key preclinical development questions to malaria drug developers outside their companies, and have also carried out specific safety and pharmacokinetics experiments on development compounds and donated the results to the projects.

Fourth, academics are finding new ways to collaborate to facilitate the progress of antimalarial drugs. One example is the recently formed Structure-guided Drug Discovery Coalition (see Further information). Funded by the Bill and Melinda Gates Foundation, this group brings effort and funding from structural genomics groups in Canada, the USA and the UK, and links them to experts in chemistry and pharmaceutical drug development to accelerate the translation of structure to drugs. Another example is open-source drug discovery, wherein results are published on the web for all groups to use (Open Source Malaria; see Further information).

## Conclusions

The fight against malaria will be a long one; ambitious targets of a 90% reduction in deaths by 2030 have already been proposed by the WHO, indicating that new medicines will be needed for at least another two decades. How urgently we need new medicines depends partly on how good our stewardship of the existing ones is: monotherapies, incorrect dosing and counterfeits still plague malaria treatment. The initial impression from the malaria portfolio is that it is relatively strong and has improved dramatically over the past 5 years. However, it is important to attempt to separate the medicines in the pipeline into the different product types to see the true picture.

First, for blood-stage treatments, there are four molecules currently in Phase II studies, which initially gives the optimistic impression that a new medicine will soon be launched. However, our estimate of the probability of success for each combination is only 34%10. In this review, it was calculated that, on average, 18 candidate molecules would be needed for every new blood-stage medicine launched. Based on this and the ten new molecules in the translational part of the portfolio, the pipeline is in good shape to deliver one new candidate medicine. This medicine would be most useful in regions where the current drugs are failing because of resistance. Use of a new medicine in an eradication agenda places increased stringency on safety margins. If these medicines are to be used to treat asymptomatic subjects to destroy the reservoir of parasites, the risk:benefit ratio needs to be similar to that of a vaccine — a much more difficult goal to achieve — with a consequently lower success rate for drug development. Requiring a single-dose therapy rather than a 3-day regimen also decreases the number of drugs that will be launched but should increase the therapeutic success of those drugs.

Second, for blocking transmission the WHO treatment guidelines recommend low-dose primaquine. The current key challenge here is not discovery but rather making paediatric tablets of low-dose primaquine available. For the future, most of the blood-stage compounds in development show at least partial inhibition of transmission in membrane feeding assays, and therefore it is likely that new combinations will be more effective at blocking transmission than current drugs. The search for new molecules that solely block transmission is therefore less of a priority, although ideally new blood-stage combination therapies should be selected to include transmission-blocking properties.

Third, for prevention of the relapse of P. vivax, although the data on tafenoquine are remarkable, there remains the issue of haemolysis in G6PD-deficient subjects. Because there are no new anti-relapse scaffolds in the pipeline, identification of new scaffolds, and indeed new assays, is important, as is investment in a robust diagnostic test for evaluating the G6PD status of patients in the field.

Finally, for chemoprevention, some of the newer molecules, such as KAF156 (also known as GNF156) and DSM265, show interesting preclinical activity against liver schizonts, but this remains to be confirmed in studies in humans. Given that chemoprotection includes protecting expectant mothers, increased resources over the next few years are needed for studies of safety in pregnancy and models that predict this safety.

In summary, although there has been a dramatic improvement in the pipeline of new antimalarial molecules over the past decade, the glass is still rather empty. Attrition in development is a fact of life, and the contents of the glass will tend to evaporate unless they are constantly topped up by the discovery of new molecules. Resistance already exists to the mainstays of therapy: both artemisinin derivatives and the 4-aminoquinoline or aminoalcohol partners. The discovery of new antimalarials is in a race against the spread of such resistance from the Mekong to the major endemic regions of Africa. There has been a massive investment in malaria control in the Mekong region, where the drug resistance has arisen, and this has bought us time in which to develop the next generation of medicines. There is a rich diversity of mechanisms of action among the new medicines entering the pipeline, and all these new compounds are checked for activity against existing clinical isolates and against experimentally produced resistant parasite lines. As much as possible, the pipeline is being protected against future resistance.

Beyond discovering good drugs, there are numerous other obstacles to making the right drug available for the right patient. Keeping the cost down is always a challenge, and the inclusion of price constraints and access clauses into drug discovery grants for malaria is an important first step. Ensuring availability of multiple suppliers is also critical in the long run. Counterfeit drugs remain a problem: they put the lives of patients at risk and also increase the threat of resistance to the active ingredients. More needs to be done to address this issue. There is often too long a gap between the development of the adult dose of new drugs and development of the paediatric version. This gap is being addressed by paediatric ACTs, but early assessment of the correct dose in the paediatric target population is going to be an important challenge over the next decade. Additionally, still too little is known about drug–drug interactions, especially those with drugs for other frequently occurring diseases in the target population (such as HIV and tuberculosis). Deploying new medicines in the areas that most need them is not always straightforward. Review and approval of regulatory files in disease-endemic countries can take several years after approval by a stringent authority. Continued investment is needed in regulatory support, in addition to the development of networks of in-country regulatory authorities, so that the file does not have to be reviewed by every single country on its own.

Against this complex background, the key is to be able to provide the front-line countries with a strong portfolio of new medicines: in a single word, choice. The portfolio of new chemical entities in development against malaria has been transformed in the past decade. One challenge for the next decade will be to ensure that these are registered and rolled out to the countries that need them the most. This objective must be achieved without decreasing investments in drug discovery or the identification of new targets. New classes of medicines are needed to address the more subtle problems such as transmission-blocking and prevention of the relapse of liver-stage parasites. Furthermore, medicines that are safe enough to be used routinely for chemoprevention in vulnerable patient populations are still needed. If we can launch a handful of new classes of medicines over the next decade, without losing the focus in drug discovery, then it may be audacious enough to claim that we have reached, to quote Sir Winston Churchill, “... perhaps, the end of the beginning”.