Challenges and strategies for developing efficacious and long-lasting malaria vaccines

New knowledge and strategies are emerging to enable the development of an efficacious and long-lasting vaccine against malaria. Although there has been major recent progress in malaria vaccine development, substantial challenges remain for achieving highly efficacious and durable vaccines against Plasmodium falciparum and Plasmodium vivax malaria. Greater knowledge of mechanisms and key targets of immunity are needed to accomplish this goal, together with new strategies for generating potent, long-lasting, functional immunity against multiple antigens. Implementation considerations in endemic areas will ultimately affect vaccine effectiveness, so innovations to simplify and enhance delivery are also needed. Whereas challenges remain, recent exciting progress and emerging knowledge promise hope for the future of malaria vaccines.


INTRODUCTION
Malaria remains as one of the world's leading health issues, respon sible for >200 million clinical cases and up to 500,000 deaths annually, and is a leading cause of death among young children (1). The strong need for malaria vaccines with high efficacy to achieve control and elimination is recognized by key global organizations. This need is heightened by the spread of antimalarial drug and insecticide resist ance (1), recent rebound increases in malaria in many regions, and stalled progress toward reducing the global malaria burden (1). The World Health Organization (WHO) and partners set a strategic goal of developing malaria vaccines with >75% efficacy against clinical malaria and suitable for use in all malariaendemic areas by 2030 (duration of protection demonstrated over ≥2 years and booster doses required no more frequently than annually) (2).
However, achieving vaccines with sustained high efficacy has been an enduring challenge. Animal models have provided a proof of concept for various experimental vaccines, but very few have shown strong efficacy in human clinical trials. While multiple Plasmodium spp. can cause human malaria, Plasmodium falciparum is responsi ble for most of the cases and deaths, particularly in Africa, and has been the major focus of vaccine development. Plasmodium vivax is the second major cause of malaria, but few vaccines have been de veloped and tested for this species, in part, because of the relatively recent recognition of the severe forms of this infection and because of the challenges in maintaining P. vivax in culture in vitro. Other species-Plasmodium knowlesi, Plasmodium ovale, and Plasmodium malariae-are responsible for a small burden of disease and are not currently a major focus of vaccine development.
The complexity of malaria is a major challenge for vaccine develop ment; with over 5000 genes, there are hundreds of potential targets and vaccine candidates over different life stages (Fig. 1). Vaccine strat egies can be broadly grouped by the malaria parasite life stage they target. Preerythrocytic vaccines target sporozoites, which are inoc ulated by a mosquito and travel to the liver to initiate infection in hepatocytes, and/or parasiteinfected hepatocytes. Preerythrocytic vaccines are attractive because they can prevent initial infection and thereby prevent clinical illness and malaria transmission. After rep lication in hepatocytes, merozoite forms are released into the blood where they infect red blood cells (RBCs) and undergo asexual replica tion, releasing new merozoites that infect RBCs. Bloodstage vaccines generally target the merozoite form and aim to prevent replication and the development of clinical illness. Reducing bloodstage para sitemia may also reduce the transmission of malaria. Transmission occurs when some intraerythrocytic parasites differentiate into sexual forms, known as gametocytes, which can be taken up by mosquitoes where they undergo sexual replication, and are subsequently trans missible to humans. Transmissionblocking vaccines (TBVs) target gametocytes and parasite stages in the mosquito midgut to prevent infection of mosquitoes. However, the success of this strategy has not yet been demonstrated in clinical trials or human population studies.

CURRENT STATE OF MALARIA VACCINES TESTED IN CLINICAL TRIALS
After successful preclinical development, numerous approaches have been used to evaluate vaccines for efficacy in clinical trials, using dif ferent outcome measures. Demonstrating efficacy against malaria in children under conditions of natural exposure is required, but early clinicalphase trials may first test for efficacy in preventing infection or clinical illness or reducing parasitemia in adults in endemic re gions. A small number of vaccines have shown efficacy in phase 2 trials. However, in general, efficacy has been modest and has lacked sus tained protection. Because of the cost, complexity, and size of phase 2 field trials, there has been an increasing use of human challenge models whereby vaccinated adults undergo controlled human ma laria infection (CHMI), delivered via mosquito bite or intravenous injection. While many candidate vaccines have been evaluated in preclinical models, we will focus our review on those vaccines that have been tested for efficacy in human trials (Table 1).

Pre-erythrocytic vaccines
RTS,S is the most advanced malaria vaccine candidate and is based on a viruslike particle (VLP) containing central repeat and Cterminal epitopes of the major sporozoite surface antigen, circumsporozoite protein (CSP), with the aim of generating immunity that would pre vent infection and subsequent malaria. A series of promising phase 1 and phase 2 trials led to a large phase 3 trial of >15,000 infants and young children across 11 sites in subSaharan Africa. In children 5 to 17 months old at first vaccination, vaccine efficacy assessed over 12 months of followup was 50.4% (95% confidence interval, 45 to 54%; intention totreat analysis) and vaccine efficacy against severe malaria was 45.1% (95% confidence interval, 23 to 60%). Among infants 6 to 12 weeks of age, vaccine efficacy was 30.1% (95% confidence interval, 23 to 36%). Over 3 to 4 years of follow up, vaccine efficacy was 28 and 18% without booster, in children and young infants, respectively, and 36 and 26% in those who received a booster dose at ~18 months (3). A key limitation with RTS,S is waning of vaccine efficacy over time, and there fore, identifying strategies to generate more sustained efficacy is a prior ity. Statistical analyses suggested that vaccine efficacy was higher initially (60 and 40% at 3 months for the children and infants, respectively) but quickly waned to limited efficacy by 18 months (4). There was high variability in vaccine efficacy between study sites over the 18month followup, ranging from 40 to 77% among children and 0 to 49%  Fig. 1. Plasmodium life cycle and targets for vaccine development. Different life cycle stages of Plasmodium, the protozoan parasite that causes malaria. These stages include sporozoites that infect hepatocytes and develop inside them (pre-erythrocytic stages), merozoites that infect RBCs and develop into trophozoites and schizonts (blood stages), gametocytes that develop inside RBCs and are taken up by female mosquitoes during a blood meal, and gametes and ookinetes that develop in the mosquito gut (transmission stages). The parasite target antigens of the three major types of vaccines-pre-erythrocytic, blood stage, and transmission blocking-are indicated in Table 2. Pre-erythrocytic vaccine candidates target sporozoites in the blood or developing parasites in hepatocytes. Blood-stage vaccines predominantly target merozoites, which are extracellular for a short time after release from the liver or schizont before they invade erythrocytes, but some target the parasite antigen PfEMP1 expressed on the surface of infected erythrocytes. TBVs induce antibodies against the transmission forms of the Plasmodium parasite, the gametes and ookinetes, to prevent infection of mosquitoes and subsequent transmission of sporozoites to another human host.
continued on next page among young infants (4). Despite efficacy being below current de sired goals, modeling studies (accounting for transmission intensity, intervention usage, vaccine coverage, and other parameters) predict that RTS,S would avert a substantial number of clinical cases and severe episodes and contribute to improving public health (5). How ever, a higher rate of meningitis and allcause mortality in girls was reported in RTS,S vaccinated children (3,6), but it is not known whether this was causally related to vaccination. RTS,S has received a positive scientific opinion from the European Medicines Agency and is entering large implementation trials in three African coun tries to further assess efficacy, safety, and operational issues.
Another vaccine approach has a viralvector heterologous prime boost using ChAd43 and modified vaccinia ankara (MVA) viral vec tors expressing TRAP antigen (which is present on sporozoites and infected hepatocytes) and multiple T cell epitopes (METRAP). This approach induced sterile protection in 21% and delayed patency in 36% of vaccinees in CHMI efficacy trials (7). Phase 2b trials demon strated a 67% risk reduction in infection among Kenyan men over shortterm followup (8), but no efficacy among Senegal adults (9). Only one vaccine against P. vivax has progressed to efficacy trials; a PvCSPbased vaccine showed no efficacy in a CHMI challenge (10).
Two whole sporozoite vaccine strategies have shown efficacy in clinical trials. The PfSPZ vaccine contains irradiationattenuated sporozoites that invade hepatocytes but do not develop to blood stages. PfSPZ provided ~80% protection against CHMI in malaria naïve adults against homologous strain challenge (11-13). However, it had lower efficacy against heterologous strains and was less immunogenic and less efficacious in field evaluations (14)(15)(16). Data from further efficacy trials are expected soon. PfSPZCVac involves inoculation with viable sporozoites in combination with antimalarial drug prophylaxis to kill parasites at early blood stages. High efficacy was seen in CHMI studies using the homologous strain (17,18). Other whole sporozoite vaccines include the use of genetically at tenuated parasites that are engineered to infect the liver but cannot progress to bloodstage infection and disease. These have been effi cacious in mice but, to date, have only been tested for safety and immunogenicity in humans (19). However, the delivery of whole parasite vaccines would require different infrastructure and opera tional approaches to the current childhood Expanded Program for Immunization (EPI). The goal of the EPI is to provide universal ac cess to relevant vaccines for atrisk populations for the control of infectious diseases. Currently, EPI vaccines are stored and trans ported in a ~4°C cold chain and administered by intramuscular or subcutaneous injection, or orally. Current liveattenuated whole sporozoite vaccines require storage and transport in liquid nitrogen vapor phase and delivery by intravenous injection, which is not supported by EPI.
Blood-stage vaccines AMA1, a key invasion protein of merozoites, is the most studied bloodstage vaccine (Table 1). To date, only one AMA1based clin ical trial has shown efficacy: FMP2.1/AS02A phase 2b trial in Malian children demonstrated 64% efficacy against vaccinelike strains but did not give significant overall protection, presumably because of antigenic diversity (20). Another AMA1based vaccine, AMAC1, had no efficacy in Malian children (21). MSP1 is a highly abundant mero zoite surface antigen, and antibodies against MSP1 are protective in animal models and associated with protection in some human cohort studies (22). However, a phase 2b trial of FMP2.1/AS02A(MSP1- 42) in Kenyan children showed no significant efficacy despite inducing high titer antibodies (23). The combination of MSP1 and AMA1 using viral vectors also showed little efficacy in a CHMI trial (24). An MSP3based vaccine elicited high magnitudes of cytophilic im munoglobulin G1 (IgG1) and IgG3 subclass responses in African children, and secondary analyses found that vaccinated children had a reduced incidence of malaria (25). The GMZ2 vaccine (MSP3 and GLURP) induced high magnitudes of functional antibodies in a phase 1 trial (26), but little efficacy in a larger trial in African children (27). ‡Efficacy against all clinical episodes of malaria. §Efficacy against CHMI (note that only protection against infection was included, in some studies there may have been a delay in parasitemia).

Vaccine
Phase; schedule Participants* Efficacy % (95% CI) †, ‡, § Reference Another multiantigen vaccine (combination B; composed of MSP2, RESA, and a fragment of MSP1) showed promise in a phase 2b trial in Papua New Guinean children, with a 62% reduction in parasite density among a subgroup of vaccinees not pretreated with antima larials (28). The protective effect appeared to be due to targeting of MSP2, and efficacy was strain specific against vaccinelike infections. The only P. vivax bloodstage candidate to undergo clinical trials, PvDBPRII, recently completed a phase 1a trial (29) and induced antigenspecific T cell responses and antibodies that inhibited PvDBP RII binding to its receptor.

Transmission-blocking vaccines
To date, no data on the efficacy of TBV candidates in clinical trials has been reported. Leading TBV candidates include Pfs230 and Pfs48/45, which are expressed by gametocytes in the human host, and Pfs25, exclusively expressed in the mosquito vector (zygote and ookinete stages). In phase 1 clinical trials, vaccines based on Pfs25 and its ortholog Pvs25 have shown induction of antibodies that block mosquito infection (30). A recent study reported that Pfs25EPA/ alhydrogel (Pfs25 conjugated to a detoxified form of Pseudomonas aeruginosa exoprotein A) induced functional transmissionblocking antibodies in healthy Malian adults. However, substantial antibody levels were only achieved after four doses, and antibody levels rapidly waned after the final dose (31). Vaccines based on Pfs230 and Pfs48/45 have entered clinical trials, but results are not yet available.

IMPROVING EFFICACY: THE FUTURE OF MALARIA VACCINES
On the basis of current results of vaccine trials, it is clear that reaching the WHO goal of a malaria vaccine with >75% efficacy in malaria endemic populations is exceptionally challenging. Fundamental issues for achieving higher efficacy include (i) generation of highly potent functional immunity, which depends on a strong knowledge of mechanisms and mediators of protective responses; (ii) selection of the right antigens and epitopes (or combinations) that mediate protective immunity; and (iii) developing strategies to overcome immune eva sion and prevent vaccine escape. Furthermore, emerging findings sug gest that vaccineinduced responses are lower in malaria exposed populations, reflected by lower vaccine efficacy than seen in malaria naïve populations (Table 1), raising the prospect of considerable im mune dysregulation in malariaexposed populations that affects the ability to generate and maintain potent protective responses (32). An important constraint to achieving higher vaccine efficacy is the lack of clear correlates of protection to inform research and de velopment activities. As such, developing immunologic correlates of protection is a priority area in the WHO malaria vaccine technol ogy roadmap. Data from studies of naturally acquired and vaccine induced immunity, as well as evidence from animal models, point to the importance of both humoral and cellular immune responses in mediating protection from malaria. However, the relative impor tance of humoral and cellular immunity, as well as specific compo nents of these responses, differs depending on the malaria life stage and by the antigenic target. Furthermore, the effects of humoral and cellular immunity on controlling parasite density differ in the kinet ics of their effects (33). Antibodies, being preformed, can have an immediate effect, provided that their function and concentration are sufficient. Cellular immunity dependent on CD4 + T cells typi cally takes 7 to 10 days to exert an effect because memory T cells need time to expand in number and then secrete effector cytokines and inflammatory molecules. Natural killer (NK) cells and  T cells can have a more immediate effect.

Mechanisms of humoral immunity
Antibodies have been identified as a key component in adaptive hu moral immunity against multiple malaria stages (Table 2). There fore, maximizing the induction of antibodies with strong functional activity will be crucial for enhancing vaccine efficacy. However, a key knowledge gap is the lack of defined immune correlates of pro tection, hampering the capacity to evaluate vaccines. Antibodies can function via multiple mechanisms, including direct inhibition (neutralization) by blocking key parasite receptorligand interactions, by interacting with Fc receptors (FcRs) expressed on phagocytes to promote cellular uptake and degradation [opsonic phagocytosis and antibodydependent cellular cytotoxicity (ADCC)], and by fixing C1q to activate the complement cascade to enhance neutralization and kill target cells (22). The functional properties of antibodies are influenced by multiple factors including isotype/subclass, epitope specificity, affinity, and glycosylation (34). Cytophilic IgG subclasses (IgG1 and IgG3) have the highest capacity to mediate Fcdependent functions (34) and have been linked with protection in studies of naturally acquired immunity (22). IgG subclass profiles differ by spe cific antigens (35) and with the use of different vaccine adjuvants (36), which could be exploited in vaccine design, formulation, and dosing.
FcRs are present on major effector cells including monocytes, macrophages, neutrophils, and NK cells, and their expression and functions differ between cell types and subtypes. Complement re ceptors are also differentially expressed across cell types, and com plement fixation can enhance or mediate phagocytosis. Furthermore, the expression and function of FcRs and complement receptors are altered by activation of immune cells, which may occur during malaria infection or other infections. Therefore, a strong under standing of the key effector mechanisms (including cells and anti bodies), and the roles of specific FcRs and complement, is essential for maximizing humoral immunity in vaccine design. Mice are com monly used as models for malaria vaccines. However, there are im munologically important differences compared to human immunity for murine IgG subclasses and their functions, FcR types and func tions and their expression on immune cells, and effector cell com position and abundance (37), as well as key biological differences in model Plasmodium spp. used in animal models. The course of in fection and overall antigenic exposure also differ between human and murine malaria. This has major implications for the application of mouse models for evaluating vaccineinduced humoral immunity and highlights that mouse models should be used judiciously to address specific questions.
Naturally acquired and vaccineinduced antibodies to multiple pre erythrocytic antigens are associated with a reduced risk of infection and malaria (38). This includes RTS,S vaccine-induced immunity, which has been shown to be associated with protection (39). Anti bodies can target the sporozoite after inoculation in the skin, after entry to the blood vessel, and during arrest at the liver sinusoid be fore hepatocyte invasion (40). Direct neutralizing antibodies have been identified in mice where they immobilized sporozoites in the dermis (41). Antibodies can also directly inhibit P. falciparum spo rozoite gliding in vitro and traversal and invasive forms of motility when tested at relatively high concentrations (19). Monoclonal anti bodies (MAbs) induced by RTS,S vaccination also protected humanized mice from infection (42). However, direct neutralizing antibodies have not been specifically associated with protection. Recent studies established that P. falciparum sporozoites are susceptible to antibody complement attack, which inhibits traversal and causes cell death (43,44). Complementfixing antibodies against CSP were associated with protective naturally acquired immunity in children (43), and complement fixation by antiGal antibodies (targeting a glycan on sporozoites) appeared to play a role in vaccineinduced immunity in a mouse model (45). Other Fcdependent functions of antibodies targeting sporozoites and the roles of phagocytes are relatively under studied. Opsonic phagocytosis of antigencoated beads by a monocyte cell line was not associated with protection in a CHMI trial of RTS,S (46). Given the abundance of neutrophils in the blood, an investiga tion of their role is warranted. Furthermore, most of the studies have focused on IgGdependent mechanisms, but recent studies have suggested that IgM induced by vaccination can play a role (47).
The key role of antibodies against blood stages was clearly estab lished by passive transfer of antibodies from immune adults to chil dren and adults with malaria, which drives parasite clearance (48). Protective antibodies target merozoite antigens (22) and antigens on the surface of infected erythrocytes (IEs) [predominantly targeting PfEMP1; (49)]. Antibodies to merozoites can act by directly blocking key receptorligand interactions to inhibit invasion of erythrocytes (22). However, recent studies suggested that most acquired human anti bodies rely on complement recruitment for effective inhibitory ac tivity, and this activity was associated with immunity in children (50). Cytophilic IgG subclasses (IgG1 and IgG3) can promote merozoite phagocytosis by monocytes, leading to their activation, and engage FcRs on neutrophils, leading to acute respiratory burst activity (51,52). Monocyte phagocytosis can lead to the secretion of molecules that in hibit parasite growth, known as antibodydependent cellular inhibition (ADCI) (53). Antibodies to infected erythrocytes are thought to act by inhibiting vascular adhesion and promoting phagocytosis by monocytes (49), which may be enhanced by complement (54). Antibodies can also inhibit parasite growth in RBCs via ADCC by NK cells (55). Neutro phils may also play a role; however, they have been less studied. How the relative importance of specific functions can vary between antigens is not well understood. It is likely that multiple effector mechanisms are involved in bloodstage immunity because directly growth inhibitory antibodies have not been consistently associated with protective im munity (22), and antibodies that promote complement activation or opsonic phagocytosis and ADCC have been correlated with protec tive immunity (22). Furthermore, some data from mouse models support important roles for FcRs and complement (56,57).
Transmissionblocking immunity is mediated by antibodies that act primarily by neutralizing the transmission stages in the mosquito midgut after being taken up in a blood meal, thereby preventing mosquito infection (58). MAbs to Pfs230 and naturally acquired human antibodies were shown to induce lysis of gametes in the presence of active complement (59,60). However, antibodies can have transmissionblocking activity without the need for comple ment fixation (60), and there is little knowledge on how functional transmission blocking antibodies or complementfixing antibodies are acquired and maintained. The roles of FcRs and innate immune †Summary of demonstrated mechanisms derived from human studies and animal models. ‡CD4 + T cell help for generation of antibodies and induction of proinflammatory cytokines.
§For the purposes of this table, leading candidates are considered to be those that have been tested in clinical trials (particularly those with demonstrated efficacy) or antigens that have substantial efficacy in a nonhuman primate model. cells remain largely unexplored question but should be prioritized in future research.

Mechanisms of cell-mediated immunity
Generating potent CD8 + T cell responses is important for achieving effective immunity against infected hepatocytes, and optimizing CD4 + T cell responses is likely to be essential for potent and longlasting immunity. Targeting preerythrocytic stages, protection induced by PfSPZ immunization in animal models is primarily mediated by CD8 + T cells (61). CD4 + T cells can also confer protection in mice (62) and provide help to CD8 + T cells against liver stages (63). The importance of CD8 + T cell responses in liverstage immunity has not been clearly established in humans, likely because of the low frequency of circulating cells in peripheral blood. The best cellular correlate of protection identified for PfSPZ vaccines in humans was frequency of the V9 V2 subset of  T cells (11).  T cells, which recognize phosphoantigens produced by the Plasmodium apicom plast (64), are important sources of interferon (IFN), which plays a role in protection (65). PfSPZCVac also induced parasitespecific CD4 + and CD8 + T cells expressing cytotoxic markers CD107a and granzyme B, respectively, which were associated with protection (66). Polyfunctional cytokineproducing memory CD4 + T cells respond ing mainly to sporozoites also correlated with protection after im munization with PfSPZCVac (18,66).
RTS,S vaccination induced CSPspecific CD4 + T cells secreting T helper 1 (T H 1) cytokines interleukin2 (IL2), IFN, and tumor necrosis factor at low frequencies (39). Field pediatric studies showed that vaccination elicited polyfunctional CD4 + T cell responses mainly confined in memory compartments (67), and T H 1 cytokine responses (assessed following vaccine antigen stimulation of peripheral blood mononuclear cells) were associated with malaria protection, where as T H 2 responses were associated with risk (68). CD8 + T cell re sponses are not prominent after immunization by RTS,S, although analysis is limited to peripheral blood cells.
Targeting blood stages, cellular responses provide important effec tor mechanisms, in addition to humoral immunity. CD4 + T cells and particularly T follicular helper cells (T FH cells) are key for germinal center reactions necessary for the generation of plasma cells, high affinity antibodies, and longlived memory B cells (69). In addition, secretion of IFN by  T cells (70),  T cells (71), or NK cells in re sponse to infected erythrocytes is an important effector mechanism for bloodstage immunity. Although CD8 + T cells are not thought to play a direct role in killing of P. falciparum blood stages, recent data suggest that they do play role against P. vivax through major histocompat ability complex I (MHCI) expressed on infected reticulocytes (72).

Optimizing adjuvants, dosing, and the use of alternative vaccine platforms
One avenue that may be leveraged to improve vaccine efficacy is optimal adjuvant usage ( Table 3), which is severely hampered by the limited number of approved adjuvants for human use. An ideal adjuvant should enhance immunogenicity without compromising vaccine tolerability or safety. Alum is the adjuvant with the most ex tensive safety record for use in human clinical trials and is commonly used in malaria vaccine trials. However, vaccines to Plasmodium antigens formulated in alum induce only moderate antibodies and poor cellular responses, with an overall T H 2skewed response (73). For example, early studies found that RTS,S with alum was not effi cacious (74) and was subsequently formulated with new adjuvants AS01/AS02 to improve efficacy. AS01/AS02 adjuvants target Toll like receptor 4 (TLR4) to stimulate the production of cytokines and costimulatory molecules, and new approaches using other TLR ag onists also show promise for enhancing immune responses (36), which are discussed below. Further understanding of immune re sponses required for protective immunity and the nature of re sponses induced by vaccination with specific adjuvants may further improve adjuvant selection in malaria vaccines.
To elicit strong CD8 + T cell responses required for protection against preerythrocytic stages, vaccine candidates have been for mulated in recombinant viruses and administered by heterologous primeboost strategies, which have been a promising approach for the METRAP vaccine (7). Some primeboost strategies have also yielded improved immune responses in animal models and humans in other infections, such as HIV and influenza (75). Other improve ments in humoral and cellular responses may be achieved with other delivery platforms or formulations, including micro needle skin patch delivery, nanoparticles, and VLPs, to improve anti gen retention and uptake by lymph nodes and antigen presentation or using specific strategies for targeting antigenpresenting cells (36). For example, antigens in larger particles have been found to prolong antigen pre sentation by dendritic cells (DCs), leading to enhanced production of T FH cells and antibodies (76). Studies with vaccines for other in fections have also pointed to the potential importance of spacing intervals between vaccine doses (75).

Selecting antigens and epitopes
Multiple criteria need to be considered to identify and prioritize an tigens and epitopes for vaccine development. These include antigen cellular location and function, abundance, polymorphisms, data from in vitro functional assays, evidence of protective associations in studies of naturally acquired immunity, and data from animal models. Vaccine antigens that have progressed in clinical trials to date have largely been proteins identified before the era of malaria genomics and proteomics. In recent years, studies have widened the search and identified numerous new targets of immunity and promising vaccine candidates (Table 2). This advance has been most notable for merozoite antigens (77), but additional potential candidates for preerythrocytic and transmission stages are now emerging. These include 6cys domain family members (such as Pf12, Pf38, Pf41, and Pfs47), which have potential as preerythrocytic vaccine targets (78), and transmissionblocking vaccine targets (58,79,80). Pf12, Pf38, and Pf41 are also expressed during asexual blood stages (81) and thus have potential in multistage vaccines. PfHAP2 was recently identi fied as a promising transmissionblocking candidate (82).
Other approaches have identified key invasion ligands essential for parasite growth. This includes PfRH5, which binds the RBC re ceptor basigin (83), as a promising bloodstage candidate, and PfRH5 vaccination was efficacious in Aotus monkeys when using a potent adjuvant (84). Understanding the importance of PvDBP in reticulo cyte invasion by P. vivax underpinned this protein as a lead vaccine candidate (85). Ongoing studies are revealing key protein functions that help downselect and refine candidates for accelerated development.
Once protective antigen candidates have been prioritized, com bination vaccines may further improve efficacy (Fig. 2). Several cur rent strategies have already attempted this by combining antigens that target multiple parasite stages, such as NYVACPf7 (Table 1) (86). Proof of concept that combination vaccines may induce higher effi cacy was recently shown in mouse models where antibodies targeting sporozoites (CSP) and a transmissionblocking candidate (ortholog of Pfs25) acted synergistically to reduce parasite density in a multi generation population model (87). Combining multiple antigens against a single parasite stage may also result in increased efficacy, which is supported by modeling studies (88). Furthermore, human cohort studies have found that antibodies to multiple merozoite Table 3. Overview of strategies for generating highly efficacious and long-lasting vaccines.

Increasing vaccine efficacy
Target key epitopes Identify functional epitopes for host cell invasion or parasite development

Identify epitopes for antibody-mediated complement and Fc-receptor functions
Structure-based vaccine design: Improved targeting of functional epitopes Other factors (e.g., microbiome, metabolomics, genomics)

Generating long-lasting protective responses
Maintenance of protective antibodies Maximize initial induction of antibodies targeting functional epitopes Induction of multiple antibody functions may maintain protective effects for longer; complement recruitment may enhance antibody function antigens have higher associations with protection than singleantigen responses (77,89). The breadth of the antibody response also ap pears important in protection induced by irradiated sporozoite vac cination (90). Whole parasite vaccines are an additional strategy for presenting multiple antigens in their native conformation. Induction of protective antibodies may also be improved by iden tifying key epitopes targeted by protective responses. Current vac cine strategies to whole antigens or whole parasites typically consist of a range of protective and nonprotective epitopes. Refining vaccine immunogens and responses to predominantly target key functional or protective epitopes and response types may enable protective im munity to be better maintained as the overall immune response wanes. This concept underlies the rationale behind structurebased vaccine design, whereby knowledge of antigen structures and func tional epitopes targeted by protective antibodies may enable enhanced immunogen design. For example, a recent study isolated antibodies from Tanzanian volunteers immunized with PfSPZVac that bound to both the CSP NANPrepeat region (included in the RTS,S vac cine) and a peptide at the Nterminal junction. These antibodies were more effective at inhibiting sporozoite infection of hepatocytes in a humanized mouse model than antibodies targeting only the NANP region (91,92). This suggests that inclusion of the Nterminal junction peptide in RTS,S may improve vaccine efficacy. Similarly, epitopes of inhibitory MAbs have been identified for several mero zoite vaccine candidates (PfRh5, EBA175, PfAMA1, and PvDBP). However, translation of knowledge of functional epitopes into more efficacious vaccines has not yet been achieved.

Overcoming evasion of vaccine-induced immunity
Polymorphisms in antigens included in vaccines facilitate immune escape, which has been a problem in several vaccines including RTS,S and vaccines based on MSP2 and AMA1; vaccine efficacy was higher against infection of vaccinelike strains compared to vaccine dissimilar strains (93). Therefore, implementation of RTS,S could shift the burden of malaria to vaccineescape strains over time, decreas ing its effectiveness. This emphasizes the need for nextgeneration vaccines that induce more potent crossstrain protective responses. Strategies to address this include the inclusion of multiple alleles of an antigen [e.g., AMA1 (94)

Fig. 2. Strategies to develop efficacious and long-lasting malaria vaccines.
Schematic representation of key steps and activities in developing future malaria vaccines that have a high level of efficacy and generate long-lasting immunity. Potential candidate antigens for vaccine development include existing candidates and new antigens being identified from candidate discovery research. Refinement of lead candidates for development involves initial down-selection of candidates, refining immunogen design, overcoming issues of antigen polymorphism and vaccine escape by the parasite, and facilitating long-lasting immunity. This may require combinations of antigens. Preclinical evaluation (in vitro and in animal models) tests vaccine efficacy and duration and identifies strategies for optimal induction and maintenance of protective responses. Clinical evaluation of vaccine efficacy can be conducted in CHMI trials and in population-based clinical trials. Throughout this process, defining mechanisms of functional protective immunity and immune longevity and how to generate sustained immunity is crucial. Establishing correlates of immunity will greatly facilitate evaluation of candidate vaccines and aid in vaccine design and refinement.
immunization with two different MSP2 alleles surprisingly shifted antibody targeting to conserved epitopes, in contrast to naturally acquired antibodies that are overwhelmingly allele specific (96).
Another approach is to focus on conserved epitopes or less poly morphic antigens. Conserved epitopes have been reported in CSP, and potentially improved targeting of these by vaccination may yield higher efficacy. PfRH5 is an example of a less polymorphic antigen, and antibodies to PfRH5 have crossstrain activity in vitro (97). Modified vaccine antigens for hepatitis C virus and influenza have shown promising crossstrain neutralization activity (98,99).
Parasitemediated immunomodulation poses an additional challenge to vaccine development for atrisk populations. Previous malaria exposure may hamper the induction of vaccine responses, evidenced by the variation in the vaccine efficacy of RTS,S among populations with different intensities of malaria exposure (4), and the reduced immunogenicity and efficacy of PfSPZ vaccines in malaria exposed populations (14,15). Specifically relevant to vac cine efficacy, DCs (the only antigenpresenting cells with the ca pacity to activate naïve T cells) are functionally compromised and undergo substantial apoptosis during clinical malaria and sub patent P. falciparum and P. vivax infection (100)(101)(102). Both malaria species may also induce changes to T FH cells that may impair B cell help (103). Specifically, parasitedriven inflammation and IFN production appears to skew T FH development to T H 1like subsets (104), which have reduced capacity to activate naïve B cells (105). These changes to T FH subsets may be linked to the induction of atypical memory B cells that have impaired function, which may be an important constraint on the development of vaccinemediated immunity (106). Changes to CD4 T cells during Plasmodium infec tion results in increased IL10-producing CD4 T cells (107), which is linked to the induction of immunoregulatory networks. In some cases, specific immunosuppressive pathways induced in malaria are also found in cancer, and in the future, it may be possible to target these to modulate responses and improve vaccine efficacy [reviewed in (108)]. Other parasite mediated cellular responses that may affect vaccineinduced immunity include changes to  T cells (109) and induction of trained innate immunity (110). Immunomodulation driven by bloodstage parasites may also disrupt preerythrocytic immunity (111).

Improving RTS,S vaccine efficacy
Given the demonstrated efficacy of RTS,S in young children, inves tigating strategies to improve efficacy and durability is a priority. Recent modeling studies suggest that increasing initial efficacy may have a more substantial decrease in childhood clinical cases and higher public health impact than increasing durability (112). New data are emerging that modifying the RTS,S vaccine schedule may improve efficacy. A regimen with a fractional (reduced) third dose given at 7 months rather than a full dose at 2 months was more effi cacious in a CHMI phase 1/2a trial in malarianaïve adult volun teers and generated increased antibody avidity and B cell somatic hypermutation (113). However, a previous study in infants found that a 027month schedule (full dose) had lower efficacy than a 012month schedule (114). Recently, a modified RTS,Slike vac cine construct was generated (R21), which, unlike RTS,S, does not require excess hepatitis B surface antigen expression for VLP for mation. Initial studies suggest that this generates higher magni tudes of antibodies and showed efficacy in preclinical studies in mice (115).
Other strategies could include adding additional alleles of CSP, because polymorphisms affect vaccine efficacy (93) including addi tional regions of CSP (91,92), or adding additional antigens. Different adjuvants may generate an IgG subclass profile with greater func tional activity. The lack of an established correlate of protection and incomplete understanding of the mechanisms of immunity remain a substantial impediment to advancing RTS,S efficacy. Systems bio logy approaches have been useful in identifying factors that predict responses with several existing vaccines and have recently been ap plied to identify transcriptional features related to RTS,S immune responses (116). Furthermore, understanding the potential impact of host factors on vaccine immunogenicity and basis for the wide diversity in vaccine responses seen among vaccinated children would enable approaches to generate more consistent protective responses and improved boosting. In addition to previous malaria exposure and immunomodulation, these may include nutrition, microbiome, and genetic factors.

GENERATING LONG-LASTING IMMUNITY
Identifying and developing strategies to generate more durable and longlasting protective immunity is an additional high priority ( Table 3). All malaria vaccines to date have shown relatively short lived protective efficacy. RTS,S has shown moderate to high initial vaccine efficacy that wanes quickly (3), suggesting that strategies to increase duration of responses would be valuable.

Memory B cells and long-lived antibody-secreting cells
The immunologic basis for durable immunity and how to generate sustained protection by vaccines remains poorly understood. For antibody responses, longlasting protective immunity requires the induction of memory B cells that can mount a recall response upon reinfection and the generation of longlived antibodysecreting cells (ASCs) that maintain circulating antibodies. Antibodies can also be maintained by the reactivation of memory B cells by polyclonal stim ulation via microbial products such as lipopolysaccharide or un methylated singlestranded DNA, which activates B cells via TLR4 and TLR9, or by bystander T cell help into shortlived ASCs that boost antibody production (117). Maximizing the initial induction of antibody functional activity, including optimal targeting of the most important epitopes, will be important, and the generation of multifunctional antibodies may help maintain efficacy over time. Studies show that recruitment of complement factors enables inhib itory or neutralizing activity at low antibody concentrations (43,50), which may help maintain immunity as antibodies wane.
Antibody responses to malaria antigens are broadly regarded as relatively shortlived in the absence of ongoing exposure. For exam ple, naturally acquired antibodies to merozoite antigens had reported mean halflives of 0.6 to 8 years (depending on antigen) compared to extremely longlived measles responses (457 years) in the same individuals (118). Models suggest that both shortlived and long lived ASCs are responsible for circulating malaria antibody titers (119). The shortlived nature of acquired or vaccineinduced re sponses suggests suboptimal induction or maintenance of longlived ASCs and memory B cells. This is in contrast to other vaccines that induce longlived antibodies, for example, halflives of 92 years after vaccinia vaccination, 11 years for tetanus, and 19 years of diphtheria, and many decades for measles (120). However, recent studies of naturally acquired immunity have shown that, despite populationlevel shortlived responses, some individual children and adults do develop very longlived antibodies and duration of re sponses differ between individuals, antigens, and response types (121), providing a proof of principle that durable immunity should be achievable by malaria vaccines.
Consistent with short duration of malaria antibodies, evidence from naturally acquired immunity suggests that the generation of memory B cell responses during malaria is suboptimal. In high transmission settings, memory responses appear to require repeated P. falciparum infection and expand gradually over many years of exposure (122). However, like longlived antibodies, robust memo ry B cell responses have been identified in some individuals from both high and low malaria transmission regions (122,123). Encour agingly, memory B cells were induced by vaccination with RTS,S/ AS01 in Gabonese children (124) and have been observed in whole parasite vaccine trials (125). Understanding the underlying basis for interindividual variation in induction and maintenance of longlived protective responses is a key next step in identifying cellular events that have the potential to be manipulated to enhance vaccineinduced immunity. Recently, specific microRNA signatures that vary be tween individuals that predicted early control of parasite growth in CHMI, the induction of CD4 + T cells and B cells, and the develop ment of antibody responses (126) were identified.
Although further insights are needed to develop effective strate gies, there are promising data on using adjuvants and TLR agonists to generate more durable immunity. A recent study in rhesus ma caques using a Pfs25synthetic particle vaccine demonstrated that the inclusion of TLR agonists R848 or CpG with the vaccine gener ated antibodies with significantly longer halflives associated with induction of type I IFN polarization (127). With influenza, the use of AS03 adjuvantinduced memory B cells, CD4 + T cells, and long lived antibodies in humans (128), and the use of MF59 adjuvant in humans, has also been associated with longlived antibodies and memory B cells (129). However, efficacy of the CSPbased vaccine R21 in mice was lower for MF59 than for other adjuvants (115). This difference from the influenza studies may possibly be due to a difference in responsiveness in human compared to animal models. Furthermore, combining multiple synergistic TLRs may increase plasma cell and antibody persistence (130).

Targeting T cell responses in memory development
T FH cells are a key T cell subset required for the development of memory B cells and longlived ASCs (131), and studies of influenza vaccination suggest that T FH subsets influence the induction of pro tective antibodies and memory B cells (69). The importance of T FH in immunity to Plasmodium has been suggested using mouse mod els (132). To date, a direct link between specific T FH activation and protective antibody induction in malaria in humans has not been established.
Although research into improving T FH activation during vacci nation is in its infancy, there are indications that this is possible. For example, Clec9A antigen targeting of DCs promotes T FH generation and antibody development in mice (133), including T FH memory responses that reactivate after secondary challenge. In addition, the TLR9 agonist CpG increased T FH activation in vitro (134). CpG has been tested in a phase 1 trial of malaria vaccine candidates AMA1C1 and MSP1-42-C1 in malarianaïve adults; the addition of CpG7909 increased antibody and memory B cell development (135). However, large gaps in our understanding of the key underlying mechanisms of CpG activity remain. Recent studies found that, while CpG en hanced the magnitude of antibody responses to protein vaccines, it also blocked the ability of antigenspecific B cells to capture, pro cess, and present antigen; to activate T cells; and thus failed to pro mote affinity maturation, including in the AMA1C1 trial (136). To harness strategies to target T FH in malaria vaccination, it is impera tive to establish which specific subsets of T FH are involved in protec tive malaria antibody induction and whether these responses are disrupted in malaria exposed populations.

Boosting of immunity by infection
In maintaining longlived vaccine responses, it would be beneficial for recurrent parasite exposure to boost immunity and help main tain vaccine efficacy. An appeal of bloodstage vaccines is that boost ing is likely to occur with repeated exposure because this is observed with naturally acquired immunity. With RTS,S, natural infections appear not to significantly boost vaccine responses and vaccine im munity wanes relatively quickly, reflecting the lower rate of acquisition of preerythrocytic immunity compared to bloodstage immunity (43). A possible explanation for this is that the number of sporo zoites inoculated by mosquitoes is simply too low to provide the required antigenic stimulus necessary to boost memory B cells, al though other factors may be important. Further research to under stand how boosting of responses by infection could be achieved might be highly valuable toward maintaining preerythrocytic im munity. A similar lack of antibody boosting was observed in an AMA1 vaccine study where monkeys were administered three doses of recombinant P. knowlesi AMA1 vaccine, each followed by an in fectious bloodstage challenge (137). Memory T cells were boosted by infection and evident for at least 200 days, highlighting differ ences across immune responses. Further investigation of possible cellular mechanisms underlying the lack of boosting (such as aner gy of malariaspecific memory B cells or failure to induce memory T FH ) needs to be examined. In contrast, a trial in European adults vaccinated with merozoite antigens MSP1 and/or AMA1 (viral vec tored) found that CHMI did lead to boosting of antibodies (138). In naturally acquired immunity, antibody concentration and rate of boosting by infection show variability depending on the antigen (139). Of the leading transmissionblocking candidates, Pfs230 and Pfs48/45 are targets of naturally acquired antibodies, whereas Pfs25 (or Pvs25) is not expressed in the human host; therefore, there is no opportunity for boosting with repeated exposure.

ASSESSING VACCINE EFFICACY IN CLINICAL TRIALS
Increasingly, vaccines are evaluated for efficacy using combinations of phase 2 fieldbased trials and CHMI trials. Phase 2 trials are ideal because they assess vaccine efficacy under conditions of natural ex posure, with relevant antigen diversity and infecting dose conducted in representative target populations. However, disadvantages include costs and logistical challenges, larger sample size requirements, co morbidities that may affect immune responses and vaccine efficacy, and inability to control timing and level of exposure. Therefore, CHMI is a valuable additional strategy to assess candidates, formu lations, and dosing before proceeding to fieldbased phase 2 trials and to identify correlates of protection. The CHMI model has been recently expanded to include assessment of transmissionblocking interventions (140). However, interpretation of earlyphase challenge data requires careful consideration (33). Although CHMI has been well established as a model for preerythrocytic vaccines, its utility for assessing bloodstage vaccines remains unclear; currently, there is a lack of data on bloodstage vaccine efficacy in CHMI versus phase 2 trials. Furthermore, the great majority of CHMI trials have been performed with only one P. falciparum strain, NF54.
An issue for consideration relates to differences in clinical thresh olds for parasite density between healthy malarianaïve adults in CHMI studies and children in malariaendemic countries, the main target population. The symptomatic threshold for children in many endemic countries is typically much higher and can vary widely be tween 100 and 20,000 parasites/l (141). In naïve volunteers, vac cine trials are halted at a low parasitemia for safety reasons (about 10 parasites /l), typically ~10 days after inoculation and at about the same time that cellular immunity is only starting to have an effect. Thus, the lack of efficacy in CHMI should only be cautiously used as evidence that the vaccine will not work in an endemic country, par ticularly if the vaccine has induced a potent cellular immune response. Symptom threshold depends on numerous factors including anti parasite immunity (that will reduce the number of parasites), anti toxin immunity, and immune tolerance, whereby inflammatory immune responses that not only limit parasite growth but also cause immunopathology are dampened. Antibodies to parasites and para site toxins or inflammatory mediators may decay at different rates after the last infection and may be differentially boosted. Maternal microchimerism, where children obtain maternal leukocytes and DNA during pregnancy or childbirth, can also alter sensitivity to malaria. Children with more maternal immune cells were more like ly to acquire malaria in early childhood but were less likely to devel op clinical symptoms, and maternal malaria was associated with increased microchimerism (142). Maternal malaria exposure also induces the differentiation of antigenspecific fetal effector T cells that are associated with protection from malaria in the first year of life (143). Recent investments in establishing CHMI capacity in malariaendemic settings will help address these issues (144).

VACCINE IMPLEMENTATION
Future implementation of malaria vaccines will occur together with other control interventions, including insecticidetreated nets, in door residualinsecticide spraying, intermittent preventive malaria treatment or chemoprophylaxis, and interventions such as mass drug administration or screenandtreat campaigns. Hence, high vaccine efficacy may not be essential if vaccines are integrated with other interventions. RTS,S may be valuable in reducing seasonal malaria given its relatively high initial, but shortterm, efficacy (145), although impacts on the broader EPI will need to be considered. A further consideration is that as transmission declines, disease bur den shifts to older children and adults (146) because of a loss of ac quired immunity. Similarly, the implementation of RTS,S or other vaccines in young children might lead to a shift in disease burden to older age groups. This may affect on the target population profile for vaccines over time.
Currently, there is a low coverage of existing malaria interven tions in many countries (e.g., only half of the atrisk population sleep under a bed net) (1), and similarly, coverage of EPI vaccines is low in many malariaendemic countries (especially measles given in the second year of life). This highlights challenges beyond vaccine development and the need for advances and innovations that would enhance vaccine implementation and integration of malaria vaccines into existing EPI schedules and malaria control interventions. Sim plified and adaptable dosing schedules, or coformulation of malaria with other vaccines, would be an advantage. Greater durability of malaria vaccine efficacy would reduce demands on health care sys tems and provide flexibility for integration into EPI. Increased vac cine thermostability accommodates interruptions in the cold chain to allow implementation under the new WHO Controlled Tempera ture Chain, which can facilitate better coverage and reduce costs (147). Vaccines requiring more frequent schedules, primeboost regimens, altered dosing regimens, or intravenous delivery all present sub stantial implementation barriers. Community perspectives on vac cine implementation and delivery are also important considerations. Research investments in these areas may be crucial for the longterm success of malaria vaccines, and considerations around future im plementation early in vaccine development may be beneficial in the long term.

CONCLUSIONS AND FUTURE DIRECTIONS
The need for a highly effective and longlasting malaria vaccine against P. falciparum and P. vivax remains strong. Recent demon stration of efficacy in endemic populations with the RTS,S vaccine and significant efficacy of several candidates in CHMI models have been important milestones in the quest for effective malaria vaccines. However, substantial challenges remain to advance beyond the gen erally modest efficacy demonstrated by existing vaccines tested in malariaendemic areas and achieve the goal of a highly efficacious and durable vaccine, preferably for both P. falciparum and P. vivax. A deeper understanding of mechanisms and key targets of immuni ty is needed to underpin this, and research to reveal new strategies for the induction of a higher level of protective functional immunity. Achieving higher efficacy may require vaccines to induce multiple functional mechanisms and may require the inclusion of multiple antigens (of the same stage or across multiple stages), whole attenu ated parasites, or innovative vaccine design to induce responses that better target specific protective epitopes compared to established strategies. A lack of correlates of protection to evaluate current vac cines and inform the development of new candidates is a constraint that hinders more expedient progress. Furthermore, most vaccines to date have induced protective immune responses of only short or modest duration, and new insights into how longlasting immunity can be generated are urgently needed. Ideally, future vaccines will be low cost and affordable and administered in practical formula tions and regimens suitable for mass vaccine campaigns and inclu sion in childhood EPI. Although challenges remain, recent exciting progress and emerging knowledge promise great hope for the future.