Dendritic Cells and Their Multiple Roles during Malaria Infection

Dendritic cells (DCs) play a central role in the initiation of adaptive immune responses, efficiently presenting antigens to T cells. This ability relies on the presence of numerous surface and intracellular receptors capable of sensing microbial components as well as inflammation and on a very efficient machinery for antigen presentation. In this way, DCs sense the presence of a myriad of pathogens, including Plasmodium spp., the causative agent of malaria. Despite many efforts to control this infection, malaria is still responsible for high rates of morbidity and mortality. Different groups have shown that DCs act during Plasmodium infection, and data suggest that the phenotypically distinct DCs subsets are key factors in the regulation of immunity during infection. In this review, we will discuss the importance of DCs for the induction of immunity against the different stages of Plasmodium, the outcomes of DCs activation, and also what is currently known about Plasmodium components that trigger such activation.


Introduction
Malaria is the most important protozoan parasitosis in humans. It afflicts millions of people annually causing an expressive burden, mainly in tropical countries. Plasmodium has a very complex life cycle, and its different stages alternate between the vertebrate host and the mosquito. The vertebrate infective stages are able to move around in the skin but also traverse and infect cells in tissues. During this journey, Plasmodium interacts with DCs that are responsible for the initiation of adaptive immune responses. The interactions among DCs and the parasite are complex and may shape the outcome of the anti-Plasmodium immune responses.

Dendritic Cells and Their Role in the Induction of Immune Responses
DCs are a distinct lineage of mononuclear phagocytic cells specialized in antigen presentation. They show excellent ability to capture, process, and present antigens to T cells [1], directly activate B cells [2], and are also involved in the amplification of innate immune responses, such as activation of NK cells [3,4]. Once in contact with a pathogen, DCs undergo a process known as maturation that culminates with efficient antigen presentation and cytokine production. Cytokines produced by DCs become part of the microenvironment that induces immune responses capable of stimulating the development of effector T lymphocytes [5]. In addition, DCs are involved in tolerance development in the thymus by negative selection of autoreactive lymphocytes [6] and in the periphery, where they present self-antigens in the absence of inflammation [7]. This entire range of DC functions is associated with their ability to recognize pathogen-or damage-associated molecular patterns (PAMPs or DAMPs, resp.) through pattern recognition receptors (PRRs) [8]. Different classes of PRRs were discovered in the last decades and include membrane anchored receptors such as tolllike receptors (TLRs) [9] and C-type lectin receptors (CLRs) [10], besides the cytoplasmic nucleotide-binding oligomerization domain-(Nod-) like receptors (NLRs), RIG-I-like and pDCs. In the blood and in lymphoid tissues, mDCs are CD3 − CD14 − CD19 − CD20 − CD56 − HLA-DR + CD11c + [27] and can be further subdivided into two subsets that express CD1c (BDCA1) or CD141 (BDCA3) [28]. On the other hand, pDCs are CD3 − CD14 − CD19 − CD20 − CD56 − HLA-DR + CD11c + CD303(BDCA2) + CD304(BDCA4) + [29]. In the skin, three populations can be found: CD1a + CD14 − DCs, CD1a − CD14 + DCs, and LCs [15,30]. Their functions in the steady state or during inflammatory/infectious diseases are beginning to be elucidated only recently [31].

Malaria
Malaria is an infection caused by protozoa that belong to the phylum Apicomplexa, genus Plasmodium. This unicellular parasite is transmitted to humans through the bite of an infected mosquito. Approximately 200 million cases of this disease are reported annually causing half a million deaths. Most of the deaths occur among children living in Africa. Malaria is prevalent in tropical and subtropical regions and is currently endemic in parts of America and parts of Asia and in Sub-Saharan Africa. In 2013, 97 countries reported endemic malaria [32].
There are five species of Plasmodium that infect humans: P. falciparum, P. vivax, P. malariae, P. knowlesi, and P. ovale [32]. The parasitic cycle begins when infected mosquitoes transfer sporozoites (the infective forms of Plasmodium) during the blood meal. These forms can remain in the skin for hours, slowly reaching circulation [33,34]. In the circulatory system, sporozoites are carried to the liver, where they invade and replicate within hepatocytes [35]. Replication gives rise to merozoites that are released from the liver in structures named merosomes and can subsequently invade erythrocytes [36]. This phase of the cycle is known as erythrocytic phase and is when malaria symptoms start. At this stage, schizogony is repeated at specific intervals depending on the species and the febrile seizures correlate with the release of merozoites in the circulation. After a few days, some merozoites that infect erythrocytes give rise to male and female gametocytes that once taken up by the mosquito continue the cycle in the invertebrate host.
Severe cases of malaria are generally caused by P. falciparum. Infection with this parasite may progress to cerebral malaria, and infected individuals often present neurological symptoms such as convulsions and coma. In addition, patients with severe malaria may also present abnormal posture, respiratory syndrome, severe anemia, and multiple organ failure [37,38]. The fact that severe malaria is usually associated with P. falciparum may be related to the potential of this species to produce hyperparasitemia. On the other hand, P. falciparum is the only species that clearly produces alterations in the microcirculation, allowing the parasite to escape destruction in the spleen. For example, erythrocytes infected with P. falciparum have the ability to adhere to the microvasculature. This phenomenon is known as cytoadherence and is mediated by molecules expressed by the infected erythrocyte that are able to bind to a series of endothelial receptors [39], such as CD36 and ICAM-1 [40,41]. In addition, infected erythrocytes are able to bind to other Conventional DCs infected and also noninfected erythrocytes, in a phenomenon known as rosetting. In this case, there is formation of cell aggregates that also interfere with the microcirculation [42]. The immunologic memory generated during infection with Plasmodium spp. is most often transient and restricted to patients living in endemic areas due to frequent exposure to the parasite by bites of infected mosquitoes [43]. In other words, naturally acquired immunity is not sterilizing and requires the persistence of the parasite to maintain the population of memory cells [44]. Several evidences suggest that naturally acquired protective immunity against malaria is obtained after successive infections [45]. Children intensely exposed to transmission have successive clinical episodes of malaria. With increasing age, clinical symptoms are less pronounced, although individuals may have high blood parasitemia [46]. In general, naturally acquired immunity is partly strain specific and primarily leads to a reduction of mortality rates and incidence of complications and later to a decrease in the incidence of disease. Finally, this naturally acquired immunity leads to a drop in parasitemia to low or even undetectable levels by conventional detection methods [47].
There is evidence that protective immune responses against malaria (sporozoites or blood stages) are initiated when antigen-presenting cells, DCs or macrophages, internalize the parasite and process and present its antigens to T cells via MHCI (through cross-presentation) or MHCII in a proinflammatory environment in which IL-12, TNF , and IFN are produced. During the blood stage infection, it was shown that these cytokines act synergistically activating macrophages to produce reactive oxygen species (ROS) and nitrogen leading to parasite death [48,49]. CD8 + T cells have also been implicated in the protective immune response against sporozoites of Plasmodium [50,51]. Several studies have shown that the CD8 + T cells play a vital role in immunity against the preerythrocytic phase of Plasmodium both in mice [52] and in humans [53]. CD8 + T cells eliminate infected hepatocytes and are capable of producing TNF and IFN [54]. The role of CD4 + T cells during Plasmodium infection has also been extensively studied in humans and mice. They seem to be particularly important during the erythrocytic phase, when IFN producing CD4 + T cells are required for the elimination of iRBCs. In addition, follicular helper T (Tfh) cells were also shown to be pivotal for the activation of antibody producing B cells [55].
Our understanding of the immunity and immunoregulation that develop during malaria is still incomplete. DCs act in the immune response to activate and/or regulate production of proinflammatory or regulatory cytokines that may have fundamental roles in regulating the acquisition of protection or in the exacerbated response observed in severe malaria patients. In the following sections, we will discuss what is known about DC participation during Plasmodium infection. Much of this knowledge was generated using mouse models that can only be infected with murine Plasmodium species. Despite that limitation, mouse models have helped us to understand how DCs influence the anti-Plasmodium response.

Role of Dendritic Cells during Plasmodium Infection
DCs ability to present antigens during malaria infection was recently reviewed [48]. As mentioned above, DCs are responsible for T cell priming and thus regulate the development of adaptive immune responses [56]. DC function has been extensively studied during infection with different species of Plasmodium, and, in some aspects, contradictory data was obtained. The reasons for such contradictions could be ascribed to the use of different Plasmodium species and stages, as well as differences in the DC activation status. We will discuss the data in more detail in the sections ahead.

DCs during Acute Plasmodium Infection in Humans.
The first study analyzing DC function directly in the context of P. falciparum infection was reported by Urban et al. These authors incubated monocyte-derived DCs with infected red blood cells (iRBCs) and showed that they were able to bind to human DCs and inhibit their maturation, reducing their ability to stimulate T cell responses [57]. These results were questioned by Elliott et al. when they showed that in fact inhibition of DC maturation was only obtained when iRBCs were used in a 100 : 1 iRBC : DC ratio and was not contact dependent. In addition, when a lower dose of iRBC (10 : 1 ratio) was used, DCs matured efficiently and activated autologous T cell proliferation [58]. In a field study with Kenyan children, BDCA3 + DCs were significantly increased during acute infection, while CD1c + DC numbers were unaltered when compared to healthy individuals [59,60].
In addition, an association between increased numbers of circulating BDCA3 + DCs and severe human malaria was also observed [60,61]. Although more prevalent, infection with P. vivax is more benign and less studied. However, a few studies were performed in an attempt to examine DC status and function during acute and symptomatic P. vivax infection. For example, the numbers of pDCs (HLA-DR + CD123 + ) and mDCs (HLA-DR + CD11c + ) were evaluated in infected individuals from Thailand and Brazil, and a decrease in the mDCs/pDCs ratio was observed in both studies [62,63]. In addition, 1/3 of the P. vivax-infected Brazilian patients showed low surface expression of CD86 [62]. A similar reduction in CD86 levels, as well as in CD83 and HLA-DR, was observed in Indonesian patients infected with P. falciparum and P. vivax [64]. This reduction correlated with an increase in DCs spontaneous apoptosis and impairment in their ability to capture, mature, and present antigens to T cells [64]. Interestingly, when patent asymptomatic patients were studied, HLA-DR expression was preserved in groups infected with either P. vivax or P. falciparum [65]. This result may indicate that DC function is preserved in patients that are infected but do no present symptoms, suggesting that functional DCs are important for the maintenance of clinical, but not parasitological, immunity.

Are DCs Capable of Inducing T and B Cell Responses during Plasmodium Infection?
Human and murine species of Plasmodium have been used in studies designed to harness the participation of DCs in the induction of immunity against the preerythrocytic stages. DCs pulsed with sporozoite extracts were shown to elicit specific killing of P. vivax exoerythrocytic stages within infected hepatocytes [66], while DCs pulsed with a well-characterized CD8 + T cell epitope derived from the P. yoelii circumsporozoite protein reduced the liver burden in BALB/c mice after a sporozoite challenge [67]. This CD8 + T cell response was abrogated when DCs were depleted in vivo [68]. Moreover, after an infectious mosquito bite, the CD8 + CD11b − DCs located in the draining lymph nodes were shown to be the DC subset responsible for CD8 + T cell priming in vivo [69,70] (Figure 2). In the P. berghei model, DCs pulsed with irradiated sporozoites were able to similarly prime central memory CD8 + T cells when compared to DCs primed with untreated sporozoites. However, irradiation enhanced sporozoites' ability to prime effector CD8 + T cells capable of producing IFN . In this particular study, the fine specificity of CD8 + T cells was not evaluated [71]. A more detailed study showed that the two major splenic DCs subsets (CD8 + CD11b − and CD8 − CD11b + ) induced IFN producing CD8 + T cells specific for the circumsporozoite protein, the major sporozoite surface protein [72]. In addition, the CD8 + T cell protective response against a genetically modified P. yoelii strain was shown to be dependent on effective DC maturation obtained through CD40 signaling [73]. The fact that DCs are able to induce CD8 + T cell responses against antigens expressed in the preerythrocytic stage seems undisputable. However, an observation made by Ocaña-Morgner et al. added another layer of complexity to the picture. These authors observed that DCs from mice previously infected with P. yoelii strain 17XNL and undergoing erythrocytic cycle presented an immature phenotype and were unable to initiate CD8 + T cell responses to subsequent liver-stage antigens [74]. An unidentified soluble factor released by the iRBC seemed to be responsible for the inhibition of DCs maturation [75]. Cross-presentation to CD8 + T cells was also inhibited during active P. berghei blood infection [76] and was mainly dependent on the CD8 + CD11b − DCs subset [77]. Taken together, these results suggest that blood stage infection with Plasmodium can impair the development of an effective CD8 + T cell response against the liver stages, which then could clear the parasite upon reinfection.
The role of DCs in the activation of CD4 + T cell responses was also evaluated during P. yoelii infection (Figure 2). DCs derived from infected mice (day 6) presented higher expression of the surface costimulatory molecules CD80 and CD40 and were able to efficiently present antigens to CD4 + T cells that in turn produced higher levels of IL-2, IFN , and TNF . Production of these cytokines required DC-derived IL-12 [78,79].
Infections with P. chabaudi have also shed light on the role of DCs during malaria. In vitro studies showed that P. chabaudi schizonts induced bone marrow-derived DCs to express MHCII and costimulatory molecules and to produce IL-6, IL-12, and TNF [80]. An in vivo study showed that DCs migrated from the marginal zone of the spleen into the T cell area within 5 days after infection, and by day 7 an increase in costimulatory molecules was observed [81]. The role of CD8 + CD11b − and CD8 − CD11b + DCs subsets in ↓ BAFF expressing DCs Figure 2: Plasmodium infection activates different DCs subsets. Infection with different stages of Plasmodium activates DCs that in turn start producing proinflammatory (IL-6, IL-12, and TNF ) but also regulatory (IL-10) cytokines. The CD8 + CD11b − DCs are responsible for priming CD8 + T cell responses against intrahepatic forms, while both cDCs subsets (CD8 + CD11b − and CD8 + CD11b + ) play a role in the activation of CD4 + T cell responses that can lead to an inflammatory or regulatory outcome, depending on the timing of infection. pDCs also play a role in the induction of a more regulatory CD4 + T cell response. The number of BAFF expressing DCs is reduced during Plasmodium infection, and that may reduce their ability to support B cell differentiation directly. Based on [69, 70, 78-80, 82, 85, 86, 91].
antigen presentation and specific CD4 + T cell activation was also investigated. Despite the fact that both subsets induced IFN production, only the CD8 − CD11b + DCs isolated at the infection peak (day 7) were able to induce proliferation of Plasmodium-specific Tg CD4 + T cells and considerable amounts of IL-4 and IL-10 [82], indicating that this subset could be responsible for the switch in the balance from the proinflammatory Th1 response seen in the first few days to a more pronounced Th2 response. However, on days 10 and 13 after P. chabaudi infection, the CD8 − CD11b + DCs were no longer able to induce CD4 + T cell proliferation or cytokine production due to downregulation of costimulatory molecules and IL-12 expression [83]. Nevertheless, the in vivo CD4 + T cell response later recovers and the mice are able to control infection, and at this stage antibodies seem to play an important role in the parasitemia control [55,84]. pDCs were also studied in this model and, despite an increase in their numbers during infection, they played no role in CD4 + T cell activation or in the control of infection [85].
On the other hand, in a P. yoelii model of infection, pDCs numbers increased by day 6 and remained high until at least day 14, and they were able to induce IL-10-expressing CD4 + T cells [86]. Taken together the results discussed above suggest that different DC subsets have different functions during the blood stage Plasmodium infection, with cDCs involved in the induction of the proinflammatory response and pDCs accounting for a more balanced response at a later stage ( Figure 2). As reported for P. falciparum [57] and P. yoelii [74], DCs were also able to bind and internalize P. chabaudi iRBC. This phenomenon was shown to be partially dependent on actin polymerization, and the iRBCs uptake was again associated with increased expression of MHCII and costimulatory molecules, IL-12 production, and stimulation of CD4 + T cell proliferation and IFN production [87]. Moreover, a recent study using intravital microscopy in mice showed that splenic DCs not only interact with CD4 + T cells in T cell rich areas and in the red pulp but also very actively phagocytose iRBCs, contributing directly to their elimination during acute infection [88].
The function of DCs in the development of cerebral malaria was also assessed using a model of P. berghei ANKA strain infection in C57BL/6 mice. In this setting, cDCs were shown to play a major role in the induction of cerebral malaria [89]. A more detailed study showed that the CD8 + and CD103 + DCs subsets are essential to induce the pathogenic CD8 + T cells responsible for lethal brain inflammation during P. berghei ANKA infection [60]. A recent set of experiments showed that the numbers of cDCs and pDCs were drastically reduced during P. berghei ANKA infection, through a mechanism that involved activation of caspase-3 and induction of DC apoptosis. In this particular case, the function of the remaining DCs was not evaluated [90].
DCs also play an important role in direct B cell activation ( Figure 2). This happens because they produce a cytokine known as B cell activating factor (BAFF) that enhances B cell differentiation and survival [2]. During P. yoelii infection, there is a decrease in the percentage of DCs expressing BAFF, resulting in a reduction of their ability to support memory B cell differentiation into antibody secreting cells [91].
Finally, besides T and B cells, DCs are also able to interact with innate immune cells such as NK and T cells during Plasmodium infection. The depletion of NK cells during P. berghei ANKA infection led to a significant reduction in DCmediated CD8 + T cell priming but did not affect CD4 + T cells. It seems that NK cells stimulate DCs to produce IL-12 that in turn is required for optimal T cell priming. The effect of DCs on NK cell function was also evaluated, and DC depletion reduced NK cell-mediated IFN responses to this Plasmodium species [92]. In the same way, T cells can also communicate with DCs and do so when they express CD40L and produce IFN that in turn enhance DC activation [49].
The data discussed above indicates that DCs may play distinct roles during Plasmodium infection, promoting either activation of protective immune responses or exacerbation of pathology.

How Do DCs Recognize Plasmodium Components?
The expression of TLRs on DCs was studied in patients infected with P. falciparum. Patients with mild and severe forms of the disease displayed increased surface expression of TLR2 and TLR4 on mDCs and decreased intracellular expression of TLR9 on pDCs, when compared to healthy controls [93]. Despite this decreased TLR9 expression on pDCs, another study showed that the TLR9-MyD88 signaling pathway was required for pDC activation upon stimulation using schizonts or soluble schizont extracts [94]. The possible ligand for this pathway will be discussed below. A study using P. chabaudi infected mice showed that DCs from MyD88 knockout mice, but not from TLR2, TLR4, TLR6, TLR9, or CD14 knockout mice, were unable to produce proinflammatory cytokines and induce CD4 + T cell responses [95]. This result indicates that the adaptor molecule MyD88 is required, but different ligands may be signalizing through different TLRs.
As mentioned before, DCs express different classes of PRRs capable of recognizing PAMPs derived from a vast array of pathogens [8]. In the case of Plasmodium, three PAMPs have been more extensively studied: hemozoin, immunostimulatory nucleic acid motifs, and glycosylphosphatidylinositol (GPI) anchors [96].
When Plasmodium invades the RBC, it degrades hemoglobin as a source of amino acids, which in turn releases heme that is potentially toxic. To survive, the parasite detoxifies heme into hemozoin using an enzyme named Heme Detoxification Protein (HDP) [97]. Contrasting results have been obtained and the role of hemozoin is still a matter of debate. Initially, hemozoin was shown to be present inside macrophages and circulating monocytes during P. falciparum infection, reducing their ability to phagocytose other particles or generate oxidative burst [98]. Human monocytes loaded with hemozoin and then in vitro differentiated into DCs presented impaired surface expression of MHCII and costimulatory molecules [99,100], while monocyte-derived DCs incubated with synthetic hemozoin upregulated costimulatory molecules and released proinflammatory (IL-6) and anti-inflammatory (IL-10 and TNF ) cytokines, but not IL-12, leading to suboptimal T cell activation [101,102]. In a P. chabaudi model, hemozoin-containing DCs were unable to fully activate T cells that in consequence did not cluster or migrate to lymphoid organ follicles [103,104].
Different results showed that P. falciparum iRBC hemozoin, but not -hematin (a synthetic hemozoin), was able to induce human monocyte-derived DCs to upregulate costimulatory molecules (CD83, CD86, and CD1a) and produce IL-12 [105]. A follow-up study by the same group showed that this effect was dependent on TLR9 activation followed by MyD88 signaling but independent of TLR2, TLR4, TLR7, or TRIF [106]. In contrast with these authors, Parroche et al. showed that hemozoin is not a direct ligand for TLR9; instead it functions as a carrier for plasmodial DNA that is phagocytosed by DCs and carried to intracellular compartments [107]. Finally, Wu et al. argued that hemozoin was not a TLR9 ligand for DCs and did not function as a DNA carrier into cells. Instead, their results showed that a protein-DNA complex was the parasite's component responsible for the DC activation through TLR9 signaling. Protein-DNA complex formation was essential for the entry of parasite DNA into DCs leading to TLR9 recognition [108]. In addition, hemozoin was also found to activate the NLRP3 inflammasome during experimental malaria infections [109,110].
P. falciparum genome is specially rich in AT-motifs. It was recently shown that these motifs have immunostimulatory properties and are able to induce type I IFNs. In fact, Plasmodium iRBCs triggered type I IFN production in macrophages [111]. DCs were not directly tested, but there is a very good possibility that they are activated in the same way.
GPI anchors are essential for Plasmodium survival and link different proteins to the parasite surface. P. falciparum merozoite-derived GPI anchors induced the production of proinflammatory cytokines and nitric oxide by macrophages [112,113] mainly in a TLR2-dependent way but also, to a lesser extent, in a TLR4-dependent way [114], which requires CD36 cooperation [115]. P. falciparum GPIs' ability to activate DCs was evaluated more recently, and a study showed that TLR2-signaling was also important for DCs activation and induction of TNF and IL-12 production. As observed for macrophages, CD36 also collaborated with TLR2 [116].
Journal of Immunology Research 7 The participation of different TLRs was also evaluated in a P. berghei ANKA model of cerebral malaria. Results showed that mice deficient in TLR1, TLR2, TLR3, TLR4,  TLR6, TLR7, or TLR9 and their adapter proteins MyD88, TIRAP, and TRIF were as susceptive to cerebral malaria as their wild-type counterparts [117,118]. In contrast to these results, another group showed the contrary: cerebral malaria pathogenesis seems to be mediated by MyD88 signaling and its absence increases survival [119]. DCIR-a CLR that recognizes carbohydrates-deficient mice are resistant to cerebral malaria and present less CD8 + T cell infiltration and inflammation in their brains when compared to wild-type mice [120].
In addition to PAMPs, Plasmodium infected erythrocytes were also shown to accumulate uric acid, a very potent endogenous danger-associated molecular pattern (DAMP). Uric acid precipitates accumulated within P. falciparum and P. vivax iRBC are released by cell rupture and induce the maturation of human dendritic cells in vitro [121]. Taken together, these results indicate that DCs sense and are able to respond to different Plasmodium PAMPs and also to DAMPs that are generated during infection.

Harnessing DCs for the Development of Vaccines against Malaria
DCs have also been studied as targets for the development of vaccines against malaria. The circumsporozoite protein (CSP)-the major sporozoite surface protein-from P. yoelii was fused to a monoclonal antibody capable of binding to a CLR-known as DEC205-expressed on the surface of the CD8 + CD11b − DCs subset. The administration of low doses of the DEC205-CSP monoclonal antibody in the presence of a DCs maturation stimulus-CD40+polyinosinic:polycytidylic acid, poly(I:C)-was able to induce IFN producing CD4 + and CD8 + T cells, besides specific anti-CSP antibodies [122]. The use of an anti-human DEC205 monoclonal antibody fused to the P. falciparum CSP together with poly(I:C) to immunize nonhuman primates elicited anti-CSP antibodies and also multifunctional CD4 + T cell responses [123]. Another approach used to target the P. yoelii CSP to DCs was to fuse its sequence to the macrophage inflammatory protein 3 (MIP3 ) that targets the CCR6 receptor present on the surface of immature DCs. Mice were immunized with a DNA plasmid encoding the MIP3 -CSP in the presence of Vaxfectin, and protection was obtained against the challenge [124]. These results indicate that DCs may be targeted with Plasmodium antigens in an attempt to induce potent immune responses and ultimately induce longlasting protection.

Concluding Remarks
The effects of Plasmodium infection on dendritic cells are broad and normally result in DC activation and induction of potent T cell responses that also lead to B cell activation and antibody production. However, evidence has been gathered that malaria may also suppress DC function. The complexity of the parasite cycle and also the different DCs subsets contribute to increase the level of difficulty in understanding the outcome of all interactions. A multidisciplinary approach to elucidate the mechanisms involved in the activation of DCs by Plasmodium is necessary. If we understand how the parasite modulates DCs, it may be possible to manipulate this information to develop an effective vaccine against malaria.