Memory CD8+ T cell‐mediated protection against liver‐stage malaria

Nearly half of the world's population is at risk of malaria, a disease caused by the protozoan parasite Plasmodium, which is estimated to cause more than 240,000,000 infections and kill more than 600,000 people annually. The emergence of Plasmodia resistant to chemoprophylactic treatment highlights the urgency to develop more effective vaccines. In this regard, whole sporozoite vaccination approaches in murine models and human challenge studies have provided substantial insight into the immune correlates of protection from malaria. From these studies, CD8+ T cells have come to the forefront, being identified as critical for vaccine‐mediated liver‐stage immunity that can prevent the establishment of the symptomatic blood stages and subsequent transmission of infection. However, the unique biological characteristics required for CD8+ T cell protection from liver‐stage malaria dictate that more work must be done to design effective vaccines. In this review, we will highlight a subset of studies that reveal basic aspects of memory CD8+ T cell‐mediated protection from liver‐stage malaria infection.

membrane-bound merosomes that burst in the bloodstream, freeing the merozoites to target a new host cell, this time infecting erythrocytes. The merozoites then go through an asexual reproduction cycle in erythrocytes that vary from 24 to 72 h depending on the species of Plasmodium. The parasite consumes the resources of the erythrocyte and most importantly changes the shape of the cell via changes to the cell membrane. This allows nutrients that are beneficial to the parasite to preferentially enter the cell but also exposes proteins and antigens from the parasite on the surface of the cell. In a susceptible individual, the number of parasites in the blood can expand anywhere from six to twenty times per asexual cycle, causing disease when the density in the blood reaches approximately 100 million parasites. 6 Liver-stage malaria is silent, with respect to host pathogenesis.
It is the release of merozoites from the liver into the bloodstream that ultimately triggers pathogenesis in humans. 7 Initial manifestations include nonspecific signs and symptoms such as headache, muscle pains, abdominal pains, and fatigue, followed by periodic fevers. "Uncomplicated" malaria symptoms can occur in 6-10 h cycles called malaria "attacks" that are characterized by chills, followed by a high fever, and then return to normal temperature. "Severe" malaria is defined as an infection complicated by organ failures and/or abnormalities in the blood and metabolism. Destruction of red blood cells can drive severe anemia, hemoglobinuria, acute respiratory distress syndrome, abnormalities in blood coagulation, low blood pressure, kidney injury, and metabolic acidosis. Cerebral malaria (CM) is a neurological complication that occurs most commonly in children during infection with P. falciparum. By definition, cerebral malaria is characterized by the patient "being in a coma for at least 1 h following a seizure or hypoglycemic episode, with asexual forms of P. falciparum in the patient's blood smears and no other explanation for a coma." 8,9 Multiple explanations have been proposed for the clinical manifestations of severe malaria including the sequestration of infected RBCs blocking cerebral microvasculature as well as the accumulation of toxic metabolic by-products and inflammatory cytokines and chemokines in the blood. 10 Mouse models of experimental cerebral malaria (ECM) have directly implicated CD8+ T cells in the pathogenesis of CNS disease. 11 The exact pathogenic mechanisms of CM in humans remain unclear. CD8+ T cells are canonically described as cytotoxic T cells, which can limit viral, bacterial, and parasitic infection by killing infected cells by direct cytolysis via perforin/granzymes and Fas/FasL interactions or by activating ancillary immune responses through secretion of cytokines such as IFNγ and TNFα. Following recognition of the specific peptide-MHC complex by the T cell receptor, antigenspecific T cells proliferate and perform their effector functions in an acute capacity. Following expansion, the effector cells will contract in numbers; though a percentage of them will persist as memory cells at a higher frequency and lower activation threshold (relative to naïve T cells) to respond if the same antigen is encountered again. 12 Although the short duration of an initial liver-stage malaria infection limits the potential for newly primed CD8+ T cells to contribute to parasite clearance, clear evidence from the literature prior to 2005 demonstrates that parasite-specific memory CD8+ T cells can limit and even eliminate liver-stage parasites if present at the time of challenge infection. [13][14][15][16][17] The use of various vaccination strategies in both mouse models and controlled human challenge studies have solidified the role of these T cells in protection from this deadly parasite.
However, multiple questions regarding the characteristics of liver protective memory CD8+ T cells remained unanswered, which hamper the development and deployment of a highly effective vaccine.
In this review, we will focus on how studies using multiple vaccination models have enabled the study of CD8+ T cell-mediated protection, with a particular emphasis on murine models of infection. We will highlight major strides and discoveries that have been made in this vein and indicate pursuits being made for the future of the field.

| MAL ARIA IN NATUR ALLY INFEC TED HUMANS
Malaria infection is the deadliest parasitic disease of humans that has only been exacerbated by service disruption due to the COVID-19 pandemic. In 2020, the WHO estimated there were 241 million cases of malaria; a significant increase from the 2019 estimate of 227 million. 1  Due to the habitat of the Anopheles spp. mosquito, malaria is considered endemic in most tropical parts of the world, resulting in half of the world's population being at risk of exposure. 18 Human disease is caused by 5 plasmodium species: P. falciparum, P. vivax, P. malariae, P. ovale and P. knowlesi. P. falciparum is the main cause of severe and fatal malaria, accounting for more than 90% of malaria mortality. This is the dominant Plasmodium species in sub-Saharan Africa and unfortunately, the species that most commonly acquires resistance to antimalarial drugs. P. vivax is more dominant in the Americas. P. malariae and P. ovale are significantly less common (<1% of isolates) and are rarely found outside of sub-Saharan Africa. P. knowlesi has been found in southeast Asia, but it is unclear if this species can be transmitted back to mosquitoes from humans, suggesting an additional nonhuman primate vector. 19 Certain factors can make an individual at higher risk for developing severe malaria. In malaria-endemic regions, severe and cerebral malaria is generally a disease that affects children younger than 5 years of age (a phenotype that is generally attributed to lacking the acquired immunity that develops over many seasons of infections).
Pregnant women are more susceptible to severe malarial disease particularly in areas where malaria transmission over years is low or unstable, for the same reason. 20 While severe and cerebral malaria are less common in individuals who have had repeated seasonal exposures, it is important to note that this repeated exposure does not provide complete protection from parasite infection or from disease.
Despite annual exposures, individuals living in malaria-endemic regions do not typically develop natural sterilizing immunity to the parasite. It is unclear the exact reason for this, but researchers have put forth several reasonable nonmutually exclusive hypotheses.
First, in natural mosquito-borne inoculation, only tens to hundreds of malaria sporozoites are deposited into the dermis of a human. 21 Of those, only relatively few will go on to productive hepatocyte infection, while the remaining are whisked away to the skin-draining lymph node (dLN) 22 or remain in the skin. 3 As a result, a very low number of hepatocytes become productively infected with the malaria sporozoites. In fact, the overall large numbers of hepatocytes in the liver relative to the few sporozoites that infect them make the odds of finding an infected hepatocyte 1 in a million in mice and 1 in a billion in humans. 23 The importance of antigen presentation in the liver during the priming of liver-stage malaria-specific CD8+ T cell responses has begun to be appreciated in the last 5 years and will be covered in more detail later in this review. Human and murine immunization studies have supported the idea that liver antigen dose is an important factor in driving liver-stage malaria-specific CD8+ T cell responses by demonstrating that the magnitude of the CD8+ T cell responses to Plasmodium infection or whole attenuated parasite immunization is highly dependent on the quantity of antigen delivered. 24,25 Indeed, early human studies with radiation-attenuated sporozoites (RAS) (discussed further below) found that nearly 1000 bites from infected mosquitoes were required to induce sterilizing immunity, 26 which would be the equivalent of 10 malaria seasons in a highly endemic region. 27 It is thought that once the sporozoites reach the liver to engage in the "silent phase" of replication, the parasite capitalizes on the ability of hepatocytes to induce tolerance to avoid elimination as it performs early replication in hepatocytes. 28 Hepatocytes express major histocompatibility complex class I (MHC-I) and intracellular adhesion molecule-1 (ICAM-1), which are both important molecules for T cell activation and have been observed interacting with naïve T cells via electron microscopy. 29 However, studies involving transplantation have demonstrated that hepatic allografts have the capacity to not only prevent but also reverse the rejection of skin allografts, demonstrating the tolerogenic capacity of the liver allograft microenvironment. 30,31 Studies from a murine hepatitis model demonstrate that hepatocytes are poor antigen-presenting cells for naïve T cells. 32 These data suggest the hypothesis that hepatocytes, the dominant cell type infected with malaria sporozoites, are poor antigen-presenting cells and are more likely to induce T cell tolerance to liver-stage malaria epitopes, rather than a productive T cell response.
Invasion of the host hepatocyte by a sporozoite involves the production of the parasitophorous vacuole (PV), a host membranebased structure that sequesters the parasite from the host cytoplasm. It is thought that liver-stage antigens produced within the protective shell of the PV, though numerous, are not always readily accessible to the host antigen processing machinery, and as such, do not drive a robust T-cell response via hepatocyte antigen presentation. This, in combination with evidence that hepatocytes preferentially prime tolerogenic T cell responses (in the context of transplantation studies) suggest that infected hepatocytes housing epitopes hidden within the PV are not the ideal APCs for priming liver-stage antimalarial T cell responses.
There is a substantial body of work that indicates that once malaria infection progresses to the blood stage, there are significant implications for the host immune system which can preclude the generation of protective immunity via natural infection. Dendritic cells (DC), a subset of professional APC are a critical cell type for priming antimalarial CD8+ T cell responses. 33 Activation and maturation of these DCs by the triggering of pattern recognition receptors (PRRs) typically induces the upregulation of MHC-I and MHC-II, surface adhesion molecules, as well as costimulatory molecules CD40, CD80, and CD86. Studies in the late 1990s initially found that co-incubation of Plasmodium-infected erythrocytes with DCs reduced LPS-mediated surface expression of any of these critical APC molecules on the DCs. 34 Additionally, these cells poorly primed CD8+ T cell responses in vitro. The exact mechanism responsible for this is actively under investigation.
Finally, the issue of the overall vastness of Plasmodium species diversity and parasite intrahost diversity represents another barrier to naturally acquired malaria immunity over repeated exposures.
Despite a relatively low genome replication error rate (99.1% fidelity), 35 Plasmodium parasites can replicate their genomes up to 10 10 times during the erythrocytic stage of infection, permitting even such a low error rate to generate intrahost diversity. 36 Various immunological and physical bottlenecks within the host occurring during replication (e.g., antibody and T cell-mediated selection and selection under chemoprophylaxis) and cross-species transmission (e.g., human, mosquito, nonhuman primate) are posited to contribute to the evolution of these parasites. Genetic diversity is reflective of the level of Plasmodium endemicity in a given area, with areas of intense Plasmodium transmission also having greater parasite diversity. 37,38 These multiclonal populations that are capable of infecting a single individual appear to enhance the acquisition of protective immunity in endemic areas as individuals harboring polyclonal parasite populations are more likely to have asymptomatic malaria infections. 39 Individuals living in areas of low transmission rates generally are infected with a monoclonal population of Plasmodium. 40,41

| MOUS E MODEL S OF MAL ARIA
The use of Plasmodium species that are specific to rodent populations has long been a crucial tool for the study of immunology using mouse models. Plasmodium berghei was first isolated from wild tree-rats in the Congo in the 1940s and was the first Plasmodium found to infect and kill mice and rats. 42,43 Interestingly, P. berghei infection was found to have a 20% mortality rate in the wild mouse populations but had a much higher and faster rate of mortality in laboratory mice. The ANKA strain of P. berghei was isolated and shown to have an erythrocytic cycle of 43 h and to preferentially develop in reticulocytes. This strain is widely used as a model of experimental cerebral malaria (ECM) where more than 80% of certain strains of mice, including C57Bl/6, will be dead (if left untreated with antimalarials) by day 10 postinfection. 44 There are several differences between human malarial brain pathology and P. berghei ANKA-induced brain pathology. Infected mice develop characteristic cerebral lesions similar to humans but where mice have monocyte-dependent cerebral vessel obstruction, humans have erythrocyte-dependent cerebral microvasculature sequestration. 45 Another difference seen in the literature is that mice that die from cerebral malaria often die of cerebral symptoms and blood-brain-barrier breakdown, whereas few humans die due to cerebral complications specifically. 46 Another commonly used strain of rodent malaria is Plasmodium yoelii. Unlike P. berghei, some P. yoelii isolates, for example, the 17xNL strain, produce a nonlethal parasitemia that is eventually cleared in experimental mice. Similar, to P. berghei, it preferentially invades reticulocytes. P. yoelii is considered a "good model" since it can recapitulate many human symptoms of P. falciparum and P. vivax malaria and is useful for the study of liver-stage malaria, blood-stage antigens, and malarial immunology in general. 47 P. yoelii sporozoites in particular are thought to be the best analog for human sporozoites and for the study of sporozoites. 48 Plasmodium chabaudi and P. vinckei are the other two species of rodent malaria parasites discovered in sub-Saharan Africa in the mid-1900s and are widely used in mouse models of human malaria.
P. chabaudi has been used to study the genetics of malaria parasites, particularly for the genetics of antimalarial resistance since P. chabaudi is inherently sensitive to chloroquine. 49,50 P. chabaudi also induces hematological responses with similarities to P. falciparuminfected humans. 51 Of great relevance, P. chabaudi is the only rodent parasite that exhibits periodic recrudescence of blood-stage infection if not treated with antimalarial drugs, which is a feature of P. falciparum infection in humans. 52 P. vinckei is the least used experimental mouse model of rodent malaria due to its lack of phenotypic and genotypic data. 53

| CONTROLLED HUMAN MAL ARIA INFEC TI ON (CHMI)
While murine models of infection have been an absolutely critical component in studying host-parasite interactions, cellular biology, and basic immune correlates of protection from malaria, CHMI presents the opportunity to evaluate vaccine efficacy in human volunteers. 54 The first practice of experimental human malarial infection occurred from 1920 to 1950, though not for the explicit study of malaria. 54 In fact, it was used during this time as an experimental treatment for neurosyphilis. Parasitized red blood cells were taken from malaria-infected individuals (as well as sporozoites from infected mosquitoes) and were purposefully administered with the goal that the fever spikes during acute malaria attacks could kill the syphilitic bacteria and slow disease progression. 55,56 It was not until 1973, however, that controlled human malarial infections were used for the explicit study of the correlates of protection against malaria.
Seminal studies from Clyde et al. were the first to demonstrate the protective capacity of radiation-attenuated sporozoite vaccination in humans. [57][58][59] Following these studies, several key advances were made in the malaria field, including the development of methods for the in vitro culture of P. falciparum 60 and the infection of female mosquitoes from feeding on these cultures. 61 The technology made available from these studies spurred a collaboration between the National Institutes of Health and the US Army and Navy to complete the first sponsored CHMI studies.

| VACCI N E S
Targeting vaccines to the malaria preerythrocytic stage is an ideal strategy for malaria vaccination, as inhibiting liver-stage development of merozoites prevents both the pathogenic blood-stage malaria as well as downstream sexual development that perpetuates transmission back into the mosquito vector. Murine models as well as human vaccination studies have provided substantial information on the correlates of vaccine-mediated protection against malaria but have also highlighted the complexities of finding an effective vaccine (Table 1 and Figure 1).

| Radiation-attenuated sporozoites (RAS)
One of the first hints that an effective antimalaria vaccine possible was in 1967, when Nussenzweig et al. reported that a single immunization of mice with P. berghei radiation-attenuated sporozoites (RAS) could prevent detectable infection with homologous nonattenuated sporozoite challenge in 63% of mice (relative to 10% in unvaccinated controls). 63 The rationale for this strategy is to irradiate wild-type sporozoites (or mosquitoes infected with those sporozoites) to induce random mutations into the parasite DNA, such that the parasites remain viable and can infect hepatocytes but replication arrests in the early liver stage. As a result, early liver-stage antigens are This study demonstrated the development of CSP-specific T cell and antibody responses, indicating this as a correlate of protection. However, studies in individuals from malaria-endemic regions revealed that similar immunization conditions yielded sterilizing immunity in less than 30% of individuals exposed to parasites in the field, 67 highlighting an obvious shortcoming in our understanding of the immune response to these vaccines in the relevant populations. Still, RAS immunization is considered the gold standard malaria vaccine.
Early murine studies primarily focused on the elicitation of antimalarial antibodies following RAS immunization as a benchmark for vaccine immunogenicity. 68 However, CD8+ T cells appear to be the main correlate of protection elicited by RAS. The earliest studies demonstrated depletion of CD8+ T cells from RAS-immunized BALB/c mice abolished protective immunity. 69 However, the depletion of CD4+ T cells from these animals did not affect RAS-induced immunity to malaria. Moreover, CD8+ T cell clones isolated from RAS-immunized BALB/c mice specific to P. yoelii CSP could be transferred into naïve mice and display CTL-mediated protection. 70 TA B L E 1 Advantages and drawbacks of the most commonly investigated malarial vaccine approaches.

| Genetically attenuated parasites (GAP)
A more strategic approach compared with random genetic mutations induced by radiation attenuation is the use of specific genetic modifications of parasites (termed genetically attenuated parasites or "GAPs"). The successful sequencing of the Plasmodium genome made the identification of candidate genes associated exclusively with liver-stage or blood-stage development possible. 71 The theory behind successful GAP is to delete a gene or genes that are specific and essential for liver-stage development that would permit initial replication in hepatocytes but would ultimately arrest prior to the blood stage. Although RAS has long been considered the gold standard in antimalarial immunity, the benefits of GAP development cannot be understated. One substantial benefit offered by the use of GAPs is that the use of parasites with the exact genotypic properties known is advantageous for the standardization and regulation of their production and deployment. Prior to irradiation, RAS are fully infectious sporozoites capable of inducing severe malaria. Irradiation must be a sufficient dose to prevent a breakthrough to blood-stage malaria, but replication must occur sufficiently to induce protective immunity.
Multiple murine models evaluating the protective capacity of GAPs demonstrated complete attenuation and protective immunity induced by immunization. 72,73 Initial studies identified the gene candidates UIS3 and UIS4 ("upregulated in infective sporozoite gene 3 and 4," respectively) as potential malarial GAP deletion targets. 74,75 Deletion of each separately resulted in attenuation and sterilizing immunity. Later, the Plasmodium 6-Cys family of proteins was also identified as potential gene deletion targets. Specifically, deletion of a member of this family (P52) in P. falciparum was found to result in abortive replication that ultimately led to the death of infected hepatocytes, which was posited to provide access to these liver-stage antigens to APCs. 76 Later, this GAP was combined as a double knockout with another 6-Cys family protein (P36), which made the replication defect even more severe. This double knockout p52−/p36-P. falciparum GAP was tested for efficacy in a human hepatocyte chimeric mouse model, which resulted in sterilizing immunity. 77 However, the first human clinical trial using p52−/p36-P. falciparum demonstrated that this particular attenuation strategy resulted in breakthrough parasitemia in 1 out of 6 trial subjects, indicating that while the replication of the parasite was hampered, it was not completely attenuated. 78 Furthermore, the Harty and Kappe labs previously demonstrated that later arresting GAPs (or GAPs that permit further progression of replication in the liver stage), generated more robust antimalarial CD8+ T cell responses. 79 This was attributed to the idea that the further along in the liver-stage replication cycle the parasites were, the more potential antigens would be produced, and effectively a larger response with a larger breadth of coverage would result The combination of these findings indicated that while GAPs appear to be a promising avenue for the development of broad and complete antimalarial immunity, identification of rational gene targets for complete attenuation remain a challenge.

| Chemoprophylaxis and sporozoite (CPS)
Another whole sporozoite vaccination strategy, termed "chemoprophylaxis and sporozoite" (CPS) treatment involves infection with nonattenuated sporozoites under "prophylactic cover" of chloroquine (CQ) treatment. 80 This vaccination method permits liver-stage replication but arrests the parasite before blood-stage replication progresses to the point of disease. Early human trials involved the exposure of volunteers to 3 monthly exposures of 12-15 mosquito bites using mosquitoes infected with P. falciparum while being administered weekly 300 mg doses of the antimalarial drug chloroquine. Eight weeks following the last immunization, the subjects were challenged with 5 bites from P. falciparum-infected mosquitoes in the absence of chloroquine cover. These initial human challenge studies reported sterile protection in 100% of vaccine recipients; a promising outcome. Interestingly, when subjects were immunized with the same regimen, but challenged with blood-stage merozoites, no immunity was observed, which supports the notion that immunity that develops under RAS, GAP, and CPS treatment is likely induced via the priming of preerythrocytic immune responses that are somewhat independent of blood-stage immunity. 81 In addition, this finding provided further evidence that the likely key to inducing sterilizing antimalarial immunity lies in priming liver-stage immunity.
While this vaccination strategy has its drawbacks (the potential for a breakthrough of infectious parasites into the blood that necessitates the use of drug treatment), this method does theoretically permit the furthest liver-stage development, expanding the breadth of antigens produced and should therefore generate the broadest immune response repertoire.

| Subunit vaccines
There exist substantial challenges that obstruct the use of RAS, GAPs, and CPS for large-scale control of malaria in endemic regions.
Primarily, this is driven by technical problems that are associated with the large-scale rearing of GMP-grade mosquitoes, production, purification, and preservation of sporozoites, the necessity of intravenous administration, and in the case of RAS, the technical need for the use of radiation. Therefore, despite their superior protective capacity, these vaccination strategies may not be the most realistic option for mass production and control in malaria-ravaged areas. In In a 2014 study, DNA vaccination followed by human adenovirus 5 (AdHu5) encoding CSP was found to be safe, and induced antigenspecific CD8+ T cells which correlated with its 27% efficacy rate upon CHMI of malaria naïve individuals, representing another potentially a promising avenue of research. 94

| A TR AC TAB LE MOUS E MODEL FOR S TUD IE S OF LIVER-PROTEC TIVE MEMORY CD8+ T CELL S
Early studies identified the CSP antigen as an inducer of CD8+ T cells and a target of protective immunity in BALB/c mice. 95 Subsequently, multiple approaches including peptide immunization, DNA vaccination, and viral vectored prime-boost immunization with the CSP antigen or T cell epitopes were evaluated in preclinical mouse studies designed to identify potential strategies for subunit vaccination of humans against malaria. 23 In addition to protection, many of these early studies used the ELISPOT assay

| Host-parasite interactions impact CD8+ T cell-mediated liver-stage immunity
The mechanisms by which CD8+ T cells mediate clearance of malaria-infected hepatocytes are not entirely clear and are largely based on RAS immunizations that elicit CD8+ T cells, CD4+ T cells, and antibody responses. Using RAS immunization, it was demonstrated that indeed antigen-specific killing was necessary and that bystander inflammation and killing did not contribute to the clearance of infected hepatocytes. 105 It has also been shown that the depletion of IFNγ abrogates vaccine-induced protection from P. berghei and P. yoelii. 15 One mechanism by which this may occur is through the IFNγ-dependent inducible nitric oxide synthase (iNOS) pathway which is toxic to exoerythrocytic parasite forms. 106 Other CD8+ T cell effector molecules involved in direct cytolysis have historically appeared to be dispensable for vaccineinduced immunity. In the absence of perforin, granzyme B, or FasL expression, RAS-immunized mice (using P. berghei) were still immune from the sporozoite challenge. 107 However, using the DC + LM immunization approach to focus solely on CSP-specific memory CD8+ T cell-mediated protection, we showed that perforin-mediated cytolysis was required for sterilizing immunity to P. yoelii, but dispensable for sterilizing immunity against P. berghei. In contrast, deletion of the IFNγ gene or antibody-mediated blockade of TNFα compromised memory CD8+ T cell-mediated protection against both parasite species. 108 Although these data cannot be directly translated to protection against P. falciparum infection in humans, they do suggest that differences in parasite virulence, perhaps reflected in the diversity of field parasites, may influences which memory CD8+ T cell effector mechanisms will successfully control liver-stage malaria.
To further address this notion, we asked whether host or parasite genetic composition would affect the threshold of memory CD8+ T cells necessary for sterilizing immunity. BALB/c mice with memory CSP-specific CD8+ T cell frequencies representing >1% of peripheral blood lymphocytes were protected from P. berghei infection. In contrast, a hard threshold for sterilizing immunity mediated by CSP-specific memory CD8+ T cells against P. yoelii could not be defined. These data were consistent with P. yoelii being more virulent as demonstrated by the necessity of perforin in control of this parasite species.
Prior research in the field had shown that various inbred mouse strains demonstrated differences in susceptibility to P. berghei and differences in capacity for protection by RAS immunization. 109

| Discriminating between protective and nonprotective epitopes
It is generally accepted that most CD8+ T cells elicited by a viral infection can detect and kill virally infected cells. In contrast, we previously showed that mice could make CD8+ T cell responses against a viral epitope that failed to contribute to protection against a bacterial pathogen expressing the same epitope. 111 Unexpectedly, mice could make CD8+ T cell responses against the bacterially expressed epitope. These data showed that at least However, while CSP-and TRAP-specific CD8+ T cells were capable of reducing liver parasite burden and providing sterilizing immunity, GAP50-and S20-specific CD8+ T cells did not lead to reduced liver parasite burden or sterilizing immunity after sporozoite challenge. Failure to control liver-stage infection correlated with the inability of GAP50-and S20-specific CD8+ T cells to recognize P. berghei-infected hepatocytes and produce IFNγ in vitro.
These data showed that some epitopes that prime CD8+ T cell responses after RAS vaccination are not targets of protective liverstage immunity. As such, this information revealed that secondary screens for protective relevance are necessary when selecting antigens for subunit malaria vaccines, a concept with direct implications for ongoing vaccine development.
Ultimately, this group of studies thematically addressed several outstanding questions in the malaria field, finding that as is often the case, immunity to malaria is a complicated beast that involves multiple host and pathogen factors. The use of the DC-LM immunization model, for the first time, permitted us to provide quantitative data to address the threshold numbers of memory CD8+ T cells needed for sterilizing protection against liver-stage malaria. In comparison with vaccine trial data at the time, our data suggested that sterilizing immunity to malaria in humans would necessitate an increased potency of subunit vaccine approaches under clinical evaluation.
In addition, we were able to define the spectrum of effector functions critical for T cell-mediated protection against malaria. These data reveal that both host and parasite genetic factors can influence the characteristics of liver-stage protective memory CD8+ T cells, suggesting that outbred human populations may pose similar challenges in developing vaccine-induced protection. Additionally, this immunization model allowed us to discriminate between protective and nonprotective CD8+ T cell specificities, highlighting the need for additional downstream selection steps to identify candidate antigens for subunit vaccines.

| WHOLE PAR A S ITE MODEL S OF CD8+ T CELL IMMUNIT Y TO LIVER-S TAG E MAL ARIA
The DC-LM prime-boost approach provides a powerful model for reductionist studies of protective CD8+ T cell immunity in mice.
However, the direct translational potential of this approach to human vaccination is limited. In contrast, and as noted above, clear C57Bl/6 mice require at least two high-dose P. berghei immunizations to exhibit sterilizing immunity to a homologous challenge.
A major challenge to furthering our understanding of protective memory CD8+ T cells induced by RAS immunization was, and remains, the paucity of precisely identified Plasmodium epitopes.
To address outstanding questions in this area, our lab has attempted to combine specific epitope information and recombinant Plasmodium expressing defined heterologous epitopes in conjunction with specific T cell receptor transgenic (TCR-tg) cells in combination with an approach we developed to identify the total CD8+ T cell response to infection in any host. We have also used these approaches to compare CD8+ T cell responses across the whole parasite immunization platforms and to describe a potential dose sparing application of RAS vaccination in combination with subunit boosting. However, this particular study did not evaluate the location of T cell priming as a variable.

| CD8+ T cell priming
The next location of antigen expression occurs in the liver. A small subset of infectious sporozoites or RAS travel to the liver to infect hepatocytes, which will become the next source of Plasmodium antigen production. At this stage, the sporozoites undergo various changes in gene and protein expression as they develop further in hepatocytes, driving a diversification of potential antigens over time. 13,79 This was eloquently shown in our lab and the Kappe lab by comparing total CD8+ T cell responses against early and late arresting GAPs in a murine vaccination model. It was found that late arresting GAPs induced larger and broader CD8+ T cell responses to malaria following immunization, in addition to driving cross-species immunity. 79 The Butler et al. 79 study also suggested that in addition to CD11c+ mediated T cell priming in the skin dLN, priming of exclusively liver-stage antigens elsewhere would likely be an important component of vaccine-induced immunity. This hypothesis was directly tested in our lab by Kurup et al. 33 In this study, we demon-

| CD8+ T cell correlates of protection following RAS immunization
In a parallel study to those evaluating differences in CD8+ T cell memory effector mechanisms between parasite species following the DC-Prime LM boost, our lab was also investigating protective mechanisms involved in different whole parasite vaccines. Several that is, a large number of CD8+ T cells is still required for protection, which likely underlies the necessity of multiple RAS immunizations to induce sterilizing immunity.
Importantly, we were able to demonstrate in the RAS system that mouse background did matter. C57BL/6 mice were still more susceptible to sporozoite challenge even after RAS immunization compared with BALB/c mice, despite C57BL/6 mice having twofold more antigen-specific CD8+ T cells. This supported our findings of host nonhematopoietic factors being important for host susceptibility. 118 In this system, BALB/c mice required only 9% of the peripheral blood lymphocytes to be antigen-specific memory CD8+ T cells for protection whereas even greater than 40% in C57BL/6 mice were still not sufficient to protect from P. yoelii sporozoite challenge. This study also showed that giving repeated homologous RAS boosters doubled the number of CSP-specific memory CD8+ T cells which led to increased protection in BALB/c mice. 118 However, when C57BL/6 mice were given RAS boosters every 60 days this did not result in protection from P. yoelii.
Interestingly, these C57BL/6 mice had more T central memory

cells (T CM ) in the blood compared to T effector memory cells (T EM )
and were also IFNγ + and TNF-α+, whereas the CD8+ T cells from BALB/c mice were mostly IFNγ single positive. When considering correlates of protective immunity, we found that effector memory phenotype and not antisporozoite antibodies correlated with protective immunity after P. yoelii RAS immunization. These experiments together showed that numerical requirements for sterilizing immunity are very high regardless of whether a single subunit vaccination or a broader range of antigens in whole parasite vaccination is used.

| Broadening the antigenic repertoire with late arresting GAPs
As noted previously, GAP vaccines have several potential strengths over RAS vaccination due to their defined genetic attenuations. Several studies had shown that early arresting GAP targeting genes such as uis3, p52, and sap1, induced CD8+ T cellmediated protective immunity. [73][74][75][76]120 However, no comparative studies had evaluated the magnitude of the CD8+ T cell response to RAS relative to GAP vaccination or the potential that GAP vaccines arresting at later liver stages induce a better protective CD8+ T cell response. This was an important question, because at the time, early arresting GAP vaccine candidates were moving to clinical trials. [121][122][123] To address this, in collaboration with the lab of Stefan Kappe, we compared CD8+ T cell responses and protection elicited by a single RAS immunization or a single immunization with either an early arresting GAP (sap1-) or a late arresting GAP (fabb/f-). 79 The comparison to RAS immunization is important because in terms of intrahepatic development, RAS arrests at undefined, but relatively early liver stages. We hypothesized that late-stage arresting GAPs would result in a broader spectrum of antigens expressed, which would lead to superior protection.
We ultimately found that utilization of the late-stage arresting GAPs induced more robust CD8+ T cell responses against malaria in BALB/c, C57Bl/6, and Swiss Webster mice. This resulted in enhanced sterilizing protection in all mouse strains mice compared with the early arresting GAP or RAS immunization ( Figure 2). F I G U R E 4 CD8+ T cell priming following whole sporozoite immunization. During a blood meal, sporozoites from an infected mosquito are deposited into the dermis. A small subset of sporozoite is taken up by CD11c+ dendritic cells in the skin resulting in a nonproductive infection. The dendritic cells traffic to the skin dLN and present the processed sporozoite antigens to naïve CD8+ T cells. Another small subset of sporozoites enters the blood and traffics to the liver where they will invade host hepatocytes. CD11c + dendritic cells pick up antigenic debris from hepatocytes and traffic it to the liver dLN to prime naïve T cells.
Taken together, this study demonstrated that late-stage arresting GAPs represent a potentially superior vaccine candidate and even have the potential to elicit cross-stage protection. 120 However, the identification of analogous late-stage liver gene targets in human Plasmodium species has proven challenging for the further development of late-stage arresting GAPs for use in humans.

| LIVER-S PECIFI C T CELL IMMUNIT Y
The liver has a very unique anatomy in that it is highly vascularized with sinusoids that allow for the transfer of molecules between hepatocytes through the fenestrated epithelium of blood vessels. Historically, most malaria studies have focused on circulating memory T cells (T circM ) because they are accessible in human blood.
Additionally, in many models of infection, the abundance of T circM is directly related to protection from reinfection.
It is now established that CD8+ T cells circulate within liver sinusoids and crawl along the epithelia in a dynamic way 125 ( Figure 5). CD8+ T cell protection is extremely important for immunity to a variety of hepatotropic infections. 23 T cells in the liver and spleen of RAS-immunized mice. 134 This subset F I G U R E 5 T cells of the immune microenvironment in Plasmodium exposed liver. The liver immune microenvironment hosts various cell types in malaria-exposed mice. Stellate cells and Kupffer cells are the major APCs of the liver. During challenge or recurrent infection, sporozoites travel via the vasculature to the liver and first infect hepatocytes. While in the sinusoids, sporozoites and infected erythrocytes may be sensed by surveilling T CIRCM and T RM . Ly-6C + T CIRM is constantly moving throughout the vasculature as they travel the whole body and respond to malaria antigens. CXCR3 + CXCR6 + T RM patrol only the sinusoidal epithelia within the liver tissue and surveil for malaria antigen to mount a CD8+ T cell response.
of malaria-specific liver-residing memory CD8+ T cells was found to have major differences in their immune function and transcriptional programming compared with CD8+ memory T cells induced by other infection models. In particular, the same group showed that this subset of cells downregulated CD62L while upregulating CD69 and CXCR6. 134,135 Liver T RM in mice is similar to other T RM subsets but does not express the alpha E integrin CD103, a common marker of T RM in epithelioid tissues. 136 Liver T RM has been established as significant in autoimmune hepatitis and is elevated in patients with severe autoimmune hepatitis, demonstrating a potential pathogenic role in some instances. 137 Liver T RM has been demonstrated to be important for controlling human and murine infections. 138 146 This expeditious gene regulation and T EM infiltration appear to occur independent of liver T RM , because depletion of these cells in DC+ LM immunized mice via anti-CXCR3 antibody before sporozoite challenge did not prevent control of infection. As a whole, while liver T RM may be essential for sterilizing immunity after RAS vaccination, it is likely that the rapidly recruited T CIRCM population contributes to parasite control. The precise requirements for T RM versus T CIRCM in protection may be dictated by their relative abundance after distinct immunization approaches. Together, these data suggest that both circulating T EM and T RM likely play a role in limiting liver-stage malaria infection, though they are able to do this independently.

| CON CLUDING REMARK S
Malaria continues to be a devastating disease that affects more than half the world's population. The quest for a rational vaccine that is cost-effective, logistically feasible, scalable, and elicits long-term sterilizing immunity continues to be of paramount importance. The past few decades have brought to light multiple immunological features that can be harnessed to create the most effective vaccine possible. The existence of multiple Plasmodium species that infect rodents in combination with technological advances, permits investigators to recapitulate various aspects of human infections, and critically, make progress in understanding immune correlates of protection. The DC-LM model has proven to be a useful model for studies of the generation, quantity, and function of antigen-specific memory CD8+ T cells in a tractable and quantifiable way. The DC-LM model not only allowed for the investigation of the number of antigen-specific memory CD8+ T cells necessary to elicit protection against malaria but allowed for the study of the specific phenotypes of the memory populations in a rapid, replicable manner for any antigen of interest. The application of new technologies to study whole parasite models of immunization has also revealed the importance of T RM and T circM in protection that are relevant for improving human vaccination strategies.
Taken together, these models have yielded important insights into the types of memory T cells that are generated after immunization, their quantity and function, as well as the best immunization regimens and combinations.
The past several years have seen more focus on liver T cells specifically in the context of malaria research and the identification of tissue-resident memory T cells as an important population of memory lymphocytes crucial in a variety of infections, particularly for liver-stage malaria. The field moving forward will need to focus on the design of a vaccine that promotes long-term sterilizing immunity and intervenes at the liver stage of infection prior to blood-stage breakthrough. Further research into the generation, enhancement, and maintenance of T RM cells in the liver will be of paramount importance as these are known to be critical for protection. It is also crucial to understand the unique T RM in the liver of malaria-experienced individuals living in endemic areas as generating long-lasting immunity in this vulnerable yet large population remains a major challenge in the field. By coupling the vaccine design tools at our disposal with our knowledge of the qualities of long-lasting CD8+ T cells required for protection, we have a greater chance of achieving our goal of an effective cross-protective malaria vaccine for both adults and children living in endemic areas.

AUTH O R CO NTR I B UTI O N S
SH, MH, and JTH wrote and edited the manuscript. All authors approved of the submitted version.

ACK N OWLED G M ENTS
The authors would like to thank the previous and current members of the Harty lab, specifically Nathan Schmidt, Noah Butler, Lecia Pewe, Lisa Hancox, Katherine Doll, Scott Anthony, Samarchith Kurup, and Mitchell Lefebvre whose work has been described in this review.

FU N D I N G I N FO R M ATI O N
This work was supported by grants from the National Institutes of Health (National Institute of Allergy and Infectious Diseases) RO1AI085515, RO1AI095178, RO1AI100527, RO1AI167847 to JTH and T32 AI007260 and 1F32AI174382 to MH.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analyzed in this study.