Human ApoE2 protects mice against Plasmodium berghei ANKA experimental cerebral malaria

ABSTRACT Cerebral malaria (CM) is a severe neurological complication of Plasmodium falciparum infection with acute brain lesions. Genetic variations in both host and parasite have been associated with susceptibility to CM, but the underlying molecular mechanism remains unclear. Here, we demonstrate that variants of human apolipoprotein E (hApoE) impact the outcome of Plasmodium berghei ANKA (PbA)-induced experimental cerebral malaria (ECM). Mice carrying the hApoE2 isoform have fewer intracerebral hemorrhages and are more resistant to ECM than mice bearing the hApoE3, hApoE4, or endogenous murine ApoE (mApoE). hApoE2 mice infected with PbA showed increased splenomegaly and IFN-γ levels in serum but reduced cerebral cell apoptosis that correlated with the survival advantage against ECM. In addition, upregulated expression of genes associated with lipid metabolism and downregulated expression of genes linked to immune responses were observed in the brain tissue of hApoE2 mice relative to ECM-susceptible mice after PbA infection. Notably, serum cholesterol and the cholesterol content of brain-infiltrating CD8+ T cells are significantly higher in infected hApoE2 mice, which might contribute to a significant reduction in the sequestration of brain CD8+ T cells. Consistent with the finding that fewer brain lesions occurred in infected hApoE2 mice, fewer behavioral deficits were observed in the hApoE2 mice. Finally, a meta-analysis of publicly available data also showed an increased hApoE2 allele in the malaria-endemic African population, suggesting malaria selection. This study shows that hApoE2 protects mice from ECM through suppression of CD8+ T cell activation and migration to the brain and enhanced cholesterol metabolism. IMPORTANCE Cerebral malaria (CM) is the deadliest complication of malaria infection with an estimated 15%–25% mortality. Even with timely and effective treatment with antimalarial drugs such as quinine and artemisinin derivatives, survivors of CM may suffer long-term cognitive and neurological impairment. Here, we show that human apolipoprotein E variant 2 (hApoE2) protects mice from experimental CM (ECM) via suppression of CD8+ T cell activation and infiltration to the brain, enhanced cholesterol metabolism, and increased IFN-γ production, leading to reduced endothelial cell apoptosis, BBB disruption, and ECM symptoms. Our results suggest that hApoE can be an important factor for risk assessment and treatment of CM in humans.

Plasmodium berghei ANKA (PbA) infection of susceptible mice, such as C57BL/6 and CBA mice, results in experimental cerebral malaria (ECM) that closely reproduces many characteristics of human CM and is an important model for studying the disease (5).PbA infection stimulates the expression of chemokines and adhesion molecules in cerebral endothelial cells (6).CXCL9 and CXCL10 are the main factors that direct CD8 + T cell migration to the brain, and the expression of adhesion molecules such as ICAM-1 and VCAM-1 is responsible for the sequestration of activated leukocytes as well as adhesion of infected red blood cells (iRBCs) to the endothelial cells in the brain (7,8).Activated CD8 + T cells within the brain secrete cytotoxic perforin and granzyme B that induce endothelial apoptosis and disrupt the blood-brain barrier (BBB) leading to petechial hemorrhage in the brain during the acute phase of ECM (9,10).In addition, functional studies using specific neutralizing antibodies or specific gene-deficient mice have confirmed that loss of CD8 + T cells significantly reduces brain damage and host mortality (11,12).Infiltration of CD8 + T cells, cytoadherence mediated by receptors such as ICAM-1 and CD36, and BBB disruption are observed in ECM and CM (13).However, the trigger factors and pathological mechanisms involved in CD8 + T cell expansion and migration to the brain have not been fully addressed.
Apolipoprotein E (ApoE) is a glycoprotein involved in cholesterol transport and lipid homeostasis in the blood and central nervous system (CNS) in humans and mice (14)(15)(16)(17).The human ApoE (hApoE) consists of three isoforms, i.e., hApoE2, hApoE3, and hApoE4, while the mouse expresses only one form of ApoE (mApoE) that has ~70% homology with hApoE (17).hApoE is well-studied for its crucial roles in the onset and development of Alzheimer's disease (AD), traumatic brain injury, and cerebrovascular disease (18)(19)(20).ApoE variants are also associated with pathologies of viral infections including hepatitis C (HCV), human immunodeficiency (HIV), and herpes simplex (HSV) (21).hApoE3 and hApoE4, but not hApoE2, were reported to bind an atypical P. falciparum ETRAMP (22), and hApoE4 allele from children of a village in Gabon was significantly associated with higher median parasite densities (23).Interestingly, hApoE2 homozygote children from Ghana appeared to be infected with P. falciparum at an earlier age than those carrying other hApoE genotypes although the observation was based on four hApoE2 homozy gote samples (24).In another study, hApoE2 carriers were under-represented in cerebral and severe anemia patients, whereas hApoE3/4 carriers are more likely to suffer from severe malaria, suggesting a protective role of hApoE2 (25).One hypothesis to explain the "contradictory" observations for hApoE2 is that hApoE2 carriers may be exposed to the parasites earlier in life and develop immunity against severe malaria earlier (26).Additionally, genetic knockout of the single-copy mouse ApoE (mApoE −/− ) significantly protected the host from ECM with reduced parasite burden (27).These results suggest that hApoE variants play different roles in malaria pathogenesis, and their roles in malaria are still controversial.Here, we investigated the effects of hApoE alleles on PbA-induced ECM using transgenic mice (ApoE-TR) expressing hApoE variants.We showed that mice expressing hApoE2 (hApoE2 mice) were significantly more resistant to ECM than hApoE3 and hApoE4 mice.Furthermore, we found that increased lipid metabolism, downregula tion of inflammatory responses, and reduced CD8 + T infiltration in the brain contributed to better survival of the hApoE2 mice.Meta-analysis on the hApoE variant frequencies in different continental human populations showed higher hApoE2 frequencies in African and Oceanian populations, suggesting malaria selection on the hApoE2 allele.

RESULTS hApoE2 protects mice from PbA-induced cerebral malaria
To better understand the role of ApoE variants in the development of CM pathology, we utilized hApoE-TR mice as well as C57BL/6 wild-type (WT) mice that carry mApoE to evaluate the host mortality and brain lesions during the PbA infection.We first deter mined whether hApoE expression in the murine host was affected after PbA infection and showed that hApoE expression levels from the brain of PbA-infected mice were comparable to those of uninfected mice (Fig. S1A through D).Most of the male WT, hApoE3, and hApoE4 mice succumbed to ECM within 6-10 days post-infection (pi), whereas most (~80%) of the hApoE2 mice survived (Fig. 1A).The hApoE2 mice survived longer than 15 days but eventually died 3-4 weeks pi (Fig. 1A).The infected mice had similar parasitemia on days 2-3 pi, but significantly lower parasitemia in hApoE2 mice compared with WT and hApoE3 mice on day 4 and/or day 6 pi (Fig. 1B).Similar results were obtained from infections of the female mice (Fig. 1C and D).We also infected mice with PbA sporozoites harvested from mosquito salivary glands.Significantly longer survival, lower parasitemia, lower parasite load in the liver, and a 2-day delay in the development of blood-stage parasites were observed in hApoE2 mice compared to hApoE3 mice (Fig. 1E through H), suggesting an effect on liver-stage development.
BBB breakdown is a hallmark feature of ECM and contributes to cerebral hemorrhage, edema, coma, and death (28).We, therefore, assessed BBB integrity using the Evans blue (EB) extravasation assay, in which the EB can bind serum albumin and penetrate the impaired brain (29).On day 6 pi, the brains of PbA-infected WT, hApoE3, and hApoE4 mice were heavily stained with diffused EB suggesting intensive vascular leakage.In contrast, there was almost no EB coloration observed in infected hApoE2 mice (Fig. 2A).Quantification of brain extracts showed that EB levels were significantly lower in the hApoE2 mice than in other mice, indicating less BBB disruption (Fig. 2B).Furthermore, hematoxylin-eosin (H&E) staining of the brain sections revealed severe hemorrhages in ECM-susceptible mice, but not in the infected hApoE2 (Fig. 2C and D).In the ECM-sus ceptible mice, many hemorrhagic foci were observed in the olfactory bulb, which is consistent with a previous observation that the olfactory bulb was the most damaged brain region during the ECM development (30).Taken together, these data suggest that hApoE2 plays an important role in protecting mice from PbA-induced ECM with reduced host fatality at the early stage of infection.

Splenomegaly and high serum IFN-γ level in hApoE2 mice infected with PbA
The spleen plays a critical role in host immunity and the clearance of parasitized red blood cells (31,32).The hApoE2 mice had significantly larger spleens (splenomegaly) than those of other mice, indicating enhanced splenic immune responses in hApoE2 mice (Fig. S2A and B).However, histological examination of splenic tissue from the infected mice showed no significant difference in hemozoin pigment accumulation or tissue damage within the spleen between hApoE2 and ECM-susceptible mice (Fig. S2C  and D).
Cytokines produced by immune cells are closely associated with the development of ECM (6,33).We next measured serum levels of cytokines/chemokines (IL-1β, IL-5, IL-6, IL-10, IL-15, IFN-γ, MCP-1, CXCL9, CXCL10, and TNF-α) of mice on day 0 (uninfected), 2, and 5 pi (Fig. S2E through N).Serum cytokines were maintained at a relatively lower level in the uninfected mice and in day 2 infected mice.Interestingly, significantly higher levels of IL-5, IL-6, and TNF-α were observed in hApoE2 than in hEpoE4 mice day 0 pi (Fig. S2I through K).On day 5 pi, the level of IFN-γ was significantly higher in hApoE2 mice compared to those in the other mice (Fig. S2E), suggesting that the higher level of serum IFN-γ could reduce the risk of CM as suggested by previous clinical studies (34)(35)(36).No significant difference in the level of IL-10 between different types of mice (Fig. S2F), although this anti-inflammatory cytokine has also been shown to contribute to the protection against ECM (37).In summary, enhanced splenomegaly and higher serum IFN-γ levels in the infected hApoE2 mice could play a role in the protection against ECM.
TNF-α treatment was shown to prevent the development of malaria pre-erythrocytic stages in vitro (38), and administration of recombinant TNF-α protected against bloodstage infection of Plasmodium chabaudi (39).The elevated level of TNF-α in the hApoE2 mice day 0 pi also suggested a role of TNF-α in protection.We treated hApoE3 mice with TNF-α (20 pg in 0.1% bovine serum albumin [BSA]), infected the mice with PbA 1 hour later, and measured parasitemia and host survival.Compared to the control group that was treated with vehicle, TNF-α treatment significantly reduced parasitemia from day 4 pi (Fig. S3A), but no significant difference in host survival rate was observed (Fig. S3B), supporting a role of TNF-α in parasitemia control.

Differential expression of genes in lipid metabolism and immune response in the brain of hApoE2 mice infected with PbA
To investigate the potential molecular mechanism of hApoE2 in the protection against ECM, we performed RNA sequencing (RNA-seq) and transcriptome analysis of the brain tissues from noninfected and PbA-infected mice on day 6 pi.Comparison of samples from noninfected hApoE2 and hApoE3 mice detected clear differences in gene expression (Fig. S4A and B; Table S1), particularly unregulated genes in pathways of superoxide anion generation and superoxide metabolic process and downregulated genes in several immune effector processes and cytokine production (Fig. S4C).For the samples from infected mice, principal component analysis (PCA) and hierarchal clustering showed an obvious separation of the transcriptome between infected hApoE2 mice and the other three types of mice (WT, hApoE3, and hApoE4) (Fig. 3A), suggesting there are distinct transcriptional profiles between hApoE2 and the other mice.Differentially expressed genes (DEGs) were compared and evaluated in three different groups: hApoE2 versus WT, hApoE2 versus hApoE3, and hApoE2 versus hApoE4.The analyses yielded the following numbers of DEGs (P adj < 0.05 and fold change > 2; Table S1): 392 genes in the hApoE2 versus WT, 474 genes in the hApoE2 versus hApoE3, and 672 genes in the hApoE2 versus hApoE4.Among the regulated genes, 202 genes, with 79 upregulated genes and 123 downregulated genes, were shared by all three groups of comparison (Fig. 3B and C).Notably, genes such as Agt, Gal, and Scd1 that have a regulatory effect on lipid esterification, transportation, and localization were upregulated, while the transcriptions of Il12rb1, Tnfrsf8, and Ccl5 that relate to leukocyte activation, proliferation, and migration were downregulated in the hApoE2 mice (Fig. 3C; Table S1).We further performed pathway enrichment analysis of the commonly regulated genes and showed significant enrichment of upregulated genes in lipid metabolism, including steroid metabolic process, lipid localization, and lipid transport (Fig. 3D).Steroid hormone helps control metabolism, inflammation, immune functions, and the development of sexual characteristics (40,41).On the other hand, the downregulated genes were enriched mainly in the host immune response, such as regulation of leukocyte activation, regulation of immune effector process, and tumor necrosis factor-mediated signaling pathway (Fig. 3D).Overall, the transcriptome analysis indicated the downregulation of inflammatory immune responses and enhanced lipid/ steroid metabolism in the brain of infected hApoE2 mice, which likely contributes to reduced brain inflammation and protection against ECM.

Higher cholesterol levels in the serum and CD8 + T cells of hApoE2 mice
ApoE is a critical component in the regulation of lipid efflux, endocytosis, and circulation (15,42).The variation in lipid metabolism between the hApoE variants depends on their affinity to different receptors (43).Additionally, a recent study showed that cholesterol could induce the exhaustion of CD8 + T cells in the tumor microenvironment (44).However, whether host lipid metabolism contributes to the hApoE2-mediated resistance to ECM remains to be determined.
To investigate the lipid circulation in PbA-infected mice, we measured serum lipid levels on day 0 (uninfected), 2, and 5 pi.The levels of total serum cholesterol, LDL-choles terol, and triglycerides were significantly higher in the serum of hApoE2 mice than those of other mice from day 0 to day 5 (Fig. 4A through D).Filipin III has been used to detect cholesterol in cells (45,46).We next determined the cellular cholesterol content in CD8 + T cells by measuring the level of Filipin III and showed that higher levels of Filipin III in CD8 + T cells either from the brain or spleen of the hApoE2 mice than those from hApoE3 mice day 5 pi (Fig. 4E through H).Many lipids are essential for multiple biological processes, including energy metabolism and immunological regulation (47,48).How the higher lipid content affects CD8 + T cell activity and ECM requires further investigation.

hApoE2 attenuates T lymphocyte migration to the brain during the PbA infection
It has been reported that cerebral pathology mediated by brain infiltrating CD8 + T cells is responsible for ECM, and CD4 + T cells affect the generation of effector CD8 + T cells in the brain (11,49).No significant differences were observed in the T cell numbers from major immune organs such as the spleen, pelvic lymph nodes, and thymus between uninfected hApoE2 and hApoE3 mice, except mesenteric lymph nodes that had a significantly higher number of T cells in hApoE3 mice than in hApoE2 mice (Fig. S5A through D).To investigate the differences in host immune response after parasite infection, we isolated leukocytes from the brain and spleen of PbA-infected hApoE2 and hApoE3 mice and quantified the CD4 + and CD8 + T cells using flow cytometer analysis.Parasite infection activates host immunity, which frequently leads to acute splenomegaly and splenic lymphocyte expansion (50,51).Although the PbA infection significantly increased the number of CD4 + T cells in the spleen on day 5 pi, there was no significant difference in splenic CD8 + T cell proliferation between hApoE2 and hApoE3 mice (Fig. 5A through C).Importantly, both the frequency and cell number of CD4 + and CD8 + T cells sequestered in the brain were significantly lower in the hApoE2 mice compared to the hApoE3 mice on day 5 pi (Fig. 5D through F).Additionally, the expression levels of CXCL9, CXCL10, and CXCR3 in the brain were significantly lower in infected hApoE2 mice than those in infected hApoE3 mice (Fig. 5G and H).These results suggest that hApoE2 mitigated the migration of CD4 + and CD8 + T cells to the brain during PbA infection, which helped to reduce BBB disruption and intracerebral hemorrhage.

hApoE2 reduces cell apoptosis in the mouse brain with PbA infection
Active CD8 + T cells were shown to promote cerebral pathology through the secretion of cytotoxic perforin and granzyme B to disrupt intercellular junctions of the endothelium, and endothelial cell apoptosis generally occurs during ECM development (9,10,52).Interestingly, the granzyme B level was significantly reduced in the brain of PbA-infected hApoE2 mice compared to those of other mice (Fig. 6A and B).The caspase cascade plays a crucial role in apoptosis, and cleaved caspase 3 (CC3) is a marker for apoptotic cells (53).A significant reduction in the level of CC3, but not the level of procaspase 3, was observed in the brains of infected hApoE2 mice compared to those from other infected mice (Fig. 6A, C, and D).Furthermore, TdT-mediated dUTP nick-end labeling (TUNEL) staining of brain sections showed that the number of apoptotic cells was significantly lower in infected hApoE2 mice than in the other infected mice (Fig. 6E and F).Moreover, immunofluorescent staining of the brain sections showed that CC3 was colocalized with CD31 (a marker of endothelial cells) in many cells from ECM-susceptible mice, but no apoptotic event was detected in vascular endothelial cells from hApoE2 mice during PbA infection (Fig. 6G).These results indicate lower levels of granzyme B and apoptosis in the brains of hApoE2 mice after the PbA infection, which in turn reduced BBB disruption and, therefore, CM pathology.

Less behavioral deficits in hApoE2 mice after the PbA infection
Along with serious BBB disruption, neurological and cognitive deficits, and behavio ral sequelae are frequently seen among CM survivors after rescue therapy (54,55).
Considering the alleviated symptoms present in infected hApoE2 mice, we conducted behavioral tests to determine whether hApoE2 mice have fewer behavioral impairments after infection.Infected hApoE2 and hApoE3 mice received pyrimethamine treatment from day 6 to 12 and were subjected to behavioral tests on day 15 after PbA infection (Fig. S6A).Indeed, the infected hApoE2 mice spent significantly more time than the infected hApoE3 mice in the light compartment of the light/dark box test, indicating less anxiety level for hApoE2 mice post-rescue treatment (Fig. S6B).Meanwhile, the capacity for spatial learning and memory was evaluated by the Morris water maze after infection.The infected hApoE3 mice were less efficient at finding the hidden platform than the uninfected hApoE3 mice, indicating a deficit in the spatial learning ability of hApoE3 mice (Fig. S6C).By contrast, PbA infection hardly affected the spatial learning of hApoE2 mice, as evidenced by the similar latency to reach the platform between uninfected and infected hApoE2 mice (Fig. S6C).Furthermore, a significant decrease in the number of target zone crossings was observed in the infected hApoE3 mice compared to the infected hApoE2 mice after the platform was removed (Fig. S6D).The infected hApoE2 mice also spent more time in the target quadrant than the infected hApoE3 mice (Fig. S6E).However, the total explored distance showed no significant difference between the hApoE2 and hApoE3 mice, thus excluding impairments in motor function (Fig. S6F).
These findings together suggest less severe behavioral impairments of the hApoE2 mice than the hApoE3 mice after the infection, which is in agreement with the greater protection of this mouse against ECM induced by PbA infection.

hApoE2 and hApoE4 variants are more prevalent in African populations
Our results above show that hApoE2 is protective against malaria including ECM, which also predicts potential selection for higher hApoE2 allele frequency in malaria-endemic regions.We next systematically reviewed published data from population-based cohorts to estimate the proportion of ApoE variants in worldwide populations.Initial literature searches of public databases yielded 3,253 studies containing genotypes of hApoE variants.After excluding duplicated data sets and removing irrelevant and ineligible publications (Fig. S7), a total of 416 studies with 1,611,214 samples were finally included in our meta-analysis (Data set 1).As predicted, the frequency of hApoE2 was significantly increased (11.75% for Africa versus 7.14% for the global average) but the frequency of hApoE3 was significantly reduced (69.8% for Africa versus 80.0% for the global average) in malaria-endemic Africa (Fig. S8), which was consistent with a protective role of hApoE2 and a detrimental role of hApoE3 in our transgenic mouse models.Surprisingly, the frequencies of hApoE4 were also significantly higher in African and Oceanian popula tions (Fig. S8), which was not supported by our experimental models.One possible explanation was that hApoE4 could be also selected by other unknown pathogens.Nonetheless, our experimental models demonstrate that hApoE2 is protective against malaria through increased levels of IFN-γ in the serum and levels of lipids within T cells, reduced T cell migration to the brain and inflammation, and better protection of BBB and brain cells.These results support and clarify the protective role of hApoE2 in malaria.

DISCUSSION
hApoE variants have been implicated in disease severity in patients infected with P. falciparum parasites (22,23,25,26); however, the roles of individual hApoE variants in malaria and the molecular mechanism of hApoE-mediated protection against CM remain largely unknown because of technical difficulties and ethical restrictions associated with studying human subjects.Taking advantage of transgenic mice expressing different hApoE variants, we evaluated the outcomes of mice carrying hApoE variants after infection with PbA and investigated the molecular mechanisms contributing to ECM.We show that hApoE2 is protective against ECM, but not hApoE3 and hApoE4, with hApoE2 mice having attenuated pathological damages and surviving longer than other mice.Compared to ECM-susceptible mice (WT, hApoE3, and hApoE4), the majority of hApoE2 mice survived the fatality window caused by ECM (days 6-10 pi) and exhibited fewer hemorrhages and cerebral injuries after the PbA infection.Particularly, hemorrhage foci that often occurred in the olfactory bulb of infected ECM-susceptible mice were absent or infrequent in those of infected hApoE2 mice.Cerebral dysfunction is associated with vascular leakage, and brain capillaries are frequently disrupted during ECM because of increased parasite adhesion and T-cell sequestration in the brain (30,56,57).The dramatically reduced BBB disruption suggests that PbA-induced cerebral pathology is significantly alleviated in hApoE2 mice.The survival advantage of hApoE2 mice against ECM is consistent with human population studies showing that individuals carrying the hApoE3/E4 genotype are more likely to develop CM than those with the other genotypes (22,23).Additionally, our meta-analysis also shows potential positive and negative selection on hApoE2 and hApoE3, respectively, supporting a protective role of hApoE2.However, we cannot rule out that other pathogens such as HSV, HCV, and HIV also contributed to the selection of specific hApoE alleles.Differential immune responses may contribute to the different cerebral pathologies observed in the hApoE mice.Host immune response is a double-edged sword against severe infectious diseases, including CM (58,59).Previous research has suggested that an over-reactive immune system mediated by CD8 + T cells plays a crucial role in the ECM process, and a properly controlled host immunity was linked to the reduction of fatal outcomes (60)(61)(62).The splenomegaly and higher level of IFN-γ in the infected hApoE2 mice suggest activation of protective immunity, and the reduced T cell infiltration and inflammation in the brain may help protect the mice from ECM.Studies on fatal malaria indicated that splenomegaly was associated with reducing mortality and triggering a protective host immune response (27,51,63).Higher serum IFN-γ level in infected hApoE2 mice was correlated with the survival advantage of PbA infection, which is consistent with clinical studies showing a higher IFN-γ level in individuals with uncom plicated malaria than in CM patients (34-36).Injection of IFN-α into PbA-infected mice resulted in a significantly increased level of IFN-γ, reduced parasitemia, and improved survival from ECM than mice receiving diluent (64).On the other hand, IFN-γ may contribute to splenic T cell priming, activate brain endothelial cells, and promote cerebral symptoms after PbA infection (65,66).It has been reported that IFN-γ controls the recruitment of pathogenic CD8 + T cells to the brain and Ifn-γR1-deficient mice are highly resistant to ECM (67).Thus, the effects of IFN-γ in cerebral pathology and host mortal ity during ECM with PbA infection are likely dose-and time-dependent.Nonetheless, the increased IFN-γ and other cytokines in the PbA-infected hApoE2 mice suggest a protective role of IFN-γ.
Our data also show that reduced hemorrhage and BBB breakdown in the brain of PbA-infected hApoE2 mice are associated with lower granzyme B levels and cell apoptosis.BBB is essential for the separation of the central nervous system from the circulatory system, and the degree of disruption to the BBB is related to the severity of brain dysfunction in neurological diseases (28,68,69).In the hApoE2 mice, the gran zyme B level in the brain was significantly lower than those of other mice.Additionally, CD31-marked endothelial cells were colocalized with CC3 in ECM-susceptible mice, but not in the hApoE2 mice.These results suggest that granzyme B from activated CD8 + T cells may play a role in BBB impairment and promote endothelial cell apoptosis, which is consistent with the observations from previous studies (10,70).hApoE plays a major role in lipid metabolism, especially cholesterol transporta tion (71,72), and increased lipid levels may impact parasite development and host immune responses.Circumsporozoite protein on malaria sporozoites can interact with low-density lipoprotein receptor-related protein for the invasion of hepatocytes (73), ApoE-enriched beta-very low-density lipoprotein and lactoferrin could inhibit sporozoite invasion of HepG2 cells (74).However, our observation of protection after injection of iRBCs should bypass the liver stages in this study.It has been demonstrated that cholesterol can induce CD8 + T cell dysfunction and exhaustion in the tumor microenvir onment (44,(75)(76)(77).The absence of ApoE in the murine host leads to increased antigen presentation and enhanced CD4 + T cell activation due to the enrichment of cholesterol in dendritic cells (78).Additionally, ApoE promotes T cell-mediated anti-tumor activity by inhibiting immunosuppressive cells (79).Furthermore, variants of hApoE have been shown to account for the different outcomes after infection with different pathogens, including the COVID-19 virus and hepatitis C virus (80,81).Our data show that the cholesterol levels in the serum and within CD8 + T cells were considerably higher in hApoE2 mice than those in ECM-susceptible mice, which were correlated with the alleviated cerebral immune response.Moreover, CD4 + and CD8 + T cell migration to the brain was significantly reduced in hApoE2 mice after PbA infection.Therefore, the increased lipid metabolism in the hApoE2 mice may suppress CD8 + T cell activation and migration as well as inflammatory responses in the brain, reducing BBB breakdown and hemorrhage.
Higher serum levels of CXCL9 were observed in the hApoE2/3/4 mice without parasite infection.The result suggests that the introduction of human ApoE genes might alter some host immune responses even though the levels of mApoE and hApoE variants were expressed at similar levels in uninfected mice.Additionally, the spleen size and the levels of most serum cytokines and specific lipids were similar in the uninfected mApoE and hApoE3/4 mice, suggesting the impacts of the replacement of mApoE by hApoE3/4 were minimal.However, significantly higher levels of some lipids in the serum and T cells were observed in the hApoE2 mice than in mApoE, hApoE3, and hApoE4 mice before and after parasite infection.The results suggest that the introduction of hApoE2, but not hApoE3 and hApoE4, changes the metabolic and immunologic potentials, leading to better protection.Although the protective role of hApoE2 observed in the transgenic mice may not reflect the true situation in human malaria, the high frequency of hApoE2 in malaria-endemic Africa suggests a selection of this ApoE variant by malaria parasites or other unknown agents.Whether hApoE2 can protect human subjects from malaria requires further clinical association studies.
In summary, our study suggests that hApoE2 protects mice from ECM through several potential mechanisms including altered lipid metabolism, suppressed CD8 + T cell activation and migration, and enhanced IFN-γ production in the circulation.Increased lipids uptake can cause CD8 + T cell dysfunction or exhaustion and prevent inflammation in the brain (82), which in turn reduces endothelial cell apoptosis, BBB disruption, and ECM symptoms (Fig. 7).The higher levels of IFN-γ and TNF-α in the serum of hApoE2 mice also suggest a strong protective immune response outside the CNS, which helps control parasite growth.Early IFN-γ production is protective against ECM (83), and TNF-α could protect against blood-stage infection of Plasmodium chabaudi (39).However, chronic high-level IFN-γ may induce the expression of adhesion molecules such as ICAM-1 as well as CXCL9 and CXCL10 that recruit CD8 + T cells to the brain during ECM (49,84).Given the association between hApoE variants and the risk of ECM, hApoE can be an important factor for risk assessment and treatment of CM in humans.For example, our observation of the higher level of lipid metabolism in the infected hApoE2 may provide a therapeutic possibility.Hyperlipidation is also potentially a central mechanism underlying hApoE2's protective effect against AD (85).Oral administration of liver X receptor agonists such as GW3965 and TO901317 to increase lipidation in the brain of mice has been shown to reduce brain Aβ load and improve cognitive performances of amyloid transgenic animals (86,87).Similarly, oral administration of retinoid X receptor agonist bexarotene upregulated ApoE and ABCA1 in mouse brains and increased ApoE levels in the cerebrospinal fluid (88).These lipidation enhancers can be tested in ECM and clinical trials for reducing or preventing symptoms of CM.

Mice, malaria parasites, and infection
Homozygous ApoE transgenic mice with the endogenous mApoE gene replaced with hApoE2, hApoE3, or hApoE4 gene were acquired from Taconic; C57BL/6 WT mice were bred in the Xiamen University Laboratory Animal Center (XMULAC).All mice used in this study were aged 6-8 weeks.The parasite strain P. berghei ANKA (PbA) was initially obtained from MR4 (http://www.mr4.org/) and preserved in our laboratory.Cryopre served parasites were thawed and injected intraperitoneally (i.p.) into donor mice before being used to infect recipient mice with an inoculum containing 5 × 10 5 iRBCs in 200 µL phosphate-buffered saline (PBS) via tail vein injection.Parasitemia was measured daily by microscopic examination of Giemsa-stained thin blood smears from day 2 pi.Infected mice were monitored for the development of neurological signs between days 5 and 10 pi.

Mosquito feeding and sporozoite infection
A colony of Anopheles stephensi mosquitoes (Hor strain) was used to generate PbA sporozoites for the inoculation of mice.All mosquitoes were kept at 23°C and 75% humidity under a 12:12 light-dark illumination cycle.Mosquitoes (~50 females) were fed on an anesthetized mouse having 0.5%-1% gametocytes for 30 min.Mosquito salivary glands were dissected on day 15 post-feeding under a light microscope and harvested for homogenization.Released sporozoites were counted using a cell counting chamber and each mouse was injected intravenously (iv) with 5,000 sporozoites.

BBB permeability assay
BBB permeability was assessed through the EB extravasation assay as described (52).Briefly, mice were injected intravenously with 200 µL of 1% Evans blue (EB) in sterile PBS on day 6 pi and allowed to circulate for 1 hour.Mice were sacrificed and their brains were collected after intracardial perfusion with 20 mL ice-cold PBS.The brains were weighed and then incubated in formamide at 4°C for 48 hours to extract EB.The concentration of extracts was measured at 610 nm using a Tecan Spark microplate reader.

Histology
Infected mice were anesthetized on day 6 pi and perfused with 20 mL ice-cold PBS and 10 mL 4% paraformaldehyde (PFA).Brains and spleens were removed, weighed, and incubated overnight at 4°C in 4% PFA.Following dehydration and paraffin-embedment processes, tissue sections (5 µm) were made and stained with a H&E solution.The brain sections were scanned under a Motic VM1 microscope, and the spleen sections were imaged with a Leica DM4B system.Images were processed and analyzed using Image J software.

Measurement of serum cytokines
Serum samples were collected from uninfected and infected mice on days 0, 2, and 5 pi.Serum cytokines were detected and quantified using the MILLIPLEX MAP Kit (Millipore, MCYTOMAG-70K) following the manufacturer's instructions.Briefly, the antibody-bead conjugates were added to a 96-well plate and mixed with serum, followed by incubation at 4℃ overnight on a plate shaker.The beads were washed 3× with washing buffer containing 0.05% Proclin from Millipore and incubated with secondary antibodies for 1 hour at room temperature.Streptavidin-phycoerythrin was added to each sample and incubated at room temperature for 30 min.Signals for IFN-γ, IL-1β, IL-5, IL-6, IL-10, IL-15, MCP-1, CXCL9, CXCL10, and TNF-α were measured using Luminex 200.

RNA preparation and transcriptome analysis
Infected mice were sacrificed on day 6 pi, and the brain tissues were collected and immediately frozen in liquid nitrogen.Total RNA was isolated using TRIzol reagent (Invitrogen, 15596018) according to the manufacturer's instructions.RNA purity was determined using the NanoPhotometer spectrophotometer (IMPLEN, CA, USA), and RNA integrity was further assessed using the RNA Nano 6000 Assay Kit of the Bioana lyzer 2100 system (Agilent Technologies, CA, USA).One microgram RNA per sample was used for sequencing library construction using the NEBNext UltraTM RNA Library Prep Kit (Illumina, San Diego, CA, USA) following the manufacturer's recommendations.Differential gene expression analysis was performed using the DESeq2 R package (1.16.1).The resulting P-value was adjusted using Benjamini and Hochberg's approach, and genes with an adjusted P-value < 0.05 were assigned as differentially expressed.Gene ontology (GO) enrichment analysis of DEGs was implemented in the cluster profiler R package after correction for gene length bias.GO terms with a corrected P-value < 0.05 were considered significantly enriched for DEGs.

Flow cytometry
Mouse brains were isolated from PBS-perfused mice, homogenized, and then digested in 5 mL RPMI medium containing 0.1 mg/mL collagenase D (Roche, 11088866001) and 1 mg/mL DNase I (Roche, 11284932001) for 30 min at 37°C on a plate shaker.The digested samples were further passed through a 70 µm pore cell strainer.Brain-seques tered mononuclear cells were harvested by centrifugation on a 30%/70% discontinuous Percoll gradient in PBS.Single-cell suspensions from mouse spleens were prepared by mechanical disruption through a 70 µm strainer followed by ammonium-chloride-potas sium (ACK) buffer to lyse the RBCs.To identify T cell subsets, cells were stained with eFluor450-CD4 (eBioscience, 48-0042-80) or FITC-CD8 (Biolegend, 100705) for 30 min at room temperature for downstream flow cytometry analysis.Cells were detected using Fortessa X-20 and analyzed by FlowJo.

TUNEL staining
The brains were dissected from infected mice and incubated in 4% PFA at 4°C overnight.Tissue slices (20 µm) were made and stained using the TUNEL Apoptosis Detection Kit (Yeasen, 40306ES20).Briefly, the samples were incubated with Proteinase K solution for 10 min after the fixation with 4% PFA.Then, the samples were stained with TdT solution containing FITC-12-dUTP Labeling Mix and Hoechst for 30 min.Images were captured on a Zeiss LSM 780 confocal microscope.

Immunofluorescence assay
Infected mouse brains were incubated overnight in 4% PFA and then embedded in an OCT matrix.Brain sections at a thickness of 20 µm were cut on a freezing micro tome (Leica, CM1950) and stained with anti-mouse CD31 (Proteintech, 65058-1-lg) and anti-cleaved caspase 3 antibodies (CST, 9661) overnight at 4°C.Samples were stained with Hoechst and detected using Zeiss LSM 780.

Mouse behavioral test
hApoE2 and hApoE3 mice were infected with PbA and treated with pyrimethamine during days 6-12 pi.On day 15 pi, anxiety-like behavior was measured through the light-dark test.The device consists of a half-light aversive compartment and a half-safe compartment, with a small door in the middle.Mice were placed at the door and the time they spent in the light compartment was registered within 5 min.The Morris water maze test was used to assess spatial learning and memory.A hidden platform was placed about 1 cm below the water surface in a black circular pool, and each mouse received constant trials for 6 days to search the platform during the training period.Then, the platform was removed during the test phase.The number of target zone crossings and time spent in the target quadrant were recorded.The data were analyzed using Noldus EthoVision XT.

Cholesterol content measurement
Serum samples were collected from uninfected and infected mice, and the cholesterol contents of the serum were measured using various detection kits (Mindray) including a Triglycerides Kit, Total Cholesterol Kit, HDL-Cholesterol Kit, and LDL-cholesterol Kit following the manufacturer's instructions.Briefly, monocytes were isolated from the spleen and brain.Cells were stained with FITC-CD8 (Biolegend, 100705) and 50 µg/mL Filipin III (Cayman, 70440) for 1 hour at room temperature.Samples were measured and analyzed using Fortessa X-20 and FlowJo, respectively.

Meta-analysis
Literature was comprehensively searched in PubMed and Web of Science with the search term "ApoE" up to 24 February 2023.The proportion of the hApoE allele was transformed with Freeman-Tukey double arcsine and the 95% CI was calculated using a random-effects model.Heterogeneity between the studies was assessed by Cochran's Q test and quantified through the I 2 statistic for P-value analyses.The subgroup analyses were conducted according to multiple variables (continent, gender, and mean age).R version 4.2.2 and the meta-analysis package were applied for statistical analyses and to generate forest plots.A P-value under 0.05 was considered statistically significant.

Statistical analysis
Statistical analyses were conducted using GraphPad Prism.Two-tailed parametric t-test or Mann-Whitney test were used in comparisons between two groups, and multiple comparisons were assessed with Tukey's multiple comparisons tests.A comparison of survivals was performed using the log-rank (Mantel-Cox) test.Data are presented as mean ± SEM; ns, not significant (P > 0.05); *P < 0.05; **P < 0.01; and ***P < 0.001.curation, Formal analysis, Funding acquisition, Project administration, Writing -original draft, Writing -review and editing

DIRECT CONTRIBUTION
This article is a direct contribution from Xin-zhuan Su, a Fellow of the American Academy of Microbiology, who arranged for and secured reviews by Sanjai Kumar, U.S. Food and Drug Administration, and Richard Colleton, Ehime University.

FIG 2
FIG 2 Evans blue leakage and intracerebral hemorrhage of the cerebral tissue in hApoE mice infected with PbA.(A) Representative brain images from PbA-infected mice injected with Evans blue on day 6 pi.Bar = 5 mm.(B) Quantification of Evans blue content from the brain of PbA-infected mice.n = 4 mice per group.(C) Representative H&E images showing hemorrhage in the olfactory bulb (Olb), cortex (Ctx), and cerebellum (CE) in PbA-infected mice on day 6 pi.The arrowheads point to hemorrhage sites.Bar = 50 µm.(D) Plots of hemorrhage sites in the Olb, Ctx, and CE from PbA-infected mice.ND, not detected; n = 5-6 mice per group.(B) and (D) Tukey's multiple comparisons test.*P < 0.05; **P < 0.01; and ***P < 0.001.

FIG 3
FIG 3 Transcriptome profiles of mouse brain infected with PbA.(A) PCA of RNA-seq data from the brains of PbA-infected WT and hApoE mice.(B) Venn diagram comparing differentially expressed genes between infected hApoE2 mice and ECM-susceptible mice.(C) Heatmap of 202 differentially regulated genes from infected hApoE2 mice and ECM-susceptible mice.Selected genes are indicated on the right side of the figure.(D) Significantly enriched pathways of upregulated and downregulated genes in panel C. PCA and enrichment analyses are as described in Materials and Methods.

FIG 6
FIG 6 Reduced apoptosis in the brain of hApoE2 mice infected with PbA.(A) Western blots of granzyme B, procaspase 3, and CC3 in the brain from uninfected and PbA-infected mice on day 5 pi.(B−D) Densitometric quantification of granzyme B, procaspase 3, and CC3 expression in panel A, respectively; band densities were normalized to β-actin.n = 3 experimental repeats.(E) Representative TUNEL-stained images of the brain showing apoptosis in PbA-infected mice.The asterisks point to TUNEL-positive cells.Bar = 100 µm.(F) Quantification of TUNEL-positive cells in panel E. n = 4 mice per group and three images per mouse were calculated.(G) Co-staining of CD31 (red) and CC3 (green) showing apoptotic endothelial cells of brain sections from PbA-infected mice.Bar = 20 µm.(B, C, D, and F) Tukey's multiple comparisons test.*P < 0.05; **P < 0.01; ***P < 0.001; and ns, not significant.