IFN-γ–Producing CD4⁺ T Cells Promote Experimental Cerebral Malaria by Modulating CD8⁺ T Cell Accumulation within the Brain

Plasmodium berghei ANKA infection in susceptible strains of mice results in the development of experimental cerebral malaria (ECM), a fatal neuropathology characterized by the sequestration of parasite-infected RBCs (pRBC) and leukocytes within the brain (reviewed in Refs. 1–3). The clinical signs of ECM, including ataxia, paralysis, coma, and ultimately, death, are analogous to those of human cerebral malaria; as in human disease (reviewed in Ref. 1), although treatment with antimalarial drugs can prevent mortality associated with ECM, surviving mice may display long-lasting neurologic deficits, including impaired memory (4, 5).

IFN-γ–dependent processes are involved in the development of cerebral pathology during P. berghei ANKA infection, as evidenced by the complete resistance to ECM of IFN-γ^−/− and IFN-γR^−/− mice on normally susceptible backgrounds (6, 7). IFN-signaling pathways are significantly upregulated within the brains of ECM-susceptible mice during infection with P. berghei ANKA, suggesting that IFN-γ may directly influence the local cerebral environment (8, 9). In support of this hypothesis, IFN-γ was shown to promote macrophage accumulation and macrophage effector functions in the brain during infection (6), as well as to direct migration of CD8⁺ T cells into the brain through CXCL9- and CXCL10-dependent pathways (7, 10-13). Although the functional role of brain-accumulating macrophages/monocytes in promoting ECM is unclear (14, 15), it is well established that T cells contribute to the initiation and/or terminal development of cerebral pathology (1, 3, 14). IFN-γ also promotes the upregulation of adhesion molecules on brain endothelial cells during infection, potentially enhancing pRBC and leukocyte sequestration within the brain vasculature and transmigration of cells into the perivascular space. (6, 10). Indeed, it was recently shown that pRBC and leukocyte accumulation is reduced in the brains of IFN-γ^−/− mice during infection (15, 16).

IFN-γ can be produced by various cell populations during malaria infection, including NK cells, NKT cells, γδ TCR⁺ T cells, and αβ TCR⁺ CD4⁺ and CD8⁺ T cells (reviewed in Refs. 17, 18). Strikingly, and consistent with the notion of temporal activation of innate and adaptive immune responses, sequential production of IFN-γ by NK cells and CD4⁺ T cells was shown to occur in vitro following exposure of PBMCs to Plasmodium falciparum parasites (19). In some malaria infections, IFN-γ production by CD4⁺ T cells may be transient as a result of changes in the immunological environment. For example, during P. chabaudi AS and P. yoelii infections, changes in the dendritic cell compartment in the later stages of infection result in reduced IFN-γ production by CD4⁺ T cells concomitant with upregulation of IL-4 and IL-10 production (20, 21). With specific relevance to P. berghei infection, an early burst of IFN-γ during the early stages of infection was associated with protection against ECM (22). Although there are many potential cellular sources of IFN-γ during P. berghei ANKA infection, it is unclear whether IFN-γ production by an individual cell population in a specific tissue location governs the development of ECM or whether sequential or concomitant induction of IFN-γ by different cell types and in different locations is required.

Although ECM is referred to as a classical IFN-γ–mediated condition, it was shown that IFN-γ–producing CD4⁺ T cells are essential for the migration and recruitment of Th17 cells into the CNS during experimental autoimmune encephalomyelitis (EAE) (23) and that Th17 cells, rather than IFN-γ–producing cells, are responsible for much of the pathology of EAE (24, 25). Thus, it is possible that IL-17 mediates the development of ECM during P. berghei ANKA infection downstream of IFN-γ production. IL-17 can exert a number of direct effects in the CNS that may contribute to neuropathology during malaria infection, including direct neuronal damage (26), disruption of the blood–brain barrier (27), and activation of astrocytes (28, 29). The role of IL-17 during P. berghei ANKA infection has not been examined in detail.

In this study, we ruled out any role for IL-17 in the development of ECM and used IFN-γ reporter mice (30) to examine the dynamics of IFN-γ production within distinct anatomical locations during P. berghei ANKA infection. We subsequently calculated the contribution of IFN-γ production by each distinct cell type to the total IFN-γ response. Despite the myriad of cells that can produce IFN-γ during infection, using an adoptive-transfer model, we show that IFN-γ production specifically by CD4⁺ T cells, and not by innate cells or CD8⁺ T cells, can promote signs of ECM in normally resistant infected IFN-γ^−/− mice. Finally, we provide mechanistic evidence that IFN-γ production exclusively by CD4⁺ T cells causes ECM by increasing CD8⁺ T cell migration and accumulation within the brain, potentially via the local upregulation of CXCL9 and CXCL10 within the brain. Our results significantly extend our understanding of the pathogenesis of ECM during P. berghei ANKA infection.

C57BL/6N (wild-type [WT]), C57BL/6 IFN-γ knockout (IFN-γ^−/−), C57BL/6 RAG-1 knockout (RAG-1^−/−), and C57BL/6 Ly5.1 (Ly5.1⁺) mice were bred in-house at London School of Hygiene and Tropical Medicine or purchased from Harlan (Oxford, U.K.). C57BL/6 IFN-γ reporter mice (YETI: 30) were kindly provided by Dr. M. Mohrs (Trudeau Institute, Saranac Lake, NY) and were bred at Harlan. Animals were maintained in individual ventilated cages. C57BL/6 IL-17 receptor knockout (IL-17R^−/−) mice, provided by Amgen USA, which are deficient in IL-17A and IL-17F responses, were bred and maintained in the Department of Pathobiology, University of Pennsylvania. C57BL/6J mice used as controls in IL-17R^−/− experiments were purchased from The Jackson Laboratory (Bar Harbor, ME) and were maintained at the University of Pennsylvania. Male and female mice were used in separate experiments between 6 and 10 wk of age.

Cryopreserved P. berghei ANKA parasites, derived from the PbA clone 15cy1, were thawed and passaged once in vivo before being used to infect experimental animals. Experimental mice were infected i.v. with 10⁴ parasitized RBCs. In some experiments, mice were injected i.p. with 250 μg anti-CD8 (53-7.62) or anti–IFN-γ (XMG1.2) on days −2, 0, 2, 4, and 6 of infection. Isotype control Abs (rat IgG2a and rat IgG1, respectively) were used to verify the specific activity of anti–IFN-γ and anti-CD8 mAbs administered in vivo. All Abs were obtained from BioXCell (West Lebanon, NH). Parasitemia was monitored daily by microscopic examination of Giemsa (BDH)-stained thin blood smears. Signs of disease were classified using the following clinical scale: 1 = no signs; 2 = ruffled fur/and or abnormal posture; 3 = lethargy; 4 = reduced responsiveness to stimulation and/or ataxia and/or respiratory distress/hyperventilation; and 5 = prostration and/or paralysis and/or convulsions. All animals were immediately euthanized when observed at stage 4 or 5. Stages 2 and 3 were classified as prodromal signs of ECM, and stages 4 and 5 were classified as ECM. Animals in stages 4 and 5 of infection invariably showed signs of cerebral pathology, including blocked vessels, hemorrhages, edema, perivascular cuffing and disruption, and damage to cerebral vasculature endothelial linings following histological examination.

Splenic single-cell suspensions were prepared by homogenization through a 70-μm cell strainer (BD Biosciences). Brains were chopped into small pieces and incubated in HBSS containing 10% FCS with collagenase/dispase (final concentration 2 mg/ml) (Sigma-Aldrich, Gillingham, U.K.) for 45 min at room temperature. The suspension was filtered through a 70-μm cell strainer before being overlaid on a 30% Percoll gradient and centrifuged at 1800 × g for 10 min. The pellet was collected, and the supernatant was discarded. Spleen and brain samples were treated with RBC lysing buffer (BD Pharmingen) to remove RBCs, washed, and resuspended in flow cytometry buffer (HBSS with 2% FCS). Absolute cell numbers were calculated using a hemocytometer, and live/dead cell differentiation was performed using trypan blue (Sigma-Aldrich). To examine T cell activation, 1 × 10⁶ cells/sample were surface stained with anti-mouse CD4 (RM4.4), anti-mouse CD8 (53-6.7), anti-mouse CD44 (IM7), anti-mouse CD62L (MEL-14), anti-mouse CD69 (H1.2F3), anti-mouse CD71 (R17217), anti-mouse CD27 (LG.7F9), anti-mouse KLRG-1 (2F1), anti-mouse CD25 (PC61), anti-mouse CCR5 (HM-CCR5), anti-mouse CXCR3 (CXCR3-173), and anti-mouse CD45.1 (A20) or permeabilized (0.1% saponin/PBS) and stained with anti-mouse granzyme B (NGZB). NK and NKT cells were identified by staining with anti-mouse CD3 (145-2C11) and anti-mouse CD49b (DX5). To assess intracellular cytokine production, 1 × 10⁶ live cells were incubated with PMA (200 ng/ml) and ionomycin (1 μg/ml) in the presence of brefeldin A (1:1000) for 5 h at 37°C, 5% CO₂. The cells were washed and stained after permeabilizing cells with anti-mouse IFN-γ (XMG1.2) and anti-mouse TNF (MP6-XT22). All Abs were obtained from eBioscience or BD Biosciences. Fluorescence minus one controls and isotype-control Abs were used to validate flow cytometric results. Flow cytometric acquisition was performed using a FACSCalibur (BD Immunocytometry Systems), LSR II (BD Immunocytometry Systems), or Miltenyi MACSQuant (Miltenyi Biotec), and all analyses were performed using FlowJo software (TreeStar, Ashland, OR).

To examine the status (permeability) of the blood–brain barrier, mice were injected i.v. with 200 μl 1% Evans blue (Sigma-Aldrich). After 1 h, mice were sacrificed, and the brains were removed following whole-body perfusion with 15 ml PBS. In separate experiments, brains were removed from mice (following perfusion) that were not injected with Evans blue and placed in 10% formol saline. Preserved samples were then sectioned and stained with H&E (Independent Histopathology Services, London, U.K.). The presence of hemorrhages, inflamed and damaged blood vessels (including occluded vessels), and perivascular cuffing was examined by microscopy.

RNA isolation from brains was performed using RNeasy isolation kits, according to the manufacturer’s instructions (QIAGEN). Isolated RNA was treated with DNAse to remove genomic DNA prior to synthesis of cDNA. cDNA expression for each sample was standardized using the housekeeping gene β-actin. Validated TaqMan gene-expression assays for β-actin, CXCL9, CXCL10, VCAM-1, and ICAM-1 were purchased from ABI Biosystems (Warrington, U.K.). PCR cycling (TaqMan, ABI 7500 fast RT-PCR) conditions were 50°C for 2 min, 95°C for 10 min followed by 40 cycles of 15 s at 95°C, completed with 1 min at 60°C. Parasite levels were determined by SYBR Green PCR using P. berghei 18s and P. berghei carbamoyl phosphate synthetase DNA primers, normalized to mouse β-actin or tyrosine 3-monooxygenase activation protein (YWHAZ), respectively. P. berghei 18s primers: 5′-AAGCATTAAATAAAGCGAATACATCCTTAC-3′ and 5′-GGAGATTGGTTTTGACGTTTATGTG-3′; β-actin primers: 5′-GTGGGCCGCTCTAGGCACCAA-3′ and 5′-CTCTTTGATGTCACGCACGATTTC-3′; carbamoyl phosphate synthetase primers 5′-TGGAATGTGTGAACATGATAAATA-3′ and 5′-TTTCCTGGCCCATTTGATAA-3′; YWHAZ primers: 5′-TTCTATTCCTCTATTTCCATGTTGG-3′ and 5′-AGGAGGAGGAGGAAGAGGAG-3′. All PCR data are presented as fold change (log₁₀) in gene expression in brains from infected IFN-γ^−/− mice that received adoptively transferred CD4⁺ T cells or infected control IFN-γ^−/− mice that did not receive CD4⁺ T cells relative to levels in brains from naive mice. Results for each gene were calculated using the Equation 2^{^(average normalized naive − normalized sample)}.

Splenic single-cell suspensions were prepared from naive WT or RAG-1^−/− mice, as described above. In separate experiments, 40 × 10⁶ WT splenocytes or 5 × 10⁶ RAG-1^−/− splenocytes (providing equivalent numbers of NK cells) were adoptively transferred into recipient mice by i.v. injection on the day before (−1) or on day 5 of P. berghei ANKA infection. Splenic CD4⁺ and CD8⁺ T lymphocytes were positively selected using anti-mouse–conjugated MidiMACS beads, according to the manufacturer’s instructions (Miltenyi Biotec). A total of 5–10 × 10⁶ purified T cell populations was adoptively transferred, depending on the experiment, into recipient mice by i.v. injection on the day before (−1) or on day 5 of P. berghei ANKA infection. The purity and phenotype of positively selected T cell populations were assessed by flow cytometry prior to adoptive transfer and were typically found to be >90%. CD4⁺ and CD8⁺ T cells were ∼95% TCR αβ⁺ and CD3⁺. Representative plots showing the purity and heterogeneous activation status of donor CD4⁺ T cells based upon CD44, CD62L, CD71, CD27, CXCR3, and CCR5 expression were provided for review during submission.

The normality of all data was tested using the D’Agostino and Pearson omnibus normality test. For comparisons between two groups, statistical significance was determined using either the t test or the Mann–Whitney U test, depending on normality of the data. For comparisons among three or more groups, statistical significance was determined using a one-way ANOVA, with the Tukey post hoc analysis for normally distributed data, or a Kruskal–Wallis test, with Dunn post hoc analysis for nonparametric data. All statistical analyses were performed using GraphPad Prism. In all cases, results were classified as significantly different when p < 0.05.

To assess the roles of IL-17 and IFN-γ in the development of ECM, we infected IFN-γ^−/− mice, IL-17R^−/− mice, and WT control mice with P. berghei ANKA and monitored various parameters of infection (Fig. 1). In accordance with previous observations (14, 31), all WT mice developed ECM and succumbed to infection by day 9 postinfection (p.i.), with a parasitemia < 35% (Fig. 1A, 1B). ECM was defined as a clinical score of 4 or 5 using the grading system described in 2Materials and Methods. In addition, WT mice showed substantial weight loss from day 7 p.i. (Fig. 1C). Importantly, the course of P. berghei ANKA infection in IL-17R^−/− mice, which are deficient in IL-17A and IL-17F responses, was indistinguishable from that observed in WT mice, with similar parasite burdens (except on day 9 p.i., when a greater parasitemia was observed in IL-17R^−/− mice) and weight loss throughout the course of infection (Fig. 1A–C). All IL-17R^−/− mice succumbed to infection by day 9 p.i.; upon histological examination, similar numbers of comparably sized hemorrhages were observed in the brains of IL-17R^−/− and WT mice (Fig. 1D). Furthermore, leukocyte and pRBC accumulation was observed in cerebral blood vessels of both WT and IL-17R^−/− mice (Fig. 1D). In contrast, IFN-γ^−/− mice were protected from the development of ECM and survived to day 27 p.i. before succumbing to hyperparasitemia and severe anemia (Fig. 1A, 1C, data not shown). Consistent with previous findings (7), minimal inflammation was observed in brains of IFN-γ^−/− mice, which were removed at the time when WT and IL-17R^−/− mice developed ECM (Fig. 1D). Thus, these data confirm the essential role of IFN-γ and rule out a major role for IL-17 in the development of cerebral pathology during P. berghei ANKA infection.

IFN-γ is crucial for the development of ECM during P. berghei ANKA infection (6, 7, 32), but the cellular sources of IFN-γ are unclear. Thus, to investigate whether IFN-γ is produced by multiple cell types in a temporal or organ-specific manner or whether one cell type dominates the response, we examined the kinetics of IFN-γ mRNA transcription in IFN-γ reporter (YETI) mice on various days of infection. YETI mice developed ECM and succumbed to infection with comparable kinetics as did WT mice (Supplemental Fig. 1). Use of reporter mice overcame problems associated with low cell recovery from the brain during infection and the need for artificial mitogenic restimulation of cells ex vivo to measure cytokine production. Positive and negative enhanced yellow fluorescent protein (eYFP) gates were defined using non-eYFP WT mice (data not shown). Prior to infection, ∼10–15% of splenic T cells, 20% of lung T cells, and 50% of brain-derived T cells expressed eYFP (Fig. 2A, 2B, data not shown); these cells were all CD44⁺ high (results not shown) and are presumably polyclonal effector/memory T cells generated under homeostatic conditions (30). The higher frequency of eYFP expression in the brain compared with spleen or lung is consistent with the notion that only effector/memory T cells perform immune surveillance within the CNS (33, 34).

There were no consistent changes in the frequencies of eYFP⁺ CD4⁺ and CD8⁺ T cells in the spleen or brain until day 7 of infection, at which point the frequencies of eYFP⁺ CD4⁺ and CD8⁺ T cells increased in both organs (Fig. 2). Similarly, the mean fluorescence intensity (MFI) of eYFP expression (measuring the number of IFN-γ transcripts and consequently their capacity to produce IFN-γ) (35) by CD4⁺ and CD8⁺ T cells was highest on day 7 within both organs (Fig. 2D, 2G). Interestingly, in the brain, eYFP expression (MFI) was significantly higher by CD4⁺ T cells compared with CD8⁺ T cells on day 7 of infection, indicating that CD4⁺ T cells were a greater source of IFN-γ on a cell-by-cell basis (Fig. 2G). Although the frequencies of splenic CD4⁺eYFP⁺ and CD8⁺eYFP⁺ T cells were similar throughout the course of infection, significantly greater absolute numbers of CD4⁺eYFP⁺ T cells were observed in the spleen at all time points compared with CD8⁺eYFP⁺ T cells (Fig. 2E). In contrast, higher numbers of CD8⁺eYFP⁺ T cells were observed in the brain on day 7 of infection (Fig. 2H). Numbers of eYFP⁺CD4⁺ and eYFP⁺CD8⁺ T cells were similar in all organs during the earlier stages of infection (Fig. 2E, 2H).

A high proportion (50–80%) of NK (CD3⁻DX5⁺) and NKT cells (CD3⁺DX5⁺) constitutively expressed eYFP in the spleen and brain of naive mice (Fig. 2A–C, 2F). Neither the frequencies of eYFP⁺ NK cells nor the MFI of expression by NK cells changed significantly during the course of infection in any of the tissues, with the exception of a transient increase in MFI in the spleen on day 5 p.i. (Fig. 2D, 2G, data not shown). The frequencies of eYFP⁺ NKT cells increased slightly on day 7 of infection in the spleen but not the brain (Fig. 2C, 2F), and the MFI of eYFP expression by NKT cells increased on days 5 and 7 in the spleen and on day 7 in the brain (Fig. 2D, 2G, data not shown). These results are consistent with previous reports showing that resting NK and NKT cells contain preformed IFN-γ mRNA and are poised for IFN-γ production (30). Nevertheless, the absolute numbers of splenic NK cells expressing eYFP did not change significantly in the brain, and declined significantly in the spleen, during infection (Fig. 2E, 2H). The absolute numbers of NKT cells expressing eYFP were low in comparison with the other populations and did not change significantly in any of the organs examined (Fig. 2E, 2H).

We next analyzed the total IFN-γ (eYFP) response within the spleen and brain during P. berghei ANKA infection to define the relative contribution of IFN-γ production from each cell type to the overall IFN-γ response (Fig. 3). CD4⁺ T cells were found to be the dominant eYFP⁺ population in the spleen throughout the course of infection, accounting for ∼50% of all eYFP⁺ cells. The NK cell contribution to the splenic eYFP response decreased significantly (from ∼25% to <5%) during infection, being replaced principally by CD8⁺ T cells. NKT cells and other cells represented ∼12% of the total splenic eYFP⁺ population throughout the course of infection (Fig. 3B). NK cells were the dominant population of eYFP⁺ cells in the brain in naive mice but were rapidly outcompeted by CD4⁺ and CD8⁺ T cells during infection; CD8⁺ T cells became the dominant eYFP⁺ population in the brain (Fig. 3B). NKT cells and other cells represented ∼15% of the eYFP⁺ population in the brain (Fig. 3B). Combined, these data show that multiple cellular populations contribute to the IFN-γ response during infection, and the relative contribution of each population varies with stage of infection and tissue location.

Because several cell populations can produce IFN-γ during P. berghei ANKA infection, we established an adoptive-transfer model to examine which population(s) were the essential source(s) of IFN-γ leading to development of ECM. To demonstrate that it is possible to promote signs of ECM in IFN-γ^−/− mice, we first adoptively transferred naive whole splenocytes from WT mice into IFN-γ^−/− mice 1 d prior to infection. One hundred percent of IFN-γ^−/− recipients of naive WT whole splenocytes developed prodromal signs of ECM (mean grade, 2.5, Table I), although none of the mice ultimately developed terminal cerebral complications. In contrast, adoptive transfer of naive splenocytes from RAG-1^−/− mice into IFN-γ^−/− mice failed to induce any signs of ECM (grade 1). Because these results suggested the importance of T cell sources of IFN-γ in the promotion of pathology, we next transferred αβ TCR⁺ T cells purified from naive WT mice into IFN-γ^−/− mice 1 d prior to infection with P. berghei ANKA. Importantly, adoptive transfer of naive WT TCR⁺ T cells also promoted prodromal signs of ECM in 100% of IFN-γ^−/− recipients (over three separate experiments), but the majority of animals did not progress to terminal ECM (mean grade, 3.1) (Table I). Interestingly, CD4⁺ T cells appeared to be the most important source of pathogenic IFN-γ; adoptive transfer of naive CD4⁺ T cells into IFN-γ^−/− recipients 1 d before infection conferred partial susceptibility to infection, with animals showing some prodromal signs of ECM (mean grade, 2.8) (Table I), whereas adoptive transfer of naive CD8⁺ T cells into IFN-γ^−/− recipients 1 d before infection did not induce any signs of ECM (grade 1).

Taken together, the above data strongly indicated that CD4⁺ T cells, and not CD8⁺ T cells or innate cells, may be the most important source of pathogenic IFN-γ during P. berghei ANKA infection. However, because of the failure to induce terminal ECM in IFN-γ^−/− mice following adoptive transfer of any population(s) of naive WT cells, we modified the model by adoptively transferring infection-derived WT leukocytes into IFN-γ^−/− recipients. Based upon previous data showing that adoptive transfer of infection-derived WT CD8⁺ T cells into naive CXCR3^−/− mice immediately prior to infection promoted the development of fatal cerebral malaria (11, 12), we hypothesized that this approach may provide greater sensitivity to assess the ability of specific IFN-γ–producing cell populations to induce terminal ECM. Because our initial results suggested the importance of IFN-γ production by αβ TCR⁺ T cells, we initially transferred WT αβ TCR⁺ T cells (1:1 ratio of CD4⁺/CD8⁺ T cells), harvested from spleens of infected donor mice on day 5 of P. berghei ANKA infection (which represents the peak of T cell expansion and activation) (31), into IFN-γ^−/− recipients on day −1 of infection. All recipient IFN-γ^−/− mice developed prodromal signs of ECM (mean clinical grade, 3.25), but only 25% of mice that received cells on day −1 subsequently developed terminal ECM (Table I). As expected, adoptive transfer of infection-derived (day +5) RAG-1^−/− splenocytes into IFN-γ^−/− recipients on day −1 of infection did not promote any signs of disease (clinical grade 1). Notably, adoptive transfer of infection-derived (day +5) αβ TCR⁺ T cells on day +5 of infection also promoted signs of disease in 100% of recipient mice (mean clinical grade, 3.75), and 50% of mice that received cells subsequently developed terminal ECM (Table I). Combined, these data indicated that IFN-γ production specifically by αβ TCR⁺ T cells during the latter stages of infection (i.e., after day 5) can cause cerebral pathology, although our data do not rule out additional roles earlier in infection. In agreement with this dual-role hypothesis, adoptive transfer of infection-derived WT αβ TCR⁺ T cells into IFN-γ^−/− mice on days −1 and +5 of P. berghei ANKA infection induced late-stage ECM in all recipients (clinical grade ≥ 4), requiring all animals to be euthanized on day 6 or 7 (depending on the experiment) (Table I).

Using the same infection-derived adoptive-transfer system, we next examined whether IFN-γ–sufficient effector CD4⁺ or CD8⁺ T cell populations were individually able to promote ECM. Adoptive transfer of infection-derived (day +5) WT CD4⁺ or CD8⁺ T cell populations on days −1 and +5 did not affect peripheral parasitemia during the early stages of infection (Fig. 4A). Importantly, all IFN-γ^−/− recipients of infection-derived WT CD4⁺ T cells developed terminal ECM and were euthanized on day 6–7 p.i. (mean grade, 4.6), showing similar physical signs and brain pathology as infected WT mice, including cortical and cerebellar edema and hemorrhage (Fig. 4B–D, Table I, Table II). Adoptive transfer of infection-derived (day +5) CD4⁺ T cells into IFN-γ^−/− mice on day 5 of infection also caused prodromal signs of ECM (mean score, 3.0) (Table I). In contrast, and somewhat surprisingly, adoptive transfer of similar numbers of infection-derived WT CD8⁺ T cells on days −1 and +5 did not induce cerebral pathology in IFN-γ^−/− mice, and all recipients survived the acute phase of P. berghei ANKA infection (Fig. 4B, Table I). To our knowledge, these experiments are the first to demonstrate that IFN-γ production solely by CD4⁺ T cells, and not by CD8⁺ T cells or innate cells, can lead to the development of terminal ECM.

To define the mechanistic basis for the induction of ECM by WT CD4⁺ T cells, we first determined the organ/tissue localization of adoptively transferred infection-derived CD4⁺ T cells in recipient IFN-γ^−/− mice. Donor Ly5.1⁺ C57BL/6 CD4⁺ T cells were clearly observed in the spleen, lung, and brain of recipient (Ly5.2⁺) IFN-γ^−/− mice on day 6 or 7 of infection (depending on experiment), when the recipient mice developed ECM (Fig. 5A, 5B). These adoptively transferred Ly5.1⁺CD4⁺ T cells maintained their ability to produce IFN-γ when restimulated in vitro (Fig. 5C, 5D). Importantly, neutralization of IFN-γ production in vivo by administration of anti–IFN-γ Ab (from days −2 to +6 of infection) completely blocked the ability of adoptively transferred CD4⁺ T cells to induce ECM in IFN-γ^−/− recipients (Fig. 5E). These data indicate that infection-derived WT CD4⁺ T cells promote ECM in IFN-γ^−/− mice by active secretion of IFN-γ, rather than through IFN-γ–independent effector functions imprinted during priming in an IFN-γ–sufficient (WT) environment.

To determine whether adoptive transfer of infection-derived IFN-γ–competent CD4⁺ T cells modified the host’s endogenous splenic T cell response, which may have contributed to the development of ECM, we assessed the ability of endogenous splenic CD4⁺ and CD8⁺ T cells of IFN-γ^−/− infected mice to produce TNF (Fig. 6). Although the role of TNF in promoting ECM in C57BL/6 mice during P. berghei ANKA infection has been questioned (36, 37), it is an informative measure of effector T cell proinflammatory functionality. Adoptive transfer of infection-derived WT (CD45.1⁺) CD4⁺ T cells did not significantly affect TNF production by host (CD45.1⁻) splenic CD4⁺ or CD8⁺ T cells (Fig. 6A–C); indeed, host splenic CD4⁺ T cells displayed an equivalent capacity to produce TNF during infection compared with adoptively transferred IFN-γ–competent CD45.1⁺CD4⁺ T cells (Fig. 6D).

In addition, the numbers of endogenous splenic CD4⁺ and CD8⁺ T cells were similar in IFN-γ^−/− mice that did (transfer) or did not (no transfer) receive WT CD4⁺ T cells (Fig. 6E, 6F); consistent with the lack of effect on TNF production, there were no significant differences in the frequencies of splenic CD4⁺ or CD8⁺ T cells that displayed an activated phenotype (CD44⁺, CD62L^low, CD69⁺, CD71⁺, or GrB⁺) (Fig. 6G–P). Together, these data demonstrate that IFN-γ–producing CD4⁺ T cells cause ECM directly rather than by modulating the endogenous splenic T cell response. Although we previously showed that IFN-γ induces contraction of the splenic T cell response during P. berghei ANKA infection by promoting T cell apoptosis (31), adoptive transfer of IFN-γ–producing CD4⁺ T cells did not reduce endogenous splenic T cell numbers (Fig. 6E, 6F), suggesting that the timing, concentration, location, or source of IFN-γ that drives the development of ECM may differ from the IFN-γ response that drives splenic T cell apoptosis.

Because it is well established that CD8⁺ T cell migration to and accumulation within the brain is required for the development of ECM in susceptible strains of mice (1, 3, 14), we hypothesized that transfer of infection-derived WT CD4⁺ T cells would enhance the accumulation of host CD8⁺ T cells in the brains of recipient IFN-γ^−/− mice. As expected, significantly higher numbers of host CD8⁺ T cells (but not host CD4⁺ T cells) were observed in brains of IFN-γ^−/− mice that received infection-derived IFN-γ–competent CD4⁺ T cells than in brains of infected IFN-γ^−/− mice that did not receive WT T cells (Fig. 7A, 7B). In addition, although adoptive transfer of WT CD4⁺ T cells did not affect the frequencies of brain-accumulating host CD8⁺ T cells that expressed CD44, adoptive transfer of WT CD4⁺ T cells did significantly increase the frequencies of brain-accumulating host CD8⁺ T cells that were CD62L^low, granzyme B⁺, and KLRG-1⁺, indicating that local CD4⁺ T cell-derived IFN-γ may influence terminal differentiation and function of CD8⁺ T cells in the brain (Fig. 7C–F). As expected, Ab-induced depletion of host CD8⁺ T cells in IFN-γ^−/− mice that had received WT effector CD4⁺ T cells prevented the development of ECM (Fig. 7G–I): anti-CD8–treated mice remained healthy until the termination of the experiment on day 14. Together, these data show that IFN-γ–producing CD4⁺ T cells induce ECM in a CD8⁺ T cell-dependent manner but, crucially, the CD8⁺ T cells themselves do not need to be able to make IFN-γ.

To determine the mechanism by which IFN-γ–competent, infection-derived CD4⁺ T cells mediate CD8⁺ T cell accumulation within the brains of IFN-γ^−/− mice during P. berghei ANKA infection, we compared the expression of CXCL9, CXCL10, ICAM-1, and VCAM-1 mRNA in the brains of IFN-γ^−/− mice that received infection-derived IFN-γ–competent CD4⁺ T cells relative to mRNA levels in control IFN-γ^−/− mice that did not receive WT CD4⁺ T cells. At the time (day 6 or 7, depending on the experiment) when IFN-γ^−/− mice that received WT CD4⁺ T cells developed ECM, significantly higher levels of CXCL9 and CXCL10 mRNA were detected in their brains than in brains of control IFN-γ^−/− mice (Fig. 7J). In contrast, ICAM-1 and VCAM-1 expression was similar in the two groups of mice (Fig. 7J). Importantly, neutralization of IFN-γ significantly reduced the expression of CXCL9 and CXCL10 in the brain of IFN-γ^−/− mice that received infection-derived IFN-γ–competent CD4⁺ T cells (Fig. 7J). To our knowledge, these results suggest, for the first time, that IFN-γ produced exclusively by CD4⁺ T cells is sufficient to modulate the cerebral environment, leading to the upregulated expression of key chemokines that control pathogenic CD8⁺ T cell accumulation in the brain (10–13).

Adoptive transfer of infection-derived WT CD4⁺ T cells did not significantly alter the level of peripheral parasitemia in recipient IFN-γ^−/− mice (Fig. 4A), suggesting that CD4⁺ T cells promoted ECM without significantly altering parasite burdens. Importantly, however, peripheral parasitemia does not necessarily reflect total tissue parasite biomass (38). Thus, we specifically examined whether adoptive transfer of IFN-γ–competent CD4⁺ T cells enhanced parasite sequestration/accumulation within the brain of IFN-γ^−/− mice. Surprisingly, parasite gene expression (detecting P. berghei 18S and Carbamoyl phosphate synthetase genes) was not significantly increased in the brains of IFN-γ^−/− mice that received IFN-γ–competent CD4⁺ T cells compared with the levels in the brains of IFN-γ^−/− control-infected mice (Fig. 8). Moreover, we did not detect differences in total parasite biomass following adoptive transfer of infection-derived CD4⁺ T cells using the well-established bioluminescence system (data not shown). Therefore, our results suggest that IFN-γ–producing CD4⁺ T cells can induce ECM development without significantly altering the overall accumulation of parasites within the brain.

ECM has historically been considered an IFN-γ–dependent immunopathological syndrome. However, the discovery of Th17 cells led to the reclassification of many conditions previously deemed to be Th1 dependent, including EAE and inflammatory arthritis (24, 25, 39). In this study, using IL-17R^−/− mice that lack IL-17A and IL-17F responses (40), we found no evidence for a role for IL-17A or IL-17F in the development of cerebral pathology during P. berghei ANKA infection, which is consistent with a recent report demonstrating that IL-17A knockout mice are susceptible to ECM development (41). These results reaffirm the pivotal and IL-17–independent role of IFN-γ in the genesis of this condition. Indeed, we did not detect Th17 responses in WT or IFN-γ^−/− mice at any point during infection (data not shown), indicating that—even in the absence of IFN-γ, which has been reported to block Th17 cell development (42)—P. berghei ANKA infection fails to generate an environment suitable for Th17 differentiation.

It is perhaps unsurprising that IFN-γ mRNA (as shown by eYFP) is expressed and/or upregulated by multiple cell populations during P. berghei ANKA infection. Nevertheless, our study demonstrates that the cell-specific contribution to the total eYFP response is governed by both the precise stage of infection and organ of location; some cell populations upregulate eYFP within tissues, whereas other eYFP⁺ populations are lost. These data have important implications for not only understanding the protective and pathogenic function(s) of IFN-γ during infection but also for defining the tissue-specific signals required for the development and/or maintenance of IFN-γ–producing cell populations. Although upregulation of eYFP does not guarantee that a cell will go on to secrete IFN-γ (eYFP is a faithful marker of transcription but not translation), because P. berghei ANKA is an acute infection, covering only the early phases of T cell activation, it is highly probable that induction of eYFP in effector T cells reflects active protein translation (35). However, if NK cells and NKT cells are unable to rapidly translate constitutively expressed IFN-γ mRNA during infection (30), it is possible that their contribution to the overall IFN-γ response during infection may be overestimated in our analysis.

More importantly than dissecting the sources of IFN-γ during P. berghei ANKA infection, we demonstrated that IFN-γ produced solely by CD4⁺ T cells is sufficient to promote the development of ECM. Although we acknowledge that we were only able to induce terminal ECM in IFN-γ^−/− mice by adoptive transfer of infection-derived CD4⁺ T cells, and not by adoptive transfer of naive CD4⁺ T cells, we showed that infection-derived RAG-1^−/− cells and infection-derived WT CD8⁺ T cells are unable to promote any signs of ECM when adoptively transferred into IFN-γ^−/− mice. Moreover, we showed that naive WT CD4⁺ T cells, but not naive RAG-1^−/− cells or naive WT CD8⁺ T cells, promote comparable levels of disease as do whole naive splenocytes when adoptively transferred into IFN-γ^−/− mice. Thus, when our naive and infection-derived adoptive transfer results are combined, our data very convincingly show that CD4⁺ T cells are the predominant source of pathogenic IFN-γ during P. berghei ANKA infection. Although previous studies showed that CD4⁺ T cells are involved in the development of ECM (7, 43-45), to our knowledge, this is the first definitive proof that they do so specifically through IFN-γ production. Our inability to induce ECM through naive cell transfers is potentially related to the specific immunological conditions induced during P. berghei ANKA infection, which possibly prevents the priming, maintenance, and survival of sufficient numbers of transferred cells in the recipient mice to cause ECM. Indeed, we and other groups have previously described the proapoptotic environment induced by P. berghei ANKA infection (31, 46), which causes the contraction of the T cell effector response.

CD4⁺ T cells, rather than CD8⁺ T cells, may be the most important source of IFN-γ during P. berghei ANKA infection, simply because—in the brain, at least—they produce much more IFN-γ on a per-cell basis than do CD8⁺ T cells. But it is also possible that CD4⁺ and CD8⁺ T cells accumulate in different anatomical locations within the brain (e.g., in the subarachnoid space or perivascular spaces, where localized IFN-γ production is more, or less, damaging). Crucially, however, we showed that IFN-γ–producing CD4⁺ T cells cannot, by themselves, induce ECM in IFN-γ^−/− mice; adoptive transfer of IFN-γ–producing CD4⁺ T cells into infected IFN-γ^−/− mice increased endogenous CD8⁺ T cell accumulation within the brain, and depletion of endogenous CD8⁺ T cells prevented the onset of ECM. Thus, our data indicate that IFN-γ–producing CD4⁺ T cells drive the development of ECM during P. berghei ANKA infection by facilitating the entry and accumulation of CD8⁺ T cells into the brain, which then mediate pathology through IFN-γ–independent mechanisms.

Although it was recently shown that CD4⁺ T cells are required for the activation of pathogenic CD8⁺ T cells during P. berghei ANKA infection (45), we do not believe that IFN-γ–producing CD4⁺ T cells simply activate CD8⁺ T cells in the lymphoid compartment or periphery during infection, allowing their migration to the CNS, because transfer of WT infection-derived CD8⁺ T cells (which were primed in the presence of IFN-γ–sufficient CD4⁺ T cells) was not sufficient to induce ECM. While it is possible that IFN-γ–producing CD4⁺ T cells directly modulate CD8⁺ T cell effector function after day 5 of infection, when the CD8⁺ T cells were purified for adoptive transfer, we do not believe this to be likely because the activation and effector functions of splenic CD8⁺ T cells are seemingly unimpaired in IFN-γ^−/− mice on day 7 of infection, when WT mice develop ECM (31). These data, when taken together with our finding in this study that adoptive transfer of WT CD4⁺ T cells into infected IFN-γ^−/− mice leads to upregulation of the chemokines CXCL9 and CXCL10 within the brain, lead us to conclude that the major role of IFN-γ–secreting CD4⁺ T cells, which accumulate in the brain through IFN-γ–independent mechanisms (31), is to modulate the localized brain environment, facilitating the migration/accumulation of CD8⁺ T cells into the brain. This hypothesis is consistent with upregulation of genes associated with IFN signaling in the brains of ECM-susceptible mice infected with P. berghei ANKA (8, 9) and with studies showing much lower expression of CXCL9 and CXCL10 in brains of ECM-resistant IFN-γ^−/− mice than in brains of susceptible WT mice during P. berghei ANKA infection (7, 10, 31). The CD8⁺ T cell effector mechanisms responsible for progression to end-stage ECM are poorly understood; the data presented in this article indicate that CD8⁺ T cells do not need to produce IFN-γ, which is in agreement with a very recent study indicating that CD8⁺ T cells do not need to express IFN-γ for their accumulation or pathogenic function within the brain (47); however, evidence from other studies indicates a role for granzyme B-dependent and perforin-mediated killing of brain endothelial cells (47–49).

We also showed that adoptive transfer of IFN-γ–producing CD4⁺ T cells does not affect total parasite transcript levels within the brain of IFN-γ^−/− mice during P. berghei ANKA infection. We also failed to detect differences in parasite biomass with the bioluminescence system, as used by Amante et al. (16) and Claser et al. (15) (data not shown but provided to reviewers during manuscript review). These results are surprising, because a number of recent studies showed that IFN-γ–dependent parasite accumulation in the brain is a key feature of ECM and is necessary for the development of pathology (15, 16, 49-51). However, there is some debate on the precise role of CD4⁺ T cells in mediating parasite accumulation within tissues during P. berghei ANKA infection; Amante et al. (16) recently suggested that CD4⁺ T cells are required for parasite accumulation within tissues, whereas Claser et al. (15) reported that CD4⁺ T cells do not facilitate parasite accumulation within the brain. Nonetheless, our results suggest that, in our ECM model, extensive parasite accumulation within the brain is not essential for the development of cerebral pathology, which may underline the heterogeneity of the ECM syndrome. However, although adoptive transfer of CD4⁺ T cells did not significantly increase total parasite levels within the brain, increased accumulation within specific and small regions of the brain would not necessarily have been detected in our analysis. In this scenario, parasite accumulation within a small number of specific and focal regions may have been sufficient to cause ECM. Crucially, we do not yet understand the precise regions of the brain that are most sensitive to parasite accumulation and inflammation during malaria infection, leading to death (52). Identifying whether (and how) P. berghei ANKA parasites sequester within the brain microvessels and the relative importance of this process for the development of ECM are ongoing areas of research in our and other laboratories.

In summary, we have examined the cellular sources of IFN-γ during P. berghei ANKA infection and, to our knowledge, highlighted for the first time that CD4⁺ T cells promote the development of ECM specifically through their production of IFN-γ. Our results suggest that IFN-γ produced by CD4⁺ T cells may act locally within the brain to modify the brain environment, which is required for the accumulation of CD8⁺ T cells within the brain, leading to ECM. These data represent a significant advancement in our understanding of the pathogenesis of ECM during P. berghei ANKA infection, which may have implications for understanding, and, therefore, either preventing or treating, human cerebral malaria.

We thank Dr. Chris Engwerda and Dr. Ash Haque (Queensland Institute of Medical Research, Brisbane, QLD, Australia) for discussion and critical reading of the manuscript.

The authors have no financial conflicts of interest.