Heterogeneous and Tissue-Specific Regulation of Effector T Cell Responses by IFN-γ during Plasmodium berghei ANKA Infection

Plasmodium berghei ANKA infection in susceptible strains of mice leads to the development of experimental cerebral malaria (ECM), a neuropathy that shares many similarities with human cerebral malaria (reviewed in Refs. 1 and 2). The pathogenesis of the terminal stages of ECM is still poorly defined, but it is clear that the prototypic type 1 cytokine IFN-γ plays a pivotal role in the development of cerebral pathology (3–5). Thus, IFN-γ–deficient (IFN-γ^−/−) and IFN-γR–deficient (IFN-γR^−/−) mice on susceptible backgrounds, including C57BL/6, fail to develop ECM (3, 4). Notably, high levels of circulating IFN-γ and upregulation of IFN-responsive genes are also correlated with development of cerebral malaria in humans (reviewed in Ref. 6).

It is currently unclear whether the resistance of IFN-γR^−/− mice to ECM is primarily due to alterations in innate cell or T cell activity or a combination of both. Both macrophage accumulation within the brain and macrophage function (TNF secretion) are reduced in P. berghei ANKA-infected IFN-γR^−/− mice compared with infected wild-type (WT) mice (3), suggesting that IFN-γ–responsive macrophages may contribute to the etiology of ECM. In contrast, IFN-γ regulates the expression of CCR2, CXCR3, ICAM-1, VCAM-1, and LFA-1 (3, 7) and the production of CCL5 (RANTES), CXCL10 (IFN-γ–inducible protein-10 [IP-10]), and CXCL9 (monokine induced by IFN-γ [MIG]) in the brain during P. berghei ANKA infection (4, 7), suggesting that IFN-γ may contribute to ECM by directing the recruitment of T cells to the brain. In support of this hypothesis, CD8⁺ T cells, primed in the spleen during P. berghei ANKA infection via cross-presentation of Ag by classical lymphoid dendritic cells (8–10), have been shown to migrate to the brain via CXCR3-CXCL10 (IP-10)-, MIG (CXCL9)-, and CCR5-CCL5 (RANTES)-dependent pathways (4, 7, 11–13). T cell migration to the brain is thus reduced in IFN- γ(R)^−/− (7). Because migration of CD8⁺ T cells to the brain is believed to be a key process in the development of ECM (reviewed in Ref. 14), through modulating parasite tissue biomass and/or causing direct endothelial cell damage (15–18), these data suggest that alterations in chemokine pathways and resultant attenuated migration of CD8⁺ T cells is a major reason for the resistance of IFN-γ^−/− and IFN-γR^−/− mice to ECM.

There is, however, significant evidence from other models that IFN-γ may directly control the activation and expansion of T cells: IFN-γ drives STAT1-dependent expression of T-bet in CD4⁺ T cells, which is the initial step in the differentiation of Th1 cells (19, 20). IFN-γR signaling is also required to repress IL-4 production by Th1 cells during recall responses (21), and IFN-γ modulates microglial activation within the brain during experimental autoimmune encephalitis, controlling both Th1 and Th2 cell activation (22). Consequently, in the absence of direct IFN-γR signaling, Th1 and CD8⁺ T cell responses are impaired (23–25). Thus, reduced accumulation of T cells in the brains of IFN-γ^−/− mice during P. berghei ANKA infection may not be solely because of altered chemotactic signals but may also be a consequence of impaired T cell activation or differentiation.

In addition to its immunostimulatory effects, IFN-γ may also suppress T cell hyperactivity by limiting CD4⁺ and CD8⁺ T cell accumulation or expansion through the induction of apoptosis (26–33) and by deletion of Ag-presenting dendritic cells (34). Although few studies have examined in detail the apoptotic pathways regulated by IFN-γ, it has been shown that intrinsic (mitochondrial) and extrinsic (TRAIL, DR5, and TNFR1) pathways of apoptosis are induced in CD4⁺ T cells by IFN-γ during bacillus Calmette-Guérin infection (28), and IFN-γ promotes caspase-8–dependent activation-induced cell death of CD4⁺ T cells in vitro following TCR stimulation (32). Interestingly, it has been suggested that IFN-γ–dependent apoptosis can occur by paracrine signaling without expression of IFN-γR on T cells (29, 33) and through autocrine cell-specific IFN-γ signaling (30). With relevance to malaria infection, Ag-specific (but not nonspecific) T cell apoptosis has been described during P. berghei ANKA, P. yoelii, and P. chabaudi AS infections (27, 35, 36). Loss of splenic Ag-specific CD4⁺ T cells was reduced when cells were adoptively transferred into mice treated with anti–IFN-γ (31), suggesting that IFN-γ was responsible for apoptosis. Importantly, however, the contribution of modified migratory behavior following treatment with anti–IFN-γ leading to changes in splenic T cell numbers was not addressed. In separate studies, Ag-specific T cell apoptosis has been shown to occur during malaria infection through distinct death receptor- and bax-dependent pathways (37).

In this study, we have compared the T cell response in WT and IFN-γ^−/− mice during P. berghei ANKA infection to characterize the alterations in T cell-mediated immune functions that underlie the resistance of IFN-γ^−/− mice to ECM. Our results show that IFN-γ is not required for the activation of splenic CD4⁺ and CD8⁺ T cells during infection but that it is essential for contraction of the splenic effector T cell population through induction of apoptosis. Moreover, we demonstrate that IFN-γ directs tissue-specific migration of CD8⁺ T cells and CD4⁺ T cell, but not CD8⁺ T cell, activation and effector function within peripheral organs. Importantly, we show that persistence of effector T cells in the spleen in IFN-γ^−/− mice is due to the absence of active environmental cues during the precise period of ECM development, rather than due to intrinsic differences in T cell programming imprinted during the early stages of infection.

C57BL/6 (CD45.2⁺), C57BL/6 IFN-γ knockout (IFN-γ^−/−), and C57BL/6 Ly5.1 (CD45.1⁺) mice were bred in-house or purchased from Harlan (Oxford, U.K.) and maintained in individually ventilated cages. Male and female mice were used in separate experiments and were between 6 and 9 wk of age. Cryopreserved P. berghei ANKA parasites were thawed and passaged once in vivo in C57BL/6 mice before being used to infect experimental animals. Experimental mice were infected i.v. with 10⁴ parasitized RBCs (pRBC). Parasitemia was monitored daily by microscopic examination of Giemsa (VWR International)-stained thin blood smears. The severity of ECM was determined using the following grading system: 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 euthanized when observed at stage 4 or 5.

Single-cell suspensions of spleen, liver and lung were prepared by homogenizing tissues through a 70-μm cell strainer (BD Falcon). Brains were cut into small pieces, aspirated through a 5-ml syringe, and incubated in HBSS containing 10% FCS with collagenase/dispase (2 mg/ml) (Sigma-Aldrich) for 45 min at room temperature. The brain suspension was then homogenized through a 70-μm cell strainer, overlaid on a 30% Percoll gradient, and centrifuged at 1800 × g for 10 min. The mononuclear cell pellet was collected. RBC were lysed in all samples by the addition of lysing buffer (BD Biosciences). 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 stained with anti-CD4 (RM4.5), anti-CD8 (53-6-7), anti-CD11a (M17/4), anti-CD25 (PC61), anti-CD27 (LG.7F9), anti-CD44 (1M7), anti-CD62L (MEL-14), anti-CD69 (H1-2F3), anti-CD71 (R17217), anti-CXCR3 (CXCR3-173), anti-CD45.1 (A20), anti-CD45.2 (104), anti-CD49D (R1-2), anti-programmed death-1 (PD-1) (RMP1-30), annexin V, and 7-aminoactinomycin D (7-AAD). Intracellular staining for granzyme B (16G6), anti–CTLA-4 (UC10-4B9), and anti-Ki67 (B56) was performed by permeabilizing cells with 0.1% saponin/PBS. To quantify macrophage/monocyte numbers, 1 × 10⁶ cells/sample were stained with anti–F4-80 (BM8), anti-CD11b (M1/70), and anti-CD3 (17A2) to exclude lymphocytes. To assess ex vivo cytokine production, 1 × 10⁶ live cells were incubated with PMA (200 ng/ml) and ionomycin (1 μg/ml) in the presence of brefeldin A (eBioscience) for 5 h at 37°C and 5% CO₂. The cells were washed, and intracellular staining with anti-mouse IFN-γ (XMG1.2) and anti-mouse TNF (MP6-XT22) was performed as above. All Abs were obtained from eBioscience or BD Biosciences. Flow cytometric acquisition was performed using a FACSCalibur or LSR-II (both BD Immunocytometry Systems), and all analysis was performed using FlowJo software (Tree Star).

TCRβ⁺ or CD4⁺ and CD8⁺ T lymphocytes were positively selected using anti–mouse-conjugated midiMACS beads (Miltenyi Biotec) according to the manufacturer’s instructions. In some experiments, purified CD4⁺ and CD8⁺ T cells were snap frozen in liquid nitrogen and stored at −80°C for real-time PCR analysis. In separate experiments, 5–7 × 10⁶ purified T cells were adoptively transferred into recipient mice on day 5 postinfection with P. berghei ANKA. The purity of positively selected T cell populations was assessed by flow cytometry prior to adoptive transfer or PCR analysis and was typically found to be >90%.

RNA isolation from purified CD4⁺ and CD8⁺ T cells was performed using RNeasy isolation kits according to the manufacturer’s instructions (Qiagen). Isolated RNA was DNAse treated to remove genomic DNA prior to synthesis of cDNA. cDNA expression for each sample was standardized using the housekeeping genes GAPDH or β-actin. Data are presented as fold change (log₁₀) in gene expression in infected IFN-γ^−/− cells/tissues relative to infected WT cells/tissues. The results for each gene were calculated using the equation 2^{^(Average normalized WT − normalized IFN-γ−/−)}. Validated gene expression assays for T-bet, RORγT, GATA-3, IL-4, IL-17, perforin, Bcl-2, BAD, BAX, BIM, CXCL9, CXCL10, RANTES, 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 and 95°C for 10 min, followed by 40 cycles of 15 s at 95°C completed with 1 min at 60°C.

Statistical significance was determined using two tailed Student t test, unless otherwise stated, with p < 0.05 taken as indicating a significant difference.

Consistent with previous studies, WT C57BL/6 mice were exquisitely susceptible to the development of ECM during P. berghei ANKA infection. In contrast, IFN-γ^−/− deficient mice were completely resistant to ECM, even though peripheral parasitemia was comparable in WT and IFN-γ^−/− mice throughout the course of infection (Supplemental Fig. 1). To determine whether T cells may be directly controlled by IFN-γ during P. berghei ANKA infection, contributing to the development of ECM, we evaluated the T cell-intrinsic expression of the IFN-γR on naive and activated (CD44⁺) CD4⁺ and CD8⁺ T cells in WT mice in various tissues during the course of infection. We observed that the frequency of CD4⁺ and CD8⁺ T cells expressing the IFN-γR increases during infection in the spleen, lung, liver, and brain (Fig. 1, Supplemental Fig. 2). In naive mice, IFN-γR was expressed predominantly by effector (CD44⁺) CD4 and CD8⁺ T cells in all organs (Fig. 1). In contrast, during malaria infection, IFN-γR was heterogeneously expressed on both naive and effector CD4⁺ and CD8⁺ T cells, and the frequency of IFN-γR⁺ cells (and the mean fluorescence intensity [MFI] of IFN-γR expression) varied significantly between the different tissues (Fig. 1).

To investigate whether the T cell response differed between infected WT and IFN-γ^−/− mice, we first examined the total numbers of splenic CD4⁺ and CD8⁺ T cells during the course of infection: the spleen is believed to be the primary site of immune priming during blood stage malaria infection (reviewed in Ref. 38). Splenic CD4⁺ and CD8⁺ T cell numbers increased at a comparable rate, and plateaued at equivalent levels, in WT and IFN-γ^−/− mice, suggesting that the initial expansion phase of the T cell response was unimpaired in the absence of IFN-γ (Fig. 2A). Indeed, splenic T cell proliferation, as measured by the frequencies of cells expressing Ki67, was comparable in WT and IFN-γ^−/− mice (results not shown but provided for review). However, the subsequent contraction of splenic CD4⁺ and CD8⁺ T cell numbers, between days 5 and 7 postinfection, was markedly attenuated in IFN- γ^−/− mice compared with WT mice (Fig. 2A). Furthermore, in the absence of IFN- γ, expression of the early activation marker CD69 was significantly attenuated on both CD4⁺ and CD8⁺ T cells; this was most evident on day 5 postinfection, and granzyme B expression was reduced on day 7 in CD4⁺ and CD8⁺ T cells (Fig. 2B, 2C). In addition, reduced frequencies of splenic CD4⁺ T cells (but not CD8⁺ T cells) expressed CTLA-4 on day 7 of infection in IFN-γ^−/− mice (Supplemental Fig. 3). Thus, these data suggest that there may be small defects in T cell function in the absence of IFN-γ. In contrast, lack of IFN-γ led to only transient differences in expression of all other examined activation markers. Similar frequencies of CD62L^low, CD44⁺, CD25⁺, CD27^low, CD11a⁺, CD49D⁺, and PD-1⁺ splenic T cells were observed in WT and IFN-γ^−/− mice throughout the course of infection (Fig. 2B, 2C, Supplemental Fig. 3; CD49D results not shown), although because of the failure of the T cell population to contract in IFN-γ^−/− mice, total numbers of activated T cells were higher in these mice than in WT mice by day 7 postinfection.

TNF is a potent proinflammatory cytokine that is induced by IFN-γ and is associated with severe malarial disease (reviewed in Ref. 38). Thus, we examined whether TNF production by CD4⁺ and CD8⁺ T cells was reduced in IFN-γ^−/− mice (Fig. 3). A large proportion (>30%) of CD4⁺ and CD8⁺ T cells from both naive WT and naive IFN-γ^−/− mice produced TNF-α after PMA/ionomycin restimulation. TNF responses were maintained throughout the course of infection and were broadly similar in WT and IFN-γ^−/− mice; however, on day 5 of infection, the proportion of CD4⁺ T cells expressing TNF and the MFI of TNF expression in CD4⁺ and CD8⁺ T cells were significantly lower among IFN-γ^−/− mice than among WT mice (Fig. 3). These data imply that in the absence of IFN-γ, T cells have a slightly reduced capacity to produce TNF-α at a key stage of P. berghei ANKA infection.

As expected, IFN-γ production was completely abrogated in IFN-γ^−/−CD4⁺ and CD8⁺ T cells, whereas CD4⁺ and CD8⁺ T cells producing IFN-γ (alone or in combination with TNF) were observed in WT mice throughout the course of infection. Of note, the CD8⁺ T cells with the highest levels (MFI) of TNF-α production also produced IFN-γ (Fig. 3B).

Perhaps surprisingly, given that IFN-γ has been shown to be an inhibitor of both Th2 and Th17 lineages (39, 40), we did not detect intracellular IL-17A/F at any time in either CD4⁺ or CD8⁺ T cells from either WT or IFN-γ^−/− mice (data not shown); we had expected that expression of Th2- and Th17-associated genes would be increased in CD4⁺ T cells from infected IFN-γ^−/− mice. Indeed, expression of GATA3, RORγT, and IL-4 mRNA was significantly higher in infection-derived IFN-γ^−/−CD4⁺ T cells than in corresponding WT CD4⁺ T cells (Fig. 4A, 4B). Nevertheless, expression of IL-17A, T-bet, and perforin mRNA was similar in WT and IFN-γ^−/−CD4⁺ T cells (Fig. 4A, 4B). Similarly, IL-4 mRNA expression was higher in CD8⁺ T cells from IFN-γ^−/− mice compared with WT mice throughout infection, and although T-bet expression was slightly increased in IFN-γ^−/−CD8⁺ T cells isolated on day 7 of infection compared with corresponding WT cells, the expression of IL-17 and perforin did not differ (Fig. 4C). Thus, in the absence of IFN-γ, there are subtle effects on polarization of—and cytokine production by—splenic CD4⁺ and CD8⁺ T cells during P. berghei ANKA infection, but there is no large-scale change in T cell phenotype that might explain the complete resistance to ECM among IFN-γ^−/− mice.

We hypothesized that migration to peripheral tissues (rather than cell death) was the cause of reduced splenic T cell numbers in WT mice on day 7 of P. berghei ANKA infection and thus that splenic T cell numbers might be maintained in IFN-γ^−/− mice because of their reduced migration. To test this, we enumerated CD4⁺ and CD8⁺ T cells in the brain, lung, and liver of infected WT and IFN-γ^−/− mice. Consistent with previous studies (4), CD8⁺ T cell numbers were significantly higher in the brains of WT mice than those of IFN-γ^−/− mice on day 7 postinfection (Fig. 5A); however, the difference in total brain T cell numbers (∼5000 CD8⁺ T cells/brain) was insufficient to account for the marked loss of T cells in the spleen in WT mice (Fig. 2B). Moreover, similar (or even higher) numbers of CD4⁺ and CD8⁺ T cells were observed in the livers and lungs of IFN-γ^−/− mice than in those of WT mice on day 7 of infection (Fig. 5B, 5C). Thus, although T cell migration does differ between WT and IFN-γ^−/− mice during P. berghei ANKA infection, it is unlikely that the loss of splenic T cells in WT mice is simply because of enhanced T cell migration to peripheral tissues.

To determine whether altered T cell migration and tissue accumulation in IFN-γ^−/− mice was due to cell-intrinsic differences in chemokine receptor expression, we compared T cell expression of CXCR3 and CCR5 [which have been shown to promote T cell migration during P. berghei ANKA infection (7, 11–13)] in WT and IFN-γ^−/− mice. Although alterations in chemokine and chemokine receptor expression have been reported in P. berghei ANKA-infected IFN-γ(R)^−/− mice (4, 7), T cell-specific CXCR3 and CCR5 expression has not been reported. We observed that CXCR3 is constitutively expressed on hepatic CD4⁺ T cells, and its expression is not affected by P. berghei ANKA infection (Fig. 5D), and that CCR5 is upregulated to a similar extent on WT and IFN-γ^−/− hepatic CD4⁺ T cells during infection. CXCR3 and CCR5 are both upregulated by infection on CD4⁺ T cells in the spleen, brain, and lung, but this upregulation is significantly attenuated in IFN-γ^−/− mice compared with WT mice (Fig. 5D, 5E). In contrast, CXCR3 expression on CD8⁺ T cells was upregulated to the same extent in WT or IFN-γ^−/− mice (in all organs) during infection (Fig. 5F). Similarly, CCR5 expression was comparable on CD8⁺ T cells in WT and IFN-γ^−/− mice within the spleen, lung, and liver; expression was, however, reduced in CD8⁺ T cells within the brain of IFN-γ^−/− mice (Fig. 5G). These data suggest that differences in CD8⁺ T cell migration in IFN-γ^−/− mice are more likely to be because of differences in chemokine production within individual tissues than because of differences in chemokine receptor expression on T cells. In support of this conclusion, we observed significantly reduced expression of CXCL9 (MIG) and CXCL10 (IP-10) in the brains of IFN-γ^−/− mice compared with WT mice on day 7 of infection (Supplemental Fig. 4).

Although similar levels of splenic T cell activation were observed in infected WT and IFN-γ^−/− mice, it remained possible that the activation status of T cells migrating to, and accumulating in, the tissues may differ between WT and IFN-γ^−/− mice. Importantly, upregulation of IFN-γR expression in peripheral nonlymphoid tissues in infected WT mice (Fig. 1) suggested that IFN-γ may be particularly important in controlling T cell effector function in peripheral tissues. As ECM and respiratory distress during P. berghei ANKA have been associated with T cell accumulation in brain and lung, respectively, we next compared expression of various activation/memory markers on CD4⁺ and CD8⁺ T cells responses within the brain and lung of infected WT and IFN-γ^−/− mice (Fig. 6). Interestingly, a smaller proportion of the brain-infiltrating CD4⁺ T cells expressed an activated phenotype, as measured by CD44 and CD62L expression, in IFN-γ^−/− mice than in WT mice (Fig. 6A), although—despite the observation that the IFN-γR is expressed to equivalent levels on CD4⁺ T cells in the lung and brain (Fig. 1)—the proportions of activated CD4⁺ T cells in the lungs did not differ between WT and IFN-γ^−/− mice. Moreover, a smaller proportion of brain-infiltrating (but not lung-infiltrating) CD4⁺ T cells expressed CD71, CD11a, and Ki67, but not GrB, in IFN-γ^−/− mice than in WT mice. In contrast, fewer CD4⁺ T cells within the lung and brain expressed CD49D in IFN-γ^−/− mice. Unexpectedly, given the marked reduction in accumulation of CD8⁺ T cells in the brain and lung in IFN-γ^−/− mice (Fig. 5), CD8⁺ T cell activation in the brain and lung was essentially unimpaired in IFN-γ^−/− mice (Fig. 6B). These data suggest that, during P. berghei ANKA infection, CD4⁺ T cells are much more dependent on IFN-γ for their activation within the brain than are CD8⁺ T cells. Furthermore, as Ki67 expression was comparable on brain-accumulating CD8⁺ T cells in WT and IFN-γ^−/− mice (Fig. 6B), our results show that enhanced accumulation of CD8⁺ T cells within the brain of WT mice is not due to increased local extralymphoid proliferation. The expression of ICAM-1 and VCAM-1 was also unimpaired in IFN-γ^−/− mice (Supplemental Fig. 4). Thus, combined, our results also suggest that the VLA-4 and LFA-1 pathways do not dominantly control CD8⁺ T cell accumulation within the brain during P. berghei ANKA infection.

A reduced proportion of brain- and lung-infiltrating CD4⁺ T cells expressed the T cell inhibitory receptors CTLA-4 and PD-1 in IFN-γ^−/− mice compared with WT mice on day 7 of infection (Supplemental Fig. 3). Similarly, reduced proportions of brain- and lung-infiltrating CD8⁺ T cells expressed CTLA-4 in IFN-γ^−/− mice compared with WT mice on day 7 of infection. In contrast, PD-1 expression by CD8⁺ T cells was unaltered in the absence of IFN-γ (Supplemental Fig. 3). These results indicate that defective CD4⁺ T cell activation within the brain of IFN-γ^−/− mice is unlikely because of increased CTLA-4 and PD-1–mediated suppression.

Because the loss of splenic T cells in P. berghei ANKA-infected WT mice did not seem to be due to increased migration of T cells to peripheral tissues, and because IFN-γ is known to control the contraction phase of the effector response by directing T cell apoptosis (26–33), we compared the frequencies of late-stage apoptotic/dead T cells, determined by costaining with 7-AAD and annexinV, in spleens of WT and IFN-γ^−/− mice. We observed no significant differences in apoptotic T cell frequencies between uninfected WT and IFN-γ^−/− mice, indicating that survival of resting T cells is unaffected by presence or absence of IFN-γ (Fig. 7A–C). The frequency of apoptotic CD4⁺ T cells increased significantly in WT mice, but not IFN-γ^−/− mice, during P. berghei ANKA infection (Fig. 7A, 7B). Similarly, significantly increased frequencies of late-stage apoptotic CD8⁺ T cells were observed in infected WT mice compared with infected IFN-γ^−/− mice and uninfected mice, although the frequencies of late-stage apoptotic CD8⁺ T cells were increased in infected IFN-γ^−/− mice compared with uninfected controls (Fig. 7A, 7C).

To determine whether reduced T cell apoptosis in IFN-γ^−/− mice was due to modulation of the intrinsic pathway of apoptosis, we examined, by real-time PCR, the expression of the antiapoptotic regulator Bcl-2 and the proapoptotic molecules BAD, BAX, and BIM in purified populations of CD4⁺ and CD8⁺ T cells (Fig. 7D, 7E). When compared with cells from WT mice, on day 5 of infection, expression of Bcl-2 was increased in CD4⁺ and CD8⁺ T cells from IFN-γ^−/− mice, suggesting that upregulation of Bcl-2 protects these cells from apoptosis. Levels of BAD and BAX expression were similar in WT and IFN-γ^−/−CD4⁺ and CD8⁺ T cells throughout the course of infection (Fig. 7D, 7E). Surprisingly, however, expression of BIM was increased in IFN-γ^−/− derived CD8⁺ T cells (but not CD4⁺ T cells) on days 5 and 7 of infection (Fig. 7E). Taken together, these data suggest that the loss of splenic effector T cells during the course of P. berghei ANKA infection in WT mice is due to IFN-γ–mediated apoptosis: increased Bcl-2 expression may prevent apoptosis in IFN-γ^−/− mice, but apoptosis in WT mice is unlikely to proceed solely through the classical intrinsic pathway.

Our data suggest that, despite marked differences in their susceptibility to T cell-dependent immunopathology, the splenic effector T cell response is broadly similar in P. berghei ANKA-infected WT and IFN-γ^−/− mice with the only substantial difference being the increased resistance to apoptosis of T cells from IFN-γ^−/− mice. To determine whether T cells maturing in the absence of IFN-γ are intrinsically resistant to apoptosis or whether their resistance is simply due to the lack of circulating IFN-γ, we purified CD4⁺ and CD8⁺ T cells from WT and IFN-γ^−/− mice on day 5 of P. berghei ANKA infection, transferred them into congenic (CD45.1⁺), infected (day 5; i.e., at the peak of T cell expansion and prior to onset of apoptosis in WT mice), or uninfected WT mice, and determined their recovery 2 d later (i.e., day 7 of infection; peak of apoptosis in WT mice) (Fig. 8). Representative dot plots showing the recovery of host (CD45.2⁻) and donor (CD45.2⁺) CD4⁺ T cells in the spleen are shown in Fig. 8A. As expected, significantly fewer donor CD4⁺ T and CD8⁺ T cells were recovered from spleens of infected recipient mice than from spleens of uninfected recipient mice (Fig. 8B, 8C). This was associated with increased recovery of donor cells from the lungs in infected mice (Fig. 8D, 8E), indicating preferential migration of donor T cells to peripheral tissues rather than the spleen during P. berghei ANKA infection. However, the increased accumulation of donor cells in nonlymphoid tissues in infected recipients was insufficient to account for the reduced recovery of donor cells in the spleen, strongly suggesting that apoptosis rather than cellular migration controlled donor cell recovery from the spleen. Importantly, CD4⁺ T cells from IFN-γ^−/− mice disappeared at the same rate as CD4⁺ T cells from WT mice once transferred into an IFN-γ–replete host. In contrast, CD8⁺ T cells from IFN-γ^−/− donors survived better in both infected and uninfected recipients than CD8⁺ T cells from WT donors did (Fig. 8C). Taken together, these data further indicate that effector CD4⁺ T cells from WT and IFN-γ^−/− mice are not intrinsically different and behave equivalently when placed in comparable (i.e., IFN-γ replete) environments. Although IFN-γ^−/− splenic CD8⁺ T cells seem slightly more resilient than WT CD8⁺ T cells, they still succumb to IFN-γ–dependent signals during infection.

Our data strongly suggest that differences in maintenance of splenic T cell populations in P. berghei ANKA-infected WT and IFN-γ^−/− mice are due to differences in an active process of cell survival and/or apoptosis beginning on or after day 5 of infection and mediated by extrinsic (environmental) signals. To test this hypothesis, we purified splenic CD4⁺ and CD8⁺ T cells from WT C57BL/6 CD45.1⁺ congenic mice on day 5 of P. berghei ANKA infection and adoptively transferred them into infected (day 5 of P. berghei ANKA infection) and uninfected, WT, and IFN-γ^−/−,CD45.2⁺ mice. The survival of CD45.1⁺ donor cells in the spleen, lung, and brain of infected and uninfected recipient mice was assessed 2 d after transfer at the stage when WT mice developed ECM (Fig. 9). Representative dot plots showing the recovery of donor (CD45.1⁺) and host (CD45.1⁻) CD4⁺ T cells from the spleen are shown in Fig. 9A. In clear support of the hypothesis, more donor CD4⁺ and CD8⁺ T cells were recovered from the spleens of infected IFN-γ^−/− mice than from those of WT mice (Fig. 9B, 9C); indeed, donor cell numbers in spleens of infected IFN-γ^−/− mice were comparable to those in uninfected mice. Because identical cells were transferred into infected WT and IFN-γ^−/− hosts, it is evident that the loss of splenic T cells in WT mice is an active process mediated by extrinsic IFN-γ after day 5 of infection. Importantly, the lower recovery of donor T cells from spleens of WT-infected mice was not due to increased T cell migration to peripheral tissues because no donor cells were recovered from the brain (data not shown) and equivalent numbers of donor CD4⁺ and CD8⁺ T cells were found in the lungs of infected WT and IFN-γ^−/− recipients (Fig. 9D, 9E). Thus, active, nonautocrine IFN-γ–dependent signals control splenic CD4⁺ and CD8⁺ T cell survival during P. berghei ANKA infection.

We have performed a detailed comparison of the effector T cell response in WT and IFN-γ^−/− mice infected with P. berghei ANKA to examine the extent to which cell-intrinsic differences in T cell function determine the resistance of IFN-γ^−/− mice to ECM. We have found that CD8⁺ T cell activation, which is an essential precursor of ECM development (reviewed in Ref. 14), occurs entirely independently of IFN-γ signaling. However, lack of IFN-γ significantly reduced cerebral migration, or accumulation, of CD8⁺ T cells; instead, in the absence of IFN-γ, CD8⁺ T cells accumulated in other organs (such as the lung). Impaired accumulation of CD8⁺ T cells within the brain of IFN-γ^−/− mice was not due to attenuated in situ local proliferation. Although it has previously been suggested that IFN-γ upregulates CCR5- and CXCR3-dependent CD8⁺ T cell migration pathways during malaria infection (4, 7), we have found that CXCR3 and CCR5 expression by CD8⁺ T cells is unaffected (spleen or lung) or only marginally affected (brain) by lack of IFN-γ. Thus, our results suggest that CXCR3- and CCR5-dependent CD8⁺ T cell migration to the lung and brain during P. berghei ANKA infection is unlikely to be controlled primarily at the level of chemokine receptor expression and is more likely dependent on the level of chemokine production within a particular tissue. If so, then IFN-γ may be an important inducer of chemokine expression in the brain, a supposition that is well supported by our data, which is consistent with published results (7), showing that production of IP-10 and MIG in the brain is abrogated in IFN-γ^−/− mice. In contrast, as VLA-4 and LFA-1 expression by CD8⁺ T cells was unimpaired in IFN-γ^−/− mice, and the expression of ICAM-1 and VCAM-1 within the brain was comparable in WT and IFN-γ^−/− mice during infection, our results strongly suggest that dysregulation of these adhesion pathways does not contribute to impaired accumulation of CD8⁺ T cells within the brain of IFN-γ^−/− mice (on C57BL/6 background) during P. berghei ANKA infection. The reason for the differences in our results compared with Van den Steen et al. (7), who observed lower expression of ICAM-1 and VCAM-1 in the brains of IFN-γ^−/− mice (DBA/2 and BALB/c backgrounds), is potentially due to the genetic background of the IFN-γ^−/− mice used in the two studies. Indeed, in the study by Van den Steen et al. (7), the fold change in expression of ICAM-1 and VCAM-1 within the brain of WT and IFN-γ^−/− mice varied between BALB/c and DBA-2 mice.

Interestingly, IFN-γ seems to play a somewhat different role in regulating CD4⁺ T cell responses to P. berghei ANKA infection. Despite lower expression of CCR5 and CXCR3, CD4⁺ T cell accumulation in the brain and lung was unimpaired in P. berghei ANKA-infected IFN-γ^−/− mice, consistent with findings in IFN-γR^−/− mice (4). This is even though fewer brain-derived CD4⁺ T cells expressed the proliferative marker Ki67 in IFN-γ^−/− mice, suggesting that IFN-γ may influence (effector or effector-memory) CD4⁺ T cell proliferation within the brain during malaria infection. Thus, it is possible that CD4⁺ T cells may exit the brain and recirculate, or the cells may die in situ in IFN-γ–competent mice. In contrast, the proportion of brain-localized CD4⁺ T cells with a highly activated phenotype was significantly reduced in IFN-γ^−/− mice compared with WT mice. Therefore, in contrast to what we saw for CD8⁺ T cells, CD4⁺ T cell migration to and/or accumulation in peripheral tissues during P. berghei ANKA is controlled by IFN-γ–independent processes, but IFN-γ is essential for CD4⁺ T cells to become fully activated. This raises the possibility that CD4⁺IFN-γ⁺ T cells may modify the cerebral environment, promoting CD8⁺ T cell migration and accumulation, which in an interdependent relationship may augment parasite biomass (15–17). In support of this, we have shown that adoptive transfer of infection-derived, IFN-γ−competent CD4⁺ T cells into IFN-γ^−/− mice increases CD8⁺ T cell accumulation within the brain and induces ECM (A. Villegas Mendez and K. Couper, manuscript in preparation) in an IFN-γ–dependent manner. IFN-γ may modulate the function of tissue-resident cells, such as astrocytes or endothelial cells (22, 41), or modulate the accumulation of other infiltrating cells such as macrophages [results not shown but provided for review and reference (3)].

In addition to its effects on T cell activation and migration, IFN-γ also regulates T cell survival in the spleen during P. berghei ANKA infection. The contraction phase of the effector T cell response was markedly attenuated in IFN-γ^−/− mice, such that by day 7 of infection, spleens of IFN-γ^−/− mice contained 2- to 3-fold more T cells than the spleens of WT mice did. Nonetheless, as peripheral parasitemia was similar in WT and IFN-γ^−/− mice, this stronger T cell response does not seem to translate into better parasite control, and thus, the resistance to ECM of IFN-γ^−/− mice cannot be simply be because of lower parasite burdens [as seems to be the case in IP-10^−/− mice (42)]. Thus, we postulate that IFN-γ^−/− mice are resistant to ECM because of lessened inflammation within the brain and resultant reduced local tissue–parasite accumulation and vascular hemorrhage. In support of this, brain parasite biomass is known to be lower in IFN-γ^−/− mice (16, 17). Our data suggest that IFN-γ promotes contraction of the effector T cell pool by inducing apoptosis, confirming earlier reports that IFN-γ induces apoptosis of malaria-specific CD4⁺ T cells (31). Nevertheless, we have extended the analysis of the role played by IFN-γ to demonstrate that IFN-γ^−/− T cells are fully susceptible to apoptosis when transferred into an IFN-γ–sufficient environment and that WT CD4⁺ T cells are resistant to apoptosis when transferred to an IFN-γ–deficient environment. In other words, extrinsic IFN-γ signaling mediates apoptosis of T cells in infected mice, irrespective of whether the T cell itself can secrete IFN-γ or not. This suggests that apoptosis results from paracrine signaling (e.g., from one T cell to another) rather than autocrine signaling. Moreover, our data suggest that IFN-γ–mediated T cell loss is unlikely to occur solely via the classical intrinsic pathway of apoptosis because T cell expression of the proapoptotic molecules BIM, BAD, and BAX did not differ significantly between WT and IFN-γ^−/− mice. One possibility is that IFN-γ upregulates TNF production and, consequently, that apoptosis is indirectly induced in WT mice by increased death receptor signaling (Trail, TNFR1, and DR4/5). Importantly, as the expression of CTLA-4 and PD-1 on splenic CD4⁺ and CD8⁺ T cells was unaltered in IFN-γ^−/− mice during infection, it is unlikely that these pathways play a large role in regulating the survival and contraction phases of the splenic T cell response during P. berghei ANKA infection. Moreover, attenuated expression of CTLA-4 and PD-1 by tissue-accumulating CD4⁺ T cells did not enhance their accumulation within nonlymphoid tissues in IFN-γ^−/− mice, also suggesting that cell-intrinsic expression of these costimulatory molecules does not significantly regulate CD4⁺ T cell migration into nonlymphoid tissues during P. berghei ANKA infection. Nevertheless, we cannot exclude the possibility that reduced CTLA-4 expression contributed to enhanced CD8⁺ T cell survival within the lungs of IFN-γ^−/− mice. The role of CTLA-4 in controlling CD8⁺ T cell survival and accumulation of CD8⁺ T cells within the brain is unclear, because although CTLA-4 expression by brain-infiltrating CD8⁺ T cells was lower in IFN-γ^−/− mice, accumulation of cells was also attenuated.

In summary, we have demonstrated heterogeneous and tissue-specific control of CD4⁺ and CD8⁺ T cell responses by IFN-γ during P. berghei ANKA infection. Overall, splenic T cell responses appeared stronger in IFN-γ^−/− mice, primarily because of reduced T cell apoptosis. Importantly, during P. berghei ANKA infection, IFN-γ appears to have very different effects on activation and migration of CD4⁺ and CD8⁺ T cells in tissues such as the lung and brain, and this is not directly related to the cell-intrinsic expression level of IFN-γR. Thus, altered accumulation of CD8⁺ T cells within the lung, liver, and brain in WT and IFN-γ^−/− mice during P. berghei ANKA infection is likely because of multiple factors, including reduced apoptosis within the spleen and tissues in IFN-γ^−/− mice combined with altered tissue-specific signals regulating cellular adhesion and chemotaxis. These observations increase our understanding of the role of IFN-γ in the pathogenesis of ECM and may have relevance for understanding the role of IFN-γ in the pathogenesis of human cerebral malaria.

The authors have no financial conflicts of interest.