Gamma Interferon Mediates Experimental Cerebral Malaria by Signaling within Both the Hematopoietic and Nonhematopoietic Compartments

ABSTRACT Experimental cerebral malaria (ECM) is a gamma interferon (IFN-γ)-dependent syndrome. However, whether IFN-γ promotes ECM through direct and synergistic targeting of multiple cell populations or by acting primarily on a specific responsive cell type is currently unknown. Here, using a panel of cell- and compartment-specific IFN-γ receptor 2 (IFN-γR2)-deficient mice, we show that IFN-γ causes ECM by signaling within both the hematopoietic and nonhematopoietic compartments. Mechanistically, hematopoietic and nonhematopoietic compartment-specific IFN-γR signaling exerts additive effects in orchestrating intracerebral inflammation, leading to the development of ECM. Surprisingly, mice with specific deletion of IFN-γR2 expression on myeloid cells, T cells, or neurons were completely susceptible to terminal ECM. Utilizing a reductionist in vitro system, we show that synergistic IFN-γ and tumor necrosis factor (TNF) stimulation promotes strong activation of brain blood vessel endothelial cells. Combined, our data show that within the hematopoietic compartment, IFN-γ causes ECM by acting redundantly or by targeting non-T cell or non-myeloid cell populations. Within the nonhematopoietic compartment, brain endothelial cells, but not neurons, may be the major target of IFN-γ leading to ECM development. Collectively, our data provide information on how IFN-γ mediates the development of cerebral pathology during malaria infection.

The functional IFN-␥R, a heterodimeric complex composed of the IFN-␥ receptor 1 (IFN-␥R1) and IFN-␥R2 chains, can be expressed on many different cell types (15,16). IFN-␥R1, the major ligand binding subunit, is ubiquitously expressed, whereas the expression of the nonbinding, signal-transducing IFN-␥R2 subunit is generally low and is tightly controlled (15,16). Consequently, IFN-␥R2 expression, rather than IFN-␥R1 expression, controls the responsiveness of cells to IFN-␥ (15,17). In nonmalaria models, it has been shown that IFN-␥ enhances antigen processing, major histocompatibility complex (MHC) and costimulatory marker expression, and cytokine production in dendritic cells and macrophages/monocytes (15,16). IFN-␥ can also act on T cells, orchestrating CD4 ϩ T cell and CD8 ϩ T cell activation and differentiation, as well as apoptosis (18)(19)(20)(21)(22)(23)(24). Moreover, IFN-␥ can directly modify the function and status of brain-resident and -specialized cell populations, including neurons, brain endothelial cells, microglial cells, and astrocytes, in a variety of inflammatory settings, including malaria (25)(26)(27)(28)(29)(30). Combined, these observations indicate that IFN-␥ may mediate ECM development by targeting a specific cell type, in a particular location, at a precise time of infection. Alternatively, it may cause cerebral pathology during malaria infection by functioning within a complex cellular network, acting synergistically on different cell types.
In this study, we have investigated the cell population(s) and compartments that IFN-␥R signals within to mediate ECM development during P. berghei ANKA infection. Utilizing novel cell-and compartment-specific IFN-␥R2-deficient mice (31), we demonstrate that IFN-␥ causes ECM by signaling within both the hematopoietic and nonhematopoietic compartments. Within the brain, IFN-␥R signaling within the two compartments was additive, which led to severe neuroinflammation. Within the hematopoietic and nonhematopoietic compartments, IFN-␥R2 expression by myeloid cells, T cells, and neurons was, individually, not required for the development of ECM. Importantly, we show that brain endothelial cells were highly responsive to IFN-␥ in combination with tumor necrosis factor (TNF). Thus, within the nonhematopoietic compartment, IFN-␥ may mediate ECM by directly targeting brain endothelial cells. The results in this study improve our knowledge of the IFN-␥R-expressing cell populations that may contribute to the IFN-␥-dependent development of cerebral pathology during malaria infection.
IFN-␥R signaling within the hematopoietic compartment does not strongly influence splenic immunity during infection. We assessed whether abrogated IFN-␥R2 signaling within the hematopoietic compartment modulated the generation and/or maintenance of innate and adaptive splenic immune responses during infection, contributing to the intermediate resistance of VAV-Cre ϩ IFN-␥R2 flox/flox mice to ECM. Within the innate compartment, the numbers of inflammatory monocytes, neutrophils, and classical dendritic cells (CD11b ϩ and CD8 ϩ ) were unaltered in the spleens of IFN-␥R Ϫ/Ϫ and VAV-Cre ϩ IFN-␥R2 flox/flox mice compared with the numbers in WT mice on day 7 of infection ( Fig. 2A). In contrast, the numbers of F4/80 ϩ macrophages were significantly increased in the spleens of IFN-␥R Ϫ/Ϫ mice compared with the numbers in VAV-Cre ϩ IFN-␥R2 flox/flox mice and WT mice on day 7 of infection ( Fig. 2A). Thus, nonhematopoietic and/or any global IFN-␥R2 signaling appears to specifically limit splenic macrophage numbers during P. berghei ANKA infection.
There was a trend toward higher numbers of splenic CD4 ϩ T cells in IFN-␥R Ϫ/Ϫ mice but not in VAV-Cre ϩ IFN-␥R2 flox/flox mice compared with the numbers in WT mice on day 7 of infection ( Fig. 2B). Similarly, splenic CD8 ϩ T cell numbers were significantly increased in IFN-␥R Ϫ/Ϫ mice but not VAV-Cre ϩ IFN-␥R2 flox/flox mice compared with the numbers in WT mice on day 7 of infection ( Fig. 2B). Splenic CD4 ϩ T cell and CD8 ϩ T cell activation was, however, largely unaltered, as determined by comparing the levels of granzyme B, KLRG-1, and Ki-67 expression in IFN-␥R Ϫ/Ϫ mice to the levels in VAV-Cre ϩ IFN-␥R2 flox/flox mice and WT mice on day 7 of infection ( Fig. 2C). Of the markers examined, only ICOS was differentially expressed by splenic CD4 ϩ T cells, but not CD8 ϩ T cells, in IFN-␥R Ϫ/Ϫ mice compared with VAV-Cre ϩ IFN-␥R2 flox/flox mice and WT mice (Fig. 2C). Combined, these data suggest that hematopoietic IFN-␥R signaling does not substantially control the splenic immune response during P. berghei ANKA infection; however, nonhematopoietic and/or synergistic global IFN-␥R signaling may affect the numbers but not the general activation of specific splenic innate and adaptive immune cell populations, most probably through promoting cellular apoptosis rather than controlling cellular migration to diverse nonlymphoid tissues (11).
IFN-␥R signaling within the hematopoietic and nonhematopoietic compartments synergistically defines the level of cerebral inflammation during infection. Next, we utilized NanoString gene expression analysis to examine how hematopoietic and nonhematopoietic compartment-specific IFN-␥R signaling coordinately shape the inflammatory landscape within the brain during P. berghei ANKA infection, promoting ECM development. As expected, the levels of expression of a number of genes associated with interferon signaling, cell adhesion/migration, cytotoxic T lymphocyte (CTL) activity, and innate/endothelial cell activity were significantly elevated within the brains of WT mice experiencing ECM compared with their expression in the brains of ECM-resistant IFN-␥R Ϫ/Ϫ mice ( Fig. 3A to E). Notably, the expression of key genes, including those encoding CXCL9, CXCL10, granzyme B, and TAP-1, all of which have been shown to be involved in ECM pathogenesis (32), was dampened rather than completely abrogated in the brains of infected VAV-Cre ϩ IFN-␥R2 flox/flox mice compared with their expression in brains from WT mice ( Fig. 3B and D). Indeed, for all genes identified as differentially expressed in the brains of infected WT mice compared with their expression in infected IFN-␥R Ϫ/Ϫ mice, including those encoding the IFN signaling molecules STAT1, IRF1, IRF7, and IFIT1, intermediate gene expression was observed in the brains of VAV-Cre ϩ IFN-␥R2 flox/flox mice ( Fig. 3A to E). Interestingly, although the numbers of intracerebral CD8 ϩ T cells were significantly decreased in infected globally IFN-␥R2 Ϫ/Ϫ mice compared with the numbers in WT mice on day 7 of infection, potentially contributing to the reduced inflammatory landscape in the brains of globally IFN-␥R2 Ϫ/Ϫ mice, the numbers of intracerebral CD8 ϩ T cells were comparable in infected VAV-Cre ϩ IFN-␥R2 flox/flox mice and WT mice (Fig. 3F). Consequently, these results suggest that ablation of IFN-␥R signaling within the hematopoietic compartment in VAV-Cre ϩ IFN-␥R2 flox/flox mice led to a generalized reduction of inflammation within the brain during infection, rather than the suppression of individual or specific modules of genes. However, the ameliorated cerebral inflammation in VAV-Cre ϩ IFN-␥R2 flox/flox mice was not simply due to reduced CD8 ϩ T cell accumulation in the brain.
IFN-␥R2 signaling in peripheral myeloid cells and T cells is dispensable for ECM development. To examine which cell populations respond to IFN-␥ within the hema- Results in all panels are the mean values Ϯ SEM of the groups, with 4 to 8 brains per group. For the data shown in panels A to E, brains were obtained from 2 independent experiments. Results in panel F are representative of two independent experiments. *, P Ͻ 0.05 for WT versus IFN-␥R2 Ϫ/Ϫ mice; #, P Ͻ 0.05 for WT versus VAV-Cre mice. Statistical significance was tested using one-way ANOVA with Tukey's post hoc test.

IFN-␥ synergizes with TNF to activate brain endothelial cells in vitro.
As neuronal (and astrocytic) IFN-␥R2 signaling was not required for ECM development, we addressed which other nonhematopoietic cell populations express IFN-␥R and may respond to IFN-␥ to promote cerebral pathology during P. berghei ANKA infection. Accumulating evidence suggests that brain endothelial cells forming the blood-brain barrier play a pivotal role in ECM pathogenesis through the production of chemokines, cross-presentation of malarial antigens, and regulation of the permeation of materials into the brain tissue (5,9,13,32,36). Brain Sca-1 ϩ CD31 ϩ endothelial cells (gated as shown in Fig. 6A, using the strategy defined in reference 37) expressed the IFN-␥R subunits in naive mice and during ECM ( Fig. 6B and C). Therefore, we hypothesized that endothelial cells were the main nonhematopoietic cell population that responds directly to IFN-␥, leading to the development of ECM. To examine this, we stimulated brain endothelial cells (bEnd5 cells), which also express IFN-␥R (results not shown), with IFN-␥, TNF, or both cytokines in combination. IFN-␥ promoted significantly higher upregulation of MHC class I expression by bEnd5 cells than TNF (Fig. 6D and E). Indeed, costimulation with IFN-␥ and TNF failed to increase MHC class I expression above that induced by IFN-␥ by itself (Fig. 6D and E). Of relevance, and as previously reported (32), MHC class I expression by brain endothelial cells was significantly upregulated in WT mice during ECM compared with its expression in naive WT mice (Fig. 6F and G). In contrast, IFN-␥ was unable to promote VCAM-1 or ICAM-1 expression or interleukin-6 (IL-6) production by brain endothelial cells by itself in vitro and instead only synergized with and amplified the expression and production induced by TNF (Fig. 6H to J). As MHC class I gene expression was significantly lower in the brains of globally IFN-␥R2 Ϫ/Ϫ mice and, to a lesser extent, VAV-Cre ϩ IFN-␥R2 flox/flox mice compared with its expression in WT mice on day 7 of P. berghei ANKA infection (Fig. 3D), these data in combination indicate that IFN-␥ may be the dominant (direct and indirect) driver of MHC class I expression by brain endothelial cells during P. berghei ANKA infection. IFN-␥ may also play a pathogenic role by synergizing with other inflammatory cytokines to amplify endothelial cell activation, leading to the development of cerebral pathology. showing MHC class I expression by brain endothelial cells from P. berghei ANKA-infected and naive mice. (H to J) MFI of VCAM-1 (H) and ICAM-1 (I) on and production of IL-6 (J) by stimulated and unstimulated bEnd5 cells, measured by ELISA. Results in panels C and G are the mean values Ϯ SEM of the groups, with 5 mice per group. Results in panels E and H to J are the mean values Ϯ SEM from three independent biological replicates and are representative of 2 independent experiments. *, P Ͻ 0.05 between defined groups. Statistical significance was tested using the Mann-Whitney test for the data in panels C and G and one-way ANOVA with Tukey's post hoc test for the data in panels E and H to J.

DISCUSSION
In this study, we have utilized novel cell-and compartment-specific IFN-␥R2 Ϫ/Ϫ mice to examine which cell type(s) IFN-␥ targets to cause ECM during P. berghei ANKA infection. We found that VAV-Cre ϩ IFN-␥R2 flox/flox mice displayed intermediate resistance to ECM, with 100% of mice developing prodromal disease and 40% of mice developing late-stage ECM. Thus, we have demonstrated that IFN-␥ signals throughout both the hematopoietic and nonhematopoietic compartments to cause ECM.
Notably, abrogation of hematopoietic compartment-specific IFN-␥R2 signaling did not significantly modulate the splenic innate or adaptive immune responses during infection. In contrast, hematopoietic and nonhematopoietic IFN-␥R2 signaling appeared to function additively to promote malaria-induced cerebral inflammation. As previously reported, IFN-␥R2 signaling was critically required for the expression within the brain of a number of chemokine genes (7,12), as well as genes controlling antigen processing and MHC class I peptide presentation. Unexpectedly, the reduced inflammatory landscape in the brains of infected VAV-Cre ϩ IFN-␥R2 flox/flox mice was not simply due to reduced CD8 ϩ T cell accumulation. Thus, our results are consistent with the model where either redundant global or nonhematopoietic compartment-specific IFN-␥R2 signaling recruits CD8 ϩ T cells to the brain during P. berghei ANKA infection, most probably through control of chemokine expression in the brain, as the expression of CXCL9 and CXCL10 was reduced but not abrogated in the brains of VAV-Cre ϩ IFN-␥R2 flox/flox mice. Nonhematopoietic and hematopoietic IFN-␥R2 signaling subsequently combine to establish an environment allowing CD8 ϩ T cells to optimally mediate their pathogenic activity. Indeed, we have previously shown that the behavior of intracerebral CD8 ϩ T cells is as important as their presence per se in promoting ECM development (38). Of note, a higher level of brain-to-brain variability in gene expression was observed in the infected VAV-Cre ϩ IFN-␥R2 flox/flox group than in the infected WT or the IFN-␥R2 Ϫ/Ϫ group. This likely reflects the heterogeneity in the susceptibilities of VAV-Cre ϩ IFN-␥R2 flox/flox mice to ECM and suggests that ECM develops following breaching of threshold levels of expression of inflammation-related genes. Although we cannot fully exclude the possibility that alterations in immune cell development contributed to the partial resistance of VAV-Cre ϩ IFN-␥R2 flox/flox mice to ECM, the T cell numbers and activation were normal in the spleens and brains of uninfected IFN-␥R Ϫ/Ϫ mice and VAV-Cre ϩ IFN-␥R2 flox/flox mice (results not shown). Thus, we believe it is unlikely that T cell development was significantly affected by global or hematopoietic compartment-specific abrogation of IFN-␥R2 expression.
We dissected the targets of IFN-␥ within the hematopoietic lineage and determined that IFN-␥R signaling within T cells (CD4-Cre) and monocytes/macrophages and neutrophils (LysM-Cre) is (individually) dispensable for the development of ECM. That IFN-␥R expression on T cells is not required for the development of ECM was perhaps unsurprising given that we have previously shown that CD8 ϩ T cell activation is largely unimpaired systemically in IFN-␥ knockout (KO) mice during P. berghei ANKA infection (11). Moreover, CD4 ϩ T cell activation is altered, to a relatively small degree, only in the brains of IFN-␥ KO mice during infection (11). Thus, T cell-intrinsic IFN-␥R signaling does not appear to contribute to the development of pathogenic T cell responses during P. berghei ANKA infection. Rather, IFN-␥ indirectly promotes the migration to and accumulation of CD8 ϩ T cells within the brain during P. berghei ANKA infection by modulating the expression of chemokines within the tissue (7,11,12).
The observation that LysM-Cre ϩ IFN-␥R2 flox/flox mice were as susceptible to ECM as WT mice indicates that IFN-␥R signaling within blood-derived monocytes, macrophages, and neutrophils is not required for the development of ECM. These results are thus consistent with the more general observations in the literature that peripheral monocytes and macrophages are not critically involved in the pathogenesis of ECM (9,39,40). It is important to note, however, that while LysM-Cre has a high gene recombination rate in peripheral monocytes, macrophages, and neutrophils, it only promotes gene recombination in 50% of microglial cells (41). Moreover, the efficiency of VAV-Cre-mediated genetic recombination in microglial cells is debated (42). Consequently, we cannot definitively extrapolate the role of IFN-␥R signaling in microglial cells in the development of neuropathology during malaria infection using the Credriving lines employed in this study. Nevertheless, utilizing CX3CR1-iDTR mice, in which diphtheria toxin (DT) treatment promotes extensive depletion of microglial cells, perivascular macrophages, and other myeloid cell populations (43), we have recently demonstrated that perivascular macrophages and microglial cells are not involved in the terminal stages of ECM development (38). Consequently, we do not believe that IFN-␥R signaling by microglial cells is critically required for ECM development. It remains to be resolved whether IFN-␥ acts in a redundant and/or additive manner throughout the hematopoietic compartment to promote ECM or whether it mediates its effects through an as-yet-unidentified hematopoietic cell population.
In terms of the nonhematopoietic cell populations that respond to IFN-␥ during P. berghei ANKA infection, our results indicate that neuronal IFN-␥R signaling is not crucial for the development of late-stage neuropathology. This suggests that the IFN-␥dependent effects on neuronal cells observed during ECM (33) must be mediated in trans via activity on other cell types. As Nestin-Cre also exerts strong effects in the astrocyte lineage cells (35), our results also suggest that astrocytic IFN-␥R signaling is dispensable for the development of ECM.
Notably, it has recently been shown that IFN-␥ promotes cross-presentation of malarial antigens by brain endothelial cells and that this is an important event in ECM pathogenesis (13). Consistent with this, we found that IFN-␥R was expressed by brain endothelial cells in naive mice and during P. berghei ANKA infection. Moreover, as has previously been reported, we found that IFN-␥ stimulation was able to promote strong upregulation of MHC class I by brain endothelial cells in vitro (13,44). In addition, the gene expression of MHC class I was lower in the brains of globally IFN-␥R2 Ϫ/Ϫ mice and, to a lesser extent, VAV-Cre ϩ IFN-␥R2 flox/flox mice than in the brains WT mice on day 7 of P. berghei ANKA infection. As the VAV-Cre line utilized in this study does not promote gene recombination in vascular endothelial cells (45), our results therefore suggest that IFN-␥ may nonredundantly directly and indirectly target brain endothelial cells to control MHC class I expression during P. berghei ANKA infection, leading to the development of ECM. Although IFN-␥, in combination with other inflammatory cytokines, also promoted VCAM-1 and ICAM-1 expression and IL-6 production, the gene expression levels of VCAM-1 and ICAM-1 were not significantly lower in the brains of global IFN-␥R2 Ϫ/Ϫ mice and VAV-Cre ϩ IFN-␥R2 flox/flox mice than in the brains of WT mice on day 7 of P. berghei ANKA infection. As it has recently been reported that the protein levels of ICAM-1 and VCAM-1 are reduced by brain endothelial cells in IFN-␥ Ϫ/Ϫ mice during P. berghei ANKA infection (46), this may suggest that there is a dichotomy in the gene and protein expression levels of ICAM-1 and VCAM-1 by brain endothelial cells during P. berghei ANKA infection or that gene expression differences specifically by brain endothelial cells in the IFN-␥R2-deficient mice were diluted within our wholebrain NanoString analysis. Thus, although we acknowledge the reductionist nature of our in vitro experiments and the fact that they do not fully recapitulate the inflammatory intracerebral environment during ECM, we believe our data, in agreement with previous studies (5,44), strongly support the hypothesis that brain endothelial cells may be the major nonhematopoietic cell IFN-␥ target leading to ECM. In future work, the relative importance of IFN-␥ targeting of brain endothelial cells in promoting ECM can be definitively assessed by utilizing Slco1c1-Cre mice (47) crossed with IFN-␥R2 flox/flox mice.
In summary, the results in this study are consistent with an overall model where IFN-␥ acts on different cell populations within the hematopoietic and nonhematopoietic cell compartments to orchestrate brain inflammation that enables CD8 ϩ T cells to manifest their pathogenic activity. It remains to be defined whether IFN-␥ acts on the diverse responding cell types in a concomitant or temporal manner to promote ECM. However, our results further argue that IFN-␥ predominantly exerts its ECM-inducing pathogenic function locally within the brain, rather than within the spleen. The results Statistical analysis. Data were checked for normality using the Shapiro-Wilk test. For two-group comparisons, statistical significance was determined using the t test (parametric data) or Mann-Whitney test (nonparametric data). For comparisons of three or more groups, statistical significance was determined using one-way or two-way analysis of variance (ANOVA) with Tukey post hoc analysis (parametric data) or the Kruskal-Wallis test with Dunn's post hoc test (nonparametric data). Results were considered significantly different when the P value was Ͻ0.05.