IFN-γ Signaling to Astrocytes Protects from Autoimmune Mediated Neurological Disability

Demyelination and axonal degeneration are determinants of progressive neurological disability in patients with multiple sclerosis (MS). Cells resident within the central nervous system (CNS) are active participants in development, progression and subsequent control of autoimmune disease; however, their individual contributions are not well understood. Astrocytes, the most abundant CNS cell type, are highly sensitive to environmental cues and are implicated in both detrimental and protective outcomes during autoimmune demyelination. Experimental autoimmune encephalomyelitis (EAE) was induced in transgenic mice expressing signaling defective dominant-negative interferon gamma (IFN-γ) receptors on astrocytes to determine the influence of inflammation on astrocyte activity. Inhibition of IFN-γ signaling to astrocytes did not influence disease incidence, onset, initial progression of symptoms, blood brain barrier (BBB) integrity or the composition of the acute CNS inflammatory response. Nevertheless, increased demyelination at peak acute disease in the absence of IFN-γ signaling to astrocytes correlated with sustained clinical symptoms. Following peak disease, diminished clinical remission, increased mortality and sustained astrocyte activation within the gray matter demonstrate a critical role of IFN-γ signaling to astrocytes in neuroprotection. Diminished disease remission was associated with escalating demyelination, axonal degeneration and sustained inflammation. The CNS infiltrating leukocyte composition was not altered; however, decreased IL-10 and IL-27 correlated with sustained disease. These data indicate that astrocytes play a critical role in limiting CNS autoimmune disease dependent upon a neuroprotective signaling pathway mediated by engagement of IFN-γ receptors.


Introduction
CNS resident cells are targets of autoimmune mediated damage but also active participants in disease development, progression and control [1,2]. However, their contributions to neuroprotection and regulation by inflammatory mediators are not well defined. CNS insults, including autoimmune disease, initiate rapid astrocyte activation characterized by cellular hypertrophy and increased intermediate filament glial fibrillary acidic protein (GFAP) expression [3][4][5]. Astrocytes form a physical barrier surrounding areas of inflammation initially limiting bystander tissue damage [6,7]. However, this barrier subsequently impedes axonal regeneration contributing to sustained disability [3,5,8]. Innate and adaptive pro-inflammatory astrocyte functions include production of pro-inflammatory cytokines, reactive oxygen species, chemokines, and matrix metalloproteinases [3][4][5]. By contrast, secretion of anti-inflammatory cytokines and scavengers of reactive oxygen species, as well as inhibition of both microglial activation and tumor necrosis factor (TNF) secretion, all support an astrocyte mediated anti-inflammatory function [3][4][5]. Therefore, astrocyte activation constitutes a ubiquitous, yet heteroge-neous response associated with both promoting and inhibiting CNS repair [3][4][5].
MS and EAE are both associated with T cells secreting IFN-c (Th1 cells) and IL-17 (Th17 cells) which play complex, not fully understood roles in disease initiation and progression [2,9]. In vivo and in vitro evidence indicates that suppression of encephalitogenic T cell proliferation within the CNS during EAE and activation of anti-inflammatory programs are in part mediated via astrocytes [4,5]. Similar to astrocytes, IFN-c mediates both proand anti-inflammatory functions during autoimmune disease [10]. Early IFN-c induced effects are pro-inflammatory; IFN-c facilitates inflammatory cell access, shapes their composition, increases major histocompatibility complex (MHC) expression, contributes to macrophage and microglia activation, and initiates oligodendrocyte death [11]. Similarly, IFN-c mediated protection during EAE is also multifaceted [12][13][14][15]. It acts as a negative regulator of neutrophil accumulation, Th17 cell activation, IL-1R signaling, protease secretion, and chemokine activity. It also inhibits proinflammatory cytokine secretion via induction of suppressor of cytokine secretion (SOCS) proteins, facilitates T cell apoptosis and protects oligodendrocytes via inducing endoplasmic reticulum (ER) stress responses [10,11,16,17]. This highlights the critical role of a single mediator in both promoting disease but also limiting inflammatory mediated damage required to initiate repair cascades. Based on the gatekeeper functions of astrocytes and the diverse biological effects of IFN-c, we set out to determine how IFN-c signaling specifically to astrocytes influences CNS autoimmune disease. The results demonstrate that IFN-c signaling to astrocytes had no profound effects on initial disease progression, but played an essential protective role during the transition from acute to chronic disease. Clinical remission induced by IFN-c signaling to astrocytes coincided with reduced demyelination, axonal degeneration, and astrocyte activation. The IFN-c receptor is expressed ubiquitously; however, these data reveal astrocytes as the primary target and mediator of the well established antiinflammatory activity of IFN-c within the CNS.

IFN-c Signaling to Astrocytes Limits Demyelination during Acute EAE
To understand the role of IFN-c signaling to astrocytes during the pathogenesis of CNS autoimmune disease, EAE was induced in transgenic mice expressing a signaling deficient dominant negative IFN-c receptor 1 specifically on astrocytes (GFAPcR1D mice) [18]. Peripheral activation of self reactive T cells in GFAPcR1D mice was similar to wt mice (data not shown). This is consistent with both the CNS restricted transgene expression in the GFAPcR1D mice as well as the similar T cell activation following peripheral immunization with a non-self antigen [18]. Following immunization the incidence of disease, initiation of clinical symptoms, and initial symptom progression were unaltered by the inability of astrocytes to respond to IFN-c ( Fig. 1A; Table 1). In addition, neither immunized group exhibited clinical symptoms of atypical EAE associated with the absence of IFN-c [19]. However, clinical symptoms in GFAPcR1D mice began to diverge from wt mice at , day 12 post immunization (p.i.) prior to the peak of clinical disease. In both groups clinical disease peaked at , day 14 p.i., but severity was increased from a score of 3.1 in wt mice to 4.0 in the GFAPcR1D group ( Fig. 1A; Table 1).
GFAPcR1D and wt mice were compared at the peak of acute disease to determine if IFN-c signaling altered astrocyte activation or CNS inflammation. Astrocyte hypertrophy and GFAP expression ( Fig. 1B) were similar in both groups, indicating no overt effects of IFN-c on initial astrocyte reactivity. Furthermore, neither the extent of inflammation nor the anatomic distribution of inflammatory cells was altered (Fig. 1B). Flow cytometry confirmed no difference in the overall extent of CD45 hi inflammatory cells recruited into the CNS ( Fig. 2A). Furthermore, percentages of CD4 + T cells ( Fig. 2A), CD8 + T cells and macrophages within the infiltrates were also similar (Fig. 3). In contrast to the association between increased EAE severity and neutrophil accumulation in the global absence of IFN-c [14,15], only a small percentage of neutrophils were identified in the CNS of both groups by flow cytometry (Fig. 3); their low presence was confirmed by the inability to identify cells with the characteristic morphology of neutrophils in the brain by histopathology (Fig. 1B). The absence of increased neutrophils in the CNS of GFAPcR1D mice during acute disease suggested that clinical disease was aggravated by a mechanism distinct from global IFN-c deficiency. In addition to the similar frequency of CD4 + T cells in the brains of the two groups ( Fig. 2A), myelin oligodendrocyte glycoprotein (MOG) reactive CD4 + T cells secreting IFN-c were also similar ( Fig. 2B). By contrast, MOG specific CD4 + T cells secreting IL-17 ( Fig. 2B) and Foxp3 + regulatory CD4 + T cells (Tregs) were decreased in GFAPcR1D mice compared to wt mice (Fig. 2C). Reduced Th17 and Tregs may be attributed to increased IFN-c [10,16]. Alternatively, as IFN-c is protective in EAE [12][13][14][15], increased disease severity may reflect reduced bioavailable IFN-c due to sequestration of IFN-c binding to the dominant negative receptor. However, measurement of IFN-c in cell free supernatants derived from dissociated brains at the peak of acute disease demonstrated protein levels of 2.160.2 ng/brain in wt mice versus 3.660.5 ng/brain in GFAPcR1D mice (n = 3; p,0.05). As overall frequencies of MOG reactive CD4 + T cells secreting IFN-c were similar, increased IFN-c in the brains of GFAPcR1D mice suggested enhanced secretion at the cellular level. Although a contribution of NK or CD8 T cells could not be excluded, these potential sources of IFN-c were unlikely due to their low frequencies (NK ,5% and CD8 + T cells , 5%, see Fig. 3) and their equivalent frequencies in the wt and GFAPcR1D mice. Furthermore, reduced Th17 cell and Treg frequencies were consistent with suppression of these populations due to elevated IFN-c [10,16,20].
Inflammation in spinal cords from the GFAPcR1D mice was also similar to wt controls at the peak of clinical symptoms (Fig. 4). Although numbers and distribution of CD4 + T cells were also similar (4.661.3/mm 2 in GFAPcR1D mice vs. 3.860.1/ mm 2 in wt mice; Fig. S1), spinal cords of GFAPcR1D mice exhibited a ,2-fold increase in demyelination (Fig. 4). Areas of demyelination encompassed 7.361.0% of spinal cord white matter in GFAPcR1D mice versus 3.860.9% in wt mice (p#0.01). Furthermore, the increase in demyelination was associated with a prominent loss of axons in GFAPcR1D mice compared to controls (Fig. 4). Despite elevated demyelination and axonal loss in the absence of IFN-c signaling to astrocytes, spinal cords showed no evidence of differential astrocyte activation by either immunohistochemistry (Fig. 4), or differences in GFAP mRNA expression during the peak of acute disease (Fig. 5). Although CCL5, IL-1 and TNF mRNA expression were increased in the spinal cords of the GFAPcR1D mice, no differences in IFN-c, iNOS, CXCL10 or IL-6 mRNA were consistent with similar inflammation (Fig. 5). These data suggest that the initial disease pathogenesis, reflected by an increased demyelination in spinal cords, but not brain, during ascending paralysis is dampened by IFN-c signaling to astrocytes.

IFN-c Signaling to Astrocytes Reduces Chronic Autoimmune Disease
Subsequent to peak disease severity the clinical symptoms in wt mice began a modest decline by day 16 p.i. (Fig. 1A). By contrast, GFAPcR1D mice not only exhibited increased severity of clinical symptoms following day 14 p.i., but the continued disease escalation was associated with increased mortality ( Fig. 1A; Table 1). Sustained morbidity and significant mortality in GFAPcR1D mice after day 30 p.i. implied a critical role for IFN-c induced astrocyte signaling in neuroprotection and limiting disability. A similar absence of clinical remission was found in a preliminary experiment comparing EAE SJL mice carrying the GFAPcR1D gene (data not shown). These data suggest that astrocyte responses to IFN-c are protective, irrespective of genetic background.
Escalating disease in GFAPcR1D mice coincided with focal areas of intense inflammation in spinal cords, which contrasted with the more diffuse inflammation in wt mice (Fig. 6). Demyelination was also increased with myelin loss encompassing 5.761.8% of spinal cord area in GFAPcR1D mice versus 2.361.4% in wt mice (p#0.05) at day 35 p.i. (Fig. 6). The demyelinated areas further exhibited enhanced axonal damage in GFAPcR1D mice (Fig. 6), supporting a correlation between enhanced tissue damage, sustained morbidity and increased mortality ( Fig. 1A; Table 1). Astrocyte activation associated with areas of myelin loss is a prominent finding in the CNS of both patients with MS and rodents with EAE. Although demyelination was increased in the CNS of GFAPcR1D mice, the extent of astrocyte activation associated with spinal cord white matter lesions was similar in both groups ( Fig. 6; ,60 GFAP + cells/mm 2 ). By contrast, the frequency of activated astrocytes in spinal cord grey matter areas that were not associated with demyelinated lesions, was increased in GFAPcR1D mice with 101.5623.0 GFAP + activated astrocytes/mm 2 versus 25.5615.5 in wt mice (p,0.001) (Fig. 7). A similar increase in astrocyte activation within grey matter distal to white matter lesions was also detected in GFAPcR1D SJL mice during chronic EAE (data not shown).

IFN-c Dependent Astrocyte Regulation of Inflammation
Flow cytometric analysis during chronic disease revealed an ,4fold increase in CD45 hi inflammatory cells confirming increased inflammation in the absence of IFN-c signaling to astrocytes (Fig. 8A). Similar to the acute disease, relative percentages of neutrophils, macrophages, CD4 + and CD8 + T cells were all similar (Fig. 3). Furthermore, no differences in expression of activation markers on CNS derived CD4 + T cells, including CD69, CD122, CD127, Fas, FasL, ICOS and CD152 were evident between the groups (data not shown). The percentages of MOG reactive Th1 cells were also not altered in the CNS of GFAPcR1D relative to wt mice (Fig. 8B) and the decreased percentage of potentially destructive Th17 cells identified during acute disease (Fig 2), was also sustained during chronic disease (Fig. 8B). Equivalent expression of IL-7, IL-23 and TGF-b mRNA (Fig. 8C) suggested that the decrease in Th17 cells was dependent upon an increase in IFN-c and not related to a defect in activation or maintenance signals [20]. Increased inflammation and sustained MHC class II expression on microglia (Fig. 8A) further suggested that IFN-c signaling to astrocytes down regulates inflammation universally without altering the composition of the CNS infiltrates. This concept is supported by sustained composition of CNS infiltrates during inflammation induced astrocyte apoptosis [21].
The inability to restrain ongoing inflammation during EAE in GFAPcR1D mice was evident at multiple levels. Astrocyte activation was sustained consistent with increased expression of GFAP mRNA. CCL2, CCL5 and CXCL10 mRNA levels were increased (Fig. 8C). The expression of mRNA encoding potentially destructive immune mediators including IL-1, IL-6, iNOS, and TNF were all increased (Fig. 8C). IFN-c was ,2-fold higher in the CNS of GFAPcR1D mice (Fig. 9). Lastly, the level of the antiinflammatory cytokine IL-10 was reduced ( Fig. 9), suggesting limited availability of IL-10 may be a key in prolonging astrocyte activation and inflammation [22]. A possible link between reduced IL-10 and the inability of IFN-c signaling to astrocytes is provided by the capacity of IFN-c to induce IL-27 in astrocytes [23], thereby promoting IL-10 production. Indeed IL-27 in the CNS of the GFAPcR1D mice was significantly reduced compared to wt mice during chronic disease (Fig. 9). By contrast, IL-12 ( Fig. 9), an indirect suppressor of CNS inflammation by promoting IFN-c production [24], was similar in the CNS of both GFAPcR1D and wt mice. Astrocytes derived from GFAPcR1D and wt mice were stimulated with IFN-c to confirm IFN-c dependent IL-27 secretion by astrocytes [25]. While IFN-c induced IL-27 secretion by astrocytes from wt mice, IL-27 secretion was significantly reduced in cultures derived from GFAPcR1D mice (Fig. 10). These data support the possibility that IFN-c mediated IL-27 secretion by astrocytes regulates EAE effector T cell function, inflammation and tissue destruction via induction of IL-10; however, this does not exclude the possibility that the inability of astrocytes to respond to IFN-c could directly or indirectly dysregulate a variety of other immunomodulatory functions [4,5,17].

Discussion
In the EAE model of MS, IFN-c functions as a proinflammatory cytokine in the early stages of disease [10,11,26], yet it also assumes a prominent anti-inflammatory role during the transition to remission [10,[12][13][14][15]. Protection has been attributed to a variety of potentially interrelated mechanisms including limiting neutrophil accumulation, Th17 cell activation, IL-1R signaling, matrix metalloproteinase secretion, pro-inflammatory cytokine secretion and chemokine activity; in addition IFN-c facilitates T cell apoptosis and protects oligodendrocytes from death via an ER stress response [10,11,17]. The activities of astrocytes during autoimmunity also range from pro-to antiinflammatory [3][4][5]. However, to what extent a direct action of IFN-c on astrocytes contributes to inhibitory mechanism is unclear. The data herein demonstrate that among the multiple  Early functions of astrocytes imposed by innate responses are largely pro-inflammatory during EAE [3,6,7,26]. For example, astrocytes contribute to the loss of BBB integrity via secretion of reactive oxygen species, chemokines and pro-inflammatory cytokines [3][4][5]. This pro-inflammatory milieu is associated with initial axonal damage prior to accumulation of self reactive T cells [27], which further promotes immune mediated damage. Blocking the pro-inflammatory transcription factor NF-kB in astrocytes improves recovery during chronic EAE [28,29]. Supporting an initial pro-inflammatory role of astrocytes dependent on innate, not IFN-c responsiveness, inhibition of IFN-c signaling to astrocytes did not influence EAE onset or incidence, initial disease progression, astrocyte activation, or BBB integrity as indicated by similar entry of inflammatory cells into the brain. By contrast, an inflammation dampening effect of IFN-c became evident during the onset of disease remission. Astrocytes limit ongoing inflammation and pathogenic processes at several levels. They not only facilitate repair by inhibiting inflammatory cell entry into the CNS parenchyma [6,7,30], but also down regulate T cell effector function and proliferation [4,5,21]. For example, EAE in GFAP deficient mice results in more severe and widespread inflammation [31]. In addition, T cell-astrocyte interactions as well as astrocyte secretion of an unidentified soluble product, distinct from nitric oxide, prostaglandins, or tryptophan metabolism, suppress T cell proliferation [4,5]. T cell proliferation was also not inhibited via defective IL-2 release, despite the suggestion that T cell-astrocyte interactions facilitate secretion of anti-inflammatory cytokines [4,5]. Lastly, astrocytes may limit inflammation by triggering apoptosis in T cells [32]. Nevertheless, the role of IFN-c signaling in these potentially anti-inflammatory, protective mechanisms is not clear.
Elevated IFN-c in the CNS of GFAPcR1D mice during both acute and chronic disease excluded limited IFN-c due to sequestration by the transgenic receptor as a mediator of increased clinical severity and mortality. Indeed, enhanced IFN-c production may reflect an attempt to compensate for the loss of IFN-c dependent astrocyte mediated control of inflammation [33]. Sustained expression of pro-inflammatory cytokines, particularly IL-6 and TNF, thus represents a primary mechanism underlying ongoing tissue destruction in GFAPcR1D mice. IL-6 is known to mediate neurological dysfunction [34] and astrocytes are the predominant source of IL-6 during CNS autoimmune disease. Furthermore, its secretion is down regulated by IFN-c induced SOCS activity [17]. On the other hand, sustained TNF implicated dysregulated activation of microglia or macrophages via increased IFN-c or IL-6, as TNF is predominantly secreted by activated CNS macrophages and microglia during EAE [35,36]. However, the down regulation of microglia activation, including TNF secretion, following interaction with activated astrocytes [37] questions this notion. While both IL-6 and TNF may thus contribute to sustained pathological changes, the source of TNF remains unclear. Similarly, astrocytes are a potential source of the IFN-c induced chemokine CXCL10 during EAE, one of the chemokines controlling T cell recruitment [38]. However, the increased expression of CXCL10 coupled with the inability of the astrocytes in the GFAPcR1D mice to respond to IFN-c suggests altered microglia activation and secretion of CXCL10 [38] or activation of CXCL10 transcription via an independent signaling pathway mediated by TNF or Type 1 interferons [39,40].
Importantly, pathology further correlated with decreases in both Tregs and IL-10, but not with increased antigen specific Th17 cells, although astrocytes are critical targets of IL-17 [41]. Although it is possible that the increased IFN-c in the CNS influenced peripheral activation of antigen specific Th17 cells, previous data demonstrated no evidence for expression of the transgene in peripheral organs, including lymphoid organs [18]. IL-10, secreted by a variety of cells types including CD4 + T cells, CD8 + T cells and Tregs [42], inhibits both acute and chronic EAE [43,44]. The decrease in this anti-inflammatory cytokine suggests that the induction of IL-27 secretion is a critical aspect of astrocyte responses to IFN-c, thus promoting IL-10 secretion by T cells [23,25]. However, our data is unable to distinguish the contribution of paracrine versus autocrine IFN-c induced signals to astrocytes on IL-10 production, as IL-10 also regulates astrocyte activation [22], and can itself be secreted by astrocytes to exert autologous functions [45].
In addition to mediating an imbalance of protective versus detrimental cytokines, our results demonstrate that IFN-c signaling to astrocytes limits the extent of astrocyte activation during chronic EAE, specifically in grey matter. Activated astrocytes, especially within and adjacent to areas of demyelination are a prominent feature of white matter plaques associated with both chronic MS and EAE [3][4][5]. Astrocytes protect neurons and oligodendroglia, but also form glial scars which inhibit regeneration after neuronal injury [3,5,8,20]. The absence of reactive astrocytes increases axonal regeneration after injury [46], consistent with the concept that limiting astrogliosis is critical for axonal regeneration after neuronal injury. While the significance of sustained astrocyte activation in grey matter tracks is unclear in our model, it is consistent with increased axonal damage, and suggests possible damage to axons outside the lesions, which may not be manifested by histological analysis.
Limiting inflammation following either infection or during an autoimmune attack is a prerequisite for the initiation of repair. This is critically important within the CNS which has limited regenerative capacity. In summary, our data identify astrocytes as prominent targets underlying IFN-c mediated suppression of chronic CNS inflammation [19]. The data further provide a link between sustained inflammatory responses, enhanced demyelination and axonal degeneration associated with loss of neurological function during EAE and chronic progressive MS. Although the inability of astrocytes to respond to IFN-c did not alter disease in the brain, engagement of the IFN-c receptor on astrocytes in spinal cord limits demyelination and functions in a neuroprotective capacity. Current therapies for MS are primarily focused on antiinflammatory and immunomodulatory approaches and have been partially successful in treating acute episodes. The identification of astrocytes as critical responders and mediators of IFN-c signaling in limiting CNS autoimmune disease may provide insights into new approaches to limit long term progression to disability.

Cytokine Determinations
Cytokines were determined by ELISA using antibody pairs and recombinant cytokine standards from BD Bioscience. IL-27 was measured using QUANTIKINE Mouse IL-27p28 Immunoassays (R&D Systems Inc., Minneapolis, MN).
Sections from at least 3 separate experiments containing at least 3 mice per group were reviewed in a blinded manner. Numbers of GFAP + cells in spinal cord were determined in 15 non-overlapping 406 fields (0.2 mm 2 ) in white matter and gray matter. Stained spinal cord sections of all 6 levels on individual glass slides were scanned (406) and digitally imaged at high resolution with an Aperio ScanScope (Vista, CA). Aperio software was used to quantify areas of demyelination within the white matter tracks of each of the 6 sections per individual mouse. For analysis of CD4 + T cells spinal cords were embedded in Tissue-Tek O.C.T. (Andwin Scientific, Tryon, NC), flash frozen in liquid nitrogen and stored at 280uC. Blocks were warmed to 220uC prior to cutting 10 mm sections by cryostat. Following fixation with acetone for 2 min at 4uC, non-specific antibody binding was blocked with Cyto Q Background Buster (Innovex Biosciences, Richmond, CA) for 15 min. Sections were stained with anti-CD4 (L3T4) antibody (Vector Laboratories) diluted in Cyto Q Immuno Diluent (Innovex Biosciences) followed by biotinylated rabbit anti-rat, peroxidase ABC reagent and visualized with NovaRED substrate (all from Vector Laboratories). Sections were counter stained with hematoxylin to visualize the nuclei. Spinal cord sections were scanned (406) and digitally imaged at high resolution with an Aperio ScanScope.

Gene Expression Analysis
Frozen tissues were homogenized in TRIzol (Invitrogen, Carlsbad, CA) and cDNA prepared as described [47,48]. Quantitative real-time PCR was performed using 4 ml of cDNA and SYBR Green Master Mix (Applied Biosystems, Foster City, CA) in duplicate on a 7500 Fast Real-Time PCR System (Applied Biosystems). Expression levels were normalized to ubiquitin or GAPDH using the following formula:

Statistical Analysis
Statistical significance was determined by two-tailed Student's t test. A value of p,0.05 was considered statistically significant. Figure S1 CD4 + T cell recruitment into the spinal cord during acute EAE. CD4 + T cells were visualized in 10 mm frozen sections of spinal cords from wt and GFAPcR1D tg mice at day 18 p.i. Immunoperoxidase stain (NovaRED chromogen with hematoxylin counterstain). Scale bars = 50 microns. (TIF)