Astrocytes promote a protective immune response to brain Toxoplasma gondii infection via IL-33-ST2 signaling

It is of great interest to understand how invading pathogens are sensed within the brain, a tissue with unique challenges to mounting an immune response. The eukaryotic parasite Toxoplasma gondii colonizes the brain of its hosts, and initiates robust immune cell recruitment, but little is known about pattern recognition of T. gondii within brain tissue. The host damage signal IL-33 is one protein that has been implicated in control of chronic T. gondii infection, but, like many other pattern recognition pathways, IL-33 can signal peripherally, and the specific impact of IL-33 signaling within the brain is unclear. Here, we show that IL-33 is expressed by oligodendrocytes and astrocytes during T. gondii infection, is released locally into the cerebrospinal fluid of T. gondii-infected animals, and is required for control of infection. IL-33 signaling promotes chemokine expression within brain tissue and is required for the recruitment and/or maintenance of blood-derived anti-parasitic immune cells, including proliferating, IFN-γ-expressing T cells and iNOS-expressing monocytes. Importantly, we find that the beneficial effects of IL-33 during chronic infection are not a result of signaling on infiltrating immune cells, but rather on radio-resistant responders, and specifically, astrocytes. Mice with IL-33 receptor-deficient astrocytes fail to mount an adequate adaptive immune response in the CNS to control parasite burden–demonstrating, genetically, that astrocytes can directly respond to IL-33 in vivo. Together, these results indicate a brain-specific mechanism by which IL-33 is released locally, and sensed locally, to engage the peripheral immune system in controlling a pathogen.

Introduction Recruitment of immune cells to the brain during infection is a highly orchestrated process, requiring concerted expression of a number of chemokines and adhesion factors at the bloodbrain barrier [1]. But the cues which precede these factors are less well understood. In particular, in many cases, it is unclear if brain resident cells possess the machinery to detect the presence of pathogens to promote the recruitment of peripheral cells. Murine infection with the eukaryotic parasite Toxoplasma gondii (T. gondii) features continual recruitment of bloodderived immune cells to the brain and serves as an excellent model for better understanding immune responses at this site.
T. gondii is a globally relevant pathogen which infects most warm-blooded vertebrates, including one-third of the human population [2][3][4]. Upon initial exposure of hosts to T. gondii through contaminated food or water [5], an early stage of infection occurs, called the acute phase, during which T. gondii disseminates throughout peripheral tissues [6]. By two-weeks post-infection, parasite has been largely cleared or controlled in most tissues, but ultimately persists in the brain of its hosts for their lifetime [2,[5][6][7][8][9]. Mortality from T. gondii infection is associated with an increased prevalence of replicating parasite in brain tissue, documented in immunosuppressed patients undergoing transplant surgeries [10], and in HIV-AIDS patients [11][12][13], highlighting the importance of the immune response in controlling T. gondii. Indeed, control of brain T. gondii infection requires a Th1-dominated immune response [2,9], whereby CD4+ and CD8+ T cells and the IFN-γ they produce are required for survival [14]. Macrophages also exhibit anti-parasitic effector mechanisms which are necessary to control the parasite [9,[15][16][17][18][19][20][21].
It is not known, however, how the parasite is sensed in the brain to create an environment that promotes immune cell entry, stimulation, and maintenance. During the acute phase of infection in the periphery, dendritic cells and macrophages can sense either the parasite itself or host signals to initiate chemokine and cytokine expression which recruits and skews a strong Th1 immune response [9,[22][23][24][25][26]. However, resident dendritic cells and peripheral immune cells do not exist in brain tissue under steady-state conditions [27,28] and it is unclear if T.gondii-specific molecular patterns are sensed in this tissue. During chronic T. gondii infection of the brain, necrotic lesions form, characterized by the presence of replicating parasite, loss of brain-resident cell markers, and infiltration of immune cells, suggestive of tissue damage and alarmin release [14,[29][30][31][32][33]. We hypothesized that indirect sensing of T. gondii infection, via recognition of host cell damage caused by the parasite, is an important step in instructing the immune response to T. gondii in the brain. Here we focused on the nuclear alarmin, IL-33, as a candidate orchestrator of the immune response to T. gondii. IL-33 is highly expressed in brain tissue [34], and IL-33 signaling has been shown to be protective against tissue pathology during chronic T. gondii infection, but the mechanism by which IL-33 signals, and the immune mechanisms underlying this protection, were not studied in detail [35].
IL-33 is categorized as an alarmin because it is known to amplify immune responses upon signaling through its receptor ST2, also known as il1rl1, without a requirement for secretion or cleavage. A role for IL-33 in stroke [36,37], neurodegeneration [38,39], EAE [40,41], and CNS infection [42][43][44] has been described, but mechanistic understanding of IL-33 signaling during brain disease is limited. It is unclear in many instances which cell type(s) is the relevant responder to IL-33 during disease in the brain. While brain-resident macrophages have been shown to express the IL-33 receptor at baseline [45], astrocytes have been shown to upregulate it during pathology [37,46]. In addition, IL-33 receptor is also expressed on immune cells which can infiltrate the brain from the blood during disease [47,48]. Here, we separate the relative contribution of astrocytes, microglia/macrophages, and infiltrating immune cells in response to IL-33 release during brain pathology, finding that IL-33 can signal strictly within the brain to promote protective immunity to T. gondii.

IL-33 is released during T. gondii brain infection and is required to control parasite
To study T. gondii brain infection, we infected mice by intra-peritoneal injection with the avirulent T. gondii strain Me49 and waited four weeks post-infection for a natural, chronic brain infection to be established. At this time point, infection of most other tissues throughout the body, also known as acute infection, has been controlled [6,7]. For all experiments, brain tissue was harvested at four weeks post-infection unless otherwise specified.
Consistent with its role as a pre-stored alarmin, IL-33 is highly expressed in the brain at baseline [34], and only mildly increases with infection (S1A Fig). Although astrocytes are the major source of IL-33 during development, oligodendrocytes become a significant contributor by postnatal day 30 [34,45,46,49]. In the T. gondii-infected brain, we found nuclear IL-33 to be expressed predominantly by mature, CC1+ oligodendrocytes and also by astrocytes, the frequencies of which varied by brain region (Fig 1A and 1B, S1B Fig). We found IL-33 to be expressed almost exclusively by oligodendrocytes in white matter tracts, such as the corpus callosum, while IL-33 expression in gray matter, such as the cortex, was split more evenly between astrocytes and oligodendrocytes ( Fig 1A and 1B). These data align with IL-33 expression in the uninfected mouse brain [46], indicating that the major sources of IL-33 in the brain do not change with T. gondii infection. Collectively, these results indicate that IL-33 is expressed by glia in the T. gondii infected mouse brain parenchyma. Importantly, we also detected IL-33 protein expression in astrocytes in healthy human brain tissue, as has been shown previously on the transcript level [50](S1C and S1D Fig).
Since IL-33 does not need to be cleaved to be active, it is classically thought of as an alarmin that is released by necrotic cell death [51,52]. During T. gondii infection, the parasite itself as well as the inflammatory environment provide opportunity for host cell damage and alarmin release. During chronic infection, T. gondii predominantly exists in the brain as an intracellular cyst form which is slow growing [6,8] (Fig 1C) and does not appear to pose an immediate risk to cells, due to a lack of observed tissue destruction and inflammation surrounding cysts ( Fig 1C). But, for reasons not fully understood, cysts can reactivate anywhere in the brain, releasing individual parasites previously contained within the cyst wall [8,30] (Fig 1C). Individual parasites can invade surrounding cells, replicate, and can lyse the cell or form a new cyst [53]. The presence of individual replicating parasites is correlated with morbidity in humans, and is thought to cause necrotic lesions in immunocompromised patients [10][11][12][13].

Fig 1. IL-33 is expressed by oligodendrocytes and astrocytes during brain T. gondii infection, is released during infection, and is required for control of parasite. (A and B)
Representatitive images (A) and percentage quantification (B) of nuclear IL-33 protein expression (red) by cell type four weeks post-infection in various brain regions by confocal fluorescent microscopy. IL-33 is costained with the nuclear oligodendrocyte marker, Olig2 (green) or activated astrocyte marker, GFAP, (white). (C and D) Representative images of T. gondii in brain tissue, either in cyst form (arrow), or reactivated individual parasites (box)(C). IL-33-positive cells are absent from inflammatory foci containing replicating parasites (D). (E and F) Extracellular IL-33 release as measured by ELISA after ex vivo incubation of all cells isolated from infected brain tissue (E), or from CSF samples, each dot represents pooled CSF from 4-5 individual mice from a separate infection, displaying four infections in total (F). (G and H) Parasite burden as measured by cyst count from brain homogenate of IL-33-deficient (G) and IL-33 receptor (il1rl1)-deficient mice (H). Statistical significance was determined by two-tailed t-test (E) or randomized block ANOVA, when results from multiple independent experiments are shown (F-H). � = p < .05, �� = p < .01, ��� = p < .001. Scale bars indicate 50μm. https://doi.org/10.1371/journal.ppat.1009027.g001

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain Interestingly, we found clusters of immune cells, including T cells and macrophages, surrounding individual replicating parasites but not T. gondii cysts (S2A and S2B Fig, Fig 1C). Therefore, we hypothesized that either lytic T. gondii replication [53] or local inflammation, can cause release of local damage signals, such as IL-33. Corroborating this hypothesis, while IL-33 tiles evenly throughout uninfected brain tissue [46], we noted focal loss of IL-33 staining, as well as oligodendrocyte and astrocyte markers, at the center of inflammatory lesions containing parasite (Fig 1D, S2C and S2D Fig).
To assess if host cell death occurs in the T. gondii infected brain which could facilitate IL-33 release, we injected mice intraperitoneally with propidium iodide (PI), which is taken up by cells with loss of membrane integrity. 24 hours post-injection, while we did not observe significant clusters of PI positivity in uninfected (naïve) brain tissue (S2E Fig), we found focal areas of PI positivity in chronically infected brain tissue near inflammatory foci, denoted by a lack of GFAP staining and DAPI clustering (S2F Fig). These results indicate that infection of the brain by T. gondii can cause damage to host cells.
To determine if IL-33 was released during brain T. gondii infection, we first utilized an exvivo assay to measure extracellular IL-33 during infection. We processed naïve and infected mouse brains down to a single cell suspension containing brain resident cells as well as immune cells and parasite. We then incubated the single cell suspensions for four hours at 37˚C before taking the supernatant and measuring extracellular IL-33 by ELISA. Strikingly, IL-33 was present in detectable quantities from supernatants of infected brain samples, but not naïve controls, suggesting that T. gondii infection has the capacity to induce IL-33 release ( Fig 1E). We then validated IL-33 release in vivo, by sampling cerebrospinal fluid (CSF) from the cisterna magna of infected mice and found detectable IL-33 levels in pooled CSF samples ( Fig 1F).
We next asked if IL-33 was required to control infection. We infected wildtype and il33 -/mice and assessed their brains for parasite burden, enumerating a significantly increased number of cysts in il33-deficient mice ( Fig 1G). We also detected an increased parasite burden in the brains of infected mice that lacked the IL-33 receptor (known as ST2 or il1rl1), by cyst count and quantitative PCR (Fig 1H, S3B Fig), demonstrating a role for extracellular IL-33 signaling during this infection. At 10 days post-infection, no parasite was detected in the peritoneal cavity of infected il1rl1 -/or wildtype animals. Therefore, il1rl1 -/mice are not delayed or defective in clearing parasite in the periphery. Additionally, parasite levels were equivalent in the brain at day 12 post infection (12DPI) in il1rl1 -/mice in comparison to wildtype mice (S3A Fig), further supporting that the control of parasites is intact during the early stages of infection. We did not find il33 -/or il1rl1 -/mice to succumb to infection. Nonetheless, these results demonstrate that extracellular IL-33 signaling plays a critical role in limiting T. gondii during chronic brain infection.

IL-33 signaling is required for adequate numbers of functional T cells during chronic T. gondii infection
In order to better understand how IL-33 signaling was protective during T. gondii infection, we focused on characterizing il1rl1 -/mice from this point onward. We first profiled recruited immune cell populations in the brain, beginning with the adaptive immune response, because a strong, Th1-biased immune response is absolutely critical for control of chronic T. gondii infection [14]. In the absence of IL-33 signaling during infection, approximately one-third fewer total T cells were present in the brain by flow cytometry (Fig 2A and 2B), suggesting an inability to recruit or maintain these cells. Approximately 96% of T cells in T. gondii infected brains were positive for either CD4 or CD8 markers (Fig 2A), and there was no selective

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain decrease in either of these subsets, but rather a general reduction in T cell counts in il1rl1 -/mice (Fig 2A and 2B). A reduction in T cell numbers could be due to reduced recruitment, reduced survival, or reduced proliferation in the absence of IL-33 signaling. Indeed, fewer CD4+ and CD8+ T cells were proliferating in the brain (Fig 2C and 2D), and fewer of these cells were functional, displaying reduced production of the critical cytokine IFN-γ by ex vivo re-stimulation with PMA/ionomycin (Fig 2E and 2F). IFN-γ is critical because it induces widespread antiparasitic changes, such as increasing chemokine production [54,55], adhesion factor expression [56], and intracellular killing mechanisms of infected cell types, such as macrophages [9]. We found that the chief sources of IFN-γ in the brain during infection were CD4+ and CD8+ T cells, with minimal contribution from NK cells, an important source of the cytokine during early stages of acute infection These results indicate that IL-33 signaling does affect the T cell response; however, it does not appear that these defects underlie defects in brain immunity to T. gondii, since activated T cell numbers and parasite burden were unaffected in the brain at day 12 in il1rl1 -/mice (S3A and S6C Figs). Plasma IFN-γ levels at day 7 were also intact in il1rl1 -/mice (S4G Fig), and are likely enough to control parasite early in infection. Furthermore, a time course showed progressive decreases in IFN-γ+ T cell numbers in il1rl1 -/mice in comparison to wildtype mice (S6 Fig,  Fig 2F), demonstrating brain-specific impacts on immunity in the absence of IL-33 signaling.
It is important to mention that decreased immune activation in il1rl1 -/mice, such as decreased IFN-γ production, during the acute or chronic stage of infection, could be beneficial in protecting against immunopathology. Indeed, il1rl1 -/mice are protected from weight loss, early mortality, and intestinal pathology during the acute stage of infection when mice are infected orally with T. gondii [58]. Infecting with 10 cysts of Me49 parasite intraperitoneally, we do not observe significant mortality or pathology in wildtype mice, but nonetheless, IL-33 signaling could potentiate immunopathology during more severe infections.

IL-33 signaling is required for the recruitment and anti-parasitic function of myeloid cells during brain T. gondii infection
We were also interested in the impact of IL-33 signaling on the myeloid cell lineage since macrophages cluster tightly around replicating parasite in brain tissue (S2B Fig) and express iNOS, a key molecule involved in the control of T. gondii [9]. We assessed numbers of CD11b+ myeloid cells in infected il1rl1 -/mice by flow cytometry, using CD45hi expression to differentiate infiltrating myeloid cells from CD45int microglia (Fig 3A and 3B). Many of the CD45hi CD11b+ cells recruited to the T. gondii-infected brain are Ly6C+ CD11c-and Ly6G-, indicating that a large portion of recruited myeloid cells are Ly6Chi monocytes and Ly6Clo monocyte-derived macrophages [59]. Importantly, infected il1rl1 -/mice displayed a reduced frequency and number of CD45hi myeloid cells, by approximately half, in the brain during chronic infection, while CD45int cell numbers held constant (Fig 3A and 3B).

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain To more specifically assess monocyte recruitment, we used CCR2-RFP reporter mice to visualize monocytes in the brain during chronic infection. Monocyte-derived cells are critical for control of brain T. gondii infection, as anti-CCR2 antibody administered during chronic infection results in rapid mortality [59]. By immunohistochemistry, we observed large numbers of CCR2+ cells throughout the brain 4 weeks post-infection in wildtype mice ( Fig 3C). Importantly, we crossed CCR2-reporters to a il1rl1 -/background, which revealed a marked reduction in CCR2+ cells (Fig 3D and 3E). These results confirmed that there is a defect in monocyte recruitment in the absence of IL-33 signaling during infection.
Next, we assessed the anti-parasitic function of the myeloid compartment in the absence of IL-33 signaling, by focusing on inducible nitric oxide synthase (iNOS) expression. Synthesis of nitric oxide by host cells can deprive the parasite of essential amino acids and prevent parasite growth in vitro [60]. In vivo, iNOS knockout mice succumb to infection during the early chronic phase [16]. Of the myeloid cells that were able to infiltrate the brain, fewer of these cells were making iNOS in the absence of IL-33 signaling (Fig 3F-3I). Interestingly, we did not detect significant iNOS positivity in any tissues, aside from the brain, in acute or chronic infection (S5 Fig). This finding is supported by literature demonstrating that iNOS is particularly important during the chronic phase, but not the acute phase of T. gondii infection [16], for reasons that are unclear. No defects in myeloid cell number were found at baseline in uninfected il1rl1 -/mice, in peripheral tissues during acute infection, including in the peritoneum, blood, and spleen, or in peripheral tissues during chronic infection (S5 Fig). In sum, these results, in conjunction with T cell deficits, demonstrate that IL-33 signaling impacts the presence and function of immune cell populations that are necessary for controlling T. gondii infection in the brain.
Finally, because we noted defects in both IFN-γ+ T cells and iNOS+ macrophages in il1rl1 -/mice, we wanted to better understand the kinetics of the reliance of immune cell populations on IL-33 signaling. Thus, we performed a time course and assessed T cell and myeloid cell numbers and activation in the brain at various stages of infection. We found that at day 12 post-infection, when significant immune populations are first present in the brain, immunity was largely intact in il1rl1 -/mice, measured by IFN-γ+ T cells and INOS+ myeloid cells (S6A-S6E Fig). By day 21, IFN-γ+ T cell numbers were significantly reduced in il1rl1-/-mice in comparison to wildtype mice, and infiltrating myeloid cell numbers were reduced as well (S6F- S6I  Fig). However, at day 21 post-infection, iNOS frequencies were intact in il1rl1 -/mice (S6J Fig), in contrast to a significant decrease of iNOS frequency in il1rl1 -/mice at day 28 post infection ( Fig 3F and 3G). These results indicate that iNOS expression is one of the last phenotypes to be affected by IL-33 signaling.

IL-33 signaling induces expression of factors involved in recruitment of immune cells to the brain
Next, we asked if IL-33 signaling was changing the environment within the brain to make the tissue conducive to immune cell recruitment. Initiation and maintenance of an immune

IL-33 signaling is required for the recruitment and anti-parasitic function of peripheral myeloid cells in the brain during chronic T. gondi infection. (A and B) Representative flow cytometry plots (A) and quantification (B) of CD11b + myeloid cells in the brain 4 weeks post infection. CD45hi expression was used to differentiate infiltrating myeloid cells from CD45int microglia. (C-E) Visualization (C and D)
, and quantification (E) of infiltrated CCR2+ monocytes by fluorescence confocal microscopy of infected CCR2-RFP reporter mice. il1rl1-deficient mice were crossed to CCR2-RFP mice to assess the contribution of IL-33-il1rl1 signaling to monocyte recruitment. (F and G) Representative flow cytometry plots (F) and quantification (G) of iNOS+ CD45hi, CD11b+ infiltrating cells in the brain. (H and I) Visualization (H) and quantification (I) of the size of iNOS foci in brain tissue by fluorescence confocal microscopy. Statistical significance was determined by randomized block ANOVA when two or more experiments were pooled (B, G, I), and by a two-tailed t-test (E). � = p < .05, �� = p < .01, ��� = p < .001. Scale bars indicate 2000μm (C,D) and 50μm (H). https://doi.org/10.1371/journal.ppat.1009027.g003

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain response in the brain requires expression of cytokines, chemokines, adhesion factors, and factors which promote the entry of and maintain proliferation of immune cells [1]. It is well known that T. gondii infection induces many of these factors [54][55][56]61],which we validated by whole brain qRT-PCR. We find a profound increase in brain ccl2, cxcl9, cxcl10, ccl5, cxcl1, vcam, and icam expression over uninfected controls (Fig 4A). When we assessed the same genes in infected il1rl1-deficient mice, we saw significantly decreased expression of the chemokines ccl2, cxcl10, and cxcl1 (Fig 4B), along with smaller reductions in the expression of the adhesion factors vcam and icam (Fig 4B). Interestingly, ccl2 and cxcl10 expression has been attributed to astrocytes by in situ hybridization during T. gondii infection [55]. Although cxcl1 has not been extensively studied in chronic T. gondii infection, it has also been reported in astrocytes during neuroinflammation [62]. However, we did not observe an effect of IL-33 signaling on cxcl9 expression, a chemokine which is made by PU.1-expressing cells rather than astrocytes [63] (Fig 4B). These results suggest that IL-33 signaling, either directly or indirectly, could be inducing chemokine expression in astrocytes during T. gondii infection.
We next wanted to identify the cellular source of chemokine in infected brain tissue to better understand where IL-33 was exerting its effects. We focused on studying the expression pattern of a chemokine whose transcript levels were altered to the greatest degree in the absence of IL-33 signaling, the monocyte chemoattractant ccl2. We used immunofluorescence microscopy to image the brain tissue of chronically infected ccl2-mCherry reporter mice [64]. We observed ccl2-mCherry expression in "hotspots" throughout the brain, implicating a local response to signals in brain tissue (Fig 4C). We validated by immunohistochemistry that ccl2 expression was greatly reduced in il1rl1 -/mice by crossing our ccl2-mCherry reporters to an il1rl1-deficient background (Fig 4C-4E). Specifically, ccl2 foci in il1rl1 -/mice were much reduced in size compared with wildtype infected mice (Fig 4D and 4E), while ccl2 expression was unchanged between groups at baseline in uninfected mice (S7A Fig). Of note, we also observed astrocytes to co-express IL-33 and ccl2, but only approximately 25% of ccl2+ astrocytes in the infected cortex expressed IL-33 (S7B and S7C Fig). Although multiple signals undoubtedly converge to induce chemokine expression during infection, including other innate cytokines [65], these data suggest that IL-33 is a major contributor to the induction of local ccl2, supporting a role for IL-33 in inducing chemokine production to promote immune cell entry to the brain.
To get a better understanding of which cells might be responding directly or indirectly to IL-33 to induce ccl2 expression, we next assessed which cells were producing the ccl2, and found that ccl2-mCherry in infected brain tissue co-localized with both GFAP+ astrocytes and Iba1+ macrophages (Fig 4F). Approximately 75% of cells expressing ccl2 were astrocytes, 22% were Iba1+ macrophages, and 3% of cells did not co-stain with either of these markers ( Fig  4G). These results were validated by assessing ccl2 expression in brain cells magnetically enriched for either astrocytes or macrophages. To enrich for astrocytes, we used a previously validated protocol, which first removes CD11b+ cells by negative magnetic isolation, followed RNA from these cell populations showed a 10-fold higher ccl2 expression in astrocytes compared with macrophages during chronic infection ( Fig 4H). Thus, because ccl2 expression is dependent on IL-33 signaling, and because astrocytes are the major producer of ccl2 in the infected brain, we hypothesized that IL-33 can signal directly on astrocytes to induce chemokine expression, or signals through an intermediate cell type to impact astrocytic chemokine expression.

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain

IL-33 signals on a radio-resistant cell type to control chronic T. gondii infection
In order to further understand IL-33 signaling in the brain, we began to narrow down the cell type(s) that directly responds to IL-33 to support protective immune responses. In studies concerning IL-33 during CNS disease, genetic evidence for a responding cell type to IL-33 is lacking. This is likely due in part to the fact that several brain-resident cells can express ST2 (il1rl1) during disease, such as microglia, astrocytes, and endothelial cells [37,[67][68][69], but a wide range of immune cells can also express the receptor, including but not limited to type 2 innate lymphoid cells (ILC2s), regulatory T cells, Th2 cells, mast cells, and macrophages [48,70,71]. During T. gondii infection, we did not detect ST2 expression on the immune cells present in highest numbers in the infected brain-effector T cells and monocyte-derived myeloid cells (S9 Fig). We did, however, detect ST2 expression on ILC2s, which decreased significantly in frequency from uninfected to infected brain tissue, as well as mast cells, and regulatory T cells (S9 Fig).
To determine if IL-33 signals on a radio-sensitive or a radio-resistant cell type to exert its effects during chronic T. gondii infection, we lethally irradiated wildtype and il1rl1 -/mice and reconstituted these mice with bone marrow from either wildtype donors or il1rl1 -/donors ( Fig 5A). Blood was assessed for reconstitution prior to infection (Fig 5B-5D), and in all cases, immune cells were >90% positive for the congenic CD45 marker of donor mice, confirming successful reconstitution of the chimera (Fig 5B-5D). We note that irradiation did alter the severity of infection, resulting in high parasite burdens in all groups, and a high number of infiltrating cells (Fig 5E-5G) Nonetheless, these experiments revealed that il1rl1-deficiency on radio-resistant cells, or in il1rl1 -/recipient, but not donor mice, recapitulated the phenotype of global knockouts, including reduced infiltrating myeloid and T cell numbers, and parasite burden, whereas il1rl1 expression on donor bone marrow, was dispensable (Fig 5E-5G). These data demonstrate that IL-33 signals predominantly on non-hematopoietic cells to control brain T. gondii infection, thus largely ruling out immune cells as responders.
Importantly, microglia are not fully replenished by bone marrow-derived cells post-irradiation. Therefore, we cannot rule out microglia as responding to IL-33 in the brain. In addition, microglia highly express il1rl1 at baseline [37,45,46,67], prior to chronic infection. To better understand which radio-resistant cells were capable of responding to IL-33, we assessed il1rl1 expression on microglia/macrophages, and ACSA-2+ astrocytes, which have also been shown to express the IL-33 receptor, especially during disease [37,46,67]. To do this, we magnetically enriched for these cell types from naïve and chronically infected brain tissue, and found that microglia/macrophages cells express il1rl1 at high levels in uninfected mice, but downregulate the IL-33 receptor 20-fold upon infection, while astrocytes express low levels at baseline and increase receptor expression with infection (S10A Fig). This result indicates a change in capability of cells which are able to respond to IL-33 prior to and following infection. Other brain resident cells, such as endothelial cells, neurons, and oligodendrocytes may express the IL-33 receptor during T. gondii infection and contribute to immunity. We chose to focus on microglia/macrophages and astrocytes because they have been previously reported to be the predominant cells types that express the IL-33 receptor [37,45,46,49,67,68,72], are major producers of IL-33-dependent chemokine during T. gondii infection [55] (Fig 4G and 4H), and are required for control of T. gondii infection [73,74].

IL-33 signaling on astrocytes, but not microglia/macrophages, potentiates immune responses and limits parasite burden
To determine if IL-33 directly signals on microglia/macrophages or astrocytes, we crossed il1rl1 fl/fl mice [75], to either constitutive CX3CR1cre mice or GFAPcre mice, respectively. We

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain observed that deletion of il1rl1 from macrophages did not affect any major phenotype observed in il1rl1 -/knockout mice, including brain myeloid cell number and function, T cell number, or parasite burden (Fig 6A-6C). We observed minor defects in CD4+ T cell number in these mice, including CD4+ T cell proliferation (Fig 6A). When we deleted il1rl1 from astrocytes (S10B Fig), while myeloid cell recruitment and activation was relatively unaffected, the adaptive immune response was impacted, including reduced T cell proliferation and IFN-γ production, in mice with IL-33R-deficient astrocytes (Fig 6D and 6E). In contrast with CX3CR1cre mice, GFAPcre il1rl1 fl/fl mice exhibited stronger CD8+ T cell proliferation deficits than CD4+ T cells (Fig 6D), perhaps suggesting an impact of IL-33 signaling on antigen presentation. Importantly, aligned with a decrease in T cell-derived IFN-γ, which is critical for control of infection [14], GFAPcre il1rl1 fl/fl mice displayed increased parasite burden (Fig 6F). GFAPcre il1rl1 fl/fl mice did not have defects in immune cell compartments in other tissues,

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain such as the spleen, displaying increased immune cell numbers in the spleen during chronic infection (S10C Fig), to a level comparable with whole-body il1rl1-deficiency. Collectively, these results show that astrocytes respond to a damage signal, IL-33, to promote a protective immune response to T. gondii in the brain.

Discussion
We observe that brain resident cells can respond to IL-33, a damage-associated molecular pattern (DAMP), to initiate protective immunity during chronic T. gondii infection. This work begins to uncover the mechanisms by which the presence of T. gondii can be sensed specifically in brain tissue, which is devoid of peripheral immune cells in the healthy state. Indeed, much of the work on sensing of T. gondii has focused on immune cell recognition of T. gondii in peripheral tissues, during the acute phase of infection. It has been shown that T. gondii profilin protein, a pathogen-associated molecular pattern (PAMP), can be sensed directly by TLR11 and TLR12-expressing dendritic cells in the periphery, which are capable of initiating protective Th1 immunity by producing IL-12 [9,22,23]. In humans, however, TLR11 is a pseudogene and TLR12 is not expressed [76]. Thus, more recent studies have focused on innate sensing mechanisms with potential relevance to humans, including NOD-like receptor sensing of T. gondii [24,25], as well as alarmin recognition, such as S100A11 [26], by macrophages. Since we find IL-33 to be highly expressed in human brain tissue, we hypothesize that our results will be relevant to control of human brain infection. Furthermore, our results shed light on the importance of continual DAMP recognition to contain infection long after adaptive immune responses are first primed in peripheral tissues.
We found nuclear IL-33 to be expressed by oligodendrocytes and astrocytes in the mouse T. gondii-infected brain, and we were able to detect extracellular IL-33 in the cerebrospinal fluid during chronic infection. Since extracellular IL-33 is negatively regulated in a number of ways to limit inflammation [51,77,78], this result suggests high levels of IL-33 release during brain T. gondii infection. Our results also raise the question of how IL-33 is liberated from astrocytes and oligodendrocytes during disease. Although our results suggest that host cell death, indicated by propidium iodide staining, is occurring in the T. gondii infected brain which could release IL-33, the possibility of secretion of IL-33 from living cells cannot be ruled out [71]. Mechanisms of glial death during disease have not been extensively studied in vivo, and it is unclear if the parasite itself, or if secondary, toxic inflammation can mediate the release of host damage signals. It is also unclear how long-ranging the effects of alarmin signaling are. Work on IL-33 in cardiac muscle [75], as well as the retina [79], has indicated that IL-33 often signals very locally, acting in either an autocrine or paracrine manner. For example, Müller cells, specialized astrocytes in the retina, have been implicated in the release and sensing of IL-33 in the retina [79]. Our work further corroborates local IL-33 signaling, as IL-33-dependent ccl2 expression is found adjacent to necrotic lesions containing replicating T. gondii. Thus, IL-33 may be released by astrocytes or oligodendrocytes, and can signal on astrocytes, which could be the same cell or an adjacent cell, to induce chemokine expression necessary for the recruitment of anti-parasitic immune cells to the brain.
In the lung, gut, and skin, IL-33 has historically been shown to signal almost exclusively on immune cells to orchestrate immunity [48]. However, recent work demonstrates that IL-33 can signal on non-hematopoietic cells-including cardiomyocytes in the heart [75] and endothelial cells in culture [69,80]. We found, consistent with a previous report in stroke [37], that IL-33 signals on a radio-resistant responder during brain T. gondii infection. Tissue microenvironment and context likely influence whether IL-33 signaling on immune cells is relevant. Peripheral immune cells are not present in healthy brain tissue due to the blood brain barrier,

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain thus necessitating the capability of brain resident cells to respond to damage. Additionally, the environmental milieu specific to CNS T. gondii infection promotes a robust Th-1skewed immune response necessary for intracellular parasite killing, characterized by T-cell derived IFN-γ and macrophage-derived iNOS [7,9]. Thus, the IL-33-dependent mechanisms, specifically those activating type 2 immune responses that drive asthma, allergy, and the expulsion of helminth infections [47,70], may not be relevant to brain infection with an intracellular pathogen.
When we examined brain-resident cell expression of il1rl1, we found that microglia express high levels of il1rl1 in healthy brain tissue, but they downregulate the receptor 20-fold upon infection. A downshift in microglia inflammatory genes has been recently reported by our laboratory during T.gondii infection [73], and it is unclear what factors drive this process. However, these results do not rule out the importance of IL-33 signaling on microglia in other settings. Indeed, during development, microglia expression of il1rl1 is required for microglial phagocytosis of synapses [45]. While we did not find any major impact of microglia/macrophage il1rl1 expression on control of chronic T. gondii infection, it is possible that microglia respond to IL-33 at earlier time points, such as when the parasite first reaches the brain. If microglia normally respond to IL-33 earlier in infection, our results indicate that astrocyte responses to IL-33, or additional MyD88-dependent signals, may compensate for the lack of il1rl1-signaling in microglia.
In contrast to microglia/macrophages, we found that astrocytes increase expression of the IL-33 receptor upon infection. Astrocytes are critical for controlling CNS infection and potentiate inflammation. For example, astrocyte responses to IFN-γ are necessary to control T. gondii infection [74]. We found that IL-33 receptor expression on astrocytes was required for adequate promotion of an anti-parasitic immune response in the brain, including proliferation and IFN-γ production in T cells, which is critical for control of T. gondii. This is the first time, to our knowledge, that astrocytes have been directly implicated in an IL-33 response via genetic manipulation. Our results also raise questions regarding the role of astrocytes in supporting T cell responses to infection and whether these responses are direct or indirect. IL-33 signals through the adaptor MyD88 [70], and could influence the expression of numerous molecules that support T cell responses, including trophic factors, TCR engagement, chemokines, or blood brain barrier opening for T cell entry.
Notably, immune defects in GFAPcre il1rl1 fl/fl mice did not match the magnitude observed in il1rl1 -/mice, and did not significantly impact myeloid cell recruitment. These results either implicate compensation by other cell types when il1rl1 is deleted from a single population, or the presence of additional il1rl1-expressing radio-resistant cell types not considered here, such as neurons, oligodendrocytes, and endothelial cells, which have recently been shown to express il1rl1 during disease and can be directly activated in response to IL-33 in vitro [69,80]. Additionally, it is possible that a decrease of circulating IFN-γ at day 12 post infection in il1rl1 -/mice has lasting impacts on chemokine and adhesion expression responsible for recruitment of myeloid cells. Nonetheless, our results indicate that astrocytes are capable of responding to extracellular IL-33 during infection. This finding, in addition to the detection of IL-33 release in the brain, has broad implications for sensing of tissue damage in other disease states, as definitive evidence that alarmins can signal strictly within brain tissue to initiate protective immunity.
Our results underscore the importance of the signaling of one alarmin in controlling chronic T. gondii infection. Little is known about the direct recognition of T. gondii in brain tissue and we implicate the response to infection-induced damage as a major mechanism of innate immune activation in the brain throughout chronic infection. Numerous canonical alarmins are expressed in the brain and their receptors are differentially expressed by cell type.

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain Thus, there is likely a complex integration of multiple PAMP and DAMP signals that is required to allow peripheral immune cells to overcome the blood brain barrier, infiltrate the tissue parenchyma, and carry out effector function. Identifying the cell types that can respond to specific DAMPs in the brain will be key not only to understanding immunity to CNS infection but any neuroinflammatory condition with hallmarks of cell death, including neurodegeneration.

Ethics Statement
Animal care and procedures adhered to the regulations of the Institutional Animal Care and Use Committee (IACUC) at the University of Virginia under protocol number 3968 and at Cornell University under protocol number 2014-0119. For human tissue samples, consent was not obtained due to anonymity of samples.

Contact for reagent and resource sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Tajie Harris (tajieharris@virginia.edu). . il33 -/animals were obtained with a Materials Transfer Agreement from Amgen by Elia Tait Wojno, currently at the University of Washington and previously at Cornell University. All IL-33 -/animals were housed at Cornell University. il1rl1-floxed embryos were received from KOMP repository (RRID MMRRC:048182-UCD). Importantly, we received notice after fully breeding these mice to cre-expressing mice, that the deposited floxed embryos contain a copy of an inducible, cardiomyocyte-driven cre [75], unknown to the KOMP repository. Upon receiving notice, we did detect the presence of MYH6cre ERT2 alleles throughout our il1rl1 fl/fl colony. Importantly, we deleted il1rl1 under the control of constitutive cre drivers and never administered tamoxifen to these mice. We performed statistical tests to determine the effect of MYH6cre expression, and did not find evidence of an effect in experiments where the MYH6cre was present in some animals and not others (S11 Fig). Moreover, in the experiments reported here (Fig 6), all animals were positive for the MYH6cre allele. All animals were housed in a UVA specific pathogen-free facility with a 12h light/dark cycle. Mice were age and sex matched for each experiment, and were sacrificed in parallel. Animals were infected with T. gondii at 7 to 9 weeks of age and were housed separately from breeding animals.

Human brain tissue
Healthy human brain samples from adult patients were obtained from the UVA Human Biorepository and Tissue Research Facility. Samples were preserved on paraffin embedded slides. Patient identification and medical background was withheld and therefore IRB approval was not required.

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain

Parasite strains
The avirulent, type II Toxoplasma gondii strain Me49 was used for all infections (gift from Christopher Hunter, University of Pennsylvania). T. gondii cysts were maintained in chronically infected (1-6 months) Swiss Webster (Charles River) mice. To generate cysts for experimental infections, CBA/J (Jackson Laboratories) mice were infected with 10 cysts from brain homogenate of Swiss Webster mice by i.p. injection in 200μl PBS. 5-30 cysts from 4 weekinfected CBA/J brain homogenate were then used to infect animals in all experiments.
Human tissue immunofluorescence. Slides containing 4 μm sections of human brain tissue were received from the UVA Biorepository and Tissue Research Facility and deparaffinized in a gradient from 100% xylene (Fisher) to 50% ethanol (Decon Laboratories). Slides were then washed in running water and distilled water. Antigen retrieval was performed by incubating slides in antigen retrieval buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) in an Aroma digital rice cooker for 45 min at 95˚C. Slides were then washed in running water followed by PBS-TW. Slides were then incubated with primary and secondary antibodies as described above for mouse brain tissue. Prior to imaging, Autoflourescence Eliminator Reagent was applied per the manufacturer's instructions (EMD Millipore Cat#2160).
Propidium Iodide injection. Adult naïve or 4-week infected mice were injected intraperitoneally with 20mg/kg propidium iodide (Invitrogen, cat#P1304MP). 24 hours post injection, the mice were sacrificed and their brains were fixed in 4% PFA and imaged for endogenous fluorescence by confocal microscopy.

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain Tissue processing and flow cytometry. Whole PBS-perfused brains were collected into 4 mL of complete RPMI (cRPMI)(10% fetal bovine serum, 1%NEAA, 1%Pen/Strep, 1%Sodium Pyruvate, 0.1%-β-mercaptoethanol). Papain digestion was performed for the chimera experiment. To perform papain digestion, brains were cut into 6 pieces and incubated in 5 mL HBSS containing 50U/mL DNase (Roche), and 4U/mL papain (Worthington-Biochem, Cat#LS003126) for 45 min at 37˚C. Tissue was triturated first with a large bore glass pipette tip, and twice with a smallbore pipette tip every 15 min. In all other experiments collagenase/dispase was used to digest brain tissue. To perform collagenase/dispase digestion, perfused brains were minced using a razor blade and passed through an 18-gauge needle. Brains were then digested with 0.227mg/mL collagenase/dispase (Sigma-Aldrich Cat#11097113001) and 50U/mL DNase1 (Millipore Sigma Cat#10104159001) for 1h at 37˚C. Following digestion, homogenate was strained through a 70 μm nylon filter (Corning, Cat#352350). Samples were then pelleted and spun in 20 mL 40% Percoll (GE Healthcare Cat#17-0891-09) at 650g for 25 min. Myelin was aspirated and cell pellets were washed with cRPMI. Finally, samples were resuspended in cRPMI and cells were enumerated. Meninges were collected from peeling from the brain and scoring from the skull cap and pooled. Meninges were passed through an 18-gauge needle five times and mashed through a 70μm filter. Samples were spun at 1500rpm for 10 min, and resuspended. Spleens were collected into 4 mL cRPMI and macerated through a 40 μM nylon filter (Corning, Cat#352340). Samples were pelleted and resuspended in 2 mL RBC lysis buffer (0. 16
For qRT-PCR assessment of parasite burden, brain tissue was placed in complete RPMI post-harvest, minced with a razor blade, and passed through an 18-gauge needle five times. Brain homogenate (300μL) was then put in a tube with zirconia/silica beads (Biospec, Cat#11079110z) and homogenized in a Mini-bead beater (Biospec) machine. Homogenate was then incubated at 65C for 3 hours with Lysis buffer from the Genomic DNA Isolate II Genomic DNA Kit (Bioline, Cat#BIO-52067). DNA was isolated from brain samples following the manufacturer's instructions. A 2X Taq-based mastermix (Bioline Cat#BIO-86005) was used for all reactions and run on a CFX384 Real-Time System (Bio-Rad), using T. gondii genomes as a standard curve, ranging in 10-fold dilutions from 300,000 genomes to 30 genomes. T. gondii genomes were isolated from cultured, infected human foreskin fibroblast cells. DNA from brain samples was diluted to 500ng/well. The Taqman assay used to detect T. gondii has been described previously [81].

T. gondii cyst counts
Brain tissue (100 mg) was minced with a razor in 2mL cRPMI. Brain tissue was then passed through an 18-gauge and 22-gauge needle. 30 uL of resulting homogenate was pipetted onto a microscope slide (VWR) and counted on a Brightfield DM 2000 LED microscope (Leica). Cyst counts were extrapolated for whole brains.

Bone marrow chimera
Wildtype B6.SJL-Ptprc a Pepc b /BoyJ (C57BL/6 CD45.1) and il1rl1 -/-C57BL/6 CD45.2 mice were irradiated with 1000 rad. Irradiated mice received 3x10 6 bone marrow cells from CD45.1 and CD45.2 donors the same day. Bone marrow was transferred by retro-orbital i.v. injection under isoflurane anesthetization. All mice received sulfa-antibiotic water for 2 weeks post-irradiation and were given 6 weeks for bone marrow to reconstitute. At 6 weeks, tail blood was collected from representative mice and assessed for reconstitution by flow cytometry. Mice were then infected for 4 weeks prior to analysis.

Brain homogenate ex vivo supernatant collection
Brains were harvested from naïve and infected mice. Brain tissue was processed down to a single cell suspension as outlined in "tissue processing and flow cytometry". Cells from half a brain were incubated in a 96-well plate in 200μL for 4 hours at 37˚C. Cells were then pelleted at 1500rpm for 5 min, and supernatant was stored at -80˚C for measurement of IL-33 by ELISA.

CSF collection
Naïve and infected mice were anesthetized with ketamine/xylazine, their necks were shaved and obstructing skin and tissue surrounding the dura of the cisterna magna was removed. A

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain small glass capillary was inserted into the dura of the cisterna magna, allowing approximately 10μL of CSF to fill the capillary. CSF from 4-5 mice was pooled and frozen at -80˚C.

IL-33 ELISA
The IL-33 Quantikine ELISA kit was purchased from R&D Systems (Cat#M3300) and manufacturer instructions were followed. 50μL of either brain homogenate supernatant (ex-vivo assay) or CSF was used as sample volume. 4-5 CSF samples from individual animals were pooled to reach 50μL sample volume. IL-33 standards were serially diluted to a minimum of 30pg/mL. Final values of standards and samples were read on a spectrophotometer at 450nm.

IFN-γ and IL-12 ELISAs
For ELISAs from day 7 post-infection, tail blood was taken from infected animals and incubated with 30U/mL heparin (Sigma, H3149-10KU) before being spun at 2000g for 10 min at 4˚C to obtain plasma. For the IFN-γ ELISA done at 12 days-post infection, blood was taken from cardiac puncture and allowed to sit for 30 min at room temperature before spinning at 2000g for 10 min at 4˚C to obtain serum. Quantikine ELISA kits for IFN-γ and IL-12p40 were purchased from R&D Systems (Cat# MIF00, Cat#M1240) and manufacturer instructions were followed, using 50μL of plasma or serum samples diluted twenty-fold. Final values of standards and samples were read on a spectrophotometer at 450nm.

ACSA-2/CD11b magnetic enrichment
Brain tissue was harvested from mice and processed down to a single cell suspension as described in "tissue processing and flow cytometry" methods, except myelin removal beads (Miltenyi Cat#130096733) were used in place of Percoll. The Anti-ACSA2 and CD11b Microbead kits, purchased from Miltenyi (Cat#130097678, Cat#130093634), were used to enrich for astrocytes and macrophages, respectively, over magnetic columns (Miltenyi Cat#130-042-401). In the case of enriching for astrocytes, both kits were used to first remove CD11b+ cells and then enrich for ACSA-2+ cells. Double-purifications were performed in all cases, which was optional, but recommended, by Miltenyi's protocol. Cells were pelleted at 1500rpm and stored in 300μL Trizol (Ambion) at -80˚C until further use.

Statistical analyses
Statistical analyses comparing two groups at one time point were done using a student's t-test in Prism software, v. 7.0a. Statistical analyses comparing more than two groups within the same timepoint or infection were done using a one-way Anova. In instances where data from multiple infections were combined, all from the same time-point post infection, a randomized block ANOVA was performed using R v. 3.4.4 statistical software to account for variability between infections. Genotype was modeled as a fixed effect and experimental day as a random effect. P values are indicated as follows: ns = not significant p>.05, � p < .05, �� p < .01, ��� p < .001. The number of mice per group, test used, and p values are denoted in each figure legend. Data was graphed using Prism software, v.7.0a.
Supporting information S1 Fig. IL-33 expression in mouse and human brain tissue. (A) Real time PCR for il33 transcript from whole brain homogente at 4 weeks post infection compared to naïve brain tissue. (B) Colocalization, denoted by gray arrows, of nuclear IL-33 protein (red) with mature oligodendrocytes, marked by nuclear Olig2 expression (green), and CC1(white) by confocal

PLOS PATHOGENS
Astrocyte responses to IL-33 promote immunity to T. gondii in the brain fluorescence microscopy of infected mouse brain tissue. (C and D) Confocal fluorescence microscopy of nuclear IL-33 stain present in astrocytes (C) but not oligodendrocytes (D) in the temporal lobe of human brain tissue from patients that did not succumb to toxoplasmic encephalitis (healthy). Statistical significance was determined by randomized block ANOVA (A), which shows data pooled from two independent experiments � = p < .05, �� = p < .01, ��� = p < .001. Scale bars indicate 50μm.  Table. Custom primers for transmembrane il1rl1. For specific detection of the IL-33 transmembrane receptor, rather than soluble ST2. (XLSX) acknowledge the support we received from core facilities at the University of Virginia, including the Biorepository and Tissue Research Facility, the Flow Cytometry Core, and the Research Histology Core.