CD8 T Cells and STAT1 Signaling Are Essential Codeterminants in Protection from Polyomavirus Encephalopathy

A comprehensive understanding of JCPyV-induced PML pathogenesis is needed to define determinants that predispose patients to PML, a goal whose urgency is heightened by the lack of anti-JCPyV agents. A handicap to achieving this goal is the lack of a tractable animal model to study PML pathogenesis. Using intracerebral inoculation with MuPyV, we found that MuPyV encephalitis in wild-type mice causes an encephalopathy, which is markedly exacerbated in mice deficient in STAT1, a molecule involved in transducing signals from type I, II, and III IFN receptors. CD8 T cell deficiency compounded the severity of MuPyV neuropathology and resulted in dramatically elevated virus levels in the CNS. These findings demonstrate that STAT1 signaling and CD8 T cells concomitantly act to mitigate MuPyV-encephalopathy and control viral infection.

ABSTRACT

JC polyomavirus (JCPyV), a human-specific virus, causes the aggressive brain-demyelinating disease progressive multifocal leukoencephalopathy (PML) in individuals with depressed immune status. The increasing incidence of PML in patients receiving immunotherapeutic and chemotherapeutic agents creates a pressing clinical need to define biomarkers to stratify PML risk and develop anti-JCPyV interventions. Mouse polyomavirus (MuPyV) CNS infection causes encephalopathology and may provide insight into JCPyV-PML pathogenesis. Type I, II, and III interferons (IFNs), which all signal via the STAT1 transcription factor, mediate innate and adaptive immune defense against a variety of viral infections. We previously reported that type I and II IFNs control MuPyV infection in non-central nervous system (CNS) organs, but their relative contributions to MuPyV control in the brain remain unknown. To this end, mice deficient in type I, II, or III IFN receptors or STAT1 were infected intracerebrally with MuPyV. We found that STAT1, but not type I, II, or III IFNs, mediated viral control during acute and persistent MuPyV encephalitis. Mice deficient in STAT1 also developed severe hydrocephalus, blood-brain barrier permeability, and increased brain infiltration by myeloid cells. CD8 T cell deficiency alone did not increase MuPyV infection and pathology in the brain. In the absence of STAT1 signaling, however, depletion of CD8 T cells resulted in lytic infection of the choroid plexus and ependymal lining, marked meningitis, and 100% mortality within 2 weeks postinfection. Collectively, these findings indicate that STAT1 signaling and CD8 T cells cocontribute to controlling MuPyV infection in the brain and CNS injury.

IMPORTANCE A comprehensive understanding of JCPyV-induced PML pathogenesis is needed to define determinants that predispose patients to PML, a goal whose urgency is heightened by the lack of anti-JCPyV agents. A handicap to achieving this goal is the lack of a tractable animal model to study PML pathogenesis. Using intracerebral inoculation with MuPyV, we found that MuPyV encephalitis in wild-type mice causes an encephalopathy, which is markedly exacerbated in mice deficient in STAT1, a molecule involved in transducing signals from type I, II, and III IFN receptors. CD8 T cell deficiency compounded the severity of MuPyV neuropathology and resulted in dramatically elevated virus levels in the CNS. These findings demonstrate that STAT1 signaling and CD8 T cells concomitantly act to mitigate MuPyV-encephalopathy and control viral infection.

INTRODUCTION

Interferons (IFNs) exhibit a range of antiviral activity through induction of a host cell-intrinsic antiviral state and by regulating innate and adaptive immunity (1–4). Interferons (IFNs) are classified into three types: type I, II, and III. Type I IFNs, consisting of IFN-α, -β, -ω, -ε, and -κ, signal through the heterodimeric IFN-αβ receptor (IFNAR) complex (5). Nearly all brain-resident neurons and glial cells are capable of producing and responding to type I IFNs (6). In contrast, IFN-γ, the sole member of the type II IFN family, signals through the IFN-γR and IFN-γ production is restricted to T cells, NK cells, and NKT cells (7, 8). Type III IFNs, consisting of IFN-λ1, -λ2, -λ3, and -λ4, signal through the heterodimeric IFNLR (9, 10); only cells of epithelial and endothelial lineage produce type III IFNs (11). Although members of the IFN family signal through distinct receptors, transduction is conveyed via Janus kinase (JAK)/signal transducer and activator of transcription (STAT) signaling pathways (8). All members of the IFN receptor family signal through STAT1. Activated STAT1/STAT2 heterodimers and STAT1 homodimers translocate into the nucleus where they bind interferon-sensitive response elements and gamma-activated sequences in the promoter regions of IFN-related genes (8). Other cytokine signaling pathways also utilize STAT1, including of common γ (γ_c) and β chain cytokine receptor families (12). The cellular distribution of IFN receptors and constellation of JAK and STAT signaling molecules orchestrate cell-intrinsic immune programs and tailor tissue-specific responses to pathogens. The transcription of IFN-related genes induces potent antiviral defenses against many viral infections.

Fourteen polyomaviruses (PyVs) are constituents of the human virome (13–15). In immunocompetent individuals, PyV infection is asymptomatic and persists lifelong, primarily in the urogenital tract, in the central nervous system (CNS), and in hematopoietic cells (13, 16). Three PyVs have to date been shown to be part of the skin microbiome, suggesting skin as an additional site of PyV persistence (17). Dampened immune status (e.g., HIV/AIDS, immunomodulatory therapies for autoimmune/inflammatory disorders, hematological malignancies, organ transplantation, the elderly) predisposes individuals to PyV-induced diseases, including BKPyV-associated kidney transplant nephropathy, Merkel cell virus carcinoma, and JCPyV-induced progressive multifocal leukoencephalopathy (PML) (13, 18). PML is an often-fatal disease characterized by subcortical demyelination presaged by loss of JCPyV-infected astrocytes and oligodendrocytes (17, 19). JCPyV CNS infection has also been reported to cause encephalopathy, cerebellar granule cell neuropathy, and meningitis (20). Even in these at-risk populations, PML is rare, suggesting that other yet to be defined determinants underlie PML development. The tight species specificity of PyV has handicapped progress in elucidating the pathogenesis of human PyV diseases and immunological mechanisms controlling infections and pathological sequelae.

Mouse polyomavirus (MuPyV) and JCPyV have similar genomic and structural features and both establish persistent, asymptomatic infections in their natural hosts that are controlled by adaptive immunity (21, 22). We reported that IFN-γ and type I IFNs mediate viral control in the kidney and spleen (23, 24). Here, we sought to determine whether type I, II, or III IFNs or STAT1 confer protection against MuPyV pathogenesis in the brain. Using intracerebral (i.c.) inoculation with MuPyV, we found that STAT1 signaling contributed to controlling acute and persistent viral infection, but absence of type I, II, or III IFN receptors individually did not affect persistent MuPyV infection. Furthermore, type I, II, or III IFN signaling did not alter MuPyV-specific CD8 T cell responses or protection against MuPyV-induced encephalopathy. Loss of STAT1 signaling, however, led to unchecked immune infiltration and hydrocephalus. Notably, CD8 T cell deficiency per se did not affect virus control in the CNS or neuropathology, but depletion of CD8 T cells in STAT1-deficient mice led to dramatically higher CNS virus infection and early mortality. These findings indicate that STAT1 signaling and CD8 T cells act in concert to control PyV replication in the brain and to mitigate CNS injury.

RESULTS

STAT1 signaling controls acute MuPyV infection in the brain and spleen.

Type I, II, and III IFNs have been shown to control many neurotropic and gliatropic viral infections (e.g., coronavirus, WNV, and measles virus) in mice (25–30). In addition, patients with mutations in type I, II, or III IFNs are susceptible to HSV encephalitis and mycobacterial infections (31). To determine whether IFN signaling affects MuPyV replication in the brain, wild-type (WT) mice, mice deficient in receptors for type I IFNs (IFNAR^−/−), type I and III IFNs (IFNLR1^−/− IFNAR^−/−), and type II IFNs (IFN-γR^−/−) and mice deficient in STAT1 (STAT1^−/−) were infected i.c. with MuPyV. During acute (day 8 postinfection [p.i.]) MuPyV infection, STAT1^−/− mice had significantly higher viral load in the brain than WT, IFN-γR^−/−, IFNAR^−/−, and IFNLR1^−/− IFNAR^−/− mice (Fig. 1A). In addition, IFNAR^−/− and IFNLR1^−/− IFNALR^–/– mice had similar viral loads, suggesting that type I IFNs are sufficient in controlling MuPyV infection in the brains of IFNLR1^−/− IFNALR^–/– mice (Fig. 1A). Similarly, type I IFN contributed to viral control in the spleen during acute infection, as demonstrated by significantly higher viral loads in the spleens of IFNAR^−/− and IFNLR1^−/− IFNALR^–/– mice than in WT mice (Fig. 1B). During persistent viral infection (day 30 p.i.), STAT1^−/− mice had significantly higher viral loads in the brain than did WT, IFNAR^−/−, IFN-γR^−/−, and IFNLR^−/− IFNAR^–/– mice (Fig. 1C). STAT1^−/− mice did not have higher viral loads in the spleen (Fig. 1D), suggesting that STAT1 does not mediate viral control systemically. It merits noting that the apparently dispensable need for type II IFNs to control MuPyV encephalitis contrasts with our previous work showing an increased MuPyV load in the kidney during persistent infection in IFN-γR^−/− mice (23). These results suggest that IFN-γ differentially controls infections in a tissue context-dependent fashion, a possibility supported by evidence of heterogeneity in levels of IFN-γR expression in different tissues (32). Taken together, these results indicate that STAT1-mediated signaling exerts anti-MuPyV control in the brain during acute and persistent infection.

FIG 1.

MuPyV load in the brains during acute and persistent infection. (A to D) Real-time PCR analysis of viral genome copies in brains and spleens at days 8 (A and B) and 30 (C and D) p.i. Means ± the SD of 7 to 15 mice per group from three to four independent experiments are shown. *, P < 0.05; **, P < 0.01 (one-way ANOVA with Tukey’s multiple-comparison test and Grubb’s test).

STAT1 signaling protects against brain pathology during MuPyV encephalitis.

Aberrations in the homeostatic balance of type I, II, and III IFNs exacerbate progression of many brain diseases in humans, including Alzheimer’s disease and progressive and relapsing-remitting multiple sclerosis, due to altered immune surveillance of the brain (33, 34). To determine whether loss of IFNs affected MuPyV pathology, we performed histopathological analyses on the brains and assayed blood-brain barrier (BBB) permeability. MuPyV-infected STAT1^−/− mice had significantly increased lateral ventricle size compared to infected WT mice (Fig. 2A to C). Using LFB staining intensity as a readout of demyelination, we found no change in demyelination in the cingula, external capsules, and dorsal hippocampal commissures of MuPyV-infected WT, IFNAR^−/−, IFNLR1^−/− IFNAR^−/−, IFN-γR^−/−, and STAT1^−/− mice compared to sham-inoculated mice (dotted line) (Fig. 2D). Recognizing that differences in hydrocephalus can reflect deficits in endothelial barrier integrity and that IFNs have been shown to regulate BBB integrity in other mouse-viral CNS infection models (27, 35, 36), we next quantified permeability of the BBB by extravasation of intraperitoneally (i.p.) administered sodium fluorescein dye into the brain at day 9 p.i. The BBB permeability was higher in infected IFNAR^−/−, IFNLR1^−/−IFNAR^−/−, and STAT1^−/− mice than in infected WT mice and in sham-injected mice, demonstrating that type I (and potentially type III) IFNs and STAT1 signaling are required to sustain BBB integrity during acute MuPyV infection (Fig. 2E). Notably, IFN-γ was dispensable for maintenance of BBB integrity because IFN-γR^−/− mice had fluorescein accumulation in the brain similar to that observed in infected WT and sham-inoculated mice (Fig. 2E). Interestingly, we found that MuPyV preferentially binds cells lining the ventricles (Fig. 2F), which is consistent with reports that JCPyV binds the choroid plexus and vascular endothelium on human brain sections (37). Together, these results indicate that type I/III IFNs and STAT1 signaling are important for maintaining BBB integrity and limiting the development of hydrocephalus during MuPyV encephalitis.

FIG 2.

STAT1 signaling affects MuPyV-induced brain pathology. (A and B) Representative LFB-PAS images of sham-inoculated (A) and MuPyV-inoculated (B) WT, IFNAR^−/−, IFNLR^−/− IFNAR^−/−, IFN-γR^−/−, and STAT1^−/− mouse brains obtained –2.70 mm caudal to bregma at day 30 p.i. Due to the similarities between the sham inoculated brains of each genotype (see panel A), we grouped the sham-inoculated mice together for further analyses. (C) Percent of lateral ventricle volume from brain sections obtained –2.70 mm caudal to bregma at day 30 p.i. The dotted line refers to sham-inoculated mice combined from all groups. (D) Percent area myelinated in the cingulum, dorsal hippocampal commissure, and external capsule obtained –2.70 mm caudal to bregma at day 30 p.i. (E) BBB permeability was measured 9 days p.i. by the accumulation of sodium fluorescein dye in the brain. Data are normalized to the lipopolysaccharide-positive control, and the serum fluorescein levels in individual mice are shown. (F) MuPyV (red) binding to cells lining the lateral ventricle, DAPI (blue). Treatment with neuraminidase prevents virus binding (bottom panel). Means ± the SD of 5 to 12 mice per group from two to three independent experiments are shown. *, P < 0.05; ***, P < 0.005; ****, P < 0.001 one-way ANOVA compared to WT with Tukey’s multiple-comparison test.

Astrocyte-intrinsic STAT1 signaling does not affect viral load or encephalopathy.

In human glia chimeric mice, JCPyV was found to productively infect astrocytes (38). We recently reported that astrocytes isolated from brains of MuPyV-infected mice express early region viral transcripts (39). Using GFAP-Cre STAT1^fl/fl mice, we sought to determine whether astrocyte-intrinsic STAT1 signaling conferred protection against MuPyV-induced CNS disease (40). Histological analyses showed that infected WT and GFAP-Cre STAT1^fl/fl mice had similar ventricle volumes and LFB intensities in the cingulum, external capsule, and dorsal hippocampal commissure (Fig. 3A to C). WT and GFAP-Cre STAT1^fl/fl mice also had similar viral loads in the brain and periphery (Fig. 3D and E). These data suggest that STAT1 signaling in astrocytes per se fails to control MuPyV infection or brain pathology.

FIG 3.

STAT1 deficiency in astrocytes does not affect brain pathology. (A) Representative LFB-PAS images of sham and MuPyV-inoculated WT and GFAP-cre STAT1^fl/fl mice obtained –2.70 mm from bregma at day 30 p.i. The asterisk indicates cutting a artifact. Data are representative of three independent experiments. (B) Percent of lateral ventricle in total brain from brain sections obtained –2.70 mm caudal to bregma at day 30 p.i. The dotted line refers to sham-inoculated mice. (C) Percent area myelinated in the cingulum, dorsal hippocampal commissure, and external capsule –2.70 mm caudal to bregma at day 30 p.i. (D and E) Real-time PCR analysis of viral genome copies in brain (left) and spleen (right) tissues at day 30 p.i. Means ± the standard errors of the mean (SEM) of three mice per group from two independent experiments are shown. All data were determined to be not significant according to one-way ANOVA with Tukey’s comparison test.

Mice deficient in STAT1 have increased neuroinflammation.

Previous work has shown that a loss of STAT1 causes aberrant inflammation (41). Blinded histopathologic evaluation by a veterinary pathologist of WT and STAT1^−/− mouse brains revealed that STAT1^−/− mice had significantly increased inflammation and neurodegeneration than sham-inoculated mice (Fig. 4A and B). Because STAT1^−/− mice had extensive hydrocephalus and significantly higher viral loads, we sought to determine whether the brain pathology may be linked to excessive immune activation. At day 30 p.i., STAT1^−/− mice, had increased numbers of CD11b^hi CD45^hi Ly6G⁺ Ly6C⁺ neutrophil-like and CD11b^hi CD45^hi Ly6G^– Ly6C^hi monocyte-like cells in the brain compared to WT mice during persistent viral infection (Fig. 4C to E). At day 30 p.i., blinded analysis by a veterinary pathologist revealed that WT mice had low-level infiltration of monocytes and lymphocytes into the brain parenchyma (Fig. 4F). STAT1^−/− mouse brains had increased inflammatory cell infiltrates within and surrounding the lateral ventricles, dorsal third ventricle, hippocampus, and dentate gyrus at 30 days p.i. (Fig. 4F). STAT1^−/− mice also had focal loss of ependymal lining at sites of myeloid cells infiltration (Fig. 4F). Myeloid cells, such as neutrophils, have been implicated in the pathogenesis of various neurodegenerative diseases, suggesting that the increased infiltration of myeloid cells may exacerbate pathology in STAT1^−/− mice (42).

FIG 4.

STAT1 deficiency increases neuroinflammation and accumulation of innate immune cells in the brain during persistent infection. (A and B) Mouse brains were rated for the degree of inflammation (A) or neurodegeneration (B) by a pathologist blinded to the identity of the experimental groups. Brains were removed at 30 days p.i. Means ± the SD of two to six mice per group from two independent experiments are shown. (C and D) Numbers of Ly6G⁺ Ly6C⁺ (C) and Ly6G^– Ly6C⁺ (D) cells at day 30 p.i. (E) Gating strategy for cells in C and D. (F) Representative H&E-stained sections (400× original magnification) from mock- and MuPyV-inoculated WT and STAT1^−/− mice at day 30 p.i. Means ± the SD of five to eight mice per group from three independent experiments are shown. *, P < 0.05; **, P < 0.01; ***, P < 0.005 (one-way ANOVA with Tukey’s multiple-comparison test).

Brain-resident cells, such as microglia and astrocytes, can also contribute to an inflammatory state during viral infection. STAT1 activation by IFN signaling promotes the formation of proinflammatory, M1-like microglia during insult, suggesting that a deficiency of STAT1 may induce an anti-inflammatory phenotype in microglia (43). STAT1^−/− mice, however, had more intense Iba1 staining and increased numbers of Iba1⁺ cells in the brain, suggesting that cells of macrophage lineage, including microglia, are activated (Fig. 5A and B to E). This effect seems to be microglia-specific because there was no change in the number of GFAP⁺ cells or the intensity of staining (Fig. 5C to E). Thus, STAT1 deficiency promotes infiltration of myeloid immune cells and persistent activation of microglia and/or macrophages, further demonstrating increased damaging neuroinflammation with MuPyV infection in the absence of STAT1.

FIG 5.

STAT1^−/− mice have activated microglia in the brain during persistent MuPyV infection. (A and B) Intensity (A) and number (B) of Iba1⁺ cells at day 30 p.i. (C and D) Intensity (C) and number (D) of GFAP⁺ cells at day 30 p.i. (E) Representative images of Iba1 (top) and GFAP (bottom) sham- and MuPyV-inoculated WT and STAT1^−/− mice at 30 days p.i. Means ± the SD of five to eight mice per group from two independent experiments are shown. **, P < 0.01; ***, P < 0.005; ****, P < 0.001 (one-way ANOVA with Tukey’s multiple-comparison test).

STAT1 deficiency results in elevated accumulation of functional CD8 T cells during MuPyV encephalitis.

Previous studies have shown that STAT1 can paradoxically both orchestrate an anti-inflammatory signaling program through activation of suppressor of cytokine signaling and promote proinflammatory effects on innate and adaptive immune response (12, 44). Our previous work has shown that the adaptive immune response, particularly the CD8 T cell response, is an important component of MuPyV control (45, 46). We found that STAT1^−/− mice persistently infected with MuPyV had increased numbers of brain-infiltrating CD8 T cells (Fig. 6A). However, the frequency of CD8T cells specific for of D^bLT359, the immunodominant CD8 T cell epitope in B6 mice (45), was comparable in STAT1^−/− and WT mice (Fig. 6B), suggesting nonspecific CD8 T cells are recruited to the brain in the absence of STAT1 signaling. A similar frequency of brain-infiltrating CD8 T cells in persistently infected WT and STAT1^−/− mice produced IFN-γ and tumor necrosis factor alpha (TNF-α) and exhibited degranulation in response to LT359 peptide stimulation (Fig. 6C).

FIG 6.

STAT1 signaling limits CD8 T cell accumulation during MuPyV encephalitis. (A) Numbers of total CD8 T cells from brains at day 30 p.i. (B) Frequency of D^bLT359-specific CD8 T cells from the brain at day 30 p.i. (C) Frequencies of IFN-γ⁺, IFN-γ⁺ CD107⁺, and IFN-γ⁺ TNF-α⁺ CD8 T cells from the brain at day 30 p.i. after ex vivo stimulation with LT359 peptide. (D) Frequency of CD103⁺ D^bLT359-specific CD8 T cells in the brain at day 30 p.i. (E) CD69 geometric mean fluorescence intensity (gMFI) on D^bLT359-specific CD8 T cells from brain at days 8 and 30 p.i. Means ± the SD of eight to eleven mice per group from two to three independent experiments are shown. **, P < 0.01; ***, P < 0.005; ****, P < 0.001 (one-way ANOVA with all groups compared to WT and Tukey’s multiple-comparison test).

We next sought to determine whether the formation of resident memory CD8 T cells (T_RM) was intact. CD103 and CD69 are regarded as the canonical surface markers of CD8 T_RM, and we have previously shown that 80 to 95% and 40% of D^bLT359-specific CD8 T cells in the brain express CD69 and CD103, respectively, during persistent MuPyV infection (47–50). We found that persistently infected mice deficient in STAT1 signaling had similar frequencies of CD103⁺, but fewer CD69^hi D^bLT359-specific CD8 T cells in their brains (Fig. 6D and E). Previous work has shown that CD69 is required for the development of T_RM, suggesting that STAT1^−/− mice have altered brain T_RM (bT_RM) development (51, 52). Taken together, these results suggest that STAT1 limits accumulation of functional CD8 T cells that express the T_RM phenotype.

CD8 T cells prevent MuPyV lethality in STAT1^–/– mice.

CD8 bT_RM represents an autonomous barrier against reinfection and excessive immune activation in the brain (53). Moreover, virus-specific CD8 T cells are a dominant source of IFN-γ, which, while exerting antiviral activity, is also injurious to CNS-resident cells (54). To address the discrepancy between high viral loads and worse brain pathology in STAT1^−/− mice, despite the increased numbers of functional CD8 T cells in the brain during persistent infection, we depleted CD8 T cells from STAT1^−/− mice prior to infection in order to analyze the contribution of CD8 bT_RM and their effector function to neuropathology and viral control after i.c. MuPyV infection. CD8 T cell depletion caused significantly more weight loss in STAT1^−/− mice (Fig. 7A). This weight loss was not due solely to loss of CD8 T cells in STAT1^−/− mice because sham-inoculated mice did not lose weight following anti-CD8β MAb-mediated depletion (data not shown). Strikingly, MuPyV infection proved lethal for the CD8 T cell-depleted STAT1^−/− mice with 100% mortality by 2 weeks p.i. (Fig. 7B). At terminal endpoints, STAT1^−/− mice had dramatically increased viral loads in the brain and spleen in the absence of CD8 T cells (data not shown). This result was confirmed in a separate cohort of mice intentionally euthanized at day 11 p.i. (Fig. 7C). Histological examination of brains at this time point showed enhanced VP1 staining in the cells lining the ventricles, denoting lytic viral infection (Fig. 7D). The ependymal cells lining the lateral and third ventricles were vacuolated and necrotic, with pyknotic nuclei. STAT1^−/− mice lacking CD8 T cells also had dense VP1⁺ foci located in the internal capsule that were absent in WT mice and STAT1^−/− mice with an intact CD8 T cell response (Fig. 7D). The leptomeninges in STAT1^−/− mice lacking CD8 T cells were also infected and contained more infiltrating mononuclear cells (Fig. 7E). A moderate level of hydrocephalus was also appreciated in STAT1^−/− mice lacking CD8 T cells, suggesting that damage to the ependyma as a result of MuPyV infection leads to terminal viral pathology. These results demonstrate that CD8 T cells and STAT1 signaling together provide essential antiviral defense to polyomavirus CNS infection.

FIG 7.

T cells are protective in STAT1^−/− mice against lethal MuPyV infection. (A) Percent body weight of STAT1^−/− and WT mice after rat IgG or αCD8 treatment on days 4 and 1 prior to MuPyV infection and weekly thereafter. (B) Survival curve of STAT1^−/− and WT mice after MuPyV infection. (C) Real-time PCR analysis of viral genome copies in spleen at day 11 p.i. (D and E) VP1 IHC with hematoxylin counterstain of indicated groups in the lateral ventricles (D) and meninges (E) at day 11 p.i. Scale bars represent 100 μm (50×) or 25 μm (200×). Means ± the SD of three to five mice per group from two to three independent experiments are shown.

DISCUSSION

Using the MuPyV infection model, we demonstrate that signaling via STAT1 prevents virus-induced CNS injury and controls viral infection and does so in concert with CD8 T cells. Although mice deficient in type I IFN, type II IFN, or type I and III IFN receptors developed MuPyV-induced pathology, they did so equivalently to that of WT mice. STAT1 deficiency, however, resulted in significant hydrocephalus and elevated viral loads in the brain. Mice deficient in STAT1 mounted a sustained inflammatory response to MuPyV CNS infection consisting of brain-infiltrating inflammatory myeloid cells, T cells, and activated microglia. CD8 T cells on their own did not contribute significantly to controlling CNS virus loads or neuropathology, but depletion of CD8 T cells in STAT1^−/− mice caused dramatically higher viral infection, profound hydrocephalus and complete mortality within 2 weeks of inoculation. Together, these findings support the conclusion that STAT-1 signaling and CD8 T cells act as key comediators of protection against MuPyV CNS infection and pathology.

We previously reported that IFN-γ controlled viral burden in the kidney; however, IFN-γ was found here not to contribute to viral control in the brain (Fig. 1) (23). These results reinforce the concept that the type of immune mechanisms brought to bear to control viral infections may be a function of infected organs and tissues. For viral clearance in the brain, CD8 T cells may engage cytolytic (e.g., perforin-granzyme exocytosis, Fas) and noncytolytic (e.g., IFN-γ) effector mechanisms. Ample evidence supports the idea that effector mechanism(s) in the brain, an organ replete with sizeable populations of non/poorly renewable cells, may be intimately linked to host cell tropism and viral pathogen life cycle (e.g., productive-lytic, productive-noncytopathic, and latent), such as perforin-mediated killing of vesicular stomatitis-infected neurons (55), and perforin- and IFN-γ-dependent control of lymphocytic choriomeningitis virus (LCMV)-infected meninges and ependyma (53, 56, 57), as well noncytopathic clearance of LCMV-infected microglia (58). Our finding that the control of MuPyV encephalitis is codependent on STAT1 signaling and CD8 T cells speaks to an essential interplay between innate and adaptive immune responses in the PyV-infected CNS.

In contrast to WT mice, cells of macrophage lineage, which includes microglia, were chronically activated in the brains of STAT1^−/− mice during persistent MuPyV infection (Fig. 5). Microglia express numerous Toll-like receptors (TLRs) and pattern recognition receptors, which allows them to respond rapidly and efficiently to CNS viral infection (59). In this connection, Ingenuity Pathway analyses of NanoString assay gene expression inflammation panels indicate that TLR signaling pathways are upregulated during MuPyV encephalitis (39). After stimulation by pathogens, microglia secrete proinflammatory factors such as glutamate, reactive oxygen species, cytokines, and chemokines that activate other glial cells, such as astrocytes, and mobilize the peripheral immune response (60). Although glial cell activation has beneficial effects, such as pathogen clearance, uncontrolled and persistent activation can exacerbate brain damage via sustained production of inflammatory, neurotoxic mediators and protracted immune activity (59). Thus, the activation of microglia in STAT1^−/− mice may create a positive-feedback loop that perpetuates brain-resident and immune cell activation and increases brain pathology.

The pronounced hydrocephalus in MuPyV-infected STAT1^−/− mice (Fig. 2) raises the possibility that signaling via STAT1 engages antiviral defense mechanisms in ependymal cells. In this connection, meningeal cells and primary human choroid plexus epithelial cells, a type of modified ependymal cell, support productive JCPyV infection (37, 61), and JCPyV large T antigen- and VP1 capsid-expressing choroid plexus cells are detected in brain tissue sections from PML patients (62). Our work further supports these findings and additionally suggests that CD8 T cells hold ependymal polyomavirus infection in check in the absence of STAT1 signaling. Ependymal cells are a ciliated and terminally differentiated single cell layer of cuboidal epithelial cells that line each brain ventricle, as well as the central canal of the spinal cord; the adherens and tight junctions of ependymal cells constitute the brain parenchyma-cerebrospinal fluid (CSF) barrier (63, 64). The tight apposition of vasculature and modified ependyma in the choroid plexus may provide a conduit for virus to transit from the blood to the CSF, with subsequent productive infection of ependyma as a beachhead to invade the brain parenchyma. Loss of endothelial barrier integrity in STAT1^−/− mice (Fig. 2 and Fig. 7) is associated with increased MuPyV viral load, which provides further support that MuPyV potentially enters the brain via the hematologic route by crossing the blood-CSF barrier, as proposed for JCPyV CNS infection (37).

Disruption of ependymal cilia or adhesion by brain insults such as trauma or viral infection can cause hydrocephalus (63). For example, infection of ependymal cells by HSV-1 increases hydrocephalus and ependymal cell loss (65). Alternatively, the localization of immune cells at the ependymal barrier suggests that the hydrocephalus in STAT1^−/− mice may also be a consequence of the antiviral immune response. Hydrocephalus has also been associated with abnormal neurogenesis, most likely resulting from the loss of trophic support from ependymal cells to progenitor cells in the subventricular zone (SVZ) (63). These pluripotent progenitor cells of the SVZ can differentiate into neurons and oligodendrocyte precursor cells, and their loss has been shown to decrease remyelination in inflammatory conditions (55). Previous work has shown that ependymal cells can respond to type I and II IFNs in vitro (64). Despite the predominant role of ependymal cells in brain damage during viral infection, the intrinsic antiviral defenses employed by ependymal cells remain largely unknown. Data presented here suggests that a loss of STAT1 contributes to a loss of integrity in ependymal cells and the development of extensive hydrocephalus.

Despite the large number of functional CD8 T cells in the brains of MuPyV-infected mice, STAT1 deficiency rendered mice susceptible to viral-brain pathology, underscoring the importance of the combination of intact CD8 T cell responses in conjunction with STAT1 signaling in controlling viral replication and pathology. STAT1 deficiency results in an unrestrained immune response, especially by neutrophils and CD8 T cells, in a number of different viral infections (41, 66–70). Entry of neutrophils and CD8 T cells in the brain is a hallmark of neuroinflammation (71). The infiltration of neutrophils contributes to demyelination and cell loss in the CNS, as demonstrated by decreased cuprizone-induced demyelination and picornavirus-induced hippocampal pathology in the absence of neutrophils (42, 72–74). Neutrophils also facilitate recruitment of virus-specific lymphocytes into the infected CNS via permeabilization of the BBB (75). CD8 T cells contribute to CNS demyelination in response to infection with the JHM strain of mouse hepatitis virus (76). Alternatively, CD8 T cells, especially CD8 bT_RM, provide lasting immune surveillance of the CNS (53, 77). Collectively, these studies implicate STAT1 signaling functions as a homeostatic immune checkpoint for immune cell CNS infiltration.

In summary, we found that an intact CD8-STAT1 axis is necessary to confer protection against MuPyV encephalitis and encephalopathy. Moreover, motor, behavioral, and cognitive studies would be expected to uncover further deficits in MuPyV-infected STAT1^−/− mice. Our findings raise the provocative idea that fortifying STAT1 signaling may bolster CD8 T cell-mediated control of PyV infections in the brain.

MATERIALS AND METHODS

Mice.

Adult (6- to 12-week-old) female and male C57BL/6 (B6) mice were purchased from the National Cancer Institute (NCI, Frederick, MD). IFNLR1^−/− IFNAR^−/− mice were obtained from Sergei Kotenko (Rutgers New Jersey Medical School). IFNAR^−/− and IFN-γR^−/− mice were obtained from Ziaur Rahman (Penn State College of Medicine). STAT1^−/− mice were obtained from Christopher Norbury (Penn State College of Medicine) with an approved materials transfer agreement from Rutgers New Jersey Medical School. All knockout mice are on a B6 background. Mice were bred and housed in accordance with the guidelines of the NIH Guide for the Care and Use of Laboratory Animals and the Institutional Animal Care and Use Committee at the Penn State College of Medicine. For survival studies, mice were euthanized when their body weight declined >20%, per approved IACUC protocols.

Viruses and infections.

MuPyV.A2 was prepared in baby mouse kidney cells as described previously (78). Mice were inoculated i.c. with 3 × 10⁵ PFU MuPyV.A2 in 30 μl as described previously (45).

CD8 T cell depletion.

Mice were injected i.p. with 250 μg of rat anti-CD8β or ChromoPure whole rat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) beginning 4 days before infection and continuing weekly. Depletion was confirmed in peripheral blood by flow cytometry-based cell number assay using Absolute Count Standard (Bangs Laboratories, Fishers, IN).

T cell isolation and flow cytometry.

Mononuclear cells were isolated from brains by collagenase-DNase digestion and Percoll gradient centrifugation as described previously (45). Mononuclear cells were isolated from spleen as described previously (79). After isolation, cells were stained with fixable viability dye (eBioscience, San Diego, CA), APC-D^bLT359 tetramers (NIH Tetramer Core Facility, Atlanta, GA), and the following surface antibodies: CD8α (clone 53-6.7; eBioscience), CD44 (clone IM7; eBioscience), PD-1 (clone RMPI-30; BioLegend), CD103 (clone M290; BD Horizon), CD69 (clone HI.2F3; BioLegend), CD127 (clone A7R34; BioLegend), KLRG1 (clone 2F1; BD Biosciences), Ly6C (clone HK1.4; BioLegend), Ly6G (clone 1A8;BioLegend), MHC II (clone M5/114 15.2; eBioscience), CD11b (clone M1/70; BD Biosciences), streptavidin (BioLegend), and CD45 (clone 30-F11; BD Biosciences). In addition, cells were stained with the following biotinylated surface antibodies: Thy1.2 (clone 53-21; eBioscience), CD19 (clone 6D5; BioLegend), and NK1.1 (clone PK136; BioLegend). For intracellular cytokine stimulation, assays were performed as described previously (80). Intracellular staining included IFN-γ (clone XMG1.2; BioLegend), TNF-α (clone XMG1.2; BioLegend), CD107a (clone 1D4B; BD Biosciences), CD107b (clone ABL-93; BD Biosciences). Samples were acquired on a BD LSRFortessa (BD Bioscience, San Jose, CA) or BD LSR II (BD Biosciences) and analyzed using FlowJo software (FlowJo, LLC, Ashland, OR).

Quantitative PCR.

For quantifying viral genome DNA copies, real-time PCR was performed on samples, as described previously (46).

Histological processing and analysis.

Mice were sedated using ketamine and immediately perfused with 10 ml of 10% heparin in phosphate-buffered saline (PBS), followed by 10 ml of 10% neutral buffered formalin (NBF). The heads were placed in 10% NBF overnight, and the brains were excised the next day. The brain was sectioned into 7-μm-thick slices using a standard microtome by the Comparative Medicine Histology Core. Formalin-fixed paraffin embedded sections (FFPE) were stained with Luxol Fast Blue-Periodic acid-Schiff (LFB-PAS) for visualization of myelin fibers and hematoxylin and eosin (H&E) for overall histopathologic characteristics as described previously (81). LFB-PAS-stained sections were digitally imaged on a Keyence BZ-X710 all-in-one fluorescence microscope and stitched together using ImageJ software (National Institutes of Health, Bethesda, MD). Analysis of myelin was performed and analyzed as described previously (82) or by using the LFB-PAS deconvolution function in ImageJ. In brief, total brain sections were deconvoluted and the density of blue staining was quantified for the entire brain section. For ventricle size, the pixel area of the left and right lateral ventricles and dorsal third ventricle was divided by the total pixel area of the brain section or the cortex using Adobe Photoshop (San Jose, CA). Ventricle size was then expressed as a percentage of the total brain area or cortex area. For immunohistochemistry (IHC), 10-μm FFPE brain sections were deparaffinized and rehydrated prior to antigen retrieval in 10 mM sodium citrate buffer (pH 6.0). Sections were permeabilized with 1% Triton X-100, stained with primary antibodies anti-GFAP (Dako, Carpinteria, CA) or anti-Iba (Dako) for 1 h at room temperature or anti-VP1 (I58; gifted by Bob Garcea, University of Colorado Boulder, Boulder, CO) overnight at 4°C and then stained with secondary biotinylated goat anti-rabbit (Vector, Burlingame, CA) for 1 h at room temperature, followed by avidin-conjugated horseradish peroxidase (Vectastain Elite ABC kit; Vector, Burlingame, CA). Staining was developed using the Vector NovaRED peroxidase substrate kit (Vector).

Brain and meningeal inflammation, and neurodegeneration were evaluated by a pathologist (H. M. Atkins) blinded to treatment groups. Bregma stereotactic coordinates were used to group brains sections to within 1 mm of each other. Severity of inflammation and neurodegeneration was scored (0 to 4) using the following criteria: 0, no deviation from normal; 1, a slight change that barely exceeds normal limits; 2, identified but limited in severity; 3, prominent but still potential for increased severity; and 4, a lesion that completely occupies a region of tissue. Microglia and astrocyte numbers were evaluated based on relative prevalence of positive cells within 100 μm surrounding the lateral and third ventricles, with 0 indicating no positive cells and 4 representing too many cells to count. Immunohistochemistry relative staining intensity was also scored using a four-point scale (0 to 4), with 0 corresponding to no staining and 4 to the most profound staining.

For virus binding to brain sections, mice were perfused with 10 ml of 10% heparin in PBS, followed by 10 ml of 4% paraformaldehyde. Brains were postfixed in 4% paraformaldehyde for 6 h and then sucrose dehydrated in 30% sucrose. Then, 12-μm sections of brain and spleen were taken on a Leica biosystem cryostat (model CM1850; Buffalo Grove, IL). Sections were treated with type II neuraminidase (Sigma-Aldrich, St. Louis, MO) at a 1/200 dilution in PBS supplemented with 1 mM CaCl₂ and 1 mM MgCl₂ or with buffer alone at 37°C for 30 min. Sections were incubated with virus at 1 × 10⁷ PFU/ml for 1.5 h at room temperature. Virus was stained with antibody to VP1 (I58) and detected with an Alexa Fluor 647-conjugated secondary antibody (Jackson ImmunoResearch, West Grove, PA). Images were acquired on a Leica DM4000 fluorescence microscope.

Blood-brain barrier permeability assay.

BBB permeability was measured as described previously (27, 36).

Statistical analysis.

Experimental data were analyzed using Prism 6.07 (GraphPad, La Jolla, CA) and Mann-Whitney tests, one-way analysis of variance (ANOVA), and two-way ANOVA with Tukey or Sidak’s multiple-comparison test. Outliers were excluded using Grubb’s test. Error bars indicate means ± the standard deviations (SD). All experiments were replicated independently.

Data availability.

All data generated or analyzed in this study are included in this published article.