Astrocyte-Derived CXCL10 Drives Accumulation of Antibody-Secreting Cells in the Central Nervous System during Viral Encephalomyelitis

Abstract

Microbial infections of the central nervous system (CNS) are often associated with local accumulation of antibody (Ab)-secreting cells (ASC). By providing a source of Ab at the site of infection, CNS-localized ASC play a critical role in acute viral control and in preventing viral recrudescence. Following coronavirus-induced encephalomyelitis, the CNS accumulation of ASC is chemokine (C-X-C motif) receptor 3 (CXCR3) dependent. This study demonstrates that CNS-expressed CXCR3 ligand CXCL10 is the critical chemokine regulating ASC accumulation. Impaired ASC recruitment in CXCL10^−/− but not CXCL9^−/− mice was consistent with reduced CNS IgG and κ-light chain mRNA and virus-specific Ab. Moreover, the few ASC recruited to the CNS in CXCL10^−/− mice were confined to the vasculature, distinct from the parenchymal localization in wild-type and CXCL9^−/− mice. However, neither CXCL9 nor CXCL10 deficiency diminished neutralizing serum Ab, supporting a direct role for CXCL10 in ASC migration. T cell accumulation, localization, and effector functions were also not affected in either CXCL9^−/− or CXCL10^−/− mice, consistent with similar control of infectious virus. There was also no evidence for dysregulation of chemokines or cytokines involved in ASC regulation. The distinct roles of CXCL9 and CXCL10 in ASC accumulation rather coincided with their differential localization. While CXCL10 was predominantly expressed by astrocytes, CXCL9 expression was confined to the vasculature/perivascular spaces. These results suggest that CXCL10 is critical for two phases: recruitment of ASC to the CNS vasculature and ASC entry into the CNS parenchyma.

INTRODUCTION

The site of infection and access of neutralizing antibody (Ab) contribute to viral control during both primary and secondary infections. This is particularly important in the central nervous system (CNS), which is a target for numerous acute encephalitic viral infections, as well as a reservoir of latent/persistent infections (1). Sequestration from the circulation by the specialized structure of the blood brain barrier makes Ab passage from the serum to the CNS inefficient. While control of acute neurotropic infections of the CNS is largely T cell mediated, local Ab, produced by Ab-secreting cells (ASC), is required for ultimate clearance and/or prevention of viral recrudescence (2–5). Continuous intrathecal Ab production thus provides an effective nonlytic mechanism of sustained viral control within the CNS. This is supported by complement-independent, Ab-mediated control of mouse hepatitis virus (MHV), Sindbis virus, and rabies virus (6–8). The CNS as a survival niche for ASC is well established in experimental animal models, as well as humans. However, the mechanisms of ASC accumulation and replenishment may vary in distinct inflammatory conditions and involve both direct ASC migration from the periphery and ectopic follicle formation (5, 9–12).

As B cells differentiate into ASC and emigrate from lymphoid follicles, they upregulate the surface expression of chemokine (C-X-C motif) receptor 4 (CXCR4) and CXCR3. While CXCR4 interactions with its ligand, CXCL12, mediate homing to and survival of ASC in the bone marrow (13), CXCR3 signaling is associated with ASC migration to inflamed tissue, including the CNS (14). The CXCR3 ligands CXCL9, CXCL10, and CXCL11 are strongly induced in the CNS following viral infection, as well as during nonmicrobial inflammation (15–22). An essential role for CXCR3 in ASC accumulation in the CNS was recently demonstrated using CXCR3-deficient (CXCR3^−/−) mice infected with gliatropic as well as dual hepato- and neurotropic strains of coronavirus (23, 24). Both virus strains cause acute encephalomyelitis that resolves into a persistent infection associated with chronic, ongoing immune-mediated myelin loss (23, 25). Although control of infectious virus in the CNS is T cell mediated (26–28), continuously produced antiviral IgG within the CNS is required to prevent viral recrudescence (4, 24, 29). CXCR3 deficiency did not impair T cell-mediated antiviral activity or serum neutralizing Ab but did severely abrogate virus-specific ASC accumulation in the CNS (24).

The present study sought to define the relative contributions of CXCR3 ligand(s) CXCL9 and CXCL10 to ASC accumulation and regional localization. During infection with the gliatropic JHM variant V2.2-1 of mouse hepatitis virus (JHMV), CXCL10 and, to a lesser extent, CXCL9 mRNA expression is sustained after initial viral control (24, 30), although both peak concomitantly with peak T cell infiltration and gamma interferon (IFN-γ) secretion (31, 32). Neither CXCL9^−/− nor CXCL10^−/− JHMV-infected mice exhibited deficits in virus-specific serum Ab or CNS T cell activity, similar to CXCR3^−/− mice (24). However, both ASC and virus-specific IgG within the CNS were significantly impaired in CXCL10^−/− but not CXCL9^−/− mice. Furthermore, the few ASC recruited to the CNS in CXCL10^−/− mice were confined to the vasculature, distinct from their parenchymal distribution in wild-type (WT) and CXCL9^−/− mice. The parenchymal accumulation of ASC in WT mice was directly linked to the predominant CXCL10 expression by astrocytes, while CXCL9 expression was confined to the microvasculature. These results demonstrate a prominent, nonredundant role of CXCL10 in both ASC recruitment and entry into the inflamed CNS parenchyma during viral encephalomyelitis.

MATERIALS AND METHODS

Mice, virus infection, and virus titers.

C57BL/6 mice were purchased from the National Cancer Institute (Frederick, MD). CXCL9^−/−, CXCL10^−/−, and CXCR3^−/− mice on the C57BL/6 background were previously described (24, 33). All mice were housed under pathogen-free conditions at an accredited facility at the Cleveland Clinic Lerner Research Institute. Mice were infected at 6 to 7 weeks of age by intracranial injection with 1,000 PFU of the gliatropic JHMV variant (34). Animals were scored daily for clinical signs of disease as follows: 0, healthy; 1, ruffled fur and hunched back; 2, hind limb paralysis or inability to turn to upright position; 3, complete hind limb paralysis and wasting; and 4, moribund or dead. All animal experiments were performed according to guidelines approved by the Cleveland Clinic Lerner Research Institute Institutional Animal Care and Use Committee. Virus titers within the CNS were determined in clarified supernatants by plaque assay using the murine delayed brain tumor (DBT) astrocytoma as described previously (34). Plaques were counted after 48 h of incubation at 37°C.

Gene expression and cytokine analysis.

Snap-frozen spinal cords from individual mice (n = 3 to 9) were placed into 1 ml TRIzol (Invitrogen, Grand Island, NY) and homogenized using a TissueLyser and stainless steel beads (Qiagen, Valencia, CA). RNA was extracted according to the manufacturer's instructions. DNA contamination was removed by treatment with DNase I for 30 min at 37°C (DNA-free kit; Ambion, Austin, TX), and cDNA was synthesized from RNA using Moloney murine leukemia virus (MMLV) reverse transcriptase (Invitrogen), oligo(dT) primers, and random primers (Promega, Madison, WI). Quantitative real-time PCR was performed using 4 μl of cDNA and SYBR green master mix (Applied Biosystems, Foster City, CA) in duplicate on a 7500 fast real-time PCR system (Applied Biosystems). PCR conditions were 10 min at 95°C followed by 40 cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s. Real-time primers for transcripts encoding a proliferating-inducing ligand (APRIL), B cell maturation antigen (BCMA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), JHMV nucleocapsid, interleukin-10 (IL-10), IL-21, transmembrane activator and calcium modulator ligand interactor (TACI), CXCL9, CXCL10, CXCL11, and CXCL12 were previously described (30, 35). GAPDH, IFN-γ, CXCL13, B cell-activating factor of the tumor necrosis factor (TNF) family (BAFF), CCL19, CCL21, and IgG mRNA levels were determined using Applied Biosystems gene expression arrays with universal TaqMan fast master mix (Applied Biosystems) in duplicate. Primer and probe sequences for κ-light chain mRNA detection were as described previously (10). The PCR conditions were 20 s at 95°C, followed by 40 cycles at 95°C for 3 s and 60°C for 30 s. The transcript levels were calculated relative to the levels of the housekeeping gene GAPDH using the following formula: 2^{[CT(GAPDH) − CT(target gene)]} × 1,000, where C_T (threshold cycle) is determined as the threshold cycle at which the fluorescent signal becomes significantly higher than the background.

The CXCL9 and CXCL10 protein levels in clarified CNS supernatants were assayed by using enzyme-linked immunosorbent assay (ELISA) kits (MCX900 and SMCX100) according to the procedures supplied by the manufacturer (R&D Systems, Minneapolis, MN). Optical densities were read at 450 nm in a Bio-Rad model 680 microplate reader and analyzed using Microplate Manager 5.2 software (Bio-Rad Laboratories, Hercules, CA).

Serum and CNS Ab.

The levels of virus-specific IgG within the CNS were determined in clarified spinal cord supernatants by ELISA as described previously (24). Briefly, 96-well plates were coated with 100 μl of a serum-free supernatant derived from JMHV-infected DBT cells and incubated overnight at 4°C. Plates were washed with phosphate-buffered saline (PBS)-Tween 20, and nonspecific binding was blocked with 10% fetal calf serum (FCS) in PBS overnight at 4°C. Samples were added and incubated overnight at 4°C. After washes, bound IgM and IgG were detected using biotinylated goat anti-mouse IgM (Jackson ImmunoResearch, West Grove, PA) or goat anti-mouse IgG2a (Southern Biotech, Birmingham, AL). Secondary Ab was detected using streptavidin-horseradish peroxidase (HRP) (BD Bioscience) followed by 3,3′,5,5′-tetramethylbenzidine (TMB reagent set; BD Bioscience). Optical densities were read at 450 nm in a Bio-Rad model 680 microplate reader and analyzed using Microplate Manager 5.2 software (Bio-Rad Laboratories). Data are expressed as arbitrary units/spinal cord where 1 arbitrary unit = 0.1 A₄₅₀. Levels were calculated using the following formula: (A₄₅₀/0.1) × dilution factor × volume of PBS used to homogenize spinal cord. Background levels from naïve mice were subtracted.

Neutralizing Ab was measured as described previously (36). Briefly, triplicates of serial 2-fold dilutions of heat-inactivated serum from individual mice (n = 5) were incubated with 50 PFU of JHMV in 96-well plates for 90 min at 37°C. DBT cells (8 × 10⁴ cells/well) were then added, and plates were incubated at 37°C for 48 h. The neutralization titers represent the log₁₀ of the highest average serum dilution that inhibited cytopathic effect.

Histology.

Spinal cords from PBS-perfused mice were snap-frozen in Tissue-Tek optimum cutting temperature compound (Sakura Finetek, Torrance, CA). Distribution of CD138⁺ ASC and CD4⁺ and CD8⁺ T cells was determined by staining acetone-fixed, 10-μm, longitudinal cryosections with Ab specific for CD138 (Abcam, Cambridge, MA), CD4, or CD8 (BD, Bioscience), respectively, followed by anti-mouse or anti-rat IgG. Immunoreactivity was detected using either the Vector Mouse on Mouse (M.O.M.) peroxidase immunodetection kit (Vector Laboratories, Burlingame, CA) or Vectastain ABC kit. Viral nucleocapsid protein was detected in paraffin-embedded sections by immunoperoxidase staining using anti-JHMV mouse monoclonal Ab (MAb) J.3.3 as described previously (37). Digital images were captured using a Leica DMLB light microscope (Leica Microsystems, Bannockburn, IL) equipped with a SPOT Insight camera (SPOT Imaging Solutions, Sterling Heights, MI). Sections were scored in a blinded manner, and representative fields identified based on the average score of all sections in each experimental group.

For detection of CXCL9 and CXCL10, longitudinal sections were fixed with 4% paraformaldehyde for 20 min, treated with 1% Triton X-100 in PBS at room temperature, blocked for 30 min, and then stained with rabbit anti-mouse laminin (Cedarlane Laboratories, Ontario, Canada), rat anti-mouse CD138 (BD Bioscience), goat anti-mouse CXCL9 (AF-492-NA; R&D Systems), goat anti-mouse CXCL10 (AF-466-NA; R&D Systems), rat anti-mouse glial fibrillary acidic protein (GFAP) (2.2B10; Invitrogen), rabbit anti-mouse ionized calcium binding adaptor molecule-1 (Iba-1) (Wako, Richmond, VA), anti-mouse CD31 (550274; BD Bioscience), or anti-JHMV monoclonal Ab overnight at 4°C. Secondary fluorescently labeled Ab were added, and the slides incubated for 1 h at room temperature. Sections were mounted with ProLong Gold antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen) and analyzed using a Leica DM2500 microscope.

Flow cytometric analysis.

For phenotypic analysis, CNS-derived mononuclear cells were isolated from pooled brains (3 or 4 mice per time point) as described previously (37). The expression of cell surface markers was determined by staining with Ab specific for CD4 (L3T4), CD8 (53-6.7), and CD45 (30-F11) (BD Bioscience). Virus-specific CD8 T cells were identified using H2-D^b/S510 major histocompatibility complex (MHC) class I tetramers (38). Cells were analyzed on a FACSCalibur flow cytometer (BD, Mountain View, CA) using FlowJo 7.1 software (Tree Star, Ashland, OR). Cell numbers were calculated based on cell yields and percentages of gated live cells.

Statistical analysis.

The results are expressed as the mean ± SEM for each group of mice. In all cases, a P value of <0.05, determined by the unpaired t test, was considered significant. Graphs were plotted and statistics assessed using GraphPad Prism 3.0 software.

RESULTS

Impaired Ab production within the CNS of CXCL10^−/− mice.

ASC recruitment into the CNS is CXCR3 dependent in C57BL/6 mice following infection with a gliatropic strain of coronavirus, as well as with a dual hepato- and neurotropic strain (23, 24). To determine whether the CXCR3 ligands CXCL9 and CXCL10 play interchangeable roles in mediating ASC migration, gliatropic-JHMV-infected CXCL9^−/− and CXCL10^−/− mice were analyzed for humoral responses in the periphery and CNS. Neutralizing Ab titers in serum were similar between infected WT, CXCL9^−/−, and CXCL10^−/− mice (Table 1), confirming that neither CXCL9 nor CXCL10 deficiency affected peripheral ASC activation following JHMV infection, similar to the results for CXCR3^−/− mice (24). Based on the high Ig heavy and light chain expression in ASC compared to that of non-Ab-secreting B cells, IgG and κ-light chain transcripts were assessed in the CNS. Total IgG and κ-light chain mRNA were barely detectable and similarly low at day 7 postinfection (p.i.) in all groups (Fig. 1A and B). Both transcripts increased in all groups by day 14 p.i., although the increase was more modest in CXCL10^−/− mice. By day 21 p.i., mRNA increased further, by ∼10-fold, in WT and CXCL9^−/− mice but only reached 25% of the WT values in CXCL10^−/− mice. Nevertheless, progressive low-level ASC accumulation was evidenced by levels of transcript expression similar to those in WT mice at day 14 p.i. Moreover, the expression of receptors that support ASC survival and differentiation, namely, BCMA and TACI (39–42), were also significantly upregulated at day 21 p.i. in infected WT and CXCL9^−/− mice (Fig. 1C and D). In contrast, they remained near naïve levels throughout infection in CXCL10^−/− mice (Fig. 1C and D). The kinetics of mRNA expression thus indicated similar ASC recruitment in WT and CXCL9^−/− mice but lagging levels in the absence of CXCL10. To verify that Ig mRNA reflects reduced virus-specific Ab production, CNS supernatants from day 21 p.i. were assessed for virus-specific IgM and IgG. Virus-specific IgM and IgG levels were similar between WT and CXCL9^−/− mice (Fig. 1E and F), with IgG exceeding IgM levels (24). In contrast, Ig levels were significantly lower in CXCL10^−/− mice. These data implied that ASC migration and virus-specific-Ab production within the CNS are CXCL10 dependent.

Table 1.

Normal serum neutralizing Ab titers in CXCR3 ligand-deficient mice

Day p.i.	Mean neutralization titer ± SEM in indicated micea
	WT	CXCL9−/−	CXCL10−/−
7	1.2 ± 0.04	1.2 ± 0.08	1.3 ± 0.07
14	2.0 ± 0.08	2.0 ± 0.09	1.9 ± 0.05
21	2.3 ± 0.07	2.2 ± 0.04	2.7 ± 0.07

^aNeutralization titers represent the log₁₀ of the highest average serum dilution that inhibited viral cytopathic effect. n = 5 individual mice per time point.

Fig 1.

Ab production within the CNS is impaired in the absence of CXCL10. The relative transcript levels of IgG (A), κ-light chain (B), BCMA (C), and TACI (D) in spinal cords of naïve and infected WT, CXCL9^−/−, and CXCL10^−/− mice were assessed by real-time PCR. Data depict the mean ± SEM relative to GAPDH mRNA of two or three independent experiments with at least 3 mice per time point per experiment. Significant differences between WT and CXCL9^−/− or CXCL10^−/− mice are denoted as follows: *, P < 0.05; **, P < 0.005; ***, P < 0.001. NS, not significant (P > 0.05). Day 7 p.i. expression levels were as follows: IgG, 0.26 ± 0.09 in WT, 0.17 ± 0.04 in CXCL9^−/−, and 0.17 ± 0.03 in CXCL10^−/− mice (A); κ-light chain, 0.45 ± 0.07 in WT, 0.48 ± 0.10 in CXCL9^−/−, and 0.63 ± 0.03 in CXCL10^−/− mice (B). (E and F) Virus-specific IgM and IgG2a levels in clarified supernatants from spinal cord (SC) homogenates at day 21 p.i. were assessed by ELISA. Arbitrary units reflect total Ab levels per spinal cord from 4 to 6 mice. Significant differences between WT and CXCL9^−/− or CXCL10^−/− mice are denoted as follows: **, P < 0.005; ***, P < 0.001. NS, not significant (P > 0.05).

CXCL9 and CXCL10 do not alter CNS chemokine or cytokine expression.

To exclude the possibility of chemokine or cytokine dysregulation in CXCL10^−/− and CXCL9^−/− mice, the transcript levels of CXCR3 ligands, as well as other chemokines and survival factors associated with B cell migration and survival, were assessed. CXCL9 mRNA was ∼50% lower in infected CXCL10^−/− mice than in WT mice at day 7 p.i. (Fig. 2). However, the relative difference in CXCL9 expression subsequently diminished, with indistinguishable CXCL9 mRNA in CXCL10^−/− and WT mice by day 21 p.i., the peak of ASC CNS accumulation (12, 24, 30, 36) (Fig. 2). In contrast, CXCL10 transcripts in the CNS of infected CXCL9^−/− mice showed no differences compared with the levels in infected WT mice (Fig. 2). The expression of the CXCR3 ligand CXCL11, although not functional in C57BL/6 mice (14), was also similar in all groups of infected mice (Fig. 2). The expression of CXCL12 mRNA, which regulates ASC recruitment and survival in bone marrow (13), did not exceed basal levels throughout infection in any group (Fig. 2). Potential alterations in the chemokines CXCL13, CCL19, and CCL21, all involved in regulating B cell movement within lymphoid tissue and ectopic follicles (43), were also assessed within the CNS. While infection increased CXCL19 mRNA to a similar extent in all groups (Fig. 2), CCL21 mRNA appeared to drop transiently at 7 days p.i. but was restored to basal levels by 14 days p.i. in all groups (Fig. 2). Unexpectedly, however, mRNA expression of CXCL13, which guides B cells into B cell zones within lymphoid tissue (44), was 10- to 20-fold higher in both CXCL9^−/− and CXCL10^−/− mice than in WT mice (Fig. 2). Lastly, to determine whether CXCL10 deficiency negatively affects factors associated with ASC survival, transcripts encoding APRIL and BAFF (39, 40, 45–47) were assessed. APRIL and BAFF mRNA increased in the CNS of infected CXCL9^−/− and CXCL10^−/− mice with similar kinetics as in WT mice, with the most-extensive upregulation of APRIL at day 14 p.i. and sustained elevated BAFF expression throughout days 7 to 21 p.i. (Fig. 2). With the exception of elevated CXCL13, these results indicate that the expression of chemokines and survival factors involved in B cell and ASC regulation is not altered in the absence of CXCL9 or CXCL10; moreover, CXCL9 cannot compensate for the absence of CXCL10 in ASC accumulation.

Fig 2.

Chemokine and cytokine expression in spinal cords of infected WT, CXCL9^−/−, and CXCL10^−/− mice. Relative transcript levels in spinal cords of naïve and infected WT, CXCL9^−/−, and CXCL10^−/− mice were assessed by real-time PCR. Data depict the means ± SEM relative to GAPDH mRNA of two or three independent experiments with at least 3 mice per time point per experiment. Significant differences between WT and CXCL9^−/− or CXCL10^−/− mice are denoted as follows: *, P < 0.05; ***, P < 0.001. NS, not significant (P > 0.05).

T cell function in the CNS does not require CXCL9 or CXCL10.

A role for T cells in promoting ASC recruitment to and survival within the CNS is poorly explored. Anti-CXCL10 Ab treatment reduced T cell recruitment following a heterologous coronavirus infection, resulting in uncontrolled virus replication and death prior to ASC recruitment (48). In contrast, CXCR3 deficiency did not impair T cell accumulation or initial viral control in the CNS following the gliatropic virus infection used herein (24). Nevertheless, CD4 T cell-derived IL-21, a cytokine that sustains CD8 T cell function and promotes ASC differentiation (49–51), was reduced by 50% in CXCR3^−/− mice relative to its WT levels (24). To ascertain whether defective ASC accumulation in the absence of CXCL10 is due to impaired T cell function in the CNS, T cell infiltration and antiviral and ASC-promoting cytokines were analyzed. Flow cytometry revealed similar CNS recruitment of CD4, total CD8, and virus-specific CD8 T cells (data not shown). Comparable transcript levels of antiviral IFN-γ, IL-21, and the anti-inflammatory and humoral immunity promoting factor IL-10 (Fig. 3) further indicated that neither the absence of CXCL9 nor that of CXCL10 altered T cell functions in the CNS. Notably, during JHMV infection, IFN-γ is secreted by both CD4 and CD8 T cells (31, 52, 53), while IL-21 and IL-10 are produced by CD4 T cells (30, 35). Despite ∼5-fold increased viral titers in CXCL10^−/− mice at day 7 p.i., infectious virus was controlled with similar kinetics and efficiency in all groups by day 21 p.i. (Fig. 3D). The analysis of spinal cords, the primary site of virus persistence, further revealed reductions of viral RNA in WT, CXCL9^−/−, and CXCL10^−/− mice by day 21 p.i. (Fig. 3E). The onset of encephalomyelitis and clinical disease were similar in all groups (data not shown). These data demonstrate that CXCL10 promotes early virus control but that neither the absence of CXCL9 nor that of CXCL10 significantly impairs T cell migration or antiviral function in the CNS prior to ASC accumulation.

Fig 3.

Antiviral T cell function in the CNS is independent of CXCL9 and CXCL10. Relative transcript levels of IFN-γ (A), IL-21 (B), and IL-10 (C) in spinal cords of naïve and infected WT, CXCL9^−/−, and CXCL10^−/− mice were assessed by real-time PCR. Data depict the means ± SEM relative to GAPDH mRNA of two or three independent experiments with at least 3 mice per time point per experiment. (D) Viral titers determined by plaque assay are expressed as the means ± SEM of two independent experiments with at least 3 mice per time point per experiment. Data are on a log₁₀ scale. Significant differences in viral titers between WT and CXCL10^−/− mice are denoted as follows: *, P < 0.05. (E) Viral nucleocapsid mRNA in spinal cords was determined by real-time PCR. Data are the means ± SEM relative to GAPDH mRNA of two or three independent experiments with at least 3 mice per time point per experiment.

CXCL9 and CXCL10 have regionally distinct localization in the CNS.

The apparently distinct functions of CXCL9 and CXCL10 in regulating ASC accumulation were unanticipated based on both their abilities to promote ASC transmigration in vitro (54). Differential in vivo recruitment may therefore be attributed to distinct quantities or localization. CXCL10 mRNA expression was ∼3-fold higher than CXCL9 mRNA in the CNS of WT mice at day 7 p.i.; however, mRNA levels were similar at days 14 and 21 p.i. (Fig. 2). Enhanced expression of CXCL10 at earlier time points may therefore result in more regionally concentrated and prolonged presentation of CXCL10. The measurement of both chemokines in clarified CNS supernatants by ELISA indicated peak values at day 7 p.i., with overall levels of CXCL10 exceeding those of CXCL9 (Table 2). However, both CXCL9 and CXCL10 protein declined rapidly by day 10 p.i., distinct from the sustained mRNA expression (Fig. 2). These data suggest that chemokine secretion is rapidly reduced and/or that extracellular chemokine is bound by surrounding matrix or consumed and thus not available to detection by ELISA. CXCL9 and CXCL10 production in the CNS has been associated with different cellular sources (15–22), suggesting distinct anatomical presentation. CXCL10 mRNA is prominently associated with astrocytes during viral as well as nonmicrobial neuroinflammation (15, 16, 18), whereas CXCL9 mRNA is associated with microglia (16–18). Thus, astrocyte-mediated CXCL10 expression may provide a stronger signal at the glia limitans to attract ASC into the parenchyma. To test this hypothesis, spinal cords from JHMV-infected mice were analyzed for chemokine expression prior to the emergence of ASC at day 14 p.i (12, 30, 36). CXCL10 was indeed most prominently expressed in GFAP⁺ astrocytes at day 7 p.i., with limited expression by Iba-1⁺ microglia/macrophages (Fig. 4). CXCL10 expression remained prominent in astrocytes at day 10 p.i., coinciding with initial ASC recruitment, and was still detectable at day 21 p.i. Notably, during viral persistence at day 21 p.i., CD138⁺ ASC located proximal to CXCL10⁺ astrocytes (data not shown). Lastly, the absence of CXCL10 expression by virus-infected cells supported sparse astrocyte infection (55) and IFN-mediated CXCL10 expression (56–58). In contrast to CXCL10, CXCL9 was exclusively associated with blood vessels (Fig. 5) and appeared to be contiguous with the endothelial matrix, as shown by dual labeling of CXCL9 and laminin (Fig. 5). The endothelial marker CD31 further revealed that some but not all CXCL9 colocalized with endothelial cells (Fig. 5). CXCL9 was also detected in few perivascular Iba-1⁺ microglia/macrophages but not in parenchymal microglia/macrophages and did not colocalize with GFAP⁺ astrocytes (Fig. 5). These data clearly uncover highly restricted, nonoverlapping expression patterns of CXCL9 and CXCL10 in vivo.

Table 2.

CNS levels of CXCL9 and CXCL10

Day p.i.	ng/spinal cord (mean ± SEM) ofa:
	CXCL9	CXCL10
Naïve	BD	BD
5	1.8 ± 0.6	79.5 ± 7.6
7	9.5 ± 1.6	143.9 ± 11.8
10	2.1 ± 0.6	8.2 ± 1.7

^aCXCL9 and CXCL10 were assessed by ELISA. n = 3 individual mice per time point. BD, below detection level.

Fig 4.

Astrocytes express CXCL10 during JHMV infection. Spinal cord sections from infected WT mice at day 7 p.i. were analyzed for CXCL10 expression (A to C, left). CXCL10 is identified with anti-CXCL10 Ab staining (red). Astrocytes (A), microglia/macrophages (B), and viral nucleocapsid protein (C) were visualized with anti-GFAP, anti-Iba-1, or anti-JHMV J.3.3 Ab, respectively (green, middle). DAPI staining (blue) was used to identify cell nuclei. Merged images are shown at right. Bar, 50 μm.

Fig 5.

CXCL9 is primarily detected at blood vessels. Spinal cord sections from infected WT mice at day 7 p.i. were analyzed for CXCL9 expression (A to D, left). CXCL9 was identified with anti-CXCL9 Ab staining (red). Laminin (A), endothelium (B), microglia/macrophages (C), and astrocytes (D) were visualized with antilaminin, anti-CD31, anti-Iba-1 Ab, or anti-GFAP Ab, respectively (green, middle). DAPI staining (blue) was used to identify cell nuclei. Merged images are shown at right. Bar, 50 μm.

CXCL10 controls ASC trafficking into the CNS parenchyma.

To determine whether CXCL9 and CXCL10 deficiency affected the anatomical distribution of ASC, spinal cords of persistently infected mice were analyzed for the frequency and distribution of ASC. The number and distribution of CD138⁺ ASC were similar when comparing WT and CXCL9^−/− mice (Fig. 6), with CD138⁺ ASC predominantly in demyelinated lesions and adjacent white matter. In contrast, almost no CD138⁺ ASC were detected in the white matter of CXCL10^−/− mice (Fig. 6). Moreover, the few CD138⁺ ASC were confined to the vasculature, often in small clusters (Fig. 6). Dual labeling for CD138 and laminin confirmed that the few ASC present in CXCL10^−/− mice were perivascular (Fig. 6). Reduced numbers of ASC, as well as vascular retention in the CNS of CXCL10^−/− mice, suggests that CXCL10 is instrumental not only in recruiting ASC but also in mediating their migration into the parenchyma.

Fig 6.

Parenchymal trafficking of CD138⁺ ASC is specifically impaired in CXCL10^−/− mice. (A) Longitudinal sections of spinal cords from infected WT (day 28 p.i.), CXCL9^−/− (day 28 p.i.), and CXCL10^−/− (day 25 p.i.) mice were analyzed for CD138⁺ cells. CD138⁺ cells were identified with anti-CD138 Ab (brown) with hematoxylin counterstain. Images are representative of areas reflecting the average blinded score value of each group. (B and C) Longitudinal sections of spinal cords from infected WT and CXCL10^−/− mice at day 21 p.i. analyzed for CD138⁺ cells and laminin. CD138⁺ cells were identified with anti-CD138 Ab (green). Laminin was visualized with antilaminin Ab (red). DAPI staining (blue) was used to identify cell nuclei. Merged images are shown. Bar, 50 μm.

CXCR3 ligands also act as cues for T cell migration and prolonged interaction with target cells within inflamed tissue (59, 60). However, neither the absence of CXCL9 nor that of CXCL10 affected the anatomical localization of CD4⁺ or CD8⁺ T cells within the CNS during virus-induced encephalomyelitis (Fig. 7). Similar to WT mice, both populations were detected in the white matter (Fig. 7). The numbers of virally infected cells and their distribution were also comparable in WT and CXCL10^−/− mice (data not shown), consistent with the similar viral mRNA (Fig. 3). Essentially all positive cells were detected in the white matter and exhibited small round morphology consistent in appearance with oligodendrocytes (data not shown). These data suggest that differences in ASC localization were not attributable to distinct distribution of either viral antigen or T cells.

Fig 7.

T cell trafficking within the CNS is CXCL9 and CXCL10 independent. Longitudinal sections of spinal cords from infected WT (day 28 p.i.), CXCL9^−/− (day 28 p.i.), and CXCL10^−/− (day 25 p.i.) mice were analyzed for CD4⁺ and CD8⁺ T cell distribution using anti-CD4 or anti-CD8 Ab (brown) with hematoxylin counterstain. Bar, 200 μm.

Impaired yet sustained Ig expression in CXCL10^−/− mice coincides with ongoing virus control.

The ongoing viral control in the CNS of infected CXCL10^−/− mice, despite significantly impaired ASC accumulation, differs from the lack of control in CXCR3^−/− mice. Virus-specific IgG levels in the CNS were therefore directly compared to determine whether this discrepancy reflected higher local Ab production in CXCL10^−/− than in CXCR3^−/− mice. Indeed, IgG was ∼6-fold higher in CXCL10^−/− than in CXCR3^−/− mice at 21 days p.i. (Fig. 8A). Similarly, the CNS expression of IgG and κ-light chain mRNA was increased 5- and 8-fold, respectively, in CXCL10^−/− versus CXCR3^−/− infected mice (data not shown). These data indicate that virus-specific IgG coincides with total Ig mRNA expression. To assess whether the low, yet elevated levels of virus-specific IgG in CXCL10^−/− versus CXCR3^−/− mice are sufficient to sustain viral control, WT and CXCL10^−/− mice were monitored during persistence. IgG and κ-light chain transcripts dropped less than 2-fold in WT mice and were sustained at similarly low levels in CXCL10^−/− mice between days 21 and 63 p.i. (Fig. 8B and C). Persisting viral mRNA in the spinal cord were comparable between WT and CXCL10^−/− mice at day 21 p.i. and progressively declined through day 63 p.i. in both groups (Fig. 8D). However, after 35 days p.i., viral RNA transcripts declined more slowly in the absence of CXCL10, as indicated by 6- and 8-fold higher viral mRNA levels at days 49 and 63 p.i. (Fig. 8D). Notably, however, infectious virus remained below detection in CXCL10^−/− mice (data not shown). These results suggest that a few ASC can be highly effective at sustaining protective ASC responses in the CNS.

Fig 8.

Local CNS Ab production is less impaired in CXCL10^−/− mice than in CXCR3^−/− mice. (A) Virus-specific IgG2a in clarified supernatants from spinal cord homogenates of infected WT, CXCL10^−/−, and CXCR3^−/− mice at day 21 p.i. was assessed by ELISA. Arbitrary units reflect total Ab levels per spinal cord from three individual mice per time point. Significant differences between WT and CXCL10^−/− mice are denoted as follows: ***, P < 0.001. Significant differences between CXCL10^−/− and CXCR3^−/− mice are denoted as follows: ###, P < 0.001. (B to D) Relative transcript levels of IgG (B), κ-light chain (C), and viral nucleocapsid mRNA (D) in spinal cords of infected WT and CXCL10^−/− mice were assessed by real-time PCR at the indicated time points. Data depict the means ± SEM relative to GAPDH mRNA with at least 3 mice per time point. Significant differences between viral nucleocapsid expression between WT and CXCL10^−/− mice are denoted as follows: *, P < 0.05.

DISCUSSION

Infections of the CNS in humans, as well as experimental models, are commonly associated with the presence of effector T cells and ASC or elevated Ig in cerebrospinal fluid. Although numerous studies have characterized chemokines and their receptors involved in T cell accumulation in the CNS, the chemokines driving ASC recruitment to the CNS are less well understood. Local ASC provide a protective role in reducing viral replication when T cells alone fail to eliminate virus. A better understanding of events underlying ASC CNS accumulation may thus lead to more-effective vaccination strategies. CXCR3 was recently identified as the chemokine receptor mediating ASC trafficking into the CNS in two different models of neurotropic coronavirus infection (23, 24). Here, we demonstrate a dominant role of the CXCR3 ligand CXCL10 in regulating CXCR3-mediated ASC recruitment to the CNS following infection with a gliatropic coronavirus. The absence of CXCL10 not only diminished overall ASC recruitment but trapped the few migrating ASC within the CNS vasculature. These data suggested that CXCL10 both acts to recruit ASC to the vasculature and mediates entry into the CNS parenchyma.

The preferential CNS recruitment of ASC by CXCL10 rather than CXCL9 correlated with their distinct localization patterns. CXCL10 colocalized primarily with astrocytes, and its expression was more abundant and dispersed than that of CXCL9. CXCL10 production by astrocytes is consistent with in situ CXCL10 mRNA detection in astrocytes during MHV infection and experimental allergic encephalomyelitis (16, 20). Moreover, the inability to colocalize CXCL10 with JHMV-infected cells suggests that CXCL10 is induced in astrocytes in response to proximally secreted IFN-α/β, IFN-γ, and/or TNF (58). In contrast, during West Nile virus encephalitis, neurons are the predominant sources of CXCL10 and also the major cell type supporting virus replication (19). CXCL10 is also in neurons during murine cerebral malaria, with only rare expression by endothelial cells and no expression by astrocytes (61). However, the prominent effect of CXCL10 on overall ASC recruitment, not only parenchymal entry, suggests that CXCL10 is also presented at vascular sites to promote local recruitment. The mechanisms underlying distinct CXCL10 production in these models may reside in the relative magnitudes of direct microbe-induced signals versus lymphocyte-derived cytokines, such as IFN-γ. Thus, under conditions of high IFN-γ levels, astrocytes may be the main CXCL10 producers, irrespective of virus tropism. In contrast, if local IFN-γ production is less prominent, the response of individually infected cell populations may dominate.

In contrast to CXCL10, CXCL9 was only detected in the microvasculature. This observation is consistent with CXCL9 detection on endothelial cells during cerebral malaria (61), as well as vascular localization in Rickettsia rickettsii infection and primary CNS lymphoma (22, 62). The vascular CXCL9 presentation also suggests local production of IFN-γ, a major inducer of CXCL9, by early infiltrating perivascular CD4 T cells (32) responding to perivascular antigen-presenting cells (63). The association of CXCL9 with perivascular cuffing further supports the concept that infiltrating cells may produce CXCL9 for presentation by nearby endothelial cells (64). However, the inability to detect CXCL9 in microglia/macrophages during JHMV infection was surprising, as these myeloid cells are capable of producing CXCL9 during CNS inflammation (16–18, 20). Furthermore, IFN-γ is abundantly expressed in the CNS during JHMV infection (31) and sufficient to induce major histocompatibility complex class II on microglia and major histocompatibility complex class I on oligodendrocytes (65–67). These data suggest that CXCL9 detection in microglia may be hampered by limited chemokine production or rapid release. Indeed, preliminary analysis of purified microglia from infected spinal cords revealed that both CXCL9 and CXCL10 are upregulated at the transcriptional level (data not shown), supporting a contribution of microglia to CXCR3 ligand production. The absence of CXCL9 expression by astrocytes is consistent with their lack of PU.1, a key determinant for CXCL9 expression by myeloid cells (17).

Impaired ASC recruitment in the absence of CXCL10 could not be attributed to reduced expression of other chemokines or cytokines known to affect ASC development or survival. However, elevated CXCL13 in both CXCL9^−/− and CXCL10^−/− but not CXCR3^−/− infected mice (24) was surprising, as CXCL13 upregulation is associated with ectopic follicle formation and ASC accumulation during neuroborreliosis (68). While CXCL13 is expressed by stromal cells in lymphoid tissue, both dendritic cells and macrophages may be additional sources during inflammation (43). Although the increased CXCL13 suggested increased B cell recruitment, we did not detect ectopic follicle-like structures (data not shown). Moreover, the similar and the significantly decreased Ig production in CXCL9^−/− and CXCL10^−/− mice, respectively, relative to that in WT mice, argues that CXCL13 does not promote ASC accumulation in this model, similar to results from Sindbis virus infection and autoantigen-induced CNS inflammation (69).

The lack of adverse effects on virus control in CXCL9^−/− and CXCL10^−/− mice was consistent with unaltered T cell function in the CNS. The distinct findings using CXCL10 Ab-mediated disruption may reside in the use of a different MHV variant and dose, as well as experimental approaches (48, 70, 71). However, finding no evidence for either loss of viral control during persistence or increased morbidity, despite significantly reduced Ig production in infected CXCL10^−/− mice relative to the Ig production in infected WT and CXCL9^−/− mice, was unanticipated given that virus was increased in infected CXCR3^−/− mice (24). These contrasting results may be partially explained by the elevated virus-specific Ab in the CNS of CXCL10^−/− relative to that in CXCR3^−/− mice. Furthermore, the confined localization of T cells to white matter, distinct from the widespread T cell distribution in CXCR3^−/− mice, may limit disease severity (21, 24). Lastly, CXCR3 expression on neurons as well as microglia may have physiological roles in modulating neuronal function, neuron-glia interactions, and neuroprotection (72–79). Specifically, CXCR3 expression on neurons mediates increased susceptibility to neuronal death in vitro following West Nile virus infection (79). While this would suggest a detrimental rather than protective function of CXCR3 signaling on neurons, our data suggest that CXCL9 and CXCL10 do not play a major role or compensate for each other in influencing CNS-resident glia or neurons during JHMV infection.

In summary, the results demonstrate a crucial role for CXCL10 in promoting ASC accumulation within the CNS during a viral infection. Furthermore, astrocytes as prominent sources of CXCL10, as well as ASC survival factors BAFF and APRIL, may be strategically relevant for focal ASC retention. Whether activated astrocytes play additional roles in directly promoting B cell differentiation, as recently described for CXCL10-expressing macrophages (80), remains to be investigated. Overall, the data highlight distinct regulation of T cell and ASC migration to the CNS and provide novel insights into intervention strategies to increase protective humoral immunity during CNS infections or impede detrimental B cells implicated in autoimmunity (81).