Astrocytes lure CXCR2-expressing CD4⁺ T cells to gray matter via TAK1-mediated chemokine production in a mouse model of multiple sclerosis

Significance

Multiple sclerosis (MS) is a chronic neurological disease characterized by demyelination and neuronal damage. While T cells are consistently detected in the gray matter of human MS samples, they are rarely seen in the gray matter of the most common mouse model of MS. Here, we show a modified mouse model that is characterized by a high degree of neuronal damage, T cell infiltration, and reactive gliosis in spinal cord gray matter. Using conditional knockout mice, we show that T cell migration to spinal cord gray matter depends on T cell expression of CXCR2 and astrocyte expression of TAK1-mediated chemokines such as CXCL1.

Abstract

Multiple sclerosis (MS) is a chronic neurological disease of the central nervous system driven by peripheral immune cell infiltration and glial activation. The pathological hallmark of MS is demyelination, and mounting evidence suggests neuronal damage in gray matter is a major contributor to disease irreversibility. While T cells are found in both gray and white matter of MS tissue, they are typically confined to the white matter of the most commonly used mouse model of MS, experimental autoimmune encephalomyelitis (EAE). Here, we used a modified EAE mouse model (Type-B EAE) that displays severe neuronal damage to investigate the interplay between peripheral immune cells and glial cells in the event of neuronal damage. We show that CD4⁺ T cells migrate to the spinal cord gray matter, preferentially to ventral horns. Compared to CD4⁺ T cells in white matter, gray matter-infiltrated CD4⁺ T cells were mostly immobilized and interacted with neurons, which are behaviors associated with detrimental effects to normal neuronal function. T cell-specific deletion of CXCR2 significantly decreased CD4⁺ T cell infiltration into gray matter in Type-B EAE mice. Further, astrocyte-targeted deletion of TAK1 inhibited production of CXCR2 ligands such as CXCL1 in gray matter, successfully prevented T cell migration into spinal cord gray matter, and averted neuronal damage and motor dysfunction in Type-B EAE mice. This study identifies astrocyte chemokine production as a requisite for the invasion of CD4⁺T cell into the gray matter to induce neuronal damage.

Multiple sclerosis (MS) is a prevalent, chronic neurologic autoimmune disease that results in accumulating disability. Disease onset usually occurs at 20–50 y of age and is characterized by symptoms of numbness, pain, fatigue, and/or visual impairment (1–3). Within 15–25 y of onset, 50% of MS patients require assistance with walking (4, 5) and 50% of MS patients report neurocognitive impairment (6). Accumulation of debilitating symptoms is attributed to an episodically inflamed central nervous system (CNS) as a result from recurrent attacks by immune cells (7).

Demyelinated lesions are the classical hallmark of MS (8, 9); thus, the disease is historically considered a disease that primarily affects white matter of the CNS. In the past 20 y, mounting evidence suggested that inflammatory lesions in the CNS are not restricted to white matter but are also observed in CNS gray matter (9–11). In addition to myelin loss, gray matter lesions present with neuronal damage characterized by axonal transection, synaptic loss, and even neuronal loss (12–16). Neuronal damage is proposed to underlie the permanent and irreversible neurological dysfunctions in persons with MS (17, 18).

The infiltration of antigen-specific lymphocytes such as T cells is implicated in CNS gray matter damage observed in MS (19) and an established mouse model, experimental autoimmune encephalomyelitis (EAE) (7–9). In the classical EAE model, T cells are mainly restricted to white matter of the spinal cord (20, 21) and are rarely found in spinal cord gray matter, with few exceptions (22). How T cells arise in CNS gray matter during MS pathogenesis is poorly understood. To mediate neuronal damage, T cells must be trafficked from lymphoid organs of peripheral tissues, such as lymph nodes and spleen, before transmigrating into the CNS. Such migration can occur via a vascular route through the blood–brain barrier, blood-cerebrospinal fluid, or meningeal lymphatic system (23, 24). Lymphocyte infiltration into the CNS is a tightly regulated process that is controlled by multiple factors that are cell-intrinsic or cell-extrinsic, including blood–brain barrier status, adhesion molecule expression, and presence of migratory cues (24, 25). During neuroinflammation, invading immune cells and local reactive glial cells create signaling gradients by secreting chemoattracting small peptide mediators to attract pathogenic cells to sites of inflammation. CNS-resident astrocytes have been identified as a key producer of important chemokines to induce the migration of T cells (26, 27).

Here, while exploring the drivers of severe neuronal damage in spinal cord gray matter of mice induced to have a neurodegenerative form of EAE (termed Type-B EAE), we made the serendipitous observation that Type-B EAE is characterized by massive infiltration of CD4⁺ T cells into gray matter of the spinal cord. Accumulation of CD4⁺ T cells in spinal cord gray matter was prevented by genetic ablation of T cell CXCR2. Additionally, genetic ablation of astrocyte TAK1, an upstream molecule of CXCR2 ligand CXCL1, successfully prevented T cell migration to spinal cord gray matter, neuronal damage, and motor dysfunction.

Results

Severe Neuronal Damage in Gray Matter of Type-B EAE Mice.

C57BL/6J (B6) mice received two types of myelin oligodendrocyte glycoprotein (MOG) peptide-specific EAE induction: “Type-A” EAE that consisted of the widely used one-time EAE immunization (28) with low-dose heat-killed Mycobacterium tuberculosis (Mtb, 200 μg per mouse) or “Type-B” EAE that consisted of two immunizations with high-dose heat-killed Mtb (400 μg per mouse) (Fig. 1A) (29). To evaluate motor paralysis that correlates with spinal cord inflammation in EAE (30, 31), clinical signs were scored daily as follows: 0.5, partial tail limpness; 1, tail limpness; 1.5, reversible impaired righting reflex; 2, impaired righting reflex; 2.5, paralysis of one hindlimb; 3, paralysis of both hindlimbs; 3.5, paralysis of both hindlimbs and one forelimb; 4, hindlimb and forelimb paralysis; and 5, death. As previously observed (29), clinical severity defined by motor paralysis started around 10 d postimmunization (dpi), peaked at 21 dpi, and tapered into disease remission in Type-A EAE mice (Fig. 1B). On the other hand, clinical severity of Type-B EAE mice did not taper and instead chronically increased up to 60 dpi (Fig. 1B).

Fig. 1.

Severe neuronal damage in gray matter of Type-B EAE mice. (A) Schematic for two EAE mouse models. (B) Daily EAE disease scores of Type-A and Type-B EAE mice up to 60 dpi (Type-A EAE: n = 6; Type-B EAE: n = 6). (C) Representative brightfield images of Golgi-Cox–stained lumbar spinal cord hemisections (Left) and quantification (Right) of percentage area stained over ROI area marked by red dotted lines (GM, gray matter) at 28 dpi (n = 6 per condition). Data are a compilation of two independent experiments. (D) Size of Golgi-Cox⁺ cells in vGM (n = 3 per condition). (E) Representative 3D images of Golgi-Cox–stained neuron dendrites in vGM (Left). Quantitative analysis of dendritic spine density (Right) at 28 dpi (n = 3). Statistical analysis: two-tailed Student’s t test (B), Mann–Whitney u test (C–E). *P < 0.05, ns, not significant. Schematic cartoons in (A) are credited to https://biorender.com/.

To determine correlation between duration of motor dysfunction and degree of neuronal damage in the CNS, we examined neuronal morphology in gray matter of the lumbar spinal cord of Type-A and Type-B EAE mice at 28 dpi using Golgi-Cox staining (32). By 28 dpi, chronic disease is expected to produce neurodegeneration or axonal loss detectable by histopathology. In addition, this timepoint captured significantly different EAE scores between Type-A and Type-B EAE animals, thus permitting the evaluation of cellular mechanisms underlying diseases severity. A significant, ∼34% reduction in Golgi-Cox staining was observed in the gray matter of lumbar spinal cord (L4–L6) of Type-B EAE mice when compared to Type-A EAE mice (Fig. 1C). Interestingly, a significant, ∼25% reduction in Golgi-Cox staining was observed in the ventral gray matter (vGM) of Type-B compared to Type-A EAE mice (Fig. 1C). In contrast, the reduction (∼9%) in Golgi-Cox staining observed in dorsal gray matter (dGM) of Type-B compared to Type-A EAE mice was not significant (Fig. 1C). Analysis of cell sizes of Golgi-impregnated cells in the vGM revealed enlargement, which reflects cell swelling in Type-B EAE mice when compared to Type-A EAE mice (Fig. 1D). Because dendritic spine loss and abnormalities are reported to manifest in EAE (33, 34), we next quantified spine density on dendrites in the vGM using Golgi-Cox–stained samples reconstructed in three dimensions (3D) via confocal reflection microscopy (32). The number of vGM dendritic spines was significantly reduced by 33% in Type-B EAE mice compared to Type-A EAE mice (Fig. 1E). Together, these results suggest that Type-B EAE triggers neuron loss and neuron morphological abnormalities in vGM of lumbar spinal cord. This provides a likely explanation for our observation of persistent motor dysfunction in these mice (Fig. 1A). Further, these results suggest that Type-B EAE is a useful model to identify the mechanism of neuronal damage in spinal cord gray matter compared to the conventional Type-A EAE model.

CD4⁺ T Cell Migration to the Gray Matter in Type-B EAE Is Dependent on CXCR2 Expression.

CD4⁺ T cells are key immune cells for the induction of motor dysfunction and CNS neuronal damage during EAE (29, 35, 36). We evaluated CD4⁺ T cell infiltration as visualized via anti-CD4 immunofluorescence in the spinal cord of Type-A and Type-B EAE mice at 28 dpi. CD4⁺ T cells were twofold more abundant in the gray and white matter of Type-B EAE mice compared to Type-A EAE mice (Fig. 2A and SI Appendix, Fig. S1A). Strikingly, while CD4⁺ T cell accumulation in gray matter was sparse and evenly distributed between dGM and vGM in Type-A EAE mice, CD4⁺ T cells were selectively and significantly enhanced in vGM of Type-B EAE mice (Fig. 2A). Additionally, a significant increase in CD4⁺ T cells was observed in ventral white matter (vWM) of Type-B EAE mice compared to Type-A EAE mice, but not in dorsal white matter (SI Appendix, Fig. S1A).

Fig. 2.

CD4⁺ T migration to gray matter is dependent on CXCR2 chemokine signaling in Type-B EAE mice. (A) Representative micrographs of CD4⁺ T cells in lumbar spinal cord sections labeled with CD4 antibody at 28 dpi (Left) and quantification of CD4⁺ T cell count in gray matter (GM), dGM, and vGM (Right) (Type-A EAE: n = 5; Type-B EAE: n = 4). (B) Representative confocal images with overlaid T cell trajectories (corresponding to 40 min of imaging) in vGM at 28 dpi. Average cell velocity (μm/min) is graphed (>30 cells analyzed per condition). Data are representative of two independent experiments. (C) Representative merged 3D micrograph of CD4⁺T cell (red) and Golgi-Cox–stained neuron (purple) in vGM of Type-B EAE lumbar spinal cord using conventional and reflection confocal modalities, respectively. (D) Representative merged 3D micrograph of fluorescent CD4⁺T cell and TUJ1+ neuron in ventral gray matter of Type-B EAE lumbar spinal cord. (E) Representative images (Left) and quantification (Right) of CD4⁺ T cells in vGM of the lumbar spinal cord of control Cxcr2 ^fl/fl and cCxcr2^−/− mice at 28 dpi (n = 4 per condition). (F) Daily EAE disease scores of Type-B Cxcr2 ^fl/fl and cCxcr2^−/− mice up to 28 dpi (Cxcr2 ^fl/fl: n = 8; cCxcr2^−/−: n = 6). Statistical analysis: Mann–Whitney u test (A, B, and E), two-tailed Student’s t test (F). *P < 0.05.

To study T cell dynamics in the spinal cord during EAE, we evaluated T cell movement in explanted lumbar spinal cord using EAE-induced lymphocyte protein tyrosine kinase (Lck)-iCre-Tomato-floxed mice in which T cells endogenously express red fluorescence protein (RFP). RFP expression in gray and white matters was measured in ex vivo spinal cord tissue isolated at 28 dpi using time-lapse z-stack confocal imaging. Mean velocity of T cells in vGM was significantly slower than in vWM in both Type-A and Type-B EAE mice (Fig. 2B and SI Appendix, Fig. S1B). This finding suggests that T cells may have more lasting interactions with other cells, such as motor neurons, in the vGM compared to vWM. Reduced velocity of T cells in the gray matter may be indicative of cell death (37). Therefore, we applied dead cell dye to ex vivo spinal cord slice of Lck-iCre-Tomato-floxed mice. RFP⁺ T cells in ex vivo spinal cord slice were not positive for dead cell staining (SI Appendix, Fig. S1C). In contrast, in a control experiment for dead cell staining, RFP⁺ T cells were positive for dead cell staining in spinal cord slices treated with high concentrations of anti-CD3 and anti-CD28 antibodies to induce T cell death (SI Appendix, Fig. S1C). This result suggests that the immobilization of RFP⁺ T cells in the spinal cord gray matter is not due to cell death.

To determine if T cells in the gray matter interact with CNS neuronal populations, we first conducted CD4 immunohistochemistry in Golgi-Cox samples. By combining confocal fluorescence and reflection microscopy (32), CD4⁺ T cells were seen to adhere to Golgi-stained neurons in vGM of Type-B EAE mice at 28 dpi (Fig. 2C). Adherence of CD4⁺ T cell to neurons was also observed via immunohistochemistry with CD4 and TUJ1, a neuron marker (Fig. 2D). Yet, CD4⁺ T cells were not observed to adhere to microglia (indicated by IBA1 staining) or astrocytes (indicated by glial fibrillary acidic protein [GFAP] staining) (SI Appendix, Fig. S1D). Thus, T cells may exhibit long-term cell-to-cell interactions with neuronal bodies.

We next sought to identify necessary chemotactic signals needed to facilitate CD4⁺ T cell migration into vGM of the spinal cord in Type-B EAE mice. Because CXCR2 is abundantly expressed by CD4⁺ T cell in Type-B but not Type-A EAE mice (29, 38), we evaluated the role of CXCR2 signaling in CD4⁺ T cell migration to vGM using T cell-specific conditional knockout of CXCR2 (cCxcr2^−/− mice) by breeding LckiCre mice with Cxcr2-floxed mice. The number of CD4⁺ T cells in vGM was significantly reduced by 50% in Type B EAE-induced cCxcr2^−/− mice compared with Type-B EAE-induced control (Cxcr2-floxed) mice (Fig. 2E). Furthermore, similar to a previous report (29), Type-B EAE-induced cCxcr2^−/− mice displayed less severe motor paralysis symptoms when compared to Type-B EAE-induced control mice (Fig. 2F). This result suggests that CD4⁺ T cell expression of CXCR2 is essential for CD4⁺ T cell migration to vGM of spinal cord and symptom development in Type-B EAE mice.

CXCR2 Ligands CXCL1 and CXCL5 Are Up-Regulated in Gray Matter Astrocytes within Spinal Cord Ventral Horn in Type-B EAE Mice.

We next investigated whether the migration of CXCR2-expressing CD4⁺ T cells in Type-B EAE-induced mice was dependent on signaling cues derived from spinal cord. CD4⁺ T cells were isolated from Type-B EAE mice at 14 dpi. By this timepoint antigen-specific T cells are clonally proliferated and express CXCR2, which allows abundant isolation of MOG antigen-specific and CXCR2-expressing CD4⁺ T cells (29). Isolated splenic CD4⁺ T cells (10⁵ cells) were intrathecally transferred into the spinal cord of Type-B or Type-A EAE-induced Tcra^−/− mice, which intrinsically lack T cells, at 14 dpi (Fig. 3A). Approximately twofold more Type-B EAE-derived CD4⁺ T cells migrated to vGM of the spinal cord in Tcra^−/− mice with Type-B EAE when compared to Type-A EAE (Fig. 3B). This suggests that Type-B EAE induced changes in the spinal cord, independent of T cells, which were essential for T cell migration to vGM. Symptoms of paralysis were not observed in Tcra^−/− mice with Type-A or Type-B EAE induction before CD4⁺T cell transfer (SI Appendix, Fig. S2A); this finding aligns with EAE as a T cell-driven disease model. In contrast, post-CD4⁺ T cell transfer, EAE disease score was significantly higher in Type-B EAE-induced Tcra^−/− mice with CD4⁺ T cell transfer than Type-A EAE-induced Tcra^−/− mice with CD4⁺ T cell transfer (SI Appendix, Fig. S2A). Together, these findings suggest that the presence of signaling cue(s) derived from Type-B EAE induction in Tcra^−/− mice is necessary for T cell migration to vGM and the development of prolonged disease.

Fig. 3.

CXCR2 ligands are up-regulated in gray matter astrocytes in spinal cord ventral horns. (A) Schematic for intrathecal passive EAE model. (B) Representative images (Left) and quantification (Right) of tdTomato CD4⁺T cells in vGM of the spinal cord of Type-A EAE-Tcra^−/− and Type-B EAE-Tcra^−/− mice that received passive intrathecal injections of CD4⁺ T cells (10⁵ cells) isolated from spleen of Lckicre-tdTomato Type-B EAE mice at 14 dpi (n = 3 per condition). (C and D) Representative micrographs of astrocytes and monocyte/microglia in vGM of lumbar spinal cord sections derived from Type-A EAE-Tcra^−/− and Type-B EAE-Tcra^−/− mice before CD4⁺ T cell transfer. Bar graphs show glial reactivity measured by integrated density in vGM (Type-A EAE-Tcra^−/− mice: n = 5; Type-B EAE-Tcra^−/− mice: n = 4). (E and F) Representative micrographs of astrocytes (E) or microglia/macrophages (F) in vGM of spinal cord sections derived from Type-A EAE and Type-B EAE B6 mice at 28 dpi. Bar graphs show glial reactivity measured by integrated density in vGM (Type-A EAE: n = 5; Type-B EAE: n = 4). Data are representative of two independent experiments. (G) Representative micrographs of CXCL1 in vGM. Bar graph shows chemokine ligand expression levels measured by integrated density in vGM (Type-A EAE: n = 5; Type-B EAE: n = 4). (H) Representative micrographs of CXCL1 (red) costained with GFAP (green) and IBA1 (purple) or TUJ1 (green) in vGM of Type-B EAE mice at 28 dpi. Statistical analysis: Mann–Whitney u test. *P < 0.05. Schematic cartoons in (A) are credited to https://biorender.com/.

To identify cellular players involved in attracting and maintaining T cell localization to the gray matter, we investigated glial activation status in vGM of EAE-induced Tcra^−/− mice at 14 dpi without T cell transfer. By applying immunohistochemistry to lumbar spinal cord sections, we found that astrocytes are significantly more reactive in vGM of Type-B EAE Tcra^−/− mice compared to Type-A EAE Tcra^−/− mice without CD4⁺ T cell transfer, as indicated by heightened GFAP immunoreactivity (39) (Fig. 3C). There was no significant difference in microglia/macrophage density and reactivity, as indicated by IBA1 immunoreactivity, in vGM between the two groups (Fig. 3D). These findings identify astrocytes as being a potential key cellular player in the event of T cell localization in gray matter.

We next asked whether Type-B EAE differentially alters the astrocyte component compared to Type-A EAE in B6 mice. Significantly greater integrated fluorescence density of GFAP staining was observed in vGM of Type-B EAE compared to Type-A EAE mice (Fig. 3E). Interestingly, in contrast to the absence of heightened microglia/monocyte reactivity in Type-B EAE Tcra^−/− mice without CD4⁺ T cell transfer, significantly greater IBA1 integrated fluorescence density was observed in vGM of Type-B EAE compared to Type-A EAE mice (Fig. 3F). Further morphometric analysis of IBA1 soma characteristics revealed that microglia/macrophage soma in Type-B EAE B6 mice are rounder than those of Type-A EAE condition, which implies microglial activation or monocyte infiltration (40) (SI Appendix, Fig. S2B). No change in soma size was found (SI Appendix, Fig. S2B). These results suggest that CD4⁺ T cells are required for microglial/macrophage activation or recruitment in Type-B EAE mice. Together, our results demonstrate that Type-B EAE enhanced astrocyte reactivity in vGM independent of the presence of T cells, which may be linked to local T cell invasion and aggregation.

Local activation of astrocytes in the vGM of Type-B EAE B6 and Tcra^−/− mice suggests that these astrocytes may secrete chemotactic ligands to encourage T cell homing. To identify vGM-derived factors expressed in Type-B EAE that facilitate CD4⁺ T cell localization to vGM, we evaluated CXCR2 ligand expression in the spinal cord of Type-B and Type-A EAE mice. CXCR2 ligands CXCL1 and CXCL5 were significantly up-regulated in vGM of the spinal cord of Type-B compared to Type-A EAE mice (Fig. 3G and SI Appendix, Fig. S2C). CXCL1 was expressed by astrocytes, but not microglia or neurons, as indicated by coimmunostaining of CXCL1 with GFAP and IBA1 or with TUJ1, a neuron-specific tubulin marker (Fig. 3H). These results suggest that migration of CXCR2-expressing CD4⁺ T cells to vGM may be mediated by increased site-specific CXCR2 ligands expressed by astrocytes.

Astrocyte-Targeted TAK1 Depletion Prevents CD4⁺ T Cell Migration to Gray Matter, Neuron Damage, and Motor Dysfunction in Type-B EAE Mice.

Transforming growth factor β-activated kinase 1 (TAK1) is upstream of CXCL1 and CXCL5 chemokine production in astrocytes (41, 42). To determine the role of astrocyte chemokine ligand production in CD4⁺ T cell trafficking to vGM of the spinal cord, we used a GFAP⁺ cell conditional knockout of TAK1 (cTak1^−/− mice) by breeding GFAP-Cre mice with Tak1-floxed mice. GFAP-Cre mouse lines are widely used to target astrocytes for genetic manipulation (43, 44). cTak1^−/− mice showed a significant ∼40% reduction of CXCL1 expression in astrocyte in vGM of Type-B EAE mice (Fig. 4A). cTak1^−/− mice also showed significantly less severe EAE symptoms upon Type-B EAE induction compared to control mice (Fig. 4B), suggesting reduced inflammation in the spinal cord. Disease amelioration was not observed in cTak1^−/− mice with Type-A EAE (SI Appendix, Fig. S3). This suggests that astrocyte TAK1 signaling may be a key disease mechanism in Type-B but not Type-A EAE.

Fig. 4.

Astrocyte-specific TAK1 depletion prevents CD4⁺ T cell migration to gray matter, neuron damage, and motor dysfunction in Type-B EAE mice. (A) Representative micrographs (Left) and quantification (Right) of CXCL1 in vGM. Bar graph shows CXCL1 expression levels measured by integrated density in vGM. (B) Daily EAE disease scores up to 28 dpi. (C) Representative micrographs (Left) and quantification (Right) of CD4⁺T cells in vGM of spinal cord hemisections at 28 dpi. (D) Representative brightfield images of Golgi-Cox–stained vGM (Left) and quantification (Right) of percentage area stained over ROI area at 28 dpi. (E) Representative 3D images of Golgi-Cox–stained neuron dendrites in vGM (Left). Quantitative analysis of dendritic spine density (Right). Control Type-B EAE: n = 5; cTak1^−/− Type-B EAE: n = 4 for all experiments. Statistical analysis: Mann–Whitney u test. *P < 0.05.

While high numbers of CD4⁺ T cells are observed in vGM of Tak1-floxed control mice with Type-B EAE, a significant 80% reduction in CD4⁺ T cell localization in vGM was observed in cTak1^−/− Type-B EAE mice (Fig. 4C). Consistent with our observation that neuronal damage in the ventral horn was correlated with disease severity, area percentage of Golgi-Cox staining and number of dendritic spines in the vGM were approximately two times significantly greater in cTak1^−/− mice with Type-B EAE when compared to Tak1-floxed control mice with Type-B EAE (Fig. 4 D and E). These results suggest that TAK1-mediated chemokines such as CXCL1 and CXCL5 expression in astrocytes is necessary to facilitate CD4⁺ T cell migration into vGM of the spinal cord of Type-B EAE mice and produce local neuronal damage.

Discussion

Our studies highlight an important role of CD4⁺ T cell chemokine receptor and astrocyte-derived chemokine ligands in determining spatial distribution of infiltrating peripheral CD4⁺ T cells in gray matter of the CNS. Type-B EAE presents with prolonged motor dysfunction without remission, and CD4⁺ T cells were highly augmented in lumbar spinal cord gray matter, specifically the ventral horns, during chronic disease. Importantly, lumbar spinal cord ventral horns host lower motor neurons that innervate hindlimb muscles. Additionally, lesions are predominantly seen in the lumbar region of the spinal cord when using the classic EAE paradigm (45, 46). However, based on recent observations of diffuse T cell localization in whole spinal cord during chronic disease (47), we postulate that T cell localization in the gray matter could also occur in spinal cord regions outside of lumbar regions. This hypothesis requires experimental validation. Here, we show that the abundance of CD4⁺ T cells in gray matter spatially correlates with neuronal damage pathology, defined here as the reduction of neuronal bodies, swelling of neurons, and loss of dendritic spines. Increased neuronal cell size was demonstrated by our group and others as potentially reflecting cell swelling, which is indicative of cell death processes such as necrosis (38, 48, 49). While it is tempting to speculate that an increased cell size may be reconciled with the reduction in dendritic spine density, this remains to be experimentally addressed.

Several studies demonstrated correlations between white matter lesions and gray matter atrophy, suggesting that degeneration of neurons can be a result of white matter demyelination via retrograde degeneration (50). Another hypothesis suggests that white matter and gray matter damage are partially independent events. Postmortem studies indicate that loss of neurons occurs within both gray matter demyelinated lesions and normal-appearing gray matter (51, 52), suggesting that neuronal loss might occur independent of demyelination. Several mechanisms are proposed to explain how T cells damage neurons. These include neurotoxic cell contact-dependent interactions, such as IFNγ-mediated induction of calcium-permeable complexes between IFNγ receptor and glutamate receptor 1 expressed on neurons; induction of cellular apoptosis by Fas–FasL interactions; and initiation of neurite retraction by semaphorin glycoproteins (29, 53). Our previous report demonstrated that CD4⁺ T cells from Type-B EAE mice have significantly increased expression of semaphorin 6B, which induces neuronal damage in the form of neurite retraction in a coculture system with primary neurons (29). Using mice that express RFP in LCK⁺ T cells, we found that T cells significantly accumulate in gray matter of Type-B EAE mice. Notably, T cells analyzed for velocity in this experiment were not limited to CD4⁺ T cells and may have included other LCK-expressing cells such as CD8⁺ T cells. Because we found that immobilized T cells interact with neurons in vGM of Type-B EAE mice, these cell interactions may account for neuronal damage.

It is increasingly accepted that astrocyte and microglial activation are necessary and crucial for host defense and neuron survival, whereas glial overactivation has deleterious effects. In the context of CD4⁺ T cell-driven EAE pathology, microglial cells are massively activated in response to proinflammatory cytokines released from T lymphocytes, which, in turn, further potentiate site-specific inflammation and neurodegeneration (54). For example, oxidative burst by activated microglia has a major role in the induction of demyelination and axonal injury in MS lesions (55, 56). Our findings show that microglia are in a hyperactivated state in the presence of abundant CNS-infiltrated T cells, indicated by increased soma roundness and cellular aggregation. Interestingly, our studies using Tcra^−/− Type-B EAE mice revealed that CD4⁺ T cells are required for microglial activation in vGM. Because we did not observe direct interaction between microglia and CD4⁺ T cells in the vGM, we suspect that CD4⁺ T cells may not directly interact with microglia, but instead influence microglial effector function via secretion of cytokines and other microglia-activating factors. Thus, we hypothesize that activated microglia in vGM likely contributes to neuronal damage observed in vGM of Type-B EAE mice.

Several studies provided data that indicate infiltration of activated peripheral immune cells is associated with astrocytes up-regulating chemokines during neuroinflammation (57). Steelman and coworkers reported that a TAK1 signal is essential to generate a range of chemokine ligands, including high-affinity CXCR2 ligand such as CXCL1 and CXCL5, in astrocytes in vitro (41). While genetic deletion of TAK1 is considered to be limited to astrocytes, whose hallmark filament protein is GFAP, GFAP expression in adult mouse spinal cord has been observed in neuronal populations, specifically, neural stem cells, where it is believed to be vital for neurogenesis (58, 59). Thus, we cannot exclude the possibility that the effect of TAK1 deletion on T cell migration is partially dependent on neuronal expression of TAK1. Nonetheless, it is important to note that TAK1 signaling in neuronal stem cells has not been associated with chemotactic functions, whereas TAK1 signaling in astrocytes largely influences production of chemokine ligands (41). We also showed CXCL1 is not colocalized with neurons in the ventral gray matter of lumbar spinal cord in Type-B EAE mice. Thus, we attribute the reduction in T cell localization in spinal cord gray matter of GFAP-specific Tak1^−/− mice to reduction in astrocyte chemokine ligand production downstream of TAK1 signaling. Astrocytes are activated by Type-B EAE induction in the absence of T cells and activated microglia, suggesting that other cellullar players are involved in astroyte activation. Recently, we reported that Type-B EAE induces up-regulation of reactive oxygen species and proinflammatory cytokines in neutrophils (60), which can result in astrogliosis (41). Therefore, neutrophils may be a critical population involved in activating astrocytes in Type-B EAE.

Chemokine ligands for CXCR2 are commonly considered to be neutrophil chemoattraction factors because CXCR2 is abundantly expressed by neutrophils (61). The Type-B EAE model provides a unique opportunity to study the effects of TAK1-mediated chemokine ligands on migration patterns of CXCR2-expressing CD4⁺T cells. Based on our previous finding of increased CXCL1 levels in serum and the surprising up-regulation of CXCR2 on CD4⁺ T cells (61), we speculate that CXCR2 migration plays a role in the accumulation of immune cells in the CNS. Indeed, T cell-specific knockout of Cxcr2 successfully inhibited T cell infiltration into gray matter. To assess whether the presence of Type-B EAE-related intrinsic factors in the inflamed spinal cord was necessary for migration of CD4⁺ T cells to gray matter, we performed passive transfer of CD4⁺ T cells isolated from Type-B EAE donor mice to Type-A or Type-B EAE-induced recipient Tcra^−/− mice via an intrathecal route. We chose intrathecal route instead of the more commonly used intravenous route to bypass the process of T cell migration from the periphery to the CNS in order to avoid the possibility of cell property changes in the periphery. Indeed, while T cells were abundant in gray matter of ventral horns of Type-B EAE recipient mice, they were sparse in Type-A EAE recipient mice. These results indicate that localization of T cells in gray matter is dependent on T cell expression of CXCR2 and signaling factors present in Type-B EAE mice.

We previously demonstrated that serum CXCL1 levels significantly increase during disease development in Type-B but not Type-A EAE mice (29, 38). In the current study, we show that CXCL1 and CXCL5 are highly up-regulated in gray matter of lumbar ventral horn in Type-B EAE mice, which have interferon-beta–insensitive disease (29, 38). Incidentally, CXCL1 messenger RNA was reportedly up-regulated in persons with interferon-beta–insensitive MS, whereas CXCL5 was recently reported to be a serum biomarker associated with MS relapse (62, 63). In this study, we focused on the migration-inducing capacity of astrocyte-derived chemokine ligands in mediating T cell accumulation in gray matter. While not previously investigated, it is possible that the role of CXCR2 on CD4⁺ T cells expands beyond cellular trafficking. For instance, we and others showed that activation of CXCR2 by recombinant CXCL1 stimulates transcription of genes related to oxidative stress in neutrophils (38, 64).

In conclusion, we found that Type-B EAE is characterized by up-regulated TAK1-mediated CXCL1 expression by astrocytes in the gray matter of ventral horns. This expression is necessary for CD4⁺ T cell localization in gray matter via CXCR2-dependent migration. In summary, our autoimmune gray matter inflammation model lends itself to the study of autoimmune-mediated neurodegeneration and may open new avenues for the development of individual prognostic profiles and therapeutic measures for patients with MS.

Materials and Methods

Animals.

Healthy male and female C57BL/6J (Jackson Laboratories, no. 000664) mice were used for experiments. Tcra mice (no. 002116), Cxcr2^fl mice (no. 024638), tdTomato^fl mice (no. 007908), Lckicre mice (no. 012837), Gfapcre mice (no. 024098), and Tak1^fl mice (no. 011039) of C57BL/6J background were purchased from The Jackson Laboratories. The studies described here were approved by University of Illinois at Urbana–Champaign Institutional Animal Care and Use Committee (protocol no. 19171). Additional details are stated in SI Appendix.

Active EAE Induction.

Six- to eight-week-old mice were randomly subjected to either Type-A EAE or Type-B EAE. Type-A EAE induction consists of one subcutaneous (s.c.) injection of 100 μg of myelin oligodendrocyte glycoprotein (MOG)_35–55 peptide (MOG_35–55, United Peptides) emulsified in complete Freund’s adjuvant (no. F5881, Sigma) including 200 μg per mouse heat-killed Mycobacteria (Mtb, no. DF3114-33-8, Fisher). Type-B EAE consists of two s.c. injection of 100 μg of MOG_35–55 emulsified in complete Freund’s adjuvant including 400 μg per mouse Mtb at 0 and 7 dpi. Pertussis toxin (no. 181, List Biological Laboratories, Inc., 200 ng per mouse) was i.p. injected on 0 and 2 dpi for Type-A EAE and 0, 2, and 7 dpi for Type-B EAE. Determination of EAE behavioral score can be found in SI Appendix.

Passive EAE Induction by Intrathecal Transfer of CD4⁺ T Cells.

Six- to eight-week-old Tcra^−/− mice were randomly subjected to Type-A or Type-B EAE induction. Six- to eight-week-old B6 mice were subjected to Type-B EAE simultaneously. At 14 dpi, B6 mice with Type-B EAE are killed to allow harvesting of splenic CD4⁺ T cells. Detailed methods on splenic immune cell isolation are provided in SI Appendix. Isolated CD4⁺ T cells (10⁵ cells per mouse) were adoptively transferred by intrathecal injection to Tcra^−/− recipient mice with Type-A or Type-B EAE induction at 14 dpi. Animals were monitored daily until 14 d-postpassive transfer.

Golgi-Cox Confocal Microscopy and Analysis.

Animals were perfused with 4% paraformaldehyde (PFA) with no postfixing then processed according to FD Rapid Golgi staining kit protocol. To obtain spine density data, 50 µm of Golgi-Cox–stained spinal cord sections were imaged using a Nikon A1 confocal scanning microscope under Confocal Reflection Super-Resolution modality and dendritic spines were analyzed using filament tracer autopath function (Imaris). To obtain percentage area stained and size of Golgi-Cox⁺ cell values, 10× brightfield micrographs of selected region (ventral region or dorsal region) were analyzed using ImageJ (at least four images per region per animal). To costain Golgi-Cox and CD4 immunohistochemistry, 50 µm Golgi-Cox–stained spinal cord sections were subjected to immunohistochemistry with extended incubation times. Detailed methods are provided in SI Appendix.

CD4⁺ T Cell Count z-Stack Acquisition and Analysis.

Fifteen-micrometer-thick z-stacks (distance between each of 17 images in a stack was 0.975 μm) of halved spinal cord lumbar L4–L6 coronal sections were taken on a Nikon confocal A1 fluorescent microscope using a 20× objective. In all analyses, six halved spinal cord lumbar coronal sections were analyzed based on selected areas per tissue section. Region of interests (ROIs) were manually traced using NIS-Elements software (Nikon) by a single-blinded experimenter. CD4⁺ T cells per selected region were manually counted by the same blinded experimenter in serial view. Additional details are stated in SI Appendix.

Immunohistochemistry and Antibodies.

Animals were anesthetized at 28 dpi (unless otherwise mentioned) by isoflurane and intracardially perfused with phosphate buffer saline (PBS) followed by 4% PFA. Lumbar spinal cord was carefully removed and postfixed in 4% PFA for 24 h at 4 °C. Tissue was then submerged in 30% sucrose solution for 12–24 h before being embedded and snap-frozen in optimum cutting temperature gel on pulverized dry ice. Sample was stored at −80 °C until it was cut into 30-µm coronal sections (Cryostar NX50). Additional details are stated in SI Appendix.

Visualization and Image Analysis.

Images of half spinal cord lumbar coronal sections at L4–6 were taken on a Nikon confocal A1 fluorescent microscope using 20× objective with numerical aperture of 0.75. All pictures in each set of micrographs were acquired using the same laser intensity and gain values. In all analyses, five halved spinal cord lumbar coronal sections were analyzed based on selected areas per tissue section (two per section, three sections per mouse). Measurement of integrated density was obtained from selecting ROI on raw images followed by the measurement of integrated density/ROI area using National Institutes of Health ImageJ software. For analysis of IBA1+ cell morphology, cell soma and roundness were conducted in multiple 20× fields (two per section, two sections per mouse) using the ImageJ MorpholibJ plugin (40, 65). Grayscale attribute filtering was applied with Operation set to “Opening,” Attribute set to “Area,” Minimum value set to “25 pixels,” and Connectivity set to “8.” Morphological filter was subsequently applied with Operations set to “Opening,” Element set to “Octagon,” and Radius (in pixels) set at “1.” A standardized threshold was applied across all ROIs. Measurements of “soma area” and “roundness” were obtained using “analyze particles” with measurements set to include “area, shape description, fit eclipse.”

Ex Vivo Time-Lapse Imaging of T Cells in Lumbar Spinal Cord.

Mice were anesthetized using isoflurane, and lumbar spinal cords were excised distally. Excised spinal cord tissue from LckCre-Tomato^fl mice was immediately placed in 4 °C DMEM with 10% fetal bovine serum (FBS) to prevent hypoxia. Spinal cord was cut (2 mm thick) coronally at L4 and L6 using a box cutter, encased in 1% low-melting-point agarose solution in PBS, and mounted on the glass bottom of a 25-mm glass button dish. Prewarmed DMEM media at 37 °C with 1% FBS was added to the dish. T cells were visualized with a confocal system (Nikon A1) using TRITC excitation. XYZ-stacks were typically collected using a 10× objective within a scan field of 512 × 512-pixel resolution and a z-plane distance of 2 μm for 40 min. Nikon NIS Elements data files were processed in ImageJ. A maximum-intensity projection of each z-stack was generated. TrackMate plugin (66) for ImageJ was used to perform single-particle tracking of T cells over a 40- to 45-min interval. Detailed method is provided in SI Appendix. To confirm T cell viability in spinal cord tissue in ex vivo study, we used Live/Dead fixable blue dead cell stain kit (Thermofisher) to stain dead cells. As positive control for dead cell staining, we cultured spinal cord tissues in PBS for 3 d, then stimulated T cells with high doses of αCD3 (10 µg/mL) and αCD28 (30 µg/mL) for 3 h.

Statistics.

Sample sizes, numbers of animals, and factors used for statistical evaluations are indicated in figure legends and Methods. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those generally employed in the field (29, 67, 68). Datasets comprising two groups for comparison were tested for normal distribution (Shapiro–Wilk test, Kolmogorov–Smirnov test), and statistical significance using unpaired two-tailed Student’s t tests for parametric data, or Mann–Whitney u tests for nonparametric data, as detailed in figure legends. Statistical analyses and graphical presentations were computed with GraphPad Prism software (GraphPad). All bar graphs are presented as the mean ± SEM.

Data Availability

All study data are included in the article and/or supporting information.