Meningeal B Cell Clusters Correlate with Submeningeal Pathology in a Natural Model of Multiple Sclerosis

Multiple sclerosis (MS) is a neuroinflammatory disease characterized by demyelination and axonal loss that can lead to major neurologic impairment (1). MS onset is more prevalent in young adults, especially women (2). Most patients initially experience relapsing-remitting MS (RRMS) characterized by peripheral leukocyte infiltration that leads to the formation of perivascular demyelinating lesions (3). A smoldering course often emerges termed secondary progressive MS (4). About 10% of patients progress from onset without obvious relapses (referred to as primary progressive MS) (5). The progressive injury in both secondary progressive MS and primary progressive MS is thought to be mediated by chronic CNS-compartmentalized inflammatory and degenerative mechanisms. The overall MS spectrum is pathologically characterized by leukocyte infiltration associated with oligodendrocyte damage, demyelination, and axonal injury (as the pathologic consequence of relapse biology), as well as injury in the subpial cortex, intracortical gray matter, and leukocortical areas (considered major substrates for progressive MS) (6, 7). The submeningeal cortical injury has been associated with the presence of adjacent meningeal immune cell infiltrates containing T cells, myeloid cells and numerous B cells/plasma cells (8, 9). Such B cell–rich meningeal infiltrates have been shown to correlate with accelerated progressive disease (10, 11). Additionally, these leptomeningeal aggregates can display the arrangement of a tertiary lymphoid organ (TLO)–like structure (8, 9). TLOs are abnormal lymph node–like formations arising at sites of lingering inflammation, such as chronic infection, and autoimmune-related diseases including MS (12). TLOs constitute satellite lymphoid structures able to promote inflammatory processes that in MS are thought to sustain the intrathecal differentiation, maturation, and persistence of pathogenic B cells (13). As B cells can remain for years to decades within the inflamed CNS, their contribution to propagating neuroinflammation has recently gained large attention, including through their interaction with T and myeloid cells as well as their established Ab-independent functions (14).

Similar to MS, granulomatous meningoencephalomyelitis (GME) is an idiopathic spontaneous neuroinflammatory disease with female predilection occurring in dogs of young and middle age (15–17). The distribution of inflammation can be disseminated, focal, or confined to the optic nerve, and the localization of inflammation correlates with presenting clinical signs (18, 19). Clinical onset can be acute, though disease progression can be delayed with immunosuppressive therapy (18, 20). Antemortem diagnosis of GME is difficult and is typically made by excluding other etiologies and by a combination of imaging and clinical pathologic diagnostics. Cerebrospinal fluid analysis of pleocytosis with >50% mononuclear cells combined with negative serology for infectious diseases, is supportive of a GME diagnosis (20). Magnetic resonance imaging (MRI) is frequently used to localize areas of inflammation, which are characterized by single or multiple hyperintense lesions with irregular borders, as well as variable enhancement of parenchyma and/or meninges following gadolinium administration (19, 20). Histologically, GME has been characterized by inflammation within the meninges and white matter of the forebrain, brainstem, and/or spinal cord (16, 21). Lesions are composed of multifocal to regionally diffuse perivascular cuffs of macrophages, lymphocytes, and plasma cells (18, 22). Inflammation is predominant in the white matter but can extend into the gray matter (23). Although leptomeningeal immune cell infiltration has been noted (24), little is known about the nature of the immune cells, and their association with neuropathological changes.

Numerous studies have demonstrated that clinical and pathological manifestations of certain neurologic diseases are similar in humans and domestic animal species (25, 26). Because of this, the potential of validating novel therapeutics in canine clinical trials prior to or in tandem with clinical trials in humans is gaining widespread recognition and participation (27, 28). Given the genetic and immunologic similarities between humans and dogs, along with advancements in analytical techniques for canine tissues, the list of therapeutics validated by these types of comparative clinical trials applied in spontaneous, natural models of human disease continues to grow (29, 30).

We hypothesize that GME recapitulates certain immunological and neuropathological processes of MS, particularly at the leptomeningeal level, and as such may serve as a useful natural model of progressive disease. The GME cases studied here displayed areas of both focal and disseminated leptomeningeal enhancement (LME) on imaging. These correlated with heavy leptomeningeal infiltration characterized by large collections of B cells displaying lympho-organogenic features, properties not found in parenchymal infiltrates. Notably, the leptomeningeal B cell infiltration correlated with areas of submeningeal cortical demyelination emulating pathological aspects of progressive MS. As a whole, these findings suggest that in GME, leptomeningeal inflammation drives pathology within the underlying neuroparenchyma and indicates GME might serve as a novel, naturally occurring model to study compartmentalized neuroinflammation and the resulting neuropathology seen in progressive MS.

The Section of Neurology of the Ryan Veterinary Hospital of the School of Veterinary Medicine, University of Pennsylvania (Penn Vet) identified potential GME cases based on presenting clinical signs, cerebrospinal fluid analysis, and diagnostic MRI. The anatomic pathology database was also searched from January 2008 through December 2018 for cases with multifocal brain disease with a diagnosis including GME. Thirteen GME cases with cortical leptomeningeal inflammatory infiltrates not contiguous with invasive inflammatory lesions of an initial 27 cases of GME were included in this study (Table I). Ten dogs had MRI performed as part of their diagnostic workup, and six of these dogs had magnetic resonance images available for review. Postcontrast LME was scored on a scale of 0 to 3, with 0 = no enhancement, 1 = mild enhancement seen in normal animals, 2 = moderate increase in enhancement, 3 = severe increase in enhancement. T1-weighted (pre- and postcontrast), T2-weighted, and fluid-attenuated inversion recovery images were evaluated. Additionally, MRI was available for review from two cases each of necrotizing meningoencephalitis (NME) and canine distemper virus (CDV). Formalin-fixed and paraffin-embedded (FFPE) blocks from 4 cases each of NME and CDV in dogs aged 3–10 y old were evaluated histologically. MRI images and histology were evaluated and compared with the GME cases. FFPE blocks sampled from the cerebrum from four neurologically normal dogs spanning the same age range (3–11 y old) of GME cases were used as comparative controls for immunohistochemistry (IHC) analysis that is described below.

Tissue samples were fixed for at least 24 h and FFPE blocks were sectioned 5-µm thick, mounted on glass slides, and stained with H&E. Six to nine brain and spinal cord sections were examined from each case, including sections from the level of the cerebral hemispheres and lateral ventricles, hippocampus and thalamus, midbrain, cerebellum, and spinal cord. For evaluation of leptomeningeal inflammatory infiltrates and demyelination, rostral cerebral sections from the level of the frontal lobe to the hippocampus at the interthalamic adhesion, dependent upon which FFPE blocks were available for each case, were sampled.

To analyze markers sensitive to formalin fixation, frozen blocks of CNS tissue were processed as previously published (31). In brief, frozen blocks of brain tissue from four GME dogs were sectioned with a Leica cryostat into 8-μm thick slices. Frozen sections were fixed in acetone 10 min at –20°C. All following incubations were done at room temperature unless noted otherwise. Sections were blocked with serum 10% from the species of the secondary Ab for 90 min. For double immunofluorescence staining, the sections were incubated with the primary Abs (Supplemental Table I) overnight at 4°C. The secondary Abs were diluted in 3% serum and incubated for 60 min. Finally, the sections were mounted in mounting media and Hoescht (1/5000) (Thermo Fisher Scientific). For negative controls, the primary Abs were omitted. Immunostaining of FFPE sections was performed as previously described (32). Slides containing 5-μm CNS sections were deparaffinized, hydrated, and placed in Dako Ag retrieval solution to be heated in a pressure cooker for 22 min and allowed to cool for 30 min. After blocking, primary and secondary Abs were added as described above for the frozen sections. Immunostained sections were imaged using an Olympus IX83 set up equipped with a motorized X, Y, Z stage and a spinning disk confocal head (X-Light V2; CrestOptics, Rome, Italy) using a Hamamatsu R2 cooled CMOS camera (Hamamatsu, Hamamatsu City, Japan) operated by the MetaMorph software (Molecular Devices, Sunnyvale, CA). Imaging processing and analysis was performed as previously described (33, 34). Paraffin embedded canine thymus and brain, and frozen tonsil and brain were sectioned and used to standardize the staining of CNS/immunological markers.

Tissue sections were processed for IHC as previously described (32, 35). In brief, cortical areas of all 13 GME cases and four unaffected dogs were evaluated with Abs to CD3 (T cells) (M7254, Ms mAb; Agilent Dako), Multiple Myeloma 1 (MUM1/IRF4) (plasma cells) (M7259, Ms mAb; Dako), CD79b (B cells) (no. 96024, Rb mAb Cell Signaling Technology), Iba1 (macrophages/microglia) (019-19741, Rb pAb; WAKO), and myelin-oligodendrocyte glycoprotein (MOG) (myelin) (CST, #96457, Rb mAb). All of these stains were performed using an LEICA Bond RXm automated IHC stain according to the manufacturer’s recommendations. Isotype control Abs were used as negative controls. Positive controls were normal canine brain tissue and lymph node. All GME and control IHC slides stained immunohistochemically were scanned at 20× using a Leica Aperio AT2 slide scanner (Leica Biosystems, Buffalo Grove, IL) and image acquisition was performed with ImageScope (Leica Biosystems, Buffalo Grove, IL). For the quantification of the MOG staining to assess demyelination, image analysis was performed using a positive pixel count algorithm using ImageScope software. A single algorithm was developed using appropriate thresholds for the staining intensity, and was applied to specific annotated regions on each slide.

Scoring of leptomeningeal and parenchymal inflammation was based on previously described methods (4). We evaluated slides of cerebral cortex stained for B cells (CD79b), T cells (CD3), and used MUM1 to label plasma cells. Leptomeningeal inflammatory infiltrates were scored using a four-tiered scale (absent = no inflammatory cells present; mild = fewer than 20 diffusely distributed inflammatory cells or leptomeningeal perivascular cuffing up to 3 layers; moderate = 20–50 diffusely distributed inflammatory cells or leptomeningeal perivascular cuffing three to seven layers; marked >50 diffusely distributed inflammatory cells or leptomeningeal perivascular cuffing greater than seven layers). Parenchymal cortical inflammation was scored using a similar scale: 0 = absent, 1 = mild (vessels with one cuff), 2 = moderate (many vessels with two cuffs), 3 = marked (scattered or many vessels with >3 cuffs).

Data were analyzed using GraphPad Prism v.8 software. Unpaired student t test was used for comparisons between number of cell populations within leptomeningeal infiltrates and for determining the differential expression of MOG in subcortical areas of controls versus GME cases. Two tailed (95%) Pearson correlation analysis was used to determine the association between the percentage of MOG expression (demyelination index) and distinct neuropathological parameters including leptomeningeal B cell accumulation.

T1W hyperintensity following gadolinium administration is indicative of a break in the CNS barriers from various underlying causes including neovascularization, inflammation, cerebral ischemia, and CNS pressure changes (36). In GME, MRI has become an important tool to diagnose the disease when combined with cerebrospinal fluid analysis and neurologic assessment (18). In our cohort, MRI of six dogs were available for review by a board-certified veterinary neurologist (C.H.V.). We found dogs with GME had mild to strong contrast enhancing, poorly delineated and ill-defined lesions involving the white and gray matter (Fig. 1), as has been previously described (37). However, we also noted the leptomeninges commonly displayed areas of focal or multifocal contrast enhancement in 6 out of 10 of the dogs examined, which has not been commonly reported in GME (Fig. 1, Table I). The brains of two dogs each with NME and CDV, imaged during the same time period, exhibited no LME (data not shown). Histological evaluation of the contrast enhancing lesions confirmed correlative inflammatory lesions in the leptomeningeal space and also in the neuroparenchyma. In general, the intensity of LME was increased in GME cases with more severe leptomeningeal inflammation; however, the limited number of cases with imaging available for enhancement scoring (6 out of 13 cases) precluded statistical analysis of this trend. This relatively consistent finding of LME led us to evaluate both the neuroparenchymal and leptomeningeal infiltrates, as well as the associated neuropathological changes in the abnormal regions detected with MRI in the GME cases.

We examined 13 cases demonstrating the classic histologic changes characteristic of GME. Brain sections, including leptomeninges, were evaluated by two board-certified veterinary anatomic pathologists (M.E.C., M.D.S.). As is seen in GME, all cases exhibited cuffing of leptomeningeal and parenchymal blood vessels morphologically characterized by lymphocytes as well as variable numbers of aggregated and/or intermixed mononuclear cells composed of macrophages and plasma cells (Fig. 2). The leptomeningeal clusters ranged from small perivascular aggregates composed primarily of small lymphocytic cuffs (Fig. 2A) to further progressed aggregates, with moderate numbers of lymphocytes admixed with macrophages and plasma cells (Fig. 2B). In the most developed aggregates, large numbers of tightly packed lymphocytes resembling TLO-like structures expanded the leptomeninges with macrophages accumulating peripherally (Fig. 2C).

To address the specificity of the leptomeningeal neuropathology seen in GME, we also examined two spontaneous cases of neuroinflammatory diseases that occur in canine patients and were previously considered potential natural models for MS (17, 38, 39). We found that these two canine diseases, NME and CDV (n = 4 for each), demonstrated different inflammatory patterns and different associated neuropathology. NME featured malacia and loss of cortical gray and subcortical white matter with large neuroparenchyma aggregates of activated macrophages/microglia and minimal to mild perivascular cuffing by lymphocytes (Supplemental Fig. 1A–C). In CDV, there was moderate to marked perivascular cuffing by lymphocytes in the parenchyma with only mild inflammation in the meninges and myelin loss was generally confined to cerebellar white matter (Supplemental Fig. 1D–1F). Thus, the unique histopathological features detected in GME indicate robust immune cell infiltration in the perivascular neuroparenchyma and, specifically, within the leptomeningeal space.

Given the aggregation of lymphocytes within the leptomeninges (Figs 2A–C), we assessed the immune cell composition in this compartment. We found large numbers of CD20⁺ B cells in leptomeningeal aggregates with a distribution and spatial arrangement confirming the histopathological TLO-like phenotype (Fig. 2D). In contrast to the leptomeningeal B cell organization, the perivascular cuffs within the neuroparenchyma lacked the lymphoid-like arrangement and were composed of a smaller proportion of B cells compared with the leptomeninges (Supplemental Fig. 2A). Scattered CD3⁺ T cell infiltration similar to autoimmune-driven delayed-type hypersensitivity responses has been described in GME lesions (22, 24). The leptomeningeal TLO-like aggregates presented significantly lower numbers of CD3⁺ T cells than B cells and were distributed randomly between the B cell aggregates (Fig. 2D). Of note, within the B cell clusters studied we did not observe distinct B cell and T cell areas. CD3⁺ T cells also distributed sparsely within perivascular cuffs in the parenchyma (Supplemental Fig. 2B). To discriminate the type of T cell infiltrating the leptomeninges, we interrogated the presence of cytotoxic T cells. We found that most of the T cells scattered between the CD20⁺ B cells were CD3⁺CD8⁺, in both the leptomeningeal space (Fig. 2E) and neuroparenchyma (Supplemental Fig. 2C), indicating that CD8 T cells are the predominant T cell population over CD4 T cells in GME infiltrates. We could not directly assess the CD4⁺ T cell compartment as none of the Abs tested detected canine CD4. The leptomeningeal TLO-like aggregates in GME were characterized by the presence of few Ki67⁺ B cells (Fig. 2F), indicating a limited proliferation rate and the likely importance of a persistent peripheral pool of B cells to support these structures. In contrast, B cells accumulating perivascularly in the neuroparenchyma lacked Ki67⁺ expression (Supplemental Fig. 2D). These findings indicate that B cells represent an important component of GME immunopathology and that the leptomeningeal compartment is conducive to their aggregation as seen in MS.

In MS, leptomeningeal B cell–rich clusters vary from rudimentary clusters of cells to relatively organized TLO-like structures (9). In these GME cases, we found a similar gradation of leptomeningeal B cell aggregation (Figs. 2A–C). Highly organized TLOs feature segregated T and B cell zones containing germinal center (GC) B cells, follicular dendritic cell (FDC)/follicular reticular cell (FRC) networks, high endothelial venules (HEVs) and a reticular network supporting the homing and activation of T and B cells (40). GC B cells are characterized by the expression of the markers GL7 (41) and BCL6 (42). As we were unsuccessful in staining canine tissues with these markers, we relied on CD81 for the identification of GC B cell–like cells (43, 44). Leptomeningeal B cell aggregates displayed CD81 expression within TLO-like aggregates (Fig. 3A), whereas CD81 expression was sparse in between B cells within parenchymal infiltrates (Supplemental Fig. 2E). To address the involvement of FDCs and FRCs, we used CD271, a marker expressed by both cell types in secondary lymphoid organs (45, 46). The leptomeningeal TLO-like aggregates exhibited enrichment in CD271⁺ cells displaying a reticular morphology (Fig. 3B) as compared with parenchymal perivascular infiltrates (Supplemental Fig. 2F). Of note, some of the CD271 enrichment around parenchymal vasculature (blood–brain barrier) might reflect labeling of pericytes (47). Immune cell recruitment at sites of inflammation and their interaction with tissue-resident and stromal cells driving TLO formation promotes the release of CXCL13, CCL19, and CCL21. These chemokines promote the recruitment and spatial organization of immune cells into TLOs (12). The B cell chemoattractant CXCL13 was prominently expressed in the TLO-like leptomeningeal aggregates within the B cell clusters and meningeal blood vessels (Fig. 3C). Likewise, the T cell and dendritic cell chemoattractant CCL21 was also expressed within those aggregates and by meningeal vessels (Fig. 3D). These findings highlight the important role of the blood–meningeal barrier (BMB) in supporting immune cell aggregation in the leptomeninges.

The development of HEVs is a defining feature distinguishing organized TLOs from other types of inflammatory infiltrates given their critical role in supporting naive cell migration (48). HEVs specifically express peripheral node addressin (PNAd), a molecule that serves as the ligand for l-selectin/CD62L, and normally absent in other types of blood vessels even within lymphoid organs (49, 50). In GME, PNAd expression appears to be luminal and restricted to small vessels (IsoB4⁺) within leptomeningeal TLO-like aggregates (Fig. 3E). These HEVs display a staining pattern consistent with the well-described plump morphology of mature HEVs (51) and were mostly localized in the periphery of the B cell aggregates as previously reported (52). The organized spatiotemporal distribution of B cells and other cellular components making up TLO-like structures depends on the discrete arrangement of extracellular matrix components (53). Analysis of collagen type IV demonstrated an organized stroma interweaving and extending in between immune cells and presenting a radiocentric pattern displaying higher matrix deposition in proximity to meningeal blood vessels (Fig. 3F). This structural conformation is similar to the bundled and aligned collagen conduits present within lymph node follicles and TLO-like structures in the periphery and in the CNS (54–56). Despite the technical limitations in studying specialized populations (GC B cell–like cells and FDCs/FRCs) associated with chronic inflammatory processes, our findings indicate the formation of TLO-like aggregates in the leptomeninges of GME dogs and suggest these structures might exacerbate compartmentalized neuroinflammation.

As meningeal inflammation in patients with MS has been associated with submeningeal cortical demyelination, neuroaxonal loss, and an accelerated and progressive course of MS (4, 7, 57), we evaluated leptomeningeal immune cell infiltrates and subjacent regions of cortical gray matter. We found that the leptomeningeal inflammatory cells were distributed diffusely (several individualized cells loosely arranged) or in a focal manner (dense aggregates of cells) in all the GME cases studied (Fig. 4A). Most GME cases (n = 10) exhibited both focal and diffuse immune cell infiltration, whereas two cases demonstrated only a diffuse pattern and one case presented focal immune cell infiltration without diffusely distributed cells (Fig. 4A). Whereas the focal aggregates were composed of B cells, T cells, and plasma cells (Figs 4B–D), the diffuse infiltrates contained mainly B cells with scattered T cells and only rare plasma cells (Supplemental Figs 3A–C), mirroring the compositional immune findings shown in (Fig. 2. Within the cortical gray matter underlying these leptomeningeal inflammatory aggregates, we defined the degree of demyelination by evaluating the expression of MOG, one of the main structural components of myelin (58). We examined the level of myelin expression in subpial, intracortical, leukocortical, and subcortical white matter regions relative to sections from control dogs (Figs 5A, 5B). We detected decreased myelin expression in the subpial and leukocortical regions when considered individually (p = 0.007 and 0.008, respectively) and a notable trend in the intracortical region (p = 0.053) (Supplemental Fig. 3D). In contrast, the expression of myelin in the subcortical white matter did not differ between GME and controls (Supplemental Fig. 3D). The overall percentage of myelin staining in the combined regions of cortex underlying leptomeningeal aggregates was significantly decreased in GME relative to controls (p = 0.005) (Fig. 5C). Thus, we demonstrated submeningeal demyelination in regions of the cerebral cortex previously described as affected in MS (4, 11, 59). Notably, we found a significant correlation between the severity of B cell leptomeningeal aggregates and the extent of cortical demyelination (Fig. 5D), further reflecting the remarkable neuropathological similarities between GME and MS. We also noted that cortical blood vessels within the parenchyma were often cuffed by lymphocytes and macrophages, similar to what is seen in MS (60), though their presence did not correlate with cortical demyelination (Fig. 5E). Similar to leptomeningeal inflammation, B cells tended to predominate relative to T cells within these perivascular cuffs (Fig. 5E). Additionally, we found evidence of cortical glial nodules characterized by aggregates of activated macrophages/microglia (Fig. 5F) as described in MS (4, 61). However, the presence of these nodules were not correlated with cortical demyelination (Fig. 5F). These findings indicate that in GME, leptomeningeal B cells play a role in the underlying neuropathology seen in the cortex as it has been postulated in MS. This demonstrated relationship makes GME a compelling model replicating definitive aspects of MS neuropathology.

GME has long been considered a T cell–mediated autoimmune disease of unknown origin (22, 62). In this study we present evidence supporting the importance of B cells to the pool of leukocytes infiltrating the CNS, particularly at the leptomeningeal level. MS and its animal models, like experimental autoimmune encephalomyelitis (EAE), were similarly long considered to be T cell–mediated diseases (6). However, the ability to substantially limit new disease activity in MS using anti–B cell therapies has highlighted the important contribution of B cells to neuroinflammation (63–65). In addition to the contribution of peripheral B cells to MS relapses, their long-term persistence in leptomeningeal infiltrates is considered a potential driver of progressive MS (4, 8, 9, 11, 66). In this study, postcontrast meningeal enhancement by MRI was present in GME. This pattern of enhancement corresponded with leptomeningeal B cell accumulations, which in turn correlated with submeningeal cortical demyelination and closely resembles that of progressive MS (8, 9, 11). Thus, we propose that B cell–rich meningeal immune cell aggregates may potentiate chronic inflammation in GME and contribute to submeningeal CNS injury, as has been postulated to occur in MS (14, 67, 68).

The pattern of leukocyte infiltration in the GME leptomeninges demonstrated enrichment of B cells organized at times as TLO-like structures. In MS, meningeal infiltrates rich in B cells are associated with accelerated progressive clinical disease as compared with progressive cases lacking these structures (11, 59). Attempts to recapitulate TLO-like development have been carried out using the most common model of MS, murine EAE (54, 69–71). EAE exhibits great variability depending upon the mouse strain, Ag used, as well as immunization protocol, and the observed disease/neuropathology does not recapitulate important features of MS (72, 73). Additionally, although other canine neurologic diseases previously postulated to mimic MS do demonstrate meningeal inflammation (NME) and demyelination (CDV), the specific pattern of inflammation and demyelination found in GME closely mirrors the compartmentalized immunopathology seen in MS. Furthermore, GME affects both the brain and the spinal cord and prevails in young female dogs, as is seen clinically in MS. The consistent pattern of B cell–rich leptomeningeal aggregation that correlates with underlying cortical demyelination suggests that this compartmentalized inflammation exacerbates the clinical course of GME as occurs in MS. A TLO-like pattern of infiltration arises at predefined locations as a result of the structural and chemical support provided by stromal and endothelial cells, which in the CNS, are enriched within the leptomeninges (69, 74–76). Correspondingly, we found that chemokines responsible for immune cell aggregation/organization within TLO-like structures were enriched in the leptomeningeal space in GME. CCL21 and the B cell chemoattractant CXCL13 might sustain the migration of lymphocytes into the leptomeninges across the BMB, a vascular bed more supportive of immune cell interactions than its parenchymal counterpart, the blood–brain barrier (77–81). In addition, the presence of HEVs further supports this process and suggests that naive lymphocytes can migrate into the leptomeninges and TLO-like aggregates. As TLOs are considered focal sites that promote lymphocyte activation, Ag presentation, somatic hypermutation, and class switching of B cells (48, 82), our findings are consistent with the postulation that leptomeningeal TLO/TLO-like formations exacerbate and sustain inflammatory responses within this niche.

The overrepresentation of B cells within the GME leptomeningeal clusters and the relative abundance of CD8⁺ T cells over CD4⁺ T cells challenges the prevailing view that CD4⁺ T helper responses drive the pathogenesis of GME (83) and further highlights immunopathological similarities between GME and MS. As we have found in GME, CD8⁺ T cells also outnumber CD4⁺ T cells in MS lesions (84–86). These similarities contrast EAE in which the immunization protocol evokes T cell responses heavily biased toward inflammatory CD4⁺ T cells with relatively little involvement of CD8⁺ T cells (73, 87). In terms of the inflammatory response, analysis of GME brains indicates high levels of IL-17 and IFN-γ expression as compared with other neuroinflammatory canine diseases (83). Given that most T cells in the GME brains are cytotoxic, we speculate that IFN-γ–producing CD8⁺ T cells (Tc1) and IL-17–producing CD8⁺ (Tc17) cells represent important drivers of disease. This immune phenotype has also been reported in MS in which Tc1 and Tc17 cells are enriched in the cerebrospinal fluid of patients and their intrathecal frequencies correlate with disability (88, 89). Moreover, Tc1 and Tc17 cells are found within MS lesions (90, 91), and aside from IFN-γ and IL-17 production, they can also display a cytotoxic phenotype (92, 93). We postulate that the increased migration of B cells and CD8⁺ T cells across the BMB (79, 80), the stromal meningeal microenvironment (8, 9), and the compartmentalized nature of the inflammatory CD8⁺ T cell–B cell response reported in MS (60), is mirrored by the leptomeningeal immune cell infiltrates present in GME. These findings suggest that CD8⁺ T cells migrate to and accumulate in the meningeal compartment in which B cells then stimulate their activation via cytokines and/or through Ag presentation. These interactions likely drive CD8⁺ T cell effector phenotypes, which in turn induce submeningeal demyelination and axonal pathology. Studies in MS have attributed these neuropathological changes to B cells, Tc1, and Tc17 cells releasing soluble mediators known to be cytotoxic for oligodendrocytes and neurons, resulting in demyelination and neuronal death (4, 94–98). Future studies of GME could aim to fully characterize the cytokine composition present in cerebrospinal fluid and in serum in an effort to establish diagnostic and prognostic biomarkers for GME and MS alike.

In MS, leptomeningeal B cell/TLO-like aggregates are associated with submeningeal cortical lesions (8, 99). Our findings in this study suggest that a similar B cell driven destruction of the underlying cortex occurs in GME. The distribution of meningeal inflammatory cells (in both focal and diffuse patterns) and the associated cortical demyelination noted in the GME cases reviewed in this study, are similar to those described in MS (7, 57, 61, 100, 101). Most prior studies examining pathologic lesions of GME have concentrated on the neuroparenchymal inflammatory infiltrates with little emphasis on meningeal inflammation (62, 83). In this study we demonstrate that the extent of cortical demyelination, considered the major substrate for progressive MS (6, 7), correlates with the severity of leptomeningeal B cell aggregation. Similar to lesions in early MS (4), the subpial lesions in GME are not fully demyelinated, suggesting that GME might be modeling the more acute and/or regenerative process associated with early stages of disease progression. Although this contrasts with the fully demyelinated cortical lesions reported in progressive MS (57, 102), other features of progressive MS (especially the presence of leptomeningeal aggregates of B cells associated with these regions of demyelination) are present in GME. As subpial cortical demyelination correlates with an accelerated and progressive course of MS (4, 7, 57) and because no other MS model replicates this chronic compartmentalized immunopathology (103), GME holds a tremendous potential to advance our understanding of the mechanisms driving disease progression and to develop and test novel therapeutics. In addition to demyelination, subpial cortical neuronal loss is also seen in progressive MS (104), an aspect not evaluated in our GME cohort and to be addressed in future studies. Finally, in two cases, we observed areas in which leptomeningeal infiltrates invade/displace the underlying cortex. This might indicate a more active/acute immune–mediated inflammatory process in which GME could differ from MS or might represent earlier stages of aggressive leptomeningeal driven pathology as has been reported in acute MS (4).

The underlying etiologic trigger(s) of inflammation have not been identified in MS nor in GME. The CD8⁺ T cell representation in both diseases suggests the involvement of a viral trigger. Previous infection with EBV is one of the strongest environmental risk factors linked with MS development (105). An underlying infectious cause for GME has been sought after for decades to no avail (106–108). Canid herpesvirus 1 is a well-known cause of encephalitis in young dogs, and the pathogenesis of latent infection is not well investigated (109). Furthermore, the expression of viral-induced epitopes has been described in GME, suggestive of an underlying infectious cause (110). Dysregulation of latent EBV infection may induce viral reactivation and neuropathology in MS patients (111–113). EBV-transformed B cells preferentially accumulate in the leptomeninges in which they may activate CD8⁺ T cells to promote chronic neuroinflammation (111, 114, 115). The accumulation of B cells and CD8⁺ T cells in GME recapitulates elements of the leptomeningeal immune cell composition observed in MS and EBV infection and may offer a clue to an underlying etiology of GME.

Even though no single model can, by itself, recapitulate the entire spectrum of clinical and pathological features of MS, we believe GME can serve as a naturally occurring model to study the compartmentalized meningeal inflammation and resulting neuropathology seen in progressive forms of MS. Further studies aimed at elucidating the mechanisms driving progressive submeningeal injury may contribute to the development of therapeutic strategies for both GME and MS. Given that GME demonstrates B cell leptomeningeal enrichment that correlates with LME, this constitutes a significant advancement to establish biological proof of principle to target leptomeningeal inflammation. As the dog is often the large animal species used for the final safety and pharmacokinetics evaluation of novel drugs before approval for human use (116, 117), GME could offer more immediate translational benefits in the hope of halting chronic aspects of the disease and ultimately improve treatment options for GME and MS patients.

We thank the anatomic pathology and neurology residents as well as histology technicians of the Penn Vet for help in procuring and processing tissues used in this study. We also thank the clinical veterinary neurologists for contributions in the clinical diagnosis of these canine patients. We are indebted to Brian and Caris Chan for the kind gift of the confocal microscope use in this study to C.H.V. and Penn Vet. Finally, we highlight the expert input from Dr. Michael May (Biomedical Sciences, Penn Vet) regarding immune features of chronic inflammation and from Dr. Zissimos Mourelatos, Dr. MacLean Nasrallah, and Dr. Edward Lee (Perelman School of Medicine at the University of Pennsylvania) regarding feedback on the histopathological characterization of leptomeningeal immune cell infiltrates.

The authors have no financial conflicts of interest.