The pathology of central nervous system inflammatory demyelinating disease accompanying myelin oligodendrocyte glycoprotein autoantibody

Myelin Oligodendrocyte Glycoprotein (MOG) is located on the surface of central nervous system (CNS) myelin and oligodendrocytes and is a marker of oligodendrocyte maturation [6]. It may act as a cell adhesion molecule, a regulator of microtubule stability, and a mediator of myelin and immune interactions [32]. MOG is a part of the immunoglobulin superfamily with an extracellular domain, a transmembrane domain, a cytoplasmic loop, and a cytoplasmic tail [32, 55]. Its location on the surface of myelin led to initial speculation of its immunogenicity in multiple sclerosis (MS) but results were inconsistent, reflecting detection of non-conformational MOG epitopes and older generation assay techniques (enzyme-linked-immunosorbent-assay [ELISA]) [3]. Initial work using assays with MOG in its native conformational form gave the first clue that MOG-IgG may represent a biomarker of an inflammatory CNS demyelinating disease distinct from MS [53]. This was followed by the use of improved cell-based assay techniques using HEK 293 cells transfected with the correctly folded MOG. MOG-IgG has been shown in multiple studies from independent researchers to be a specific serum biomarker of a subset of central nervous system (CNS) inflammatory demyelinating diseases (IDD), clinically distinct from MS and aquaporin-4 (AQP4) antibody (Ab) seropositive neuromyelitis optica spectrum disorder (NMOSD) [9, 29, 34, 45, 51, 53, 62, 63]. Clinical attacks are typically more severe than in MS, with relapses often rendering a patient blind from optic neuritis (ON), wheelchair-dependent from transverse myelitis (TM) or encephalopathic from an ADEM-like presentation [7, 15, 45]. However, recovery is better than with AQP4 Ab seropositive NMOSD [15]. MOG-antibody associated disorders (MOGAD) may follow a relapsing or monophasic course. Children are predisposed and more likely to have a monophasic course [23, 29, 31, 34, 76]. Clinical diagnostic criteria have been proposed requiring one or more of ON, TM, ADEM or other brain or brainstem involvement suggestive of demyelination, MOG-IgG seropositivity and exclusion of other etiologies [45]. The pathophysiology of MOGAD has not been well elucidated. Currently, pathological studies are limited to case reports, some of which have suggested an overlap with pattern II MS pathology characterized by deposition of complement and immunoglobulin in active lesions [13, 21, 25‐27, 35, 66, 67, 75]. Defining the immunopathologic features of MOGAD is necessary to improve understanding of its pathogenesis, which is a potential first step towards developing novel treatments. Herein, we describe an international collaborative biopsy and autopsy series of MOGAD.

The MOG Ab is of IgG1 isotype and positivity was confirmed and end-titers determined using serum samples analyzed with a live-cell-based fluorescence activated cell sorting (FACS) assay with full length MOG in its conformational form either at the Mayo Clinic Neuroimmunology Laboratory (Rochester, MN, USA) with a cut off for positivity of an IgG binding index of ≥ 2.5, or the Division of Neuropathology and Neurochemistry at the Medical University of Vienna (Austria) with a cut off for positivity of > 1:160 as previously described [24, 45]. Persistent seropositivity was defined as a positive MOG Ab (≥ 6 months after onset) or serial samples seropositive ≥ 6 months apart.

Clinical details were collected by (R.H., E.P.F., A.S.L.C.) and obtained from the relevant electronic medical record by retrospective chart review at each institution along with clinical records shared by outside facilities. Detailed clinical histories are presented for both autopsy cases. For the biopsy cohort we recorded details of the demographics (age, sex), clinical phenotype of relapses including transverse myelitis [15], optic neuritis [7], ADEM-like brain presentation [45, 81], and NMOSD [79] or combinations thereof. For those with other presentations we included descriptions of their clinical manifestations. Acute and maintenance immunotherapy treatments administered were recorded. Disability based on expanded disability status scale (EDSS) score was determined at last follow-up [38]. Cerebrospinal fluid (CSF) was analyzed for white blood cell count, protein and oligoclonal bands. Retrospective review of magnetic resonance imaging of the brain was performed on 12 biopsy cases with available scans, or the written interpretation of the radiologist recorded when not available (4 biopsy cases). Lesion characteristics, location, and pattern of gadolinium enhancement were determined.

Twenty two patients had only a brain biopsy performed (frontal 9; temporal 2; parietal 2; cerebellum 1; fronto-parietal 1; parieto-occipital 1; thalamus 1; brain [not otherwise specified] 5). All 22 patients had an inflammatory central nervous system disorder with detailed clinical information available in 18 (82%). The median age at neurologic symptom onset was 10 years (range, 1–66) and 12 of 22 (55%) were children (< 18 years). In those with exact dates available, the median time from disease onset to biopsy was 7 months, (0–516). The median duration from symptom onset to last follow-up was 43 months (range, 3–516). The demographics, clinical presentation and course, cerebrospinal fluid (CSF) findings and MRI details are summarized in Table 1. Representative MRI scans of biopsied cases are shown in Fig. 1(b–d). The median MOG-IgG titer at Mayo Clinic was 1:100 (range, 1:20–1:10,000) and at the University of Vienna was 1:2560 (range, 1:80–1:5120). Among 14 patients with serial samples, twelve relapsing patients had persistent MOG Ab seropositivity, whereas two patients with a monophasic course were transiently seropositive.

Among 14 patients with details available, treatments utilized acutely included one or more of: intravenous steroids 13; plasma exchange 7; and intravenous immune globulin (IVIg) 6. All 14 treated patients showed some improvement after acute treatment. Eleven of 12 patients with details available received one or more maintenance immunotherapies or immunomodulatory medications including: oral prednisone 7; rituximab 6; mycophenolate 3; azathioprine 2; IVIg 2; cyclophosphamide 2; glatiramer acetate 2; interferon-β1a 2; ocrelizumab 1; natalizumab 1; or fingolimod 1.

Among biopsied cases, the available tissue was distributed as follows: 8 cases had white matter tissue only, 2 had cortex only, 11 had both white matter and cortex, and 1 case had white matter and deep gray matter tissue. Of the total of 20 biopsy cases with white matter tissue available for analysis, 6 cases showed no white matter lesions while 14 showed white matter demyelinated lesions. Multiple stages of white matter demyelination could be present within a given person. Eleven biopsies showed early active demyelination, 1 showed late active demyelination, and 4 showed inactive demyelination. The full spectrum of pathological features observed in the MOGAD autopsies was represented in the biopsy cohort, and summarized in Table 2. Representative examples of biopsied white matter and cortex are illustrated in Figs. 4 and 5, respectively. MOGAD biopsy pathology was dominated by a transitional pattern with perivenous and confluent white matter demyelination seen in 50% of the biopsies and a higher incidence of intracortical demyelinated lesions (64% of biopsies), compared to that seen in MS patients with similar disease duration. Abundant myelin-laden macrophages were present within active demyelinating white matter lesions, but the accumulation of activated microglia at the lesion border, typical of slowly expanding / smoldering white matter MS lesions, was not observed. Within the cortex, microglial infiltration often extended beyond the cortical demyelinating lesion (Fig. 5l) with cortical microglial aggregates seen in one biopsy (Fig. 5m). Variable perivascular and parenchymal lymphocyte infiltration was present in the PPWM as well as in all demyelinating stages, most prominent among early active lesions (Fig. 6). In contrast to MS, [49, 71], lymphocyte infiltrates were dominated by CD4+ T-cells, with a median perivascular CD4/CD8 ratio of 2.89 (range, 1.26–29). All cortical demyelinated lesions contained perivascular and parenchymal lymphocytic infiltrates, most prominent within intracortical demyelinated lesions (Fig. 6). In biopsies containing meninges, meningeal inflammation was topographically associated with the presence of cortical demyelination (Fig. 6). A low to moderate number of granulocytes (both eosinophils and neutrophils) were present across demyelinating stages, including in the PPWM, as well as in cortical demyelinated lesions, most pronounced within intracortical demyelinated lesions (Fig. 6).

Immunopattern classification was interpretable among 8 of 11 early active biopsies. Complement deposition (C9neo antigen) within macrophages and on myelin sheaths, typical of pattern II MS, was present in 8/8 active biopsies, including a single biopsy characterized by an overlap of immunopattern II and pattern III, defined by a preferential loss of myelin associated glycoprotein (MAG) and oligodendrocyte apoptosis (Fig. 4j, k, m, n). Axons were generally intact with few axonal spheroids apart from a single biopsy (Fig. 4l). AQP4 was preserved in all cases and dystrophic astrocytes were absent.

MOGAD pathology is characterized by the coexistence of perivenous and confluent primary demyelination with partial axonal preservation and reactive gliosis in the white and gray matter, with particular abundance of intracortical demyelinating lesions. This occurs on the background of CD4-dominated T cell and granulocytic inflammatory infiltrates. Contrary to classical AQP4-IgG seropositive NMOSD, in MOGAD the expression of AQP4 is preserved.

As described already in the earliest reports of MS pathology [40], the demyelinated lesions arise around focal perivenous inflammatory aggregates as acute plaques, which further expand by a zone of lesional activity surrounding the inactive core (chronic active lesions; [36]). A subset of these chronic active MS lesions are smoldering plaques with a rim of activated microglia at the edge and very sparse macrophages with early myelin degradation products [20, 36], which enlarge slowly over months or years [2, 10]. Thus, chronic active focal white matter lesions are typical for multiple sclerosis pathology and present at all disease stages [20]. However, with the exception of a single lesion with a macrophage rim in a single case (Fig. 3a, e), slowly expanding demyelinated plaques are absent in our current case series of MOGAD cases. The MOGAD cases show an ADEM pattern of demyelination, characterized by widespread and diffuse inflammation and demyelination around small venules, which in some instances form confluent plaques by fusion, but not by radial expansion [81].

The coexistence of both perivenous and confluent demyelination seen within our MOGAD cohort was previously described in a subset of pathologically defined ADEM [80, 81], where it was associated with a greater risk of relapse. Although these may have been cases of MOGAD (30–50% of clinical ADEM cases are MOG-IgG seropositive) [63], the absence of available sera to confirm a diagnosis in this published series precludes making any definitive conclusions and makes comparisons difficult to prior cohorts, in which serological testing was not undertaken.

Cortical demyelination is common among the MOGAD cohort, and, as observed in MS, topographically associated with meningeal inflammation [47]. In addition, subpial cortical demyelination, similar to that seen in MS, is present. Thus subpial demyelination is present only in MS, ADEM, and MOGAD, while it is not found in any other inflammatory, demyelinating, metabolic or neurodegenerative disease studied up to now [17, 52]. However, in contrast to MS, intracortical rather than leukocortical demyelinated lesions predominate. Although two prior MOGAD biopsied cases reported cortical encephalitis without demyelination, this may have potentially been due to a very early stage of the disease process or an effect of treatment prior to biopsy [21, 26].

The CD4-dominated inflammatory infiltrate in MOGAD differentiates this disease from MS, in which inflammatory infiltrates are mainly composed of CD8+ T-cells, as shown in large cohort studies including all disease stages [49, 71]. Higher numbers of CD4+ T-cells in MS cases have so far only been described in single, atypical cases [61]. Furthermore, in MS there are high numbers of B-cells in the inflammatory infiltrates and the CSF, as well as a profound further differentiation of B-cells into plasma cells, which produce antibodies, reflected in the high frequency of oligoclonal bands in MS. This is particularly the case in patients with fulminant disease courses, where B-cells even outnumber T-cells in the lesions [49]. Given that one of the MOGAD autopsies and some MOGAD biopsies were treated with B cell depleting agents (rituximab or ocrelizumab), we are unable to specifically comment on the ratio of B cells and T cells in the lesions. However, given the low frequency of oligoclonal bands in MOGAD, it suggests that most of the pathogenic antibodies (as in AQP4-IgG seropositive NMOSD), come from the blood and only a small proportion from local intrathecal synthesis.

Both autopsies, and all 8 biopsy cases that could be reliably immunopattern classified, demonstrate evidence of complement deposition within active white matter lesions, consistent with prior reports [12, 27]. This is not surprising as MOG-IgG is of IgG1 isotype and capable of activating complement. Although a perivascular pattern of complement deposition is seen in the autopsy cases, the absence of complement deposition on astrocytes of the glia limitans is distinct from AQP4-IgG seropositive NMOSD pathology, where the complement deposition colocalizes to sites of AQP4 expression. These findings suggest an antibody-mediated mechanism in MOGAD, and this view is supported by experimental studies, showing the induction of demyelination by patient derived anti-MOG antibodies after direct injection into the brain tissue [64] or after their injection into the cerebrospinal fluid of animals with T-cell mediated autoimmune encephalomyelitis [68]. However we do not observe a selective loss of the MOG protein, which would only be seen if the antibody strips the outer myelin lamella, leaving the compact myelin intact. This does not occur even in the experimental models of MOG-EAE. It is plausible that when the antibody binds to the outer myelin surface, the myelin is either completely destroyed by complement or by antibody dependent cellular cytotoxicity (ADCC) phagocytosis. A selective loss of the MOG antigen would only occur when the binding of the MOG-antibody leads to endocytic internalization of the antigen into a cell, which still remains viable, albeit functionally compromised. This is a major mechanism by which AQP4-IgG targets the AQP4-antigen [42].

A single MOGAD biopsy case demonstrated an overlap of immunopattern II and III, with the presence of complement deposition, as well as MAG loss. The variable destruction of oligodendrocytes seen in MOGAD pathology, with evidence of oligodendrocyte apoptosis, could contribute to MAG loss, as this occurs in the setting of metabolic abnormalities in the oligodendrocyte unable to maintain its distal periaxonal cell processes, leading to a dying back oligodendrogliopathy [48]. Therefore, MAG loss is not specific to MS, but rather seen in a number of disorders including stroke, progressive multifocal leukoencephalopathy, and MS [1]. Monoclonal MOG Ab 818C5 induces β-tubulin dephosphorylation, retraction of oligodendrocyte processes and alters oligodendrocyte cytoskeletal structure, therefore it is plausible that a humoral process in MOGAD targeting oligodendrocytes could explain the observed overlap of immunopattern II and III in this single biopsy case [11, 16, 50].

The clinical characteristics of MOGAD biopsied patients in this study are consistent with what has been reported in the literature of MOGAD with similar demographics (median age of 10 years, similar proportion of males and females), characteristic clinical phenotypes (ADEM-like presentation and optic neuritis dominating, followed by transverse myelitis and NMOSD), typical CSF findings (a high median white cell count [> 100/µL] and low frequency of oligoclonal bands [< 10%]), clinical course (optic neuritis dominating) and overall good long-term outcome in most biopsied cases (median EDSS of 2 at last follow-up). The autopsy cases also have features that fall within the realm of what has been reported with MOGAD, with seizures and encephalopathy being recognized manifestations, fulminant cases previously reported, and a predominantly cortical encephalitis (as in autopsy 2) [13, 22, 54]. The white and gray matter lesions on MRI and absence of CSF oligoclonal bands in the autopsied cases is also typical [63]; however, the hemorrhagic transformation and fulminant course resulting in brain herniation in autopsy case 2 is atypical and may relate to the concomitant ischemic changes. Confirmation of the MOG-antibody seropositivity in 2 independent laboratories however ensures this was a case of MOGAD.

Heterogeneity in age at onset, disease course (monophasic versus relapsing), therapies utilized, clinical course, pathology and antibody titers in MOGAD patients in this study is consistent with the literature on MOGAD [34], but clear differences remain from MS and AQP4-IgG-seropositive NMOSD that suggest MOGAD is a distinct entity. The wide range in age of onset of MOGAD is similar to what is observed in other defined subsets of CNS inflammatory demyelinating diseases (both MS and aquaporin-4-IgG seropositive NMOSD can impact children or the elderly) [60, 73, 74]. There are however unequivocal major differences between MOGAD and MS in presenting manifestations (ADEM is common in MOGAD but very rare in MS), clinical course (a progressive course is not reported with MOGAD but the majority of MS patients eventually develop a progressive course) [30], frequency of CSF oligoclonal bands (8% of MOGAD in our study [similar to prior reports] [15, 34] versus MS where it occurs in 88%) [14], and there are a number of MRI differences in the spinal cord, brain and optic nerve that have been reported [7, 15, 33]. The response to treatments is difficult to determine in this study as well as prior studies of MOGAD, given their retrospective nature, the heterogeneity of treatments used and the lack of a control group. There is an urgent need for prospective randomized placebo-controlled trials in MOGAD, which may give additional insight into the underlying pathophysiology.

The serologic detection of MOG-IgG is useful diagnostically and the features of the antibody may provide insight into pathogenesis. A recent large study of MOG antibodies in MS showed it is detected in only 0.2%, highlighting its utility in discriminating from MS [8]. While in our study we used two sites for MOG-IgG detection, our multi-center collaborative studies have shown the live-cell based MOG assay cut-offs show consistent results across centers for a non-MS phenotype [77]. The frequency of MOG-IgG and AQP4-IgG coexistence was reported at 0.06% which suggests a distinct immunopathogenesis to MOGAD and AQP4-IgG seropositive NMOSD, rather than MOG-IgG representing an epiphenomenon [37]. The presence of transient seropositivity is associated with a monophasic course in this study and is similar to prior reports [23]. Higher titers have also been associated with an increased tendency towards a relapsing course but evaluation in our study is limited by differing assay cut-offs utilized at the different sites [23, 45].

The majority of MOGAD cases in this series have a relapsing clinical course, and none developed a progressive course at last follow-up. A pathological substrate of progression in MS is the presence of slowly expanding (smoldering) lesions in the white matter [19], which are largely absent in MOGAD. Although our cohort is skewed toward early disease, a secondary progressive course typical of MS rarely if ever occurs in MOGAD [30]. As a subset of smoldering lesions have an iron ring in MRI [2, 10], it would be helpful to determine whether MOGAD patients ever develop iron ring lesions.

From a clinical and radiologic perspective, MOGAD more closely resembles AQP4-IgG seropositive NMOSD, than MS. This is evident by the clinical, radiologic and CSF findings that accompany MOGAD which include a high frequency of bilateral optic neuritis, more severe episodes (frequent blindness/paraplegia), longitudinally-extensive lesions in the optic nerve and spinal cord, lesions adjacent to the 3rd ventricle, central rather than peripheral spinal cord lesions, CSF white cell count often > 50 µL and CSF unique oligoclonal bands rarely detected [15, 29, 33]. However this study shows, despite this clinico-radiologic overlap, the pathology of MOGAD is distinct from AQP4-IgG seropositive NMOSD. The presence of hypertrophic reactive astrocytes, preserved to increased AQP4 expression [67, 82], occasional Creutzfeldt cells and frequent cortical demyelination in MOGAD, are helpful pathologic discriminators from AQP4-IgG seropositive NMOSD, which is characterized by AQP4 loss, dystrophic astrocytes, and absence of cortical demyelination [58, 59]. Our findings support most prior MOGAD pathology cases which similarly reported no AQP4 loss [12]. However, two MOGAD pathology cases reported variable AQP4 loss [66] and a single case with dual MOG-IgG and AQP4-IgG seropositivity also reported AQP4 loss [13].

The pathogenic significance of human anti-MOG antibodies in the induction of demyelination has been shown in several recent experimental studies [55, 64, 68]. Although the pathophysiological mechanisms of MOG-antibody associated inflammatory demyelinating disease in humans are so far not fully established, they have been extensively characterized in experimental models of MOG induced experimental autoimmune encephalomyelitis (MOG-EAE). Circulating MOG antibodies alone do not induce inflammation, demyelination and neurodegeneration in normal rodents [43]. However, MOG antibodies, when directed against a conformational epitope, expressed on the surface of oligodendrocytes and myelin sheaths [4], may trigger demyelination in vitro or in vivo after injection into the cerebrospinal fluid or the central nervous system (CNS) tissue, but require the presence of complement or of cytokines that induce macrophage activation and antibody dependent cellular cytotoxicity [44, 72]. In vivo, circulating anti-MOG antibodies become pathogenic when they reach the brain in the context of inflammation, mediated through MHC Class II restricted CD4+ T-lymphocytes [43]. The pathology in the brain and spinal cord in co-transfer models of encephalitogenic T-cells and anti-MOG antibodies depend upon the quantitative balance between T-cell mediated inflammation and the demyelinating antibodies. Thus, a massive T-cell mediated inflammation in the presence of circulating anti-MOG antibodies results in an ADEM-like inflammatory disease, with perivenous (sometimes confluent) demyelination, while in the presence of a mild T-cell mediated inflammation with high titers of circulating antibodies focal (MS-like) demyelinating plaques are induced [41].

Experimental demyelination by anti-MOG antibodies in the context of T-cell mediated inflammation is accomplished either through complement activation or through antibody dependent cellular cytotoxicity. The extent of complement activation in the lesions depends upon the isotype of the antibodies, but demyelination may also be induced by anti-MOG antibodies with poor complement activation properties through interaction with activated effector cells, such as macrophages [56, 57]. In addition, anti-MOG antibodies can also enhance CD4+ T-cell mediated inflammation through augmented antigen presentation, when inflammation is induced by T-cells recognizing MOG [18]. When inflammation is induced by T-cells directed against other CNS antigens, MOG antibodies trigger demyelination, but do not augment the T-cell mediated inflammation. These data, obtained with monoclonal rodent antibodies, have recently been confirmed in experimental transfer studies using patient derived anti-MOG antibodies [68].

MOG-EAE induced by active sensitization with full length recombinant MOG has shown that the clinical presentation and the topographic distribution of lesions in the central nervous system depend upon the genetic background of the animals. Thus, MOG-EAE in Lewis rats mainly induces an inflammatory disease with limited perivenous demyelination, in BN rats an NMOSD like phenotype, in DA rats large confluent demyelinated lesions in all brain regions and in Wistar rats an inflammatory demyelinating pathology with prominent induction of cortical lesions [69, 70]. Fine mapping of gene regions, associated with these different disease phenotypes, identified genes located in the major histocompatibility complex (MHC) region as the main culprits. However, this does not necessarily mean that these phenotypes are related to differences in antigen presentation, since additional genes such as the MOG gene itself or immune effector genes are also located within this region of the genome [78]. Since MOG is expressed on mature oligodendrocytes, but not on oligodendrocyte progenitor cells, rapid recruitment of new oligodendrocytes occurs in the lesions, which is associated with rapid and complete remyelination [6, 65].

As the MOGAD biopsies reflect the full spectrum of pathological features seen in the autopsy cases, the presence of either perivenous alone, or an overlap of perivenous and confluent white matter demyelination, in the setting of prominent intracortical demyelination on brain biopsy, as well as an inflammatory reaction in which CD4+ T-cells outnumber CD8+ cells in the presence of granulocytes, and preserved AQP4 expression, should prompt consideration for MOG-antibody testing. The pathology is helpful in differentiating MOGAD from other CNS demyelinating diseases. Experimental studies in MOG-EAE reflect the broad range of the clinical and pathological phenotypes of MOGAD, as described in our present study, and provide potential mechanistic explanations for their occurrence in different patients.