Human antibodies against the myelin oligodendrocyte glycoprotein can cause complement-dependent demyelination

In the last decades, several studies reported that autoantibodies against the myelin oligodendrocyte glycoprotein (MOG) are associated with inflammatory demyelinating diseases of the central nervous system (CNS), such as acute disseminated encephalomyelitis (ADEM), monophasic or recurrent isolated optic neuritis (ON) or transverse myelitis (TM), pediatric multiple sclerosis (MS), aquaporin-4 (AQP4)-seronegative neuromyelitis optica spectrum disorders (NMOSD), and N-methyl d aspartate receptor (NMDAR) encephalitis with overlapping demyelinating syndromes [1‐16]. Although MOG antibodies are rarely found in MS, MOG-seropositive patients show neuropathological features that are similar to MS pattern II pathology with demyelinated lesions, preserved axons and complement-mediated pathology [17‐20]. MOG is a very well-studied antigen in experimental autoimmune encephalomyelitis (EAE), which is often used as a model for MS. These animal models have been developed using active immunization with MOG, transfer of MOG-specific T cells and antibodies, and mice transgenic for MOG-specific T and B cell receptors [21‐27]. In EAE, it was demonstrated that MOG antibodies are strongly pathogenic only after inflammation and blood-brain barrier leakage induced by encephalitogenic T cells. Co-transfer of the mouse monoclonal anti-MOG antibody 8-18-C5 leads to severe complement-dependent lesions in a myelin basic protein (MBP)-induced EAE model in Lewis rats [21, 22, 28] and in vitro brain cell cultures [29].

Nevertheless, the effects of human MOG (hMOG) antibodies and their exact role in demyelinating diseases are still not fully understood. While hMOG antibodies can initiate both complement-dependent cytotoxicity (CDC) and antibody-dependent cellular cytotoxicity (ADCC) in an Fc-dependent manner [30‐34], in vitro studies have demonstrated MOG antibody binding causes disturbances of thin filaments and microtubule cytoskeleton in oligodendrocytes [35] and cellular cytotoxic effects [2] independent of effector function. Previous in vivo studies have indicated that MOG antibodies trigger the activation of MOG-specific T cells by facilitating opsonization and accumulation in antigen-presenting cells in the CNS and periphery. This process then fosters T cells to recognize their autoantigen and activate in an Fc-dependent manner [36, 37]. Administration of hMOG antibodies into a MBP-T cell-mediated EAE model in Lewis rats resulted in minor demyelination and axonal loss in the absence of excessive complement activation [38], and intracerebral injection of hMOG-IgG into murine brain resulted in complement-independent reversible demyelination and axonal loss [39].

It has become increasingly clear that only MOG antibodies recognizing correctly folded and glycosylated MOG are disease relevant. Although MOG is highly conserved between different species, previous studies have shown that single amino acid exchanges between mouse MOG (mMOG) and hMOG led to a loss of binding ability of most hMOG antibodies to mMOG [40]. These experimental observations and the polyclonal nature of serum IgG reinforce that not all hMOG autoantibodies will be reactive to rodent MOG [40, 41]. To our knowledge, these species-specific recognition patterns have not been systematically evaluated when analyzing the pathogenicity of human MOG antibodies. Therefore, our study aimed to investigate whether hMOG autoantibodies carefully selected for binding to both hMOG and rodent MOG can induce demyelinating pathology in rodents. We screened hMOG-reactive patient sera for reactivity to mMOG and rat MOG (rMOG), analyzed binding patterns in different brain tissues, and selected a set of high and low titer IgG samples with different binding patterns to rodent MOG to perform ex vivo and in vivo studies for demyelinating activity.

Only a limited fraction of human MOG antibodies reveal reactivity to either mMOG or rMOG by CBA. Reactivity to mMOG (60%) was more frequently seen than rMOG (18%). The present investigation is the largest study on species-specific reactivity of human MOG antibodies conducted so far, confirming results of prior studies with a smaller sample size [40, 41]. Similarly, only a subset of hMOG autoantibody-positive samples was reactive on mouse and rat brain tissue. Only 52% of the mMOG-positive samples but 100% of the rMOG-positive samples (CBA, cutoff ≥ 1:160) showed a positive binding pattern to mouse and rat myelin tissue, respectively. In contrast, 88% hMOG-positive samples bound to human brain sections. Differential binding to mouse, rat and/or human tissue was strongly influenced by antibody titers to mMOG, rMOG, and hMOG. Specificity of the tissue binding was confirmed by pre-adsorbing serum samples with MOG-transfected cells. In case of a positive tissue binding with hMOG-, mMOG-, or rMOG-negative samples, the staining signal could not be diminished by pre-adsorption. In these cases, we cannot differentiate between low-level MOG binding, binding to alternative myelin antigens, or background signal. Staining could be produced by MOG antibodies with epitope specificity to epitopes revealed only after tissue fixation. Conversely, the loss of myelin staining observed for a number of hMOG-positive samples is likely due to the loss of conformational epitopes due to tissue fixation and dehydration. The presence of autoantibodies to alternative myelin and neuronal targets have been reported in the CSF of MS patients [49, 53], and these antibodies may contribute to tissue reactivity both in hMOG-negative and hMOG-positive samples.

The differences in antibody binding to hMOG, mMOG, and rMOG are surprising considering the rather small differences in the amino acid sequences of the three proteins (Additional file 7). A previous study has already established the importance of the C-C′ loop in the recognition of human-specific MOG antibody responses [40] which might also be responsible for the differences in antibody recognition seen in our study. In contrast, antibodies also reactive with mMOG and probably also rMOG mainly recognize the F-G loop, which is also the epitope recognized by the monoclonal antibody 8-18-C5. Further, it is still unknown whether there are differences in the affinity of human antibodies to different MOG epitopes. Whereas it is well known that the affinity of the monoclonal antibody 8-18-C5 is higher than that of other murine MOG antibodies [31], the affinity of human MOG antibodies has not yet been analyzed because human monoclonal MOG antibodies needed for methodological reasons are not available yet.

Serum IgG from only one adult NMOSD patient with high titers to mMOG, rMOG, and hMOG caused obvious myelin loss as evidenced by MBP degradation and signs of axonal disturbances and swellings. Interestingly, this patient had the highest titer anti-MOG IgG of subclass 3 (Additional file 5). Demyelinating effects were only shown by the addition of human complement and could not be replicated without complement. This rare pathology contrasts with other patient samples as well as previous studies that showed minimal complement-mediated pathology using human MOG autoantibodies [38, 39]. Weak complement activation has also been observed in a relapsing encephalomyelitis patient positive for MOG antibodies with complement activation only present at areas of active demyelination [17]. These studies showed that MOG IgG are only to some extent able to activate the complement pathway and therefore mediating pathologic effects; however, our study revealed a major contribution of complement-mediated pathology in this sample, which is in line with several reports observing complement deposition in brain tissue of patients with MOG antibodies [18‐20]. Furthermore, this single MOG-IgG-positive patient may target a specific epitope that allows efficient complement activation. Although we cannot completely exclude whether additional autoantibodies in this sample are responsible for the observed pathology, pre-adsorption with hMOG lead to significant reduction of myelin staining, indicating minimal contribution of other autoantibodies against myelin.

While antibodies to MOG are mainly from the IgG1 subtype and therefore able to activate complement in vitro [31, 54] and in vivo [31], the observed effects of hMOG autoantibodies are typically minimal and reversible [38, 39]. Two further antibody injection experiments in rodents did not lead to pathogenic effects in connection with human IgG positive to MOG (A. Saiz, personal communication). In contrast, the monoclonal MOG antibody 8-18-C5 causes severe demyelinated areas in the organotypic brain model [55]. Hence, it appears that MOG antibody-mediated effects in ex vivo mouse models are strongly dependent on high titers to rodent MOG and the high affinity recognition of specific MOG epitopes that facilitate complement activation. Importantly, a recent study has demonstrated that two samples reactive with rodent MOG were able to induce ADCC and CDC in MOG and MBP EAE after injection in the cerebrospinal fluid (Edgar Meinl, personal communication).

In a next step, we were interested whether the anti-MOG antibodies demyelinating in vitro could also cause demyelination in vivo, by antibody-transfer in MBP-specific T cell-induced EAE. This model has already proven to be a suitable animal model to analyze human autoantibody-mediated effects in vivo [28, 56]. In this context, adoptive transfer of MBP-specific T cells trigger cellular and humoral CNS autoimmunity by activating the complement machinery and Fc gamma III receptor-positive cells, thereby boosting ADCC- and CDC-mediated lesions in an γ-IFN dependent way in Lewis rats [52]. However, none of the hMOG-positive samples showed any pathological effects in the EAE model, including sample MOG 7 which was reactive to rMOG and caused demyelination in murine organotypic brain culture. One potential explanation for this discrepancy could be low antibody binding affinity in the in vivo rat EAE model. Since antibody titers in rat serum to hMOG and rMOG were relatively low compared to the titers seen in the patients, the concentration of MOG antibodies may have been insufficient to induce pathologic effects. To account for the possibility of an insufficient concentration of antibodies, we also employed a fourfold higher amount of h/rMOG-positive IgG (40 mg); however, we did not observe an increase in demyelination. Similar to the results observed in cerebellar slice cultures, the monoclonal MOG antibody 8-18-C5 applied in milligram amounts led to substantial myelin loss, macrophage activation, and complement deposition.

In addition to CDC and ADCC, human MOG antibodies have recently been shown to trigger disease severity in an EAE model due to the activation of MOG-specific T cells by fostering autoantigen recognition via antigen-presenting cells. MOG antibody-mediated demyelination was not observed in these studies [36, 37]. Thus, hMOG antibody-mediated CDC may not play a significant role in this rodent model. To conclude, the important question whether human MOG antibodies are pathogenic in vivo is still unsolved. Since the complement-mediated demyelination is difficult to study in mice [57], a human MOG transgenic rat model might be needed to analyze effector mechanisms of human MOG antibodies. This would be also very important because cross-reactivity to rodent MOG is not necessarily related with pathogenicity per se. We still cannot exclude, whether MOG antibodies solely reactive to human MOG could also induce demyelination in an animal model more resembling the human species. However, this study showed that cross-reactivity to rodent MOG is a very important factor to analyze potential MOG antibody-mediated effector mechanisms in rodent models.

Our study has several limitations. First, the majority of patients were children, and most of them had a benign monophasic disease course. Only very few adult patients with severe disease courses were included. Indeed, the only sample able to induce pathogenic effects in the organotypic brain slice model was from an adult patient. Second, the purification and injection of total patient IgG may have resulted in insufficient MOG-IgG concentration to induce pathological effects ex vivo and in vivo. One possible approach to circumvent problems with minor antibody concentrations is to directly inject those antibodies intrathecally or intracerebrally. High efficacy has been shown recently, by injecting low doses of 8-18-C5 and complement into the corpus callosum of mice leading to complement-dependent demyelination [58]. Third, as already mentioned above, cross-reactivity to rodent MOG is not necessarily related with pathogenicity in rodents per se and the pathogenic epitope might only be present in human MOG. Forth, we cannot totally exclude, whether additional autoantibodies may contribute to pathology in MOG-seropositive diseases. Future studies may require purified MOG autoantibodies or recombinant antibodies derived from MOG-reactive human plasma cells. Finally, only a subset of samples could be analyzed for the in vivo pathology due to limited amounts of patient material.

Only a small proportion of antibodies to hMOG are reactive to rodent MOG as analyzed by CBA and tissue IHC. While MOG mAb 8-18-C5 caused vigorous complement-mediated demyelination ex vivo and in vivo, serum IgG from only a single MOG-seropositive patient with high-titer antibodies to mMOG and rMOG was able to induce complement-dependent tissue injury in an ex vivo organotypic brain slice model. The results suggest that high-titer species-specific MOG autoantibodies recognizing unique epitopes on native MOG may be mandatory for CDC in vivo. Future studies should therefore use hMOG antibody-positive samples with high titers and affinity to rodent MOG epitopes or monoclonal MOG antibodies with defined epitope recognition and affinity before testing them in vivo for pathogenic potential.