We describe the largest series to date of MOG-IgG-positive patients with brainstem involvement, thereby expanding the clinical spectrum of MOG-IgG autoimmunity. The 15 cases presented here were identified as part of a large European cohort (n = 50) of MOG-IgG-positive patients reported in part 1  and part 2  of this series, suggesting a relatively high frequency (30 %) of brainstem involvement among patients with MOG-IgG-associated myelitis and/or ON. Similarly, AQP4-IgG-positive NMO, also initially thought to affect mainly the optic nerves and spinal cord, was later reported to involve the brainstem in up to one third of all cases [16, 20, 27, 28]. We detected MOG-IgG by means of new-generation cell-based assays (CBA) employing recombinant full-length human MOG instead of enzyme-linked immunoassays, which are prone to both false-negative and false-positive results and which are no longer recommended for routine clinical diagnosis of MOG antibodies ; CBAs were also used for detecting AQP4-IgG [24, 25]. It is a further potential strength of this study that all patients analyzed were of Caucasian descent. This may be important since genetic factors are thought to play a role in NMO and related disorders .
The cases described here underline our finding from part 2  that MOG-IgG-related CNS autoimmunity is not mostly a mild and monophasic disease, as suggested by some earlier studies with smaller sample sizes and shorter observation periods [4, 6], but can in fact take a relapsing and severe disease course with potentially life-threatening complications: brainstem involvement caused respiratory impairment in two patients in our series and was fatal in one of them. Respiratory insufficiency due to brainstem inflammation has previously been identified as the main cause of death in AQP4-IgG-positive NMO .
Besides central hypoventilation, the patients’ symptoms included dysarthria, dysphagia, impaired cough reflex, sensory impairment due to trigeminal nerve damage, facial nerve palsy, trigeminal hypesthesia, oculomotor nerve palsy and diplopia, nystagmus, internuclear ophthalmoplegia (INO), vertigo, hearing loss, balance difficulties, gait and limb ataxia, and, of particular note, intractable nausea and vomiting (INV).
Brainstem symptoms included intractable nausea and vomiting
INV in patients with brainstem encephalitis is caused by lesions in the dorsal medulla oblongata (area postrema). While presence of an area postrema syndrome (APS) has hitherto been considered to have high predictive value for a diagnosis of AQP4-IgG-positive NMO [15, 31–35], our study demonstrates that area postrema lesions and INV also occur in MOG-IgG-positive patients. However, this is not totally unexpected. First, MOG is expressed throughout the entire CNS and, accordingly, inflammatory lesions were found in almost all areas of the CNS in our cohort ; there is a priori no reason why the area postrema in particular should be spared. Second, it has been speculated that the circumventricular organs, including the area postrema, which lack a proper blood-brain barrier (BBB), may be sites prone to entry of AQP4-IgG into the CNS; assuming MOG-IgG are pathogenic (as suggested by recent studies [2, 14]), this would also be relevant for MOG-IgG.
Inflammation occurred in all areas of the brainstem
Apart from the pons, which was affected in 11/13 (84.6 %) patients, MRI brainstem lesions were most commonly located in the medulla oblongata (8/14 or 57.1 %), in the mesencephalon (lesions in the cerebral peduncles in 2/14 or 14.3 %), and in the cerebellar peduncles (5/14 or 35.7 %).
Additional involvement of the diencephalon and the cerebellum
In several cases brain involvement on MRI was not restricted to the brainstem. Concomitant diencephalic (pulvinar) and cerebellar lesions – and thus rhombencephalitis – were present in 2/13 (15.4 %) and 1/14 (7.1 %), respectively. In case 3, an additional lesion was present in the cerebellum which was accompanied by limb ataxia and marked gait ataxia. Cerebellar gait and stance ataxia was also noted in case 2 as well as cerebellar dysarthria. Similarly, patient 9 in part 2  presented with a (clinically silent) lesion in the cerebellar white matter; however, that patient had no lesions in the brainstem. Cerebellar symptoms have very rarely been described in AQP4-IgG-positive patients, too . However, not all patients with ataxia had cerebellar lesions; some patients in this and in the total cohort  presented with sensory ataxia and/or unsteady gait due to paresis caused by acute myelitis.
Supratentorial brain lesions were present in 10/15 (66.7 %) patients. Lesions were found in the deep (including the periventricular) white matter (including in the corona radiata), the sub-/juxtacortical white matter (including in the insula), and, in a single patient, to some extent the cerebral cortex; furthermore, in the corpus callosum and, in one case, the basal ganglia and thalami (see part 2 of this series for exemplary MRI images ). In patient 3, brain and brainstem lesions were extensive and even resulted in impaired consciousness. Similarly, extensive confluent supratentorial lesions were seen in patient 7. Barkhof’s MRI criteria for MS were met in 4/15 (26.7 %) patients.
Acute brainstem involvement was associated with ON and/or myelitis in most cases
Isolated attacks of brainstem encephalitis, i.e., attacks that were not accompanied by clinical symptoms of ON or myelitis, were very rare in the total cohort of 50 MOG-IgG-positive patients , accounting for only 5/276 (1.8 %) documented attacks. However, they were relatively common if only patients with a history of at least one brainstem attack are taken into account (5/21 [23.8 %] attacks, 4/15 [26.7 %] patients).
In most patients, however, brainstem attacks were at least once associated with clinical myelitis (13/15 [86.7 %] patients), ON (6/15 [40 %]) and/or supratentorial encephalitis. In 5 out of 14 patients (36 %) with a history of simultaneous brainstem encephalitis and myelitis and available MRI data, lesions contiguously stretched from the medulla oblongata into the cervical cord at least once, similar to what can be seen in AQP4-IgG-positive patients. In case 2, inflammation affected almost the entire neuraxis, including the lumbar, thoracic, and cervical spinal cord as well as areas in the brainstem and in the supraventricular white matter. This was associated with exceptionally high MOG-IgG serum titers (1:10,240) and presence of detectable MOG-IgG levels also in the CSF.
Some patients met Wingerchuk’s criteria for NMO(SD)
At last follow-up, 12/15 (80 %) patients with brainstem encephalitis had a history of both ON and myelitis. Two patients had a history of myelitis but not of ON, and one had a history of recurrent ON (rON) but not of myelitis. All but 2 (86.7 %) had a relapsing disease course at last follow-up. Of those 12 patients, 8 (66.7 %) met the 2006 Wingerchuk criteria for NMO ; in the remaining 4 patients, the criteria were not fulfilled due to the presence of brain MRI lesions at onset meeting Paty’s criteria and/or because criteria for LETM were not met. On the understanding that MOG-IgG-positive ON and/or myelitis are not considered “alternative diagnoses”, i.e., based solely on clinicoradiologic findings, 9/15 (60 %) met the revised 2015 consensus criteria for NMOSD.
Brainstem involvement may be asymptomatic
While overall there was a high degree of agreement between MRI findings and clinical presentation, in 5/15 (33.3 %) patients asymptomatic brainstem lesions were detected by MRI at least once, suggesting that subclinical brainstem encephalitis is not uncommon in MOG-IgG-related CNS autoimmunity. The true prevalence of brainstem involvement in MOG-IgG-positive patients may thus be higher than expected based solely on clinical presentation. Similarly, evidence for subclinical involvement of the brain, the spinal cord, or the optic nerves, as detected by MRI or electrophysiology, has been found in some of our MOG-IgG-positive patients (see part 2 of this series for details ).
CSF and MRI findings may mimic infectious CNS disease
In one case, leptomeningeal contrast enhancement was noted at the time of disease onset. This is of particular interest in the light of recent studies indicating the existence of a lymphatic system of the CNS within the meninges . Meningeal involvement has been reported in neuromyelitis optica [38–43], MS, and other disorders of putative autoimmune etiology and may indicate a new path for immune cell entry into the CNS. Moreover, this finding is of potential differential diagnostic relevance, since it may wrongly suggest acute infectious meningitis, all the more as CSF findings compatible with an infectious etiology were found in some patients: CSF pleocytosis was present in 11/14 (78.6 %) patients (50/μl, range 8-360) with available data and comprised neutrophil granulocytes in 7/11 (63.6 %) (accounting for 3-34 % of all white CSF cells) or eosinophil granulocytes in 2/6 (33.3 %), increased lactate levels in 2/3 (66.7 %) patients, BCSFB dysfunction in 5/12 [41.7 %], and CSF-restricted OCB, a mainstay of classical MS, were absent in 10/14 (71.4 %). Together with fever and other signs of systemic infection – disease onset was preceded by acute infections in at least three patients (purulent tonsillitis, ENT infection, and bronchopulmonary infection, respectively) – these CSF findings are compatible with early bacterial or viral meningoencephalitis and thus bear the risk of delayed diagnosis of autoimmune encephalitis, as was the case in one of the patients reported here. Neutrophilic and eosinophilic pleocytosis [20, 44, 45], elevated lactate levels [45, 46], and missing OCB [45, 47] are also features of AQP4-IgG-positive NMO. Elevated neutrophil counts have been previously reported also in MOG-IgG-positive pediatric patients . Moreover, attacks are preceded by acute infection in 20-30 % of cases of AQP4-IgG-positive NMO [20, 49, 50], suggesting that infection-related immunologic changes or infection-related BBB disruption [45, 47] may trigger disease activity.
Post-infectious onset and the role of BBB damage
In one of the three patients with post-infectious onset (case 6), interruption of long-term MTX treatment for RA due to infection was followed by the first attack a few weeks later, leaving the possibility that MOG-IgG were already present before clinical onset but were able to enter the CNS to a pathologically relevant degree only after infection-related BBB breakdown. In fact, a markedly elevated QAlb as well as Gd entry into the lesion was noted during acute brainstem encephalitis in that case. In AQP4-IgG-positive NMO, the autoantibody has indeed been retrospectively demonstrated in samples obtained months or years before disease onset [51–53]. Overall, 8/9 patients showed Gd enhancement during disease activity, and 3/7, including the single patient without Gd enhancement, had an increased QAlb, indicating possible BBB disruption. Primary or secondary impairment of the BBB function may be an important step in the pathogenesis of MOG-IgG-positive brainstem encephalitis, since it may allow MOG-IgG to enter the CNS.
In one patient, the first symptoms developed within 2 weeks after vaccination for diphtheria, tetanus, pertussis, polio, and influenza, and thus within a time window considered to be compatible with a post-vaccination reaction. Of note, we describe a second adult patient with disease onset shortly after vaccination against diphtheria, tetanus and pertussis in part 2 of this series . The association of MOG-IgG seropositivity with infection and vaccination deserves to be investigated in more detail. MOG-IgG have also been reported in children with acute disseminated encephalomyelitis , a condition with suspected post-infectious or post-vaccinal etiology [54, 55]. Postvaccinal onset of NMO has been described also in a few AQP4-IgG-positive patients, although a causal link has not yet been proven [56, 57].
MOG-IgG positivity associated with a mature teratoma
Both MOG-IgG-positive encephalomyelitis and AQP4-IgG-positive NMOSD are not usually found in a paraneoplastic context. It is therefore of potential interest that one of the patients described above had a history of mature teratoma that was removed just 2 months before onset of her CNS disorder. In addition, a ganglioneuroma was detected after disease onset in the same patient. It remains unknown whether this patient’s tumors and MOG-IgG seropositivity were related. Teratomas have been shown to play a role in the pathogenesis of NMDAR encephalitis, another autoantibody-related disorder of the CNS . However, NMDAR antibodies were negative in our patient. While, to the best of our knowledge, the presence of MOG in teratomas has not been investigated so far, expression of CNPase, an oligodendrocyte marker, has been described in mature teratomas , and several reports on oligodendrogliomas arising in mature teratomas exist [60–63]. Ectopic expression of MOG by the patient’s tumor therefore cannot be completely ruled out.
Influence of age and sex
The median age at first clinical brainstem attack was 31 years (range 18-53 years) in the present, adult cohort. This did not differ from the age of disease onset in the total MOG cohort (see part 2 ) (31 years, range 6-70; N = 50). Remarkably, in 7/15 (46.7 %) cases brainstem inflammation (as evidenced clinically or by MRI) was already present at or very shortly after disease onset. In the remainder, the median time between disease onset and first clinically apparent brainstem encephalitis was just 0.8 years. However, this interval varied widely, with brainstem lesions occurring for the first time only 7, 17, and 41 years, respectively, after the first attack in three patients. Together, these findings suggest that the presence or absence of brainstem involvement is not an effect of age or disease duration.
The sex ratio in our cohort (1:2.8) did neither differ from that previously reported in AQP4-IgG-negative NMO patients with and without brainstem lesions (1:2; n = 27)  nor from that in the total MOG cohort (1:2.8; n = 50)  (but is in stark contrast to the sex ratio of ~1:9 reported in AQP4-IgG-positive patients ), indicating that female gender is probably not a risk factor for the development of brainstem involvement in patients with MOG-IgG-positive ON and/or myelitis.
Good recovery from acute brainstem attacks after IVMP and/or PEX
While many patients seemed to benefit from immunotherapy, treatment responses and long-term outcomes differed markedly. Treatments applied during acute brainstem attacks (with and without concomitant ON and/or myelitis) included IVMP, oral steroids, and PEX. Overall, treatment outcome was available for 18 attacks with clinical brainstem involvement. Treatment with IVMP (with and without oral tapering) or with PEX (with or without additional IVMP) were followed by good partial recovery after 6 and 3 brainstem attacks, respectively, and by full recovery after 3 and 3 brainstem attacks, respectively. No treatment was given for the last (and subsequently fatal) attack in patient 1, which occurred during palliative care, for two brainstem attacks in case 5, which remitted spontaneously, and for the only attack in case 5 (initially considered to be of infectious origin due to granulocytic pleocytosis), which partially remitted.
PEX treatment was beneficial in patient 1 during most attacks, including three brainstem attacks and one attack in which IVMP had led only to partial recovery. Similarly, PEX also resulted in full or almost full remission when used to treat attacks with brainstem involvement in patients 6 and 8 (used in combination with IVMP), as well as in patient 9 after failure of high-dose IVMP therapy.
Importantly, however, PEX treatment could not prevent relapses 1-3 months later in case 1, including the fatal brainstem attack in that patient, as well as a relapse of ON within 2 months in patient 9. Preliminary findings from our laboratory (S.J., unpublished data) show that anti-neural autoantibodies may remain detectable or reappear soon after five to seven plasma exchanges, raising the question of whether PEX treatment may be discontinued too early in some cases. Alternatively, T cell-mediated mechanisms may play a more important role in patients who do not sufficiently respond to PEX.
Both IVMP and PEX were also not always effective when used to treat attacks other than brainstem encephalitis in this cohort and in the total cohort (see part 2 for detailed analysis ).
Long-term immunotherapy did not prevent brainstem attacks in some patients
Immunosuppressive or immunomodulatory (IS/IM) drugs used in our patients included mitoxantrone, AZA, rituximab, natalizumab, IFN-beta, and IVIG. Treatment responses varied considerably inter- and intraindividually. Patient 2 suffered from serious neurologic impairment during the acute attack and during the following months, but immunotherapy with IVMP, oral steroids, PEX, and, for 12 months, IVIG was eventually followed by almost complete remission and no more attacks. Pathophysiologically, it is of interest that immunotherapy and clinical recovery were paralleled by a significant drop in MOG-IgG titers from 1:10,240 to 1:640 in this patient. By contrast, IS/IM therapy was not effective in preventing brainstem encephalitis in several other cases. Of particular note, rituximab was followed by severe clinical and radiologic deterioration with myelitis and active (yet asymptomatic) lesions in the pons and medulla oblongata within a few weeks after infusion in patient 1, which is reminiscent of the possibly BAFF-mediated deterioration reported in some NMO patients after commencement of rituximab . Moreover, a massive brainstem attack with dysarthria, dysphagia, left oculomotor and facial nerve palsy and new T2 lesions in the pons, pedunculus cerebellaris, cerebral crus and medulla oblongata occurred just 4 months after the last rituximab infusion. In case 5, one brainstem attack took place while the patient was being treated with AZA, one during treatment with IFN-beta, and one 9 months after commencement of natalizumab therapy. Patient 13 developed an attack of simultaneous myelitis and brainstem encephalitis four months after the first natalizumab infusion and another one (with lesions in the medulla oblongata) while on treatment with glatiramer acetate. In case 7, a severe attack involving the brainstem, supratentorial brain, and spinal cord occurred 4 weeks after commencement of AZA treatment; similarly, patient 8 experienced several relapses, including a brainstem attack, while on treatment with AZA. Of note, 14/34 relapes (in 10/17 AZA treated patients) were documented in the total cohort that took place during the latency period of AZA (months 1-6) (see part 2 of this series ). Of those, around 40 % occurred in patients not co-treated with oral steroids during that period. This suggests that co-treatment should be considered during the latency period of AZA treatment in MOG-IgG-positive patients, provided contraindications have been excluded.
Disease exacerbation after IFN-beta
In common with other patients described in part 2 of this series , patient 4 was initially diagnosed with MS. Accordingly, she was treated with IFN-beta-1a i.m. However, commencement of IFN-beta treatment was associated with marked disease exacerbation, characterized by new brainstem and spinal cord lesions and a new clinical attack. Similarly, patient 13 developed three attacks of myelitis and/or optic neuritis while on treatment with IFN-beta 1a i.m. or, later on, IFN-beta 1a s.c. Disease exacerbation following IFN-beta administration has also been reported in AQP4-IgG-positive NMO [65–69] and likely reflects differences in the immunopathogenesis of MS and NMO. This observation is of high potential interest, since initial misdiagnosis as classical MS – and, in consequence, mistreatment with IFN-beta – might be even more common in MOG-IgG-positive patients than in AQP4-IgG-positive patients given the high rate of brain involvement in that condition . Falsely classified AQP4-IgG- and, possibly, also MOG-IgG-positive patients might account for some of the occasional IFN-beta non-responders observed in MS studies. Larger studies on the efficacy of IFN-beta in MOG-IgG-positive patients treated with this substance in the past seem warranted, as does retrospective testing for MOG-IgG of samples from IFN-beta non-responders identified in past clinical trials.
Long-term prognosis differed widely but did not depend on brainstem damage in most cases
Cases 1 and 2 illustrate that the prognosis differs widely among MOG-IgG-positive patients: while brainstem encephalitis led to respiratory insufficiency in both patients, it was fatal in the former case and remitted almost completely in the latter. Unexpectedly, residual neurologic impairment in our patients was mostly not related to brainstem damage. The median EDSS at last follow-up in patients with a disease duration of >24 months (n = 12) described here was 3 and ranged between 0 and 10; only 4 patients had an EDSS >3 at last follow-up (EDSS 4, 7.5, and 10 after 123, 225, and 507 months, respectively).
Brainstem involvement was associated with a more aggressive disease course
However, the median EDSS at last follow-up in patients with brainstem involvement and a disease duration of >24 months was still higher (median 3, range 0-10, n = 12) than among all patients from the total cohort  who had no history of clinical or subclinical brainstem involvement at last follow-up and an observation time of ≥24 months (median EDSS 2; n = 23; p < 0.04), as were the total number of attacks at last follow-up (median 7.5, range 1-27, vs. median 3, range 1-28), the number of myelitis attacks (median 2, range 0-11, vs. median 0.5, range 0-3), the proportion of patients who had experienced both attacks of ON and attacks of myelitis at last follow up (75 % vs. 41.7 %), the median annualized relapse rate (1.32 vs. 0.59, p < 0.03), and the proportion of patients with additional supratentorial brain lesions (75 % vs. 30.4 %; p < 0.02). As observation times did not differ significantly between these two subgroups (median 69.5, range 34-507, vs. median 70, range 26-394), brainstem involvement seemed be a risk factor for a more severe disease course. This has potential therapeutic implications and should be addressed in future prospective studies.
We acknowledge some limitations of our study. Firstly, the retrospective design is a potential limitation. However, prospective studies would be difficult to perform due to the very low prevalence of the disease. Moreover, reliable tests have become available only recently; accordingly, only retrospective long-term data are currently available. Furthermore, the number of patients included and the number of items documented in the present study were high and data loss relatively low. Secondly, the multicenter design, which was necessary given the low prevalence of the condition, could be a limitation. However, the study design also strongly reduced the risk of selection bias, which was acknowledged as a possible limitation by the authors of previous large single-center studies in the field of NMO [30, 36]. Moreover, all patients were documented at university centers providing a similar standard of tertiary care. Thirdly, we cannot fully exclude a potential referral bias, since MOG-IgG testing may have been ordered particularly in patients presenting with ON and/or myelitis based on the previous literature. It is therefore conceivable that MOG-IgG-positive patients with isolated brainstem and/or brain involvement are underrepresented in our study. Finally, from a pathophysiological point of view it is a possible limitation that we cannot formally prove that the antibody was already present at disease onset in all cases, since routine MOG-IgG testing was not available in the past. However, MOG-IgG was present already at disease onset in all patients with available data in the main cohort, as reported in part 1 of this series : 2 MOG-IgG positive sera were taken within the first week (at 2 and 4 days) after disease onset, 10 within the first month (median 10 days, range 2-31), and 18 within the first 3 months (median 26 days, range 2-85). The median MOG-IgG titer at disease onset was 1:2560 (range 160-20480; N = 18).