Case 1 – Fatal rhombencephalitis in a patient with recurrent ON and LETM

A previously healthy 44-year-old Caucasian woman first developed right-sided ON in November 2003. Lumbar puncture (LP) demonstrated a normal CSF white cell count as well as normal CSF protein, glucose, and lactate levels and negative oligoclonal bands (OCB). Brain MRI revealed asymptomatic lesions in the right cerebellar hemisphere and in the cerebral peduncle of the midbrain. Magnetic resonance imaging (MRI) of the spinal cord was not performed. After treatment with intravenous methylprednisolone (IVMP), vision returned to normal. Extensive laboratory tests did not reveal any infectious or rheumatologic cause of the patient’s symptoms.

Up to 2010 ON attacks occurred once to twice annually, but vision always returned to normal after IVMP therapy. However, after another attack of ON in August 2010, vision in the right eye dropped permanently to 0.85 on the Snellen chart despite IVMP therapy.

In January 2011, the patient developed a first attack of myelitis with accompanying brainstem encephalitis. Symptoms included right hemiparesis, dysarthria, and dysphagia. MRI showed no supratentorial brain lesions but a T2 lesion in the medulla oblongata and a cervical longitudinally extensive transverse myelitis (LETM) lesion extending from C2 to C4 with gadolinium (Gd) enhancement. LP again demonstrated a normal CSF white cell count, normal CSF protein level, and negative OCB. Treatment with IVMP was followed by complete remission. In June 2011, after another attack of myelitis with paraparesis, treatment of which with IVMP had resulted in incomplete remission of the symptoms, treatment with glatiramer acetate (GLAT) for suspected MS was initiated.

After a relapse-free period of 1 year, two further relapses (ON in May 2012 and myelitis with paraparesis in July 2012) followed. While treatment with IVMP led only to partial recovery, plasma exchange (PEX) treatment (five exchanges) was followed by complete clinical remission after both relapses. After the second relapse, GLAT was stopped (July 2012) and natalizumab started. However, treatment with natalizumab was discontinued after two infusions due to recurrent headaches.

In September, October, and December 2012 three further relapses of myelitis with paraparesis occurred. PEX (five exchanges; no IVMP) resulted in partial recovery in all cases. In January 2013, the patient received a first infusion of mitoxantrone (12 mg/m², nadir 3000/μl), and after a further myelitis relapse with paraparesis in February 2013, a second mitoxantrone infusion was given (8 mg/m², nadir 2800/μl). Complete recovery was achieved following PEX (five exchanges; no IVMP). After a further flare-up of myelitis later the same month with paraparesis and Gd enhancement at levels T2, T3, T5, and T9, treated with a cycle of five plasmaphereses with incomplete remission in February 2013, a third infusion of mitoxantrone with 8 mg/m² was given. A serum sample taken at that time later tested positive for MOG-IgG in a live-cell CBA [2] (1:1280). Early in 2013, two new ON relapses occurred and were treated with a cycle of PEX (five exchanges), again with incomplete remission. MRI now showed Gd enhancement in large parts of the thoracic spinal cord. An expanded disability status scale (EDSS) score of 6 was documented at that time. MOG-IgG were retrospectively positive at a titer of 1:640.

Rituximab (500 mg) was infused for the first time in May 2013, followed 2 weeks later by a second infusion of 500 mg rituximab. Fifteen days after the second infusion the patient’s paraparesis worsened (EDSS 8) and dysarthria re-occurred. Cerebral and spinal MRI showed active lesions with Gd enhancement at the C4/C5 level, in the pons, and in the medulla oblongata. After PEX (five exchanges; no IVMP), the patient was able to walk again and her dysarthria completely remitted. Two new myelitis relapses in July and August 2013 with paraparesis (EDSS 7.5) and dysarthria were again treated with PEX (five exchanges; no IVMP), which was followed by partial recovery. MRI showed a new inactive lesion in the pons in July 2013 and in August 2013 Gd enhancement extended from C2 to C5. After a second PEX cycle, urosepsis occurred and was treated successfully with antibiotics. The next relapse occurred in November 2013 with a left ON (visual acuity of 0.05), which remitted incompletely after one cycle of PEX (five exchanges) (EDSS 7.5).

In January 2014, massive deterioration of the brainstem symptoms occurred, with dysarthria, dysphagia, left oculomotor and facial nerve palsy, and new T2-hyperintense lesions in the pons, pedunculus cerebellaris, cerebral crus, and medulla oblongata. No supratentorial lesions were seen on brain MRI. Spinal MRI demonstrated a diffuse, asymptomatic T2-hyperintense lesion from C2 to C4 as well as patchy Gd enhancement in the pons, medulla oblongata, and the entire cervical spinal cord. The patient was treated with two courses of five PEX each (no IVMP) and initially improved. In the following month, however, clinical symptoms deteriorated again and the patient developed bilateral blindness and central hypoventilation. She was transferred to palliative care and died in May 2014.

The total number of attacks in this patient was 25, 15 of which occurred under various immunomodulatory and immunosuppressive therapies. These included around 13 relapses of ON and 12 of myelitis; the brainstem was clinically affected during 5 attacks. This corresponds to an annualized relapse rate of 2.38, though 15/25 relapses occurred during the last 24 months before the patient’s death.

Case 2 – Single episode of post-infectious whole-spine myelitis with severe brainstem and brain involvement yet complete recovery

An 18-year-old Caucasian woman had purulent tonsillitis in June 2010 and was treated with amoxicillin for 5 days. About 1 week later she developed intermittent fever (up to 39 °C), general malaise, headaches, meningism, fluctuating diplopia on right gaze, and subtle psychomotor slowing, for which she was admitted to the Department of Neurology, Charité – University Medicine Berlin. The day after admission, her condition deteriorated and she additionally noted weakness of her legs, urinary incontinence, respiratory impairment with difficulty in coughing, and mild dysphagia. On neurologic examination the patient was awake and oriented. Cranial nerve examination showed no abnormalities, but she had marked dysarthria and reported fluctuating horizontal diplopia. Her legs were plegic and she had reduced pain and touch sensation below T10. There was also British Medical Research Council (BMRC) grade 4 distal weakness in both arms. Deep tendon reflexes were preserved throughout. Except for a mature cystic ovarian teratoma without signs of malignancy, which had been removed 2 months before, her past medical history was unremarkable. N-Methyl-D-aspartate receptor (NMDAR) antibodies were negative. Spinal MRI demonstrated a prominent, longitudinally extensive, T2-hyperintense and centrally located lesion without contrast enhancement extending over almost the entire spinal cord, as well as swelling of the cord (Fig. 2). Cranial MRI showed T2-hyperintense lesions in the frontal and parietal subcortical white matter, the pontine tegmentum, and the cerebellar peduncles without contrast enhancement (Fig. 1). Thoracic and abdominal MRI revealed a right retroperitoneal mass with a diameter of 5.6 cm, which was resected 5 months later and found to be a ganglioneuroma. CSF analysis on the day of admission demonstrated an elevated total cell count with 306 white blood cells/μl (reference range <5/μl) with 80.2 % lymphocytes, 8.2 % monocytes, 7 % neutrophils, 2.4 % eosinophils, and 2.2 % activated lymphocytes, but no plasma cells or tumor cells. CSF lactate was elevated at 28.4 mg/dl (reference range <20 mg/dl). There was moderate blood--CSF barrier (BCSFB) dysfunction with an albumin CSF/serum ratio (QAlb) of 30.6 × 10^-3 (age-adjusted upper limit of normal 5.2 × 10^-3). Accordingly, total CSF protein was elevated (213 mg/dl; reference range <45 mg/dl). She had no local IgG synthesis and no CSF-specific OCB.

Routine serum chemistry including C-reactive protein (CRP) showed normal results. A complete blood count revealed mild leukocytosis (12.44/nl; reference range <11/nl). Extensive microbiological (herpes simplex virus type 1 and 2, varicella zoster virus, Epstein--Barr virus, cytomegalovirus, human herpes virus 6, enteroviruses, influenza A virus, adenovirus, tick-borne encephalitis virus, human immunodeficiency virus, human T-cell leukemia virus type 1, Borrelia, Treponema pallidum, Brucella, Bartonella, Mycoplasma, Chlamydia, Mycobacterium tuberculosis, Aspergillus) work-up detected no signs of acute CNS infection.

Antinuclear antibodies were detectable at a titer of 1:1280, but antibodies to double-stranded DNA, extractable nuclear antigens, onconeuronal antigens (Hu, Yo, Ri, CV2/CRMP5, Ma2/Ta, amphiphysin), cardiolipin, β2-glycoprotein, phosphatidyl serine, gangliosides (GD1a, GD1b GM1, GM2, GM3, GQ1b, GT1b [IgG and IgM in each case]), and aquaporin-4 were negative, as were anti-neutrophil cytoplasmic antibodies. Screening of serum and CSF for antibodies against NMDA receptors (IgG, IgA, IgM), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, glycine receptor, myelin-associated antigen, and glutamic acid decarboxylase was likewise negative.

However, testing of a serum sample obtained on the day of admission in a live-cell CBA demonstrated high-titer serum IgG antibodies against MOG (1:10240). Anti-MOG antibodies were also detectable in the CSF (1:64), but there was no evidence of intrathecal production of anti-MOG antibodies, with an anti-MOG antibody index (AI) of 0.3 (reference range <4) [3].

The patient was initially treated with ceftriaxone and ampicillin, which was discontinued after a bacterial infection was ruled out, as well as with IVMP (1 g/day for 6 days). Corticosteroids were thereafter administered orally (prednisolone 100 mg/d), slowly tapered, and eventually stopped in October 2010. Starting on the day after admission, she was additionally treated with a total of 10 PEX. MOG-IgG was retrospectively detectable also in PEX plasma (titer 1:320 after three exchanges). Thereafter, therapy with intravenous immunoglobulins (IVIG, 2 g/kg body weight) was initiated in July 2010. IVIG therapy was continued at 5-week intervals until June 2011, with dose reduction to 1 g/kg body weight from January 2011 onwards. The patient gradually recovered and was able to walk with a walking frame by August 2010. Follow-up spinal MRI in August 2010 demonstrated resolution of the longitudinal extensive lesion, but there were several residual patchy hyperintense lesions within the spinal cord. In a serum sample obtained at the same time, the anti-MOG antibody titer was clearly lower (1:640) than the initial titer, and anti-MOG antibodies were no longer detectable in CSF.

The patient did not receive any further immunomodulatory or immunosuppressive therapy. At the last follow-up examination, 28 months after disease onset and 16 months after discontinuation of IVIG, she reported no further relapses and had no impairments in activities of daily living except for residual unsteadiness when walking for a longer time, difficulties on running for more than 5 min, and slightly increased urinary frequency (EDSS 1). Neurologic examination showed a very mild residual right-sided pyramidal syndrome with brisk tendon reflexes, but was otherwise unremarkable.

Case 3 – Recurrent ON with rhombencephalitis and extensive brain involvement with poor outcome

A Caucasian woman experienced a first episode of bilateral ON in 1985 at age 28 and had three further attacks of bilateral and two of unilateral ON up to 1999, leaving her with residual visual acuity of <0.2 in both eyes. CSF and brain MRI was normal. In October 2003, the patient presented with intractable headaches lasting for several days, followed by mild impairment of consciousness. Cerebral computed tomography showed no abnormalities, while CSF analysis disclosed mixed lymphocytic and neutrophilic pleocytosis (1024 cells/μl) along with increased total protein (1350 mg/l) and mildly elevated lactate (3.6 mmol/l). CRP and leukocytes were elevated at 18 mg/l and 20.9 Gpt/l, respectively. There was no evidence of an infectious etiology, and symptoms resolved after combined treatment with dexamethasone 10 mg 4× per day for 4 days and ceftriaxone 4 g daily for 10 days. However, within 2 weeks the patient developed bilateral intention tremor of the upper extremities, marked gait ataxia, and right-sided visual worsening. Cranial MRI revealed extensive T2 hyperintense lesions in the brainstem, including around the fourth ventricle and in the supratentorial white matter, the corpus callosum, and, to some extent, the cerebral cortex. None of these abnormalities showed Gd enhancement. The CSF had 50 white cells/μl, mostly lymphocytes with some neutrophils (6 %) and occasional eosinophils (2 %). Treatment with IVMP (1 g/d for 5 days) followed by oral tapering of steroids for 6 weeks resulted in marked improvement of symptoms.

Six months later, the patient noted visual worsening in the left eye and recurrence of gait ataxia. Repeat cranial MRI depicted progressive T2-hyperintense lesions in the entire pons, especially around the fourth ventricle and the foramina of Luschka, extending bilaterally into the white matter of the left cerebellar hemisphere and into the crus cerebri with Gd enhancement. There was also involvement of the left subcortical white matter (adjacent to the temporal horn), the corpus callosum, and the juxtacortical regions of both parietal lobes, along with contrast enhancement in some of these lesions. Cerebral angiography showed no abnormalities. CSF analysis showed 5 cells/μl with a predominance of lymphocytes and some neutrophils and eosinophils. There was no evidence of intrathecal immunoglobulin synthesis. Anti-Hu and anti-Yo antibodies were negative. The patient received 1000 mg methylprednisolone i.v. daily for 5 days and subsequently long-term oral steroids, and this coincided with partial amelioration of both clinical symptoms and radiologic findings. Nightmares and visual hallucinations resolved along with rapid tapering of steroid doses. For long-term immunosuppression, overlapping treatment with azathioprine (AZA) (150 mg/day) was initiated in August 2004. However, five more relapses occurred over the following 101 months despite AZA treatment. At last follow-up, the EDSS of 7.5 and bilateral optic nerve atrophy (involving the chiasm) was documented. Retrospective testing of a serum sample taken in 2010 revealed high-titer MOG-IgG (1:20,480). MOG-IgG seropositivity was confirmed in a second sample obtained 22 months later.

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 [17]; 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 [17] 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 [20]. However, not all patients with ataxia had cerebellar lesions; some patients in this and in the total cohort [17] 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 [17]). 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 [17], 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 [36]; 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 [17]).

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 [37]. 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 [48]. 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 [17]. 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 [1], 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 [58]. 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 [59], 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 [17]) (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) [20] nor from that in the total MOG cohort (1:2.8; n = 50) [17] (but is in stark contrast to the sex ratio of ~1:9 reported in AQP4-IgG-positive patients [20]), 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 [17]).

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 [64]. 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 [17]). 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 [17], 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 [1]. 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 [17] 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.

Limitations

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 [3]: 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).