CNS aquaporin-4 autoimmunity in children

The astrocytic water channel aquaporin-4 (AQP4) is the antigenic target of the neuromyelitis optica autoantibody marker (NMO-IgG).^1–8 Clinical accompaniments of this autoantibody have provided novel insight into a previously unappreciated spectrum of CNS inflammatory disorders.⁹ In addition to supporting the diagnosis of neuromyelitis optica (NMO) in patients with optic neuritis and longitudinally extensive transverse myelitis (extending over three or more vertebral segments), seropositivity predicts recurrence in patients with a single episode of longitudinally extensive transverse myelitis or recurrent optic neuritis.^10,11 Furthermore, brain lesions are detected by MRI in 60% of seropositive patients and brain involvement is sometimes symptomatic¹²; most common are encephalopathy, intractable nausea, hiccups, diplopia, and vertigo.^13,14 In recognition of the broad clinical spectrum of NMO-related disorders, diagnostic criteria were revised in 2006 to include (but not require) NMO-IgG seropositivity.^15,16

As in adults, NMO (and NMO-IgG seropositivity) in the pediatric population is usually associated with a severe, relapsing inflammatory CNS disease,¹⁷ but is sometimes monophasic.¹⁸ The frequency of NMO-IgG in children with inflammatory disorders of the CNS is 78% for relapsing NMO, and 20% for partial forms of NMO (relapsing optic neuritis and relapsing longitudinally extensive transverse myelitis [a single case]).¹⁷ NMO-IgG has not been detected in children with multiple sclerosis (MS), acute disseminated encephalomyelitis (ADEM), monophasic optic neuritis, or monophasic transverse myelitis, or in healthy children.¹⁷ Our Clinical Neuroimmunology Laboratory’s consultative practice has revealed that seropositive children who have NMO, or an inaugural or partial manifestation of NMO, present with symptomatic brain lesions more commonly than adult patients. This report describes, in a consecutive series of NMO-IgG-positive children, the clinical, radiologic, and serologic associations that accompany AQP4 autoimmunity. We propose that CNS AQP4 autoimmunity in children is a distinct biomarker-defined entity.

METHODS

This study was approved by the Institutional Review Board of Mayo Clinic, Rochester (IRB 07-004089), and includes patients identified in the Clinical Neuroimmunology Laboratory as seropositive for NMO-IgG and age 18 or younger at onset of neurologic disease. The detection assay (indirect immunofluorescence assay on a substrate of normal mouse CNS and kidney tissues²) was performed in two circumstances: 1) prior to 2005, children were identified clinically as having NMO or a high risk syndrome, and an immunofluorescence pattern consistent with NMO-IgG (n = 14); 2) 2005–2007: service laboratory evaluation for NMO-IgG (n = 74). Ten of the 88 patients were seen in the Department of Neurology at Mayo Clinic and were evaluated by one of the authors.

Additional serologic testing, done when adequate serum was available, included 1) immunoprecipitation of green fluorescent protein (GFP)-tagged AQP4 (produced in stably transfected human embryonic kidney cells [HEK 293]¹). Fluorescence remaining bound to washed Protein-G agarose immunoprecipitant was measured spectroscopically in a 96 well plate (Genios Pro, TECAN, Research Triangle Park, NC) and was expressed quantitatively in terms of pmol of AQP4 bound per liter of serum by reference to a GFP protein standard (Sigma-Aldrich, St Louis, MO); 2) radioimmunoprecipitation of neuronal voltage-gated cation channels (calcium channels [P/Q-type and N-type] and α-dendrotoxin-sensitive potassium channels), acetylcholine receptors of skeletal muscle and ganglionic neurons, and glutamic acid decarboxylase [GAD65]¹⁹; 3) ELISA detection of striational antibodies; and 4) indirect immunofluorescence detection of antibodies reactive with neuronal nuclear antigens (antineuronal nuclear antibodies, types 1, 2, and 3), neuronal cytoplasmic antigens (antibodies specific for amphiphysin, collapsin response-mediator protein-5, Purkinje cell cytoplasm types 1, 2 and Tr), and anti-glial/neuronal nuclear (AGNA-1).²⁰

We used noncompetitive, sandwich enzyme immunoassays (EIAs) to detect non–organ-specific antinuclear antibodies (ANA) and antibodies to extractable nuclear antigens (ENA) (BioRad Diagnostics Inc., Hercules, CA). Antigens used in the generic ANA were isolated from Hep-2 cell nuclei. A negative ANA result was defined as <1.0 EIA units. Individual ANA specificities (SS-A/Ro, SS-B/La, Sm, U1RNP, Scl 70, and Jo 1) were detected by the use of sequential immunoassays as follows. A screening assay for ENA antibodies that employed a combination of six individual ENA antigens adsorbed as a cocktail to microtiter wells was used to identify sera that had one or more antibodies. Presumptive positive results were defined as >20.0 EIA units. Individual ENA antibody specificities were identified in positive sera by testing in follow-up assays that were specific for single specificities, e.g., SS-A/Ro. Noncompetitive EIA methods (Phadia Diagnostics, Kalamazoo, MI) were also used to detect phospholipid autoantibodies (IgG, IgM, and IgA anticardiolipin). Positive results were defined as >10.0 PLU. We used latex agglutination assays to detect thyroid autoantibodies (peroxidase and thyroglobulin).²¹

The 58 patients for whom clinical information was available were evaluated by a pediatric neurologist or demyelinating disease subspecialist. Detailed clinical information was obtained by review of case records²⁸ or by physician telephone interview.³⁰ For patients with adequate clinical information, Expanded Disability Severity Scale (EDSS) scores²² were assigned by the treating physician or by a neurologist author.

RESULTS

Pediatric cases account for 5% of the total NMO-IgG seropositive patients currently identified in the Mayo Neuroimmunology Laboratory. Clinical information was available for 58 of 88 seropositive pediatric patients (66%); 88% were girls. Ethnic backgrounds are summarized in table e-1 on the Neurology® Web site at www.neurology.org. The four most common ethnicities, in decreasing order of frequency, were African American, 20 (34%); Caucasian, 16 (27%); Hispanic, 6 (10%); and Native American, 5 (9%).

Clinical course.

Attacks were recurrent in 54 patients (93%), with new symptoms occurring 1 month or more after the preceding attack. Median follow-up was 12 months (range 1–120 months). Median duration between attacks was 3 months (range 1–87 months). The illness was monophasic in 4 patients (12%), but follow-up periods were very short for those patients (median, 4 months). All but 1 of 58 patients (98%) had at least one attack of optic neuritis or transverse myelitis: 48 (83%) had at least one episode of optic neuritis (median two episodes, range 1–14), 45 (78%) had at least one episode of transverse myelitis (median two attacks, range 1–14), and 22 (38%) had recurrent transverse myelitis. Thirty-nine patients (87%) with episodes of transverse myelitis experienced weakness, and 6 (13%) had only sensory symptoms. The median longitudinal extent of attack-related spinal cord MRI abnormalities was 10 vertebral segments (detailed data available for 37 patients). The timing of MR imaging relative to attack-evolution and treatment was unclear for two patients whose spinal cord lesions extended <3 vertebral segments. Attack-related symptoms involved the brain or brainstem in 26 patients (45%); table 1 shows representative MRI findings. A single patient (figure 3A, patient 5) had headache and had acute obstructive hydrocephalus secondary to aqueductal stenosis (presumed antecedent periaqueductal inflammation). Another patient had hearing loss; its origin (central or peripheral) was not ascertained (data not available for audiology and brainstem evoked potentials). MR head imaging data (table 1; figures 1–3) were available for 56 patients: 38 (68%) had brain parenchymal abnormalities, 19 (34%) had postcontrast enhancement of the optic nerve, 5 had chiasmal enhancement (1 with bilateral optic tract enhancement).

View this table:

• View inline
• View popup
• Download powerpoint

Table 1 MRI brain abnormalities (38 patients)and associated symptomatology (26 patients)

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Representative MRI abnormalities in seven children with hemispheric involvement typical of the spectrum identified in children with CNS AQP4 autoimmunity

White arrows indicate abnormality of fluid attenuated inversion recovery (FLAIR) or T2 signal. (A) Patient 1 (8-year-old girl presenting with encephalopathy) has T2 signal abnormality radiating in a tentacle-like fashion along the Virchow-Robin spaces from the lateral ventricle through parietal subcortical white matter. (B) Patient 2 (12-year-old girl presenting with delirium followed by coma, described previously¹⁴) has confluent predominantly white matter abnormalities. (C-I) Patient 3 (15-year-old girl presenting with delirium and aphasia) has confluent predominantly white matter abnormalities; 1 year later bitemporal neocortical lesions are seen (C-II); the patient is cognitively normal other than mild expressive aphasia (C-III). (D-I) Patient 4 (8-year-old girl presenting with encephalopathy and seizures) has tentacle-like white matter abnormalities radiating from the lateral ventricles into frontal and parietal white matter. Follow-up MRI 8 years later (D-II and D-III) demonstrates persisting FLAIR abnormalities and global atrophy. (E-I) Patient 5 (17-year-old girl with 11-year history of demyelinating disease) has confluent T2 abnormalities of the genu more than splenium of the corpus callosum; image 3 years later reveals resolution of the genu lesion (E-II) and hydrocephalus (shunted; see figure 3, A and B). (F-I) Patient 6 (15-year-old boy presenting with encephalopathy, focal motor seizures, and hemiparesis) has FLAIR signal abnormality in the left mesial temporal region and left cerebral peduncle. Images 3 years later (F-II) demonstrate persisting FLAIR abnormality and left hippocampal atrophy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Representative MRI abnormalities in three children showing spectrum of optic chiasm/diencephalic/peri IIIrd ventricular involvement in regions known to highly express aquaporin-4

White arrows indicate enhancement, abnormality of fluid attenuated inversion recovery (FLAIR), or T2 signal. Dashed black lines represent anatomic level in relationship to the diagram. (AI) Patient 7 (12-year-old girl presenting with bilateral optic neuritis) has enhancement of optic nerves, chiasm, and optic tracts on T1 (postgadolinium). FLAIR abnormality extends into the hypothalamus (AII) and peri-IIIrd ventricular diencephalon (AIII). (B) Patient 3 (15-year-old girl; see also figure 1, C through I) also develops nausea, syndrome of inappropriate antidiuretic hormone secretion, hyponatremia (111 mmol/L), and menstrual irregularities and has a hypothalamic lesion (FLAIR). (C) Patient 6 (15-year-old boy; see also figure 1, F through I) has FLAIR abnormalities in the left thalamus and right hypothalamus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Representative MRI abnormalities in seven children showing spectrum of brainstem, periaqueductal, and peri-IVth ventricular and spinal cord involvement in regions known to highly express aquaporin-4

White arrows indicate abnormality of fluid attenuated inversion recovery or T2 signal. Dashed black lines represent anatomic level in relationship to the diagram. (A) Patient 5 (17-year-old girl; see also figure 1, E through I) has bilateral lesions in the superior cerebellar peduncles (A-I), a white matter lesion in the cerebellum (A-II), and bilateral lesions in the internal capsule (A-III). Four years later the patient presented with symptomatic obstructive hydrocephalus (A-IV, see also figure 1, E-II; internal capsule lesions had resolved). (B) Patient 8 (11-year-old girl) has longitudinally extensive T2 abnormality in the central spinal cord as is typically found in adult neuromyelitis optica (in this instance 11 vertebral segments). (C-I) Patient 9 (17-year-old girl presenting with intractable vomiting, quadriplegia, and tonic limb spasms) and (C-II) Patient 11 (17-year-old girl presenting with intractable vomiting) have longitudinally extensive lesions spanning brainstem and central cervical and thoracic cord.

DISCUSSION

This is the first study to describe a spectrum of inflammatory CNS disorders in childhood and adolescence which is unified serologically by detection of an autoantibody reactive with the astrocytic AQP4 water channel. Although optic neuritis or longitudinally extensive transverse myelitis occurs in 97% of cases, in 45% the entity of pediatric CNS AQP4 autoimmunity extends beyond clinical and radiologic abnormalities in the optic nerve and spinal cord, and 16% of cases present with a cerebral syndrome. By contrast, brain lesions in adult patients are almost always asymptomatic.¹⁴ It is noteworthy that brain lesions in adults with NMO exhibit the same immunopathologic characteristics as spinal cord lesions.^23,24 The pattern-specific loss of AQP4 immunoreactivity that distinguishes NMO lesions immunopathologically from other inflammatory CNS demyelinating disorders is consistent with a pathogenic role for a complement-activating AQP4-specific autoantibody.²⁵ Demonstrations that NMO-IgG in vitro has a selective pathogenic effect on cell membranes expressing AQP4 in high density,²⁶ and that antibody-depleting therapies (IVIg,²⁷ plasmapheresis,²⁸ azathioprine,²⁹ and the B cell-specific monoclonal antibody rituximab³⁰) are beneficial, support the hypothesis that IgG is the proximal cause of a spectrum of CNS inflammatory demyelinating disorders targeting astrocytic AQP4.

The diverse symptomatology observed in children with CNS AQP4 autoimmunity includes brainstem syndromes and cerebral dysfunction. Brain MRI abnormalities in adults and children tend to localize in periependymal areas^12,14 which normally are enriched in AQP4.³¹ The MRI abnormalities we encountered included clinically silent lesions limited to the immediate periventricular regions; hypothalamic, thalamic, and diencephalic lesions; long spindle or radial white matter signal change extending from the lateral ventricles into the cerebrum (unlike the short, pericallosal-confined Dawson’s fingers of MS); white matter abnormalities, extensive and confluent, or discrete juxtaposed to or extending from the lateral ventricles; and brainstem and cerebellar lesions adjacent to the aqueduct of Sylvius and fourth ventricle.

Endocrinopathies and disorders of water balance, which are recognized associations of CNS AQP4 autoimmunity, are attributable to hypothalamic involvement.³² The finding of hydrocephalus documented in one patient of this report (and in two earlier reported patients¹⁴) is of interest, though the mechanism and potential relationship to AQP4 autoimmunity remains unclear.

The clinical and radiologic findings we documented in eight NMO-IgG seropositive children of this study overlap with criteria proposed recently for diagnosis of ADEM: initial polysymptomatic encephalopathy accompanied by focal or multifocal hyperintense lesions predominantly affecting white matter.³³ Although histopathologically ADEM and CNS AQP4 autoimmunity are distinct entities,³⁴ the occurrence in both disorders of encephalopathy (eight patients of this report), thalamic involvement (five patients), as well as optic neuritis, and longitudinally extensive transverse myelitis, may make their clinical differentiation difficult. Seropositivity for AQP4 autoantibody aids their distinction when clinical uncertainty exists.

One or more additional autoimmune disorders accompanied CNS AQP4 autoimmunity in 42% of patients of this report and other autoantibodies were detected in 76% of patients. The coexistence of multiple autoantibodies and autoimmune disorders in children and adults with NMO-related disorders (e.g., Graves disease, rheumatoid arthritis, SLE, SS) is a unifying characteristic of these disorders that distinguishes them from MS. The lack of NMO-IgG in patients with SLE or SS except for those with coexisting longitudinally extensive transverse myelitis, optic neuritis, or both, is consistent with NMO, SLE, and SS coexisting in some patients, rather than NMO being a complication of SLE or SS.³⁵

A detailed analysis of treatment effects in this study was limited by its retrospective nature and dependence on physician-reported outcomes. Acute attacks often responded to steroid treatment, and plasma exchange and IVIg were both beneficial auxiliary therapies in severe cases refractory to steroid therapy. A reduction in attack frequency was recorded in several patients treated with antibody-lowering maintenance therapies, such as azathioprine (especially with oral prednisone), rituximab, and mycophenolate mofetil. Earlier observational studies in patients with NMO have reported beneficial outcomes for these treatments.^29,30,36 Determination of optimal therapeutic protocols for CNS AQP4 autoimmune disorders will require randomized controlled trials.

Disability was severe and developed rapidly in the pediatric subjects of this report. Only 6% were normal on neurologic examination at the end of the follow-up period (median 12 months). A median EDSS score of 4.0 (fully ambulatory without aid) does not reflect severity of visual impairment. More than 50% of patients had persistent visual impairment, and 27% had blindness that was bilateral and complete, or near complete. The prognosis reported for childhood-onset MS from a longitudinal study (median interval from onset to EDSS 4.0, 20 years³⁷) was more benign than the outcome we have documented for childhood CNS AQP4 autoimmunity.

The relatively short follow-up period in our study (median 12 months) limits its interpretation because patients with partial forms of NMO may later fulfill NMO diagnostic criteria. Other limitations include the bias inherent in serologic ascertainment and lack of clinical information for 30 patients. Recently reported serologic results for children ascertained by clinical presentation with an inflammatory demyelinating CNS disorder demonstrated the specificity of NMO-IgG for NMO or partial forms of this disorder.¹⁷ It remains to be determined whether the 22% rate of seronegativity for NMO-IgG in children with relapsing NMO¹⁷ reflects immunotherapy, assay insensitivity, or a uniform clinical phenotype with heterogeneous immunopathogenesis.

ACKNOWLEDGMENT

The authors thank Deborah Bradshaw, MD, Upstate Medical University, Syracuse, NY, Majeed Al-Mateen, MD, Mary Bridge Children’s Hospital and Health Center, Tacoma, WA, and Lauren Krupp, MD, National Pediatric Medical Center, Stony Brook University MS Center, Stony Brook, NY, for providing detailed clinical information on individual cases. T.L. thanks Elizabeth Barnes, RN (Texas Children’s Hospital) for data management.