Glial fibrillary acidic protein immunoglobulin G as biomarker of autoimmune astrocytopathy: Analysis of 102 patients

OBJECTIVE: A novel autoimmune central nervous system (CNS) disorder with glial fibrillary acidic protein (GFAP)-IgG as biomarker was recently characterized. Here, 102 patients with GFAP-IgG positivity are described. METHODS: The 102 included patients had: (1) serum, cerebrospinal fluid (CSF), or both that yielded a characteristic astrocytic pattern of mouse tissue immunostaining; (2) confirmation of IgG reactive with specific GFAP isoforms (α, ɛ, or κ) by cell-based assays; and (3) clinical data available. Control specimens (n = 865) were evaluated by tissue (n = 542) and cell-based (n = 323) assays. RESULTS: Median symptom onset age was 44 years (range = 8-103), and 54% were women. The predominant phenotype (83 patients; 81%) was inflammation of meninges, brain, spinal cord, or all 3 (meningoencephalomyelitis). Among patients, highest specificity for those phenotypes was observed for CSF testing (94%), and highest sensitivity was for the GFAPα isoform (100%). Rare GFAP-IgG positivity was encountered in serum controls by tissue-based assay (0.5%) or cell-based assay (1.5%), and in CSF controls by cell-based assay (0.9%). Among patients, striking perivascular radial enhancement was found on brain magnetic resonance imaging in 53%. Although cases frequently mimicked vasculitis, angiography was uniformly negative, and spinal imaging frequently demonstrated longitudinally extensive myelitic lesions. Diverse neoplasms encountered were found prospectively in 22%. Ovarian teratoma was most common and was predicted best when both N-methyl-D-aspartate receptor-IgG and aquaporin-4-IgG coexisted (71%). Six patients with prolonged follow-up had brisk corticosteroid response, but required additional immunosuppression to overcome steroid dependency. INTERPRETATION: GFAPα-IgG, when detected in CSF, is highly specific for an immunotherapy-responsive autoimmune CNS disorder, sometimes with paraneoplastic cause. Ann Neurol 2017;81:298-309.