BackgroundNeurological immune-related adverse events (irAE-n) are severe and potentially fatal toxicities of immune checkpoint inhibitors (ICI). To date, the clinical significance of neuronal autoantibodies in irAE-n is poorly understood. Here, we characterize neuronal autoantibody profiles in patients with irAE-n and compare these with ICI-treated cancer patients without irAE-n.MethodsIn this cohort study (DRKS00012668), we consecutively collected clinical data and serum samples of 29 cancer patients with irAE-n (n = 2 pre-ICI, n = 29 post-ICI) and 44 cancer control patients without irAE-n (n = 44 pre- and post-ICI). Using indirect immunofluorescence and immunoblot assays, serum samples were tested for a large panel of neuromuscular and brain-reactive autoantibodies.ResultsIrAE-n patients and controls received ICI treatment targeting programmed death protein (PD-)1 (61% and 62%), programmed death ligand (PD-L)1 (18% and 33%) or PD-1 and cytotoxic T-lymphocyte-associated protein (CTLA-)4 (21% and 5%). Most common malignancies were melanoma (both 55%) and lung cancer (11% and 14%). IrAE-n affected the peripheral nervous system (59%), the central nervous system (21%), or both (21%). Prevalence of neuromuscular autoantibodies was 63% in irAE-n patients, which was higher compared to ICI-treated cancer patients without irAE-n (7%, p <.0001). Brain-reactive autoantibodies targeting surface (anti-GABABR, -NMDAR, -myelin), intracellular (anti-GFAP, -Zic4, -septin complex), or unknown antigens were detected in 13 irAE-n patients (45%). In contrast, only 9 of 44 controls (20%) presented brain-reactive autoantibodies before ICI administration. However, seven controls developed de novo brain-reactive autoantibodies after ICI initiation, therefore, prevalence of brain-reactive autoantibodies was comparable between ICI-treated patients with and without irAE-n (p = .36). While there was no clear association between specific brain-reactive autoantibodies and clinical presentation, presence of at least one of six selected neuromuscular autoantibodies (anti-titin, anti-skeletal muscle, anti-heart muscle, anti-LRP4, anti-RyR, anti-AchR) had a sensitivity of 80% (95% CI 0.52-0.96) and a specificity of 88% (95% CI 0.76-0.95) for the diagnosis of myositis, myocarditis, or myasthenia gravis.ConclusionNeuromuscular autoantibodies may serve as a feasible marker to diagnose and potentially predict life-threatening ICI-induced neuromuscular disease. However, brain-reactive autoantibodies are common in both ICI-treated patients with and without irAE-n, hence, their pathogenic significance remains unclear.