SARS-CoV-2 and autoantibodies in the cerebrospinal fluid of COVID-19 patients: prospective multicentre cohort study

Disease mechanisms underlying neurological and neuropsychiatric symptoms after coronavirus disease 2019 (COVID-19), termed neuro-COVID, are poorly understood. Investigations of the cerebrospinal fluid (CSF) for the presence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RNA and antibodies, as well as autoantibodies against neuronal surface antigens, could improve our understanding in that regard. We prospectively collected CSF and blood from patients investigated by lumbar puncture for neurological or neuropsychiatric symptoms during or after COVID-19. Primary outcomes were the presence of (i) SARS-CoV-2 RNA in CSF via polymerase chain reaction (PCR), (ii) SARS-CoV-2 immunoglobulin G (IgG) anti-S receptor-binding-domain antibodies via the Euroimmun and Wantai assays and (iii) IgG autoantibodies against neuronal surface antigens using commercial cell- and tissue-based assays (Euroimmun). Secondary outcomes were (i) routine CSF investigations and (ii) correlation between SARS-CoV-2 antibody levels in CSF with serum levels, blood–brain barrier permeability and peripheral inflammation. We obtained CSF from 38 COVID-19 patients (mean age 56.5 ± 19.2 years, 53% women) who developed neurological and neuropsychiatric symptoms. CSF pleocytosis (>5 cells) was observed in 9/38 patients (23.7%), elevated CSF protein (>0.50 g/L) in 13/38 (34.2%) and elevated CSF/serum albumin ratio in 12/35 (34.3%). PCR for SARS-CoV-2 RNA in CSF was negative in all. SARS-CoV-2 CSF antibodies were detected in 15/34 (44.1%; Euroimmun assay) and 7/31 (22.6%; Wantai assay) individuals, but there were no signs of intrathecal SARS-CoV-2 IgG production. SARS-CoV-2 CSF antibodies were positively correlated with serum levels (R = 0.93, P < 0.001), blood–brain barrier permeability (R = 0.47, P = 0.006), peripheral inflammation (R = 0.51, P = 0.002) and admission to the intensive care unit [odds ratio (OR) 17.65; 95% confidence interval (CI) 1.18–264.96; P = 0.04; n = 15]. Cell-based assays detected weakly positive NMDAR, LGI1 and CASPR2 antibodies in serum of 4/34 (11.8%) patients but not in CSF. The tissue-based assay showed anti-neuronal fluorescence in CSF from one individual, staining for Purkinje cells. In summary, whereas we did not detect active SARS-CoV-2 infection in the CSF, SARS-CoV-2 antibodies were prevalent. The absence of intrathecal antibody production points towards blood–brain barrier impairment as the origin of CSF SARS-CoV-2 antibodies. In contrast, CSF autoantibodies against neuronal surface antigens were rare. There was no evidence for a clinical correlate of these antibodies. We conclude that, rather than specific autoimmune neuronal injury, non-specific effects of critical illness including an impaired blood–brain barrier are more likely to contribute to neuro-COVID.