Background
A subset of people who recover from acute COVID-19 develop a range of emergent or persistent symptoms lasting months to years, collectively termed Long COVID (LC). We investigated whether autoimmune antibodies in the CNS associate with neurologic Long COVID symptoms (n-LC).
Methods
Participants with LC (WHO criteria) and cognitive symptoms or headache enrolled in The COVID Mind Study (n-LC, n=38). Controls (n=16) were healthy pre-pandemic participants (n=13) or post-COVID without any LC symptoms (n=3). Cerebrospinal fluid (CSF) was collected for research. Mouse brain sections were immunostained with CSF from n-LC and controls at 1:4 dilution and visualized with anti-human IgG to screen for anti-neural autoantibodies. CSF IgG was incubated with a human proteome-wide phage display library (PhIP-Seq) to identify candidate antibody targets enriched in n-LC compared to controls (enrichment threshold of sum rpk>100, fold change>10, z-score>3). We orthogonally validated candidate autoantigen by split-luciferase-binding assay (SLBA) and HEK293T cell overexpression.
Results
14/38 n-LC samples showed positive staining on mouse brain sections, with moderate to strong cytoplasmic and nuclear staining, specifically in the thalamus, cortex, hippocampus, and brainstem. 6/13 never-COVID and 3/3 post-COVID controls displayed similar cytonuclear staining patterns in mostly all brain regions. Using PhIP-Seq, we identified human peptides bound to antibodies present in n-LC CSF. Only the top 40 peptides that enriched in positively stained n-LC samples above negatively stained n-LC and control CSF are plotted in the heatmap (Figure1a). We chose Sox5 for further validation because, only Sox5 peptides showed enrichment in the highest number of participants (n=3), while most other peptides enriched in only one participant. Moreover, unlike other top peptides, Sox5 has CNS-specific functions involving neuronal signaling. Also, logistic regression showed SOX5 as the most discriminatory gene between positive and negatively stained n-LC samples. However, enriched SOX5 autoantigens did not validate with SLBA and cell-based overexpression assay.
Conclusions
Whole human proteome antigen discovery with CSF from individuals with and without n-LC did not identify a shared autoantigen in n-LC. There was no difference in the frequency and pattern of immunostaining on rodent brain slices between n-LC and controls. Our preliminary work did not identify a unified autoantibody signature in the CSF of patients with n-LC.