Divergent and self-reactive immune responses in the CNS of COVID-19 patients with neurological symptoms

Abstract

Individuals with coronavirus disease 2019 (COVID-19) frequently develop neurological symptoms, but the biological underpinnings of these phenomena are unknown. Through single-cell RNA sequencing (scRNA-seq) and cytokine analyses of cerebrospinal fluid (CSF) and blood from individuals with COVID-19 with neurological symptoms, we find compartmentalized, CNS-specific T cell activation and B cell responses. All affected individuals had CSF anti-severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibodies whose target epitopes diverged from serum antibodies. In an animal model, we find that intrathecal SARS-CoV-2 antibodies are present only during brain infection and not elicited by pulmonary infection. We produced CSF-derived monoclonal antibodies from an individual with COVID-19 and found that these monoclonal antibodies (mAbs) target antiviral and antineural antigens, including one mAb that reacted to spike protein and neural tissue. CSF immunoglobulin G (IgG) from 5 of 7 patients showed antineural reactivity. This immune survey reveals evidence of a compartmentalized immune response in the CNS of individuals with COVID-19 and suggests a role of autoimmunity in neurologic sequelae of COVID-19.

Keywords: COVID-19; SARS-CoV-2; autoimmunity; cerebrospinal fluid; neurological infection.

Graphical abstract

Figure 1

Distinct immunological landscape of CSF and PBMCs in individuals with COVID-19 with neurological symptoms (A) Schematic of the study design. CSF and blood were collected from individuals with COVID-19 and healthy control individuals. PBMCs and CSF cells were isolated, along with the CSF supernatant and plasma, for downstream analysis. (B) UMAP (Uniform Manifold Approximation and Projection) projections of 10x single-cell RNA sequencing of CSF and PBMCs of individuals with COVID-19 and healthy control individuals. (C) Venn diagram depicting upregulated interferon-stimulated genes (ISGs) and non-ISGs in dendritic cells in COVID-19 CSF compared with healthy control CSF based on the Interferome database. (D) Gene Ontology enrichment of genes upregulated in NK cells of individuals with COVID-19 in the CSF and peripheral blood. (E) Heatmap depicting cell-cell interactions between innate immune cells and adaptive immune cells by clustering shown in (B). The difference in interaction strength (COVID-19 interaction minus control interaction) is color coded and derived from log-scaled interaction counts using the CellphoneDB repository of ligands, receptors, and their interactions. Single-cell RNA-seq is derived from a total of 16 libraries plus 8 additional controls from Gate et al. (n = 3 for control CSF and PBMCs, n = 5 for COVID-19 CSF and PBMCs, and n = 8 from Gate et al.⁸).

Figure 2

Transcriptional characterization of T cells in CSF and PBMCs of individuals with COVID-19 (A) Reclustered UMAP projection of combined CSF and peripheral blood T cells, demonstrating CD4 and CD8 T cell subsets (two KLRG1+ clusters are distinguished by GZMB and IFNG expression; Figure S3). (B) Pie charts depicting the relative population frequency of different T cell subtypes found in CSF and PBMCs of control individuals and those with COVID-19. (C) Venn diagram depicting genes upregulated (adjusted p < 0.05) in CSF of individuals with COVID-19 compared with PBMCs of individuals with COVID-19 in Th1 and Th2 CD 4 T cells. (D) Gene Ontology analysis of genes that are upregulated in Th1 and Th2 cells, as depicted in (C). (E) Quad-Venn diagram of genes upregulated in CSF of individuals with COVID-19 compared with CSF of control individuals in CD8 T cells. Genes shared by the three effector CD8 T cell subtypes are circled. (F) Gene Ontology analysis of genes shared between the three effector CD8 T cell subtypes in (F). (G and H) Heatmap of Luminex-based cytokine profiling of CSF (G) and plasma (H) from individuals with COVID-19 and control individuals showing cytokines that were increased significantly increased in individuals with COVID-19 compared with control individuals (n = 6 CSF, n = 5 plasma). For each cytokine, two-tailed p values were calculated using Student’s t test. Data for each row were mean centered; each column shows data from one sample.

Figure 3

Localized central nervous system B cell responses in individuals with COVID-19 (A) Frequency of B cells as a percentage of all CSF cells in control individuals and those with COVID-19. Colors represent different individuals. (B) Re-clustered UMAP projection of B cells from CSF and blood. (C) Heatmap showing antibody binding in plasma (left) and CSF (right) to nine peptides from immunogenic regions of S, N, and ORF3a as well as whole S and N protein along with the RBD of the S protein. All data are represented as fold change of the fluorescent anti-IgG antibody signal over intra-assay negative controls. HC, healthy control. (D) Epitope frequency was ranked in each sample individually, and a difference in rank number for each cluster was graphed to determine CSF-enriched (positive) or plasma-enriched (negative) antibody epitopes. Two-tailed unpaired t test, ∗∗p < 0.01.

Figure 4

CSF antibodies reflect localized CNS infection (A) Mice were transduced with AAV-hACE2 intrathecally and intratracheally for expression in the brain and lungs, and SARS-CoV-2 was introduced intranasally to establish brain and lung infection. Mouse brains, lungs, CSF, and serum were collected on day 0 (before infection) and on days 3 and 7 after infection, and qPCR was performed to detect SARS-CoV-2 RNA. (B) Schematic of the experimental procedure for (C). Mice were given localized AAV-hACE2 to overexpress human ACE2 in the lungs (top), brain and lungs (center), or brain only (bottom). After 2 weeks, mice were infected with SARS-CoV-2. (C) ELISA against SARS-CoV-2 spike protein was performed with lung homogenates, serum, brain homogenates, and CSF. n = 5 for all three conditions. Two-tailed unpaired t test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005, ∗∗∗∗p < 0.0001) and one-way ANOVA were performed (lungs, p < 0.0001; serum, p = 0.002; brain, p = 0.0082; CSF, p = 0.0016).

Figure 5

Antigenic specificity of CSF- and PBMC-derived monoclonal antibodies (A) Bar graph depicting the frequency of the top five most expanded clones in PBMCs and CSF of affected individual 1. (B) Graph depicting overlap of clones found in CSF and PBMCs of an affected individual. Green indicates clones only found in CSF, orange clones shared between the CSF and PBMC, and red clones unique to PBMCs. The yellow box indicates clones that would fall under the top 10 most frequent clones in each compartment. (C) Heatmap showing CSF-derived (mAbs C1–C5) and PBMC-derived (mAbs P1–P3 and P5) mAb binding to nine peptides from immunogenic regions of S, N, and ORF3a as well as whole S and N protein along with the RBD of the S protein. mAb numbers correspond to the clone numbers from (A) and (B); PBMC clone 4 (mAb P4) did not express well as a mAb and was not used for subsequent studies. mAbs were screened in technical replicates. Heatmap values are mean fold change of the fluorescent anti-IgG antibody signal over intra-assay negative controls. (D) Sagittal mouse brain sections were immunostained with mAbs 1–9, and a representative whole-brain sagittal image is shown for PBMC-derived mAbs (mAb 7) and CSF-derived mAbs (mAb 4). An anti-hemagglutinin (anti-HA) antibody in the same IgG1 backbone was used as a negative control. Scale bars, 500 μm. (E) Select regions of immunostaining from mAbs 1–4. (i) mAb 1 immunostaining of cerebellar Purkinje cells (arrow) and the overlying molecular layer. (ii) mAb C2 immunostaining of cortical neuropil and occasional staining of neuron-like somata (arrow). (iii) mAb C3 immunostaining of large cells within the hilus of the hippocampus. (iv) mAb C4 immunostaining of mitral-like cells of the olfactory bulb (arrow). (v) mAb C4 immunostaining of pyramidal neurons (arrow) in CA3 of the hippocampus. (vi) mAb C4 immunostaining of neuronal cell bodies in layer II of the cortex (arrow). Scale bars, 10 μm.

Figure 6

Autoantibodies in the CSF of individuals with COVID-19 (A) Sagittal mouse brain sections were immunostained with CSF at 1:10 dilution (green) and the nuclear stain DAPI (blue). Anti IgG secondary only antibody negative control (left) and 4/6 control CSF (example CTRL 6, center) were not immunoreactive. In contrast, 5 of 7 COVID-19 CSF samples were immunoreactive (example CASE 3, right). Scale bars, 100 μm. CTRL, control. (B) Binary matrix indicating anatomic immunoreactivity of COVID-19 CSF at 1:10 dilution. (C) Select examples of COVID-19 CSF anatomic immunostaining of the hippocampus (n = 3; arrows, CA3; left column; scale bar, 100 μm), cerebrovasculature (top panel, second column arrow indicates endothelial staining [scale bar, 50 μm]; bottom panel arrow indicates a perivascular cell [scale bar, 10 μm]), olfactory bulb (n = 3, two shown; third column, top panel shows neuron-like cells; bottom panel, mitral cells; scale bars, 10 μM), and cortical neuron-like cells (n = 4, two cases shown; fourth column; scale bars, 10 μM). (D) Heatmaps of sequence-sharing peptides mapping to IFT88 (case 1, top) and THAP3 (case 3, bottom) that were enriched by CSF, shown with their corresponding enrichment by plasma. Rows, individual peptides; left two columns, technical replicates for case 1 (top) and case 3 (bottom). For COVID-19 and control columns, cell values represent the mean of log₁₀(fold change enrichment) of technical replicates. For case 1 and case 3, candidate IFT88 and THAP3 peptides, respectively, were enriched more by CSF than plasma. (E) The HEK293 overexpression cell-based assay was performed in technical replicates. A representative example demonstrates that case 1 CSF is immunoreactive to overexpressed RFP-IFT88 (CSF, green; anti-RFP, red; anti-IFT88 antibody, magenta). Scale bars, 10 μM. (F) Western blot validation of anti-THAP3 autoantibodies in CSF of case 3. CSF IgG (green) and anti-FLAG (red) recognize the same ∼25-kDa band in THAP3-overexpressing (OE) lysate but not untransfected (UN) lysate (arrow).