Elevated neuroinflammation, autoimmunity, and altered IgG glycosylation profile in the Cerebral Spinal Fluid of Severe COVID-19 patients

Abstract

Background and Objectives:

A spectrum of neurologic complications associated with COVID-19 are well documented. While neuroinflammation in the brain of COVID-19 patients likely contributes to these complications, the mechanisms of neuroinflammation and correlates of neurologic complications remain elusive, especially since the etiologic pathogen of COVID-19, SARS-CoV-2, minimally invades the CNS. This study aimed to evaluate markers of neuroinflammation, IgG glycosylation patterns indicative of pro- or anti-inflammatory state, and prevalence of brain auto-reactive antibodies in the CSF of COVID-19 patients and their relationship to brain neuropathology.

Methods:

We evaluated the CSF of 11 deceased unvaccinated COVID-19 donors and 13 matched non-COVID-19 controls. Markers of neuroinflammation, IgG glycosylation patterns, and brain auto-reactive antibodies were assessed, along with their correlation to brain neuropathology. Statistical analyses were performed to compare groups and assess relationships between variables, using non-parametric tests and bootstrap analysis.

Results:

COVID-19 CSF showed higher levels of neopterin and ANNA-1, markers of neuroinflammation and autoimmunity, respectively, and lower IFN response compared to non-COVID-19 donors. In brain regions of high microglial activation, IL4 and RANTES were significantly increased. SARS-CoV-2 was undetectable in the CSF and brain of COVID-19 donors, yet anti-SARS-CoV-2 CSF antibodies were detected. Fucosylated IgG were associated with Spike IgG, CSF protein, and soluble CD14, whereas afucosylated bisecting IgG were inversely correlated with Spike IgG. Sialic acid containing IgG were positively correlated with IL1β and TNFα. These associations were not found in non-COVID-19 donors. Inflammatory agalactosylated fucosylated IgG (G0F) were associated with infiltrating CD4+ T cells in the brains of COVID-19 donors. COVID-19 donor CSF displayed higher levels of auto-reactive antibodies to human brain antigens compared to non-COVID-19 donors and donors with positive autoantibodies showed higher levels of neopterin.

Discussion:

These data describe increased neuroinflammation and autoreactive antibody markers in the CSF of COVID-19 donors and suggest that IgG glycosylation and autoimmunity may contribute to COVID-19 pathology, highlighting potential mechanisms underlying the neurologic complications associated with COVID-19.

Background:

Infection with SARS-CoV-2, the etiologic pathogen of COVID-19, is associated with both acute and delayed neurologic complications .^1,2 The mechanisms for these complications remain unclear, given inconsistent evidence of SARS-CoV-2 neuroinvasion in humans .^3–8 Most post-mortem analyses of brain and CSF from COVID-19 donors do not detect viral RNA or proteins.^9–11 Laboratory studies using in vivo and in vitro models demonstrating infection may not represent the human condition by either overexpressing receptors for entry (ACE2) or describing its effects in immature primary cell lines.^12–15

Neuropathological case studies of COVID-19 brains continue to exhibit instances of ischemic infarcts, hemorrhages, and inflammation of the olfactory bulbs and medulla without consistent viral presence.^17–19 To date, this suggests systemic inflammation in COVID-19 mediates neural damage through a dysregulated brain-periphery axis. More recently, the presence of intrathecal autoreactive antibodies following COVID-19 infection has been reported.^20–27 While some otherwise healthy individuals have pre-existing intrathecal autoantibodies, studies suggest they are induced and enriched in COVID-19. These autoantibodies have been shown to target multiple human antigens involved in viral defense and neurogenesis.^24,28,29

IgG glycosylation patterns are implicated in autoimmune and metabolic diseases, as well as bacterial and viral responses.^30–35 IgG subtypes are post-translationally modified with N-glycans at the asparagine-297 residue on the fragment crystallizable (F_c) portion of the antibody (Supplementary Fig. 1). These modifications include a combination of additional galactose, sialic acid, core fucose or bisecting arms.³⁶ These various modifications form immune complexes and interact with F_c receptors to regulate cytokine and chemokine production. For example, bisecting IgG glycans without core fucose (afucosylated) are associated with increased antibody-dependent cellular cytotoxicity (ADCC).³⁷ Those with increasing galactose and sialic acid exhibit more anti-inflammatory effects and improved half-life in serum.^38–40 Conversely, the lack of sialic acid promotes inflammation. Modifications with galactose positively correlate with disease severity. Core fucosylated antibodies with lower degrees of galactose are more inflammatory versus antibodies with increasing galactose.³¹ As such, IgG glycosylation patterns can potentially be used as biomarkers for inflammatory or protective immune responses.

Glycosylation patterns in the sera of COVID-19 patients have been characterized, although without clear results. For example, lower proportions of fucosylation were reported among hospitalized COVID-19 intensive care unit patients compared to outpatient controls.^42,43 However, additional studies found no differences between fucosylation levels between disease severity.^44,45 These inconsistent results may partially be explained by the transient nature of IgG glycosylation pattern following disease and an inability to accurately describe COVID-19 severity clinically.

There are no studies to date evaluating the IgG glycosylation pattern in the CSF of COVID-19 patients. Glycosylation patterns have previously been described in multiple sclerosis (MS) and demyelinating neuropathy. Increased bisecting modifications were reported in MS patients, who also showed a reduction of afucosylated and galactosylated species.⁴⁶ In patients with demyelinating neuropathy, sialylation modifications abrogated complement-dependent inflammatory effects by inhibiting F_c and C1_q myeloid cell binding.⁴⁷ In this study, we evaluated the glycosylation pattern of bulk CSF IgGs from COVID-19 donors and matched non-COVID-19 controls, determined the presence of several neuroinflammatory markers in the CSF, and described the neuropathology from paired brains. We also screened for autoreactive antibodies from the CSF of COVID-19 donors using an ex vivo mouse model.

Materials and methods

Ethical statement:

Donor CSF and brains were collected in accordance with United States federal, state, and institutional review board approved guidelines on human research. Human tissue sections were collected per IRB approved protocol at Vanderbilt University (IRB 192003). The brains were placed in a biorepository in the Rush Alzheimer’s Disease Center following protocols approved by a Rush University Medical Center Institutional Review Board. All animal procedures followed approved protocol at Rush University Medical Center Institutional Animal Care and Use Committee (IACUCC 20–064) and were conducted in accordance with National Institutes of Health guidelines for housing and care of laboratory animals. Rush University Medical Center is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-International.

CSF collection and brain harvest: COVID-19 donors:

CSF from 13 COVID-19 donors was obtained by inserting a needle through the corpus callosum into the lateral ventricle of the brain to extract between 8–10 mL of fluid, which was then frozen at −80°C until further processing. All brain autopsies were performed from multiple sites in Nashville with a previously described standardized method.⁴⁸ After brain removal, brain was fixed for at least 30 days in 4% paraformaldehyde and transferred to Rush Alzheimer’s Disease Center (RADC). Brains were collected between 2020–2021, after the WHO declared COVID to be a pandemic (March 11, 2020). Non-COVID-19 donors: CSF from 11 non-COVID-19 donors was obtained from Rush Alzheimer’s disease center (RADC), a long-standing NIA funded center with longitudinal clinical assessment of AD and non-AD participants. Paired CSF and brains were provided to clinically match co-morbid non-COVID-19 conditions of the COVID-19 cohort in this study. More information about RADC is found at https://www.rushu.rush.edu/research/departmental-research/rush-alzheimers-disease-center. All CSF samples included paired brain pathology assessments by board certified neuropathologists.³⁷

SARS-CoV-2 RNA measurement in CSF:

RNA from CSF donors was isolated using miRNeasy RNA kit (Qiagen, Germantown, MD). Presence of SARS-CoV-2 viral RNA was evaluated with CDC 2019-Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel primers and probes (Cat: 2019-nCoVEUA-01; Atlanta, GA, USA) and iTaq one-step RT-PCR mix (Bio-Rad, Vancouver, WA, USA). Briefly, 8 μL of isolated RNA from CSF was included with primer/probe mix and one-step RT-PCR master mix. Primers against SARS-CoV-2 nucleocapsid (N1, N2) and RNase P (internal control) were included for analysis (Supplementary Table 1). RNA extraction and RT-PCR were performed in accordance EUA protocol for SARS-CoV-2 diagnostic panel (https://www.cdc.gov/coronavirus/2019-ncov/lab/virus-requests.html). Per diagnostic criteria, any cycle threshold values above 37 were considered not positive for the gene of interest.

SARS-CoV-2 viral protein measurement in brain:

Immunostaining for SARS-CoV-2 nucleocapsid (N) protein (dilution, 1:500; Sino Biological, Wayne, PA; Cat #40143-R001) was performed on 6 μm sections of the olfactory bulb and medulla by using Leica Bond RXM autostainer with standard bond epitope antigen retrieval method (citrate-based pH 6.0 epitope retrieval solution).

SARS-CoV-2 Spike (Trimer) IgG ELISA:

Whole CSF from donors was centrifuged 1,500 x g for 10 minutes before diluting 1:25 and 1:100 with assay buffer per manufacturer protocol. An additional 1:10 dilution was made to load sample on SARS-CoV-2 Spike Trimer coated plate (Cat: BMS2325; Thermo Fisher, St. Louis, MO). Briefly, wells were pre-washed with buffer and loaded with pre-diluted samples. Assay buffer was added, and wells incubated for 30 minutes at 37°C. Wells were aspirated and washed with a wash buffer. HRP conjugate was added and allowed to incubate for 30 minutes at 37°C. The wells were aspirated and washed before addition of substrate. After 15-minute incubation, absorbance at 450 nm was read (BioTek, Winooski, VT). ELISA samples were loaded in duplicate and repeated. Corresponding standard curves were generated for each replicate and sample readout were obtained for values within respective dilution within standard curve from 4-parameter logistic regression. Final values represent the interpolated value from the average of samples ran in duplicate with corresponding 95% confidence intervals from regression analysis. Lower limit of detection: 144 units/mL. Upper limit of detection: 150,000 units/mL.

Detection of SARS-CoV-2 S1 and nucleocapsid (NC) proteins in CSF:

S1, Whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before 1:100 and 1:1000 dilutions per manufacturer protocol (Cat: KBVH015–10; Krishgen Biosystems, Mumbai, India) were made. Briefly, samples were loaded onto plates and allowed to incubate for 1 hour at 37°C. Wells were washed before addition of and incubation with HRP conjugate for 1 hour at 37°C. Wells were washed again, and substrate solution was added. Sample with substrate were incubated at RT for 30 minutes before adding stop solution and readout at 450 nm (BiotTek, Winooski, VT). A standard curve was generated from kit provided reagents and a 4-parameter logistic curve was used for sample interpolation. Lower limit of detection: 60 ng/mL. Nucleocapsid (NC):Whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes. Samples loaded on plate were undiluted, 1:10, and 1:100 diluted and allowed to incubate for 1 hour at 37°C before washing and addition of HRP conjugate per manufacturer protocol (Cat: KIT40588; Beijing, China). Samples were incubated with HRP conjugate for 1 hour at 37°C before addition of substrate. Substrate was incubated with sample at 37°C for 15 minutes before addition of stop solution and readout at 450 nm (BioTek, Winooski, VT). Lower limit of detection: 34.61 pg/mL.

Clinical assessment:

Patients were admitted or transferred to hospitals in the middle Tennessee area (e.g., Vanderbilt, Nashville Veterans Affairs Hospital, Summit Regional Hospital, Alive Hospice) and clinical records were extracted when available. Written notes were assessed for demographics, history of vascular risk factors, vascular diseases, and neurological illness.

Neuropathology assessment:

All cases were systematically and uniformly reviewed by a board-certified neuropathologist for accurate neuropathologic assessment, blinded to clinical neurologic status, CSF biochemical data. The fixed brain was bisected into hemispheres which were cut coronally into 1 cm slabs. Tissue blocks from the multiple brain regions, including midfrontal, middle temporal, entorhinal with amygdala, inferior parietal, calcarine, anterior cingulate, posterior cingulate, anterior temporal tip, and inferior orbital frontal cortices, hippocampus, basal ganglia, thalamus, frontal periventricular, parietal periventricular, anterior watershed, posterior watershed, cerebellum, olfactory bulb, substantia nigra, midbrain caudal, pons, and medulla regions were collected unilaterally and fixed for 48–72 hours in 4% paraformaldehyde, embedded in paraffin, and cut into 6-μm sections. Additional tissue blocks with suspected gross pathology, such as infarcts and hemorrhages were taken and confirmed their presence and age by microscopic examination. Cerebrovascular pathology: Arteriolosclerosis was assessed in the basal ganglia and atherosclerosis in vessels at the base of the brain using a semiquantitative grading system from 0 (none) to 6 (severe) as described previously. Cerebral amyloid angiopathy (CAA) was assessed semi quantitatively in meningeal and intracortical vessels from four cortical sections immunostained for β-amyloid.⁵⁰ For this study, the vessel disease was consider “present” when pathologic changes in the vessels were assessed as moderate-to-severe.⁴⁹ The presence and age (acute, subacute, and chronic) of brain macroscopic infarcts and hemorrhages (subdural/subarachnoid/parenchymal) were confirmed by microscope.⁴⁹ In addition, all 6 μm H&E stained sections of the tissue blocks described above were reviewed under a microscope to identify the presence and age of microscopic infarcts and hemorrhages, in the absence of visualized gross infarct in the region. Infarcts and hemorrhages were categorized into recent and old where recent included those with age acute or subacute changes and old including those chronic infarcts or hemorrhages deemed by the neuropathologist to be at least 3 to 6 months in age. Other microscopic histopathology was also recorded from all H&E slides from all blocked regions. Neurodegenerative pathology: Neurodegenerative pathology for Alzheimer’s disease (AD), Parkinson’s disease (PD), Lewy body disease (LBD), and limbic predominant age-related TDP-43 encephalopathy-neuropathologic change (LATE-NC) was assessed in decedent brains. Briefly, the National Institute on Aging-Alzheimer’s Association criteria was used for the neuropathologic assessment of AD that requires an “ABC” score.⁵¹ The “ABC” score is derived from Thal Phases of Aβ deposition (“A”),⁵¹ Braak staging for neurofibrillary degeneration (“B”),⁵² and CERAD neuritic plaque score (“C”).⁵³ A pathologic diagnosis of AD included intermediate and high AD neuropathologic change cases. Immunohistochemistry with alpha-synuclein (Zymed LB 509; 1:50; pSyn, 1:20,000; Wako Chemicals) was performed to detect LBs from seven brain regions (midfrontal, middle temporal, entorhinal, inferior parietal, and cingulate cortices, amygdala and substantia nigra). A modified McKeith criteria was used to assign the case with either nigral predominant, limbic, neocortical, or amygdala predominant type Lewy body disease.⁵⁴ LATE-NC was assessed from 8 brain regions by immunohistochemistry with phosphorylated monoclonal TAR5P1D3, pS409/410 antibodies (dilution 1:100 from Ascenion, Munich, Germany, and dilution 1:400 from Millipore Sigma, Burlington, MA).⁵⁵

Microglial activation and brain CD4 + T cell infiltration

Microgliosis:

Microglial activation severity was determined by a semi-quantitative method based on enlargement of cell soma and thickening of processes (indicators of activation) in the olfactory bulb and/or medulla. Specifically, immunostaining with HLA-DP, DQ, DR mouse monoclonal antibody, clone CR3–43 (dilution, 1:100; Agilent technologies, CA, USA; catalog M077501–2-HLA DP DQ DR) was performed on 6 μm sections of the olfactory bulb and medulla. The degree of microgliosis was classified as none, mild, moderate, or severe, using a semi-quantitative approach based on enlargement of cell soma and thickening of processes. For analytical analyses, we dichotomized microgliosis into low and high where high included those with severe microgliosis present in at least one of the assessed brain regions and low included those having either mild or moderate microgliosis. CD4 + and CD8 + T cells: Infiltration assessed from meninges, parenchyma, and perivascular location of medulla region immunostained for CD4 rabbit monoclonal antibody, clone EPR6855 (dilution, 1:500; Abcam, Waltham, MA, USA; catalog, Ab133616,) and for CD8 mouse monoclonal antibody, clone 4B11 (dilution, 1:75; Leica Biosystems Inc, Buffalo Grove IL, USA; catalog, CD8–4B11-L-CE). A semi-quantitative grading system from 0 (none) to 5 (severe) was used to assess the severity of CD4 + and CD8 + T lymphocyte infiltration as none, sparse, mild, moderate, or severe in three locations (meninges, parenchyma, and perivascular). Finally, we used the maximum scores from the three locations scores to categorize into none (0), minimal-mild (1–2), and moderate-to-severe (3 and more). Expert raters, including board-certified neuropathologist, examined all tissue samples simultaneously. These raters were unaware of other pathology and demographic information, ensuring a blinded evaluation.

IgG isolation:

Whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before a 1:1 dilution with PBS was made. Diluted CSF was buffer exchanged with PBS by centrifugation at 1,000 RCF for 1 minute with 40kDa Zeba desalting columns (Thermo Scientific, St. Louis, MO) before incubation for two hours in end-over-end rotisserie at 4°C with 150 μL Protein G Agarose beads per manufacturer protocol (Thermo Scientific, St. Louis, MO). The bead mixture was then centrifuged through Pierce Spin-X column and washed with 1% Triton X-100 three times. The washed bead mixture was incubated with IgG Elution Buffer (Thermo Scientific, St. Louis, MO) for 10 minutes at 4°C. IgGs were centrifuged at 1,500 RCF for 1 minute and collected in 10% vol/vol 1M Tris-HCl pH 8.00. Two elutions were performed. Eluates were buffer exchanged with 40 kDa Zeba desalting columns (Thermo Scientific, St. Louis, MO). Concentration of recovered IgGs was obtained via absorbance at 280 nm.

IgG glycan sample preparation and glycosylation analysis:

Purified IgG was next treated with dithiothreitol (DTT) for 5 minutes at 100 °C to break up the heavy and light chains. The heavy chain of IgG was resolved on a 4–15% gradient polyacrylamide gel (Bio-Rad, Hercules, California) and the gel band was excised following Simply Blue SafeStain (Thermo Fisher, Waltham, MA). Next, in-gel digestion was performed on the excised heavy chain gel bands. Each gel band was washed in 100 mM ammonium bicarbonate (AmBic)/acetonitrile (ACN) and reduced with 10 mM DTT at room temperature (RT) for 45 minutes. Cysteines were alkylated with 50 mM iodoacetamide in the dark for 45 minutes at room temperature. Gel band was washed in 100 mM AmBic/ACN prior to adding 600 ng Lys-C for overnight incubation at RT. Supernatant containing peptides was transferred into a new tube. Peptides were extracted at RT for 10 minutes with gentle shaking in 50% ACN/5% formic acid (FA), and peptides solution was collected. Further, peptides were extracted with 80% ACN/5% FA and followed with 100% ACN. The peptides were lyophilized and reconstituted in 5% ACN/0.1% FA. Peptides were analyzed by LC-MS/MS using a Dionex UltiMate 3000 Rapid Separation nanoLC and a Q Exactive^™ HF Hybrid Quadrupole-Orbitrap^™ Mass Spectrometer (Thermo Fisher Scientific Inc, San Jose, CA). The peptide samples were loaded onto the trap column, which was 150 μm x 3 cm in-house packed with 3 μm C18 beads. The analytical column was a 75 μm x 10.5 cm PicoChip column packed with 3 μm C18 beads (New Objective, Inc. Woburn, MA). The flow rate was kept at 300 nL/min. Solvent A was 0.1% FA in water and Solvent B was 0.1% FA in ACN. The peptide was separated on a 60-minute analytical gradient from 5% to 50% of Solvent B. The mass spectrometer was operated in ‘Full MS scan’ mode. The source voltage was 2.50 kV and the capillary temperature 320 °C. Full MS scans were acquired from 400–2000 m/z at 60,000 resolving power and automatic gain control (AGC) set to 3×10⁶. MS data was processed using Skyline (Version 21.1). The integration and correction for the chromatographic peaks of 18 glycans were performed manually. Three of the most intense precursor ions of each glycan were selected, summed, and exported as the quantitative value of corresponding glycan. Proportions of each glycan pattern were grouped by fucosylation (F), sialylation (S), bisection (B), and galactosylation (G) numbers. Additional classifications of agalactosylated (G0), monogalactosylated (G1) and di-galactosylated (G2), as well as G-ratio (total galactosylated:agalactosylated) was also calculated (Supplementary Table 2).

Cytokine/chemokine and inflammatory marker measurements:

CSF from donors were loaded on Milliplex MAP Human Cytokine/Chemokine Magnetic Bead Panel in accordance with manufacturer protocol (Cat: HCYOMAG-60K-19C; MilliporeSigma, Burlington, MA). Standard curves were created to interpolate values from either 4-parameter logistic or cubic spline curves. Lower limits of detection in pg/mL: IFNα2: 2.9, IFNγ: 0.8, IL-10: 1.1, IL-1RA: 8.3, IL-1a: 9.4, IL-1β: 0.8, IL-2: 1.0, IL-3: 0.7, IL-4: 4.5, IL-5: 0.5, IL-6: 0.9, IL-7: 1.4, IL-8: 0.4, IP-10: 8.6, MCP-1: 1.9, MIP-1α: 2.9, TNFα: 0.7, VEGF: 26.3, RANTES: 1.2. Other neuroinflammatory markers not included in the Milliplex were calculated by individual ELISAs: Type 1 antinueronal nuclear antibody (ANNA-1): whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before preparing dilutions of 1:1,000, 1:5,000, and 1:10,000. Briefly, samples and standards were loaded onto precoated wells and allowed to incubate for 90 minutes per manufacturer protocol (Cat: MBS2533381; MyBioSource, San Diego, CA). Biotinylated detection antibody was added after initial incubation and allowed a further 1 hour at 37°C incubation before washing. Following, HRP conjugate was added to samples and incubated for 30 minutes at 37°C. Washing was repeated before addition of substrate and incubation for 15 minutes at 37°C. Stop solution was added and readout was obtained at 450 nm (BioTek, Winooski, VT). Values were interpolated from 4-parameter logistic regression from kit provided standards. Lower limit of detection: 3.75 ng/mL. Soluble CD14: whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before 1:50 and 1:200 dilutions were made. Briefly, assay diluent was added to wells before addition of dilutions (Cat: DC140; R&D Systems, Minneapolis, MN). Samples were allowed to incubate for 3 hours at RT before washing. Antibody conjugate was added and incubated at RT for 1 hour. Samples were washed and substrate solution was added, incubated for 30 minutes at RT, and stopped with stop solution before readout at 450 nm (BioTek, Winooski, VT). A 4-parameter logistic curve was calculated to interpolate unknown sample values from kit provided standards. Lower limit of detection: 125 pg/mL. Soluble CD163: whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before 1:10 and 1:100 dilutions were made. Briefly, assay diluent was added to wells before addition of dilutions (Cat: DC1630; R&D Systems, Minneapolis, MN). Samples were allowed to incubate for 2 hours at RT before washing. Antibody conjugate was added and incubated at RT for 2 hours. Samples were washed and substrate solution was added, incubated for 30 minutes at RT, and stopped with stop solution before readout at 450 nm (BioTek, Winooski, VT). A 4-parameter logistic regression was calculated to interpolate unknown sample values from kit provided standards. Lower limit of detection: 0.613 ng/mL. Neopterin: whole CSF from donors was centrifuged at 1,500 RCF for 10 minutes before diluting 1:5 and 1:50. Briefly, standards and samples were loaded onto precoated plates (Cat: MBS760192; MyBioSource, San Diego, CA). Enzyme conjugate and neopterin antiserum were also added to wells with samples before 90-minute incubation at RT in the dark. Samples were washed before the addition of substrate. Substrate incubated for 10-minutes at RT in the dark before addition of stop solution and plate readout at 450 nm. Sample concentration was interpolated from logit-log regression from kit provided standards. Conversion between nmol/L to ng/mL was calculated per manufacturer protocol. Lower limit of detection: 0.7 nmol/L.

Self-reactive antibody screen and immunofluorescence:

We used a protocol using ex vivo mouse brain slices to detect human autoreactive antibodies to neuronal antigens via indirect immunofluorescent assay ¹. Briefly, 12-week-old C57BL/6 mouse brain sections were cryosectioned at 10 μm increments and transferred to SuperFrost Plus (Fisher) microscope slides. Slides were fixed in a 63% PIPES buffer and 37% Paraformaldehyde (4%) solution for 10 minutes at room temperature. Following a wash in 1X-PBS, the slides were blocked with blocking buffer (10% normal donkey serum, 0.01% Sodium Azide, and 0.1% Triton X-100 in 1X- PBS) for 20 minutes at room temperature. Tissues were incubated with 1:10 diluted whole CSF in blocking buffer to act as primary overnight at 4°C as previously described.²⁹ Controls were incubated with staining buffer (1:1 diluted blocking buffer with 1X- PBS). An unstained, secondary only, and two human-specific primary antibodies (GFAP and Iba) were used as negative controls (1:500). Following incubation with primary antibody, slides were washed 2 times in 1X- PBS. Secondary antibody to human (donkey anti-human AF-647 conjugated) were applied to the appropriate samples at 1:500 for 30 minutes at room temperature. Slides were then washed 2 times in 1X- PBS and counterstained with DAPI for 10 minutes before coverslips were applied. Then, slides were sealed with nail polish and stored at 4°C until they were imaged under the Keyence Microscope BZ-Series (Keyence, Osaka, Japan). Tissues were imaged at 40x magnification at multiple points in the CA3 region of the hippocampus. Analysis was performed using the hybrid cell count function on the BZ-X800 Analyzer to count the number of anti-human IgG positive cells divided by the number of DAPI-stained nuclei. The percentages were calculated over five random 40x magnification fields of view for each donor.

Statistical analysis:

Statistical analysis and graphs were performed on GraphPad Prism 9.4.1 (GraphPad, San Diego, CA). Kruskal-Wallis analysis of variance (ANOVA) or Mann-Whitney U tests were performed with α=0.05 for significance. Spearman correlation coefficients were performed in R (R Foundation for Statistical Computing, Vienna, Austria). Bootstrap analysis was performed with 25,000 simulations on 11 Spearman ranks for corresponding p-value distributions.⁵⁶ Correlation coefficients above 0.60 were statistically significant under α=0.05 within the Bootstrap parameter (Supplementary Table 3). Dunn’s (Bonferroni) corrected t-test for multiple comparison post-hoc pairwise analysis was performed between variables.

Results

Cohort characteristics

We evaluated CSF and paired brains from 11 post-mortem COVID-19 donors (median age 67 years, IQR 61–77) and 13 non-COVID-19 donors (median age 77 years, IQR 70–81) (Table 1). Non-COVID-19 donors were matched to COVID-19 donors on neurodegenerative disease (Alzheimer’s and Lewy body dementia, P=1.00). No differences between history of infarcts (gross or micro) were noted (P=0.68). History of vascular diseases were common between COVID-19 and non-COVID-19 donors. All but three participants had vascular findings such as cerebral arteriosclerosis, atherosclerosis, or amyloid angiopathy. A detailed neuropathological findings of brain autopsies of the COVID-19 donors has been published.⁵⁷

Table 1.

Descriptive and clinical characteristics of cohort.

	COVID-19 N (Col %)	Non-COVID-19 N (Col %)	Fisher Exact Test P-Value
Total	11 (46)	13 (54)
Demographic characteristics
Age (median, IQR)	67 (61–77)	77 (70–81)	0.1833 a
Male sex	7 (64)	5 (38)	0.4136
White	9 (81)	13 (100)	0.2755 b
Black or African American	1 (9)	0 (0)
Other race	1 (9)	0 (0)
Hispanic/Latino ethnicity	2 (18)	13 (100)	0.1993
Clinical characteristics
Cardiovascular risk factor present c	9 (82)	12 (92)	0.5761
Gross infarcts present	3 (27)	5 (38)	0.6792
Micro infarcts present	3 (27)	5 (38)	0.6792
Alzheimer’s Disease	1 (9)	2 (15)	1.0000
Lewy Body Dementia	3 (27)	4 (31)	1.0000
Neocortical type	2 (18)	2 (15)
Limbic type	1 (9)	1 (8)

^a Unpaired, two-sample t-test

^b Chi-square statistic

^c Includes clinical history of hypertension, diabetes, dyslipidemia, obesity, coronary artery disease, cardiovascular disease, or congestive heart failure

Anti-SARS-CoV-2 IgG is present in the absence of SARS-CoV-2 virus in the CSF of COVID-19 donors.

To assess SARS-CoV-2 infection in the CNS, we measured SARS-CoV-2 viral RNA using the novel COVID-19 diagnostic panel from the Centers for Disease Control and Prevention with an internal control. SARS-CoV-2 RNA was not detected in any samples of CSF measured. All CSF samples were also negative for SARS-CoV-2 S1 antigen and nucleocapsid antigen. Brain samples from COVID-19 donors were also negative for nucleocapsid in olfactory bulbs and medulla sections. However, ten of the 11 CSF samples evaluated were positive for anti-SARS-CoV-2 Spike Trimer IgG, suggesting a previous infection in the CNS that was cleared, or potential antibody penetrance to the brain either under homeostatic condition or an inflammatory state (Fig. 1A).

Figure 1. Anti-SARS-CoV-2 Spike Trimer IgG and glycosylation profile of IgGs in CSF of COVID-19 and non-COVID-19 donors.

(A) Anti-Spike Trimer levels in the CSF were quantitated by ELISA. Data shown represents the average interpolated value of samples ran in duplicate with corresponding 95% confidence intervals from 4-parameter logistic curve. Sample 10 reached the upper limit of detection and does not have a corresponding 95% confidence interval. Sample 8 was below the lower limit of detection and does not have an interpolated value. (B) Percentages of bulk glycans with core fucose, bisecting (afucosyl-bisecting and fucosyl-bisecting, and sialylation. (C) Ratio of agalactosylated (G0) bulk glycans divided by digalactosylated (G2) glycans and presented as G-ratio. (D) Global glycan profile of COVID-19 and non-COVID-19 donor CSF. Group mean and standard error of the mean (SEM) are plotted. Mann-Whitney U Test between COVID-19 and non-COVID-19 glycan patterns are reported. #: P<0.05, ##: P<0.01.

Bulk IgG glycosylation profile in COVID-19 and non-COVID-19 donors:

To characterize the glycosylation pattern of IgGs from the CSF of COVID-19 and non-COVID-19 donors, bulk IgGs were isolated and analyzed by mass spectrometry.⁵⁸ Proportions of core fucosylation, sialylation, and bisections (afucosylated and fucosylated) were calculated from glycoproteomic analysis for both COVID-19 and non-COVID-19 donors (Fig. 1B). COVID-19 donors showed higher levels of total sialylation compared to non-COVID-19 controls (Fig. 1B, P=0.01). The G ratio designates the relative abundance of galactose residues on the glycan structures of the antibody, which is calculated by dividing all IgG species with no galactose (G0) by those with two galactose moieties (G2). COVID-19 donors showed a decreased G-ratio compared to non-COVID-19 donors (Fig. 1C, P=0.01) although both groups would be considered inflammatory (G ratio greater than 2). Species of G2F, G1, and G1FS1 were higher in COVID-19 donors compared to non-COVID-19 donors (P=0.003, P=0.01, and P=0.004 respectively). G2FS1 glycoform trended higher in COVID-19 compared to non-COVID-19 donors (P=0.055). Conversely, non-COVID-19 donors had higher levels of G0N and G2FS2 compared to COVID-19 donors (P=0.007 and P=0.02, respectively) (Fig. 1D). These findings indicate that although there are subtle glycosylation differences between COVID-19 and non-COVID-19 CSF, both would be considered inflammatory.

Differential expression of markers of autoantibodies, immune activation, and cytokines/chemokines in the CSF of COVID-19 donors:

Multiplex immunoassays were performed to assess the levels of soluble markers of inflammation and autoimmunity in COVID-19 and non-COVID-19 donors. Type-1 antineuronal nuclear antibody (ANNA-1) was higher in COVID-19 compared to non-COVID-19 donors (Fig. 2A, P<0.0001). ANNA-1 is a human autoantibody implicated in paraneoplastic neurological syndromes and small-cell lung carcinomas.⁵⁹ Further, COVID-19 donors had increased levels of neopterin, a marker of macrophage activation and gliosis (P=0.0009). These levels were significantly elevated in COVID-19 donors, despite no difference in total CSF protein between COVID-19 and non-COVID-19 donors (P=0.33). Conversely, levels of IFNγ, IFNα2, TNFα, IL-15, and IL-1α were all decreased in COVID-19 donors compared to non-COVID-19 donors (Fig. 2B, P<0.0001, P<0.0001, P=0.015, P<0.0001, and P<0.0001, respectively). These findings indicate that some inflammatory markers are elevated in COVID-19 while other cytokine responses are lower, especially those relating to innate sensing which are consistent with reports that SARS-CoV2 can suppresses innate immune responses.^28,60

Figure 2: Neuroinflammatory and autoimmune markers and cytokine/chemokine levels in COVID-19 and non-COVID-19 CSF matched controls.

(A) Neuroinflammatory and autoimmune markers (neopterin, ANNA-1) and total CSF protein (mg/mL) between COVID-19 and matched non-COVID-19 donors were evaluated. (B) Cytokine levels (IFNα2, IFNγ, TNFα, IL-15, IL-1α) between COVID-19 and non-COVID-19 donors were evaluated. Minimum to maximum points is shown for each neuroinflammatory marker with interquartile range. Mann-Whitney U Test between COVID-19 and non-COVID-19 levels are reported with corresponding exact P-values. Abbreviations: ANNA-1 antineuronal nuclear antibody-1.

Relationship between IgG glycosylation and neuroinflammatory markers in CSF of COVID-19 donors

Spearman correlations were calculated between each IgG glycosylation pattern and neuroinflammation markers to assess for association. To confirm statistical significance among Spearman ranks with a small sample size, an 11-rank Spearman Bootstrap with 25,000 simulations was performed (Supplementary Fig. 2). Spearman coefficients greater than [0.60] were confirmed to be statistically significant. In COVID-19 donors, total CSF core fucosylated glycan patterns were positively associated with CSF anti-Spike IgG (Fig. 3B, r=0.75), whereas afucosyl-bisecting glycan patterns were negatively associated with CSF anti-Spike IgG (r= −0.74). Notably, core fucosylated glycan patterns were positively associated with CSF soluble CD14 among COVID-19 donors (r=0.63). Core CSF fucosylation and total CSF protein were also positively associated in COVID-19 donors (r=0.82). Sialylated glycan patterns were positively associated with CSF IL-1β and CSF TNFα (r=0.63 and r=0.61, respectively). Lastly, neuroinflammatory markers such as neopterin and soluble CD14 were positively associated with CSF IP-10 and total protein in COVID-19 donors (Fig. 3C). Specifically, levels of CSF neopterin were positively associated with CSF IP-10 in COVID-19 (r=0.92). Further, CSF soluble CD14 was associated with both CSF IP-10 in COVID-19 (r=0.64) and total CSF protein in COVID-19 (r=0.89). Importantly none of these correlations were observed in non-COVID-19 donors. Supplementary Table 3 shows all correlation coefficients performed between COVID-19 and non-COVID-19 donors. These findings indicate that IgG glycan patterns are linked to some neuroinflammatory markers in the CSF of COVID-19 patients, which suggests that specific glycosylation changes in IgG could play a role in the neuroinflammatory processes associated with COVID-19, potentially serving as biomarkers for disease severity and targets for therapeutic intervention.

Figure 3: Spearman correlations between IgG glycosylation profiles and inflammation markers.

(A) Individual pairwise Spearman correlations were calculated between glycosylation patterns and inflammation markers. A 25,000 simulation 11-rank Spearman Bootstrap was performed to determine coefficients greater than |0.60| reach α=0.05 level of significance (Supplementary Table 2). (B) Select pairwise Spearman correlations between neuroinflammatory marker and glycan pattern reaching α=0.05 level of significance is reported between COVID-19 and non-COVID-19 donors. (C) Select pairwise Spearman correlations between neuroinflammatory markers reaching a=0.05 level of significance is reported between COVID-19 and non-COVID-19 donors. All correlation coefficients are listed in Supplementary Table 3.

Relationship between IgG Glycan Patterns, Neuroinflammation, and Neuropathological Findings in COVID-19 Patients:

Select clinical characteristics and neuropathological findings were investigated to assess the relationship between IgG glycan patterns and neuroinflammation. Within the olfactory bulbs and the medulla, significantly higher levels of CSF IL-4 and RANTES were associated with increased microglial activation (Fig. 4A). IFNα2 trended towards an inverse association with microglial activation, although this relationship did not reach significance. Donors with evidence of a recent hemorrhage had elevated levels of IP-10 and neopterin compared to those without hemorrhages (Fig. 4B). Lastly, the individual proportion of G0F species in CD4+ T-cell infiltration was different between low, moderate, and severe infiltration, however no pairwise differences between groups was noted due to small sample size (Fig. 4C). No other noted brain pathology (e.g.: recent infarcts, chronic infarcts, or neurodegenerative pathology) were associated with IgG glycan patterns or neuroinflammatory markers measured in the CSF.

Figure 4: Cytokine, chemokine, and glycosylation patterns in clinical outcomes and pathological findings in COVID-19 donors.

(A) Levels of IFNα2, IL-4, and RANTES were evaluated between low or high microglial activation in the medulla and olfactory bulb of COVID-19 donors. (B) Levels of IP-10 and neopterin in acute hemorrhage events in COVID-19 donors. (C) Agalacostylated core fucose glycan patterns are reported between CD4+ T-cell infiltration levels in the medulla. (D) Representative immunohistochemistry with HLA-DP, DQ, DR showing low (1–4) and high (5–8) levels of microglia in the medulla (1, 3, 5, 7) and olfactory region (2, 4, 6, 8). Scale bar: 50 mm (3, 4, 7, 8) and 4 mm (1, 2, 5, 6). (E) Representative immunohistochemistry with CD4 antibody from three donors represents none, minimal-mild, and moderate-severe CD4+ T cell infiltration level within meninges, parenchyma, and perivascular spaces. Scale bar: 50 mm. Minimum to maximum points are shown with interquartile range. Mann-Whitney U test between cytokine and chemokine markers with α=0.05 of significance. Kruskal-Wallis ANOVA (P=0.0121) with post-hoc pairwise Dunn’s test with no significant differences between none, mild, and severe CD4+ T cell infiltration in the medulla with agalactosylated core fucose glycan pattern.

Microgliosis was considered low in six donor brains (n=6/11; 55%) and high in the other five (n=5/11; 45%) (Table 2; Fig 4D). In addition, microglial nodules were present in two brains (n=2/11; 18%). Most of the brains had minimal-mild infiltration of CD4+ T cells (n=8/11; 73%) and CD8+ T cells (n=10/11; 91%) and only one decedent had moderate-to-severe CD4+ and CD8+ T cells in the brain. Two donor brains had no presence of CD4+ T cell infiltration (n=2/11; 18%) (Table 2; Fig 4E).

Table 2.

Select clinical and pathological findings in COVID-19 donors

	Present N (Row %)	Not present N (Row %)
Dementia	1 (9)	10 (91)
Gross and/or micro infarcts	8 (73)	3 (27)
Chronic	3 (27)	8 (73)
Acute	7 (64)	4 (36)
Parenchymal and/or subarachnoid hemorrhages	3 (27)	8 (73)
Old	0 (0)	11 (100)
Recent	3 (27)	8 (73)
Microglial activation in medulla and olfactory bulb
Low	6 (55)
High	5 (45)
CD4+ T cell infiltration in parenchymal, meningeal, and/or perivascular space
None	2 (18)
Minimal-mild	8 (73)
Moderate-severe	1 (9)
CD8+ T cell infiltration in parenchymal, meningeal, and/or perivascular space
None	0 (0)
Minimal-mild	10 (91)
Moderate-severe	1 (9)

COVID-19 donors show increased levels of CSF-derived autoreactive brain antibodies:

To assess if isolated antibodies from the CSF exhibit self-reactivity against brain target proteins, CSF from COVID-19 and non-COVID-19 donors was incubated with C57/B6 mouse brain sections before the application of a human-specific fluorophore conjugated secondary antibody, as previously described.²⁹ A secondary antibody only control was used as negative controls and demonstrated lack of fluorescent signals (Fig. 5A). All COVID-19 donors (n=11) were evaluated for self-reactive antibodies and a subset of non-COVID-19 donors (n=8) were processed and underwent hybrid cell counting. Representative images of COVID-19 (Fig. 5B) and non-COVID-19 (Fig. 5C) donors are shown. Among COVID-19 donors, 82% of donors showed positive anti-human antibody fluorescence (n=9/11), whereas only 25% of non-COVID-19 donors showed positive anti-human antibodies (n=2/8). Of the two non-COVID-19 donors who exhibited CSF antibody autoreactivity in the brain, one had accompanying neocortical type Lewy body dementia. COVID-19 donors had significantly higher levels of positive anti-human antibody cells compared to non-COVID-19 donors (Fig. 5D, P=0.0359), determined by five random fields of view per donor and reported as mean and standard error per donor. Positive anti-human antibody cell percentage was also evaluated between Lewy body dementia (LBD), cerebrovascular disease (CVD), or Alzheimer’s Disease (AD) pathology status. No significant differences were found (Fig. 5E, P=0.96, P=0.55, P=0.16, respectively). Lastly, the CSF of donors with autoantibodies showed higher levels of neopterin compared to CSF of donors without autoantibodies (Fig. 5F, P=0.0130).

Figure 5: Self-reactive antibodies in the CSF of COVID-19 and non-COVID-19 donors.

Coronal sections of 12-week-old C57BL/6 mouse brains were incubated with (A) anti-human secondary antibody only (AlexaFluor-647 channel) or 2-(4-amidinophenyl)-1H -indole-6-carboxamidine (DAPI) staining only. Coronal sections of 12-week-old C57BL/6 mouse brain sections from CA3 region of hippocampus were incubated with donor CSF and anti-human secondary antibody. Representative images from (B) seven COVID-19 CSF donors and (C) six non-COVID-19 donors are shown. Scale bar in overlay: 100 μm. (D) Hybrid automatic cell counting was performed for five random fields of view per donor and summarized as percent positive self-reactive cells per donor. (E) Percent positive self-reactive cells per donor are evaluated between donors with neurodegenerative diseases Lewy body disease (LBD), cerebrovascular disease (CVD), and Alzheimer’s Disease (AD). Mean and standard error of the mean are reported between groups. Mann-Whitney U test performed with α=0.05 level of significance.

Discussion

Neuroinflammation is considered to be a contributing factor in COVID-19 associated neurologic complications, however, its etiology is less clear. In this study, we investigated the brains and matched CSF of 11 COVID-19 donors and 13 non-COVID-19 donors for markers of neuroinflammation, autoantibodies, IgG glycosylation pattern, and neuropathology of paired samples.

Although we did not detect SARC-CoV2 in the brain or CSF, we found anti-Spike IgG in the CSF of COVID-19 donors. The detection of SARS-CoV-2 CSF antibodies without detecting the virus itself in the brain could indicate several possibilities, including a compromise in the Blood-Brain Barrier (BBB) where the virus has breached the BBB to some extent, allowing antibodies to enter the brain but the absence of detectable virus in the brain may indicate that the virus has not actively replicated within the brain tissue itself, or it has been cleared before it could be detected. Alternatively, the presence of anti-Spike antibodies in the CSF could be a result of systemic immune response rather than local CNS infection. In some cases, the antibodies may have been produced in response to viral antigens elsewhere in the body but have entered the CSF through passive diffusion or active transport mechanisms. The last possibility for this discrepancy is latent infection or a viral reservoir, which is unlikely given that SARS-CoV-2 is not a latent virus.

In COVID-19 donors, we found significantly increased neuroinflammatory markers despite equal CSF protein levels. Soluble markers of inflammation within COVID-19 and non-COVID-19 CSF showed increased markers of inflammation and autoreactivity such as neopterin and ANNA-1 but decreased interferon levels in COVID-19 CSF. Both ANNA-1 and neopterin were increased in COVID-19 compared to non-COVID-19 donors. Notably, type I and type II interferon levels of IFNα2 and IFNγ were significantly decreased in COVID-19 donors compared to non-COVID-19 donors. This adds to increasing evidence of preferential downregulation of type I interferons in COVID-19 and its prognosis for more severe disease.⁶⁰ Levels of TNF-α, a potent antiviral cytokine, were also lower in COVID-19 donors. Coordinated cytokine immune response to virus includes type I and II interferons with TNF-α. Interferon and TNF-α expression has been previously shown to be dependent.⁶¹ A decrease in each cytokine in our COVID-19 donors suggests viral mediated downregulation or exhausted immune response in these critically ill donors.

Fucosylated IgG glycans were associated with soluble CD14 in the CSF of COVID-19 donors but not among non-COVID-19 donors, suggesting this particular IgG glycosylation pattern may be indicative of macrophage/microglia activation as CD14 is shed from these myeloid cells. Other soluble markers of inflammation such as IP-10, and neopterin were associated in COVID-19 donors but not in non-COVID-19 donors. Further, inflammatory short chain glycosylation patterns were associated with infiltrating CD4+ T cells in the medulla. Lastly, we found evidence of autoantibodies at higher levels in COVID-19 CSF compared to matched non-COVID-19 CSF. For donors with autoantibodies, CSF levels of neopterin were higher compared to donors without autoantibodies. Together, these findings support increasing evidence of inflammation in COVID-19 brains and their correlates of immune dysregulation, including autoimmunity.²¹

IgG glycosylation pattern can inform inflammatory states in disease, with longer sugars and sialyation patterns linked to anti-inflammation while shorter sugars and less sialyation linked to inflammatory state. Of interest, the G ratio, which designates the extent of galactose on antibodies was lower in COVID-19 than non-COVID-19 donors in the CSF. Much of what we know about association between glycosylation/sialylation states of antibodies and inflammation/anti-inflammation states are based on antibodies in plasma. To date, there are no data to determine whether these same patterns have the same correlates in the CSF as in the plasma. As such, without functional assays it is difficult to determine whether those CSF glycosylation patterns are linked to an anti or a pro inflammatory state in the CSF. At least in the plasma, increasing glycan modifications offer more protective immune modulation by decreasing inflammation, while core fucosylated and agalactosylated IgGs are associated with systemic inflammation.^33,38,41,47 Recently, plasma IgG glycosylation pattern was assessed among a cohort of COVID-19 patients with mild or severe disease as well as among those vaccinated against SARS-CoV-2.⁶² Inflammatory glycans were associated with severe disease and anti-inflammatory glycans were associated with mild disease. Interestingly, among vaccinated individuals, a decrease in inflammatory glycans was associated SARS-CoV-2 neutralization in an in vitro assay.

We found that although both COVID-19 and non-COVID-19 had an inflammatory glycan pattern, the non-COVID-19 had a higher ratio of agalactosylated glycans (G0) to digalactosylated glycans (G2). This G ratio is usually denoted as inflammatory, however, it also suggests an overall decreased humoral immune response, especially in cases of rheumatoid arthritis compared to spondylarthritis and irritable bowel.^63,64 Our COVID-19 donors had marked immune response from current infection and prolonged ICU stay, which may partially explain increased galactose compared to non-COVID-19 donors without humoral immune responses in the absence of infectious process. Further, sialylation and afucosyl-bisecting patterns have been shown as anti-inflammatory and specific to enveloped viruses.^65,66 We similarly report an association between afucosyl-bisecting IgG modifications and COVID-19 severity. Specifically, our COVID-19 donors with increasing afucosyl-bisecting antibodies had decreased anti-Spike IgG levels (Fig. 3A).

Studies have reported changes in IgG glycosylation patterns with increasing age.^{2, 3} These findings suggest age influences the glycosylation pattern of IgG antibodies, potentially affecting their structure, function, and interactions with immune cells and pathogens. Our COVID-19 donors consisted of individuals of older age (>65), as did the non-COVID-19 donors. However, controlling for age did not alter the associations reported here, suggesting that factors other than age is contributing to these associations. Additionally, extrafollicular B cell activation is reported among severe COVID-19 patients, which correlated with higher antibody secretion.⁶⁷ Yet, this feature of B cells is found among individuals with severe COVID-19, indicating that it is not necessarily protective.

We found varying levels of CD4+ and CD8+ T cells among COVID-19 donors, which is consistent with previous studies that have documented the presence of CD4+ and CD8+ lymphocytes in COVID-19 post-mortem brains.^68–70 Additionally, we found varying levels of microgliosis among COVID-19 brain donors. However, increased levels of agalactosylated glycan patterns were found as CD4+ lymphocyte presence in the medulla increased, however, given that the limited number of cases exhibiting no CD4+ or moderate-to-severe CD4+ lymphocyte presence, this observation is not robust and requires further expanded studies to confirm a statistically significant correlation. Neopterin, which is secreted by macrophages in response to IFNγ, sCD14, and sCD163, and marks microgliosis/macrophage activation, and IP-10 and were significantly increased in donors with acute hemorrhages. These findings suggest a significant activation of macrophages in brains of COVID-19 donors and potentially associated clinical outcomes. These observations suggest possible mechanisms of inflammatory responses arising independent of productive infection of SARS-CoV-2 in the brain.

We also found that CSF-derived antibodies were more self-reactive in COVID-19 donors compared to non-COVID-19 donors. Secondary analyses evaluating self-reactivity between neurodegenerative diseases such as Alzheimer’s, Lewy body dementia, and cerebrovascular disease confirmed this finding. Our results add to increasing evidence of autoantibodies found in COVID-19 patients and support increasing evidence it is linked to SARS-CoV-2 infection status in contrast to existing neurodegenerative disease status.²⁷

Deciphering CSF IgG glycome in relationship to CSF antibody autoreactivity and inflammatory markers in neurological complications of COVID-19, including long-haul COVID-19 may be informative of disease processes that is more likely to be an aberrant immunologic response rather than direct effects of the virus on brain cells. Understanding these mechanisms can inform therapeutic interventions.

Supplementary Material

Supps

Supplementary Figure 1 Glycosylation modifications of IgGs. (A) N-297 glycosylation spectrum of inflammatory (G0F) to anti-inflammatory glycosylation profile (G2FS2). (B) Glycosylation profile afucosylated bisecting glycoform (G0N), which has increased antibody effector function. Abbreviations: GlcNAc: N-Acetylglucosamine, Fuc: Fucose, Man: Mannose, Gal: Galactose, NANA: N-Acetylneuraminic acid (sialic acid).

Supplementary Figure 2 Spearman 11-Rank Bootstrap. (A) Permutation distribution of Spearman correlation coefficient in 11-rank Spearman Bootstrap with 25,000 simulations.

Supplementary Figure 3 Individual donor glycosylation profiles. Glycosylation patterns were determined and plotted for each donor. Proportions for total core fucose (fucosylated), sialic acid (sialylated) modifications, and bisecting modifications with or without core fucose (fucosyl- or afucosyl-bisecting) are reported.

Supp table

Supplementary Table 1: CDC 2019-nCoV EUA Primers/Probe Sequences

Supplementary Table 2: Glycoform analysis

Supplementary Table 3: Group means and SEM for glycans between COVID-19 status.

Supplementary Table 4: Spearman Correlation Coefficient

Abbreviations:

SARS-CoV-2

Severe acute respiratory syndrome coronavirus

CAA

Cerebral amyloid angiopathy

LATE-NC

limbic predominant age-related

TDP-43

encephalopathy-neuropathologic changes

AD

Alzheimer’s Disease

IgG

Immunoglobulin G

ANNA-1

type-1 antineuronal nuclear antibody

Availability of data and materials:

Data will be available per request at www.radc.rush.edu.