Clonally expanded B cells in multiple sclerosis bind EBV EBNA1 and GlialCAM

Abstract

Multiple sclerosis (MS) is a heterogenous autoimmune disease in which autoreactive lymphocytes attack the myelin sheath of the central nervous system. B lymphocytes in the cerebrospinal fluid (CSF) of patients with MS contribute to inflammation and secrete oligoclonal immunoglobulins^1,2. Epstein-Barr virus (EBV) infection has been epidemiologically linked to MS, but its pathological role remains unclear³. Here we demonstrate high-affinity molecular mimicry between the EBV transcription factor EBV nuclear antigen 1 (EBNA1) and the central nervous system protein glial cell adhesion molecule (GlialCAM) and provide structural and in vivo functional evidence for its relevance. A cross-reactive CSF-derived antibody was initially identified by single-cell sequencing of the paired-chain B cell repertoire of MS blood and CSF, followed by protein microarray-based testing of recombinantly expressed CSF-derived antibodies against MS-associated viruses. Sequence analysis, affinity measurements and the crystal structure of the EBNA1-peptide epitope in complex with the autoreactive Fab fragment enabled tracking of the development of the naive EBNA1-restricted antibody to a mature EBNA1-GlialCAM cross-reactive antibody. Molecular mimicry is facilitated by a post-translational modification of GlialCAM. EBNA1 immunization exacerbates disease in a mouse model of MS, and anti-EBNA1 and anti-GlialCAM antibodies are prevalent in patients with MS. Our results provide a mechanistic link for the association between MS and EBV and could guide the development of new MS therapies.

Extended Data Figure 1:. Analysis of B cell phenotypes in MS blood and CSF.

a-l, Flow cytometry data, a,b, representative flow cytometry plots are shown for a, blood and b, CSF. c, plasmablasts as percent of all B cells in MS blood and CSF, means ± SD of n=9 patient samples, **P = 0.004, two-tailed Mann-Whitney test. d, non-plasmablast B cell subsets as percent of all B cells in blood (red) and CSF (blue), means ± SD of n=8 patient samples, ***P = 0.0006, two-tailed Mann-Whitney test, Holm-Sidak corrected for multiple comparisons. e, Integrin alpha-4 expression in non-plasmablast B cells (red) and plasmablasts (blue), mean MFI ± SD of n=9 patient samples, ****P < 0.0001, **P = 0.0013, two-way ANOVA, Tukey adjusted for multiple comparisons, f, representative histogram showing integrin alpha-4 expression in non-plasmablast B cells (red) and plasmablasts (blue) in blood (top panel) and CSF (lower panel), g, HLA-DR expression in non-plasmablast B cells (red) and plasmablasts (blue) in blood and CSF, mean MFI ± SD of n=9 patient samples, ****P < 0.0001, ***P = 0.0002, two-way ANOVA, Tukey adjusted for multiple comparisons, and h, representative histogram showing HLA-DR expression in non-plasmablast B cells (red) and plasmablasts (blue) in blood (top panel), and CSF (lower panel). i, HLA-DR expression in patients carrying HLA-DRB1*15:01 (HLA-DR15, n=5) vs. other HLA-genotypes (non-HLA-DR15, n=4) in i, blood, and j, CSF, mean MFI ± SD, significance levels calculated with two-way ANOVA, k,l, Immunoglobulin classes in k, non-plasmablast B cells and l, plasmablasts in blood (red) and CSF (blue), mean MFI ± SD of n=9 patient samples, ****P < 0.0001, two-way ANOVA, Holm-Sidak adjusted for multiple comparisons. Plasmablasts, PB; unswitched memory B cells, UM; switched memory B cells, SM; double negative B cells, DN.

Extended Data Figure 2:. Extended BCR repertoire data.

a-i, Single-cell BCR repertoire sequencing data, a, individual repertoires from all CSF B cells (top row) and subdivided into CSF plasmablasts (middle row) and non-plasmablast B cells (bottom row) of n=9 MS patients. b, Individual repertoires of all CSF B cells (top row) and subdivided into CSF plasmablasts (middle row) and non-plasmablast B cells (bottom row) of n=3 control patients. Numbers indicate number of sequences, inner circle: colored wedges represent clonal expansions and grey area represents singleton antibody sequences, outer circle: immunoglobulin classes, red: IgG, blue: IgA, green: IgM, sequence locations in outer circle correspond to inner circle. No non-plasmablast B cells were sorted for MS12 and C5. Only plasmablasts were sorted for MS39. c, Clonality, percent of clonal sequences in CSF B cells are shown, comparing BCR repertoires of control patients (n=3) to MS patients (n=9). Data corresponds to data shown in (Fig. 1b) and is separated into immunoglobulin classes IgG (left), IgA (center), and IgM (right). Means ± SD of individuals’ repertoires are shown. d, Immunoglobulin class distribution, percent of IgG (left), IgA (center), and IgM (right) of all CSF B cells are shown for n=3 control patients and n=9 MS patients. Means ± SD of individuals’ repertoires are shown. e, IGHV and IGLV cumulated mutation count in plasmablasts in blood (red) vs. CSF (blue), means ± SD of n=9 patients samples. f, Mean HC CDR3 lengths (amino acid sequences) of plasmablasts in blood (red) vs. CSF (blue), means ± SD of n=9 patient samples. g-i, Immunoglobulin gene distribution in blood vs. CSF plasmablasts for g, IGLV, IGKV1–33, ****P < 10⁻⁶, IGLV3–21, ****P = 3×10⁻⁶ according to unpaired two-tailed Student’s t tests, Holm-Sidak adjusted for multiple comparisons, h, IGHJ, and i, IGLJ. Each dot represents the usage of one gene across n=9 MS patient repertoires in the respective compartments. Linear regression lines and 95% confidence interval are shown. j, Mass spectrometry data of purified CSF immunoglobulins, showing variable chain sequences that could be uniquely identified in singleton BCR sequences vs. plasmablast sequences, peptide-spectral matches (PSM) cutoff ≥10, means ± SD of n=9 MS patients, **P = 0.0012. k,l, Same mass spectrometry data set as in (j), showing variable chain sequences that could be uniquely identified in non-plasmablast BCR sequences vs. plasmablast sequences, means ± SD of n=7 MS patients, k, PSM cutoff ≥1, **P = 0.007, l, PSM cutoff ≥10, *P = 0.037. l, m, Single-cell sequencing efficacy in non-plasmablast B cells (red) vs. plasmablasts (blue) in CSF. Fraction of sequences that passed filter thresholds are shown as percentages of the number of sorted cells in the respective group, means ± SD of n=8 patient samples (no non-PB value for MS39). c,d,j-l, P according to unpaired two-tailed Mann-Whitney test. d-i, P according to unpaired two-sided Student’s t-test. Immunoglobulin heavy V gene, IGHV; Immunoglobulin heavy J gene, IGHJ; Immunoglobulin light V gene, IGLV; Immunoglobulin light J gene, IGLJ; peptide-spectral matches, PSM.

Extended Data Figure 3:. Phylogenetic trees of B cells from MS blood and CSF.

Blood plasmablasts (top rows) and CSF B cells (bottom rows) of n=9 MS patients and CSF B cells of n=3 control patients are shown. Each node represents the full-length heavy chain and light chain sequence of a single B cell. Trees are binned according to their IGHV families and genes, then the concatenated heavy chain and light chain sequences are clustered. IgG (red), IgA (blue), IgM (green). Smaller brighter circles indicate singleton B cells, larger darker circles indicate clonal expansions. Arrows indicate sequences that were expressed as mAbs, numbers indicate V-gene mutation loads in heavy and light chains. Immunoglobulin heavy V gene, IGHV.

Extended Data Figure 4:. Polyreactivity of recombinantly expressed antibodies.

a, ELISA data showing reactivity of recombinant mAbs against LPS (top), human insulin (middle), and dsDNA (bottom). Reactivity is represented in the order of decreasing reactivity to LPS in MS mAbs and control mAbs, respectively. Measurements were carried out in duplicates at 0.1, 1, and 10 μg/ml mAb concentrations and the area under the curve (AUC) for each mAb is shown from one experiment. Commercial anti-LPS antibody (cyan), MS39p2w174 (red), germline (orange), control mAbs (blue).

Extended Data Figure 5:. MS CSF mAb reactivity to EBV and GlialCAM antigens.

a, mAb reactivities to EBV virus lysates and recombinant EBV proteins as well as to other virus lysates. Z-scores for each antigen are shown, measurement of one microarray experiment, measured in 8 technical replicates. b, mAb reactivities to LPS, Insulin, and dsDNA to assess polyreactivity. Z-scores of area under the curve (AUC) of ELISA measurements at antibody concentrations of 0.1, 1, and 10 μg/ml are shown, each measurement was carried out in duplicates. c, mAb reactivities to GlialCAM proteins, peptides, and phosphorylated or citrullinated peptides. Mean reactivities (mean fluorescence intensity counts) are shown from one microarray experiment, measured in 8 technical replicates. Immediate early latency stage protein, IE; early, E; late, L; intracellular domain, ICD; extracellular domain, ECD; phosphorylated Serine, pSer; citrulline residue, Cit; _B - _E: duplicate probes of same / similar lysates and proteins (different preparations or batches).

Extended Data Figure 6:. MS CSF mAb reactivity to EBV peptides.

a, mAb reactivities to EBV peptides. Z-scores for each antigen are shown, measurement of one microarray experiment, measured in 8 technical replicates. Intracellular domain, ICD; extracellular domain, ECD; peptide mix, PM.

Extended Data Figure 7:. mAb reactivity to EBV peptides and extended structural data for the EBNA1_AA386–405 / MS39p2w174-Fab complex.

a, mAb reactivities of selected reactive mAbs against the selected reactive peptide antigens. Z-scores for each antigen are shown, measurement of one microarray experiment, measured in 8 technical replicates. b, ELISA-based alanine-scan on EBNA1_AA386–405, corresponding to (Fig. 2e). Mean OD (450 nm) ± SD from three independent experiments, each carried out in triplicates. c, 20x image of protein crystals in hanging drop. d, Asymmetric unit containing two peptide-Fab complexes in a diagonal orientation, heavy chain (red/pink), light chain (blue/cyan), peptide (black/gray). e, EBNA1_AA386–405 peptide and its 2mFoDFc map (contoured at 1σ) are shown, depicted on heavy chain (cyan) and light chain (pink) in surface representation. f,g, Amino acid sequences of variable regions of f, mAb MS39p2w174 heavy chain and g, light chain. Bold font: CDR, regular font framework regions. Of the germline variable genes (bottom rows), only residues that differ from MS39p2w174 sequence are shown, red: residues that closely interact with EBNA1_AA386–405 according to crystal structure. dots: gaps introduced during IMGT GapAlign for alignment and numbering purposes, numbers: residue numbers according to IMGT unique numbering. Intracellular domain, ICD; extracellular domain, ECD; heavy chain, HC; light chain, LC; complementarity determining region, CDR; framework region, FR; germline, GL.

Extended Data Figure 8:. Extended characteristics of GlialCAM_AA370–389 and immunofluorescence stainings with MS39p2w174.

a, Expression pattern of GlialCAM in human tissues (proteinatlas.org). b, Phage display PhiP-Seq data, showing alignment of Pro/Arg-rich region and adjacent residues of all phage display peptides enriched above 100 / 10⁵ reads. c, Immunofluorescence of mouse brain slices stained with (i) control antibody, and (ii-iv) MS39p2w174 (green) and DAPI (blue). (i,ii) full brain, scale bars: 2000 μm, (iii) magnification of hippocampus with prominent MS39p2w174 staining, scale bar: 400 μm, and (iv) olfactory bulb with prominent MS39p2w174 staining in the olfactory nerve (oln), glomerular (gl), and external plexiform layers (epl), but not the mitral (ml), internal plexiform (ipl), or granule cell (gcl) layers, scale bar: 100 μm. d, Immunofluorescence staining of primary rat oligodendrocytes with isotype control antibody (top panel) and MS39p2w174 (bottom panel). e, K562 cells in culture, wildtype (left) and transduced with full-length GlialCAM (right). f, Immunofluorescence with MS39p2w174 on WT K562 cells (top) and GlialCAM-tg K562 cells (center and bottom). White arrow: single K562 cell, orange arrow: high intensity MS39p2w174 staining on the cell boarder between transgenic K562 cells in bulks. d,f, Scale bars: 40 μm. c-f, representative micrographs of at least two experiments. g, Overview of phosphorylated residues in GlialCAM, identified by mass spectrometry (phosphoSite.org). The two phosphorylated serine residues of interest are indicated with arrows. h, ELISA, measuring binding of MS39p2w174 to native and citrullinated GlialCAM_AA370–389 peptides, means of n=2 independent experiments, each carried out in triplicates. Wildtype, WT; extracellular domain, ECD; intracellular domain, ICD; phosphorylated serine, pSer; citrulline residue, Cit.

Extended Data Figure 9:. Plasma reactivity against EBNA1 and GlialCAM proteins and peptides in healthy control individuals and MS patients.

a, ELISA measurement of antigen-specific IgG reactivity against peptides EBNA1_AA386–405, GlialCAM_AA370–389, phosphorylated GlialCAM_{AA370–389 pSer376}, 2x-phosphorylated GlialCAM_{AA370–389 pSer376 pSer377}, and scrambled peptide control in plasma samples of healthy control individuals (n=50) and MS patients (n=71). Means ± SD in each patient group is shown. Representative OD (450 nm) measurements of two independent experiments, each carried out in duplicates. ** P < 0.01, *** P < 0.001 according to two-tailed Mann-Whitney test, Tukey corrected for multiple comparisons. b, ELISA measurements of antigen-specific IgG reactivity against GlialCAM full-length protein, GlialCAM_AA370–389, and phosphorylated GlialCAM_{AA370–389 pSer376} in plasma samples of a separate cohort of healthy control individuals (n=31) and MS patients (n=67). Means ± SD across patient groups are shown. Representative OD (450 nm) measurements of two independent experiments, each carried out in duplicates. * P < 0.05, ** P < 0.01 according to two-tailed Mann-Whitney test, Tukey corrected for multiple comparisons. c, ELISA measurements of mAB MS39p2w174 binding to EBNA1_AA386–405, without interference as well as blocked with scrambled peptide control, EBNA1_AA386–405, and GlialCAM_{AA370–389 pSer376}, as a positive control to (Fig. 3q). Mean OD (450 nm) ± SD of quadruplicate measurements from n=1 experiment are shown. * P < 0.05, ** P < 0.01, *** P < 0.001 according to one-way ANOVA, Tukey corrected for multiple comparisons.

Extended Data Figure 10:. T cell response against EBNA_AA386–405.

a, ELISA data showing mouse plasma IgG responses against PLP_AA139–151 at the indicated timepoints pre and post EAE induction, for scrambled peptide immunized mice (blue, n=10) and EBNA1_AA386–405 immunized mice (red, n=10). Mean OD (450 nm) fold change ± SD, significance levels according to unpaired two-tailed Mann-Whitney test. Means ± SD, representative of three independent experiments, each carried out as triplicate measurements. b, T cell proliferation measurement by ³H-thymidine incorporation in splenocytes and lymph node cells of mice immunized with scrambled peptide (blue) and EBNA1_AA386–405 (red). Cells from n=10 mice per group were pooled and mean counts per minute (cpm) ± SD of triplicate measurements are shown. c-h, ELISA measurements of cytokines in cell culture supernatant of mouse splenocytes and lymph node cells of mice immunized with scrambled peptide (blue) or EBNA1_AA386–405 (red) and re-stimulated with the indicated peptides. Cells from n=10 mice per group were pooled and mean cpm ± SD of six replicate measurements are shown. ****P = 8.9×10⁻⁵, unpaired two-tailed Student’s t test, Holm-Sidak corrected for multiple comparisons. c, IFN-γ, d, TNF, e, IL-12, f, IL-10, g, IL-6, h, IL-17, *P < 0.05, significance levels according to unpaired two-tailed Mann-Whitney test, Holm-Sidak corrected for multiple comparisons. i, Representative Luxol Fast Blue stained spinal cords from scrambled peptide group (top panel) and EBNA1_AA386–405 group (bottom panel). Scale bars left images: 200 μm, right images: 50 μm. j, Statistical evaluation of Luxol Fast Blue scores, means of at least 4 coronal spinal cord sections per mouse and means ± SD for each group (n=9) are shown. ****P < 0.0001, unpaired two-tailed Mann-Whitney test. k,l, Flow cytometry data of PBMC from healthy control individuals (n=6, blue) and MS patients (n=7, red), showing percent of k, IFN-γ+ and l, IL-17+ CD4+ T cells in all CD4+ T cells. Mean MFI ± SEM are shown for the respective groups. Significance levels were assessed by two-way ANOVA, followed by FDR calculation using the two-stage step-up method of Benjamini, Krieger and Yekutieli, * FDR < 0.1. m, Flow cytometry data, representative dot plots are shown for two individuals from the data set presented in Fig. 4f. Healthy control individual (left) and MS patient MS16 (right). Expression levels of Granzyme-B (GZMB) and IFN-γ are presented under the indicated stimulations.

Figure 1:. B cell repertoires in MS blood and CSF.

a-c, Single-cell BCR repertoire sequencing data, a, individual repertoires from paired blood plasmablasts (top row) and CSF B cells (second row) of MS patients and CSF B cells of control patients (bottom row) (n=9 MS patients, n=3 control patients), numbers indicate number of sequences, inner circle: colored wedges represent clonal expansions and grey area represents singleton antibody sequences, outer circle: immunoglobulin classes, red: IgG, blue: IgA, green: IgM, sequence locations in outer circle correspond to inner circle. b, Clonality analysis, percent of clonal sequences in CSF B cells are shown, comparing BCR repertoires of control patients (n=3) to MS patients (n=9). Means ± SD of each group are shown. **P = 0.0091, two-tailed Mann-Whitney test. c, IGHV gene distribution in blood vs. CSF plasmablasts, each dot represents the usage of one IGHV gene across n=9 MS patient repertoires in the respective compartments. Linear regression line and 95% confidence interval is shown, IGHV1–2, ***P = 5.6×10⁻⁴; IGHV4–59, ***P = 9.2×10⁻⁴; IGHV3–7, *P = 0.025, unpaired two-tailed Student’s t tests, Holm-Sidak corrected for multiple comparisons. d, Mass spectrometry data of purified CSF immunoglobulins, showing variable chain sequences that were uniquely identified with mass spectrometry in the singleton BCR sequences vs. clonal sequences (peptide-spectral matches (PSM) cutoff ≥1), means ± SD of n=9 MS patients. ***P = 0.0002, two-tailed Mann-Whitney test.

Figure 2:. MS CSF B cell mAb reactivity to EBV proteins and interaction of MS39p2w174 with EBNA1_AA386–405.

a, Protein microarray data showing MS CSF mAb reactivities (z-scores) to viral lysates and EBV proteins, b, MS CSF mAB reactivities to EBNA1 peptides. Selected mAbs with highest reactivities are shown (n=36 (a) and n=37 (b) out of n=148). Data from one experiment carried out in 8 technical replicates. IE: immediate early, E: early, and L: late lytic stage, red: mAb MS39p2w174 and antigen EBNA1 / MS-associated region. c, Western blot analysis of recombinant EBNA1 (full-length and truncated proteins), stained on separate blots with commercial anti-C-terminal EBNAI antibody (top panel) and MS39p2w174 (bottom panel), molecular weight marker in kDa. d,e, ELISA data, d, overlapping peptide scan of MS39p2w174 binding EBNA1 peptides (20mers, 13AA overlap), means ± SD of n=4 independent experiments, each measured in duplicates. e, Alanine-scan, EBNA1_AA386–405 logo representation showing the contribution of each residue to binding of MS39p2w174. f-j, Crystal structure of MS39p2w174 in complex with EBNA1 AA386–405. f, Cartoon representation, showing EBNA1_AA393–401 in the binding groove. Peptide (red), HC (green/blue), LC (purple/yellow/orange), CDR loop colors correspond to annotations in (g). g, View of the binding groove from the top. Surface representation of the Fab with EBNA1_AA386–402 in stick representation. h, Cartoon and stick representation outlining close interactions. Major H-bond forming residues are represented as sticks. H-bonds < 3.2 Å are represented as black dashed lines. i, Magnification of peptide in the hydrophobic cage, j, magnification of region around Arg396 to emphasize polar contacts of HC residues with Arg396 and Arg397. k,l, Bio-layer interferometry measurement of MS39p2w174 (red) and germline (blue) affinity to EBNA1 full-length protein. k, K_D in M, mean K_D ± SD, calculated from four serial dilutions from one representative experiment out of three independent experiments, **P = 0.0043, unpaired two-tailed Student’s t test. l, association and dissociation curves corresponding to (k).

Figure 3:. Molecular mimicry between EBNA1 and GlialCAM.

a, Protein microarray data showing top 16 results of HuProt array for MS39p2w174, compared to 3 control mAbs, sorted from top to bottom by the ratio of MS39p2w174 / average of controls reactivity (left column, min: 89, max: 911). Raw counts are shown in the four columns on the right (min: 1, max: 36450). b, ELISA showing binding of MS39p2w174 and two control mAbs to recombinant proteins EBNA1_AA328–641 as well as GlialCAM full-length (AA34–416) and ICD (AA262–416). Means ± SD of n=3 independent experiments are shown, each carried out in triplicates. c, Western blot analysis of recombinant GlialCAM (full-length, ICD, and ECD), stained on separate blots with commercial anti-GlialCAM antibody (top panel) and MS39p2w174 (bottom panel), molecular weight marker in kDa. d, Fluorescent immunohistochemistry of mouse brain regions, stained with MS39p2w174, (i) cerebellum, scale bar: 150 μm, (ii) inferior colliculus (ic) and occipital cortex (occ) with perivascular glia (unfilled arrows) and glia limitans staining (filled arrows), scale bar: 60 μm, (iii) dentate gyrus with perivascular staining (unfilled arrows), scale bar: 40 μm. e,f, Bio-layer interferometry measurement of MS39p2w174 (red) and germline (blue) affinity to EBNA1 full-length protein. k, K_D in M, mean K_D ± SD, calculated from three serial dilutions from one representative experiment out of three independent experiments, *P = 0.0124, unpaired two-tailed Student’s t test, f, association and dissociation curves corresponding to (e). g, Alignment of AA sequences of EBNA1_AA386–405 and GlialCAM_AA370–389 and pointing out the central epitope region. h,i, Prediction of disorder with PONDR for h, EBNA1 and i, GlialCAM. High scores indicate disorder, red areas: epitope regions. j, ELISA data showing binding of MS39p2w174 to EBNA1_AA386–405 and GlialCAM_AA370–389 non-phosphorylated and phosphorylated at the indicated serine residues, means from n=2 independent experiments are shown, representative for 5 experiments, each carried out in duplicates. pSer / Sp: phosphorylated serine. k,l, Bio-layer interferometry measurement of MS39p2w174 affinity to GlialCAM 20mer peptides. K_D in M, mean K_D ± SD, calculated from four serial dilutions from one representative experiment out of three independent experiments, ****P < 0.0001, unpaired two-tailed ANOVA, Tukey corrected for multiple comparisons. pSer: phosphorylated serine residues. l, association and dissociation curves corresponding to (k). m, Protein microarray data showing mAb reactivities (MFI) to GlialCAM proteins, peptides, and phosphorylated peptides, as well as cross-reactivities to EBNA1 and other EBV proteins. Data from one experiment carried out in 8 technical replicates. ICD: intracellular domain, ECD: extracellular domain, pSer: phosphorylated serine residues. n-p, ELISA data showing human plasma reactivities against n, EBNA1 protein AA328–641, ****P < 0.0001, unpaired two-tailed Mann-Whitney test, o, EBNA1_AA386–405, **P < 0.0044, unpaired two-tailed Mann-Whitney test, and p, GlialCAM protein, ***P < 0.0002, unpaired two-tailed Mann-Whitney test, means ± SD from n=20 control individuals and n=36 MS patient samples. q, ELISA measurements of plasma antibody reactivity against EBNA1_AA386–405, without interference as well as blocked with scrambled peptide control, EBNA1_AA386–405, and GlialCAM_{AA370–389 pSer376}. Mean OD (450 nm), normalized to unblocked sample, ± SD of n=9 patient samples, each measured in quadruplicates. e,k,n-p, P values according to unpaired two-sided Student’s t-test, q, one-way ANOVA, Dunnett corrected for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 10⁻³, ****P < 10⁻⁴.

Figure 4:. Anti-EBNA1_AA386–405 immunization exacerbates autoimmune-mediated demyelination in vivo.

a,b, ELISA data showing mouse plasma IgG response at the indicated timepoints pre and post EAE induction, for scrambled peptide immunized mice (blue, n=10) and EBNA1_AA386–405 immunized mice (red, n=10) against a, EBNA1_AA386–405, mean OD (450 nm) fold change ± SD, ****P < 0.0001, unpaired two-tailed Mann-Whitney test, Holm-Sidak corrected for multiple comparisons, and b, GlialCAM ICD, mean OD (450 nm) fold change ± SD, **P = 0.0022, *P = 0.0152, unpaired two-tailed Mann-Whitney test, Holm-Sidak corrected for multiple comparisons. Means ± SD, representative of three independent experiments, each carried out as triplicate measurements. c, EAE scores of mice immunized with scrambled peptide (blue, n=9) and EBNA1_AA386–405 (red, n=7), 3 weeks prior and on the day of EAE immunization (day 0), means of clinical scores ± SEM, *P<0.05 unpaired two-tailed Mann Whitney test. d,e, Spinal cord histology, d, representative H&E-stained spinal cords from scrambled peptide group (top panel) and EBNA1_AA386–405 group (bottom panel). Scale bars left images: 200 μm, right images: 50 μm. e, Statistical evaluation of H&E score, means of at least 4 coronal spinal cord sections per mouse and means ± SD for each group (n=9), **P = 0.0012, unpaired two-tailed Mann-Whitney test. f, Flow cytometry data of PBMC from healthy control individuals (n=6, blue) and MS patients (n=7, red), showing percent of IFN-γ+ GZMB+ CD8+ T cells in all CD8+ T cells. Means ± SEM are shown for the respective groups. Significance levels were assessed by two-way ANOVA, followed by FDR calculation using the two-stage step-up method of Benjamini, Krieger and Yekutieli, * FDR < 0.1.