Permissive central tolerance plus defective peripheral checkpoints licence pathogenic memory B cells in CASPR2-antibody encephalitis

Abstract

Autoimmunity affects 10% of the population. Within this umbrella, autoantibody-mediated diseases targeting one autoantigen provide a unique opportunity to comprehensively understand the developmental pathway of disease-causing B cells and autoantibodies. While such autoreactivities are believed to be generated during germinal centre reactions, the roles of earlier immune checkpoints in autoantigen-specific B cell tolerance are poorly understood. We address this concept in patients with CASPR2-autoantibody encephalitis and healthy controls. In both groups, comparable and high (~0.5%) frequencies of unmutated CASPR2-reactive naïve B cells were identified. By contrast, CASPR2-reactive memory B cells were exclusive to patients, and their B cell receptors demonstrated affinity-enhancing somatic mutations with heterogenous binding kinetics. These effector molecules possessed epitope-dependent pathogenic effects in vitro neuronal cultures and in vivo. The unmutated common ancestors of these memory B cells showed a distinctive balance between strong CASPR2 reactivity and very limited binding across the remaining human proteome. Our results are the first to propose mechanisms underlying autoantigen-specific tolerance in humans. We identify permissive central tolerance, defective peripheral tolerance and heterogenous autoantibody binding properties as sequential pathogenic steps which licence CASPR2-directed pathology. By leveraging the basic immunobiology, we rationally direct tolerance-restoring approaches in CASPR2-antibody diseases. This paradigm is applicable across autoimmune conditions.

Figure 1:. Central and peripheral immune tolerance in CASPR2-antibody encephalitis and healthy controls

(A) Left: Representative flow cytometry cell gating strategy to isolate memory (red) and naïve (blue) B cells for bulk (top) and single (bottom, dotted line) B cell cultures. Representative fluorescence microscopy images of culture supernatant detection of secreted CASPR2-reactive IgG or IgM, using a live cell-based assay with CASPR2-EGFP expressing HEK293T cells. DAPI = nuclear stain; scale bar = 10μm. Right: mRNA was extracted from CASPR2-reactive single B cell cultures to amplify and clone paired heavy and light chain BCR sequences. These were expressed in HEK293S cells to secrete CASPR2-reactive IgG (blue) or IgMs (pink). (B) The proportion of bulk B cell culture wells containing CASPR2-reactive IgM or IgG from patients (n=6) and healthy controls (HC; n=5). Black bar depicts median value. (C) Donut plots visualizing the frequency and clonality of CASPR2-reactive BCRs in single cell cultures. Numerator total number of CASPR2-reactive BCRs, denominator = total number of cells screened. The absolute percentage of CASPR2-reactive BCRs is then shown for two patients (P1 and P2) and two HCs (H5 and H6).

Figure 2:. CASPR2 autoreactivity is enhanced by somatic hypermutation

(A) Immunoglobulin heavy and light chain variable region mutation counts across B cell subsets in both patients (CASPR2-Ab-E) and healthy controls (HC). (B) Raw SPR traces representing the soluble extracellular domain of human CASPR2 binding to immobilized CASPR2 memory mAbs (upper row) and their corresponding UCAs (middle row). E06 UCA mAb did not express. UCA binding as an IgM was tested via a live cell-based assay with “+” indicating CASPR2-reactivity (bottom row). (C) K_D (M) quantification of mAbs via surface plasmon resonance (left). Non-significant K_D Pearson’s correlations with heavy (middle) and light (right) chain mutation count.

Figure 3:. CASPR2 mAbs bind distinct conformational epitopes in native tissues

(A) No differences in number of CASPR2 peptides immunoprecipitated by peptide phage display versus isotype control mAb. (B) Representative immunohistochemistry staining (all inlays = isotype control mAb) of CASPR2 mAbs on fixed murine hippocampal brain tissue with hippocampus visualized (upper left; mAb=E07) in wild type (WT) and CASPR2 knockout (−/−; upper middle) tissue. Representative immunofluorescent staining using E08 on live hippocampal neurons (upper right) fixed rat cerebellum (lower left; DAPI = nuclear counterstain), live human iPSC-derived sensory neurons (lower middle; mAb=E08), and live mouse dorsal root ganglia (lower right; mAb=E08). Costaining markers to identify cell types include microtubule associated protein 2 (MAP2) and beta-tubulin-III. Brain tissue scale bar = 500 μm, all others = 10 μm. (C) End-point titrations (1: dilutions) of binding across human and mouse CASPR2 live cell-based assays. mAbs are colored as in subpanel F. (D) Cartoon representation of CASPR4 single domains knocked-into full length human CASPR2-EGFP (top). Heatmap depicts CASPR4 knock-in domains that abrogated mAb binding (bottom). (E) Discoidin (Disc) and Laminin3 (L3) domain amino acid sequences from human and mouse, showing non-conserved amino acids in green. The tan and pink highlight Disc and L3 domains throughout the figure, respectively. Predicted CASPR2 protein structure (right, alphafold ID 5Y4M) showing non-conserved amino acids (green), Disc and L3 domains. (F) Binding competition map (left) demonstrating displacement of a prebound mAb (x-axis) by a competing mAb pre-conjugated with Alexafluor 594 (y-axis). Percentage inhibition is defined as percentage reduction of fluorescence intensity of the respective mAb compared to isotype control and represented by a forced directed network of epitope binning (right). Arrowhead indicates direction of binding competition and line thickness and intensity denote percentage inhibition. Shapes denote patient sample.

Figure 4.. Diverse pathogenic potentials of CASPR2 mAbs

(A) Representative images used to calculate colocalisation of pHrodo-conjugated IGG4 mAbs and CASPR2-eGFP reflecting receptor internalization by CASPR2-expressing HEK293T cells (left). Internalisation quantified by mean pHrodo fluorescence intensity mAbs in the presence and absence of dynamin inhibition (Dynasore; right). (B) Pearson’s correlation of receptor internalization to mAb binding parameters; bottom x-axis label denotes internalization . Background colors indicate the 3 epitope pockets as presented in Fig.3; top x axis label annotates epitope. (C) Representative images of CASPR2 and MAP2 expression after 4 hour application of IgG1 (open circles, left) or IgG4 mAbs (closed circles); Scale bar = 5 μm. CASPR2 puncta intensity summarized on the right with *p<0.05, non-parametric Kruskal-Wallis with Dunn’s multiple comparison post-hoc test. mAbs colored as in A. (D) Representative images (left) and quantification (right) of synaptic AMPAR and PSD95 expression similar to (C). (E) Representative tracings (top), single average event and cumulative probability (middle), and amplitude quantifications (bottom) of AMPAR-mediated mEPSC recordings of pyramidal neuronal cultures. * = p<0.05, Kruskal-Wallis test with Dunn’s multiple comparison post-hoc test. (F) Cartoon depiction of intracerebral mAb injection into bilateral hippocampal CA3 regions (left). Open field (middle) and light-dark box (right) behavioral test performance was assessed at 6 and 9 hours post-injection, respectively. *=p<0.05, t-test. (G) Summary model cartoon.

Figure 5:. Distinct CASPR2-reactive BCRs mature in affected patients

(A) Pie charts depicting the percentage of VH family usage in each labeled population. (B) Heavy chain germline VH and JH usage (top) and light chain germline VL and JL (bottom) usage within the indicated B cell populations. Circle size corresponds to frequency. Venn diagrams (right) depict the absolute number of unique V-J pairs by population, p<0.0001, 2x2 contingency analysis comparing naïve groups. (C/D) Heavy chain CDR3 amino acid length (C) and CDR3 net charge (D) did not statistically differ (ANOVA, p>0.05). (E) Pie charts demonstrating similar ratios of kappa and lambda light chain usage by population.

Figure 6:. High CASPR2 avidity and otherwise low self autoreactivity facilitate clonal escape

(A) Heat maps depicting binding by CASPR2-reactive Fab segments, -IgG or -IgM, red = positive binding. * = p<0.05 for fab and IgG frequency distributions. Populations labeled in green depict healthy control, populations labeled in red depict CASPR2-Ab E throughout the figure. (B) CASPR2-IgM mAb end point dilution in a live cell based assay. All mAbs expressed as class IgM to prevent class confound * = p<0.05, ** = p<0.005 (C) Quantification of number of peptides enriched in phage immunoprecipitation, ****=p<0.0001, Wilcoxon unpaired. (D) Summary cartoon model.