Fig 2: Patients with LGI1-/CASPR2-AIE show trans-compartment loss of innate-like MAIT cells and shifts in the NK cell subpopulations. (A) UMAP showing the subclustering of the mixed T-cell cluster (cluster 10, Fig. 3). The expression levels of marker genes for each cell type shown are visualized in Supplementary Fig. 8A. (B) Heat map comparing selected cell-type abundances (of the subclustering shown in Fig. 4A) between AIE patients, IIH and MS controls across CSF and PBMC compartments. Colours indicate an increase (red) or decrease (blue), and asterisks indicate significance (Supplementary Fig. 8B and C). (C–E) Flow cytometry validation: cell proportions of dnTc (CD3+CD4−CD8−) (C), CD56dim NK cells (CD3−CD56dim) (D) CD56bright NK cells (E), in the CSF (cohort 2). (F) Cell proportion of MAIT cells in PBMCs in another independent flow cytometry cohort (cohort 3). Frequency respective to parent gate. Gating schemes are shown in Supplementary Figs 2 (cohort 2) and 9 (cohort 3). AIE = autoimmune encephalitis; FND = functional neurological disorders; IIH = idiopathic intracranial hypertension; MAIT = mucosal-associated invariant T cell; MS = multiple sclerosis; NK = natural killer cell; PBMC = peripheral blood mononuclear cell.

Fig 3: T-cell repertoires in CSF of LGI1-/CASPR2-AIE patients show differential changes preferentially in the CD4 TCM clusters, clonal expansion of activated CD4 TCM and CD8 TCM clusters, and blood–CSF-spanning T-cell clones. (A) UMAP plot based on the T-cell subclustering (of all T-cell clusters from Fig. 1B) of all PBMCs and CSF cells. (B and C) Unbiased, cluster-free analysis of differential T-cell abundance was analysed in a pairwise fashion using the DAseq tool and visualized in shades of red for increases and shades of blue for decreases; comparison of LGI1 and IIH CSF T cells (B) and CASPR2 and IIH CSF T cells (C). (D) T-cell clone sizes in CSF from AIE patients projected onto the UMAP plot. (E and F) Visual representation of T-cell receptor clones in CSF (blue) and PBMC (red) of two representative patients with LGI1-AIE (E) and CASPR2-AIE (F). In the network, each dot indicates one clone (identical VDJ genes and CDR3 regions). The area is proportional to the clone size. The pie charts indicate the percentage of CSF cells and PBMCs within each clone. AIE = autoimmune encephalitis; IIH = idiopathic intracranial hypertension; MAIT = mucosal-associated invariant T cell; NK = natural killer cell; PBMC = peripheral blood mononuclear cell; TCM = central memory T cell; UMAP = uniform manifold approximation and projection; VDJ = variable–diversity–joining.

Fig 4: B-lineage cells in CSF are preferentially plasmablasts, clonally expanded, and express IgG1/2 and IgG4 heavy chains in LGI1-/CASPR2-AIE. Single-cell transcriptomes of all B-cell clusters (from Fig. 1B) in CSF and PBMCs from LGI1, CASPR2, IIH and MS patients were subclustered and analysed. (A) UMAP plot with B-lineage subclusters of CSF cells from AIE patients. Cluster 4 contains only PBMCs and is therefore not shown here. Expressions of selected marker genes are shown in Supplementary Fig. 3B and C. Cluster 8 is composed of ‘contaminating’ T cells. (B) UMAP plot illustrating the distribution of B-cell clone sizes in CSF from AIE patients. (C) UMAP depicting the immunoglobulin subtype transcribed by CSF cells from AIE patients. (D) Comparison of the frequency of transcribed immunoglobulin heavy chains (IGHG1 or IGHG2 versus IGHG4) in antibody-secreting B lymphocytes (ASCs) from CSF in all AIE patients. Only samples with >30 cells were considered. Significance was tested with a Mann–Whitney U-test. (E and F) Visual representation of B-cell receptor (BCR) clones (and clonotypes connected by lines) in the CSF (blue) and PBMCs (red) of two representative patients with LGI1-AIE (E) and CASPR2-AIE (F). Each dot indicates the amount of identical BCRs belonging to one clone. The area of each dot is proportional to the clone size. Clonotypes (connected by lines) were defined as identical VDJ gene segments, identical CDR3 length and a CDR3 nucleotide sequence with two or fewer mismatches. Asterisks mark clones that were verified experimentally as autoantigen specific. (G) Circos plot highlighting BCR clones composed of mixtures of plasmablasts, plasma cells and memory B cells. The line thickness indicates the number of cells. Clones consisting of only one cell type are marked in grey. AIE = autoimmune encephalitis; PBMC = peripheral blood mononuclear cell; UMAP = uniform manifold approximation and projection; VDJ = variable–diversity–joining.

Fig 5: Single-cell transcriptomics and flow cytometry identified expansion of plasma cells as a hallmark of LGI1-/CASPR2-AIE. (A) Sankey diagram showing overlap between LGI1-AIE, CASPR2-AIE, non-inflammatory-disease controls (IIH in cohort 1, functional disorder in cohort 2 and healthy controls in cohort 3) across the four sample cohorts. IHC = immunohistochemistry analysis of formalin-fixed, paraffin-embedded (FFPE) autopsy brain tissue. (B) Uniform manifold approximation and projection (UMAP) plot depicting the cell-type clusters of CSF cells. The second cell cluster included only PBMCs and is therefore shown only in Supplementary Fig. 1D. (C–F) Comparison of the relative cell-type abundance between LGI1 and IIH (C), CASPR2 and IIH (D), LGI1 and MS (E) and CASPR2 and MS (F). (G) Flow cytometry validation: relative percentage of plasma cells (%CD3−CD19+CD138+) quantified as percentages of all lymphocytes in CSF cells of the second cohort. Statistical significance was determined by the Kruskal–Wallis test with Dunn’s post hoc test and adjusted with the Benjamini–Hochberg method. FACS = fluorescence-activated cell sorting; FND = functional neurological disorders; IIH = idiopathic intracranial hypertension; MAIT = mucosal-associated invariant T cell; MS = multiple sclerosis.