Single-cell multi-omics reveals compartmentalized heterogeneity, clonal dynamics, and B cell interplay of CD8+ T cell subsets in anti-NMDAR encephalitis

To the Editor: Anti-N-methyl-D-aspartate receptor encephalitis (NMDAR-E) is a severe autoimmune disorder characterized by pathogenic autoantibodies targeting the glutamate ionotropic receptor NMDA type subunit 1 (GluN1) of NMDA receptors, manifesting clinically as psychiatric disturbances, seizures, and cognitive deficit.[1] Although current research predominantly focuses on B cell-derived autoantibody mechanism,[2,3] the contributions of T cell immunity remain incompletely characterized. Emerging evidence implicates T cell involvement in NMDAR-E pathogenesis, including histopathological observations of T cell infiltrates in patient brain parenchyma[4] and hippocampal CD8+ T cell accumulation in murine models.[5] To systematically investigate CD8+ T cell heterogeneity and function, we performed integrated single-cell RNA sequencing (scRNA-seq) and T cell receptor sequencing (scTCR-seq) profiling, combined with flow cytometry and functional co-culture assays, using peripheral blood (PB) and cerebrospinal fluid (CSF) samples from NMDAR-E patients and controls [Supplementary Figure 1A, https://links.lww.com/CM9/C548]. All samples were collected at West China Hospital, Sichuan University (January 2021–January 2023) from patients fulfilling definitive NMDAR-E diagnostic criteria. The study protocol (West China Hospital Institutional Review Board No. ChiCTR1900027074) complied with the Declaration of Helsinki principles, with informed consent obtained from participants or legal guardians. Detailed methodologies are provided in Supplementary Materials, https://links.lww.com/CM9/C548. Through integrated analysis of 45,088 single-cell transcriptomes (including 18,421 CSF and 26,667 PB immune cells) and 14,536 CSF cells from idiopathic intracranial hypertension (IIH) patients (GSE138266 dataset), we identified 59,624 single cells and classified them into 8 immune cell clusters [Supplementary Figure 1B–E, Supplementary Table 1, https://links.lww.com/CM9/C548]. Focusing on T cell subsets, we analyzed a total of 47,925 T/Natural Killer (NK) cells derived from CSF (NMDAR-E: 13,279 cells; IIH: 11,687 cells) and PB (NMDAR-E: 17,380 cells; controls: 5579 cells). Through integrated single-cell transcriptomic profiling, we delineated T/NK cells into 10 functionally distinct subpopulations [Figure 1A, Supplementary Figure 1F, https://links.lww.com/CM9/C548].Figure 1: Transcriptomic and immunomic profiles of T cell subsets in PB and CSF of NMDAR-E patients. (A) t-SNE plot displays 10 distinct cell clusters, color-coded and annotated based on merged scRNA-seq data from PB (NMDAR-E, n = 4; HC, n = 2), and CSF (NMDAR-E, n = 3; IIH, n = 6) samples. (B) Volcano plots illustrate the comparison of T/NK cell subpopulations in CSF and PB between NMDAR-E patients and controls. Plots were generated using a beta-binomial regression model, with the x-axis showing log10 fold change and the y-axis showing –log10 P values. The horizontal line indicates the significance threshold. (C) Representative contour plots and summary scatter plots show the percentages of CD8+ Temra and CD8+ Tem cells in CSF. (D) Representative contour plots and summary scatter plots show the percentages of CD8+ Temra, CD8+ Tn, CD8+ Tcm, and CD8+ Tem cells in PB. Two-tailed unpaired Student’s t-test (C), one-way ANOVA test (D), * P <0.05,† P <0.01, ‡ P <0.001. AE: Anti-NMDAR encephalitis; CSF: Cerebrospinal fluid; CD45RA: CD45 isoform RA; CCR7: C-C chemokine receptor type 7; gdT: Gamma delta T; HC: Healthy control; IIH: Idiopathic intracranial hypertension; MAIT: Mucosal-associated invariant T; NMDAR-E: Anti-N-methyl-D-aspartate receptor encephalitis; NK: Natural killer; ns: Not significant; PB: Peripheral blood; scRNA-seq: Single-cell RNA sequencing; Temra: Terminal effector memory RA+ T; Tem: Effector memory T; Tn: Naïve T; Tm: Memory T; Tcm: Central memory T; Tna: Non-activated T; Ta: Activated T; Treg: Regulatory T; t-SNE: T-distributed stochastic neighbor embedding.CD4+ T cells were stratified into naïve (CD4+ Tn) and memory (CD4+ Tm) subsets based on differential expression of lymphoid homing receptors (CCR7, SELL) and the transcription factor LEF1. CD8+ T cells were divided into nonactivated (CD8+ Tna, high in CCR7, SELL) and activated subsets (CD8+ Ta_1, CD8+ Ta_2, and CD8+ Ta_3). CD8+ Ta_1, expressing effector memory genes (GZMK, GZMH, and PRF1), was identified as activated effector memory T cells (activated Tem). CD8+ Ta_2, with high expression of cytotoxic genes (GZMB, GZMH, PRF1, and GNLY) and terminal differentiation marker TOX, was characterized as terminally differentiated effector T cells (terminal Teff). CD8+ Ta_3, expressing GZMK and NKG7 but lacking GZMB, PRF1, and CCR7, was classified as resting effector memory T cells (resting Tem). In addition, other immune subsets were identified, including γδ T cells (TRDV2 and TRGV9), regulatory T cells (Tregs; FOXP3 and CTLA4), mucosal-associated invariant T cells (MAITs; SLC4A10), and natural killer cells (NKs; GNLY and NKG7). scRNA-seq analysis indicated a higher proportion of CD8+ Ta_1 (activated Tem) and CD8+ Ta_2 (terminal Teff) subsets in the PB and CSF of NMDAR-E patients compared to controls [Figure 1B, Supplementary Figure 1G, https://links.lww.com/CM9/C548]. Flow cytometry using CD45 isoform RA (CD45RA) and C-C chemokine receptor type 7 (CCR7) classified CD8+ T cells into subpopulations [Supplementary Figure 2A, https://links.lww.com/CM9/C548]. However, confirming significant increases in CD8+ terminal effector memory RA+ T cells (Temra; corresponding to CD8+ Ta_2 in scRNA-seq) in CSF, and elevated CD8+ Temra (Ta_2) and effector memory T cells (Tem; Ta_1 or Ta_3) in PB during acute and relapse phases [Figure 1C and D]. These findings suggest compartment-specific expansion of activated CD8+ T cell subsets in NMDAR-E. Differential expression gene (DEG) analysis revealed that CD8+ Ta cells in the PB of NMDAR-E patients specifically upregulated cytotoxic effector genes (GZMH and NKG7), suggesting enhanced cytotoxic potential in peripheral immune responses [Supplementary Figure 3A, B, https://links.lww.com/CM9/C548]. In contrast, CD8+ T cells in the CSF exhibited a distinct transcriptional profile: beyond activating cytotoxicity-related genes (GZMK, PRF1, and GZMA), they simultaneously upregulated T helper type 1 cell (Th1) polarization (TBX21 and CXCR3), interferon signaling (IFNGR1 and IRF7), TCR signaling (DUSP2 and GIMAP1), and immune homeostasis regulators (CTLA4, LAG3, CD69, and CXCR6) [Supplementary Figure 3C, D, https://links.lww.com/CM9/C548]. This dual functional signature implies that CSF CD8+ T cells not only mediate pathogen/aberrant cell clearance but also maintain central nervous system (CNS) immune homeostasis via tissue residency (CD69 and CXCR6) and coinhibitory checkpoints (CTLA4 and LAG3). ScTCR-seq revealed an elevated proportion of clonotypic T cells in the PB of NMDAR-E patients compared with controls, with further clonal expansion observed in the CSF [Supplementary Figure 4A, B, https://links.lww.com/CM9/C548]. High-clonality CD8+ T cells in the PB were predominantly CD8+ Ta_2, while in the CSF, they were primarily CD8+ Ta_1 [Supplementary Figures 4C–E, https://links.lww.com/CM9/C548]. Clonally expanded T cells in the CSF upregulated genes associated with neuroinflammatory modulation (CCL3L1, CCL4, and CCL5) and non-apoptotic injury (GZMK and GZMA), while their PB counterparts overexpressed genes mediating apoptotic cytotoxicity (GZMB and NKG7) and effector cell recruitment (CCL5) [Supplementary Figure 4F, G, https://links.lww.com/CM9/C548]. These findings demonstrate spatial functional divergence between CSF- and PB-derived CD8+ T cell clones, characterized by compartmentalized chemokine and cytotoxic specialization within central and peripheral immune responses. Critically, shared TCR clonotypes between CSF and PB subsets [Supplementary Figures 4H, I, https://links.lww.com/CM9/C548] implied a common antigen-driven origin, with clonally expanded T cells likely migrating from the periphery to the CNS while adapting compartment-specific effector programs. To investigate potential CD8+ T cell-mediated immunoregulation in NMDAR-E, we established an ex vivo co-culture system using CD8+ T cells and autologous CD19+ B cells isolated from peripheral blood mononuclear cells (PBMCs) of patients and controls. Although initial apoptosis rates of CD8+ T cells and B cells showed no intergroup differences at 24 hours, by day 7, patient-derived CD8+ T cells induced significantly higher apoptosis in total B cells and key subsets–including antibody-secreting cells (ASCs), memory B cells (Bm), CD27– B cells, and double-negative (DN) B cells–compared to control cultures [Supplementary Figure 5, https://links.lww.com/CM9/C548]. These observations suggest a potential role for CD8+ T cells in modulating B cell activity during NMDAR-E progression. Emerging evidence suggests that immunosuppressive CD8+ T cell subsets may regulate autoreactive immunity by targeting pathogenic cells, as exemplified by Li et al[6] in autoimmune disorders. Building on this paradigm, we identified a distinct killer-cell immunoglobulin-like receptor (KIR)+CD8+ T cell subset in NMDAR-E patients, characterized by high expression of inhibitory KIRs (KIR3DL1/KIR2DL3) and an activated state across both CSF and PB compartments [Supplementary Figure 6A–C, https://links.lww.com/CM9/C548]. Flow cytometry confirmed elevated KIR+ CD8+ T cells in PB of severe NMDAR-E patients, with further expansion of KIR+CD8+ Temra during relapses [Supplementary Figure 6D–F, https://links.lww.com/CM9/C548]. scTCR-seq revealed clonal expansion of KIR+CD8+ T cells in patients [Supplementary Figure 7A–G, https://links.lww.com/CM9/C548]. Peripheral clones upregulated NK-associated receptors (KLRF1 and KLRB1) and cytotoxic effectors (GZMB, GNLY, and PRF1), aligning with a potent cytolytic phenotype. CSF clones enriched naïve/memory-related genes, suggesting tissue adaptation while retaining cytotoxic potential. Cellular interaction analysis demonstrated enhanced connectivity between KIR+CD8+ T cells and naïve/memory B cells (Bn/Bm) in patients [Supplementary Figure 7H, https://links.lww.com/CM9/C548]. Receptor-ligand profiling highlighted KIR-MHC-I axis activation, with KIR+CD8+ T cells engaging B cell-expressed MHC-I molecules (HLA-A, HLA-C, and HLA-E) via inhibitory KIRs (KIR3DL1 and KIR2DL1) and activating NK receptors (KLRC2 and KLRK1) [Supplementary Figure 7I, https://links.lww.com/CM9/C548]. Collectively, KIR+CD8+ T cells in NMDAR-E exhibit dual functionality: clonally expanded PB subsets mediate B cell clearance via GZMB/PRF1-dependent cytotoxicity, whereas CSF-resident clones may balance immunoregulation through KIR-MHC-I interactions, potentially preventing neuronal damage while surveilling autoreactive B cells. In summary, we integrated scRNA-seq, scTCR-seq, and flow cytometry to characterize T cell heterogeneity and clonal dynamics in NMDAR-E. Compartment-specific expansions of activated CD8+ T cells occurred in both CSF and PB, but with distinct clonal TCR repertoires and upregulated genes. Ex vivo, CD8+ T cells induced higher B cell apoptosis versus controls, with KIR+ subsets showing enhanced B cell engagement. These findings link CD8+ T cell activation/clonal expansion (especially KIR+) to disease, potentially via B cell modulation. Further studies should define antigenic drivers and CD8+ T-B cell interaction mechanisms. Acknowledgements Thanks to Dr. Yuan Liang for his help in our data analysis. Funding This work was supported by grants from the Natural Science Foundation of Chongqing (No. 2024NSCQ-MSX0894), the China Postdoctoral Science Foundation (No. 2024M763885), the Shapingba district technology innovation project (No.2024170), the National Science Foundation of China (Nos. 81571272 and 82071459), and 1.3.5 project for disciplines of excellence from West China Hospital of Sichuan University (Nos. ZYGD23032 and ZYGD22003). Conflicts of interest None. Data sharing The raw sequencing data for scRNA/TCR-seq are deposited in the Genome Sequence Archive (GSA-Human: HRA008307) at China National Center for Bioinformation (https://ngdc.cncb.ac.cn/gsa-human).