Abstract

   We presented a patient with refractory anti-GABA[A]-R encephalitis, and
   constructed libraries for single-cell sequencing from the patient’s
   peripheral blood mononuclear cells (PBMCs), cerebrospinal fluid cells,
   as well as four healthy volunteer’s PBMCs. A distinct group of
   monoclonal CD8^+ T cells and an abnormal JAK-STAT signaling pathway was
   implicated in the disease. The cross-reactive protein LIM-domain-only
   protein 5 (LMO5) identified in the patient’s thymoma, prompted the
   activation of the specific CD8^+ T cells. Furthermore, in vitro
   analysis revealed the involvement of the JAK-STAT pathway in
   LMO5-induced CD8^+T cell activation, a process effectively suppressed
   by tofacitinib, which improved the patient’s clinical outcome.

   Subject terms: Clinical genetics, Molecular neuroscience

Introduction

   Encephalitis linked to antibodies against the neuronal g-aminobutyric
   acid type A (GABA[A]) receptor (GABA[A]-R) constitutes a rare,
   distinctive, and severe form of autoimmune encephalitis [[58]1, [59]2].
   The etiology and pathomechanisms of anti-GABA[A]-R encephalitis remain
   elusive. While immune therapy shows favorable responses in some cases,
   a subset of patients experiences poor outcomes [[60]2], indicating a
   pressing need for personalized medicine in the management of
   anti-GABA[A]-R encephalitis.

Methods and materials

Statement of ethics

   The protocol of study was approved by the Ethics Committee of the Third
   Affiliated Hospital of Sun Yat-Sen University ([2021]02-014-01) and
   carried out in accordance with the principles described in Declaration
   of Helsinki of 1975.

Index patient and clinical history

   The patient, a 42-year-old male, initially presented with psychiatric
   disorders followed by loss of consciousness, fever, and seizures,
   persisting for a month despite treatment with steroid and multiple
   antiepileptic drugs (AEDs). Upon admission to our hospital, he was in a
   comatose state with a tracheotomy, exhibiting a Glasgow Coma Scale
   (GCS) score of 8 (E4V2M2), a modified Rankin Scale (mRS) of 5, and
   Clinical Assessment Scale in Autoimmune Encephalitis (CASE) score of 20
   (Fig. [61]1A, Supplemental table [62]1, the scores of the patient were
   assessed while maintaining blinding procedures). Pseudomonas aeruginosa
   pneumonia was identified through sputum culture and next-generation
   sequencing (NGS). Cerebrospinal fluid (CSF) analysis indicated elevated
   white blood cells (WBC) at 5/μL and total protein at 371 mg/dL. In the
   CSF, viral, fungal, and bacterial polymerase chain reactions and
   cultures were negative, and no malignant cells were found. CSF was also
   collected for PACEseq metagenomic next-generation sequencing (mNGS)
   detection, while the result was negative.

Fig. 1. Clinical course, treatment history, and scRNA-seq analysis in PBMCs
derived from the index patient.

   [63]Fig. 1
   [64]Open in a new tab

   A Clinical presentation and progression prior to and after intervention
   using the JAK inhibitor tofacitinib. IVIG, intravenous immunoglobulin;
   AED, antiepileptic drug including Pirenpanet, Valproate, Midazolam,
   Clonazepam, topiramate, Levetiracetam. B Brain magnetic resonance
   imaging (MRI) of the patient prior to or following tofacitinib
   treatment. Elevated T2/FLAIR signal intensities in extensive
   cortical-subcortical-basal ganglia (a) with leptomeningeal enhancement
   in the gadolinium-enhanced T1-weighted image (b, red arrows) before
   treatment. After tofacitinib treatment, the images have clearly
   improved (c, d). C PET-CT images from the patient. PET-CT images
   illustrating abnormal metabolism in the mass on the left side of the
   upper mediastinum, as indicated by the arrow. D H&E staining of thymoma
   and immunohistochemistry of CK-20, TdT-20, CD1a, CD3, CD20, and Ki67
   within thymoma. E The t-SNE plot projection including 6338 cells from
   anti-GABA[A] receptor encephalitis patient peripheral blood mononuclear
   cells (PBMCs) and 16,551 cells obtained from healthy volunteer (HV)
   PBMCs, shaded based on their cell type. F Violin plots depicting the
   normalized expression levels of cluster-defining genes across
   individual clusters. The columns represent selected marker genes, and
   the rows represent clusters with the same color as identified in E. G
   t-SNE plot illustrating the number of differentially expressed genes
   (DEGs) across patient and HV cells within each cluster. DEGs were
   identified using a Wilcoxon test, with a threshold of |log[2] fold
   change| >0.25; adjusted p-value <10^−12. H GO enrichment analyses of
   the DEGs. Bar plot showing the upregulated GO term in Effector CD8 T
   cells, and notable genes are labeled in red. The p-value was derived
   using a hypergeometric test. I Volcano plot depicting DEGs that are up-
   or down-regulated in Effector CD8 T cells, with relevant DEGs labeled.
   J Effector CD8 T cells (n = 2890 cells) were projected onto a t-SNE
   plot colored by origin (left). Expression of selected genes was
   projected onto the t-SNE plot (top), and the violin plots show the
   expression levels (bottom).

   Positive serum and CSF anti-GABA[A] antibody (titer, serum 1:30, CSF
   1:30) were detected using a cell-based assay with HEK293 cells
   transfected with plasmids encoding GABA[A]. Other autoimmune
   encephalitis antibodies and serological tests for anti-myelin
   oligodendrocyte (MOG), glial fibrillary acidic protein (GFAP), and
   aquaporin 4 (AQP4) were negative (Supplemental Fig. [65]2A, B).
   Coagulation analysis revealed abnormal activated partial thromboplastin
   time (APTT, 142.3 seconds) and deceased Factor XII viability (0.1%,
   normal range 70–150%). Whole genome sequencing of DNA in the patient’s
   white blood cells detected an F12 mutation (Supplemental Fig. [66]4;
   Supplemental table [67]6).

   Brain magnetic resonance imaging (MRI) revealed increased T2/FLAIR
   signal with extensive cortical-subcortical-basal ganglia involvement
   and leptomeningeal enhancement (Fig. [68]1A(a, b)).
   Electroencephalogram (EEG) indicated the presence of rhythmic periodic
   sharp-slow waves in the bilateral temporo-frontal areas (Supplemental
   Fig. [69]1A). Positron Emission Tomography with Computed Tomography
   (PET/CT) scans indicated increased fluorodeoxy-D-glucose (FDG) uptake
   in the left-sided mediastinal mass, revealing a mediastinal tumor (Fig.
   [70]1C).

   Due to the patient’s severe condition and coagulation abnormalities,
   surgical intervention for the mediastinal tumor was deemed intolerable.
   Treatment involved prednisolone, intravenous immunoglobulin (IVIG), and
   anti-epileptic drugs (AEDs). However, the patient exhibited limited
   response to immune therapy and subsequently developed gastrointestinal
   bleeding (Fig. [71]1A). As gastrointestinal bleeding and severe
   pneumonia, the steroid treatment also was interrupted.

   Considering the patient’s low peripheral blood B cell levels, with a
   CD3-CD19 + B cell to lymphocyte ratio of 3.53%, which falls below the
   normal range of 5–18%, coupled with the presence of severe pneumonia,
   it was determined that rituximab would not be administered.

   Given the potential of scRNA-seq to enhance the understanding of
   pathological processes in human diseases, scRNA-seq was performed
   during the subsequent treatment process.

Single-cell isolation and library development

   Libraries for scRNA-seq from the patient peripheral blood mononuclear
   cells (PBMCs) and CSF cells, as well as one healthy volunteer
   (HV)-PBMC, were constructed using a Single Cell 5′ Library and Gel Bead
   kit (10× Genomics, cat# 1000014) and Chromium Single Cell V(D)J
   Enrichment Kits, Human (10× Genomics, cat# 1000005, 1000016) using the
   manufacturer’s instructions.

   The remaining three healthy volunteer (HV)-PBMC cell suspensions were
   resuspended in a Labeling Buffer with PBS containing 1% BSA. Human
   TruStain FcX™ Antibody (Biolegend, cat# 422301) was added and
   incubated, and various hashtag antibodies were added to each sample.
   Following incubation, the samples were washed, resuspended, and mixed
   using a Single Cell 3′ Library and Gel Bead kit (10× Genomics, cat#
   1000092) and Chromium Single Cell B Chip kit (10× Genomics, cat#
   1000074). The cell suspension was loaded onto a Chromium single-cell
   controller (10× Genomics) to construct single-cell gel beads within the
   emulsion (GEMs). Isolated cells were lysed, and the released RNA was
   barcoded via reverse transcription within individual GEMs. Reverse
   transcription was conducted using a C1000TM Touch Thermal Cycler (Bio
   Rad) at 53 °C for 45 min, followed by 85 °C for 5 min, and a hold at
   4 °C. Complementary DNA was produced and amplified, and cDNA was
   separated for 3′ gene expression library and cell surface protein
   library development using SPRIselect.

   The libraries were sequenced utilizing an Illumina Novaseq6000
   sequencer and a paired-end 150-bp (PE150) approach.

ScRNA-seq analysis

   Single-cell RNAseq data were treated using CellRanger (v.5.0.0) with
   the GRCh38 human reference. Downstream analyses were performed using
   filtered gene expression matrices in R (v.4.0.4) utilizing the Seurat
   package (v.4.0.5). Briefly, after removing doublet or negative cells
   with the HTODemux function, for quality control, cells with (1) gene
   count <500; (2) UMI count <800; (3) mitochondrial gene proportion >20%,
   were removed as low-quality cells. For identification and comparison of
   shared cell types, we used the SCTransform and IntegrateData standard
   workflows for normalization, variable gene selection, integration,
   dimensionality reduction, and clustering, as outlined online
   ([72]https://satijalab.org/seurat/articles/integration_introduction).
   For clustering, we employed FindNeighbours and FindClusters functions
   on 30 PCA components using a resolution of 0.7, and t-Distributed
   Stochastic Neighbor Embedding (t-SNE) dimensionality reduction was
   conducted using the first 30 PCA components to acquire a 2-dimensional
   representation of the cellular states. To characterize the clusters,
   the FindAllMarkers function was employed to identify cluster-specific
   gene expression, based on the clusters classified and annotated.
   Differential gene expression was examined using the FindMarkers
   function in Seurat alongside the Wilcoxon Rank Sum test with default
   parameters, and the Benjamini-Hochberg method was utilized to assess
   the false discovery rate (FDR). DEGs were filtered with a minimum
   log2(fold change) of 0.25 and a maximum FDR value of 0.05, and
   enrichment analysis for the functions of the DEGs was performed using
   the clusterProfiler package.

TCR V(D)J sequencing and analysis

   Single-cell TCR V(D)J data were handled using CellRanger (v.5.0.0) with
   GRCh38 VDJ for demultiplexing, gene quantification, and TCR clonotype
   assignment. A TCR diversity metric containing clonotype frequency and
   barcode information was generated. Using the barcode information, T
   cells with prevalent TCR clonotypes were projected onto a t-SNE plot.

Whole genome sequencing (WGS)

   WGS was performed on white blood cells from the patient, and was
   performed with 50× coverage using BGI MGISEQ-T7 and standard protocols.
   Sequence reads were mapped to the human reference genome Hg38. SNPs and
   InDels were identified using GATK (Genome Analysis TK 4.2.3.0).

LMO5-TCR docking analysis

   ColabFold, ZDOCK, PyMol, and other tools were utilized to simulate
   docking between LIM-domain-only protein 5 (LMO5) and the top 10 TCR
   clonotypes to assess if an interaction between LMO5 and TCR was
   possible. Specifically, ColabFold was used to calculate the protein
   structures of TCR by combining Mmseqs2 with AlphaFold2. Protein docking
   was predicted with ZDOCK, and the outcome was visualized using PyMol.

Dot blot

   A nitrocellulose membrane premoistened with tris-buffered saline (TBS)
   was utilized in a Bio-Dot apparatus (706,545, Bio-Rad, Richmond, CA,
   USA), and 100 μL of recombinant LMO5 protein (0.18 μg/mL; Alpalife,
   Shenzhen, Guangdong, China) was added. After the peptide solution was
   completely filtered through the membrane, the wells were blocked using
   100 μL of 2% bovine serum albumin (BSA; 155,681, Jackson
   ImmunoResearch, West Grove, PA, USA) in TBS-0.05% Tween-20 (TBST) at
   room temperature until draining of the blocking solution. Thereafter,
   100 μL of serum (1:1000 in TBST), or100 μL of anti-LMO5 antibody
   (serving as a positive control, 1:1000 dilution in TBST, ab178695,
   abcam) was added onto the membrane for 1 h at room temperature. The
   wells were washed three times using 400 μL of TBST. Subsequently, the
   wells were incubated with 200 μL of HRP-conjugated anti-human IgG
   secondary antibody (1:10000; bs-0297G-HRP, Bioss, Shanghai, China),
   anti-rabbit IgG secondary antibody (1:2000; 111-035-003, Jackson
   ImmunoResearch) for 1 h at room temperature. The membrane was removed
   from the apparatus and rinsed with TBST three times (10 min each). Dots
   were assessed using Immobilon Western chemiluminescent HRP substrate
   (WBKLS0500, Merck, Darmstadt, German), and images were obtained with an
   ImageQuant LAS 4000 (GE Healthcare Life Sciences, Issaquah, WA, USA).

Hematoxylin and eosin (H&E) staining

   Paraffin-embedded tissues were sectioned into 3 μm thick slices. H&E
   staining was conducted based on the standard protocol. Briefly,
   sections underwent the following steps: dewaxing (xylene, 2 × 2 min),
   rehydration (90% ethanol for 10 min, 80% ethanol for 10 min, and 75%
   ethanol for 10 min), hematoxylin staining (5 min, G1004, Servicebio,
   Wuhan, Hubei, China), differentiation (2 s, G1039, Servicebio), bluing
   (2 seconds, G1040, Servicebio), dehydration (85% ethanol for 5 min and
   95% ethanol for 5 min), eosin staining (5 min, G1001, Servicebio),
   dehydration (anhydrous ethanol for 3 × 5 min), and clearing (xylene, 2
   × 5 min). Sections were then sealed using neutral gum (10,004,160,
   Sinopharm, Beijing, China) and photographed using an Olympus IX-71
   inverted fluorescent microscope (Olympus, Tokyo, Japan).

Immunohistochemistry

   Paraffin-embedded sections (3 µm thick) were dewaxed using xylene and
   heated within 0.01 M sodium citrate buffer (pH = 6.0) in a microwave at
   60 °C for 20 min for antigen retrieval. Tissue sections were blocked in
   0.3% H[2]O[2] for 10 min to stop endogenous peroxidase activity,
   followed by incubation with a primary antibody against LMO5 (1:100;
   10892-2-AP, proteintech), CK-20 (1:100; AQ20029, aq-meditech), TdT-20
   (ready-to-use; RMA-1044, MXB Biotechnologies), CD1a (1:50; MAB-0336,
   MXB Biotechnologies), CD3 (1:100; AQ20001, aq-meditech), CD20
   (ready-to-use; Kit-0001, MXB Biotechnologies), or Ki67 (1:100;
   MAB-0672, MXB Biotechnologies). The target was assessed using a DAB
   staining kit (GK600710, Gene Tech, Shanghai, China).

Western Blot

   The diluted serum (1:6) or CSF was combined with 4× SDS sample loading
   buffer and denatured at 95 °C for 10 min. The samples were separated
   using sodium dodecyl sulfate-polyacrylamide gel electrophoresis
   (SDS-PAGE) and transferred onto a polyvinylidene difluoride (PVDF)
   membrane. The PVDF membrane was blocked using 5% nonfat dry milk in
   TBST for 2 h at room temperature, then incubated with anti-LMO5
   antibody (1:1000) overnight at 4 °C, and incubated with anti-rabbit IgG
   secondary antibody (1:5000; BA1054, BOSTER) for 1 h at room
   temperature. The bands were visualized using BeyoECL plus (P0018,
   Beyotime Biotechnology), and images were acquired utilizing Tanon-4100
   (Tanon, China). ImageJ software was employed to quantify the band
   intensity.

Immunoprecipitation

   Immunoprecipitation was performed using a commercialized
   immunoprecipitation kit (IEMed-K206, IEMed) according to the
   manufacturer’s instructions. Briefly, 50 mL AB binding buffer was added
   into a microcentrifuge tube containing 10 mL Protein A/G beads. The
   tube was placed on a magnetic stand for 1 min. The supernate was
   discarded and the beads were incubated with 100 mL of AB binding buffer
   and 2 mg of human IgG on a rotating platform for 30 min at room
   temperature. Then, the beads were collected using the magnetic stand
   and incubated with denatured serum or CSF on a rotating platform for
   1 h at room temperature. Following washing with IP washing buffer,
   proteins were eluted from the beads with 100 mL of WB elution buffer.

ELISA test

   96-well microplates were coated with anti-LMO5 antibody (10 mg/mL;
   ab178695, abcam) for 16 h at 4 °C. The wells were rinsed with PBS (wash
   buffer) three times and incubated with 100 mL LMO5 recombinant protein
   (at concentrations of 600 ng/mL, 300 ng/mL, 150 ng/mL, 75 ng/mL,
   37.5 ng/mL, 18.8 ng/mL, 9.4 ng/mL), CSF, or serum at 37 °C for 1 h.
   After washing with 200 mL of wash buffer three times, the plates were
   sequentially incubated with an anti-LMO5 antibody (A7549, ABclonal) and
   horseradish peroxidase (HRP)-conjugated secondary antibody (AS014,
   ABclonal) at 37 °C for 1 h. A sample of 100 mL TMB substrate (AR1104,
   Boster) was added to each well, and the plate was incubated in the dark
   at 37 °C for 15 min. Subsequently, 50 mL of stop solution (AR1105,
   Boster) was employed to stop the reaction. Optical densities (ODs) were
   measured at 450 nm using a Victor X5 Series Multilabel Plate Reader
   (PerkinElmer, USA).

Lymphocyte transformation test (LTT) and flow cytometry examination

   Isolated PBMCs were resuspended in PBS and incubated with 5 mM CFSE
   ([73]C34554, Invitrogen) for 20 min at room temperature. Complete
   medium containing RPMI-1640 medium, fetal bovine serum (FBS, 10%),
   penicillin (100 U/mL), and streptomycin (100 mg/mL) was added to the
   reaction for 5 min at room temperature to remove the free dye. An
   aliquot of 2.5 × 10^5 PBMCs was cultivated in a 96-well plate in 200 mL
   of complete medium with or without 50 ng/mL of LMO5 or 40 nM
   tofacitinib. Following incubation at 37 °C for 6 days without changing
   the medium, the PBMCs were stained using 7-AAD Viability Staining
   Solution (420,403, BioLegend) for 10 min at room temperature and then
   incubated with cell-surface antibodies for 30 min on ice. The following
   monoclonal antibodies were used: anti-CD3-PE-Cy7 (300,420, BioLegend),
   anti-CD8-Alexa Fluor® 700 (561,453, BD Biosciences). Cells were
   identified using an LSRfortessa flow cytometer (BD Biosciences) and
   data were assessed using FlowJo software (FlowJo).

Immunofluorescence assay

   A sample of 2.5 × 10^5 isolated PBMCs was grown in a 96-well plate in
   200 mL of the complete medium combined with or without 50 ng/mL of LMO5
   or 40 nM tofacitinib. Following incubation at 37 °C for 4 days without
   recycling the medium, the PBMCs were obtained and resuspended in 500 mL
   of PBS. An aliquot of 2 mL of cell suspension was pipetted onto a slide
   and allowed to dry. The cells were permeabilized and blocked in PBS
   containing 5% normal goat serum (SL038, Solarbio) and 0.3% Triton X-100
   (V900502, Vetec) for 1 h at room temperature. The cells were incubated
   with a primary antibody against CD8 (1:100; ab237709, Abcam) and
   Phospho-Stat3 (1:50; 4113S, CST) overnight at 4 °C. Following 3 washes
   in PBS, the sections were incubated with Alexa Fluor® 488-conjugated
   anti-rabbit IgG (H + L) (1:500; 4412S, CST) and Alexa Fluor®
   594-conjugated anti-mouse IgG (H + L) (1:500; 8890S, CST), for 1 h at
   room temperature. Additionally, DAPI (1:2000; Sigma, 32,670) was
   employed to stain cell nuclei. Slides were mounted using Fluoro-Gel
   (17985-10, Electron Microscopy Sciences, Hatfield, PA, USA) and imaged
   using an Olympus IX71 fluorescence microscope (Olympus).

Statistical analysis

   The Shapiro-Wilk test was utilized to assess the normality of the data.
   If the data exhibited normal distribution, the means of two groups were
   compared using Student’s test; otherwise, the Mann-Whitney U test was
   employed. For the comparison of three groups, one way analysis of
   variance (ANOVA) with post hoc Tukey’s tests were conducted. A p-value
   or adjusted p-value below 0.05 was deemed to indicate statistical
   significance.

Results

Clinical outcomes and responses to tofacitinib

   Our scRNA-seq analysis suggested that the JAK-STAT pathway could be a
   potential therapeutic target for the index patient. With informed
   consent from his family, the patient was initiated on 5 mg/day of
   tofacitinib, a JAK1 and JAK3 inhibitor. After confirming the absence of
   toxicity, the dosage of tofacitinib was increased to 10 mg/day two
   weeks later. Two months into the tofacitinib treatment, the patient
   exhibited a favorable response (Fig. [74]1A): the CASE score decreased
   to 8 (Supplemental table [75]1), brain lesions reduced in MRI (Fig.
   [76]1B(c, d)), brain periodic spikes disappeared in EEG (Supplemental
   Fig. [77]1B), GABAAR-IgG titers decreased in serum and CSF
   (Supplemental Fig. [78]2C). Following family consent, the mediastinal
   tumor was surgically resected, histological examination confirmed the
   diagnosis of thymoma (B1 type) (Fig. [79]1D). Subsequently, the patient
   continued to received treatment with prednisolone and tofacitinib, and
   was followed up for 14 months, during which the clinical condition
   slowly improved (Supplemental video [80]1–[81]8).

scRNA-seq analysis uncovers cell-mediated immunity activation in
anti-GABA[A]-R encephalitis

   Following unsupervised clustering and t-distributed stochastic neighbor
   (t-SNE) plot analyses (Fig. [82]1E), cluster identities were determined
   by the expression of established markers (Fig. [83]1F, Supplemental
   table [84]2), indicating successful capture of major peripheral blood
   mononuclear cell (PBMC) subsets. Projection of the numbers of
   differentially expressed genes (DEGs) between patient- and four healthy
   volunteers (HVs)-derived cells revealed broad transcriptomic
   alterations observed in T cells and monocytes. Notably, the most
   prominent transcriptomic changes were observed in the effector CD8^+ T
   cell cluster (Fig. [85]1G).

   To understand the biological significance of transcriptional changes in
   the effector CD8^+ T cell cluster, we conducted pathway enrichment
   analysis with DEGs. This analysis revealed enrichment of pathways
   related to T cell activation and interferon-gamma signaling, driven, in
   part, by the upregulation of JAK2, JAK3, and STAT1 (Fig. [86]1H, I).
   Dysregulation of JAK-STAT signaling has been strongly associated with
   immune activation [[87]3]. Subclustering of the effector CD8^+ T cell
   resulted in the segregation of the patient and HV clusters,
   demonstrating distinct transcriptomic differences (Fig. [88]1J). This
   further validated that the expression of the JAK2, JAK3, and STAT1 was
   upregulated in the patient PBMC cluster. Additionally, JAK2, JAK3, and
   STAT1 were also upregulated in the effector CD8^+ T cell cluster of CSF
   (Supplemental Fig. [89]3A–F, Supplemental table [90]3). Consequently,
   the JAK-STAT signaling pathway emerged as a potential therapeutic
   target.

Single-cell T cell receptor (TCR) sequencing demonstrates unique TCR usage in
effector CD8^+ T cells of anti-GABA[A]-R encephalitis

   To investigate the clonality of T cells in PBMCs and CSF from the
   patient, we conducted single-cell TCRβ CDR3 sequencing using the
   Chromium Single Cell Immune Profiling platform. The results revealed
   that, in the patient, the top TCR clone “CDR3:
   CVVSEYSGGGADGLTF_CASSLSGTGYGELFF” (referred as clonotype A T cells)
   constituted more than 16% of PBMC T cells, approximately eight times
   that of the second-highest clone (Fig. [91]2A, Supplemental table
   [92]4). In contrast, in HV, the top clone comprised only 7% of PBMC T
   cells, less than twice that of the second-highest clone. This indicated
   the presence of a unique TCR clone in the patient’s PBMC T cells.
   Additionally, clonotype A T cells were also found in the patient’s CSF,
   suggesting that these cells might have crossed the blood-brain barrier
   to enter the CSF (Fig. [93]2B). When projected onto the t-SNE plots,
   clonotype A T cells in PBMC were predominantly identified as effector
   CD8^+ T cells (Fig. [94]2C). Similarly, clonotype A T cells in the CSF
   were also characterized as effector CD8^+ T cells (Fig. [95]2D).
   Comparative analysis revealed the upregulation of JAK2, JAK3, and STAT1
   in clonotype A T cells compared to other effector CD8^+ T cells in PBMC
   (Fig. [96]2E, G), and CSF (Fig. [97]2F, H). These findings underscored
   that clonotype A T cells were the primary cell type exhibiting the
   highest JAK-STAT signaling activation.

Fig. 2. Unique TCR utilization in effector CD8 T cells of an anti-GABA[A]
receptor encephalitis patient.

   [98]Fig. 2
   [99]Open in a new tab

   A Bar plot illustrating the distribution of the Top 1 clonotype (named
   clonotype A, red) of Patient PBMCs in Patient CSF, and HV PBMC. B
   Alluvial plot illustrating the top 20 clonotypes between patient PBMC
   and patient CSF. C The t-SNE plot projection of PBMCs from Patient and
   HV, with cells in clonotype A labeled in red and Effector CD8 T cells
   indicated with a dashed line. D The t-SNE plot projection of CSF cells
   from Patient and HV, with cells in clonotype A labeled in red, and
   Effector CD8 T cells indicated with a dashed line. E Volcano plot
   illustrating DEGs that are up- or down-regulated in Clonotype A of PBMC
   Effector CD8 T cells, with notable DEGs labeled. F Volcano plot
   illustrating DEGs that are up- or down-regulated in Clonotype A of CSF
   Effector CD8 T cells, with notable DEGs indicated. G GO enrichment
   analyses of DEGs in E Bar plot showing up-regulated GO terms in
   clonotype A cells and notable genes indicated in red. The p-value was
   derived using a hypergeometric test. H GO enrichment analyses of the
   DEGs in F Bar plot illustrating the up-regulated GO terms from
   clonotype A cells and notable genes labeled in red. The p-value was
   derived using a hypergeometric test.

Tofacitinib suppresses CD8^+ T cell proliferation caused by LMO5 derived from
the thymoma of an anti-GABA[A]-R encephalitis patient

   It has been reported that LIM-domain-only protein 5 (LMO5), acting as a
   tumor antigen associated with cell-cycle regulation and tumor growth,
   cross-reacts with autoantibodies to GABA[A-]R(5). To investigate this
   relationship, we conducted dot blot analysis to examine the interaction
   between autoantibodies in the patient’s serum and LMO5. As depicted in
   Fig. [100]3A, autoantibodies in the patient’s serum specifically
   recognized LOM5. Subsequently, we confirmed the presence of LOM5 in the
   resected thymoma through immunohistochemistry (Fig. [101]3B). Moreover,
   we observed higher levels of LMO5 in the patient’s serum and CSF
   respectively (Fig. [102]3C, D; Supplemental Fig. [103]5).

Fig. 3. Inhibition of LMO5-induced CD8 T cell proliferation through the use
of tofacitinib.

   [104]Fig. 3
   [105]Open in a new tab

   A Dot blot using the serum of the patient and HV to identify LMO5-IgG.
   Commercial LMO5 antibodies were used as the positive control; NC was
   the negative control. B Immunohistochemistry of LMO5 in the thymoma of
   anti-GABA[A] receptor encephalitis. C Western blot of LMO5 on serum
   from HVs and the patient and CSF from control and the patient (n = 3);
   immunoprecipitation results from a LMO5 immunoblot after human IgG
   immunoprecipitation (n = 3). D Concentration of LMO5 in serum and CSF
   via ELISA (n = 3). E The 3D molecular structures of docking result,
   computationally predicted by ZDOK, between LOM5 (green) and TRA (red),
   and TRB (blue) from clonotype A. F Immunofluorescence of phosphorylated
   Stat3 (p-Stat3, red) and CD8 (green) in PBMC treated with LMO5
   (50 ng/mL) or LMO5 (50 ng/mL) plus tofacitinib (40 nM). Nuclei were
   counterstained using DAPI (blue) (n = 3). G Statistical examination of
   relative mean fluorescence intensity of p-Stat3 in CD8-positive cells
   in F. **p < 0.01, one-way ANOVA using a post hoc Tukey’s test. H Flow
   cytometry detection of CD8+ T cell proliferation through a lymphocyte
   transformation test (LTT). CFSE-labeled PBMCs were cultured alongside
   LMO5 (50 ng/mL) or LMO5 (50 ng/mL) plus tofacitinib (40 nM) for
   six days (n = 3). I Quantification of CD8+ cell proliferation in H.
   *p < 0.05, **p < 0.01, one-way ANOVA using a post hoc Tukey’s test.

   To delve into the role of LMO5 in anti-GABA[A]-R encephalitis, we
   explored its potential binding to TCR in clonotype A T cells using ZDOK
   (Fig. [106]3E; binding score provided in Supplemental table [107]5).
   Given the activation of the JAK-STAT signaling pathway in clonotype A
   CD8+ T cells, we hypothesized that LMO5 might influence this pathway in
   CD8^+ T cells.

   Figure [108]3F, G illustrate that LMO5 enhanced the expression of
   phospho-Stat3 (p-Stat3), a key indicator of the JAK-STAT signaling
   pathway, in CD8^+ T cells, suggesting that LMO5 could activate the
   JAK-STAT signaling pathway. Furthermore, Tofacitinib reversed the
   increased expression of p-Stat3. Activation of the JAK-STAT signaling
   pathway can lead to the proliferation of CD8^+ T cells, intensifying
   inflammatory reactions. In line with this, we observed that LMO5
   induced the proliferation of CD8^+ T cells (Fig. [109]3H, I). Notably,
   Tofacitinib suppressed the proliferation of CD8^+ T cells induced by
   LMO5 in vitro.

Discussion

   In this study, scRNA-seq was conducted on PBMC and CSF of a 42-year-old
   male patient with refractory anti-GABA[A]-R encephalitis and thymoma.
   The analysis identified the JAK-STAT signaling pathway as a potential
   target. The patient received the JAK inhibitor tofacitinib, resulting
   in a notable improvement in clinical outcomes. This suggests that
   tofacitinib may have an intervention effect on the pathogenesis of
   anti-GABA[A] receptor encephalitis, highlighting the potential of
   scRNA-seq analyses in advancing personalized medicine for such
   patients.

   While previous studies primarily focused on humoral immunity
   [[110]4–[111]6], a group provided evidence for a significant role that
   CD8+ T cells may play in the pathogenesis [[112]7]. Recently,
   tissue-resident CD8+ T cells were found to be essential drivers of
   chronic CNS autoimmunity [[113]8]. Our scRNA-seq analysis uncovered
   broad transcriptomic changes in various T cell types, with the most
   significant alterations observed in the effector CD8+ T cells cluster.
   Pathway enrichment analysis revealed upregulated genes associated with
   the JAK-STAT signaling pathway, T-cell activation, and inflammatory
   response signaling in these effector CD8+ T cells, consistent with
   prior reports linking the JAK-STAT signaling pathway to T-cell
   activation [[114]9, [115]10]. Additionally, single-cell TCR sequencing
   identified a highly monoclonal TCR repertoire in CD8^+ T cells from the
   patient. Upregulation of JAK2, JAK3, and STAT1 was also identified in
   the specific monoclonal effector CD8^+ T cells. Therefore, we have
   demonstrated that a group of specific monoclonal effector CD8^+ T cells
   and abnormal JAK-STAT signaling pathways are involved in the disease.

   Further investigation into potential upstream molecules activating the
   JAK-STAT signaling pathway led us to explore LOM5, a tumor antigen.
   Previous reports indicated LOM5’s cross-reactivity with autoantibodies
   to anti-GABA[A]-R [[116]4]. In our study, elevated expression of LOM5
   was identified in peripheral blood, CSF, and thymoma from the patient.
   Bioinformatic analysis predicted LMO5’s binding to monoclonal TCR,
   suggesting its role in driving the immune process. LOM5-induced T cell
   proliferation, as demonstrated by the lymphocyte transformation test,
   supported the hypothesis that CD8^+ T cell-mediated early antitumor
   responses may release intracellular proteins including LOM5, triggering
   the production of GABA[A]-R autoantibodies.

   Furthermore, we observed higher p-Stat3 expression in CD8^+ T cells
   cultured with LOM5, which was reduced in CD8^+ T cells treated with
   tofacitinib. This led to the hypothesis that LMO5 induces CD8^+ T cell
   activation through the JAK-STAT signaling pathway, and tofacitinib
   suppresses LMO5-induced CD8^+ T cell proliferation in vitro.
   Collectively, our results suggest that the JAK-STAT signaling pathway
   may serve as a potential therapeutic target for Anti-GABA[A] receptor
   encephalitis.

   Notably, targeting the JAK-STAT signaling pathway with JAK inhibitors,
   such as tofacitinib, has shown success in various autoimmune and
   inflammatory diseases [[117]11, [118]12]. The favorable response to
   tofacitinib in our refractory case aligns with the emerging use of JAK
   inhibitors in treating rare cases of autoimmune or inflammatory
   diseases [[119]11, [120]13, [121]14]. The successful intervention with
   tofacitinib in our case also demonstrated aberrant JAK-STAT activity in
   anti-GABA[A]-R encephalitis.

   In conclusion, our results for the first time revealed specific
   monoclonal effector CD8^+ T cells and abnormal JAK-STAT signaling
   pathway activation in refractory anti-GABAA-R encephalitis. The
   successful intervention with tofacitinib supports the consideration of
   the JAK-STAT pathway as a potential therapeutic target in this disease.
   Nonetheless, these findings warrant further investigation in a larger
   patient cohort to validate their broader applicability.

Supplementary information

   [122]Supplementary information^ (1.7MB, pdf)
   [123]Supplementary table 1^ (11.7KB, xlsx)
   [124]Supplementary table 2^ (597.5KB, csv)
   [125]Supplementary table 3^ (1,019.4KB, csv)
   [126]Supplementary table 4^ (1.7MB, xlsx)
   [127]Supplementary table 5^ (40.4KB, xlsx)
   [128]Supplementary table 6^ (950KB, xls)
   [129]Supplementary movie 1^ (1.3MB, mp4)
   [130]Supplementary movie 2^ (1.7MB, mp4)
   [131]Supplementary movie 3^ (3MB, mp4)
   [132]Supplementary movie 4^ (4.3MB, mp4)
   [133]Supplementary movie 5^ (1.3MB, mp4)
   [134]Supplementary movie 6^ (4.5MB, mp4)
   [135]Supplementary movie 7^ (6.2MB, mp4)
   [136]Supplementary movie 8^ (4.5MB, mp4)

Acknowledgements