TOX deficiency facilitates the differentiation of IL-17A-producing γδ T cells to drive autoimmune hepatitis

Abstract

The specification of the αβ/γδ lineage and the maturation of medullary thymic epithelial cells (mTECs) coordinate central tolerance to self-antigens. However, the mechanisms underlying this biological process remain poorly clarified. Here, we report that dual-stage loss of TOX in thymocytes hierarchically impaired mTEC maturation, promoted thymic IL-17A-producing γδ T-cell (Tγδ17) lineage commitment, and led to the development of fatal autoimmune hepatitis (AIH) via different mechanisms. Transfer of γδ T cells from TOX-deficient mice reproduced AIH. TOX interacted with and stabilized the TCF1 protein to maintain the balance of γδ T-cell development in thymic progenitors, and overexpression of TCF1 normalized αβ/γδ lineage specification and activation. In addition, TOX expression was downregulated in γδ T cells from AIH patients and was inversely correlated with the AIH diagnostic score. Our findings suggest multifaceted roles of TOX in autoimmune control involving mTEC and Tγδ17 development and provide a potential diagnostic marker for AIH.

Introduction

Autoimmune hepatitis (AIH) is a severe inflammatory disease arising from an autoimmune response against liver autoantigens in genetically susceptible individuals [1]. The hallmarks of AIH are histological monocyte infiltration in the liver; the presence of elevated immunoglobulin G (IgG) and autoantibodies (autoAbs), including anti-nuclear Ab (ANA); and manifestations of liver damage. Dysregulation of immune tolerance plays a key role in the development of AIH [2]; however, its exact mechanisms have not been fully clarified.

The establishment of self-tolerance and the stepwise development of T-cell progenitors originating from CD4^–CD8^– double-negative (DN) thymocytes, which can be subcategorized into DN1-4 subsets by tracking the expression of the surface markers CD25 and CD44 and that later differentiate into CD4⁺CD8⁺ double-positive (DP) cells in the thymus, depend on the crosstalk between thymic parenchymal and epithelial cells [3]. Cortical and medullary thymic epithelial cells (cTECs and mTECs) regulate the positive selection of thymocytes and the diversity of the T-cell repertoire and eliminate autoreactive T cells by expressing tissue-restricted antigens (TRAs). mTECs also regulate the production of thymic regulatory T cells (tTregs) and IL-17A-producing γδ T (Tγδ17) cells to suppress immune hyperactivation and intensive inflammation [4, 5]. TOX has been demonstrated to function as a transcriptional and epigenetic coordinator in the maintenance of exhausted T cells during chronic infection and tumorigenesis [6, 7]. In addition, TOX regulates thymic development of CD4⁺ T lineage cells and induces CD8⁺ T-cell-mediated autoimmunity [8, 9]. However, the roles of TOX in other lineage commitments and self-tolerance are not well understood.

Thymic γδ T cells comprise preprogrammed IFN-γ-producing γδ T (Tγδ1) and Tγδ17 subsets [10], and Tγδ17 cells play pathogenic roles in various autoimmune diseases [11–13]. IL-7 expression restricted by the autoimmune regulator (Aire) in mTECs facilitates the Tγδ17 fate [5], whereas IL-15R signaling limits Tγδ17 effector bias during thymic development [14]. Stronger TCRγδ signaling drives the Tγδ1 fate, while weaker signaling facilitates the Tγδ17 fate [15]. However, recent data have reported that the Tγδ17 fate is programmed by regulatory networks consisting of critical transcription factors [16–19]. Whether and how TOX regulates Tγδ17 differentiation is unknown.

Liver-resident components control the modulation of intrahepatic γδ T cells to regulate regional immunity. Hepatic γδ T-cell progenitors have the potential to differentiate into Tγδ1 cells through an extrathymic developmental pathway [20]. Liver-resident Tγδ17 cells are maintained through lipid antigens presented by CD1d-expressing hepatocytes [21]. Hepatic γδ T cells play crucial roles in maintaining hepatic physiological homeostasis and modulating the progression of hepatic pathology [22]. Importantly, exacerbation of liver diseases, including AIH, nonalcoholic fatty liver disease, liver fibrosis and cirrhosis, and liver cancer, has been linked to IL-17A production in hepatic γδ T cells interacting with other intrahepatic cells [21, 23–25].

Here, we reveal that TOX is an essential regulator of Tγδ17 differentiation and is responsible for the maintenance of self-tolerance and prevention of AIH. Specifically, TOX serves as a stabilizer of T-cell factor 1 (TCF1), a negative regulator of Tγδ17 fate [18], through direct interaction to prevent its degradation. Thus, TOX is required for immune homeostasis, in turn preventing Tγδ17-mediated AIH pathogenesis.

Materials and methods

Clinical sample acquisition

Liver tissues and peripheral blood were obtained from AIH patients who underwent surgery at the Affiliated Drum Tower Hospital of Nanjing University Medical School (Nanjing, China). Samples from patients with benign hepatic lesions who underwent liver resection were collected as controls. All participants signed a written informed consent form. The study was approved by the Institutional Ethics Committees of Nanjing Drum Tower Hospital and performed in accordance with the Declaration of Helsinki and government policies.

Mice

Tox^f/f and Tox^f/fCd4^Cre mice were generated as previously described [7]. B6. Tox^f/fPdcd1^Cre mice were obtained by crossing Tox^f/f mice with Pdcd1^Cre mice obtained from Dr Xingxu Huang (ShanghaiTech University, China). The primers used for genotyping are listed in Supplementary Table S1. B6.CD45.1⁺, CD45.2⁺, and Rag1^–/– mice were purchased from The Jackson Laboratory. Male mice at the indicated weeks of age were used for the experiments, and mutant mice were compared with their littermate controls. Mice were housed under specific pathogen-free conditions on a 12-h light-dark cycle at 22–24 °C and used in accordance with the Animal Care and Use Committee guidelines of the Affiliated Drum Tower Hospital of Nanjing University Medical School.

Cell lines

HEK-293T and Jurkat cells purchased from ATCC were maintained in DMEM supplemented with 10% FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin (Invitrogen, CA, USA) in a humidified atmosphere at 37 °C with 5% CO₂. OP9 cells expressing the Notch ligand Delta-like 1 (OP9-DL1), a kind gift from Dr Linrong Lu (Zhejiang University, China) [26], were maintained in MEMα (Gibco, CA, USA) supplemented with 20% FBS in a humidified atmosphere at 37 °C with 5% CO₂.

Preparation of cell suspensions and flow cytometry

Thymi and spleens harvested from mice were mechanically disrupted and further dissociated by filtering through a 70-μm cell strainer (BD Falcon, CA, USA). Ficoll density gradients (GE Healthcare, Uppsala, Sweden) were used for the isolation of lymphocytes from PBMCs. The cell pellets were resuspended in RBC lysis buffer and incubated for 5 min. Lymphocytes from mouse livers were isolated as previously described [27]. Briefly, mice were anesthetized, and livers were perfused in situ via the portal vein with 50 mL of liver perfusion medium followed by 50 mL of liver digestion medium (Invitrogen). Livers were dissected and placed into dishes with a complete RPMI 1640 medium. The cells were centrifuged at 50 × g for 3 min to pellet hepatocytes. The supernatants were transferred to new tubes and then centrifuged at 320 × g for 5 min. For isolation of liver-infiltrated lymphocytes from AIH patients, tissues were minced into small pieces and digested with collagenase type II (100 U/mL) and DNase I (20 μg/mL) at 37 °C for 30 min. The liver digests were filtered through a 70-μm cell strainer, and the cells were centrifuged at 600 × g at 20 °C for 30 min on a gradient consisting of 80 and 40% Percoll solutions (GE Healthcare). Cells were obtained from the interface of the Percoll layers. Thymic epithelial cells (TECs) were enriched using a Percoll gradient as described previously [28].

For surface molecule staining, cells were incubated with the following fluorochrome-conjugated antibodies or conjugates: anti-TCRγδ (GL3), anti-CD3e (BM10-37 or SP34-2), anti-CD8a (53-6.7), anti-EpCAM (G8.8), anti-CD69 (H1.2F3), anti-RANKL (IK22-5), anti-CD27 (LG.3A10), anti-CD45R/B220 (RA3-6B2) and Annexin V (all from BD Biosciences, San Diego, CA, USA); anti-CD4 (GK1.5), anti-TCR Vγ1.1 (4B2.9), anti-c-Kit (ACK2), anti-CD44 (IM7), anti-CD45 (I3/2.3), anti-MHC-II (M5/114.15.2), anti-CD24 (M1/69), anti-CXCR5 (L138D7), anti-Fas (SA367H8), anti-CD62L (MEL-14), anti-NK1.1 (PK136), anti-TCRγδ (B1), and anti-TCRδ2 (B6) (all from BioLegend, San Diego, CA, USA); anti-TCR Vγ2 (UC3-10A6), anti-PD1 (J43), anti-CD16/32 (93), anti-GL7 (GL7), anti-TCR beta (H57-597), anti-CD40L (MR1), and anti-TCRδ1 (TS8.2) (all from eBioscience, San Diego, CA, USA); anti-CD25 (REA568), anti-CCR6 (R6H1), anti-CD45.1 (A20), and anti-CD45.2 (REA1223) (all from Miltenyi, Bergisch Gladbach, Germany); and UEA-1 (Vector Laboratories, San Francisco, CA, USA). For intracellular staining, cells were stimulated for 4–6 h with Leukocyte Activation Cocktail, treated with a Fixation/Permeabilization Solution Kit (BD Biosciences), and stained with the following antibodies: anti-IFN-γ (XMG1.2) and anti-CD107a (1D4B) from BD Biosciences, anti-IL-17A (BL168) and anti-IFN-γ (4 S.B3) from BioLegend, and anti-IL-17A (eBio17B7) from eBioscience. Transcription factors were stained using a Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and the following antibodies: anti-Ki-67 (16A8) from BioLegend; anti-RORγt (AFKJS-9), anti-EOMES (Dan11mag), and anti-FOXP3 (FJK-16s) from eBioscience; and anti-TOX (REA473) from Miltenyi. Live cells were identified based on staining with Fixable Viability Stain 620 (FVS620, BD Biosciences). Samples were acquired, recorded, or sorted in a FACS Aria III Cell Sorter (BD Biosciences), and data were analyzed with FlowJo software (TreeStar, OR, USA).

Reaggregated thymus organ culture (RTOC)

RTOC was performed as previously described [29]. Briefly, thymus lobes from E15.5 mice were transferred to a nucleopore filter/sponge (Millipore, MA, USA) in RPMI 1640 medium supplemented with 10% FBS and 1.35 mM 2-deoxyguanosine (dGuo; Sigma‒Aldrich, Saint Louis, MO, USA) for 6 days to eliminate endogenous T-cell progenitors. After the removal of dGuo by diffusion in PBS at 37 °C, the thymus lobes were dissociated in 0.25% trypsin (Sigma‒Aldrich) and 0.02% EDTA (Sigma‒Aldrich) in PBS. The reaggregates of FACS-sorted stromal cells (1 × 10⁵) and the indicated thymocytes (1 × 10⁵) obtained from wild-type (WT) C57BL/6 mice were centrifuged in volumes of 2 μL and cultured on the surface of filters for 7 days.

OP9-DL1 coculture system

FACS-sorted DN2 (c-Kit⁺CD4^–CD8^–CD3^–CD25⁺CD44⁺) cells isolated from 2-week-old mice were plated onto OP9-DL1 monolayers in MEMα containing 20% FBS (Gibco), 1 ng/mL IL-7 (PeproTech), and 5 ng/mL Flt3L (R&D Systems). The cells were cultured in media in the presence or absence of recombinant mouse IL-15 (rmIL-15, 10 ng/mL, BioLegend, CA, USA) and analyzed after 7 days of coculture.

Generation of mixed thymocyte chimeras

Thymocytes from postnatal day 3 WT (CD45.1/2⁺) and the indicated mutant (CD45.2⁺) mice were obtained and mixed at a ratio of 1:1, and a total of 10 × 10⁶ cells were transferred into lethally irradiated (1200 rads) WT (CD45.1⁺) recipient mice. The recipient mice received 5 × 10⁶ WT CD45.1⁺ bone marrow cells the next day and were reconstituted for 8 weeks.

In vivo treatment

For the blocking experiments, an anti-IL-17 neutralizing antibody (10 μg, clone 50104, R&D Systems, MN, USA) or rat IgG2a isotype control (clone 54447) in 100 μL of sterile PBS was administered by intraperitoneal (i.p.) injection twice a week. The treatment was performed from 2 to 32 weeks of age. For the stimulation experiments, mice were administered an i.p. injection of PBS or recombinant mouse RANKL (rmRANKL, 20 ng/g body weight, R&D Systems) twice a week. Some mice were treated with PBS or rmIL-15 (10 ng/g weight) by i.p. injection every 2 days. The treatments were performed from the day of birth to 3 weeks of age.

Plasmid construction, lentivirus production and transduction, and intrathymic (IT) injection

The lentiCRISPRv2-eGFP vector for CD40L disruption was constructed by inserting a P2A-eGFP cassette between the MluI and BamHI restriction sites in lentiCRISPRv2 puro (Addgene #98290) using a ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China). Short guide sequences were inserted between the BsmBI restriction sites. The sequences of the sgRNAs were as follows: sgCD40L-1, TTTATCCTTGCTGAACTGTG; sgCD40L-2, GCAATTTGAAGACCTTGTCA; sgCD40L-3, TGAACTGTGAGGAGATGAGA. Lentivirus was produced via triple transfection of HEK-293T cells with a lentiviral vector, the packaging plasmid psPAX2, and the envelope plasmid pMD2.G at a 5:3.75:1.25 ratio using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol and were collected 48 h after transfection for infection of γδ T cells by centrifugation for 45 min at 400 × g for further analyses.

To obtain the lentiviral vector encoding TCF1 and eGFP, the TCF1/P2A/eGFP fragment was inserted into the EcoRI/NotI sites in pCDH-CMV-MCS-EF1-Puro. Lentivirus was produced as described above. Viral titers, presented as transduction units (TUs), were determined by assessing the transduction of Jurkat cells with serial dilutions of virions concentrated using a PEG Virus Precipitation Kit (Abcam, MA, USA). Virions in a volume of 20 µL containing 2 × 10⁸ TUs were injected directly into the thoracic cavity above the sternum of 1-week-old mice using a 0.3-mL, 31-gauge, 8-mm insulin syringe. The IT-reconstituted mice were sacrificed 4–5 weeks after injection for analyses.

Adoptive transfer assay

Splenocytes were isolated from 4-week-old Tox^f/fPdcd1^Cre, Tox^f/fCd4^Cre, and control mice and transferred (10 × 10⁶) into 5-week-old Rag1^–/– mice via tail vein injection. For cotransfer of B and γδ T cells, splenic B cells were isolated from WT mice by using anti-CD19 beads and a magnetic cell separator (Miltenyi), and γδ T cells pretreated with the indicated lentivirus from TOX-deficient or control mice were sorted by flow cytometry. Then, reconstitution of Rag1^–/– mice was performed with a ratio of 1:15 (1 × 10⁶ and 15 × 10⁶ cells, respectively) via tail vein injection. For cotransfer of αβ T cells and γδ T cells, purified CD4⁺ T cells, CD8⁺ T cells, and γδ T cells isolated from the indicated mice were pooled at a ratio of 20:10:1 (6 × 10⁶, 3 × 10⁶, and 3 × 10⁵ cells respectively), resuspended in 200 μL of PBS and transferred into Rag1^–/– mice via tail vein injection. Recipient mice were analyzed 5 weeks after injection.

AIH histopathology

For assessment of liver inflammation, the left liver lobes from the indicated mutant and control mice were excised, fixed with 4% formalin, and embedded in paraffin. Liver sections were stained with H&E. AIH was evaluated by pathologists and scored using light microscopy according to a modified Scheuer scoring scale, which included scores for portal inflammation (0, normal; 1, a few cells in portal tracts; 2, lymphoid aggregates in some portal tracts; 3, lymphoid aggregates in most portal tracts), interface hepatitis (0, none; 1, focal in scattered portal tracts; 2, focal in most portal tracts or continuous in some portal tracts; 3, continuous in most portal tracts), and lobular inflammation (0, none; 1, scattered foci of lobular-infiltrating lymphocytes; 2, numerous foci of lobular-infiltrating lymphocytes; 3, panlobular-infiltrating lymphocytes).

Serological quantification of transaminases and immunoglobulin

The serum levels of ALT and AST were measured by standard procedures in the Clinical Laboratory Center of the Affiliated Drum Tower Hospital of Nanjing University Medical School. Serum IgM, IgG, and ANA were quantified with the corresponding ELISA kits (Alpha Diagnostic International, TX, USA) at 1:2000, 1:500, and 1:100 dilutions, respectively. ELISA plates were read in a Spark Multimode Reader (Tecan, Mannedorf, Switzerland). Quantification of samples was performed with a standard curve.

Quantitative assessment of fibrosis

Liver sections were stained using a Picrosirius Red Stain Kit (Abcam) to measure the collagen content according to the manufacturer’s instructions. Assessment of collagen staining was performed with an Eclipse E100 birefringence microscope (Nikon, Tokyo, Japan), and the staining intensity was analyzed with ImageJ software. The average staining intensity in five random fields of view was calculated for each mouse.

Multiplex immunofluorescence staining (mIHC) analysis

Murine and human sections were evaluated with Tyramide SuperBoost kits (including Alexa Fluor 647–conjugated, Alexa Fluor 594–conjugated, Alexa Fluor 555–conjugated, and Alexa Fluor 488–conjugated tyramide) (Thermo Fisher Scientific, MA, USA) according to the manufacturer’s protocol. Briefly, paraffin-embedded tissue specimens were deparaffinized at 60 °C for 30 min and rehydrated with xylene and a graded alcohol series. After antigen retrieval, incubation with 3% hydrogen peroxide, and blocking, tissue sections were incubated first with the corresponding primary antibodies and then with a poly-HRP-conjugated secondary antibody and Alexa Fluor-conjugated tyramide reagent. Finally, the HRP reaction was stopped, and the tissue sections were subjected to multiplex immunostaining for the detection of other signals. The slides were incubated with anti-CD3 (1:200, clone SP7), anti-keratin 5 (1:200, clone EP1601Y), anti-keratin 8 (1:200, clone EP1628Y), anti-CXCR5 (1:200, clone EPR23463-30), anti-TOX (1:200, clone NAN448B), anti-CD19 (1:200, clone EPR23174-145), anti-IFN-γ (1:200, polyclonal), anti-IL-17A (1:200, polyclonal), and anti-PD1 (1:200, polyclonal) antibodies from Abcam; anti-CD4 (1:100, clone D7D2Z), anti-CD8α (1:400, clone D4W2Z), or anti-EpCAM (1:400, clone E6V8Y) from Cell Signaling Technology (Danvers, MA, USA); anti-TCRγδ (1:200, clone 5A6. E9) from Thermo Fisher Scientific; anti-Aire (1:200, clone C-2) from Santa Cruz (Dallas, TX, USA); or anti-TCR beta (1:200, clone 8A3) from Novus Biologicals (Centennial, CO, USA). Nuclei were counterstained with DAPI. Images were acquired with a THUNDER Imaging System (Leica, Wetzlar, Germany).

Immunohistochemical (IHC) analysis

Immunohistochemical analysis was performed with an UltraSensitive SP (Mouse/Rabbit) IHC Kit (Maixin Biotech, Fuzhou, China) according to the manufacturer’s protocol. Briefly, tissue sections were first deparaffinized and rehydrated, and antigen retrieval was then performed. The slides were incubated with 3% hydrogen peroxide and blocking buffer for 10 min and 1 h, respectively. The tissue sections were incubated first with an anti-CCL20 antibody (1:200, polyclonal, Abcam) overnight at 4 °C and then with a biotin-conjugated secondary antibody, and two incubations with a streptavidin-conjugated tertiary antibody were then performed for 20 min and 10 min at room temperature. Antigens were detected with 3,3’-diaminobenzidine solution.

Immunoblot analysis

Cells were lysed in ice-cold RIPA buffer containing protease and phosphatase inhibitors (Beyotime, Nantong, China) for 30 min, and the protein concentration was measured using a BCA assay (Beyotime). The lysate supernatants were heated to 95 °C with loading buffer for 10 min, separated by SDS‒PAGE, and transferred to PVDF membranes (Millipore). The membranes were blocked with 5% milk in TBST and probed with the following antibodies: anti-TOX (1:1000, clone NAN448B) and anti-ERK2 (1:1000, clone E460) from Abcam and anti-FLAG (1:1000, clone D6W5B), anti-His (1:1000, clone D3I1O), anti-TCF1 (1:1000, clone C63D9), and anti-GAPDH (1:1000, clone D16H11) from Cell Signaling Technology. The membranes were visualized with an enhanced chemiluminescence detection kit (Thermo Fisher Scientific) and a ChemiDoc MP Imaging System (Bio-Rad, CA, USA).

Coimmunoprecipitation (Co-IP)

DN2 thymocytes or HEK-293T cells transfected with the indicated plasmids were washed with PBS buffer twice, and co-IP was then performed with a Direct Magnetic IP/Co-IP Kit (Thermo Fisher Scientific) following the manufacturer’s instructions. Briefly, ice-cold IP lysis buffer was added to the cell pellets, and the lysates were incubated on ice for 5 min with periodic mixing. The supernatants were incubated with antibody-bound beads for 2 h at room temperature. A small portion of each supernatant was saved as input. The following antibodies were used: anti-FLAG (1:50, clone D6W5B), anti-His (1:50, clone D3I1O), anti-TOX (1:100, clone E6G5O), and anti-TCF1 (1:50, clone C63D9) from Cell Signaling Technology. After the final wash with wash buffer, the beads were resuspended and boiled at 95 °C for 5 min. Samples were analyzed by immunoblotting as described earlier.

Detection of autoantibodies

The presence of autoAbs was investigated by incubating sections of tissues from Rag1^–/– mice with a 1:40 dilution of sera from 4–5-week-old Tox^f/fPdcd1^Cre or control mice, followed by incubation with a FITC-conjugated anti-mouse IgG secondary antibody and visualization by fluorescence microscopy. In another approach, lysate supernatants of tissue samples from Rag1^–/– mice were used for the detection of autoAbs, as described in the “Immunoblot analysis” section. The membranes were probed with a 1:100 dilution of sera from the indicated mice for 1 h and horseradish peroxidase-labeled anti-mouse IgG for 1 h and were then visualized with chemiluminescence reagents.

Quantitative real-time PCR (qRT–PCR)

RNA from FACS-sorted cells was isolated with an RNeasy Mini Kit (QIAGEN, Dusseldorf, Germany) following the manufacturer’s instructions. Template RNA was treated with DNase to avoid genomic DNA contamination. Reverse transcription reactions were performed using PrimeScript RT Master Mix (Takara, Dalian, China). ChamQ Universal SYBR qPCR Master Mix (Vazyme) was used to perform quantitative real-time PCR on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems, CA, USA). Expression levels were calculated relative to Gapdh. The primer pairs used are listed in Supplementary Table S1.

Genomic DNA extraction and PCR

Genomic DNA from sorted cells was purified with a Genomic DNA Mini Kit (Invitrogen) following the manufacturer’s instructions and used for subsequent PCR. Genomic PCR for Tox was performed using the primers listed in Supplementary Table S1.

Mass spectrometry

Proteins from DN2 thymocyte lysates were immunoprecipitated with an anti-TOX antibody, separated by SDS‒PAGE, and then visualized by Coomassie brilliant blue staining. In-gel protein digestion and LC‒MS/MS were performed as previously described [30]. The acquired LC‒MS/MS spectra were mapped to the Mus_musculus_UniProt database. The raw data have been deposited in Mendeley Data: 10.17632/h263bgxmb5.2.

Reanalysis of single-cell RNA sequencing data

Single-cell RNA-seq data for the thymus at postnatal day 0 were obtained from GEO under accession number GSE107910. Single-cell RNA-seq analysis was performed as previously described [31]. Briefly, t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction was performed on the calculated principal components (PCs) to obtain a two-dimensional representation. For cell cluster identification, the top 10 PCs were used for t-SNE visualization. STAR Aligner (V2.4.2), Picard tools (V1.96), and Drop-seq tools (V1.0) were used to convert raw FASTQ files into gene expression matrices.

RNA sequencing and bioinformatics analysis

Total RNA from FACS-sorted thymocytes was extracted with TRIzol reagent (Invitrogen) and then purified with DNase I (Qiagen). The RNA-seq library for the RNA samples was constructed by using a NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, MA, USA). The library was sequenced using an Illumina NovaSeq 6000 instrument, and the clean paired-end reads were mapped to Ensembl_release100. Pathway analysis was performed to determine the pathways significantly enriched with the differentially expressed genes according to the KEGG database. The gene set variation analysis (GSVA) package in R was used to explore the changes in specific pathways. Gene set enrichment analysis (GSEA) was employed to verify the biological processes associated with the differentially expressed genes. Raw data were deposited in the GEO database (GSE187504).

Quantification and statistical analysis

Statistical analyses were performed in GraphPad Prism version 7.00 (GraphPad Software, CA, USA). The data are presented as mean ± SD as indicated in the figure legends. A two-tailed Mann–Whitney U test was used to calculate P values for unpaired samples. To analyze three or more groups, the Kruskal–Wallis test was used with Dunn’s multiple comparison test. The log-rank test was performed to analyze Kaplan–Meier survival data. Spearman correlation analysis was used to analyze correlations. P values <0.05 were considered significant.

Results

TOX controls positive selection and CD4⁺ thymocyte production

To assess the role of TOX in αβ/γδ lineage decisions, we crossed floxed Tox mice with Pdcd1-Cre transgenic mice (Tox^f/fPdcd1^Cre), which were more appropriate than Lck^Cre mice [32, 33], and thus deleted Tox in most CD4^–CD8^– DN, CD4⁺CD8⁺ DP, CD4⁺ single-positive (SP), CD8⁺ SP, and γδ T cells (Supplementary Fig. S1A, B). In WT mice, TOX and PD1 expression coincided starting at the DN2-3 progenitor stage (Fig. 1A and Supplementary Fig. S1C), where commitment to the γδ or αβ lineage is made through γδ-selection or β-selection, respectively. Upon Tox inactivation after the DN stage, we observed reductions in CD4⁺CD8^lo and CD4⁺ SP cells, accompanied by increases in CD4^loCD8^lo double-dull cells, but no differences in DP, CD8⁺ SP, and gross thymocytes (Fig. 1B). The frequencies of the TCR-β^hi and positively selected CD4⁺CD8^lo subsets were lower in Tox^f/fPdcd1^Cre mice (Fig. 1C, D). These observations demonstrate that TOX deficiency inhibits CD4⁺ thymocyte production by impairing positive selection at the CD4⁺CD8^lo stage.

Fig. 1.

TOX controls positive selection and αβ/γδ lineage commitment. A Representative dot plots of TOX and PD1 expression in WT thymocyte subsets. B Cell counts of thymocyte subsets in Tox^f/f and Tox^f/fPdcd1^Cre mice (n = 10). C Frequencies of thymocyte subsets in Tox^f/f and Tox^f/fPdcd1^Cre mice gated on the CD45⁺ or TCR-β^hi subset (n = 6). D Representative dot plots of CD69 and TCR-β expression in thymocyte subsets from Tox^f/f and Tox^f/fPdcd1^Cre mice (n = 6). E, F Frequencies of the DN1-4 (E), DN1.0, and DN1.5 (F) subpopulations among DN thymocytes from Tox^f/f and Tox^f/fPdcd1^Cre mice (n = 6). G, H Representative dot plots of γδ T cells gated on DN thymocytes (G) and cytokine expression in γδ T cells (H) from Tox^f/f, Tox^f/fCd4^Cre, and Tox^f/fPdcd1^Cre mice (n = 6). The data are presented as mean ± SD; NS, P > 0.05; **P < 0.01; ***P < 0.001; Mann–Whitney U test (B–F); Kruskal–Wallis test (G, H). All data in the figure are from 2–3-week-old mice

TOX regulates γδ lineage commitment and Tγδ17 differentiation

To determine whether TOX regulates early thymocyte differentiation, we assessed the distribution of the DN1-4 subsets. The percentage of the DN CD25⁻CD44⁺ (DN1) subset was increased and that of the DN CD25⁺CD44⁺ (DN2) subset was reduced in Tox^f/fPdcd1^Cre mice (Fig. 1E and Supplementary Fig. S1D). The DN1 population can be further divided into two distinct subpopulations, DN1.0 (CD25⁻CD44⁺) and DN1.5 (CD25^loCD44⁺), based on CD25 expression. The DN1.5 population, actually constituting the γδ T lineage during progression [34], showed a higher frequency in Tox^f/fPdcd1^Cre mice (Fig. 1F and Supplementary Fig. S1E). Deletion of Tox after the DN stage (early stage) in Tox^f/fPdcd1^Cre mice increased the thymic γδ T-cell population, while deletion after the DP stage (late stage) in Tox^f/fCd4^Cre mice was associated with normal proportions and numbers (Fig. 1G). Although the maturation of γδ thymocytes in both strains of TOX-deficient mice was comparably promoted, PD1 downregulation was only observed in Tox^f/fPdcd1^Cre mice (Supplementary Fig. S1F). Given that Tγδ17 cells are negative for PD1 expression [35], we speculated that the differentiation pattern of γδ cells was skewed toward Tγδ17 subsets in Tox^f/fPdcd1^Cre mice. We next evaluated γδ effector subsets in TOX-deficient mice. In Tox^f/fPdcd1^Cre mice, polarization toward Tγδ17 subsets was nearly absolute and more obvious than that in Tox^f/fCd4^Cre mice, while Tγδ1 thymocytes were reduced in Tox^f/fPdcd1^Cre mice (Fig. 1H and Supplementary Fig. S1G). No differences in the thymic T helper 1 (Th1) or Th17 proportions were observed in these mice (Supplementary Fig. S1H). Of note, TOX-deficient Vγ2⁺ cells (IL-17A producers, Heilig & Tonegawa nomenclature [36]) and Vγ1.1⁺ cells (IFN-γ producers) showed higher Tγδ17 frequencies (Supplementary Fig. S2A). These results suggest that early-stage TOX expression is an essential factor for γδ T-cell differentiation and Tγδ17-cell commitment.

Characteristics of the thymic immune microenvironment were altered by TOX deficiency in Tox^f/fPdcd1^Cre mice

Stromal cells have been reported to participate in the regulation of Tγδ17 specialization [10]. To investigate the function of TECs in this process, we first showed that PD1 is not expressed in TECs (Supplementary Fig. S2B, C). Moreover, TOX expression was not observed in mTECs or cTECs, indicating that no direct effect on the development of TECs was induced by Pdcd1-Cre-mediated TOX deletion (Supplementary Fig. S2D). Single-cell RNA sequencing data from the murine thymus [31] suggested relatively high expression of TOX and PD1 in γδ T (sc-7 and -13) and NK (sc-10) cells (Supplementary Fig. S2E–G). We verified that PD1 was expressed in half of the γδ T cells and that TOX was deleted in Tox^f/fPdcd1^Cre γδ T cells (Supplementary Fig. S2H). Thymic NK cells from Tox^f/fPdcd1^Cre mice showed unchanged TOX expression and secretory function, although some cells expressed PD1 (Supplementary Fig. S2I, J). In addition, Tregs expressed PD1 but exhibited TOX levels comparable to those in Tox^f/fPdcd1^Cre mice and control mice; thus, the reductions in tTregs in both strains of TOX-deficient mice resulted from the compromised negative selection of Tregs but not TOX deficiency in Tregs (Supplementary Fig. S2K, L), supporting the findings of a previous study [8].

Insufficiency of RANKL secretion from CD4⁺ thymocytes constitutes an overlapping pathway for mTEC and Tγδ17 dysregulation under TOX deficiency at distinct stages

Expansion of mTECs requires interaction with positively selected CD4⁺ thymocytes [37]. The scattered distribution of mTECs was observed in Tox^f/fPdcd1^Cre mice, with undefined corticomedullar junctions and reductions in functional and mature mTECs (Fig. 2A–C). Decreased proliferation but not apoptosis was the cause of the limited cellularity of mTECs (Fig. 2D and Supplementary Fig. S3A). The mRNA levels of Aire, Aire-dependent TRAs (Spt1, Spt2, Ins2, and Mup1), and an Aire-independent TRA (Csn2) were decreased in Tox^f/fPdcd1^Cre mice, indicating defects in mTEC maturation (Fig. 2E).

Fig. 2.

TOX deficiency after the early stage drives Tγδ17 expansion by regulating mTEC and γδ T progenitor development. A Representative H&E and multiplex immunofluorescence (mIHC) staining for Aire (red) and keratin 5 (green) in the thymus in 2–3-week-old Tox^f/f and Tox^f/fPdcd1^Cre mice. The medullary area and cellularity of Aire⁺ mTECs were measured (n = 8). Scale bar, 100 µm. m medulla, c cortex. B mIHC staining for UEA-1 (red) and keratin 5 (green) and the cellularity of UEA-1⁺ cells in the thymus in 2–3-week-old Tox^f/f and Tox^f/fPdcd1^Cre mice (n = 8). Scale bar, 100 µm. C Frequencies of cTECs (UEA-1^–MHC-II⁺), immature (UEA-1⁺MHC-II^int), and mature (UEA-1⁺MHC-II^hi) mTECs gated on the CD45^– subset in the thymus in 2–3-week-old Tox^f/f and Tox^f/fPdcd1^Cre mice (n = 6). D Proliferation was evaluated by Ki67 staining in total, immature and mature mTECs gated as in (C) (n = 6). E The mRNA levels of Aire, Spt1, Spt2, Ins2, Mup1 and Csn2 in mTECs (n = 6). F The mTEC numbers in RTOCs reaggregated for 7 days with stromal cells from E15.5 Tox^f/f and Tox^f/fPdcd1^Cre mice and the indicated thymocytes from 6-week-old WT mice were measured by flow cytometry (n = 4). G–I Newborn mice were treated with PBS or rmRANKL (20 ng/g weight) by i.p. injection twice a week for 3 weeks. The numbers of total mTECs (G) and Aire⁺ mTECs (H) and the mRNA level of Il15 (I) were measured. J Cytokine expression in γδ T cells from Tox^f/f, Tox^f/fCd4^Cre, and Tox^f/fPdcd1^Cre newborn mice after treatment with PBS or rmIL-15 (10 ng/g weight) every 2 days for 3 weeks. Tox^f/f mice served as controls (n = 6). The data are presented as mean ± SD; NS, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; Mann–Whitney U test (A–F); Kruskal–Wallis test (G–J)

To study the mechanism of impaired mTEC development, RTOC experiments were performed. Only CD4⁺ thymocytes induced significant mTEC expansion and increased Aire and TRA expression, but these cells were not required for cTEC cellularity (Fig. 2F and Supplementary Fig. S3B–E). Defects in medulla formation similar to those observed in Tox^f/fCd4^Cre mice confirmed the requirement of CD4⁺ thymocytes for mTEC maturation (Supplementary Fig. S3F, G).

RANKL exhibits preferential expression in CD4⁺ thymocytes and transduces key signals for the development of mTECs [37]. Although few CD4⁺ thymocytes exhibited comparable RANKL expression, administration of rmRANKL triggered the expansion and maturation of mTECs in both strains of TOX-deficient mice (Fig. 2G, H and Supplementary Fig. S3H, I). Notably, Tox^f/fCd4^Cre and Tox^f/fPdcd1^Cre mice had sequentially higher Tγδ17 frequencies, and completely and partially normal percentages of Tγδ17 cells, respectively, were restored by rmRANKL administration (Supplementary Fig. S3J). This observation indicated a supplementary role of TOX after the DN stage relative to the DP stage of Tγδ17 lineage commitment. The tTreg reduction in TOX-deficient mice was also rescued by rmRANKL administration, suggesting that the tTreg alteration is attributed to compromised negative selection of mTECs (Supplementary Fig. S3K).

TOX defects after the DN stage drive enhanced Tγδ17 expansion and partially share the bias by reducing IL-15 secretion from mTECs with TOX loss after the DP stage

Aire influences the production of Tγδ17 cells by regulating cytokine production [5]; thus, we suggested the possibility that TOX might partially regulate Tγδ17 differentiation in an mTEC-dependent manner. As expected, mTECs in both strains of TOX-deficient mice exhibited much lower Il15 levels but higher Il7 levels than those in control mice (Supplementary Fig. S3L). rmRANKL application increased Il15 levels and reduced Il7 levels (Fig. 2I and Supplementary Fig. S3M).

To confirm the hypothesis that the effect of TOX on γδ T-cell specialization after the DP stage is attributed to suppressed IL-15 production, we subjected both strains of TOX-deficient mice to rmIL-15 administration. Tox^f/fPdcd1^Cre mice administered rmIL-15 still showed an increased percentage of the Tγδ17 subset but a reduction compared to that in control mice, unlike the normal level in Tox^f/fCd4^Cre mice administered rmIL-15 (Fig. 2J). To exclude the extrinsic effect on thymic Tγδ17 development, we used the OP9-DL1 coculture system, which showed normal Tγδ17 subsets in the cocultures from Tox^f/fCd4^Cre mice, while both cocultures from Tox^f/fPdcd1^Cre mice showed a comparable but higher Tγδ17 frequency (Supplementary Fig. S3N). Tγδ17 differentiation was comparable between WT and Tox^f/fCd4^Cre mixed thymocyte chimeras but was reinforced in Tox^f/fPdcd1^Cre donors, suggesting a nonredundant role of early-stage TOX in Tγδ17 commitment (Supplementary Fig. S3O, P). Collectively, we clarified the dual-stage roles of TOX in Tγδ17 lineage commitment—late developmental stage loss of TOX in thymocytes drives CD4⁺ thymocyte insufficiency-mediated Tγδ17 expansion by impairing the development of IL-15-producing Aire⁺ mTECs, whereas early-stage TOX loss additionally restricts Tγδ17 commitment of progenitors in a cell-intrinsic manner.

Absence of TOX leads to fatal AIH

The balance of IL-17A and IFN-γ production controls inflammation and autoimmunity [10]. The increase in Tγδ17 cells prompted us to evaluate the role of TOX in autoimmune disease. We first examined the immune changes in the periphery. PD1 was expressed in most γδ T cells and a few CD4⁺ T cells but not in B cells (Supplementary Fig. S4A, B). While Tox^f/fPdcd1^Cre mice showed a higher frequency of follicular helper T (T_FH) cells with constitutive PD1 expression [38], no difference in the number of T_FH cells was observed (Supplementary Fig. S4C, D). In addition, reductions in Tregs and CD4⁺ T cells were observed (Supplementary Fig. S4E, F).

Tox^f/fPdcd1^Cre mice showed marked weight loss with ruffled fur and poor survival, developing severe AIH with lymphocyte infiltration (Fig. 3A–C). Leakage of autoreactive DP cells was detected in AIH lesions, and these cells exhibited a higher frequency of an activated phenotype, as did T cells (Fig. 3D, E). Although Tox^f/fCd4^Cre mice exhibited better survival and more normal weights than Tox^f/fPdcd1^Cre mice (Supplementary Fig. S4G, H), both exhibited inflammatory infiltrates in multiple organs and increased ALT, AST, IgM, IgG, and ANA levels (Fig. 3F–K and Supplementary Fig. S4I–N). Extensive autoAbs against self tissues were detected (Fig. 3L). Hierarchical autoAb production and germinal center (GC) formation provided an explanation for the pathological differences in TOX-deficient mice (Supplementary Fig. S4O, P). These data indicate that the stage of TOX deficiency determines AIH severity.

Fig. 3.

Absence of TOX leads to fatal AIH. A Representative images and weights of Tox^f/f and Tox^f/fPdcd1^Cre mice at 6 weeks of age (n = 12). B Survival analysis of Tox^f/f mice and Tox^f/fPdcd1^Cre mice (n = 20). C Representative H&E and mIHC staining for CD4 (purple) and CD8 (green) and quantification of T cells in the indicated livers (n = 8). Scale bar, 200 µm. D Representative mIHC staining for CD4 (green) and CD8 (purple) and absolute numbers of CD4⁺CD8⁺ T cells in the indicated liver regions (n = 8). Scale bar, 20 µm. E Representative plots of CD44 and CD62L expression in CD4⁺, CD8⁺ and CD4⁺CD8⁺ T cells from the indicated livers (n = 8). F Representative H&E staining for evaluation of inflammatory infiltrates in multiple organs. Scale bar, 200 µm. G–K The serum levels of ALT, AST, IgM, IgG, and ANA were measured (n = 10). L Detection of serum autoAbs against multiple organs in the indicated mice. Scale bar, 200 µm. The data are presented as mean ± SD; NS, P > 0.05; **P < 0.01; ***P < 0.001; Mann–Whitney U test (A, G–K); Mantel–Cox log-rank test (B); Kruskal–Wallis test (C). The data in the figure are from 4–5-week-old mice unless otherwise stated

TOX-associated AIH is attributed to selective accumulation of Tγδ17 cells in the liver

To evaluate the role of γδ T-cell commitment in AIH pathogenesis, the compositions of peripheral γδ T-cell subsets were analyzed. Although the frequencies of γδ T cells in the spleen and peripheral blood of TOX-deficient mice were comparable, Tox^f/fPdcd1^Cre mice showed an increase in infiltrates in the liver, with obvious conventional T-cell infiltrates (Fig. 4A and Supplementary Fig. S5A). Notably, the γδ T cell population was skewed toward the Tγδ17 subset in Tox^f/fPdcd1^Cre mice (Fig. 4B).

Fig. 4.

TOX-associated AIH is attributed to selective accumulation of Tγδ17 cells in the liver. A Frequency of γδ T cells gated on T cells in the spleen, peripheral blood, and liver of the indicated 4–5-week-old mice (n = 6). PBMCs peripheral blood mononuclear cells, LILs liver-infiltrating lymphocytes. B CCR6 and CD27 expression and cytokine expression in the indicated liver-infiltrating γδ T cells from 4–5-week-old mice (n = 6). C Survival analysis of mice (2 weeks old) treated with αIL-17 or IgG2a up to 32 weeks of age (n = 6). D Cytokine expression in liver-infiltrating γδ T cells from the indicated mice (2 weeks old) treated with αIL-17 or IgG2a for 6 weeks (n = 6). The data are presented as mean ± SD. NS, P > 0.05; *P < 0.05; **P < 0.01; Kruskal–Wallis test (A, B, D); Mantel‒Cox log-rank test (C)

To confirm the role of Tγδ17 cells in AIH pathogenesis, mice were administered an IL-17-neutralizing Ab and exhibited improved survival and reduced inflammatory infiltrates in the liver (Fig. 4C and Supplementary Fig. S5B). Compared with control IgG, the anti-IL-17 antibody decreased the frequency of Tγδ17 cells infiltrating the liver in both Tox^f/fCd4^Cre and Tox^f/fPdcd1^Cre mice, possibly suggesting a reduction in Tγδ17 recruitment mediated by IL-17 blockade in the liver microenvironment (Fig. 4D). To investigate whether chemotaxis of γδ T cells is a response to selective pathogenesis in the liver, the expression of the CCR6 ligand CCL20 in multiple organs was evaluated. CCL20 expression was selectively elevated in the livers of Tox^f/fPdcd1^Cre mice (Supplementary Fig. S5C, D). These data suggest that CCR6⁺ Tγδ17 cells are recruited to the liver via CCL20 and play an essential role in AIH pathogenesis.

Transfer of γδ T cells from Tox^f/fPdcd1^Cre mice reproduces AIH in Rag1^–/– mice

The T and B cells infiltrating the livers of Tox^f/fCd4^Cre and Tox^f/fPdcd1^Cre mice showed abnormal activation; thus, we first assessed whether the transfer of splenocytes into Rag1^–/– hosts can reproduce AIH. The recipient mice developed hepatitis, and inflammatory cells infiltrated the portal area and invaded the parenchyma 5 weeks after the transfer of Tox^f/fPdcd1^Cre splenocytes (Fig. 5A–C). The severity of chronic hepatic inflammation and tissue damage was also indicated by the increased collagen deposition surrounding the hepatic vessels and elevated ALT and AST levels, respectively (Fig. 5D, E).

Fig. 5.

Transfer of γδ T cells from Tox^f/fPdcd1^Cre mice reproduces AIH in Rag1^–/– mice. A, B Representative images of morphology, H&E staining, mIHC staining for TCR-β (red) and TCRγδ (green), and picrosirius red staining in the liver (A) and hepatic inflammation scores (B) in 5-week-old Rag1^–/– mice 5 weeks after adoptive transfer of splenocytes from the indicated 4-week-old mice (n = 10). Scale bars, 1 cm (morphology), and 50 µm (staining). C, D Quantification of the indicated cells (C) and collagen deposition (D) in the liver from recipients treated as described in (A) (n = 6). E The serum levels of ALT and AST in recipients (n = 6). F Schematic diagram of adoptive transfer from 4-week-old mice to Rag1^–/– mice, along with representative images of H&E staining, mIHC staining for TCR-β (red), and TCRγδ (green), and picrosirius red staining in the liver 5 weeks after the indicated transfer. Scale bar, 50 µm. The data are presented as mean ± SD; NS, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; Kruskal–Wallis test (B–E)

GC formation mediated by CD40 signaling correlates with the autoimmune response [39], and γδ T cells have been reported to facilitate B-cell differentiation via CD40 signaling [40]. γδ T cells from Tox^f/fPdcd1^Cre mice showed an elevated level of CD40L compared with those from Tox^f/fCd4^Cre and control mice (Supplementary Fig. S6A), which might account for the differential autoAb production among mice of the three genotypes. To confirm the effect of γδ T cells on CD40L expression, TOX-deficient γδ T cells were transduced with a lentivirus for CD40L disruption and were then transferred with WT B cells into Rag1^–/– recipients. Compared to control cells (sgCD40L-GFP^-), CD40L-deficient (sgCD40L-GFP⁺) γδ T cells decreased the GC B-cell proportion and ANA production to the levels observed in control mice (Supplementary Fig. S6B, C).

As mentioned above, the frequency of Tregs was decreased in both the thymus and periphery. To exclude the effect of Tregs, we transferred pooled inocula containing WT CD4⁺ T cells and the indicated CD8⁺ and γδ T cells at orthotopic ratios into Rag1^–/– recipients and analyzed the mice 5 weeks after reconstitution. Compared with mice that received control γδ T cells and mice that received Tox^f/fCd4^Cre γδ T cells, mice that received Tox^f/fPdcd1^Cre γδ T cells exhibited more obvious liver inflammatory damage, lymphocyte infiltration, and collagen deposition (Fig. 5F and Supplementary Fig. S6D–F). The frequency of T_FH cells was unchanged by the transfer of TOX-deficient γδ T cells (Supplementary Fig. S6G). TOX deficiency in CD8⁺ T cells did not influence the severity of AIH (Supplementary Fig. S6D–G). These data suggest that TOX is involved in AIH induction by regulating both Tγδ17 fate and humoral immunity.

TOX prevents lysosomal degradation of TCF1 in thymic DN2 progenitors to regulate γδ T-cell commitment

To further understand the underlying mechanism by which TOX regulates thymic lineage commitment, DN thymocytes from Tox^f/f and Tox^f/fPdcd1^Cre mice were subjected to RNA sequencing (Supplementary Fig. S7A). Loss of TOX affected the pathways associated with autoimmune diseases and resulted in enrichment of the IL-17 signaling pathway (Fig. 6A, B). The increase in Tγδ17 cells among DN thymocytes and lack of CD4⁺ thymocytes induced by TOX deficiency resembled the phenotype of mice lacking TCF1 [18, 41], suggesting the requirement of TOX for TCF1 activity. Unexpectedly, the level and stability of Tcf7 mRNA were comparable between control and TOX-deficient DN2 thymocytes (Fig. 6C and Supplementary Fig. S7B). However, TOX affected the TCF1-associated pathways, and the TCF1 protein level was reduced in Tox^f/fPdcd1^Cre DN2 thymocytes, indicating that TOX regulates TCF1 expression at the posttranscriptional level (Fig. 6D, E). Mass spectrometry following co-IP with an anti-TOX antibody in DN2 thymocytes indicated the binding of TOX to TCF1 during thymic development (Fig. 6F). Exogenous and endogenous TCF1/TOX complex formation and colocalization were observed (Fig. 6G–I). Truncation mapping showed that the N-terminal domain of TCF1 is required for its interaction with TOX and that the NHP6B domain of TOX is required for its interaction with TCF1 (Fig. 6J, K). To identify the protein degradation pathway of TCF1, we treated TOX-deficient DN2 progenitors with inhibitors of autophagosome formation (3-MA), proteasomal degradation (MG-132), and lysosomal degradation (BafA1). Treatment with BafA1 largely restored the TCF1 protein level, suggesting its predominant degradation by lysosomes (Fig. 6L, M). These data suggest that TOX stabilizes the TCF1 protein by preventing its lysosomal degradation in the complex.

Fig. 6.

TOX prevents lysosomal degradation of TCF1 in thymic DN2 progenitors to regulate γδ T-cell lineage commitment. A, B Top 20 enriched KEGG terms (A) and enrichment of IL-17 signaling-associated genes as demonstrated by GSEA (B) in the differential transcriptome profiles from the indicated DN2 thymocytes. C Tcf7 mRNA levels in the indicated DN2 thymocytes (n = 6). D GSVA showed enrichment of TCF1-associated genes in the differential transcriptome. E Immunoblot analysis of TCF1 protein expression in the indicated DN2 thymocytes (n = 6). F Mass spectrometry analysis showed that TOX interacted with TCF1 in DN2 thymocytes. G His-TCF1 co-IP products in HEK-293T cells were analyzed by immunoblotting after transfection of the indicated plasmids. H, I Co-IP (H) of endogenous TCF1 and TOX and localization (I) of TOX (green) and TCF1 (red) in DN2 thymocytes (n = 3). Scale bar, 5 μm. J, K Domain mapping analysis in HEK-293T cells transfected with full-length FLAG-TOX plus truncations of His-TCF1 (J) or with full-length His-TCF1 plus truncations of FLAG-TOX (K). Diagrams of the plasmid construction and co-IP analysis are shown. L, M Immunoblot analysis of TCF1 protein expression in the indicated DN2 thymocytes treated overnight with MG-132 (5 μM), 3-MA (5 mM), or BafA1 (1 μM) (L) or with MG-132 or BafA1 in the presence or absence of CHX (50 μg/mL) (M) (n = 3). N Cytokine expression in the indicated eGFP^– and eGFP⁺ γδ thymocytes from IT-reconstituted mice (n = 6). The data are presented as mean ± SD; NS, P > 0.05; *P < 0.05; **P < 0.01; Mann–Whitney U test (A, E); Kruskal–Wallis test (L–N). The data in the figure are from 3-week-old mice

To confirm the mechanism by which TOX regulates the early development of thymocytes, we subjected mice to intrathymic (IT) injection of lentiviruses expressing both TCF1 and eGFP. IT injection resulted in a certain percentage of eGFP⁺ cells in the DN2 populations (Supplementary Fig. S7C). Analysis of eGFP⁺TCRγδ⁺ cells in TOX-deficient mice showed that the percentages of the Tγδ17 and CD62L^hiCD44^lo subsets were changed to the levels observed in control mice (Fig. 6N and Supplementary Fig. S7D). Notably, compared with eGFP^– lymphocytes, their eGFP⁺ counterparts in peripheral blood showed an increased percentage of TCR-β⁺ cells and a reduced percentage of TCRγδ⁺ cells (Supplementary Fig. S7E). The developmental defect in CD4⁺ SP cells was ameliorated, with a higher frequency of CD62L^hiCD44^lo subsets among peripheral eGFP⁺TCR-β⁺ subsets (Supplementary Fig. S7F, G). These results suggest that TOX normalizes αβ/γδ lineage commitment by stabilizing TCF1.

γδ T cells with low expression of TOX are enriched in AIH patients and predict diagnostic accuracy

To determine the clinical relevance of TOX, we analyzed the characteristics of γδ T cells in a cohort of AIH patients. Although the number of total T cells was comparable (Fig. 7A), elevated percentages and numbers of γδ T cells were observed in the peripheral blood of AIH patients (Fig. 7B, C). Notably, γδ T cells from AIH patients produced higher levels of IL-17A and RORγt but lower levels of IFN-γ and Eomes (Fig. 7D–G). AIH patients showed lower TOX levels in peripheral γδ T cells (Fig. 7H). In addition, the TOX levels in peripheral γδ T cells were inversely correlated with the proportions of IL-17A⁺ and RORγt⁺ subsets (Fig. 7I, J). We further analyzed alterations in TCRδ chain usage in γδ T cells. AIH patients showed a higher Vδ1⁺ cell proportion than controls (Supplementary Fig. S7H). IL-17A⁺ γδ T cells were mostly restricted to the Vδ1⁺ subset (Supplementary Fig. S7I), and the Vδ2⁺ subset showed much higher TOX levels than the Vδ1⁺ subset (Supplementary Fig. S7J). TOX expression was inversely correlated with the AIH diagnostic score (Fig. 7K, L). In addition, Tγδ17 but not Th17 cells were enriched in AIH lesions (Fig. 7M and Supplementary Fig. S7K), and there was little colocalization of TOX and IL-17A (Supplementary Fig. S7L). Higher expression of CCL20 was observed in the livers of AIH patients than in those of controls (Supplementary Fig. S7M). These results indicate that TOX expression in γδ T cells is associated with the Tγδ17 frequency, inflammatory state, and clinical diagnosis of AIH.

Fig. 7.

γδ T cells with low expression of TOX are enriched in AIH patients and predict diagnostic accuracy. A Frequency of CD3⁺ T cells in peripheral blood from healthy controls and AIH patients (n = 10). B, C Frequency and absolute number of γδ T cells in peripheral blood (n = 10). D–G Cytokine expression and TF expression in the indicated γδ T cells in peripheral blood (n = 10). H MFI of TOX in the indicated γδ T cells in peripheral blood (n = 10). I, J The correlation between the TOX level and intracellular IL-17A production (I) or intranuclear RORγt expression (J) in γδ T cells from peripheral blood (n = 20). K Individuals with lower levels of TOX in peripheral γδ T cells showed higher AIH diagnostic scores (n = 10). L The correlation between the TOX level in γδ T cells and the AIH diagnostic score was determined by Spearman correlation analysis (n = 20). M Representative mIHC staining for TCRγδ (green), IFN-γ (red), and IL-17A (cyan) and the proportions of Tγδ17 cells in the liver from healthy controls and AIH patients (n = 20). The white arrows indicate Tγδ17 cells. Scale bar, 10 µm. The data are presented as mean ± SD; NS, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; Mann–Whitney U test (A, C, E, G, H, K, M)

Discussion

Immune tolerance requires an appropriate program of lineage commitment and negative selection. Herein, we propose a dual-stage role of TOX in Tγδ17 differentiation and subsequent AIH development. Unlike in Tox^f/fPdcd1^Cre mice, in which TOX is absent after the DN stage, in Tox^f/fCd4^Cre mice, the mTEC-mediated Tγδ17 commitment bias is attributed solely to blockade of CD4⁺ commitment by TOX deficiency after the DP stage, and early thymic development is intact. Upon TOX loss after the DN stage, signatures impacting both mTEC and DN subsets more broadly perturb thymic immune homeostasis and increase the severity of AIH (Supplementary Fig. S8). Thus, we hypothesize that TOX is a cross-cutting factor driving Tγδ17-dependent AIH. In the absence of specific biomarkers, the diagnosis of AIH remains a challenging problem. AIH must be considered in the differential diagnosis of all adult patients with liver diseases presenting similar clinical, biochemical, serological, and histological features [1]. A specific subset of memory CD4⁺ and CD8⁺ T cells associated with active AIH has been identified [42]. However, the contributions of other participants have remained elusive. Our data suggest that γδ T cells play crucial roles in AIH-associated inflammation and that TOX levels in peripheral γδ T cells can facilitate the accurate diagnosis of AIH.

The γδ T lineage, which is distinct from conventional T cells primarily in its activation mode and response speed, has been appreciated for its contribution to the production of inflammatory cytokines such as IFN-γ and IL-17A during various autoimmune responses. Previous studies have indicated that IFN-γ signaling is tolerogenic for autoimmunity, whereas IL-17A plays a pathogenic role [43, 44]. Inflammation induced by Tγδ17-cell expansion in TOX-deficient mice is the essential pathological mechanism underlying the development of fatal AIH. Although it has been reported that T cells play the predominant role in AIH induction, the types of effector cells involved and the mechanisms by which dysregulated T cells trigger AIH remain largely unclear. γδ T cells have been indicated to play supportive or modulatory roles in several models of induced or spontaneous autoimmunity [45, 46]. Although AIH patients exhibit elevated numbers of γδ T cells in peripheral blood and liver tissues [47], no study has investigated the role and unique characteristics of this subset. Interestingly, CCL20, the cognate ligand of CCR6, was selectively expressed during hepatic inflammation, supporting the findings of a previous study [48]. This explains the Tγδ17-cell enrichment and localized lesions in liver tissues. However, it is not clear why elevated CCL20 expression is limited to the liver, but this specificity may be associated with communication between hepatocytes and tissue-resident macrophages in the inflammatory environment [49]. In addition, γδ T cells initiate the T_FH program by releasing Wnt ligands and involve the GC reaction-associated interplay of costimulatory molecules to regulate self-reactive humoral immunity [50]. Our results showed that CD40L in γδ T cells is the key molecule participating in the formation of GC B cells and the overproduction of autoAbs. Blunted expansion of tTregs was observed in TOX-deficient mice, supporting previous reports that RANK signaling dependent on Aire⁺ mTECs drives the development of tTregs [4]. Furthermore, adoptive transfer experiments elucidated the dominant role of Tγδ17 cells rather than Tregs in AIH pathogenesis.

TOX is a critical factor for mTEC development and is associated with an impact on CD4⁺ thymocyte production. TOX reportedly plays an essential role in thymocyte selection and CD8⁺ T-cell induction in EAE [9, 51]. Of note, recent studies have attracted attention to CD8⁺ T-cell exhaustion via TOX in the context of both cancer and chronic infection [6, 52]. However, the role of TOX in the pathogenesis of AIH has not been elucidated. It is increasingly appreciated that the cellularity of mature mTECs is essential for the establishment of central tolerance, while disruption of the medullary architecture and defective TRA expression result in autoimmunity [3]. RANKL signals from lymphoid tissue inducer cells and dendritic epidermal T-cell progenitors have been confirmed to direct mTEC maturation during embryogenesis [53, 54], and RANKL-expressing CD4⁺ thymocytes govern the development of Aire⁺ mTECs via receptor crosstalk in the adult thymus. Although defects in TOX did not alter the frequency of RANKL⁺ CD4⁺ thymocytes, an obvious decline in the CD4⁺ thymocyte number was observed. Therefore, TOX influences mTEC development and subsequent autoimmunity. Our work expands on the critical roles of TOX during the stage of central tolerance establishment.

TOX regulates γδ T-cell specification by preventing lysosomal degradation of TCF1. The γδ TCR signal strength plays a role in driving the adoption of the Tγδ1 or Tγδ17 fate [15]. However, discrete transcription factors, including SOX13, independently control the fate of distinct γδ T subsets [55]. TCF1 is linked to skewing the Tγδ17 effector fate by docking at Rorc loci and to promoting the CD4⁺ T-cell fate by acting upstream of ThPOK [18, 41]. In developing γδ T cells, c-Maf is reported to inhibit TCF1 from binding to Rorc to maintain Tγδ17 identity [17]. However, the mechanism by which TCF1 activity is regulated at the posttranscriptional level remains unknown. This study suggests that TOX interacts with and stabilizes the TCF1 protein, thereby maintaining the balance of γδ T-cell commitment. In addition, TCF1 limits the formation of IL-17A-producing CD8⁺ T cells via repression of the MAF-RORγt axis [56]. Thus, it is possible that TOX and TCF1 function as conserved regulatory axes in type 17 specialization. On the other hand, context-dependent TOX functions may be governed by differential epigenetic programming, as we found that the subset-regulatory role of TOX is limited to γδ T cells and does not influence CD4⁺ T cells. Further studies are needed to delineate the unknown mechanism from the perspectives of epigenetic remodeling and transcriptional programs.

Collectively, our data extend the understanding of the function of TOX in γδ T-cell fate specification, which has been proposed to be associated with thymocyte selection and CD8⁺ T-cell exhaustion. Our work suggests that TOX in thymocyte progenitors stabilizes the TCF1 protein, thereby regulating central tolerance, Tγδ17 lineage specification, and subsequent AIH development. We provide molecular insights into how the roles of TOX in mTEC maturation and Tγδ17 differentiation constitute central mechanisms integrating immune tolerance in AIH.

Author contributions

QH, YLu, WT, RJ, and BS designed the experiments. QH, YLu, WT, RJ, WY, YLiu, MS, FW, HZ, and NW performed the experiments, provided reagents, and/or analyzed the data. QH, YLu, WT, and ZD discussed and interpreted the data. QH and BS wrote and revised the paper. BS conceived, supervised, and revised the study.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-022-00912-y.