Sex was not a variable considered in this study. The human primary data were pooled from both sexes. Our study also examined splenocytes from male and female animals; similar findings were reported for both sexes.

T cells were isolated from the spleen of TCR Vγ9Vδ2 tg and recombination-activating gene two (RAG-2)-deficient mice that express solely the human TCR Vγ9Vδ2 complexed with mouse CD3 (21). Mice were kept and bred at the animal facility of the Department of Biomedicine of the University Hospital Basel. Spleens were collected, and CD3⁺ cells were enriched using the MojoSort Mouse CD3 T cell isolation kit (Biolegend, #480023) according to manufacturer instructions. T cells were rested overnight in complete medium containing 10% FCS and used for activation assays.

Animal research was conducted under license 35328-2093, approved by the Authority of the Canton Basel-City.

The following tumor cell lines were purchased from the American Type Culture Collection (ATCC): A375 (human melanoma, CRL-1619, RRID: CVCL_0132) and HEK 293 T (human embryonic kidney, ACC 635, RRID: CVCL_0063). The following cell lines were previously generated in our laboratory: A375 β2m⁻, A375 β2m⁻ MR1, A375 β2m⁻ MR1 K43A (9, 35), J.RT3-T3.5, Jurkat-derived cells expressing a luciferase reporter gene under the control of NFAT, TCRαβ- and β2m-deficient (JKT) (36). Tumor cells were cultured in RPMI 1640 medium (Bioconcept, #1-41F01-I), 1mM sodium pyruvate (Bioconcept, #5-60F00-H), 1x non-essential amino acids (Bioconcept, #5-13K00-H), 1x stable glutamine (Bioconcept, #5-10K50-H), and 50 µg/ml of kanamycin (Bioconcept, #4-08F00-H) (complete medium) supplemented with 10% heat-inactivated fetal calf serum (FCS, BioConcept, #2-01F10-I). Cell lines were not authenticated and routinely confirmed to be absent of mycoplasma contamination by PCR.

A375 β2m knock-out (A375 β2m⁻) cells were used to generate by CRISPR/Cas9 a BTN3A1-deficient cell line. For this purpose, A375 β2m⁻ cells were transduced with Lenti Cas9-Blast plasmid (Addgene, #52962) and selected for 10 days with 10 µg/ml Blasticidin (Gibco, #A1113903) before being transduced with lentiGuide-Puro (Addgene, #52963) with gRNAs targeting BTN3A1 (5’-CCAGAGGTGGATCGCCGCCC-3’ and 5’-GGCACTTACGAGATGCATAC-3’). After 96 h of selection with 2 µg/ml Puromycin (InvivoGen, #ant-pr-1), cell lines were generated from single-cell clones by limiting dilution. DNA was extracted by each cell clone with a Macherey-Nagel Tissue DNA extraction kit (#740952.250) and BTN3A1 gRNA target locus was amplified with Q5 polymerase (NEB, #M0492S) and the following primers: BTN3A1_For 5’-TCCTCTGAGATTTTAGCATGAG-3’ and BTN3A1_Rev 5’-TGGCAATGACTAGGAATTGG-3’. PCR products were sequenced using the Sanger method, and effective BTN3A1 gene inactivation was confirmed by comparing the PCR product with the wild-type BTN3A1 sequence using Benchling [Biology Software (2023)].

PBMCs were isolated by density-gradient centrifugation using Lymphoprep (Stemcell, #07851). The used T cell clones were already established in our laboratory or generated from bulk lines, cultured, and expanded as described in (11, 21). Ex vivo TCR γδ cells were negatively enriched from PBMCs using EasySep™ Human Gamma/Delta T Cell Isolation Kit (Stemcell, #19255) and used for proliferation and activation assays. TCR Vγ9Vδ2 cells were FACS-sorted from PBMCs using anti-Vδ2 (Clone B6; Biolegend) and anti-Vγ9 (Clone B3; Biolegend) mAbs. T cells were cultured in a complete medium supplemented with 5% human AB Serum (Blood Donation Center of the University Hospital of Basel) and 100 U/ml recombinant human IL-2 (Peprotech, #200-02).

RNA from T cell clones was extracted with Nucleospin RNA (Macherey-Nagel, #740955.50), and cDNA was synthesized with SuperScript III reverse transcriptase (Invitrogen, #18080093). TCR Vγ9 and Vδ2 chain transcripts were amplified with gene-specific primers, followed by Sanger sequencing and respective analysis using ImMunoGeneTics (http://www.imgt.org). The transcripts were amplified using primers containing cloning adaptors and cloned into a lentiviral vector (Addgene, #52962) using the In-Fusion HD Cloning Kit (Takara, #639649). The vectors containing TCR Vγ9 or Vδ2 sequences and the lentivirus packaging plasmids pMD2.G (Addgene, #12259), pMDLg/pRRE (Addgene, #12251), pRSV-REV (Addgene, #12253) and the pAdVAntage (Promega, #E1711), were co-transfected into HEK 293 T LX cells using Metafectene PRO reagent (Biontex, #T040-1.0). JKT cells were transduced with lentiviral supernatants containing TCR Vγ9 and Vδ2 sequences. JKT cells expressing the different TCRs were sorted based on CD3 surface expression and used for functional assays.

A375 cells were tested for MR1 surface expression by flow cytometry. Briefly, cells were plated at 10⁵/well and incubated with 50 µM Zol (Sigma-Aldrich, # SML0223-50MG), 30 µM Ac-6-FP (Schircks Laboratory, #11.418), or vehicle (PBS) for 4 h at 37°C. After incubation, cells were first blocked for 15 min at 4°C for unspecific binding using 50% human AB serum, 0.1% BSA, 0.02% NaN₃ in PBS, and then stained with anti-human MR1 mAbs (clone 26.5; Biolegend, #361108) for 20 min at 4°C. Cells were washed and resuspended in PBS with 0.5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich, #MBD0015). Flow cytometry analysis was performed on a Cytoflex (Beckman Coulter).

A375 WT and A375 β2m⁻ MR1 cells were exposed to 30 µM Ac-6-FP, 50 µM Zol, or vehicle (PBS) for 4 h at 37°C in complete medium supplemented with 10% FCS. RNA was isolated from 10⁶ cells with NucleoSpin RNA Mini Kit (Macherey-Nagel, # 740955) following the manufacturer’s instructions, and 500 ng of RNA were used to synthesize cDNA with the PrimeScript Reverse Transcriptase kit (Takara, #2680Q). Quantitative PCR reactions were performed with Power Sybr green PCR master mix (Applied Biosystems, #4367659) using 15 ng of cDNA per reaction with the following primers for MR1 (forward: 5’-TGGCAGCAGATGTTCAAGGTGG-3’, reverse: 5’-GAAATCCTGTGGTGCTTCCATCC-3’) and using ribosomal protein S18 (RPS18) as housekeeping gene (forward 5’-GCAGAATCCACGCCAGTACAAG-3’, reverse 5’-GCTTGTTGTCCAGACCATTGGC-3’). MR1 relative expression was calculated using the ΔΔCt method (37).

Soluble recombinant human β2m-MR1 Fc-Streptag and β2m-MR1 K43A Fc-Streptag (an MR1 mutant form with lysine in position 43 mutated into an alanine) were prepared as described (11). Briefly, A375 cells secreting the recombinant proteins were cultured for 50 h, then the supernatant was collected. Soluble MR1 was quantified by ELISA as described (11) and stored at 4°C until further use to coat magnetic beads.

A second form of soluble MR1 protein was generated by assembling in the bacterial expression vector pET23d(+) DNA-Novagen (Sigma-Aldrich, #69748-M) a hybrid construct containing the nucleotide sequences coding for β2m, followed by a sequence encoding a (G₃S)₃ flexible linker and the soluble portion of MR1 (MR1*01 allele, GenBank accession number NM_001531). As described, the protein was produced in bacteria as inclusion bodies, refolded, and purified without ligand addition (11). This molecule was used in Fluorescence polarization assays. The same protein was also refolded in the presence of the following ligands: Ac-6-FP, (2E)-3-({9-[3,4-dihydroxy-5-(hydroxymethyl)oxolan-2-yl]-9H-purin-6-yl}amino)prop-2-enal (M₁Ado), (2E)-3-({9-[4-hydroxy-5-(hydroxymethyl)oxolan-2-yl]-9H-purin-6-yl}amino) prop-2-enal (M₁dA) and 3H,10H-pyrimido[1,2-a]purin-10-one (M₁Gua) as described (11). These MR1-bearing ligand monomers were used to generate MR1 tetramers.

MagStrep “type 3” XT Beads (IBA, #2-4090-010) were washed with PBS and incubated with 30 µg/ml of Streptag-soluble MR1 at 37.5 pg/bead for 18 h at 4°C under shaking conditions. Control beads were incubated with supernatant derived from A375 cells, which did not produce soluble MR1. This group is indicated as MR1-negative beads. Then, MR1-coated beads were washed with PBS and used for activation assays or TCR tetramer staining experiments.

Activation assays were done in 384-well plates, co-culturing human or mouse T cells (2 x 10⁴/well) with A375 (4 x 10⁴ cells/well). Human T cells were also cultured with MR1-coated beads (8 x 10⁴ cells/well) for 18 h. All assays were performed in a final volume of 80 μl. T cell activation was evaluated after 18 h by measuring human IFN-γ released in the supernatant by ELISA and measuring surface expression of CD69, CD25, and CD137 markers by FACS. In some cases, proliferation and activation assays were performed in the presence of anti-human MR1 mAbs (clone 26.5; Biolegend, #361102), anti-human HLA-A, B, C mAbs (clone W6/32; Biolegend, #311427), or IgG2a isotype-matched mAbs (clone MOPC-173; Biolegend, #400202), all at 30 µg/mL.

JKT activation assays were performed in an opaque 384-well plate (Corning, #3570) (3 x 10⁴ T cells/well and 3 x 10⁴ A375 cells/well) for 8 h.

APCs were treated with Zol (50 µM), Ac-6-FP (30 µM), IPP (10 µM) (Sigma-Aldrich, #I0503) or 5-OP-RU (40 nM) for 4 h before co-culture with T cells or JKT cells. In some cases, APCs were incubated with agonist anti-BTN3A mAbs (clone 20.1, Invitrogen, #14-2779-82) at 1 µg/mL for 1 h before adding JKT cells. Competition assays were performed by incubating APCs with Ac-6-FP (30 µM) for 2 h before Zol incubation. Luciferase production was assessed using the Bio-Glo kit (Promega, #G7940) according to the manufacturer’s protocol, and luminescence was measured using Sinergy H1 (Hybrid Reader, Biotek).

For intracellular cytokine measurement, T cells were treated with 2 µM Monensin (Biolegend, #420701) and 20 µg/ml Brefeldin A (Biolegend, #420601) before being challenged for 12 h with MR1-coated beads or PMA (50 ng/ml, Sigma-Aldrich, #P1585) and Ionomycin (500 ng/ml, Sigma-Aldrich, #I0634).

IFN-γ release was measured by ELISA as previously described (9): human IFN-γ (Biolegend, capture MD-1 mAb, #507502; revealing biotinylated 4S.B3 mAb, #502504), HRP Streptavidin (Biolegend, #405210), and recombinant human IFN-γ (Peprotech, #300-02).

D9A10 TCR soluble ectodomains were expressed in E. coli as inclusion bodies, folded in vitro, and purified as described below. The D9A10 Vδ and Vγ genes were fused to the ones of Cα and Cβ, respectively, for the production of soluble TCR γδ ectodomains (14, 38). As described, a disulfide bond was engineered between the Cα and Cβ chains (39) to facilitate the formation of a stable heterodimeric complex. The Vγ/Cβ chain was engineered with a C-terminal AviTag. The resulting chimeric TCR chains were codon optimized for E. coli expression and separately cloned into the vector pET23d(+) DNA-Novagen (Sigma-Aldrich, #69748-M). Inclusion bodies of both chains were produced in E. coli BL21(DE3) pLysS (Thermo Fisher, #C606010). Vδ/Cα and Vγ/Cβ inclusion bodies (15 mg, each) were solubilized at 10 mg/ml in 6 M guanidine HCl, 50 mM β-mercaptoethanol, 10 mM EDTA, 50 mM Tris pH 8.1 and heated at 50°C for 30 min, under shaking. Insoluble debris were removed by centrifugation at 20,000 RCF for 10 min. D9A10 soluble monomers were folded in vitro by rapid dilution of the solubilized inclusion bodies in 500 ml refolding buffer consisting of 5 M urea, 0.4 M L‐arginine, 100 mM Tris pH 8.1, 3.7 mM oxidized glutathione, 6.6 mM reduced glutathione to reach a final concentration of to 60 mg/l. Folded monomers were purified by two ion exchange chromatography steps with HiTrap DEAE FF (Cytiva, #17515401) and Mono Q™ 5/50 GL (Cytiva, #17516601) columns. Elution fractions were analyzed by SDS-PAGE using reducing and non-reducing conditions to confirm the disulfide bond formation between the two chains. The TCR-containing fractions were pooled and biotinylated in vitro with BirA biotin-protein ligase bulk reaction kit (Avidity, Bulk BirA) and purified by size exclusion chromatography using a Superdex 75 10/300 GL column (Cytiva, #17517401).

Biotinylated TCR and Ag-loaded MR1 monomers were tetramerized using PE-conjugated streptavidin (Prozyme, catalog no. PJRS25) at a 4:1 molar ratio. For each biotinylated monomer (5µg), we added 8µg of PE-conjugated streptavidin. PE-conjugated streptavidin was prepared at 200 µg/mL in PBS and 1.6 µg were added stepwise to the biotinylated monomer in cold and dark conditions every 20 min under shaking.

Details of mAbs used can be found in Supplementary Table S1 . All mAbs were titrated to identify optimal concentrations before their use in multicolor flow cytometry. Cells were stained with Fixable Viability Kits as Zombie NIR (Biolegend, #423106), Zombie AQUA (Biolegend, #423101), or LIVE/DEAD Fixable Blue (Thermo Fisher, #L23105) in PBS for 20 min at 4°C, followed by surface staining by adding specific mAbs for 20 min in PBS, 0.5% BSA, 0.02% NaN₃ at 4°C. When mentioned, a viability assessment was done by adding DAPI after cell surface staining.

For intracellular staining, stained cells were fixed with a Fixation Buffer (Biolegend, #420801) for 15 min at RT. Cells were permeabilized using Intracellular Staining Permeabilization Wash Buffer 1X (Biolegend, #421002) for 20 min at 4°C, followed by anti-human mAbs.

Cells were first gated based on forward scatter-area (FSC-A) and side scatter-area (SSC-A), followed by gating on viable cells. Singlets were gated using SSC-A/side scatter height (SSC-H) and FSC-A/forward scatter height (FSC-H). Cells were acquired by an Aurora spectral analyzer (Cytek) or CytoFLEX flow cytometer (Beckman Coulter) and analyzed using FlowJo v10 software (LLC). TCR Vδ2 cells were sorted using a FACSMelody Cell Sorter (BD Biosciences).

MR1-coated beads were stained with D9A10 TCR tetramer (5 µg/ml) for 20 min at room temperature in PBS, followed by the addition of anti-PE Abs from the PE-positive Selection Kit (Stemcell, #17684) for an additional 20 min. Beads were washed with PBS and acquired using a CytoFLEX flow cytometer (Beckman Coulter).

Soluble recombinant single-chain MR1 (80 nM) was incubated for 18 h at room temperature with 25 nM of the reporter fluorochrome JYM20 (Wuxi, #WX10-01-245) and Zol or Ac-6-FP at different concentrations in fluorescence polarization assay buffer (20 mM TRIS pH 8.5, 2 mM EDTA, 150 mM NaCl, 0.05% Tween 20). Samples in 96 well plates (Corning, #3686) were read using a Spark multimode microplate reader (Tecan; Excitation wavelength: 535nM. Emission wavelength 595nm. Temperature 25°C).

Flow cytometry data was exported from FlowJo v10.7.1 and imported in R v4.2.0 through the Bioconductor package flowCore v2.10.0 (40), and clustering analysis was performed. Marker expression levels were transformed using the inverse hyperbolic sine transformation (asinh function within R) with a cofactor of 150. Clusters were computed via the R implementation of Phenograph (41), considering the transformed expression values for 9 markers (TCR Vδ2, CD69, CD26, CD28, CD56, CD57, CD161, NKp80, CD95) and a k of 100. A heatmap of the average expression of each marker in each cluster was produced using pheatmap v1.0.12 (42), and the clusters were visualized as colors overlaid on a Uniform Manifold Approximation and Projection (UMAP) using the R implementation of UMAP (v0.2.10.0) (43).

Blood samples from healthy donors and the investigated patient were obtained from the University Hospital Basel. The research with healthy blood samples and the prospective cohort study of the functional and genetic architecture of patients with primary immune dysregulation (FuGe-PID) have been approved by the Ethics Committee North-West & Central Switzerland (EKNZ 2017-01888 and EKNZ215-187).

All donors consented in writing to analyze their samples.

Statistical analyses and data visualization were performed using GraphPad Prism v10 (GraphPad Software, Inc.). P values are indicated in the figure panels and legends.

The assembling of figure panels was done using Affinity Designer 2.5.3 (Serif Europe).

We systematically analyzed whether TCR γδ cells react to MR1 and are present in healthy donors. To exclude the contribution of molecules other than MR1 to this reactivity, we established a new assay using beads coated with soluble MR1, produced by the melanoma cell line A375, as stimulatory reagents ( Figure 1A ; Supplementary Figures 1A, B ). The purity of the MR1 bound to the beads was confirmed by SDS-PAGE ( Supplementary Figure 1C ). Control beads were incubated with supernatant derived from A375 cells that do not produce soluble MR1. The activation assays were performed using TCR γδ cells purified from peripheral blood ( Supplementary Figure 1D ) in the presence of MR1-coated beads without any cytokine or other cell types to exclude the participation of molecules besides MR1 in the activation. The MR1-reactive T cells were detected by co-staining with anti-CD69, anti-CD137, and anti-TCR Vδ mAbs ( Figure 1B ) in five healthy donors ( Figure 1C ). We found that rare TCR γδ cells were activated using these stringent conditions, with those expressing the TCR Vδ2 and Vδ1 chains being the most abundant, whereas MR1-reactive TCR Vδ3 cells were much less frequent or even absent ( Figures 1C, D ).

Among the MR1-reactive TCR Vδ2 cells, CD4 and CD8-double negative (DN) were reduced, and in some donors, CD4⁺ or CD8⁺ cells were increased ( Figures 1E, F ). MR1-reactive TCR Vδ1 cells showed reduced CD4-CD8 DN and increased CD8⁺ cells ( Figures 1G, H ). These findings revealed the existence of MR1 self-reactive TCR Vδ2 cells and confirmed the existence of MR1-interacting TCR Vδ1 and TCR Vδ3 cells (12, 14). We focused on Vγ9Vδ2 cells, which represent the majority of the circulating TCR γδ cells.

The TCR Vγ9Vδ2 cell reactivity to MR1 was validated by sorting and cloning the cells that upregulated CD137 upon stimulation with MR1-coated beads ( Figure 2A ). The isolated clones expressed Vγ9 and Vδ2 chains, as represented by the D9A10 clone ( Figure 2B ). This clone was CD4-CD8 DN ( Figure 2C ; Supplementary Figure 2A ) and released IFN-γ ( Figure 2D ) when challenged with MR1-coated beads ( Supplementary Figures 2B, C ). This reactivity was blocked by adding anti-MR1 antibodies and not isotype-matched ones ( Figure 2D ). In contrast, a control Vγ9Vδ2 clone (D15A3) was not MR1-reactive but was highly reactive to APC pulsed with Zol ( Figure 2E ). The response of D9A10 cells to MR1-coated beads was also observed using intracellular detection of TNF-α and IFN-γ ( Figure 2F ; Supplementary Figures 2D–F ). Instead, the D15A3 cells did not produce these cytokines ( Supplementary Figure 2E ). Both T cell clones produced TNF-α and IFN-γ upon stimulation with A375 cells pulsed with Zol ( Supplementary Figure 2F ).

To further address the interaction between MR1 and the D9A10 TCR, we generated a soluble tetramerized D9A10 TCR to stain MR1-coated beads ( Figure 2G ). The purity of the sTCR was confirmed by SDS-Page and size-exclusion chromatography ( Supplementary Figures 2G, H ). This reagent bound to MR1-coated beads but not the uncoated ones, and the binding was prevented using anti-MR1 antibodies but not isotype-matched ones ( Figure 2H ). Similar results were observed in at least three independent experiments performed using different preparations of soluble MR1 ( Figure 2I ).

Further experiments investigated whether D9A10 cells reacted to β2m-deficient APC lacking or expressing MR1. We previously described the generation of A375 cells deficient in β2m (A375 β2m−) and transduced with the MR1 gene (A375 β2m− MR1) encoding the MR1*01 allele (9). These cells are positive for MR1 but not stained with an antibody reacting to all HLA A, B, and C molecules ( Supplementary Figure 2I ). The D9A10 clone did not react to A375 β2m− and A375 β2m− MR1 cells ( Figure 2J ), although it responded to MR1-coated beads ( Figure 2D ). This discrepancy could be caused by a small number of MR1-Ag complexes expressed on APC compared to beads. When the APCs were pulsed with Zol, unexpectedly, the MR1-positive APCs were more stimulatory than the MR1-negative ones ( Figure 2J ). In contrast, the D15A3 cells, which are MR1 non-reactive, were equally stimulated by Zol-pulsed APCs regardless of MR1 expression ( Figure 2K ). These findings confirmed the existence of MR1-reactive Vγ9Vδ2 TCRs. Furthermore, the MR1-reactive TCRs could increase the response to Zol-treated APCs expressing MR1.

To study the relevance of MR1-autoreactive TCR Vγ9Vδ2 cells, we examined one patient with abnormal expansion of TCR γδ cells, who reported recurring infections of unknown origin, manifesting fever, chills, and swelling of inguinal and cervical lymph nodes. The patient suffers from atopy and clinically relevant type I sensitization to house dust mites, tree, and grass pollen; however, at hospital admission, the patient had normal IgE antibody levels ( Supplementary Table 2 ). The hemogram reported no alterations in leukocyte absolute numbers or other types of circulating blood cells ( Supplementary Table 3 ), except for imbalanced T cell populations, with abnormally increased numbers of DN T cells and low CD4⁺ T cells ( Supplementary Table 4 ). In the past five years, the patient suffered sudden increases in absolute lymphocyte numbers, particularly of DN T cells, during symptomatic exacerbations of infections ( Supplementary Table 4 ). Flow cytometry studies were performed to investigate whether the lymphocytosis and the increased numbers of DN cells were associated with the expansion of unconventional T cells. An increased frequency of TCR Vδ2 cells was detected, which comprised ~43.4% of total T cells in the blood, whereas TCR αβ cells were ~53.6% ( Figure 3A ). TCR Vδ1 cells were ~0.4% of total T cells, and TCR γδ cells expressing other TCR Vδ chains were <2.2% of total T cells ( Figure 3A ). These findings were confirmed in three independent analyses ( Figure 3B ). Among circulating T cells, CD4⁺ cells were 30.9%, CD8⁺ cells were 25.6%, and DN cells were 43.3%. TCR Vδ2-expressing cells were mostly DN (93.1% of total TCR γδ cells), and only 6.7% expressed dull CD8 levels ( Figure 3C ).

We performed a multicolor flow cytometry analysis of 15 surface markers on the patient’s peripheral blood mononuclear cells (PBMCs) to assess the functional maturation of circulating T cells. Two healthy donors were investigated in parallel as controls (HC). Almost all TCR Vδ2 cells from the patient and the HCs were Ag-experienced. In all three donors, ~60.0% of cells showed the phenotype of T effector memory (T_EM) cells, and ~40.0% had a T effector terminally differentiated (T_EMRA) profile ( Figure 3D ). The cells from the patient and HCs expressed high levels of KLRG1, further emphasizing that these cells were similarly differentiated. Other markers, including CD161 and PD-1, were also equally expressed, whereas TIGIT was not expressed ( Figure 3E ).

The activation markers, CD26 and CD69, the Fas receptor CD95, and the co-stimulatory molecule CD28 were reduced in the patient compared to the HCs, whereas the NK receptors NKp80 and CD56 were increased ( Figure 3E ). Cluster analysis considering differently expressed markers showed the main dissimilarities between TCR Vδ2 cells from the patient and HCs ( Figure 3F ). Four cell populations (clusters 6, 9, 10, and 15) were more frequent in the patient sample than in controls ( Supplementary Figure 3A ). The cells in these clusters were enriched in those co-expressing NKp80, CD161, CD56; markers shared with NK cells (44); and CD26. In contrast, clusters 1, 2, 3, 4, 7, 8, and 13 were more frequent in HCs. These latter clusters expressed reduced or absent NKp80, CD161, and CD56. They were variably positive for CD26 and CD95 ( Supplementary Figure 3A ), indicating the presence of cell populations with different degrees of differentiation after stimulation.

We next investigated the frequency of T cells in the patient that were MR1-reactive. The T cells freshly isolated from patient-derived PBMCs stimulated with MR1-coated beads, as previously described ( Figure 1A ), upregulated the activation markers CD69 and CD137 in ~25.2% of the Vδ2 cells ( Figure 3G ). In control cultures with T cells alone and T cells cultured with uncoated beads, ~3.0% and 5.8% of cells showed background activations. In contrast, TCR αβ cells did not upregulate CD69 or CD137 under the same conditions ( Figure 3G ). Thus, a significant fraction of TCR Vδ2 cells were MR1-autoreactive.

To determine the functional phenotype of the MR1-reactive TCR Vδ2 cells, we measured the intracellular cytokine profile of a patient-derived TCR Vδ2 cell line upon activation. PMA/Ionomycin stimulation revealed the production of mainly Th1 cytokines (IFN-γ, TNF-α, IL-2, and GM-CSF) and IL-13 but not IL-4, IL-10, or IL-17A ( Figure 3H ; Supplementary Figure 3B ). When T cells were incubated with MR1-coated beads, 4.4% produced IFN-γ and 4.5% TNF-α. Other cytokines were undetected ( Figure 3H ; Supplementary Figure 3B ). As a control, the same TCR Vδ2 cells incubated with uncoated beads did not produce cytokines ( Figure 3H ; Supplementary Figure 3B ). Thus, the MR1-reactive TCR Vδ2 cells derived from this patient produced type 1 cytokines and showed a Th1 pro-inflammatory functional profile.

Next, we studied the MR1 recognition mechanism of TCR Vγ9Vδ2 cells and examined the relevance of MR1, TCR, BTN3A1, Ag, and the Vδ2 chain.

We next studied whether MR1 improves the response of other TCR Vγ9δ2 cells to Zol-treated cells, as observed with the D9A10 clone ( Figure 2J ). We investigated the activation of six TCR Vγ9Vδ2 clones using A375 cells pulsed with Zol in the presence or absence of anti-MR1 mAbs. The addition of anti-MR1 mAbs significantly reduced the amounts of IFN-γ released by the clones D9A10, D1C55, and D1B4 ( Figure 4A ). The effect of anti-MR1 mAbs was negligible for the response of the clones G2B9, D1B5, and D15A3 ( Figure 4A ). The presence of anti-HLA-A, B, and C mAbs, instead, did not have any effect, except for the D1B4 clone ( Figure 4A ). The measurement of antigen-presenting molecule expression on A375 wild-type cells denoted high levels of HLA-A, B, and C molecules, contrasting with the dull expression of MR1 ( Supplementary Figure 4A ), thus showing a strong effect of MR1 despite the low physiological expression.

To further investigate the role of MR1 in facilitating the activation of Vγ9Vδ2 cells, the two additional MR1-enhanced TCR Vγ9Vδ2 cell clones D1C55 and D1B4 were stimulated using A375 β2m⁻ and A375 β2m⁻ MR1 cells, as previously done with the D9A10 clone ( Figure 2J ). Zol-treated APCs expressing MR1 induced an increased response from both clones compared to A375 β2m⁻ cells ( Figure 4B ). Untreated APCs did not stimulate the tested clones. Thus, like D9A10, these T cells are activated more efficiently in the presence of MR1.

To test whether the TCRs Vγ9Vδ2 are involved in the MR1-induced activation, four TCR Vγ9Vδ2 pairs and one MAIT TCR pair were reconstituted in β2m-deficient and TCR-deficient Jurkat (JKT) cells ( Supplementary Figure 4B ). Each transduced JKT cell line was challenged with A375 β2m⁻ cells and A375 β2m⁻ MR1. JKT-D9A10, JKT-D1C55, and JKT-G2B9 showed higher responses to Zol-exposed APCs expressing MR1, while they were non-reactive or showed no different reactivity to vehicle-exposed APC ( Figure 4C ). In contrast, JKT cells expressing the MR1 non-reactive D15A3 TCR did not exhibit differential reactivities to MR1-expressing APCs with or without Zol ( Figure 4D ), confirming the observations with the T cell clones. As an additional control, JKT cells expressing a MAIT TCR (JKT-SMC3), which recognizes MR1-5-OP-RU complexes (45), did not respond to any APCs with or without Zol ( Supplementary Figure 4C ).

The effects of MR1 recognition over Zol-induced activation were also assessed in Zol-titration experiments. The response of JKT-D9A10 cells at high Zol doses to MR1-positive APC was higher than that to MR1-negative APC ( Figure 4E ), directly affecting the efficacy of the response. JKT cells transduced with the non-MR1 reactive TCR D15A3 did not show such an increment ( Figure 4E ). Also, the TCR Vγ9Vδ2 cells derived from the reported patient showed an increased response to MR1-expressing APC. Indeed, up to 36% of TCR Vγ9Vδ2 cells upregulated CD25 and CD137 markers upon challenge with A375 β2m⁻ MR1 cells compared to the 16% background activation observed in co-cultures with MR1-negative A375 cells ( Figure 4F ), confirming the presence of MR1-self-reactive cells.

We further assessed the TCR-dependent recognition of MR1, taking advantage of RAG-2-deficient transgenic (tg) mice expressing the human TCR of the D1C55 clone (21). The tg T cells do not express BTN3-like genes but react in a BTN3A1-dependent manner to pAgs (21). When tg TCR Vγ9Vδ2 cells were stimulated with A375 β2m⁻ cells in the presence of Zol, up to 7.3% of cells upregulated the expression of the activation markers CD69 and CD137 ( Figure 4G ). Upon challenge with A375 β2m⁻ MR1 cells, they showed a marked upregulation of both markers, and the double-positive cells were >30% ( Figure 4G ). In four independent experiments, the response to MR1-expressing APC was significantly higher than that to MR1-negative ones (60% vs. 40%) ( Figure 4H ). Hence, the presence of MR1 was also relevant in this system.

These findings indicated that the TCR Vγ9Vδ2 transfer recapitulates the response to MR1 and confirmed that only some TCRs Vγ9Vδ2 show this capacity. Furthermore, MR1 recognition enhanced the functional response upon TCR engagement.

>These results raised two hypotheses: the first is that the Zol-induced pAg accumulation and BTN conformational changes are necessary for the additive effects of MR1 recognition. A second possibility is that Zol treatment induces other cellular changes, which promote the response to MR1 independently of BTNs. Instead of Zol, we performed activation assays with the 20.1 monoclonal antibody, which induces a BTN conformational change and consequent T cell activation (20, 46). This exogenous treatment was performed using A375 β2m− and A375 β2m− MR1 cells to stimulate three JKT cells expressing the three MR1-reacting TCRs D9A10, D1C55, and G2B9 ( Figure 5A ). Under these conditions, there was no increased response to MR1-expressing APCs, thus pointing to Zol-induced changes independent of BTN3A.

We further dissected the role of BTN3A1 by inactivating this gene in A375 β2m⁻ cells ( Supplementary Figures 5A–C ). These cells were then transduced with the MR1-β2m linked gene construct. As expected, the BTN3A1 gene inactivation abolished the response of D9A10, D1C55 ( Figure 5B ), and D15A3 ( Figure 5C ) clones to Zol-exposed A375 β2m⁻ cells. However, upon Zol exposure, A375 β2m⁻/BTN3A1⁻ cells expressing MR1 activated D9A10 and D1C55 clones ( Figure 5B ). The clone D15A3, which is not sensitive to MR1, did not react to this stimulation ( Figure 5C ). The same experiment was also performed using TCR-transduced JKT cells. JKT cells expressing the D9A10, D1C55, and G2B9 TCRs also responded to Zol-exposed A375 β2m⁻/BTN3A1⁻ cells expressing MR1 ( Figure 5D ), whereas JKT-D15A3 did not react in any of the described conditions ( Figure 5E ).

We then tested the BTN3A1-dependence of the patient’s TCR Vγ9Vδ2 cells and found that when stimulated with BTN3A1-deficient cells, 17.6% of the T cells were activated without MR1 and up to 24% in the presence of MR1 ( Figure 5F ).

In conclusion, these findings showed that MR1 stimulation of the tested TCRs is BTN3A1-independent and suggested that Zol exposure could induce metabolic changes related to the accumulation of MR1-presented antigens.

The effects of Zol could be associated with being an MR1 ligand or with the possibility that it promotes the accumulation of compounds stabilizing MR1. This indirect effect might increase MR1 surface expression, thus activating specific TCR Vγ9Vδ2 cells.

Using a fluorescence polarization-based assay, we disproved the hypothesis that Zol is an MR1-presented Ag. Competition assays were performed with various doses of Zol as a competitor to prevent the binding of JYM20 and a fluorophore-tagged MR1 ligand (47). Zol did not compete even at doses as high as 1 mM. In contrast, Ac-6-FP, a control MR1 ligand, showed efficient competition at 1 µM ( Supplementary Figure 6A ). These findings suggested that Zol does not bind to the JYM20 binding site of MR1.

The second hypothesis was investigated using different approaches. In the first one, A375 WT and A375 β2m⁻ MR1 cells were exposed to Zol for 4 h before evaluating MR1 protein surface levels. Zol treatment induced a 1.5-fold increase in A375 β2m⁻ MR1 compared to untreated cells, whereas MR1 upregulation on A375 WT cells was not statistically significant ( Figures 6A, B ). As a control, Ac-6-FP, a highly potent MR1 binder (48), induced a 1.2-fold MR1 increase on A375 WT and 7-fold on A375 β2m⁻ MR1 cells ( Figures 6A, B ). The MR1 upregulation was unrelated to increased transcription of MR1 since both Zol and Ac-6-FP did not change the relative expression of the MR1 gene compared to vehicle treatment in A375 WT and A375 β2m⁻ MR1 cells ( Supplementary Figure 6B ). Thus, Zol did not induce increased MR1 gene transcription and promoted MR1 upregulation through another mechanism.

In the second approach, we assessed if the selected TCRs Vγ9Vδ2 recognize well-known MR1-presented Ags, such as 5-OP-RU and Ac-6-FP ( Supplementary Figure 6C ). Zol-treated A375 β2m⁻ MR1 cells induced the activation of JKT-D9A10, JKT-D1C55, and JKT-G2B9 ( Supplementary Figure 6C ) but not of JKT-SMC3 cells that express a MAIT TCR ( Supplementary Figure 6D ). In contrast, the 5-OP-RU metabolite activated JKT cells expressing the MAIT SMC3 TCR, but none of the TCRs Vγ9Vδ2 ( Supplementary Figure 6D ). Ac-6-FP did not induce a response from any TCR-transduced JKT cells ( Supplementary Figures 6C, D ). We also tested whether MR1 tetramers loaded with different antigens could bind the MR1-reactive TCRs expressed by Jurkat cells. We tested tetramers loaded with Ac-6-FP (48), 5-OP-RU (7), M₁Ado, M₁dA, and M₁Gua (11), which bind MR1 and stimulate specific TCRs.

None of the tetramers stained these cells ( Supplementary Figure 6E ), whereas they efficiently bound Jurkat cells expressing the E8 TCR, which binds MR1 independently of the antigen (36). The same tetramers did not stain the D9A10 and D15A3 clones ( Supplementary Figure 6E ), thus resembling the results observed with JKT cells expressing the same TCRs. These results indicated that the MR1-reactive TCRs Vγ9Vδ2 do not recognize MAIT Ag, Ac-6-FP, or the selected carbonyl adducts, and the tested MAIT TCR does not react to APCs exposed to Zol.

In a third type of experiment, we investigated whether MR1 reactivity is Ag-dependent using Ag competition studies. JKT cells expressing four TCRs were challenged with A375 β2m⁻ MR1 cells incubated with Ac-6-FP before Zol exposure. Ac-6-FP forms stable complexes with MR1 and efficiently competes with the binding of other molecules (48). The Ac-6-FP pre-treatment of A375 β2m⁻ MR1 cells significantly reduced the response of JKT-D9A10, JKT-D1C55, and JKT-G2B9 cells ( Figure 6C ), indicating it was competing with the Ags presented by MR1. Ac-6-FP did not affect the responses of control JKT-D15A3 cells, which are not MR1-self-reactive ( Figure 6D ), and also did not affect the response to Zol-treated MR1-negative cells ( Figures 6E, F ). This data suggests that the MR1-self-reactivity of D9A10, D1C55, and G2B9 TCRs is Ag-dependent.

In a fourth approach, competition experiments were performed using Ac-6-FP, MR1-coated beads, and staining with TCR tetramers. Ac-6-FP treatment did not change the total MR1 on the beads ( Figure 6G ). However, it significantly reduced the binding of the TCR tetramer to MR1 ( Figures 6H, I ).

These findings confirmed that the selected TCR Vγ9Vδ2 binds MR1-Ag complexes, and Ac-6-FP prevented this interaction.

In a fifth approach, we tested T cell reactivity to MR1 mutated at lysine 43, located within the antigen-binding pocket of MR1 (7). Beads coated with MR1 WT or the MR1 K43A mutant were compared in activation experiments of the D9A10 and D15A3 clones. Surprisingly, the beads coated with MR1 K43A stimulated D9A10 cells very efficiently, as detected by IFN-γ release ( Figure 7A ) and CD137 marker upregulation ( Figure 7B ). In contrast, the MR1 non-reactive D15A3 cells were not stimulated ( Figures 7A, B ). Control experiments showed that MR1 WT and K43A were equally bound to the beads ( Supplementary Figure 7A ).

The bead-induced activation of D9A10 cells was entirely blocked by the addition of anti-MR1 antibodies, not by isotype-matched antibodies, in the IFN-γ release assay ( Figure 7C ) and intracellular staining for IFN-γ and TNF-α ( Figure 7D ). The same assays showed no stimulation by MR1 K43A on the D15A3 clone ( Supplementary Figures 7B C ).In further experiments, beads coated with MR1 K43A were stained with TCR tetramers, and this interaction was blocked by anti-MR1 antibodies, not by isotype-matched ones ( Figures 7E, F ), thus further validating MR1 reactivity with the D9A10 TCR.

The stimulatory activity was next investigated using A375 β2-m-deficient cells expressing the MR1 K43A mutant ( Supplementary Figure 7D ). These APCs were potent stimulators of D9A10 cells, and the anti-MR1 antibodies, not isotype-matched antibodies, blocked IFN-γ release ( Figure 7G ) and intracellular accumulation of IFN-γ and TNF-α ( Figure 7H ). The control D15A3 clone did not respond in these assays ( Supplementary Figures 7E, F ).

Finally, we tested whether JKT cells expressing the MR1-reactive TCRs D9A10, D1C55, and G2B9 also showed a response to MR1 K43A. The JKT-D9A10 confirmed the response to this MR1 mutant, which was blocked by adding anti-MR1 and not isotype-matched antibodies ( Figure 7I ). In contrast, the other two MR1-reactive TCR cell lines and the control D15A3 did not respond ( Figure 7I ).

Together, these experiments indicate that the antigens presented by MR1 are necessary for activating select TCR Vγ9Vδ2. Some antigens do not require the presence of lysine 43 in the antigen-binding pocket, resembling what was observed with some MR1 T cell clones (9).

TCR sequencing of the clones investigated in the study all showed the use of TRGV9*01-TRGJP*01 rearrangements and different CDR3 regions ( Figure 8A ). Instead, they showed TRDV2 genes rearranged to TRDJ3*01 or TRDJ1*01 and different CDR3 regions ( Figure 8B ). The clones D1B4 and D9B2 shared an identical TCR Vγ9 chain but expressed TCR Vδ2 chains with different CDR3 regions ( Figures 8A, B ). Only the D1B4 clone was MR1-self-reactive, raising the possibility that MR1 reactivity depended on CDR3δ. The reconstitution of both TCRs in JKT cells ( Supplementary Figure 8A ) confirmed the lack of reactivity of the D9B2 Vδ2 chain ( Figure 8C ). Due to the involvement of the TCR CDR3 in Ag interactions, we tested the Ag relevance of this recognition. Competition experiments with Ac-6-FP reduced the recognition of Zol-treated A375 β2m− MR1 APCs ( Figure 8D ). Instead, in control experiments, the recognition of A375 β2m− APC was not inhibited ( Figure 8E ). This data suggests that the TCR Vδ2 chain participates in MR1 reactivity and that its CDR3 region plays a relevant role.