γδ T-cell Receptors Derived from Breast Cancer–Infiltrating T Lymphocytes Mediate Antitumor Reactivity

Triple-negative breast cancer (TNBC) has a very poor prognosis compared with other breast cancer subtypes. Despite encouraging results in TNBC patients treated with checkpoint inhibitors (1), the majority of patients at advanced stages of the disease do not respond to therapies, and the substantial genomic heterogeneity of these tumors makes it difficult to identify new therapeutic targets (2–4). One novel opportunity for therapy arises from the observation that γδ T cells infiltrate various tumors, including TNBCs (5, 6), and these infiltrates appear to be prognostically beneficial (7). However, the number and function of γδ T cells are substantially diminished in advanced cancer patients (8, 9), suggesting that the possible γδ T-cell immunosurveillance activity in early cancers may become dysfunctional at the later stages of cancer. Some studies have found that IL17-producing γδ tumor-infiltrating lymphocytes (TIL) can create a tumor-promoting environment (10, 11). In contrast, another in-depth study suggests that γδ TILs are not the main producers of IL17 in TNBC and rather support antitumor reactivity in breast cancer through innate-like receptors (ref. 12; for review see refs. 13, 14). However, failures of polyclonal γδ T cells to induce regressions in advanced cancer patients have been reported (14, 15). To overcome this system failure in advanced cancer patients, we developed the concept of TEGs (T cells engineered to express a defined TCRγδ) for cancer therapies (16–18). For example, TEGs can distinguish between healthy and malignant hematopoietic stem cells by sensing altered lipid pathways present in malignant cells through TCRγ9δ2, and by detecting subtle spatial and conformational changes of CD277 (ref. 19; for review see ref. 14).

To identify tumor-reactive TCRγδs, we analyzed TCR sequences in γδ TILs from TNBCs. We also analyzed TCRδ repertoires in TIL sequencing data sets and in sequencing data sets of peripheral blood from healthy individuals. We combined single-cell sequencing (SCS) of laser microdissected γδ TILs, targeted high-throughput sequencing (HTS) of TCRγδ repertoires, and TCRγδ extraction from bulk tumor RNA sequencing (RNA-seq) data using the MiXCR RNA-seq mode (20). Subsequently, we utilized the TEG format for functional analysis of TCRγδs and observed tumor reactivity of matched TCRγδ pairs against a variety of tumor cells.

Peripheral blood samples from 13 anonymous healthy donors were obtained via the Dutch blood bank (Sanquin). From the archive of the Institute of Clinical Pathology, Freiburg, 16 formalin-fixed paraffin-embedded tissue specimens with the diagnosis of TNBC were selected after ethics approval by local authorities (Ethics Committee University Medical Center Freiburg) according to the Declaration of Helsinki, and written informed consent was obtained from each patient. The histopathologic diagnoses were established by two independent pathologists (P. Bronsert and A. Schmitt-Gräff) according to the Union for International Cancer Control criteria. All the tumors were grade III using the modified Bloom-Richardson classification (21). Conforming to the recommended evaluation of TILs (22, 23), the hematoxylin and eosin (H&E)–stained samples contained at least 50% tumor infiltration. Subsequently, 11 tumors (5 medullary breast carcinomas and 6 invasive ductal carcinomas) were selected for further analysis based on the pattern of infiltration and the availability of the corresponding frozen tissue samples in the tumor bank of the Comprehensive Cancer Center Freiburg. All samples came from female patients with a median age of 59 years (range, 43–82 years). All samples were classified as having a basal-like subtype based on the expression of CK 5/6 or 14 and EGFR (24). The use of the TNBC patient samples was approved by the ethics committee of the Medical Center University of Freiburg.

The following cell lines were obtained from ATCC between 2010 and 2018: BT549 (HTB-122), Daudi (CCL-213), HCC38 (CRL-2314), Hela (CCL-2), HEK293T (CRL-3216), HEPG2 (HB-8065), HT29 (HTB-38), K562 (CCL-243), LS123 (CCL-255), MCF-7 (HTB-2), OvCa (HTB-161), Phoenix-AMPHO (ATCC, CRL-3213), Saos2 (HTB-85), SCC9 (CRL-1629), and U937 (CRL-1593.2). Freestyle 293-F cells were obtained from Invitrogen (R790-07). The following cell lines were kindly provided by Ilan Bank (Chaim Sheba Medical Center, Tel Hashomer, Israel) in 2014: BT-474, MDA-MB231, and T47D. The following cell lines were kindly provided by Lisa Wiesmüller (Universitätsfrauenklinik, Ulm, Germany) in 2014: MDA-MB-134, MDA-MB-436, MDA-MB468, and UACC-893. Thordur Oskarsson (Deutschen Krebsforschungszentrum, Heidelberg, Germany) kindly provided cell line MDA-MB157 in 2016. Phil Greenberg (Fred Hutchinson Cancer Research Center, Seattle, USA) kindly provided cell line LCL-TM. Barbara Seliger (University of Halle, Germany) kindly provided cell line MZ1851rc.

The following cell lines were reauthenticated using the Cell Line Authentication (CLA) Test provided by Eurofins Genomics Europe in 2017: Daudi, HEK239T, Hela, K562, LCL-TM, MDA-MB-231, MZ1851rc, Phoenix-AMPHO, Saos2, SCC9, and T-47D. The following cell lines were reauthenticated using the CLA Test provided by Eurofins Genomics Europe (Ebersberg, Germany) in 2019: BT474, Daudi, HCC38, HT29, K562, LCL-TM, MCF7, MDA-MB-231, MZ1851rc, OvCa, and Phoenix-AMPHO. Mycoplasma testing was done regularly in house using the Myco alert Mycoplasma detection kit (Lonza, LT07-318). Cell lines were used in assays from passage 4 up to 25 after thawing.

The Daudi, K562, U937, and LCL-TM cell lines were cultured in the RPMI-1640 medium (Thermo Fisher, 72400054) with 10% fetal calf serum (FCS; Sigma-Aldrich, F7524), penicillin (100 IU/mL), and streptomycin (100 μg/mL; Thermo Fisher, 15140163) at 37°C. The BT474 cell line was cultured in the RPMI-1640 medium with 10% FCS, penicillin (100 IU/mL), and streptomycin (100 μg/mL), and additionally supplemented with 1 mmol/L sodium pyruvate (Thermo Fisher, 11530396) and bovine insulin (0.01 mg/mL; Sigma-Aldrich, I6634-50MG) at 37°C. The cell lines MZ1851RC, BT549, HCC38, Hela, HEK293T, HEPG2, HT29, LS123, OvCa, Phoenix-AMPHO, Saos2, and SCC9 were cultured in DMEM with 10% FCS, penicillin (100 IU/mL), and streptomycin (100 μg/mL) at 37°C. The MCF-7, MDA-MB-134, MDA-MB-157, MDA-MB-436, MDA-MB-468, T-47D, and UACC-893 cell lines were cultured in the DMEM with 10% FCS, penicillin (100 IU/mL), streptomycin (100 μg/mL), glutamine (2 mmol/L; Thermo Fisher, 25030024), 1% nonessential amino acids (100 mmol/L; Thermo Fisher, 12084947), human insulin (4 μg/mL; Sigma-Aldrich, I2643), and EGF (10 ng/mL; Sigma-Aldrich, SRP3027) at 37°C. The MDA-MB-231 cell line was cultured in the IMDM (Thermo Fisher, 31980048) with 10% FCS, penicillin (100 IU/mL), streptomycin (100 μg/mL), and glutamine (2 mmol/L) at 37°C. Freestyle 293-F cells were cultured in FreeStyle 293 Expression Medium (Thermo Fisher, 12338026) at cell densities between 0.3 × 10⁶ and 2.5 × 10⁶ in a shaking incubator at 37°C.

Peripheral blood mononuclear cells (PBMC), T cells engineered to express a defined TCRγδ (TEG), natural killer T (NKT) cells engineered to express a defined TCRγδ (NEG), and γδT cells engineered to express a defined TCRγδ (GEG) were cultured in the RPMI-1640 medium supplemented with 5% human serum (Sanquin), penicillin (100 IU/mL), streptomycin (100 μg/mL), and 50 μmol/L 2-mercaptoethanol (Merck).

Our sequencing protocol has been modified from Mamedov and colleagues (25). Frozen PBMC samples from healthy donors were thawed and stained with human anti-CD3 eFluor 450 (eBioscience, 16-0037-81), anti-TCRαβ APC (eBioscience, 17-9986-42), and anti-TCRγδ PE (Beckman Coulter, IM1349), and the γδ fractions (CD3⁺γδ⁺) were sorted on a BD FACSAria II cell sorter. RNA was isolated from samples with ≥0.5 × 10⁶ cells with the RNeasy Mini Kit (QIAGEN, 74204) and from samples with <0.5 × 10⁶ cells with the RNeasy Micro Kit (QIAGEN, 74004). TCRδ cDNA was synthesized using SuperScript II Reverse Transcriptase (Thermo Fisher Scientific), utilizing a specific primer at the 3′ constant region and a universal template switch adaptor at the 5′ end of the V region (for the specific primers, see Supplementary Table S1), and purified using NucleoSpin Gel and PCR Clean-UP (Macherey-Nagel). The samples were amplified in a first round of PCR using Q5 High-Fidelity DNA polymerase (New England BioLabs, Inc.) and a T100 Thermal Cycler (Bio-Rad) under the following cycling conditions: 90 seconds at 98°C, 15 cycles of 7 seconds at 98°C, 20 seconds at 62°C, and 50 seconds at 72°C, followed by 10 minutes at 72°C. A specific nested primer located in the constant region and a step-out primer, which anneals to the switch adaptor, were used (Supplementary Table S1).

The resulting amplicons were loaded onto a 1.5% agarose gel, electrophoresed, and products between 400 and 600 base pairs were size selected. After purification of the gel with NucleoSpin Gel and PCR Clean-UP, the PCR products were used for a second PCR using a reverse nested primer on the constant region and a forward primer which annealed to the switch adaptor (Supplementary Table S1), using the following cycling parameters: 90 seconds at 98°C, 20 cycles of 7 seconds at 98°C, 20 seconds at 62°C, and 50 seconds at 72°C, followed by 10 minutes at 72°C. After purification with NucleoSpin Gel and PCR Clean-UP, library preparations for HTS were carried out using the HTSgo-LibrX kit with HTSgo-IndX indices following the manufacturer's (Gendx) recommendations. Subsequently, samples were cleaned up (HighPrep PCR beads, GC Biotech), and HTS was performed on a MiSeq System 500 (2 × 250 bp read lengths), Illumina. Analysis of the HTS data is described in “Analysis of HTS data sets and availability of data sets.”

RNA samples from TNBC tumors were reverse transcribed into cDNA using Superscript III enzyme (Invitrogen) and oligo(dT) primers according to the manufacturer's protocol. Multiplex primer sequences of hTRDV1, hTRDV2, hTRDV3, TRDV5/29, and hTRDC1 for δ-chain were used to prepare the HTS library (see Supplementary Table S1). Adaptor sequences were added as overhangs to facilitate the Illumina MiSeq analyses. All primers were mixed in equal amounts to achieve a final concentration of 10 μmol/L. The total volume of each PCR reaction was 20 μL consisting of 1 × PCR buffer (without MgCl₂; Invitrogen), 1.5 mmol/L MgCl₂, 10 mmol/L dNTPs (QIAGEN, 201900), 0.5 μmol/L forward primer mix, 0.5 μmol/L reverse primer, and 0.04 units of recombinant Taq DNA polymerase (Invitrogen). Cycling conditions were 3 minutes at 95°C, 30 seconds at 95°C, 30 seconds at 63°C, and 30 seconds at 72°C for 5 cycles; 30 seconds at 95°C and 35 seconds at 72°C for 20 to 25 cycles; and 4 minutes at 72°C. After electrophoresing amplicons on 1% TAE-agarose, 250-bp-sized bands were excised and purified (QIAquick Gel Extraction Kit, Qiagen). For paired-end Illumina sequencing, additional adapters (indicated below), including a flow cell binding site for sequencing on the Illumina MiSeq System and indices for demultiplexing, were introduced to the TCR amplicons. For this, the Nextera Index Kit (Illumina) primers were used in a 50-μL PCR reaction with the following components: 1× Advantage 2 PCR Buffer (Clontech), 1 μmol/L of Index 1 and Index 2 (Nextera Index Kit; Illumina), 1× dNTP Mix (10 mmol/L each; Clontech), 30 to 40 ng DNA template, and 1× Advantage 2 Polymerase Mix (Clontech). PCR products were purified using the Agencourt AMPure XP PCR Purification Kit (Beckman Coulter) according to the manufacturer's recommendation. Further preparation and loading of the library to the MiSeq System was done according to the “denature and dilution guide” provided by Illumina. Analysis of the HTS data is described in “Analysis of HTS data sets and availability of data sets.”

Serial FFPE 2-μm thick sections mounted on SuperFrost plus glass slides (R Langenbrinck GmbH) were dewaxed and rehydrated. After proper antigen retrieval in a pressure cooker with citrate buffer (pH 6) and citrate buffer (pH 6.1; Dako), blocking of nonspecific binding was performed using 5% normal goat serum in PBS (Jackson ImmunoResearch, 005-000-121). Mouse monoclonal anti-TCRγ-chain (Thermo Fisher, TCR1153) and rabbit anti-human cleaved caspase-3 polyclonal antiserum (Cell Signaling Technology, 9664) were used, as we previously reported (6). The human cytomegalovirus (CMV) detection was performed using mouse anti-CMV (clones CCH2 + DDG9, DAKO, and IS752) as described below. Alkaline phosphatase–conjugated and horseradish peroxidase–conjugated detection systems were used to visualize the primary antibodies in separate or sequential protocols for single or double staining tests with red and brown chromogen (DakoREALDetection System Alkaline Phosphatase/RED r/m and EnVisionFLEX Systems, Dako). Acidic hematoxylin was used as a counterstain.

Frozen serial sections of 5-μm thickness were mounted on SuperFrost plus glass slides, air-dried for 3 hours, and fixed with precooled acetone (−20°C) for 10 minutes. Samples were rinsed with 3× TBST for 5 minutes, and blocking of nonspecific binding was performed using 5% human serum in PBS for 30 minutes. The samples were incubated with the corresponding primary mouse anti-human mAbs: anti-TCRγ (Thermo Fisher, TCR1153) and anti-CD69 (Leica Microsystems, NCL-CD69), or goat anti-human polyclonal anti-IFNγ (R&D Systems, AF-285-NA), anti-TNFα (Novus bio, NBP1-19532), and anti-IL17 (R&D Systems, AF-317-NA) antisera. Fluorescence-conjugated secondary antibodies used to visualize the primary antibodies were rabbit anti-mouse AlexaFluor 488 (TCRγ; Thermo Fisher, A-11059), donkey anti-rabbit AlexaFluor 568 (CD4; Abcam, ab175694), donkey anti-goat AlexaFluor 594 (CD69, IFNγ, and TNFα; Abcam, AB150132), donkey anti-goat AlexaFluor 647 (IL17, pseudocolored; Abcam, AB150131). Samples were mounted using ProLong Diamond Antifade medium with DAPI (Thermo Fisher).

Samples were tile scanned using a Zeiss Axio Observer 7 with Apotome 2 system with an ERc 5s digital camera and analyzed using the AxioVision 4.8 and ZEN BLUE image software (all from Carl Zeiss). Four hundred randomly selected cells positive for TCRγ were counted manually in serial slides and evaluated for colocalization with IFNγ, TNFα, or IL17 with the ImageJ (NIH images) and QuPath (GitHub) softwares and the Bio-Formats, Stack Slicer, and Cell counter plugins (26) by two independent researchers (J. Villacorta Hidalgo and A.R. Terrizi).

The breast cancer cell lines and frozen tumor material were tested for CMV using an IHC-nested PCR according to a previously published protocol (27), and qPCR with the artus CMV TM PCR Kit (Qiagen) in a 7900HT Fast Real-Time PCR cycler (Applied Biosystems) according to the manufacturer's instructions (for primer sequences, see Supplementary Table S1).

Frozen sections (8 μm thick) were air dried overnight on MembraneSlide 1.0 PEN membrane covered slides (Carl Zeiss) fixed in precooled acetone (−20°C) for 10 minutes, washed twice with TBST, incubated first for 30 minutes with 5% human serum in PBS, and then for an additional 30 minutes at room temperature with mouse anti-human TCRγ (100 μL, 1:40; Thermo Fisher, TCR1153). To detect the mouse mAb binding, a biotinylated anti-mouse secondary antibody was used from the alkaline phosphatase detection system (Dako REAL Detection System Alkaline P/RED, rabbit-mouse), and the slides were counterstained with Mayer's hematoxylin. Samples were dried at room temperature for 2 hours, examined with an Axiovert light microscope (Carl Zeiss) for TCRγ⁺ TIL, and stored at 4°C until processing.

Laser microdissection was performed using an Axiovert microscope equipped with the PALM MicroBeam system (Carl Zeiss). Energy parameters of cutting and catapulting were established individually for each sample. Only infiltrating single lymphocytes in contact with cancer cells were selected and then microdissected. The selection and microdissection processes were performed at 40× and 63× magnification, respectively. Single cells were catapulted into the cap of an Adhesive Cap 500 opaque 500 μL tube (Carl Zeiss). Twenty microliters of a 1:10 mix containing Expand High-Fidelity Buffer and proteinase K (20 mg/mL, PCR grade), both from Roche Diagnostics) was then added for digestion. The tubes were incubated with their lids closed for 4 hours at 56°C, centrifuged for 2 minutes at 500 rpm, and heat inactivated at 95°C for 10 minutes. Additional tubes containing only membrane were used as negative controls. All PCR tubes were overlaid with mineral oil in a laminar flow hood before adding the PCR master mix.

To assess the clonality of γδT cells by the length of the CDR3 regions, we used spectratyping analysis. RNA was extracted from frozen tumor tissue using RNeasy Tissue minikit (Qiagen), and cDNA reverse transcribed by Oligo-dT-primers (First-Strand cDNA Synthesis Kit, GE Healthcare) according to the manufacturer's instructions.

A first round of PCR was performed using standard unlabeled primer pairs, followed by a second round of primer extension [run-off (RO)] with a fluorescence-labeled nested antisense oligonucleotide. The PCR reaction conditions for TCRγ and TCRδ were 3 minutes 95°C for initial denaturation, followed by 40 cycles (45 seconds 95°C, 45 seconds 60°C, 60 seconds 72°C), with a final 5 minutes 72°C extension. In the second round, 2 μL of the PCR products was used for a primer extension reaction with a single fluorescent (5′ 6-FAM-labeled) nested antisense primer from the particular constant region (RO primers). The conditions of this “run-off reaction” were 2 minutes at 94°C, then followed by 5 cycles (25 seconds 94°C, 45 seconds 60°C, 45 seconds 72°C), and finally by 5 minutes at 72°C. The products were analyzed using an ABI 3130 XL capillary sequencer (Applied Biosystems) to determine the length distribution of the fluorescent fragments for γ and δ groups. The results were analyzed using the Genescan Analysis software version 3.7 (Applied Biosystems; primer sequences in Supplementary Table S1; ref. 28).

Similar to the single-cell PCR technique previously described for the analysis of rearranged immunoglobulin genes (29), a multiplex, semi-nested, hot-start PCR was set up with 15 newly designed primers (see Supplementary Table S1). For the first round, a master mix was prepared containing 5 μL dNTP (2 mmol/L), 5 μL 10× PCR buffer (High-Fidelity System), 5 μL primer mix (2.5 μmol/L, forward and reverse primers), 3.2 μL 25 μmol/L MgCl₂, 6.5 μL H₂O, and 15 μL from the DNA digestion. A volume of 0.3 μL Expand High-Fidelity enzyme mix (3.5 units/μL) was added after the first denaturation step to a final volume of 40 μL. The cycler program was 95°C 2 minutes, 80°C pause (enzyme added), 72°C 1 minute, 39 cycles at 95°C 50 seconds, 56°C 30 seconds, 72°C 60 seconds, then 72°C 5 minutes and 10°C pause. For the second round, eight master mixes were prepared to detect TCRγ and δ chains: two mixes for TCRγ (Vγ1-8 and Vγ9) and six for TCRδ (Vδ1, Vδ2, Vδ3, Vδ4, Vδ5, and Vδ6) with 2.5 μL dNTP (2 mmol/L), 2.5 μL 10× PCR buffer, 1.25 μL of the respective Vγ and Vδ forward primers, 1.25 μL of the respective J segment mix primers (see Supplementary Table S1), 2 μL 25 mmol/L MgCl₂, 12.2 μL H₂O, 3 μL of first-round PCR product, and 0.3 μL Expand High-Fidelity enzyme mix (3.5 units/μL). The cycler program was 95°C 5 minutes, 72°C 1 minute, followed by 35 cycles at 95°C 50 seconds, 55.5°C 30 seconds, 72°C 1 minute, then 72°C 5 minutes, 15°C 5 minutes and 4°C pause. The PCR products were analyzed by 2% agarose gel electrophoresis, and the positive bands were excised and purified from the gel with the QIAEX II Gel Extraction Kit (Qiagen). The DNA was sequenced using the BigDye Terminator 3.1 system (Applied Biosystems). The sequencing reactions consisted of 1 μL BigDye, 3.75 μL 5× sequencing buffer, 0.75 μL forward primer (see Supplementary Table S1), 3 to 10 μL template, and water to a final volume of 20 μL. The cycling conditions were 96°C 5 minutes, then 24 cycles at 95°C 15 seconds, 50°C 10 seconds, 60°C 4 minutes, followed by a 10°C pause. The sequence sample was cleaned using the DyeEx 2.0 Spin Kit (Qiagen) and analyzed on the ABI 3130xl capillary sequencer (Applied Biosystems).

Codon-optimized DNAs coding for the full lengths of γ and δ TCR chains and endothelial protein C receptor (EPCR) were ordered at BaseClear Inc (see Supplementary Document, “Nucleotide sequences of synthetic DNA constructs”). The synthetic genes were flanked with a 5′ NcoI and a 3′ BamHI site for subsequent cloning into the retroviral expression vector pBullet (ref. 30; kind gift by Ralph Willemsen, Erasmus Medical Center, Rotterdam, the Netherlands), modified by inserting IRES-neo and IRES-puro cassettes (see Supplementary Document, “Nucleotide sequences of synthetic DNA constructs”). The γ TCR genes were subcloned into the pBullet-IRES-neo, and the δ TCR genes and EPCR were subcloned into the pBullet-IRES-puro as previously performed (31).

For the generation of TEGs, PBMCs from healthy donors (Sanquin) were transduced with defined TCRγ and δ chains as described (19, 32, 33). Briefly, retroviral supernatants were produced by Phoenix-AMPHO packaging cells, transfected with retroviral helper vectors, gag-pol (pHIT60; ref. 34) and ENV (pCOLT-GALV; ref. 35; both kind gifts by Ralph Willemsen, Erasmus Medical Center, Rotterdam, the Netherlands), together with pBullet retroviral constructs containing TCRγ and TCRδ of the indicated TCR (Results section) using FuGENE HD (Promega). PBMCs preactivated with anti-CD3 (30 ng/mL; clone OKT3, Janssen-Cilag) and IL2 (50 U/mL; Proleukin, Novartis, 13610880, hospital pharmacy UMCU) were transduced twice (within 48 hours) with viral supernatants in the presence of IL2 (50 U/mL) and polybrene (4 μg/mL; Sigma-Aldrich). Transduced T cells (TEG) were expanded by stimulation with anti-CD3/anti-CD28 Dynabeads (0.5 × 10⁶ beads/10⁶ cells; Invitrogen) and IL2 (50 U/mL) and incubated with geneticin (800 μg/mL; Gibco) and puromycin (5 μg/mL; Sigma-Aldrich) for 1 week. TEGs were stimulated biweekly with PHA-L (1 μg/L; Sigma-Aldrich), IL2 (50 U/mL; Proleukin, Novartis), IL15 (5 ng/mL; R&D Systems, 247-IL), and irradiated allogeneic PBMCs (12.5 × 10⁶, dose: 3,500 cGy), Daudi (2.5 × 10⁶, dose: 8,000 cGy) and LCL-TM (2.5 × 10⁶, dose: 8,000 cGy) cells. Fresh IL2 was added twice weekly. Transgenic TCR expression was routinely assessed by flow cytometry on a BD FACSCanto II using anti-TCRαβ APC (1:10; eBioscience, 17-9986-42), and anti-TCRγδ PE (1:10; Beckman Coulter, IM1349), anti-CD4-FITC (1:20; eBioscience, 11-0049-42), and anti-CD8–PerCPCy5.5 (1:1,000; BioLegend, 301032).

For the generation of NEGs and GEGs, NKT and γδ T cells of healthy donors were transduced using a similar protocol that was used for PBMCs. γδ and NKT cells were sorted prior to transduction using a MACS TCRγ/δ+ T-cell or NKT cell isolation kit (Miltenyi), respectively, a >90% γδTCR⁺ cell population and a >80% CD56⁺/CD3⁺ NKT cell population was obtained after MACS. The NEGs and GEGs were expanded with PHA-L (1 μg/L; Sigma-Aldrich), IL2 (50 U/mL), IL15 (5 ng/mL; R&D Systems, 247-IL), and irradiated allogeneic PBMCs (12.5 × 10⁶, dose: 3,500 cGy), Daudi (2.5 × 10⁶, dose: 8,000 cGy) and LCL-TM (2.5 × 10⁶, dose: 8,000 cGy) cells. Fresh IL2 was added twice weekly.

HEK293T, HeLa, K562, and U937 cells were transduced with EPCR in a similar fashion as described above. Three days after the second transduction, the transduced cells were selected with puromycin (2.5 μg/mL), and the transgenic expression of EPCR was assessed by flow cytometry on a BD FACSCanto II, cells were stained using anti-EPCR (1:50; Abcam, ab81712) and goat-anti-rat IgG AF647 (1:400; Invitrogen, A-21247).

IFNγ ELISPOT assays were performed as previously described (16, 36). Briefly, 15,000 TCR-transduced or mock-transduced T cells and 50,000 target cells (ratio 0.3:1), as indicated in the figures, were cocultured for 24 hours in nitrocellulose-bottomed 96-well plates (Millipore) precoated with anti-IFNγ (5 μg/mL; clone 1-D1K; Mabtech). Plates were washed and incubated with a second biotinylated anti-IFNγ (1 μg/mL in PBS/0.5% FCS clone 7-B6-1; Mabtech), followed by streptavidin-HRP (1:500 dilution in PBS; Mabtech). IFNγ spots were visualized with the TMB substrate (Sanquin), and the number of spots was quantified using ELISPOT Analysis Software (Aelvis).

Two TIL TCRγδ clones transduced in αβT cells, TEG-C132 and TEG-F4, were incubated alone or with target cells at an effector:target (E:T) ratio of 1:1 in the presence of CD107a-PE (BD Biosciences; clone H4A3). γδTCR monoclonal antibody/antibodies and a matched isotype control were included during the incubation at a concentration of 20 μg/mL. For TEG-C132 and TEG-F4, anti-γδTCR clone 11F2 (Antibody Chain International B.V.) and isotype control (IgG1κ) clone MOPC-21 (Abcam) were used. For TEG-C132, two additional γδTCR monoclonal antibodies were included: clone B1 (BioLegend) and clone 5A6.E9 (Fisher Scientific). After 2 hours of incubation, Golgistop (1:150; BD Biosciences, 554724) was added. After 6 hours, cells were washed in FACs buffer, containing PBS + 1% BSA (fraction V; Sigma-Aldrich, 10735094001) and stained with an antibody mix. For TEG-C132: anti-CD3 eFluor450 (eBioscience; clone OKT3), CD8 PerCP-Cy5.5 (BioLegend; clone RPA-T8), and γδTCR APC (BD Biosciences; clone B1). For TEG-F4: Vδ1 FITC (Thermo Scientific; clone TS8.2), CD8 PerCP-Cy5.5 (BioLegend; clone RPA-T8), CD4 PeCy7 (eBioscience; clone RPA-T4), and αβTCR APC (eBioscience; clone IP26). Cells were washed two times in FACS buffer and fixed in 1% paraformaldehyde in PBS. Data acquisition was done on FACSCanto and analyzed using FACSDiva software (BD).

The variable and constant domains of the TCR chains, clones C132, F4, and Zi11, were amplified from synthetic DNA encoding the full-length TCRs using gene-specific primers containing an AfeI restriction site for the forward V-gene–specific primers and a BamHI site for the reverse C-gene–specific primers (see Supplementary Table S1). The TCRδ chains were ligated into a modified pBullet vector containing a μ-phosphatase signal peptide at the 5′ end and fos zipper at the 3′ end of the construct (see Supplementary Document, “Nucleotide sequences of synthetic DNA constructs”). The TCRγ chains were ligated in a modified pBullet vector containing a μ-phosphatase signal peptide at the 5′ end and at the 3′ end, a jun zipper followed by a biotin acceptor peptide and a poly-histidine (His) tag (see Supplementary Document, “Nucleotide sequences of synthetic DNA constructs”). Synthetic DNA encoding for the bacterial biotin ligase BirA (Uniprot accession code: P06709; see Supplementary Document, “Nucleotide sequences of synthetic DNA constructs”) was also ligated in a pBullet vector containing a signal peptide. Soluble γδTCRs were expressed by Freestyle 293-F cells (Thermo Fisher) transiently transfected with plasmids containing TCRδ, TCRγ, and BirA (in a 45:45:10 ratio) using polyethylenimine (PEI) in a 2:3 ratio, 1 μg of total plasmid DNA was used to transfect 10⁶ cells. Six hours after transfection, the media were supplemented with 1% penicillin/streptomycin (Gibco) and 100 μmol/L biotin (Sigma-Aldrich; 14400).

The expression media were harvested 5 days after transfection by centrifuging the cultures for 10′ at 750 × g to pellet the cells. The supernatant was supplemented with TBS, and loaded on a 1-mL HisTrap Excel column (GE Healthcare). A multistep gradient with increasing concentrations of imidazole (Merck, 1047160250), 10 column volumes (CV) 10 mmol/L imidazole, followed by a linear gradient of 10 to 300 mmol/L imidazole in 20 CV, was used to wash and elute the soluble TCRs from the column. The eluted soluble TCRs were further purified using a 1-mL HiTrapQ column (GE Healthcare) at pH 8.2, soluble TCRs were loaded in TBS-20 mmol/L NaCl and eluted using a linear gradient of 20 to 400 mmol/L NaCl in 30 CV. Fractions were loaded on a 4–20 Mini Protean TGX Gels (Bio-Rad) using Laemmli sample buffer (Bio-Rad, 1610747) after electrophoresis gels were stained with InstantBlue Safe Coomassie Stain (Sigma-Aldrich, ISB1L). Fractions containing the soluble TCR were pooled and concentrated using a Vivaspin Turbo 4 10-KDa cutoff spin concentrator (Sartorius, VS04T01). Tetramers were prepared from TCR monomers by adding one equivalent of SA-PE (BD Pharmingen, 554061) to five equivalents of TCR in 2 steps over 5 minutes.

MDA-MB436 cells were incubated with TCR tetramers with a streptavidin concentration 3 μg/mL for 30 minutes at room temperature in the presence of isotype control (IgG1κ, clone MOPC-21; abcam) or anti-γδTCR (clone B1; BioLegend). Cells were washed two times in FACS buffer and fixed in 1% paraformaldehyde in PBS. Data acquisition was done on FACSCanto and analyzed using FACSDiva software (BD).

TCR sequence alignment, assembly, and clonotype extraction were performed using the MiXCR (version-v2.1.1) program for data set 1, 3, 4, and 5 (37). VDJtools (v1.1.4) was utilized for frequency-based correction of clonotypes (38). For data set 1, 3, and 4, only functional reads which passed a frequency filter of 0.1% were used for further analysis. For data set 5 sequences with a read count of ≥20 were used for analysis. RNA-seq data of data set 6 were analyzed with the RNA-seq mode of MiXCR. tcR R package (v2.2.1.15) was utilized for TRD repertoire overlap analysis and the estimation of VDJ gene usage.

HTS data in the standard FASTQ format from this study (data sets 1 and 3) are available via the SRA database and can be located using the NCBI BioProject accession number: PRJNA397967. HTS data in the standard FASTQ format from data sets 4 and 5 are publicly available (39, 40). Data set 6 consists of the RNA-seq data of the TCGA database. The following cancer data sets were used: COAD (colon adenocarcinoma), KIRC (kidney clear cell carcinoma), LUAD (lung adenocarcinoma), READ (rectal adenocarcinoma), SKCM (skin cutaneous carcinoma), and TNBC. Data set 6 was supplemented with amino acid CDR3 sequences of publicly available CDR3 amino acid sequences of TILs (41).

Unpaired t test statistical analysis or a one-way ANOVA followed by Tukey post hoc test was performed using GraphPad Prism software (GraphPad Inc, Version 7). Data were expressed as means ± standard deviation (SD). A value of P < 0.05 was considered statistically significant.

We analyzed γδ TILs that were in close contact with tumor cells within TNBC frozen sections of 7 patients, 3 with invasive ductal cancer (IDC) and 4 classified as medullary breast cancer (MBC; Supplementary Table S2; ref. 6). In both tumor types, the γδ TILs were frequently in close contact with apoptotic tumor cells (Fig. 1A and B), and the majority of γδ TILs stained positive for CD69, IFNγ, and TNFα, whereas less than 20% were positive for IL17 (P < 0.001, n = 7; Fig. 1C–F and H). No difference was seen in the percentage IFNγ⁺ and IL17⁺ γδ TILs between MBC and IDC tumors. However, we observed a small decrease in the percentage of TNFα⁺ γδ TILs in MBC compared with IDC tumors. Over 80% of CD4⁺TCRγδ^– cells produced IL17 and appeared to represent a major source for IL17 in the tumor parenchyma (Fig. 1G and I).

Initially, Vγ and Vδ spectratyping in TNBC patients was performed following RNA extraction from frozen tissue sections (Supplementary Fig. S1A and S1B). These analyses showed polyclonal γδ T-cell populations where the Vγ9 gene expression was reduced, implying that γδ T cells other than Vγ9/Vδ2 were predominant. HTS of the TNBC samples confirmed the diverse TCRδ repertoire (Fig. 2; data set 1; Supplementary Table S3). In order to characterize the pairs of TCRγ and δ chains of γδ T cells in contact with tumor cells, we used laser microdissection of single γδ TILs, followed by a PCR protocol to amplify their CDR3 regions. In total, 530 single γδ T cells were isolated from 11 tumors. Although some SCS reactions did not generate reliable sequencing data for both of the paired chains, 27 paired TCRγ and δ sequences from 9 different patients were identified (data set 2; Supplementary Table S4). We also obtained 63 TCRγ and 28 TCRδ nonpaired sequences (data set 2; Supplementary Tables S2 and S5). Various non-Vγ9 and non-Vδ2 genes were most prominently represented among the detected sequences from single cells (Fig. 2; data set 2). Out of the 81 unique TCRγ amino acid sequences obtained, 10 different sequences were shared among the patients, 9 of which had the same nucleotide sequence (Supplementary Fig. S2A). One TCRγδ, B9, had a TCRγ chain that has been previously identified in a CD1d-restricted TCRγδ, AU2.3 (42). Four of the 27 SCS-identified paired TCRγδ used the same Vδ5 amino acid sequence but was always paired with a different TCRγ chain. This Vδ5 sequence, previously found to be associated with CMV and tumor reactivity (43), was identical with 2 of 28 SCS-identified CDR3δ nonpaired amino acid sequences, classifying this CDR3δ sequence as coding for a public CDR3δ chain (Supplementary Fig. S2B). All the nucleotide sequences of this public Vδ5 CDR3δ chain were identical among different patients (Supplementary Table S4). None of the examined TNBC tissue samples tested positive for CMV (Supplementary Table S2), but the serostatus of CMV in our patients was not determined. One of the 27 SCS-identified paired TCRγδs had a shared (public) Vδ1 sequence, and another shared Vδ1 sequence was identified in the pool of the nonpaired TCRδ chains. The shared Vδ1 amino acid sequences also had the same nucleotide sequence. Only four of the 41 SCS-identified unique TCRδ amino acid sequences were present within HTS data belonging to the same TNBC sample, suggesting an initial repertoire focusing (39) at the tumor side, with a subsequent lack of clonal expansion of tumor-interacting γδ T cells.

To put our own data of TCRδ chain sequences derived from γδ TILs in TNBC within the context of natural repertoires observed in the peripheral repertoire, as well as from tumors, we investigated the overall usage of TCRδ chains in the peripheral blood of healthy donors and in other tumor tissues. Therefore, we analyzed the presence of TCRδ amino acid sequences across different healthy and tumor sequencing data sets. First, we analyzed the presence of shared TCRδ chains in the peripheral blood of 13 healthy volunteers by HTS (data set 3; Supplementary Table S3) and studied the most prevalent TCRδ sequences, defined as sequences with a clonal frequency of >0.1%. Of the 1401 most prevalent amino acid sequences in data set 3, 17 were shared between at least two donors. Next, we analyzed shared sequences of two publicly available TCRδ HTS data sets from peripheral blood of healthy donors. For data set 4 (Supplementary Table S3) of Ravens and colleagues, we again analyzed the sequences with a clonal frequency of >0.1% (39). Data from Davey and colleagues (data set 5; Supplementary Table S3) were pooled in the public space (40), and therefore only sequences with a read count of more than 20 were included in our analysis. The Vδ gene distribution of data sets 3 and 4 consisted mainly of Vδ2 sequences representing an unselected peripheral repertoire. Data set 5 was enriched for Vδ1⁺ γδ T cells, which explains the high percentage of Vδ1 sequences (Fig. 2). Within the different healthy donors' data sets, 186 Vδ2⁺ TCRδ amino acid sequences were shared and, thus, considered public (Fig. 3A; Supplementary Fig. S3A). Contrastingly, no shared Vδ1 sequences were identified in the peripheral blood data sets. Data set 6 (Supplementary Table S3) included TCRδ sequences retrieved from the RNA-seq data of six different cancers from The Cancer Genome Atlas (TCGA) database and from the γδ TILs CDR3δ sequences of various cancers published by Li and colleagues (41). The majority (62%) of our analyzed 1,407 γδ TIL TCRδ sequences were of the Vδ1 origin (Fig. 2). Considering the Vδ2⁺ T-cell proportion in the TIL data set (data set 6), the proportion of Vδ2⁺ in the TNBC HTS data (data set 1) was higher than expected, and more similar to the peripheral blood data sets (Fig. 2). This is compatible with TNBC's higher microvascular density in comparison with other breast tumors (44) or due to the fact that Vδ2⁺ chains can pair with other TCRγ chains than Vγ9 (40). When data set 6 with the TILs was compared with the sequences of all the other data sets, the percentage of Vδ2 TCRδ sequences in tumor tissues appeared relatively low, but 53 shared Vδ2 TIL amino acid sequences were identified within the data sets (Fig. 3A; Supplementary Fig. S3A). Finally, we compared nucleotide sequences across all of the different data sets, except data set 6, where nucleotide sequences were not available from the data of Li and colleagues (41). Across the different data sets, 222 Vδ2 amino acid sequences were shared, and 29 of those (13%) were also identical as nucleotide sequences (Supplementary Table S6). Because the pooling of individual donor data could result in underestimating shared nucleotide sequences, we also analyzed shared nucleotide sequences between different donors within data sets 3 and 4. This analysis revealed higher percentages of shared amino acid sequences with identical nucleotide sequences, i.e., 39% (data set 3) and 45% (data set 4), respectively (Supplementary Table S6).

The percentage of shared Vδ2 amino acid sequences in data set 6 (TILs) was relatively high compared with the shared sequences in peripheral blood of healthy donors [13.8% (53/385) vs. 1.5% (186/12207)]. When peripheral T cells were substantially enriched for Vδ1 γδ T cells before HTS (data set 5), 13 Vδ1 sequences could be characterized as shared between peripheral blood and γδ TILs (Fig. 3B; Supplementary Fig. S3B). One Vδ3 sequence, which was present among the γδ TIL sequences (data set 6), was also present in the TNBC TILs (Supplementary Fig. S3C). Thus, our analyses identified commonly shared Vδ amino acid sequences in the peripheral blood and tumor tissues, but the corresponding Vδ repertoires differed considerably. In the peripheral blood, the shared sequences were all Vδ2⁺, whereas tumor tissues also contained shared Vδ2^– TCR chains.

The TEG format with absence of many coreceptors usually observed on γδ T cells (19, 32, 33) allowed us to investigate whether tumor reactivity of γδ TILs was mediated through their individual TCRγδs and not by other innate immune receptors usually expressed on γδ TILs. To this end, we generated a series of 15 TEGs expressing paired TCRγ and δ chains derived from TNBC γδTILs (Supplementary Fig. S4A; Supplementary Table S4) to assess whether these chains had the potential to mediate antitumor reactivity. Because most human γδ T cells are not HLA restricted, we used several established cell lines derived from tumors, including breast cancer and leukemia/lymphomas. Activation of TEGs by tumor cell lines was measured using an IFNγ ELISPOT, where we used the following thresholds to quantify reactivity: >50 spots as low reactive and >100 spots as highly reactive. Nine of 15 TEGs recognized at least one of the cancer cell lines (Fig. 4A). Because some of the TEGs recognized tumor cell lines of hematologic origin only, we extended the tumor panel with 8 additional solid tumor cell lines for 5 TEGs. Three of 5 tested TEGs, which showed in the first screen only activity against hematologic tumors, also recognized solid tumor cell lines (Fig. 4B). Two of the 4 TEGs with the shared Vδ5 chain displayed distinct tumor reactivity patterns, indicating that tumor recognition was likely to depend on the specific paired TCRγ chain (Fig. 4A and B). Also, the TCRγ and δ chains from TEG-C132 were identical with those reported in the EPCR-reactive γδ T-cell clone (45). The EPCR reactivity of TEG-C132 was confirmed by measuring the response of TEG-C132 against a panel of naturally EPCR-positive or -negative tumor cell lines, or cell lines engineered to overexpress EPCR (Fig. 5A).

γδ T-cell activation can be influenced by multiple cell receptors (46). To test whether coreceptors expressed by γδ T cells could modulate the activity of TCRγδ derived from γδ TILs, TCRγδ complexes from TEG-C132 (functional within TEGs) and B23 (nonfunctional within TEGs), as well as the nonfunctional TCRγ9δ2 LM1 (33), were used to construct γδ T cells with a defined TCRγδ (GEG). No gain of function was observed for either receptor in the GEG format. In contrast, a slight reduction of activity was detected for GEG-C132 compared with TEG-C132 (Fig. 5A and B), possibly due to the mispairing of introduced TCRγ and δ chains with endogenous TCRγ and δ chains. To avoid the possible mispairing with endogenous TCRγ or δ chains, we engineered NKT cells with a defined TCRγδ (NEG). NKT cells were used because they express coreceptors similar to those present in γδ T cells (Supplementary Fig. S4B). NEG-C132 showed no increased reactivity compared with TEG-C132, and consistent with this, no gain of function was detected in NEG-B23 cells (Fig. 5C).

To further confirm that the observed tumor reactivity was mediated through the introduced TCRγδ, TEG-C132, TEG-F4, and TEG-Zi11 were selected for CD4⁺ TEGs. CD4⁺ TEGs lack most NK cell receptors (16). CD4⁺ TEGs showed a similar reactivity pattern as the nonsorted TEGs. Again, CD4⁺TEGs expressing the nonreactive TEG-LM1 did not show any reactivity (Fig. 6A–C), suggesting that indeed activity was mediated by the introduced TCRγ or δ chains. We next determined whether for selected CD8⁺ TEGs (C132 or F4) degranulation was observed upon stimulation with a recognized target cell line. TEG-C132 showed increased surface expression CD107 after a 6-hour incubation with HT29 cells (Fig. 6D). Activity was selectively associated with cells having the highest expression of TCRγ or δ chains (Supplementary Fig. S4C). In the very same experiments, blocking antibodies have also been added to further substantiate the claim that activity was mediated by the introduced TCRγ or δ chains. In the presence of anti-γδTCR, surface CD107 decreased up to 2-fold. Similar results were obtained for TEG-F4 (Fig. 6E). In the absence of target cells, adding blocking antibodies also associated with a slight increase in CD107, suggesting that functional blocking experiments were partially hampered by a slight activation through the very same antibodies. Therefore, we formally addressed whether the γδTCR alone would be sufficient to bind to recognized tumor cell lines by generating fluorescent tetramers from different γδTCR that have been coincubated with MM436 as the tumor target. In line with their reactivity in the TEG format, F4 and Zi11 tetramers stained MM436, whereas C132 tetramer did not. γδTCR dependency was furthermore confirmed by adding anti-γδTCR, which resulted in a complete abrogation of staining through F4 and Zi11 tetramers (Fig. 6F). Thus, we have demonstrated that SCS of γδ TILs, followed by expression of their individual TCRγδs in lymphocytes, could rapidly confirm the implied and anticipated effectiveness of γδ TIL-tumor cell interactions, whose antitumor reactivity does not depend on additional innate immune receptors.

Our report is a comprehensive analysis of TNBC-infiltrating γδ TILs, which are in close proximity to TNBC cells. We observed in this “project genesis” from archival material that the majority of γδ TILs harbored a proinflammatory phenotype and only a minority associated with the expression of IL17. By reconstructing TCRγ or TCRδ from SCS within the TEG format, we provided evidence that paired TCRγ and TCRδ chains could be active against not only breast cancer cells but also other tumor cell types. However, the lack of autologous viable tumor material did not allow to test the activity in an autologous system, and with the lack of knowledge on most tumor antigens seen by γδ TILs, we could not formally assess whether frequency of truly tumor-reactive γδ TILs was equivalent or exceeded the 10% as reported for αβ TILs (47). We observed a high frequency of reactivity of all characterized TCRγ and TCRδ chains against different and multiple tumor types, which was unique for each TCR pair. Changing one counterpart altered the recognition pattern and was not improved when additional innate coreceptors were expressed in combination with the TCR. Thus, γδ T cells have, through their individual TCRγ and δ chains, an intrinsic capacity to recognize cancer cells.

Clonal expansions have been reported in neoantigen-specific αβ TILs (48). The spectratype and HTS analyses of the γδ TILs in TNBC tumor tissue indicated a polyclonal repertoire, and the SCS detected no clear clonal dominance of tumor-reactive TCRγδs. This validates the previously reported absence of Vδ2⁺ TCR-driven lymphocyte expansion in cancer patients (8) and contrasts the expansions of Mycobacterium tuberculosis–activated Vδ2⁺ γδ T Iymphocytes (49). This apparent lack of clonal expansions is also in contrast with the potent clonal γδ T-cell responses to viral infections (39, 40). This dichotomy—i.e., the presence of clonal responses in infections but their absence in progressing cancers—might be caused by a tolerogenic tumor microenvironment. γδ TILs have been reported to be skewed toward an IL17- rather than an IFNγ-producing phenotype during the progression of neoplastic disease (10). However, in our cohort, only a minority of the TNBC γδ TILs produced IL17, which has also been observed in colon cancer (50), suggesting that different tolerogenic mechanisms (51) could be involved. Thus, although mouse model studies indicate that the early cancer stages are susceptible to γδ T-cell immunosurveillance (5, 52), our results imply that the complex tumor microenvironment might prevent an effective clonal expansion of tumor-reactive γδ TILs.

Our data allow to speculate that there might be public TCRγ or δ chains in tumor tissue which associate with tumor recognition. However, this observation needs to be seen with caution. First, a systematic bias could have occurred because the total number of sequences analyzed in peripheral blood samples was higher. Extracting TCR sequences from RNA-seq data favors shorter sequences because of the read length of the original data, and public sequences tend to be shorter in sequence (39, 40). Finally, because others report that Vδ2^– TCRs public chains are a rare event in humans (39, 40), we cannot entirely exclude that in the complex SCS procedure in paraffin-embedded tissues associated with some cross-contamination, and we, therefore, may have overestimated the number of public TCR chains. Regardless of this technical concern, pairing the very same TCRγ with different δ chains and vice versa in our data set allowed to test whether tumor reactivity of TCRγ or δ chains pairs depended on both chains. Different clones with either identical TCRγ or δ chains and matching with a different counterpart associated with a different recognition pattern of tumor cells. This is in particular interesting as we characterized one TCR chain pair that has been initially reported to react against EPCR only (45). However, recognition of tumor cells through Vδ2^– γδ TCRs has most been attributed to a recognition mode where both TCRγ and δ chains can bind to different ligands at the same time (53). For example, the TCRγ chain of the identified clone C132 has been suggested to bind to BTN-family members, whereas other TCRδ chains bind to CD1 family members. Recognition of tumor cells could then depend on varying distributions of both ligands. This might also explain why the expression of the Vδ5 chain alone was not sufficient to trigger EPCR reactivity. However, such new TCR pairs might also simply recognize only one completely different ligand.

The isolation of an EPCR-reactive Vγ4Vδ5 TCR clone from our TNBC tissues has been characterized by others (45) and reported to be associated with anti-CMV responses. CMV reactivation has been accounted to reduce relapse of leukemia after allogeneic stem cell transplantations (36, 54). The reduced risk of leukemic relapse after CMV reactivation has been attributed to NK and γδ T cells (36, 55, 56). Within this context, it is tempting to speculate that, in general, the CMV reactivation may not be clinically beneficial only for controlling leukemias (36), but could also confer some protection against certain solid tumors. However, we could not detect CMV in our TNBC sample by IHC, and there was no serum available to test the CMV serostatus of this patient.

Although ligands for most characterized Vδ2^– TCRγ or δ chains remain to be defined, testing Vδ2^– TCRγ or δ chains in the TEG format emphasized the considerable diversity within the γδ T-cell repertoire in terms of tumor reactivity, which might also be a major factor contributing to numerous failures of clinical trials using ex vivo– and in vitro–expanded γδ T cells (57). Here, we demonstrated that the natural weakness of γδ TILs could be annulled by extracting their TCRs and created potent antitumor T lymphocytes (TEG) expressing these γδ TIL-derived TCRs. The absence of HLA restriction of TCRγδ responses underscores the possibility of TEG-based therapies in allogeneic scenarios and paves a way toward a new plethora of antigens accessible to target solid tumors when utilizing Vδ2^– γδ TCRs (16, 17, 58).

In summary, we demonstrated that γδ TILs cells frequently harbored TCRγ or δ chains, which could mediate tumor reactivity. The identification of such cancer cell–sensing TCRγδs and characterization of their ligands may open novel opportunities for future cell-based cancer therapies.

J. Villacorta Hidalgo has and P. Fisch had ownership interest in patent applications by the University Medical Center, University of Freiburg. D.X. Beringer and Z. Sebestyen have ownership interest in patent applications by the UMC Utrecht. J. Kuball is a scientific advisor for, reports receiving a commercial research grant from, and has ownership interest (including patents) in Gadeta. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Janssen, J. Villacorta Hidalgo, D.X. Beringer, Z. Sebestyen, E. Spierings, R. Küppers, P. Fisch, J. Kuball

Development of methodology: A. Janssen, J. Villacorta Hidalgo, D.X. Beringer, S. Kock, M. Follo, P. Fisch, J. Kuball

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Villacorta Hidalgo, D.X. Beringer, S. van Dooremalen, E. van Diest, A.R. Terrizi, P. Bronsert, S. Kock, M. Werner, K. Heise, M. Follo, T. Straetemans, S. Ravens, R. Küppers, P. Fisch, J. Kuball

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Janssen, J. Villacorta Hidalgo, D.X. Beringer, F. Fernando, M. Werner, T. Straetemans, S.A. Kasatskaya, F.E. Frenkel, E. Spierings, M. Malkovsky, P. Fisch, J. Kuball

Writing, review, and/or revision of the manuscript: A. Janssen, J. Villacorta Hidalgo, D.X. Beringer, P. Bronsert, M. Werner, M. Follo, Z. Sebestyen, D.M. Chudakov, E. Spierings, I. Prinz, M. Malkovsky, P. Fisch, J. Kuball

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.R. Terrizi, P. Fisch, J. Kuball

Study supervision: D.M. Chudakov, P. Fisch, J. Kuball

Other (supervision on diagnostic, planning and analysis as principal pathologist): A. Schmitt-Gräff

Funding for this study was provided by ZonMW 43400003 and VIDI-ZonMW 917.11.337, KWF UU 2010-4669, UU 2013-6426, UU 2014-6790 and UU 2015-7601, UU 2018-11393, Vrienden van het UMCU, AICR 10-0736 and 15-0049, and Gadeta to J. Kuball, Lady Tata Memorial Trust and UU 2018-11393 to Z. Sebestyen, and SFB1160 Z2 to P. Fisch. J. Villacorta Hidalgo was supported by a PhD scholarship from Deutscher Akademischer Austauschdienst (DAAD). D.M. Chudakov is supported by grant of the Ministry of Education and Science of the Russian Federation (no. 14.W03.31.0005). We are grateful to Nagesha Appukudige for supporting computational analyses; Guido Kierkels for critical discussion; Markus Kühs, Katja Gräwe, and Bärbel Weinhold for guidance with tissue processing and IHC; and Elvira Myshkin, Andreas Gaa, and Sabine Glatzel for expert technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.