K562 cells were provided by Alexander Knuth (affiliated with the University Medical Center Mainz, Mainz, Germany, at the time of material transfer), HEK293T cells were provided by Nilabh Shastri (UC Berkeley, Berkeley, CA, USA) and Phoenix-Ampho cells were provided by Matthias Theobald (University Medical Center Mainz, Mainz, Germany). Patient-derived cell lines were established with informed consent and in accordance with the local ethical guidelines. Melanoma lines were A375, provided by Krishnaraj Rajalingam (University Medical Center Mainz, Mainz, Germany), D03, D05, D10, D14, D17, D18, D22, D28, D41, provided by Chris W. Schmidt (Queensland Institute of Medical Research, Brisbane, Australia), Ma-Mel-66b, Ma-Mel-86b, Ma-Mel-86c, provided by Anette Paschen (Department of Dermatology, University Medical Center Essen, Essen, Germany), MZ7-Mel#1, provided by Thomas Wölfel (University Medical Center Mainz, Mainz, Germany). Glioblastoma cell lines MZ-219-GBM, MZ-221-GBM, MZ-222-GBM, MZ-257-GBM, MZ-304-GBM, MZ-373-GBM, MZ-415-GBM, MZ-416-GBM, MZ-483-GBM and lung cancer cell lines LC-MZ-1, LC-MZ-6, LC-MZ-16 were provided by Sigrid Horn (University Medical Center Mainz, Mainz, Germany). Lung cancer cell line NW-BC-11 was provided by Elke Jäger (Northwest Hospital, Frankfurt/Main, Germany). Renal cell cancer cell lines RCC 1170, RCC 1774, RCC 1795, RCC 1846, RCC 1851, RCC 1879, RCC 1940, RCC 1973, RCC FM-KOT, RCC PB5, RCC PB6, RCC PB7 were provided by Barbara Seliger (University Medical Center Halle, Halle/Saale, Germany). Phoenix-Ampho cells were maintained in DMEM (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal calf serum (FCS; PAN Biotech, Aidenbach, Germany), 1% Penicillin/Streptomycin, 1% L-Glutamine (Sigma-Aldrich, St. Louis, MO, USA) and 25 mM HEPES buffer (Lonza, Basel, Switzerland). All other cell lines were maintained in RPMI (Thermo Fisher Scientific) supplemented with 10% FCS and 1% Penicillin/Streptomycin. T cells were maintained in Panserin 413 medium (PAN Biotech, Aidenbach, Germany) supplemented with 10% human serum (kindly provided by the blood bank of the University Medical Center Mainz, Mainz, Germany) and 1% Penicillin/Streptomycin. All cells were kept at 37°C in a humidified atmosphere.

The T-cell clones 11C/73 and 2C/165 were weekly restimulated with irradiated (100 Gy) antigen-expressing tumor cells (Ma-Mel-86c) at a stimulator-to-T cell ratio of 1:10 in the presence of 250 IU/mL IL-2 (Novartis, Basel, Switzerland).

PBMCs were isolated by Ficoll gradient centrifugation from buffy coats of healthy donors obtained from the blood bank of the University Medical Center of the Johannes-Gutenberg university Mainz. CD4⁺ or CD8⁺ T cells were isolated from PBMCs by positive selection using magnetic bead separation (Milteny, Bergisch Gladbach, Germany) and stimulated for 48 h with the anti-CD3 mAb OKT3 (30 ng/mL) (Milteny) in the presence of 600 IU/ml IL-2 (Novartis) and the irradiated (100 Gy) cells of the flow-through of the isolation procedure.

Stimulated CD4⁺ or CD8⁺ T cells were washed twice with PBS and once with Opti-MEM (Thermo Fisher Scientific) and were subsequently resuspended in 100 µL supplemented Nucleofector Solution (Lonza, Basel, Switzerland) per cuvette at a concentration of 4-5e6 T cells/100 µL. In parallel, RNPs targeting the TRAC (TCAGGGTTCTGGATATCTGT) and TRBC1/2 [TGGCTCAAACACAGCGACCT (20)] locus were formed. First, 0.6 µL crRNA (200 µM; IDT, Coralville, IA, USA) and 0.6 µL tracrRNA (200 µM; IDT) were duplexed per electroporation cuvette for 5 min at 95°C. Thereafter, 1.7 µL of Cas9 protein (61 µM; IDT) and 2.1 µL Opti-MEM was added and RNPs were generated by incubation for 20 minutes at room temperature. Thereafter, the T-cell suspension was mixed with the RNPs and 1 µL electroporation enhancer (100 µM; IDT), transferred to an 100 µL electroporation cuvette and immediately electroporated using program T-023 on a Nucleofector 2B device (Lonza). Next, 500 µL prewarmed T-cell medium was added to electroporated T cells and cells were rested for 5 min at 37°C. Subsequently, T cells were transferred to 24-well plates and cultured for 7 days in the presence of IL-2 (600 IU/mL) prior to retroviral transduction.

In brief, Phoenix amphotropic retroviral packaging cells were seeded at a density of 1.3 e6 cells per 100 mm plate. The next day, packaging cells were cotransfected with 5 µg of each of the helper plasmids pCOLT-GALV and pHIT60, and 10 µg of the retroviral pMX vector encoding for the TCR expression constructs using Fugene 6 (Promega, Madison, WI, USA) according to the manufacturer’s instructions. On the following day medium was changed to T cell medium and virus soup was harvested after additional incubation for 16 h by pelleting cellular debris. Subsequently, 1-2 e6 stimulated CRISPR/Cas9-edited T cells were spin infected with 1 mL of virus soup (90 min at 2000 rpm) in the presence of Polybrene (5 μg/ml; Sigma Aldrich) and IL-2 (600 U/mL; Novartis) and were further cultivated for 22 h. On the day after transduction infected T cells were restimulated with anti-CD3/CD28 beads (Milteny Biotec) and were selected with puromycin (1 μg/mL; Sigma Aldrich). After 7 days, transgenic T cells were incubated for 5 min on a magnet to remove anti-CD3/CD28 beads.

TCR expression constructs were generated as previously described (21). HLA-C*07:01, HLA-C*07:02 cDNA, as well as truncated CSPG4 cDNA ending with codon 562 were cloned into pcDNA3.1/V5-His TOPO (Invitrogen, Waltham, MA, USA) using standard cloning techniques as previously described (22). Plasmids were propagated in 10-beta competent E. coli (NEB, Ipswich, MA, USA) and were purified using the Midi or Maxi Plasmid Prep kits from Qiagen (Hilden, Germany).

For surface staining of tumor or T cells, cells were washed once with PBS and subsequently stained for 30 minutes at 4°C in the dark with the indicated surface antibodies diluted in PBS supplemented with 0.1% bovine serum albumin and 2.5 mM EDTA (FACS buffer). The following monoclonal antibodies were used: CD4-PE (clone 13B8.2; Beckman Coulter, Brea, CA, USA), CD8-APC (clone B9.11; Beckman Coulter), CD8-PE (clone B9.11; Beckman Coulter), anti-murine TCR-FITC (clone H57-697; Biolegend, San Diego, CA, USA), anti-human TCR-APC (clone IP26, Biolegend) and CSPG4-PE (REA1041, Milteny). For surface staining of MHC-I, tumor cells were stained by incubation with supernatant of the hybridoma W6/32 (kindly provided by Dr. P. Parham, Departments of Structural Biology and Microbiology and Immunology, Stanford University, Stanford, CA, USA) followed by staining with polyclonal goat-anti-mouse-FITC (Biozol, Eching, Germany). Flow cytometry was performed on a FACS Canto II (BD Bioscience, San José, CA, USA) and data was analyzed using FlowJo (BD Bioscience, Version 10.7).

Cytotoxicity of T-cell clones and transgenic T cells was analyzed by a bioluminescence-based lysis assay as recently described (21). In brief, Ma-Mel-86b and Ma-Mel-86c were engineered to express Firefly luciferase (Fluc) and cocultured with effector cells at an E:T ratio of 10:1 in the presence of 0.15 mg/mL D-Luciferin (Biosynth, St. Gallen, Switzerland). Subsequently, relative luminescence units (RLU) were determined over time in 3 h intervals using a FluoStar Omega plate reader (BMG Labtech, Ortenberg, Germany) and a 10 s integration time. Spontaneous cell death was measured in wells containing target cells only, maximum cell death was induced by exposure of target cells to Digitonin (Sigma) at a concentration of 30 µg/mL. Lysis was calculated using the following equation: lysis [%] = 100*[(spontaneous RLU—test unit RLU)/(spontaneous RLU—maximum RLU)].

IFNγ-ELISpot assays were performed as previously described (21). In brief, 293T (20,000 cells/well) were transfected with cDNA encoding HLA-C*07:01 or HLA-C*07:02 and either cotransfected with cDNA encoding truncated CSPG4 ending with codon 562 or pulsed with the CSPG4-derived peptides HIIFPHGSL (9-mer) or PHIIFPHGSL (10-mer) (both synthesized by Dr. Jan-Wouter Drijfhout, University Medical Center Leiden, The Netherlands) 24h after transfection of HLA-encoding cDNAs. Transfection was performed directly on ELISpot plates (Millipore, Burlington, MA, USA) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer´s recommendations. Parental T-cell clones or transgenic T cells (1000 – 10,000 (c)TCR⁺ cells/well) were then added to (peptide-pulsed) transfectants. In some experiments T cells were directly added to freshly seeded tumor cells (40,000 – 50,000 cells/well). After 20–24 h ELISpot plates were developed and IFNγ spots were visualized with an ImmunoSpot analyzer (Cellular Technology Limited, Cleveland, OH, USA).

Students t-test or one-way ANOVA was used to calculate statistical significance and tests were considered significant when the p value was < 0.05. The peptide concentration necessary to elicit a half-maximal response in ELISpot assays (pEC50) was calculated by performing non-linear regression using a variable slope model with four parameters. Data was analyzed in GraphPad Prism (Version 9.0.0). All experiments were independently performed at least twice.

We started our initial analyses of the HLA-C*07:01-restricted T-cell clones 11C/73 and 2C/165 using the cell lines Ma-Mel-86b and Ma-Mel-86c which had been established from distinct metastases of patient Ma-Mel-86 (10). HLA-C*07:01 is expressed in both cell lines, but in contrast to Ma-Mel-86c cells, Ma-Mel-86b cells lack surface expression of MHC-I due to biallelic loss of B2M [(10) and Figure 1A ]. Both cell lines are CSPG4⁺ as evidenced by surface staining ( Figure 1A ). Co-culture of the T-cell clones 11C/73 and 2C/165 with the HLA-C*07:01⁺CSPG4⁺ cell line Ma-Mel-86c resulted in release of IFNγ, while no relevant IFNγ-production was observed upon co-culture with the MHC-I^-CSPG4⁺ cell line Ma-Mel-86b in IFNγ-ELISpot assays ( Figures 1B, C ). The killing capacity of the T-cell clones was assessed in a bioluminescence-based cytotoxicity assay. Both T-cell clones efficiently lysed the HLA-C*07:01⁺CSPG4⁺ cell line Ma-Mel-86c/Fluc but not the control cell line Ma-Mel-86b/Fluc ( Figure 1D ). To compare functional avidity, we employed the T-cell clones against HLA-C*07:01 cDNA-transfected 293T cells loaded with titrated amounts of the CSPG4:554-562 peptide (9-mer). As shown in Figure 1E , both T-cell clones recognized peptide-loaded target cells in a concentration-dependent manner, displayed comparable and high functional avidities, and similar maximum IFNγ release. We found the concentration of peptide necessary for a half maximal response (pEC50) for T-cell clone 2C/165 to be slightly but not significantly lower compared to T-cell clone 11C/73 ( Figure 1F ). While we focused our analysis on the 9-mer CSPG4:554-562 peptide, we found that the 10-mer CSPG4 peptide CSPG4:553-562 was also recognized by both T-cell clones albeit with a lower functional avidity as compared to the 9-mer ( Supplementary Figure 1 ). For further validation and as a reference point for later experiments, we also extended our analyses to recognition of endogenously processed antigen and co-transfected 293T cells with both HLA-C*07:01 cDNA and titrated amounts of a CSPG4 cDNA fragment containing the peptide-coding region (codon 1-562). Both T-cell clones recognized CSPG4:1-562 cDNA-transfected targets comparably and in a dose dependent manner ( Figure 1G ).

Among HLA-C alleles, HLA-C*07:02 displays a high homology to HLA-C*07:01 (99.4%) and only differs in two residues (90 and 123) (23). To test whether the immunogenic CSPG4 peptides can also bind to HLA-C*07:02, we pulsed wildtype K562 cells (HLA-C*03:04, 05:01) or K562 cells stably expressing HLA-C*07:01 or HLA-C*07:02 with CSPG4:553-562 (9-mer), CSPG4:554-562 (10-mer) or a control peptide presented by HLA-B*15:01 [HERPUD1^G161S:154-162 (22)] and analyzed MHC-I levels by flow cytometry. Both CSPG4:553-562 and CSPG4:554-562 increased the expression levels of MHC-I in K562/HLA-C*07:01 but not in K562 wildtype or K562/C*07:02 cells. ( Figures 2A, B ). Next, we co-cultured the T-cell clones with either 293T cells transfected with HLA-C*07:01 or HLA-C*07:02 cDNA and loaded with the CSPG4:554-562 (9-mer) or CSPG4:553-562 (10-mer) peptide ( Figure 2C ) or cotransfected with CSPG4:1-562 cDNA ( Figure 2D ) or co-cultured them with HLA-C*07:02⁺CSPG4⁺ melanoma cell lines ( Figure 2E ). In contrast to HLA-C*07:01-expressing target cells, we did not observe reactivity of the T-cell clones against HLA-C*07:02⁺ target cells suggesting no relevant overlap in CSPG4 peptide binding among these highly homologous HLA-C alleles.

Given the high functional avidity and confirmed specificity of both T-cell clones, we next employed the 5’-RACE method (24) to identify their TCR sequences. Using this approach, two αβTCRs were amplified from cDNA and Sanger Sequencing revealed distinct clonotypes. The combination of V(D)J alleles and CDR3 sequences are summarized in Table 1 .

The αβTCR sequences of both clonotypes were subsequently cloned into retroviral expression constructs shown in Figure 3A to test whether PBMC-derived TCR-transduced CD8⁺ T cells can be redirected against CSPG4. Notably, our TCR-expression-constructs incorporated optimizations aimed to reduced mispairing of transgenic TCRs with the endogenous TCRs such as replacing the human constant regions with their murine counterparts (chimerized TCR; cTCR). To further reduce any residual TCR mispairing, expression of the endogenous TCRs was abrogated by electroporation of CRISPR/Cas9 ribonucleoproteins (RNPs) targeting the TRAC and TRBC1/2 loci. Following human TCR (hTCR) knockout, TCR-KO CD8⁺ T cells were retrovirally transduced with the cTCR expression constructs and were expanded by weekly stimulation with anti-CD3/CD28 beads in the presence of IL-2. Of note, in addition to transducing the parental αβTCR gene-pairs, we also combined the alpha and beta chains of the 11C/73- and 2C/165-TCRs in a cross-over fashion to assess for potential TCR hemichain dominance ( Figure 3B ). As shown in Figure 3C , transgenic T cells highly expressed the transgenic cTCRs including the cross-over cTCRs as evidenced by staining for the murine constant region while retaining a TCR-knockout phenotype. After validating cTCR surface expression, transgenic T cells were tested in IFNγ-ELISpot assays against HLA-C*07:01 cDNA-transfected 293T cells pulsed with titrated amounts of CSPG4:554-562 peptide. Transgenic T cells expressing the 11Cαβ and 2Cαβ cTCR constructs derived from the respective parental 11C/73 and 2C/165 T-cell clone displayed similar maximal IFNγ-spot counts at high concentration of the CSPG4:554-562 peptide, but the 2Cαβ cTCR showed a higher functional avidity ( Figure 3D ). Interestingly, CD8⁺ T cells transduced with the cross-over construct 2Cα-11Cβ also recognized CSPG4:554-562 peptide-pulsed targets albeit with lower functional avidity despite similar maximal IFNγ-spot counts at high peptide concentration. These observations were also reflected in different pEC50 values which, however, did not reach statistical significance ( Figure 3E ). In contrast, no reactivity was observed with 11Cα-2Cβ cTCR-transduced CD8⁺ T cells even at high peptide concentrations. None of the employed T-cell populations released relevant amounts of IFNγ alone or in the presence of unpulsed HLA-C*07:01 + 293T cells confirming specificity of the T-cell responses (not shown).

Next, we extended our analyses to recognition of endogenously processed antigen and co-transfected 293T cells with both HLA-C*07:01 and two different amounts of CSPG4:1-562 cDNA. CD8⁺ T cells transduced with the 11Cαβ and 2Cαβ TCR constructs specifically recognized CSPG4:1-562 cDNA-transfected targets ( Figure 3F ). Mirroring the peptide titrations experiments, we also observed reactivity of the cross-over 2Cα-11Cβ cTCR albeit only at 300 ng of CSPG4:1-562 cDNA per well. Again, spot counts of the 11Cα-2Cβ cTCR-transduced CD8⁺ T cells were not above background level. Analogous results were also seen when we tested the T-cell populations against HLA-C*07:01⁺CSPG4⁺ Ma-Mel-86c cells ( Figure 3G ). Finally, we assessed killing capacity of the TCR-transgenic T cells in a bioluminescence-based cytotoxicity assay. CD8⁺ T cells transduced with either parental or the 2Cα-11Cβ cTCR efficiently lysed the HLA-C*07:01⁺CSPG4⁺ cell line Ma-Mel-86c/Fluc but not the control cell line Ma-Mel-86b/Fluc ( Figure 3H ). Lysis rates of 11Cα-2Cβ cTCR-transduced CD8⁺ T cells were not above control T cells transduced with empty vector. Thus, these results validated both functionality and specificity of our TCR expression constructs and suggested that either the 2Cαβ has a dominant TCR alpha or that the 11Cαβ TCR has a dominant TCR beta chain regarding epitope recognition.

To analyze CD8 co-receptor dependence of the cTCR constructs for T cell activation, we generated PBMC-derived transgenic CD4⁺ T cells. The endogenous TCRs were again deleted by electroporation of CRISPR/Cas9 RNPs targeting both the TRAC and TRBC1/2 loci ( Figure 4A ). We observed functionality of the 11Cαβ-cTCR, the 2Cαβ-cTCR and the cross-over construct 2Cα-11Cβ cTCR when expressed in CD4⁺ T cells and tested against HLA-C*07:01 cDNA-transfected 293T cells loaded with titrated amounts of CSPG4:554-562 peptide ( Figure 4B ). In contrast to our observations with transgenic CD8⁺ T cells, even at the highest peptide concentration tested, maximum IFNγ release of the cross-over 2Cα-11Cβ-cTCR construct was lower compared to the 11Cαβ- and 2Cαβ-cTCRs. The cross-over construct 11Cα-2Cβ was, again, non-functional. These results were also mirrored when transgenic CD4⁺ T cells were tested against 293T cells co-transfected with both HLA-C*07:01 and CSPG4:1-562 cDNA ( Figure 4C ). However, in contrast to our observations with transgenic CD8⁺ T cells, CSPG4⁺ HLA-C*07:01⁺ Ma-Mel-86c cells were only recognized by the parental cTCRs ( Figure 4D ). Collectively these results suggested that T-cell activation mediated by the parental cTCRs was co-receptor independent. In contrast, the functional cross-over construct showed co-receptor dependence when tested against Ma-Mel-86c cells presumably reflecting its lower functional avidity.

According to the TCGA dataset high CSGP4 mRNA expression is seen in melanoma, glioblastoma, renal and lung cancer ( Figure 5A ). We therefore tested CSPG4 surface expression in a panel of cancer cell lines derived from these entities (n=37). CSPG4 surface expression (above an arbitrary 5% threshold) was observed in 12/12 (100%) melanoma, 5/9 (56%) glioblastoma, 1/4 (25%) lung and 0/12 (0%) renal cancer cell lines ( Figure 5B ). Based on these results we selected a panel of CSPG4⁺ cell lines shown in Figure 5C . CSPG4⁺ but HLA-C*07:01^- cells lines were retrovirally transduced with HLA-C*07:01 (MZ-222-GBM, MZ-257-GBM and MZ-LC-16). We then tested the parental T-cell clones or CD8⁺ T cells expressing the indicated transgenic cTCR for recognition of our target panel in IFNγ-ELISpot assays. We observed high IFNγ spot counts with both T-cell clones against all cell lines tested with T-cell clone 2C/165 showing higher spot counts against D10, D41 and MZ-GBM-222 compared to T-cell clone 11C/73. Transgenic CD8⁺ T cells expressing the 11Cαβ- or 2Cαβ-cTCR elicited IFNγ spot formation above background levels when co-cultured with the endogenously HLA-C*07:01⁺ melanoma cell lines D10, D41, the glioblastoma cell line MZ-257-GBM/HLA-C*07:01 and the lung cancer cell line MZ-LC-16/HLA-C*07:01 ( Figure 5D ). 2Cαβ-cTCR-transduced T cells additionally recognized the glioblastoma cell line MZ-222-GBM/HLA-C*07:01 and the endogenously HLA-C*07:01⁺ glioblastoma cell line MZ-483-GBM. Of note, we also observed reactivity of the cross-over 2Cα-11Cβ-cTCR against the transfected cell lines MZ-257-GBM/HLA-C*07:01 and MZ-LC-16/HLA-C*07:01, while no reactivity was observed with the 11Cα-2Cβ-cTCR against any of the tested cell lines. Overall, IFNγ spot counts elicited by the 2Cαβ-cTCR were comparable to the IFNγ spot counts of both parental T-cell clones.