Development of chimeric antigen receptors targeting the γδ T cell receptor Authors: Lauren C Cutmore, Norris Lam, Christina Amatya, Katherine A Weissler, Stephen M Hewitt, Danielle A Natrakul, James N Kochenderfer Published: 2026 ABSTRACT AbstractT cell malignancies expressing γδ T cell receptors (TCRs) have poor prognoses. No approved treatments specifically target γδ T cell malignancies. To address this deficiency, we aimed to develop chimeric antigen receptors (CARs) targeting the γδ TCR. We generated a panel of CARs with variations in antigen-recognition domains, hinge and transmembrane (HTM) domains, and costimulatory domains. We expressed CARs in human T cells by γ-retroviral transduction. By assessing CAR T cell cytokine release in response to γδ TCR-expressing target cells, we identified two single-chain variable fragments (scFvs) with superior antigen specificity for the γδ TCR. CARs containing one of the two scFvs, which was designated maximal gamma-delta-long (MGDL), exhibited higher transduction efficiency. We compared three MGDL-containing CARs: MGDL-28Z (CD28 HTM and costimulatory domains), MGDL-CD828Z (CD8α HTM domains and CD28 costimulatory domain), and MGDL-CD8BBZ (CD8α HTM domains and 4-1BB costimulatory domain). Levels of antigen-specific cytokine release were higher for T cells expressing MGDL-28Z compared with the other MGDL CARs. We tested T cells expressing the three MGDL-containing CARs in two murine γδ T cell leukemia treatment models. MGDL-28Z-expressing T cells caused the most anti-leukemia activity and longest mouse survival. These findings support testing MGDL-28Z-expressing T cells in a clinical trial.Keywords: CAR, immunotherapy, chimeric antigen receptor, t cell, γδ T cells, leukemia, lymphoma, cancer Introduction Based on the T cell receptor (TCR) expressed, T lymphocytes can be classified into two populations, αβ and γδ.1,2 γδ T cells are much less common than αβ T cells and constitute less than 10% of the total blood T cells.1,3 The γ and δ TCR proteins each contain a variable (V) domain and a constant (C) domain.1,2 There are two main versions of the γ constant domain and one version of the δ constant domain.4,5 Unlike the αβ TCR, antigen recognition by the γδ TCR does not require antigen presentation by major histocompatibility complex (MHC) molecules.1,2,6 γδ T cells recognize a range of targets, including but not limited to, MHC class I chain-related proteins, CD1c-expressing cells, and several bacteria-associated phosphoantigens.1,6,7 A major function of γδ T cells is early response against infections.6,7 Several malignancies can arise from γδ T cells.8 Hepatosplenic T cell lymphoma (HSTCL) is a rare malignancy in which the lymphoma cells usually express the γδ TCR.8,9,10 The prognosis of HSTCL is poor, with median overall survivals ranging from 4 to 28.3 months in four studies.9,10 Primary cutaneous γδ T cell lymphoma is a rare cutaneous lymphoma with a 5-year overall survival of 10%–20%.8 T cell acute lymphoblastic leukemia (T-ALL) expresses the γδ TCR in approximately 9%–13% of cases.11,12 γδ T-ALL has a better prognosis than other γδ T cell malignancies; the long-term overall survival of γδ T-ALL ranges from 50% to 66.7%.11,12 Monomorphic epitheliotropic intestinal T cell lymphoma is a rare γδ T cell gastrointestinal lymphoma with a median overall survival of 7–7.5 months.8,13 There are currently no approved therapies that specifically target γδ T cell malignancies, so a need for new therapies exists. There is evidence for γδ T cell involvement in autoimmune diseases such as psoriasis and multiple sclerosis.7,14 Chimeric antigen receptors (CARs) are fusion proteins that include an antigen-recognition domain, hinge and transmembrane (HTM) domains, a costimulatory domain such as 4-1BB or CD28, and a T cell activation domain such as CD3ζ.15,16,17,18,19 The choice of CAR domains determines CAR function.16,19,20,21,22,23,24,25 Effective CAR antigen-recognition domains have most often been single-chain variable fragments (scFvs) derived from monoclonal antibodies, but antigen-recognition domains can also be derived from other proteins such as nanobodies.19,23,26,27,28,29 Even when scFvs bind to the same epitope, changes in scFv variable domain amino acid sequences can affect CAR function by causing differences in binding affinity, CAR expression levels, and possibly other factors.30,31,32 The linker is a short peptide sequence that connects the variable domains within the scFv and provides flexibility.33,34 The choice of scFv linker can impact the function of CARs.22,35 One report described increased interleukin (IL)-2 release and reduced activation-induced cell death when the linker was lengthened.22 In another study, shorter linkers caused homodimerization of CARs and induced antigen-independent signaling that was beneficial to CAR function.35 The CAR antigen-recognition domain is connected to cytoplasmic domains by the HTM domains.20,21,36 Changes to the HTM domains can affect CAR-expressing T cell attributes such as cytokine release and susceptibility to activation-induced cell death.20,37 The most commonly used CAR costimulatory domains are CD28 and 4-1BB.16,17,21,38 CD28 induces rapid T cell activation, high levels of proliferation, and rapid substantial release of antigen-specific cytokines.16,38 In contrast, 4-1BB signaling induces slower activation, lower levels of T cell exhaustion markers, and possibly prolonged in vivo persistence.16,38,39 CARs with either a CD28 or 4-1BB costimulatory domain have been effective at treating B cell malignancies in humans.16,19,26 The effectiveness of CAR T cell therapies against B cell malignancies provides a rationale for developing CARs targeting γδ T cell malignancies.19,26 We confirmed the restricted expression pattern of the γδ TCR, and we identified CAR scFvs that specifically recognized the γδ TCR constant domains. Recognition of the constant domains gives these CARs the ability to recognize all γδ TCRs. We compared CARs to select the CAR with the most antigen specificity and the greatest in vivo therapeutic effectiveness. Results γδ TCR expression The γδ TCR is a promising target for CAR T cells because its expression should be restricted only to γδ T cells. Immunohistochemistry staining of all major organs confirmed γδ TCR expression only on scattered lymphocytes (Table S1). Design of CARs with 4-1BB costimulatory domains To target the γδ TCR, we utilized amino acid sequences of antibodies specific for the γδ TCR to design scFvs. Using variable domains from two antibodies led to CARs that specifically recognized the γδ TCR. In addition, CARs were constructed with variable domains from eight other antibodies, but these CARs did not specifically recognize the γδ TCR and were not further studied (Table S2). Two linkers designated 218 and (G4S)4 were assessed within the scFvs of CARs.22,33 Four scFvs that specifically recognized the γδ TCR were identified: maximal gamma-delta (MGD), MGDL, gamma delta (gd), and gdL. An “L” in scFv names stands for “long” and indicates use of the (G4S)4 linker, which is four repeats of a sequence of four glycines and one serine. scFv names without the “L” indicate use of the 218 linker. Four CARs containing CD8α HTM domains, 4-1BB costimulatory domains, and CD3ζ activation domains were constructed and assessed: MGD-CD8BBZ, MGDL-CD8BBZ, gd-CD8BBZ, and gdL-CD8BBZ (Figure 1A). CARs were encoded by a γ-retroviral vector called mouse stem cell virus-based splice-gag vector 1 (MSGV1).18 All CARs were expressed on the surface of human T cells following transduction (Figures 1B–1E; gating strategy in Figure S1). The most prominent difference in transduction efficiency among the different CARs was a higher frequency of CAR+ T cells with MGD-based scFvs versus CARs with gd-based scFvs (Figures 1B–1E). Design and function of 4-1BB-containing CARs (A) Diagrams of CARs with MGD or gd scFvs. The scFvs had either 218 or (G4S)4 linkers. The CARs contain CD8α hinge and transmembrane domains, 4-1BB costimulatory domains, and CD3ζ domains. MGD or gd light chain and MDG or gd heavy chain on the diagrams refer to the variable domains. (B–E) CAR expression was assessed 5 days after transduction by flow cytometry. Anti-218L was used to detect MGD-CD8BBZ and gd-CD8BBZ. Anti-G4SL was used to detect MGDL-CD8BBZ and gdL-CD8BBZ. Plots are gated on live CD3+ lymphocytes. Plots from one representative donor are shown. CAR expression on (C) CD3+ (D) CD3+CD4+, and (E) CD3+CD8+ cells was determined by flow cytometry. CAR expression on CD3+ cells was determined on live CD3-gated lymphocytes. CAR expression on CD4+ or CD8+ cells was determined on live CD3+CD4+ or CD3+CD8+ lymphocytes. Data are presented as the mean + SEM (n = 5 independent experiments). Statistics were by repeated measures one-way ANOVA with Tukey’s multiple comparisons test; ∗p < 0.05 and ∗∗p < 0.01. (F) T cells were cultured overnight with the indicated target cells. IFNγ ELISA was performed on culture supernatants. The ratio of IFNγ target cells to T cells alone was calculated by dividing IFNγ released from CAR T cells co-cultured with the indicated target cells by IFNγ released from CAR T cells alone. (G) T cells were cultured overnight with the indicated target cells. IFNγ ELISA was performed on culture supernatants. GC1DC-CEM cells express only a γδ TCR constant domain, whereas g9d2-CEM express a full γδ TCR. NGFR-CEM are negative-control target cells. Untransduced T cell IFNγ levels are too low to be seen on the graph. For both (F and G), results are normalized for the %CAR+ T cells. Data are presented as the mean + SEM (n = 5 different donors). For each target cell type, we made two statistical comparisons: MGD-CD8BBZ versus MGDL-CD8BBZ and gd-CD8BBZ versus gdL-CD8BBZ. ∗p < 0.05 by 2-tailed ratio paired t test. Assessment of γδ-TCR-specific cytokine release by 4-1BB-containing CARs We performed functional assays of 4-1BB-containing CARs by using 5 different target cell lines that expressed the γδ TCR (Figures S2A and S2B). Loucy, MOLT-13, and BE-13 were derived from human malignancies that naturally expressed the γδ TCR. The GC1DC-CEM cell line was transduced to express only the constant domains of the γδ TCR, and the g9d2-CEM cell line was transduced to express a complete Vγ9Vδ2 TCR. We quantified antigen-recognition specificity of MGD-CD8BBZ, MGDL-CD8BBZ, gd-CD8BBZ, and gdL-CD8BBZ by dividing interferon-γ (IFNγ) released from CAR-expressing T cells co-cultured with γδ TCR+ target cells by IFNγ released from CAR-expressing T cells cultured alone. This measurement is referred to as the ratio of IFNγ target cells to T cells alone. By this metric, the CARs with (G4S)4 linkers, MGDL-CD8BBZ and gdL-CD8BBZ, exhibited more specific target recognition compared with the CARs with 218 linkers, MGD-CD8BBZ and gd-CD8BBZ (Figure 1F). The higher ratios of IFNγ target cells to T cells alone reached statistical significance for gd-CD8BBZ versus gdL-CD8BBZ with Loucy and MOLT-13 target cells. For all other comparisons of ratios of IFNγ target cells to T cells alone, there was a consistent trend toward higher specificity for CARs with the (G4S)4 linker versus the 218 linker, but these comparisons did not reach statistical significance. The IFNγ values used to calculate the ratios in Figure 1F are shown in Figure S2C. We also tested anti-γδ CARs for the ability to recognize target cells artificially transduced to express a full γδ TCR or only the constant domains of the γδ TCR. For these experiments, we used GC1DC-CEM that expressed only γδ constant domains without variable domains. All CARs specifically released IFNγ in response to GC1DC-CEM, which demonstrated that the CARs all recognized the constant domains of the γδ TCR (Figure 1G). All CARs also released IFNγ in response to g9d2-CEM, which expressed a full γδ TCR. T cells expressing all CARs released much lower levels of IFNγ when co-cultured with γδ-TCR-negative NGFR-CEM cells compared with γδ-TCR-expressing target cells. This demonstrated the antigen specificity of the CARs. IFNγ release was statistically significantly higher for T cells expressing MGD-CD8BBZ compared with T cells expressing MGDL-CD8BBZ during NGFR-CEM co-culture, and T cells expressing gd-CD8BBZ released more IFNγ compared with T cells expressing gdL-CD8BBZ during NGFR-CEM co-culture. Due to the greater antigen specificity of CARs with the (G4S)4 linker versus the 218 linker, we used the (G4S)4 linker in all future iterations of the CARs. Expression and function of CARs with CD28 costimulatory domains We next constructed four additional CARs containing either MGDL or gdL scFvs (Figure 2A). These CARs all contained the (G4S)4 linker. In these experiments, all CARs contained a CD28 costimulatory domain with either CD8α HTM domains (MGDL-CD828Z and gdL-CD828Z) or CD28 HTM domains (MGDL-28Z and gdL-28Z). Design and function of CD28-containing CARs (A–C) (A) Diagrams of MGD and gd-based CARs containing either CD8α hinge and transmembrane domains or CD28 hinge and transmembrane domains are shown. All CARs in this figure had CD28 costimulatory domains. MGD or gd light chain and MDG or gd heavy chain on the diagrams refer to the variable domains. CAR expression on (B) CD4+ and (C) CD8+ T cells was determined by flow cytometry with an antibody against the G4S linker (anti-G4SL). CAR expression was determined after gating on live CD3+ lymphocytes and further gating on CD4+ or CD8+ events. Data are presented as the mean + SEM (n = 4 experiments with different donors). p values were determined by repeated measures one-way ANOVA with Tukey’s multiple comparisons test. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗∗p < 0.0001. Comparisons without asterisks were not statistically significantly different. (D–G) T cells expressing CARs indicated in the legend were cultured overnight with target cells indicated on the x axis. IFNγ (D and E) and IL-2 (F and G) ELISAs were performed on culture supernatants. Cytokine levels were normalized for %CAR+ T cells. Data are presented as the mean + SEM (n = 4 experiments with cells from different donors). Statistical testing was by 2-tailed ratio paired t test: ∗p < 0.05 and ∗∗p < 0.01. All CARs were expressed on the surface of CD4+ (Figure 2B) and CD8+ T cells (Figure 2C). We observed a consistently higher percentage of transduced T cells expressing MGDL CARs versus gdL CARs. For example, CD4+ cells had a higher %CAR+ cells with MGDL-28Z versus gdL-28Z (p < 0.0001), and CD8+ cells also had a higher %CAR+ cells with MGDL-28Z versus gdL-28Z. T cells expressing each of the CD28-containing CARs specifically released IFNγ when co-cultured with γδ TCR+ target cells (Figures 2D and 2E). IFNγ release was higher with CARs containing CD28 HTM domains versus CD8α HTM domains (Figures 2D and 2E). In addition, T cells expressing either MGDL-28Z or gdL-28Z did not recognize a variety of γδ TCR-negative target cells when assessed by IFNγ ELISA (Figure S3). We assessed IL-2 release by T cells expressing the four CD28-containing CARs. A consistent trend toward higher levels of IL-2 release was observed for CARs with CD28 HTM domains versus CD8α HTM domains, but this trend only reached statistical significance for two of the CAR plus target cell combinations assessed (Figures 2F and 2G). Cytokine release was higher for CARs with CD28 HTM domains versus CD8α HTM domains (Figures 2D–2G). This suggested stronger T cell activation by the CARs with CD28 HTM domains. However, we continued to carefully assess MGDL-CD828Z in case further studies uncovered an advantage for this CAR. Because of the consistently higher transduction efficiency of CARs with an MGDL scFv versus a gdL scFv (Figures 2B and 2C), we focused further work on CARs with the MGDL scFv-binding domain. Design and expression of three CARs incorporating the MGDL scFv We next compared the MGDL scFv in three CAR designs, MGDL-CD8BBZ, MGDL-CD828Z, and MGDL-28Z (Figure 3A). All CARs were expressed on the surface of CD4+ (Figure 3B) and CD8+ (Figure 3C) T cells. The CAR expression median fluorescence intensities (MFIs) are shown in Figure S4. MGDL-CD8BBZ-expressing T cells had higher CD4:CD8 ratios compared with T cells expressing CARs with CD28 costimulatory domains (Figure 3D). In vitro function of MGDL CARs with different designs (A–C) Diagrams of MGDL scFv-containing CARs. Light chain and heavy chain refer to variable domains. Expression of CARs on (B) CD4+ and (C) CD8+ live CD3+ T cells was determined by flow cytometry 5 days after transduction. CARs were detected with an antibody against the G4S linker (anti-G4SL). Statistical testing was by repeated measures one-way ANOVA with Tukey’s multiple comparisons test. (D) CD4-to-CD8 ratios of CD3+CAR+ T cells expressing the indicated CARs were determined by flow cytometry. Statistical analysis was repeated measures Friedman test with Dunn’s multiple comparisons test. For (B–D), n = 6 experiments with cells from different donors, ∗p < 0.05. Comparisons without asterisks were not statistically significant. (E and F) T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z were co-cultured with either g9d2-CEM-luc-Thy1.1 or NGFR-CEM-luc-Thy1.1 for 4 h. Degranulation of (E) CD4+ and (F) CD8+ T cells was assessed by flow cytometry for CD107a expression. %CD107a+ events are from flow cytometry plots gated on live CD3+CD4+ or CD3+CD8+ cells. The reported values are %CD107a+ events (degranulation) of CAR T cells co-cultured with g9d2-CEM-luc-Thy1.1 target cells minus %CD107a+ events of the same type of CAR T cells co-cultured with γδ-negative NGFR-CEM-luc-Thy1.1 target cells. Results were normalized for %CAR+ T cells (n = 3 experiments with cells from different donors). No statistically significant difference between the CARs in %CD107a+ events was found. (G) Five days after transduction, T cells transduced with CARs indicated on the x axis were stained for γδ TCR, CD3, CD4, and CD8. Percentages of CD3+CD4−CD8− cells that expressed the γδ TCR are shown (n = 3 independent experiments with cells from different donors). Statistical testing was by 2-tailed paired t tests; ∗p < 0.05. (H) CAR-transduced or untransduced T cells as indicated in the color legend were cultured overnight with the indicated target cells, and an IFNγ ELISA was performed on the culture supernatants. Cytokine levels were normalized for %CAR+ T cells (n = 5 experiments with cells from different donors). All data are presented as the mean + SEM. Statistical testing was by 2-tailed ratio paired t tests; ∗p < 0.05 and ∗∗p < 0.01. Untransduced T cell IFNγ levels are too low to be seen. Functional comparison of MGDL-CD8BBZ, MGDL-CD828Z, and MGDL-28Z Following a 4-h co-culture of CAR T cells with g9d2-CEM-luc-Thy1.1 or NGFR-CEM-luc-Thy1.1 target cells, surface expression of CD107a was measured on CAR T cells as a readout of degranulation. Both CD4+ and CD8+ T cells specifically upregulated CD107a in response to γδ TCR+ target cells (Figures 3E and 3F). As expected, levels of degranulation were higher for CD8+ versus CD4+ T cells. Compared with in vitro cultures of T cells expressing the anti-CD19 CAR Hu19-CD828Z, cultures of MGDL-containing CAR T cells had statistically significantly lower percentages of γδ TCR-expressing cells within CD3+CD4−CD8− populations (Figure 3G). This demonstrates that the MGDL-containing CAR T cells specifically eliminated γδ T cells from the cultures. We cultured primary human blood γδ T cells for 9 days to increase the number of cells (Figure S5A). We then co-cultured the primary γδ T cells overnight with autologous untransduced T cells or autologous T cells expressing MGDL-28Z, gdL-28Z, or a negative-control anti-CD19 CAR Hu19-CD828Z. MGDL-28Z and gdL-28Z CAR T cells specifically released IFNγ in response to autologous primary γδ T cells (Figure S5B). MGDL-28Z CAR T cells released statistically significantly more IFNγ than negative-control Hu19-CD828Z-expressing T cells in response to the autologous primary γδ T cells (Figure S5B). To demonstrate the elimination of primary γδ T cells by CAR T cells, we cultured primary blood γδ T cells for 7 days to increase cell numbers, and then we co-cultured these primary γδ T cells with autologous untransduced T cells or autologous T cells expressing MGDL-28Z, gdL-28Z, or Hu19-CD828Z. After 24 h of co-culture, we assessed γδ T cells by flow cytometry (Figure S6A). γδ T cells were depleted from cultures containing MGDL-28Z or gdL-28Z CAR T cells but remained in cultures containing negative-control Hu19-CD828Z CAR T cells or untransduced T cells (Figure S6). T cells expressing each of the three MGDL-containing CARs released IFNγ in response to GC1DC-CEM and g9d2-CEM but showed minimal IFNγ release in response to NGFR-CEM. With GC1DC-CEM and g9d2-CEM target cells, IFNγ release was statistically significantly higher by T cells expressing the CD28-containing MGDL CARs compared with MGDL-CD8BBZ (Figure 3H). Next, we assessed IFNγ release in response to cell lines that naturally express the γδ TCR (Figures 4A–4C). T cells expressing MGDL-28Z consistently released more IFNγ compared with T cells expressing either MGDL-CD8BBZ or MGDL-CD828Z when co-cultured with LOUCY, MOLT-13, or BE-13 (Figures 4A–4C). Similarly, IL-2 release was consistently higher for T cells expressing MGDL-28Z compared with T cells expressing either MGDL-CD8BBZ or MGDL-CD828Z when T cells were co-cultured with target cells naturally expressing the γδ TCR (Figures 4D–4F). Cytokine production and cytotoxicity of MGDL CARs T cells were cultured overnight with target cells, and IFNγ (A, B, and C) and IL-2 (D, E, and F) ELISAs were performed on the culture supernatants. Target cells were LOUCY (A and D), MOLT-13 (B and E), and BE-13 (C and F). Cytokine levels were normalized for %CAR+ T cells. Data are presented as the mean + SEM (n = 5 experiments with cells from different donors). Statistical analysis was by 2-tailed ratio paired t test; ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. (G–J) T cells transduced with MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z were labeled with CFSE and co-cultured with either g9d2-CEM-luc-Thy1.1 cells or NGFR-CEM-luc-Thy1.1 cells for 4 days. Fold changes in the numbers of (G) CD4+CAR+ T cells or (H) CD8+CAR+ T cells occurring between days 0 and 4 of co-culture with g9d2-CEM-luc-Thy1.1 cells are shown. Statistics were by ratio paired t test; ∗∗p < 0.01. At the end of the 4-day co-culture of the same cells as in (G and H), the CFSE MFIs of the (I) CD4+CAR+ T cells and (J) CD8+CAR+ T cells were assessed. The CFSE MFI ratios of T cells co-cultured with g9d2-CEM-luc-Thy1.1 versus NGFR-CEM-luc-Thy1.1 (CFSE MFI g9d2/NGFR) were calculated. Lower CFSE MFI g9d2/NGFR indicates more γδ-TCR-specific proliferation. Data are presented as the mean + SEM (n = 3 independent experiments with cells from different donors). Statistical analysis was by repeated measures one-way ANOVA with Tukey’s multiple comparisons test; ∗p < 0.05. (K) T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, MGDL-28Z, or negative-control CAR Hu19-CD828Z were tested in a 4-h cytotoxicity assay with the indicated target cells. Points represent mean ± SEM. %cytotoxicity of replicate wells from one representative example of two experiments. We assessed antigen-specific proliferation of T cells expressing each of the MGDL scFv-containing CARs. T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) and co-cultured with irradiated target cells for 4 days. γδ TCR+ g9d2-CEM-luc-Thy1.1 cells were used as the positive target cell line, and NGFR-CEM-luc-Thy1.1 cells were used as the negative-control cell line. At the end of proliferation assays, we quantified CAR T cells that were co-cultured with g9d2-CEM-luc-Thy1.1 cells. There was a trend toward a higher fold increase in the number of CD4+ T cells expressing MGDL-28Z versus MGDL-CD8BBZ, but this trend did not reach statistical significance (Figure 4G). In the same proliferation assays, the fold change in CAR+ T cell number was higher in a statistically significant manner for CD8+ T cells expressing either of the CD28-containing CARs versus MGDL-CD8BBZ (Figure 4H). The CFSE MFIs of CAR T cells were measured by flow cytometry to evaluate the dilution of CFSE, which indicates T cell proliferation. We calculated the CFSE MFI ratios of CAR T cells after co-culture with g9d2-CEM-luc-Thy1.1 versus NGFR-CEM-luc-Thy1.1 (CFSE MFI g9d2/NGFR). Lower CFSE MFI g9d2/NGFR values indicate higher levels of γδ TCR-specific proliferation. For CD4+ CAR T cells, there were statistically significantly higher levels of γδ-TCR-specific proliferation for both CD28-containing CARs compared with MGDL-CD8BBZ (Figure 4I). For CD8+ CAR T cells, there were non-statistically significant trends toward higher levels of γδ-TCR-specific proliferation for both CD28-containing CARs compared with MGDL-CD8BBZ (Figure 4J). T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z all specifically killed Loucy and MOLT-13 target cells during in vitro cytotoxicity assays (Figure 4K). Different scFv configurations did not improve in vitro function of CARs with MGDL-binding domains We investigated how the order of scFv heavy- and light-chain variable domains affected the function of MGDL-based CARs. We compared six CARs. MGDL-CD8BBZ, MGDL-CD828Z, and MGDL-28Z contained an scFv with the light-chain variable domain N terminal to a (G4S)4 linker that was followed by the heavy-chain variable domain. We designed CARs that were identical to MGDL-CD8BBZ, MGDL-CD828Z, and MGDL-28Z except they contained an scFv with the heavy-chain variable domain N terminal to a (G4S)4 linker that was followed by the light-chain variable domain (H1st). These CARs were designated MGDL-CD8BBZ H1st, MGDL-CD828Z H1st, and MGDL-28Z H1st (Figure S7A). All six CARs were expressed on the T cell surface after T cells were transduced with γ-retroviruses encoding the CARs (Figures S7B and S7C). MGDL-28Z H1st was associated with a lower percentage of CAR+ T cells than all other CARs with CD4+ T cells; in addition, MGDL-28Z H1st had a lower percentage of CAR+ T cells than the other CD28-containing CARs with CD8+ T cells (Figures S7B and S7C). We compared the release of IFNγ and IL-2 by T cells expressing each H1st CAR versus each CAR with the same design except the heavy-chain variable domain being C terminal to the scFv linker (Figure S8). The only statistically significant difference in cytokine release was higher IFNγ release for T cells expressing MGDL-CD8BBZ H1st versus T cells expressing MGDL-CD8BBZ when T cells were co-cultured with BE-13. The H1st versions of MGDL-containing CARs lacked advantages in transduction efficiency or cytokine release when compared with the designs with the light-chain variable domain N terminal to the scFv linker, so we did not conduct further experiments with the H1st versions. Diagrams of all anti-γδ CARs assessed in this work are in Figure S9. MGDL-28Z CAR T cells have a dose-dependent anti-malignancy effect in vivo Nod-scid common γ-chain knockout (NSG) mice were injected intravenously (i.v.) with MOLT-13luc cells that expressed luciferase for bioluminescence imaging (BLI), (Figure 5A). MGDL-28Z-expressing T cells eliminated MOLT-13luc cells in a dose-dependent manner (Figures 5B and 5C). Forty-two days after CAR T cell infusion, mean tumor burdens measured by BLI were lower for mice treated with 8 × 106 or 2 × 106 CAR T cells versus mice treated with 0.5 × 106 CAR T cells (Figure 5D). MGDL-28Z CAR T cells demonstrate a dose-dependent anti-malignancy effect against MOLT-13luc in vivo (A) Schematic of MOLT-13luc disseminated tumor model. NSG mice were injected intravenously (i.v.) with 10 × 106 MOLT-13 luciferase-expressing cells (MOLT-13luc). After 25 days, varying doses of T cells expressing the MGDL-28Z CAR were injected i.v. or mice were left untreated. Bioluminescence imaging (BLI) was performed every 3 days. (B) BLI images of mice prepared as described in (A) are shown. Units of radiance are photons/second/cm2/steradian (p/sec/cm2/sr). (C) For the same mice as in (B), the mean ± SEM radiance values of each dose level are shown for each time point (n = 5 mice per treatment group). p = 0.0084 by two-way ANOVA for comparison of the entire 8 × 106 and 0.5 × 106 curves. p = 0.0093 by two-way ANOVA for comparison of the entire 2 × 106 and 0.5 × 106 curves. (D) The mean radiance values on day 42 after CAR T cell infusions are shown for each dose level. Data are presented as the mean + SEM (n = 5 mice per treatment group). p values were determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test: ∗∗∗p < 0.001. For (C and D), there were not statistically significant differences in radiance between the 8 × 106 and 2 × 106 groups for the full curves or at day 42 after CAR T cell infusion. T cells expressing MGDL-28Z exhibited more in vivo anti-malignancy activity when compared with T cells expressing other CARs We went on to compare the in vivo function of the three MGDL scFv-containing CARs (Figure 6A). T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, and MGDL-28Z all controlled MOLT-13luc tumor growth relative to untreated mice (Figures 6B and 6C). As measured by BLI, mice treated with T cells expressing MGDL-28Z exhibited a non-statistically significant trend toward lower tumor burdens when compared with mice treated with T cells expressing MGDL-CD8BBZ (Figure 6C). In experiments with MOLT-13luc cells, no mice treated with T cells expressing MGDL-28Z or MGDL-CD8BBZ died before the end of the experiments 39 days after CAR T cell infusion, but 2 of 10 mice that received MGDL-CD828Z-expressing T cells died during the experiments. MGDL CAR T cells demonstrate anti-malignancy activity against MOLT-13luc in vivo (A) Schematic of MOLT-13luc disseminated tumor model. NSG mice were injected intravenously (i.v.) with 10 × 106 MOLT-13luc cells. After 25 days, 8 × 106 T cells expressing the indicated CARs were injected i.v. Mice were imaged every 3 days. (B) For each time point, tumor burdens of groups treated as described in (A) were assessed by BLI. Representative images from one of two independent experiments that used T cells from different donors are shown. (C) Two experiments with T cells from different donors were performed as described in (A). There were five mice per group in each experiment. The graph shows combined data from both experiments (n = 10 for all groups). Radiance values are shown for each time point. Data are presented as the mean ± SEM. The untreated curve ends 15 days after CAR T cell infusion because the values became too high fit on the graph as they continued to increase. The MGDL-CD828Z curve ends 27 days after CAR T cell infusion because this was the last day when all mice were alive in this group. A second MGDL-CD828Z mouse died before the end of the experiment. The MGDL-CD828Z mice died of unknown cause with weight loss, diarrhea, and low tumor burdens. p = 0.0508 by two-way ANOVA for comparison of the entire MGDL-28Z and MGDL-CD8BBZ radiance curves. p < 0.05 was considered statistically significant. All mice in the MGDL-28Z, MGDL-CD8BBZ, and untreated groups were still alive on day 39 when the experiment ended. Next, we assessed the MGDL CARs in the more aggressive disseminated BE-13luc model. NSG mice were injected i.v. with 4 × 106 luciferase-expressing BE-13luc cells (Figure 7A). After 5 days, mice were treated with 8 × 106 CAR T cells. Mice treated with MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z CAR T cells initially cleared their tumor burdens. As expected, T cells expressing the anti-CD19 negative-control CAR Hu19-CD828Z did not cause an anti-tumor effect (Figures 7B and 7C). On day 39 after CAR T cell infusion, the mean tumor burden measured by BLI was lower for mice treated with T cells expressing MGDL-28Z compared with those expressing either MGDL-CD8BBZ or MGDL-CD828Z (Figure 7D). Treatment of BE-13luc with T cells expressing any of the MGDL-containing CARs resulted in statistically significant improvements in survival compared with treatment with Hu19-CD828Z CAR T cells. There was also a statistically significant increase in survival for mice treated with T cells expressing MGDL-CD828Z versus MGDL-CD8BBZ, and mice treated with T cells expressing MGDL-28Z had the longest survival of all treatment groups (Figure 7E). MGDL-28Z CAR T cells demonstrated superior anti-malignancy activity against BE-13luc in vivo (A) NSG mice were injected intravenously (i.v.) with 4 × 106 BE-13luc cells. After 5 days, 8 × 106 T cells expressing the indicated CARs were injected i.v. Mice were imaged every 3 days. All data in this figure are from mice treated with this protocol. (B) BLI assessment of tumor burden for all treatment groups is shown. Images are shown from one of two independent experiments that used T cells from different donors. (C) The mean ± SEM radiance values of mice receiving the indicated CARs are shown for each time point. Data are from two independent experiments with T cells from different donors. There were 10 total mice per group. p < 0.0001 by two-way ANOVA for comparison of the entire MGDL-28Z and MGDL-CD8BBZ curves. p = 0.0016 by two-way ANOVA for comparison of the entire MGDL-28Z and MGDL-CD828Z curves. p < 0.05 was considered statistically significant. (D) Mean radiance values on day 39 were compared for mice receiving T cells expressing each CAR. Day 39 was the last day when all mice in all MGDL-containing CAR groups were still alive. Data are from the same mice as in (C). Data are presented as the mean + SEM (n = 10 mice per treatment group). Statistical testing was by ordinary one-way ANOVA with Tukey’s multiple comparisons test; ∗∗p < 0.01 and ∗∗∗p < 0.001. (E) Kaplan-Meier percent survival graph of mice treated with different CAR T cells. By log rank test, there were statistically significant differences in percent survival between the following MGDL-containing CAR groups: MGDL-CD828Z versus MGDL-CD8BBZ (p = 0.0025), MGDL-28Z versus MGDL-CD8BBZ (p < 0.0001), and MGDL-28Z versus MGDL-CD828Z (p = 0.0009). The survival graph contains data from the same mice as in (C). In all mice engrafted with BE-13luc, tumor progression was observed following treatment with CAR T cells. To investigate the mechanism for this, we performed peritoneal lavage on three mice with BE-13luc progression following treatment with MGDL-28Z-expressing T cells. By flow cytometry, we found low or absent γδ TCR expression on BE-13luc cells in all three mice (Figure S10). Of note, some BE-13luc cells lacked γδ TCR expression before they were injected into mice, so selective growth of these γδ TCR-negative cells might have occurred in mice that received MGDL-28Z-expressing T cells. In vivo assessment of anti-γδ CAR T cells We assessed CAR T cell number and phenotype in the spleens of BE-13luc tumor-bearing mice 6 days after CAR T cell infusion (Figure S11). There was no statistically significant difference in the numbers of CD3+CAR+, CD3+CD4+CAR+, or CD3+CD8+CAR+ splenic T cells when mice receiving T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z were compared (Figures S11B–S11D). Effector memory T cells are negative for C-C chemokine receptor type-7 (CCR7) and CD45RA (CCR7−CD45RA−).40 For all CARs, a mean of greater than 80% of splenic CD4+CAR+ and CD8+CAR+ T cells were effector memory T cells (Figure S12). Lymphocyte activation gene 3 (LAG3) expression was higher on CD8+ T cells expressing MGDL-CD8BBZ compared with CD8+ T cells expressing MGDL-CD828Z. Programmed cell death protein 1 (PD1) expression was higher on CD4+ T cells expressing MGDL-28Z compared with CD4+ T cells expressing either MGDL-CD8BBZ or MGDL-CD828Z (Figure S13). We also assessed CAR T cell number and phenotype in the spleens of BE-13luc tumor-bearing mice 14 days after CAR T cell infusion (Figures S14 and S15). There was no statistically significant difference in the numbers of CD3+CAR+, CD3+CD4+CAR+, or CD3+CD8+CAR+ splenic T cells between mice receiving T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z (Figure S14). Most persisting CAR T cells were CD8+. There was no significant difference in CCR7 expression on splenic CAR+ T cells between mice receiving T cells expressing MGDL-CD8BBZ, MGDL-CD828Z, or MGDL-28Z (Figures S15A and S15B). PD1 expression was higher on CD8+ T cells expressing MGDL-CD8BBZ compared with CD8+ T cells expressing MGDL-28Z (Figure S15C). TCR γ constant domain versions recognized by MGDL-28Z and gdL-28Z The TCR γ constant domain has two main versions, TCR γ constant 1 and TCR γ constant 2, that are encoded by the TRGC1 and TRGC2 genes, respectively.4,5 In order to determine if the CARs we designed could recognize both versions of the TCR γ constant domain, we transduced CCRF-CEM cells with the TRDC gene along with either TRGC1 or TRGC2; the resulting cells were designated GC1DC-CEM and GC2DC-CEM, respectively (Figure S2). We used g9d2-CEM cells, which express a complete γδ TCR, as a positive control. There was not a statistically significant difference in IFNγ release when T cells expressing MGDL-28Z were co-cultured with either GC1DC-CEM or GC2DC-CEM (Figure S16). In addition, T cells expressing gdL-28Z released similar amounts of IFNγ when co-cultured with either GC1DC-CEM or GC2DC-CEM. When MGDL-28Z-expressing T cells were co-cultured with either g9d2-CEM or GC2DC-CEM, there was a statistically significantly lower level of IFNγ release with GC2DC-CEM co-culture. In contrast, there was not a statistically significant difference in IFNγ release after co-culture of MGDL-28Z-expressing T cells with either g9d2-CEM or GC1DC-CEM (Figure S16). Discussion Restricted expression of the γδ TCR makes it an attractive target for CAR T cell therapy. We designed a panel of CARs targeting the γδ TCR and selected the optimal CAR by assessing CAR expression and function. There is one prior short report of a CAR targeting the γδ TCR, which reported data on a CAR with a different antigen-recognition domain than the CARs we selected.41 We assessed antigen-recognition domains derived from 10 antibodies (Figure 1; Table S2).42,43,44,45 Two scFvs, MGDL and gdL, were selected for further assessment. scFv selection was initially based on the ability of CAR-expressing T cells to specifically recognize the γδ TCR. Specificity was indicated by high levels of IFNγ release when CAR T cells were co-cultured with target cells expressing the γδ TCR and low levels of IFNγ release when CAR T cells were co-cultured with negative-control target cells or cultured alone (Figures 1F and 1G). The choice of scFv linker can affect CAR T cell function.22,35 Compared with scFvs containing the 218 linker, the scFvs with (G4S)4 linkers, MGDL and gdL, were associated with greater antigen specificity as assessed by IFNγ release (Figures 1F and 1G). CARs with the MGDL scFv were consistently associated with a higher percentage of CAR+ T cells when compared with CARs incorporating the gdL scFv (Figures 1C, 2B, and 2C); therefore, CARs incorporating the MGDL scFv were selected for extensive in vitro and in vivo studies. The choice of HTM domain can have a significant effect on CAR T cell function.20,37 Among CARs with CD28 costimulatory domains, CARs containing a CD28 HTM domain exhibited higher levels of IFNγ and IL-2 release as well as better in vivo activity against BE-13luc cells compared with CARs containing a CD8α HTM domain (Figures 2D–2G, 3H, 4A–4F, and 7). Compared with T cells expressing CD28-containing CARs, we noted a greater degree of non-antigen-specific proliferation by MGDL-CD8BBZ-expressing T cells when CAR T cells were co-cultured with NGFR-CEM-luc-Thy1.1, which are γδ TCR-negative cells. Other reports have noted that T cells expressing CARs with a 4-1BB domain exhibited less antigen specificity in proliferation and cytokine release when compared with T cells expressing CARs with a CD28 domain.24,46 However, CARs with 4-1BB domains have not been associated with a higher incidence of clinical problems such as severe cytokine-release syndrome or other toxicities when compared with CD28-containing CARs.16,26 T cells expressing either MGDL-28Z or gdL-28Z recognized target cells transduced to express γδ TCR constant domains without any γδ variable domains (Figures 2E and S16; Table S3). In addition, MOLT-13 expresses Vγ3Vδ1,47 BE-13 expresses Vγ8Vδ1,47 and g9d2-CEM expresses Vγ9Vδ2. Therefore, our results show that the CARs that we have developed recognize γδ TCR constant domains either alone or in the presence of different variable domains (Figures 2D–2G and S16). There are two main versions of the TCR γ constant domain,4,5 so it was important to develop anti-γδ CARs that could recognize TCRs incorporating either version. When anti-γδ CAR T cells were co-cultured with target cells engineered to express only the constant domains of the γδ TCR, either version of the TCR γ constant domain could be recognized (Figure S16). The g9d2-CEM cell line expresses a TCR γ constant 1 domain, and the MOLT-13 cell line expresses a TCR γ constant 2 domain.48 Therefore, results of experiments using g9d2-CEM or MOLT-13 as target cells demonstrated that the CARs that we developed recognized cells expressing complete TCRs incorporating either TCR γ constant 1 or TCR γ constant 2 (Figures 2D–2G). T cells expressing MGDL-28Z exhibited greater in vivo anti-malignancy activity compared with T cells expressing the other MGDL-containing CARs (Figure 7). However, we did not observe substantial differences in CAR T cell numbers or in the expression of exhaustion and memory markers on CAR T cells from mice receiving T cells expressing the different MGDL CARs (Figures S11–S15). We suspect that the superior anti-malignancy efficacy of MGDL-28Z was caused by its greater functional activity demonstrated by higher levels of in vitro cytokine release and proliferation. Despite initial clearance, BE-13luc recurrence occurred in all mice treated with MGDL-28Z (Figure 7). To investigate this, we assessed BE-13luc γδ TCR expression following recurrence. We found greatly diminished expression of the γδ TCR on BE-13luc cells at the time of relapse compared with in vitro-cultured BE13luc cells (Figure S10). This suggests at least one cause of relapse was outgrowth of γδ TCR-negative BE-13luc cells. MOLT-13 and especially BE-13 expressed low levels of the γδ TCR (Figure S2), so outgrowth of γδ TCR-negative cells is not surprising after anti-γδ CAR T cell treatment. Antigen loss is a common problem with CAR T cell treatment of B cell lymphoma.49 Antigen loss is a potential clinical problem for anti-γδ CAR T cell therapies. Antigen loss might be mitigated by selection of patients with uniformly positive γδ TCR expression on malignant cells. We expect that non-malignant γδ T cells will be eliminated from patients receiving CAR T cells targeting the γδ TCR. Anti-γδ CAR T cells eliminated γδ T cells from in vitro cultures (Figure 3G). There is no known human disease characterized by a deficiency of only γδ T cells.3 Therefore, potential adverse events resulting from the elimination of normal γδ T cells are not fully predictable. γδ T cells are thought to play an important role in early control of a variety of infections including Listeria monocytogenes.6,7 Monitoring for infections in patients following anti-γδ CAR T cell treatment will be important. Because γδ T cells make up a small fraction of the total T cells,1,3 we do not anticipate that fratricide will be a problem during CAR T cell production cultures. In conclusion, we designed a panel of CARs targeting the constant domains of the γδ TCR. We selected an optimal CAR designated MGDL-28Z. Compared with T cells expressing the other CARs that we assessed, T cells expressing MGDL-28Z had greater in vitro function and superior in vivo anti-malignancy activity. These results provide evidence supporting clinical trials to test MGDL-28Z-expressing T cells as therapy for γδ TCR-expressing malignancies and possibly some autoimmune diseases involving γδ T cells. Materials and methods Human cells Peripheral blood mononuclear cells (PBMC) from patients enrolled in National Cancer Institute (NCI) clinical trials were used. Use of patient samples for research was approved by the NIH Institutional Review Board. Design and construction of plasmids encoding CARs All CAR plasmids include the MSGV1 γ-retroviral backbone.18 Sequences of the MGD-CD8BBZ and gd-CD8BBZ CARs followed this pattern from the N terminus to the C terminus: CD8α signal sequence, single-chain variable fragment (scFv), human CD8α HTM domains, human 4-1BB cytoplasmic region, and the human CD3ζ cytoplasmic portion. The scFvs in both MGD-CD8BBZ and gd-CD8BBZ had sequences from the N terminus to C terminus of light-chain variable domain, linker, heavy-chain variable domain. MGD variable domains were from the murine 11F2 antibody.45 The gd variable domains were from the murine clone B1 antibody (BD, Part number 555715). Both antibodies underwent protein sequencing to determine the amino acid sequences of the variable domains (Rapid Novor). The 218 linker (amino acid sequence, GSTSGSGKPGSGEGSTKG) was used in both the MGD and gd scFvs.33 To construct plasmids encoding the MGD-CD8BBZ and gd-CD8BBZ CARs, DNA fragments encoding the scFvs were ligated into plasmids containing the MSGV1 vector backbone, the CD8α HTM domains, the 4-1BB moiety, and the CD3ζ domain. We also constructed CARs called MSGV1-MGDL-CD8BBZ and MSGV1-gdL-CD8BBZ with a different scFv linker. In these CARs and all other CARs in this article with an “L” in the CAR name, the 218 linker was replaced with the (G4S)4 linker (amino acid sequence, GGGGSGGGGSGGGGSGGGGS).34 MSGV1-MGDL-CD828Z and MSGV1-gdL-CD828Z were identical to MSGV1-MGDL-CD8BBZ and MSGV1-gdL-CD8BBZ, respectively, except CD28 cytoplasmic domains replaced the 4-1BB domains. MSGV1-MGDL-28Z and MSGV1-gdL-28Z were identical to MSGV1-MGDL-CD828Z and MSGV1-gdL-CD828Z, respectively, except CD28 HTM domains replaced the CD8α HTM domains. The CD28 component of these CARs started with amino acids IEVMYPPPY of the hinge domain and continued to the C terminus as previously described.21 We constructed a series of CARs with the heavy-chain variable domain first and the light-chain variable domain second in the scFv amino acid sequence. This order was the reverse of the other CARs that we constructed. Aside from the order of the variable domains, the sequences of these CARs were identical to the sequences of the corresponding CARs with the light-chain variable domains first in the amino acid sequence. H1st (heavy-chain variable domain first) was added to the names of these CARs. For all CAR plasmids, DNA fragments were synthesized by GeneArt (Thermo). All DNA analysis was performed with SnapGene (Dotmatics). We used SnapGene to optimize for triplet codons commonly expressed in Homo sapiens for all DNA constructs. For all ligations, we ligated DNA fragments into restriction-enzyme-digested plasmids by using a New England Biolabs Quick Ligation Kit. The sequences for CD8α, CD28, 4-1BB, and CD3ζ were derived from the National Center for Biotechnology Information website. A previously described Hu19-CD828Z CAR was used as a negative control.20 A previously described FMC63-28Z CAR was also used as a negative control.18 Gamma-retroviral transductions and T cell culture To produce replication-incompetent gamma-retroviruses, packaging cells were transfected with a plasmid encoding a CAR along with a plasmid encoding the RD114 envelope protein as previously described.18 T cell transductions with γ-retroviruses encoding CARs were performed as previously described 2 days after initiation of T cell cultures.18 T cells were cultured as described previously.18 In brief, PBMCs were stimulated with 50 ng/mL anti-CD3 monoclonal antibody OKT3 (OKT3, Orthoclone) in AIM V complete medium plus 300 international units (IU)/mL of IL-2 (Teceleukin, Roche). AIM V complete medium consisted of AIM V medium (Thermo), 5% human AB serum (Valley Biomedical), 100 U/mL penicillin, and 100 μg/mL streptomycin. Non-tissue-culture-treated six-well plates were coated with RetroNectin (Takara). Two mL of gamma-retroviral supernatant was applied to each well of the RetroNectin-coated plates, and plates were incubated for 2 h at 37°C. Two mL of medium containing 2 × 106 cells was added directly to the viral supernatant. IL-2 was added at 300 IU/mL. Incubation was carried out at 37°C for 16–18 h. Next, cells were washed and suspended in fresh medium plus IL-2. Cultures were maintained by suspending the cells at 0.5 × 106 cells/mL in AIM V complete media with 300 IU/mL IL-2 every 2 days. Cell lines We transduced γδ TCR-negative CCRF-CEM (T cell leukemia, ATCC) cells with a complete Vγ9Vδ2 TCR to generate g9d2-CEM. We transduced CCRF-CEM cells with the γ constant 1 domain (TRGC1) and the δ constant domains (TRDC) to generate GC1DC-CEM. We transduced CCRF-CEM cells with the γ constant 2 domain (TRGC2) and δ constant domain to generate GC2DC-CEM. We also transduced CCRF-CEM cells to express the low-affinity nerve growth factor receptor (NGFR-CEM). Transduced CCRF-CEM lines were all sorted by flow cytometry to ≥95% positive for protein expression of the transduced gene. We transduced g9d2-CEM and NGFR-CEM with a γ-retrovirus encoding an enhanced firefly luciferase gene and Thy1.1 to create g9d2-CEM-luc-Thy1.1 and NGFR-CEM-luc-Thy1.1.50 Loucy is a γδ T cell leukemia line from ATCC. MOLT-13 and BE-13 are γδ T cell leukemia lines from DSMZ. MOLT-13 and BE-13 were transduced with a γ-retroviral vector encoding an enhanced firefly luciferase gene and Thy1.1.50 The transduced cells were called MOLT-13luc and BE-13luc. We flow-sorted MOLT-13luc and BE-13luc by gating on Thy1.1 expression. Ninety-five percent of the sorted MOLT-13luc cells expressed Thy1.1. Thy1.1 was expressed by 99.9% of sorted BE-13luc cells. These cell lines lacking γδ TCR expression were used in Figure S3: 293T (embryonic kidney), K562 (myeloid leukemia), A549 (lung carcinoma), TC71 (Ewing’s sarcoma, gift of M. Tsokos, NCI), COLO205 (colon carcinoma), U251 (glioblastoma), HEPG2 (hepatocellular carcinoma), and Panc10.05, (pancreatic carcinoma). These cell lines were from ATCC unless noted. Except for primary CD34+ hematopoietic cells that were from Stem Cell Technologies, primary human cells used in the experiment reported in Figure S3 were from Lonza. Cell lines are listed and briefly described in Table S3. γδ T cell culture To expand primary γδ T cells, PBMCs were stimulated with 5 μM zoledronic acid monohydrate (Sigma-Aldrich) in AIM V complete medium plus 1,000 IU/mL IL-2 (Teceleukin, Roche). Cultures were maintained by suspending the cells at 1 × 106 cells/mL in AIM V complete media with 1,000 IU/mL IL-2 every 2 days. CAR detection on T cells and flow cytometry Expression of CARs with the (G4S)4 linker was assessed with an antibody called anti-G4S linker (anti-G4SL), (clone E7O2V, Cell Signaling Technology). An antibody designated anti-218 linker (anti-218L) (clone E3U7Q, Cell Signaling Technology) was used to detect CARs with the 218 linker. To conduct flow cytometry staining, T cells (5 × 105) were suspended in 50 μL of staining buffer (0.4% BSA, 0.1% sodium azide in PBS), and antibodies were added. T cells were also stained with the following antibodies: CD3 allophycocyanin-cyanin-7 (APC-Cy7) clone UCHT1, BD Biosciences; CD4 fluorescein isothiocyanate ([FITC] or Brilliant Violet 510) clone RPA-T4, BD Biosciences; and CD8 R-phycoerythrin (PE) cyanin-7 (PE-Cy7/eFluor450) clone RPA-T8, BD Biosciences or Thermo Scientific. For memory and exhaustion marker staining, the following antibodies were used from BioLegend: CD45RA (FITC) clone HI100, CCR7 (PE-Dazzle 594) clone G043H7, Lag-3 (Alexa Fluor 647) clone 11C3C65, and PD-1 (Brilliant Violet 650) clone EH12.2H7. Expression of the γδ TCR was detected via flow cytometry by using anti-γδ TCR FITC, clone 11F2 (Miltenyi). Expression of the Vγ9 protein was detected via flow cytometry by using anti-Vγ9 TCR allophycocyanin (APC), clone B3 (BioLegend). Expression of the Vδ2 protein was detected via flow cytometry by using anti-Vδ2 TCR R- PE, clone 123R3 (Miltenyi). Flow cytometry was performed with LSRFortessa (BD) or FACSymphony A5 (BD) cytometers. Dead cells were excluded with 7-amino-actinomycin D (7-AAD, BD Biosciences). Flow cytometry analysis for all experiments was performed with FlowJo (Tree Star, Inc.). Immunohistochemistry A series of formalin-fixed paraffin-embedded (FFPE) normal tissue sections were stained with anti-γδ TCR antibody clone H-41 (Santa Cruz). Immunohistochemistry staining was conducted by using standard procedures, and results were assessed by a pathologist. CD107a assay For each culture tested, tubes containing each target cell type were prepared. All tubes also contained CAR-transduced T cells, 1 mL of AIM-V complete medium, PE-labeled anti-CD107a antibody (Clone H4A3, Thermo Scientific), and 1 μL of Golgi Stop (monensin, BD). Tubes were incubated at 37°C for 4 h and then stained for CD3, CD4, and CD8. We stained with antibodies against Thy1.1 to detect g9d2-CEM-luc-Thy1.1 and NGFR-CEM-luc-Thy1.1 target cells, and we excluded the target cells from flow cytometry analysis. Normalization was conducted by dividing the percentage of CD4+ or CD8+ T cells that were CD107a+ by the previously determined percentages of CD4+ or CD8+ T cells that were CAR+. Cytotoxicity assay Cytotoxicity assays were conducted as previously described.18 Cytotoxicity was measured by comparing survival of Loucy or MOLT-13 γδ TCR+ target cells relative to the survival of γδ TCR-negative CCRF-CEM target cells. Both target cell types were combined in the same tubes with CAR-transduced T cells. CCRF-CEM cells were labeled with 5-(and-6)-(((4-chloromethyl)benzoyl)amino) tetramethylrhodamine (CMTMR, Thermo Scientific), and Loucy or MOLT-13 cells were labeled with CFSE (Thermo Scientific). Cultures were set up in sterile 5-mL test tubes (BD) in duplicate at multiple T cell to target cell ratios. The target cells contained in the tubes were 50,000 CFSE-labeled Loucy or MOLT-13 cells and 50,000 CMTMR-labeled CCRF-CEM cells. Cultures were incubated for 4 h at 37°C. Immediately after the incubation, 7-AAD (BD Biosciences) was added, and flow cytometry acquisition was performed. For each T cell plus target-cell culture, the percent survival of Loucy or MOLT-13 cells was determined by dividing the percent live Loucy or MOLT-13 cells by the percent live CCRF-CEM negative-control cells. Corrected percent survivals were calculated by dividing the percent survival of Loucy or MOLT-13 cells in each T cell plus target cell culture by the ratio of the percent live Loucy or MOLT-13 cells to the percent live CCRF-CEM cells in tubes containing only Loucy or MOLT-13 cells and CCRF-CEM cells without effector T cells. This correction was necessary to account for spontaneous target cell death and variation in starting cell numbers. Cytotoxicity was calculated as follows: the percent cytotoxicity of Loucy or MOLT-13 cells = 100 − corrected percent survival of Loucy or MOLT-13 target cells. Primary γδ target cell elimination assay Untransduced T cells or T cells expressing MGDL-28Z, gdL-28Z, or Hu19-CD828Z were cultured with autologous γδ T cells at either a 1:1 or 4:1 effector to target (E:T) ratio. Cell numbers used were based on %CAR+ T cells and %γδ TCR+ target cells as determined by flow cytometry. For the 4:1 E:T ratio, 0.5 × 106 γδ TCR+ target cells and 2 × 106 CAR+ effector cells were combined in 3 mL AIM V complete media in a tissue-culture-treated 12-well plate. For the 1:1 E:T ratio, 2 × 106 γδ TCR+ target cells and 2 × 106 CAR+ effector cells were combined in 3 mL AIM V complete media in a tissue-culture-treated 12-well plate. Cultures were incubated for 24 h at 37°C. Twenty-four hours after the cultures were set up, the cells were washed and stained for CD3, CD4, CD8, and 7AAD; in addition, cells were stained with the 11F2 antibody to detect γδ TCR+ target cells. Enzyme-linked immunosorbent assay Target cells were combined with CAR-transduced T cells in duplicate wells of a 96-well round-bottom plate at a 1:1 E:T ratio. The cells were cultured in AIM-V medium + 5% AB human serum without IL-2 at 37°C for 18–20 h. Following the incubation, ELISAs for IFNγ were performed on the culture supernatant by using standard methods as previously described.23 IL-2 ELISAs (R&D Systems) were performed as recommended by the manufacturer. Cytokine levels were normalized for %CAR+ T cells by dividing cytokine levels by the %CAR+ T cells determined by flow cytometry. Proliferation assays Cultures were set up in 24-well plates. Target cells included in cultures were either g9d2-CEM-luc-Thy1.1 or NGFR-CEM-luc-Thy1.1. Target cells were all irradiated, and 0.5 × 106 target cells were used. The cultures included 0.75 × 106 CAR T cells. The T cells were labeled with CFSE as previously described.18 AIM V medium + 5% human serum without IL-2 was used. Four days after initiation, live cells in each culture were counted with trypan blue for dead cell exclusion. Flow cytometry was performed after staining with antibodies against the (G4S)4 linker (CAR), CD3, CD4, and CD8. Cells were also stained with Thy1.1 to exclude target cells from flow cytometry analyses. Antigen-specific proliferation was presented as CFSE MFI of T cells stimulated with g9d2-CEM-luc-Thy1.1 divided by CFSE MFI of T cells stimulated with NGFR-CEM-luc-Thy1.1. Murine experiments Seven-to-nine-week-old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were used in all experiments in a low-pathogen environment. The doses and timing of cell injections are described in figure legends. Mice were sacrificed in accordance with NCI Animal Care and Use Committee guidelines. Animal studies were approved by the NCI Animal Care and Use Committee. Mouse imaging To conduct imaging, mice were injected with 200 μL of luciferin solution (15 mg/mL in PBS, GoldBio) and anesthetized with 2.5% isoflurane. After 10 minutes, BLI was performed using an IVIS Lumina Series III Imaging System. Ventral images were taken by using a 1-min exposure on a 24-cm field of view with a binning factor of 4. Bioluminescence was quantified over the body of the mouse without the tail in units of radiance in photons/second/cm2/steradian (p/sec/cm2/sr) using Living Image software (Revvity). Images were scaled to 106–107 or 107–108 radiance units. Persons performing injections and persons performing imaging were blinded about the intervention received by each group of mice. Mice were sacrificed in accordance with the NCI Animal Care and Use Committee guidelines. In vivo T cell number and phenotype experiments NSG mice were injected with 4x106 BE-13 cells i.v. After 5 days, 8x106 CD3+CAR+ T cells/mouse were infused i.v. Mice were sacrificed 6 or 14 days after T cell infusion, and spleens were dissected from mice. Whole spleens were dissociated with a syringe plunger. Red blood cells were lysed with ACK (ammonium-chloride-potassium) lysis buffer (Quality Biological), and 1 × 106 splenocytes per mouse were used for flow cytometry staining. To calculate the number of CAR+ T cells per spleen, the total number of live splenocytes per mouse was multiplied by the fraction of CD3+ cells and the fraction of CD3+CAR+ cells. To determine the number of CD3+CD4+CAR+ splenocytes, the total number of live splenocytes per mouse was multiplied by the fraction of CD3+ cells, the fraction of CD4+ cells, and the fraction of CD4+CAR+ cells. The number of CD3+CD8+CAR+ splenocytes was calculated in an analogous manner as the CD3+CD4+CAR+ splenocytes. Peritoneal lavage Mice were sacrificed on day 57 after T cell infusion. The skin overlying the peritoneum was removed, and the peritoneal cavity was injected with 5 mL of PBS by using a 20G needle. The abdomen was massaged to dislodge any tumor cells into the PBS. The PBS was drawn back into the syringe and collected in a 5-mL conical tube on ice. Isolated cells were stained for Thy1.1 and γδ TCR, and flow cytometry analysis was performed. Statistics Statistics are described in figure legends and were performed with GraphPad Prism v.10.1.2. p < 0.05 was considered statistically significant. Correction for multiple comparisons was only conducted when noted. Data and code availability Pertinent data are included in the article or the supplemental information. Any data can be provided upon reasonable request. GenBank accession codes of novel CARs reported in this manuscript are as follows: BankIt2979808: MGD-CD8BBZ PV931582 BankIt2979808: gd-CD8BBZ PV931583 BankIt2979808: MGDL-CD8BBZ PV931584 BankIt2979808: gdL-CD8BBZ PV931585 BankIt2979808: MGDL-CD828Z PV931586 BankIt2979808: gdL-CD828Z PV931587 BankIt2979808: MGDL-28Z PV931588 BankIt2979808: gdL-28Z PV931589 Acknowledgments We thank Shawn Farid for flow cytometry support. We thank the staff of the NIH Clinical Research Center mouse facility for mouse care. This research was supported by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH author(s) are considered Works of the United States Government. The findings and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services. All funding for this work was from NIH intramural funds project ZIABC012157-02. The funder was not involved in the design of experiments, interpretation of data, or the decision to submit the manuscript. Author contributions L.C.C. and J.N.K. conceived the study and wrote the manuscript. L.C.C., J.N.K., and S.M.H. analyzed the data. L.C.C., C.A., N.L., D.A.N., S.M.H., and K.A.W. conducted the experiments. All authors reviewed and edited the manuscript. Declaration of interests J.N.K. has received research funding from Kite, a Gilead Company, and Bristol Myers Squibb. These funders did not support the current work. J.N.K. has received royalties for CAR technology from Kite, a Gilead Company; Bristol Myers Squibb; and Kyverna. J.N.K. and L.C.C. are inventors on a patent application for the CARs reported in this manuscript. Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.omton.2026.201260. Supplemental information Document S1. Figures S1–S16 and Tables S1–S3 Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Figures S1–S16 and Tables S1–S3 Document S2. Article plus supplemental information Data Availability Statement Pertinent data are included in the article or the supplemental information. Any data can be provided upon reasonable request. GenBank accession codes of novel CARs reported in this manuscript are as follows: BankIt2979808: MGD-CD8BBZ PV931582 BankIt2979808: gd-CD8BBZ PV931583 BankIt2979808: MGDL-CD8BBZ PV931584 BankIt2979808: gdL-CD8BBZ PV931585 BankIt2979808: MGDL-CD828Z PV931586 BankIt2979808: gdL-CD828Z PV931587 BankIt2979808: MGDL-28Z PV931588 BankIt2979808: gdL-28Z PV931589