CAR Signaling Informs Mechanisms to EnhanceMetabolism and Function in γδ T Cells
Xiomar Bustos Perez, Leticia Tordesillas, Elena Martinez-Planes, Miguel G Fontela, Renata Marques Rossetti, Victoria Izumi, Bin Fang, John Koomen, Eric Welsh, Patrick Hwu, Daniel Abate-Daga
TL;DR
This study explores how γδ T cells differ from αβ T cells when engineered with CARs, identifying ways to improve their function and metabolism for better cancer therapies.
Contribution
The study reveals cell-type-specific CAR design requirements and identifies a new synthetic receptor to enhance γδ CAR-T cell performance.
Findings
γδ and αβ CAR-T cells show similar cytotoxicity but distinct metabolic and signaling profiles.
γδ CAR-T cells have reduced glycolytic and oxidative phosphorylation capacity compared to αβ CAR-T cells.
A synthetic co-stimulatory receptor improves AP-1 activation and in-vivo persistence of γδ CAR-T cells.
Abstract
γδ T cell-based immunotherapies have gained relevance as an alternative to the conventional αβ T cell products with pre-clinical data demonstrating tumor burden reduction and mitigation of tumor-induced damage. Given that most CAR constructs were optimized for αβ T cells, we hypothesized that distinct T cell types may require tailored CAR architectures to achieve optimal function. To test this hypothesis, we conducted a systematic comparative analysis between γδ and αβ T cells transduced with a second-generation PSCA-targeting CAR (PSCA-8t28z). We found that although γδ and αβ CAR-T cells exhibit comparable cytotoxicity, they differ phenotypically. Through a system level phosphoproteomic analysis, we identified 307 phospho-sites with differential abundance between γδ and αβ CAR-T cells. Pathway enrichment analysis placed glycolysis/gluconeogenesis and TCR signaling within the top…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCAR-T cell therapy research · T-cell and B-cell Immunology · Immune Cell Function and Interaction
INTRODUCTION
Gamma delta (γδ) T cell-based immunotherapies represent an emerging class of adoptive cell therapies (ACTs) with the potential to overcome several limitations associated with traditional alpha beta (αβ) T cell products(1–4). Unlike αβ T cells, γδ T cells exhibit innate-like cytotoxicity and recognize stress-induced or non-peptide antigens in a manner that is independent of classical MHC restriction(5, 6). In addition to their direct tumor-killing capabilities, γδ T cells have demonstrated the capacity for antigen cross-presentation(7, 8), further amplifying immune responses. These characteristics render them especially attractive for use in both autologous and allogeneic contexts(5, 9–11). Among γδ T cell subsets, Vγ9Vδ2 and Vδ1 T cells have shown preferential homing to inflamed or malignant tissues, as well as to select non-malignant niches, making them particularly relevant for targeting solid tumors(12, 13).
Clinical trials of unmodified γδ T cells have consistently demonstrated a favorable safety profile(14–17). However, clinical responses have been modest, underscoring the need for strategies that enhance their effector function and persistence. Genetic modification, most notably through the expression of chimeric-antigen-receptors (CARs), has emerged as a promising avenue to unlock the full therapeutic potential of γδ T cells. Indeed, γδ CAR-T cells have shown encouraging preclinical activity against a range of solid tumors, including hepatocellular carcinoma (18), ovarian (19), gastric (20), and metastatic prostate cancer (21). Beyond CAR expression, ongoing efforts aim to optimize the intrinsic features of γδ T cells, including their cytotoxicity, cytokine profile, and metabolic resilience (22).
To date, most CAR design efforts have been guided by studies in αβ T cells, which are better characterized and more readily obtained from peripheral blood(23). CAR architecture including costimulatory domains, hinge and transmembrane sequences has a profound impact on T cell phenotype and function, influencing cytotoxicity, persistence, memory differentiation, cytokine release, and cellular metabolism(24–27). CAR-T cells are meant to perform in nutrient-deprived, hypoxic tumor microenvironments. Because of that limiting environment, improving their metabolic performance has become a key objective(28–30). Approaches to enhance glycolytic capacity and oxidative phosphorylation have yielded promising results in αβ CAR-T cells(31–33), yet many of these metabolic strategies remain untested in γδ T cells.
Despite a growing interest in γδ T cell-based therapies, fundamental differences in their biology remain poorly understood(34, 35). Consequently, γδ CAR-T cell engineering has relied heavily on αβ-derived design principles, under the assumption of functional homology between these lineages. However, whether these αβ-based CAR constructs are optimal for γδ T cells is unclear. In this context, we sought to dissect the signaling and functional consequences of CAR engagement in γδ versus αβ T cells, to inform the rational design of gene modifications tailored to γδ T cells.
In this study, we provide a comparative analysis of CAR-triggered signaling in αβ and γδ T cells, particularly Vγ9Vδ2, with a focus on effector function and metabolic programming. Our results demonstrate that CAR signaling in γδ T cells lead to distinct pathways and divergent activation profiles. Notably, we identified significant differences in metabolic phenotype of Vγ9Vδ2 γδ T cells exhibiting lower glycolytic and mitochondrial activity compared to αβ T cells. We observed an improvement in mitochondrial respiration activity of expanded Vγ9Vδ2 γδ T cells following inhibition of thioredoxin-interacting protein (TXNIP), which was highly expressed in Vγ9Vδ2 γδ T cells. TXNIP inhibition during T cell expansion resulted in improved spared respiratory capacity, mitochondrial mass and CAR expression, suggesting a feasible strategy to augment γδ T cell metabolic fitness.
In parallel, we identified differential engagement of canonical TCR signaling components downstream of CAR activation, including diminished activation of key transcription factors i.e., Activator Protein 1 (AP-1) and Nuclear Factor of Activated T cells (NFAT) in γδ CAR-T cells. To remediate the lower AP-1 activation, we designed a novel chimeric switch receptor (CSR) that leverages the natural abundance of RANKL in the tumor/bone microenvironment. The CSR consists of a RANK (TNFRSF11A) ectodomain fused to a CD27 endodomain. Co-expression of this CSR in γδ CAR-T cells resulted in enhanced AP-1 transcriptional activation and improved γδ T cell survival in vivo.
Together, these findings highlight fundamental biological differences between γδ and αβ T cells and underscore the necessity for subset-specific engineering strategies. Tailoring chimeric receptor constructs to the unique signaling and metabolic profiles of γδ T cells will be essential to fully harness their therapeutic potential, especially in the context of solid tumor immunotherapy.
RESULTS
γδ CAR-T cells produce lower cytokine levels compared to αβ CAR-T cells
To define the functional properties of γδ CAR-T cells, we started by comparing their effector cytokine secretion with that of αβ CAR-T cells derived from the same donor. We used a prostate stem cell antigen (PSCA)-targeted second-generation CAR design (21) as our main experimental model. This CAR contains a CD8α-derived hinge/transmembrane domain fused to a CD28 co-stimulatory moiety and a CD3zeta activation domain (herein named 8t28z). In our previous work, we had shown that γδ and αβ CAR-T cells display comparable cytolytic strength against prostate cancer cell lines (21). Expanding on those results, we found that upon overnight coculture with PSCA-expressing C4–2B cells, CAR-induced secretion of granzyme B (GZMB), interferon-γ (IFNγ), tumor necrosis factor-α (TNFα), and interleukin-2 (IL-2) was observed for both T cell subsets. However, cytokine levels were consistently and statistically significantly lower in γδ CAR-T cells (Fig. 1A), even if their potential for polyfunctionality was comparable (Fig. S1A). Further phenotypical analysis revealed that γδ CAR-T cells express significantly higher levels of programmed cell death protein 1 (PD-1; p = 0.0477) and CD69 (p = 0.0147) than αβ CAR-T cells. In contrast, the activation marker CD25 was expressed at similar levels between subsets but appeared to be greater in CAR-expressing cells of either kind (Fig. 1B). Importantly, γδ CAR-T cells exhibited lower expression of co-stimulatory molecules CD27 (p = 0.0143) and CD28 (p = 0.0039) than αβ CAR-T cells (Fig. 1C), and these markers were not affected by CAR expression, suggesting that these differences are intrinsic to the T cell subset. Differences between γδ and αβ CAR-T cells were not explained by differences in CAR expression, which varied among donors (range: 20–72%), but was comparable between subsets. (Fig. S1B). These results indicate that while comparable in cytolytic potential, γδ and αβ CAR-T cells differ in other important functional traits, warranting a deeper mechanistic analysis of their immunobiology.
CAR activation induces different signal transduction in γδ and αβ CAR-T cells
Prompted by the phenotypical differences observed between γδ and αβ CAR-T cells, we designed an approach to capture a snapshot of the signaling events induced by CAR activation in either cell subset. C4–2B PSCA^+^ cells were cultured in SILAC (Stable Isotope Labeling by Amino Acids in Cell Culture) media until incorporation of labeled amino acids reached over 95% of the total proteome. These labeled tumor cells were cocultured with unlabeled CAR-T cells (or UT cells, as background controls), under three different conditions: 1) CAR-T cells incubated with tumor cells for 1 hour (antigen-specific CAR-driven signaling); 2) CAR-T cells mixed with tumor cells without incubation (antigen-independent CAR-driven signaling, or tonic signaling); 3) UT T cells incubated with tumor cells for 1 hour (basal phosphorylation levels; Fig. 2A, Fig. S2).
Phosphopeptide abundance was quantified via tandem mass spectrometry, and statistical comparisons among groups were calculated by considering a fold-change > 1.5 and p-value < 0.05 (Welch’s t-test). Comparing activated CAR-T cells (condition 1), we found 307 phospho-sites that had significantly different abundances; 224 of them were hypo-phosphorylated in γδ CAR-T cells (Fig. 2B). Both CD8a and CD4 were among the hypophosphorylated events in γδ CAR-T cells, consistent with lower (CD8a) or null (CD4) co-receptor expression in γδ T cells(36, 37) (Fig. 2B). Hypophosphorylated proteins included also PLCγ (PLCG1), c-Jun (JUN), GLUT1 (SLC2A1), among others. Hyperphosphorylated proteins included NFAT1 (NFATC2), p38 (MAPK14), PHAG1 (PAG1), LFA-1 (ITGAL), and (EOMES), among others. A full list of differentially phosphorylated proteins can be found in Table S1.
Pathway enrichment analysis identified 55 potential canonical pathways that were significantly overrepresented among the differentially phosphorylated events. The 30 most significant of these canonical pathways are represented in Fig. 2C. Remarkably, pathways related to glycolysis/gluconeogenesis and T cell receptor (TCR) signaling were among the most enriched, with a negative Z score, suggesting they are more activated in αβ compared to γδ CAR-T cells. Since CAR-T cell activation in this assay was driven solely through the chimeric-antigen-receptor and not the endogenous TCR, these results suggest that the same CAR construct can drive different activation signals depending on the T cell type in which it is expressed.
γδ T cells display lower metabolic activity compared with αβ T cells
Glycolysis and gluconeogenesis signaling networks ranked as the first and fourth most overrepresented pathways within differentially phosphorylated proteins (Fig. 2C). Five of the enzymes involved in these pathways were less phosphorylated in γδ CAR-T cells (Fig. 3A), along with the main glucose transporter GLUT-1 (SLC2A1, S473 p = 0.00613). Phospho-sites in aldolase (ALDOA, S39 p = 0.00062; Y204 p = 0.00929), triosephosphate isomerase (TPI1, S21 p = 0.00025), phosphoglycerate kinase (PGK1, S203 p = 0.00002), pyruvate kinase (PKM, S37 p = 0.00281), and enolase (ENO1, Y44 p = 0.03398; S419 p = 0.01140; ENO2, T265 p = 0.00094), differed significantly between γδ and αβ CAR-T cells. These differences appear to be intrinsic to the cell subset, because similar trends are observed when comparing CAR negative γδ vs. αβ T cells, or CAR-expressing cells in absence of tumor cells (Fig. 3B). We next quantified the mRNA abundance of hexokinase, phosphofructokinase, and pyruvate kinase in γδ and αβ T cells, by RT-PCR, to test if the differences in phosphorylation were due to differences in gene expression. As shown in Fig. S3A, gene expression was comparable among cell types, suggesting that the observed phosphorylation differences were due to differences in the activation status of the respective pathways. In contrast, GLUT-1 showed lower phosphorylation levels in S473, and lower protein expression (Fig. 3C). We analyzed GLUT-1 surface expression in UT and CAR-T cells, either in resting state or following activation (anti-CD3 clone: OKT3, 24h, Fig. S3B). GLUT-1 was less frequently expressed in γδ than in αβ UT T cells (p = 0.0001). Moreover, CAR expression was associated with greater expression of GLUT-1 in both subsets (UT vs CAR in γδ p = 0.0015; in αβ p = 0.0174). Despite the CAR-associated increase in GLUT-1 expression, a significant difference in expression between γδ and αβ CAR-T cells remained (p = 0.0005, Fig. 3C). After OKT3 activation, GLUT-1 expression increased slightly in every condition tested; however, differences between γδ and αβ CAR-T cells were still significant (p = 0.0107, Fig. S3A). To test whether the lower expression of GLUT-1 observed in γδ T cells was a result of ex-vivo culture, we analyzed its expression in uncultured PBMC. On average, 56.27% of CD8 T cells expressed GLUT-1, which was significantly higher than the average expression in Vδ2 T cells (1.15% p = 0.0208), Vδ1 T cells (18.87% p = 0.0250), and even CD4 T cells (0.03% p = 0.0209).
We also observed a trend (p = 0.0681) of greater expression of GLUT-1 in Vδ1 T cells compared to virtually no expression of the transporter in circulating Vδ2 T cells (Fig. 3D). To test whether the observed lower expression of GLUT-1 was compensated by expression of other glucose transporters previously reported in peripheral blood lymphocytes, we analyzed by flow cytometry the expression of GLUT-3, GLUT-4, GLUT-6, and GLUT-8 (38–40) (Fig. S3C). No statistically significant difference in the expression of either transporter among T cell subsets was observed, suggesting that γδ T cells may be less capable of uptaking glucose that αβ counterparts.
Next, we sought to delineate the functional implications of the observed differences in phosphorylation of glycolytic enzymes and GLUT-1 expression. We first measured the glycolytic rate of the different T cell subsets, using the SeaHorse XF Glycolytic Rate Assay Kit (Fig. 3E). In agreement with previous reports, we observed great variability in glycolytic rate between donors(41). Despite said variability, we found that while resting γδ and αβ UT T cells have comparable basal glycolytic rates, γδ CAR-T cells displayed significantly lower basal glycolytic rate than αβ CAR-T cells (p = 0.002 Fig. 3F). Compensatory glycolysis, considered as the maximal glycolytic capacity that the cells can achieve in extreme conditions, resulted significantly lower in both γδ UT (p = 0.0359) and CAR (p = 0.0037) T cells compared to their corresponding αβ counterparts (Fig. 3F). When T cells were stimulated with OKT3, both basal and compensatory glycolytic rates are greater in αβ T cells (Fig. S3D). Consistently, results from glucose uptake experiments showed that γδ T cells uptake less glucose than αβ T cells independently of CAR expression (γδ vs αβ UT p = 0.0071; γδ vs αβ CAR p = 0.0049, Fig. 3G).
In addition to producing energy for homeostatic processes, glycolysis is particularly relevant for the acute effector functions of T lymphocytes(42, 43). In contrast, oxidative phosphorylation is more dominant in resting lymphocytes, prominently in naive and central memory cells(44, 45). To test whether the lower glycolytic rate observed in γδ T cells was compensated by greater oxidative capacity, we quantified oxygen consumption using the Seahorse XF T Cell Metabolic Profiling Kit (Fig. 3H). We found that basal respiration, measured by oxygen consumption rate (OCR) before the electron chain is altered, was significantly lower in unstimulated γδ CAR-T cells compared to αβ CAR-T cells (p = 0.0326, Fig. 3I). This pattern was also observed for maximal respiration (p = 0.0081, Fig. 3I) and for spare respiratory capacity (SRC) (p = 0.0075, Fig. 3I). Following activation, the differences in basal respiration and SRC between T cell subsets were not statistically significant (Fig. S3E). However, both CAR (p = 0.0399) and UT (p = 0.0497) γδ T cells exhibited lower maximal oxygen consumption than αβ counterparts (Fig. S3E).
We next analyzed the mitochondrial mass using the Mito Tracker Deep Red (MTDR) dye. In line with previous findings, γδ T cells contained less mitochondrial mass compared with αβ T cells (γδ vs αβ UT p = 0.0060; γδ vs αβ CAR p = 0.0046, Fig. 3J), and neither CAR expression nor OKT3 activation (Fig. S3F) affected the mitochondrial mass content of either cell subset. Moreover, uncultured Vδ2 T cells from peripheral blood mononuclear cells (PBMCs) showed lower mitochondrial mass than CD4 (p = 0.0151), CD8 (p = 0.0389), and Vδ1 (p = 0.001) T cells (Fig. 3K), indicating that these properties are cell intrinsic and not a product of ex vivo manipulation. Finally, γδ CAR-T cells displayed lower membrane potential than αβ CAR-T cells as evaluated by TMRE staining (p = 0.0214, Fig. 3L). Altogether, these results indicate that the two major mechanisms of ATP production (namely, glycolysis and oxidative phosphorylation) are less active in γδ than in αβ CAR-T cells. These differences were potentially due to lower mitochondrial mass.
TXNIP inhibition improves γδ T cell metabolism and function
Among the proteins showing differential phosphorylation between γδ and αβ CAR-T cells (Fig. 2B), Thioredoxin-interacting protein (TXNIP) emerged as a potential actionable target to enhance T cell metabolism. TXNIP is a multifunctional protein involved in cellular oxidative stress response and glucose metabolism(46). Notably, all identified TXNIP phospho-sites were less abundant in γδ CAR-T cells (S346 p = 0.01522, S361 p = 0.03584, T348 p = 0.00557, T349 p = 0.01817, Fig. 4A). Particularly, the phosphorylation of T349 triggers the proteasomal degradation of TXNIP (47). Thus, its lower phosphorylation in γδ T cells suggested that these may accumulate TXNIP due to reduced degradation. To validate this prediction, we analyzed TXNIP expression in cultured T cells and found that γδ T cells express ~ 1.74 (± 0.22) times more TXNIP compared to αβ T cells (p = 0.0068, Fig. 4B). We reasoned that TXNIP accumulation may inhibit glycolysis and lead to mitochondrial damage, which could be mitigated through pharmacological inhibition of TXNIP (48, 49). To test this concept, we designed a new γδ T cell expansion protocol that incorporated SRI-37330, a TXNIP inhibitor(50), during ex vivo culture (Fig. 4C). After five days of γδ T cell expansion, we analyzed the expression of TXNIP mRNA by RT-PCR, observing a reduction in TXNIP transcript in the cells treated with SRI-37330 (Fig. 4D). TXNIP inhibition was associated with increased mitochondrial mass (p = 0.0243, Fig. 4E), a significant increase in spare respiratory capacity (p = 0.0172, Fig. 4G), and enhanced basal glycolytic rates (Fig. 4H). To test the impact of TXNIP on CAR-T function, we generated γδ CAR-T cells in presence of the TXNIP inhibitor during manufacture. The resulting product presented a greater percentage of CAR-expressing cells compared to the control cells expanded without the inhibitor (p = 0.0085, Fig. 4I). We then evaluated activation and effector function of these CAR-T cells by ELLA following coculture with PSCA-expressing C4–2B cells. γδ CAR-T cells expanded with SRI-37330 released significantly more TNFα, GZMB, and IL-2 (Fig. 4J), indicating an improvement in effector function associated with TXNIP inhibition. Collectively, these results suggest that TXNIP plays an important role in modulating metabolism and effector function in γδ CAR-T cells, and may represent an actionable target to enhance cellular immunotherapies.
CAR activation triggers different TCR signaling events in γδ versus αβ CAR-T cells
In addition to glycolysis and gluconeogenesis, TCR signaling was among the top ten most overrepresented canonical pathways identified in our phosphoproteomic experiment (Fig. 2C). In the context of our experiment, we attribute the activation of this pathway to the signal provided by the CAR molecules (designed to mimic TCR engagement). Upon stimulation of CAR-T cells with live tumor cells for an hour, early signal kinases such as ITK, LCK, ZAP70 presented comparable phosphorylation levels across both cell types (Fig. 5A), suggesting that proximal activation thresholds were achieved in both cell subsets. However, activation differences became evident downstream, showing significantly higher phosphorylation of transcription factor NFAT (NFATC2, S53 p = 0.01981, S73 p = 0.0319) and significantly lower phosphorylation of AP1 (JUN, S58 p = 0.00062) in γδ CAR-T cells (Fig. 5A). PAK2, a kinase involved in cell adhesion, and associated with IL-2 production(51), was hypo-phosporylated at activation site S141 (p = 0.01581) in γδ CAR-T cells (Fig. 5B), suggesting diminished PAK2 activity in this subset. In contrast, lower phosphorylation of PLCγ1 (PLCG1) at its inhibitory site S1248 (52) (p = 0.00082), suggested increased enzymatic activity in γδ CAR-T cells (Fig. 5B). PLCγ1 is a critical effector in DAG and IP_3_-mediated calcium signaling and its activity can modulate activation of transcription factor NFAT. MK14 (MAPK14) was differentially phosphorylated at its activation site Y182 (p = 0.01257, Fig. 5B), which can partially account for the hyper-phosphorylated state of NFAT found in γδ CAR-T cells (Fig. 5A). (53, 54).
The low detection of CD4 (p = 0.01389) and CD8α (CD8A, p = 0.00004) phospho-sites in γδ CAR-T cells is compatible with the lower or null expression of co-receptors in this cell subset. Additional phosphorylation differences, i.e., PAK2 S64 (p = 0.00797), JIP3 S585 (MAPK8IP3, p = 0.02531), JUN S58, PHAG1 S50 (PAG1, p = 0.00933), LAT S40/43 (p = 0.02706), NFAT1 S53/S73, GSK3β S219 (NFATC2, p = 0.01551), and SHP-1 S582 (PTPN6, p = 0.03806), were also noted (Fig. 5B), yet the functional implications of these sites remain unclear due to limited annotation in the literature.
Notably, phosphorylation at CD3ζ Y83 (CD247) was significantly elevated in CAR-T cells relative to their UT counterparts (γδ CAR vs UT p = 0.01229, and αβ CAR vs UT p = 0.00539). These results indicate that the CAR activation domain is phosphorylated following stimulation. However, we cannot rule out that this activation is due to tonic signaling, because such elevation was not significant between stimulated and unstimulated CAR-T cells (Fig. 5C). A similar pattern was observed for phosphorylation of CD3ζ Y72 (Fig. 5C). In contrast, the CD3ζ Y111 residue was significantly more phosphorylated in stimulated than in non-stimulated CAR-T cells, but only in the αβ subset (Fig. 5C). Interestingly, these phospho-sites showed comparable levels between subsets when comparing activated CAR-T cells at 60 minutes after antigen engagement (Fig. 5C).
To gain a deeper understanding of the kinetics of CAR activation, we expanded the phosphoproteomics results by conducting a targeted time-course analysis of CD3ζ phosphorylation between 1 and 60 minutes post-stimulation. We examined phosphorylation at specific tyrosine residues of CD3ζ ITAM via WB across three independent donors. Consistent with prior observations, ITAM sites phosphorylation were comparable between γδ and αβ CAR-T cells at the one-hour-mark (Fig. 5D). However, differential phosphorylation was found in Y83 (p = 0.0046 at 30 min) and Y142 (p = 0.0197 at 0 min, p = 0.03 at 10 min and p = 0.0119 at 30 min), where γδ CAR-T cells exhibited attenuated phosphorylation relative to αβ counterpart overtime (Fig. 5D). Altogether, these findings suggest that while early activation thresholds are met across both CAR-T cell types, γδ T cells appear to integrate antigen signals with different kinetics or magnitude, resulting in differential downstream signaling in pathways of proven functional relevance.
CD27 costimulation delivered through a chimeric-switch-receptor improves AP-1 activation and T cell survival
Based on the observation of reduced JUN phosphorylation in γδ CAR-T cells (Fig. 5A), we postulated that supplementation of JUN signaling may harness their functionality and/or persistence. Unlike similar approaches involving JUN overexpression (55, 56), we sought to increase JUN signaling through the delivery of complementary co-stimulatory signals. More specifically, we proposed that CD27- or 41BB-triggered signaling cascades would converge on the activation of AP-1 (JUN/FOS) and Nuclear Factor Kappa-B (NFĸB, Fig. 6A and 6B). CD27 is of particular interest, because we observed a significant expression deficit in γδ compared to αβ CAR-T cells (Fig. 1C). To achieve CSR activation within the tumor, we leveraged the overexpression of Receptor Activator of NFĸB ligand (RANKL, TNFSF11) in the bone/tumor microenvironment (57). We fused CD27 or 41BB signaling domains with the ectodomain of RANKL’s receptor, RANK (TNFRSF11A) with the goal of converting a pro-osteolytic input signal (RANKL) into a co-stimulatory output signal. We cloned each CSR in a retroviral vector, together with a truncated version of CD34 (tCD34) to be used as a marker of transduction. Through co-transduction of either CSR vector together with our CAR vector, we aimed to generate the corresponding γδ SWITCH-CAR-T cells (Fig. 6C and 6D). Co-transduction of the PSCA-8t28z CAR with either CSR vector yielded ~ 50% double-positive cells on average (n = 9 independent donors, Fig. 6E). We next tested the ability of SWITCH-CAR-T cells to bind RANKL fused to a human antibody Fc region (hu-RANKL-Fc). Although both CSRs show some level of binding to RANKL, the RANKCD27 construct displayed significantly greater percentage of positive T cells (p = 0.0051, Fig. 6F). Interestingly, we found an increased frequency of CD27^+^/CD45RA^−^ cells in the SWITCH-CAR group expressing the RANKCD27 (p = 0.057, Fig. 6G), but not the RANK-4–1BB construct. This central memory phenotype has been linked with better expansion of γδ T cells from cancer patients (14, 58, 59). Based on these results, we chose RANKCD27 CSR for further analysis.
We next tested whether CSR engagement resulted in activation of the AP-1 signaling. To that end, we used the TransAM^®^ AP-1 ELISA kit for the quantification of the relative abundance of nuclear c-Fos, JUN-D, and phosphorylated c-JUN in activated T cells. γδ SWITCH27-CAR-T cells had significantly greater abundance of all three AP-1 components compared to UT T cells (phos c-JUN p = 0.0153, c-Fos p = 0.0492, JUN-D p = 0.0201), whereas γδ CAR-T cells showed increased c-Fos only (p = 0.0047, Fig. 6H). These results confirm that the CD27 costimulatory signaling provided by the new CSR boosts JUN activity in the presence of RANKL.
Finally, we analyzed the in-vivo performance of SWITCH27-CAR-T cells using a murine model of bone-metastatic prostate cancer. Mice bearing intratibial xenografts of C4–2B cells were treated with systemic infusions of γδ CAR or γδ SWITCH27-CAR-T cells. Mice receiving untransduced (UT) T cells or PBS (Untreated) were used as negative controls (Fig. 6I). SWITCH27-CAR-T cells controlled tumor burden comparably to γδ CAR-T cells, demonstrating that a secondary receptor does not hinder CAR activity. Notably, endpoint analysis revealed a significantly greater abundance of γδ T cells in both spleen (p = 0.05) and bone marrow (p < 0.0001) on the SWITCH27-CAR-treated relative to UT-treated and CARtreated groups (Fig. 6J), suggesting enhanced in-vivo persistence mediated by CD27 costimulation. Collectively, these results demonstrate that integration of a costimulatory signal through a separate CSR can improve γδ CAR-T cell persistence without impairing antitumor function.
DISCUSSION
γδ T cells offer an attractive platform for adoptive cell therapies against solid tumors(1, 17, 60). However, the translational potential of γδ T cells remains undermined by a knowledge gap in their unique immunobiology, including their signal transduction needs (1). In this study, we conducted a systematic comparative analysis of CAR-triggered signaling pathways in γδ versus αβ T cells and unveiled functional and mechanistic differences that can inform the design of next-generation γδ CAR-T cell therapies.
Among the most notorious differences across subsets, we found that γδ CAR-T cells produce significantly lower levels of pro-inflammatory cytokines such as IFNγ, granzyme B, TNFα, and IL-2, compared to αβ counterparts despite showing comparable cytolytic potency (21). This profile suggests that γδ CAR-T cells could mediate effective tumor control with a lower risk of cytokine-associated toxicities. Activation markers such as PD-1 and CD69 are expressed at significantly higher levels in cultured γδ than in αβ CAR-T cells. This finding aligns with prior studies reporting that elevated PD-1 expression in γδ T cells correlates with diminished cytokine release, particularly IFN-γ, following interaction with PD-L1-expressing targets (61–63). Importantly, this functional attenuation does not compromise cytotoxicity against Zol-pretreated tumor cells, regardless of PD-L1 expression levels (61). In our model, UT T cells did not express PD-1, consistent with the well-documented transient nature of PD-1 expression, which typically peaks early after activation and declines over time (27). CD69 expression dynamics observed in our study also mirrored previously reported patterns (63), with Zol and IL-2 treatment inducing CD69 expression in ~ 80% of γδ T cells at day 2, declining to ~ 60% by day 7. In contrast, only ~ 20% of αβ T cells cultured from the same donor expressed CD69 (63). Beyond acute activation, CD69 expression may have implications for tissue residency. Expression of co-stimulatory receptors also differed across subsets, with lower CD28 in γδ T cells, as previously documented (35, 64, 65). Of note, we found a significantly lower expression CD27, a co-stimulatory receptor known to play a key role in the survival and polarization of Vγ9Vδ2 T cells(66). This CD27 deficiency may represent an opportunity for intervention aimed at enhancing CAR-T cell survival, as discussed below.
From the comparative signalosome analysis, we identified several differentially regulated pathways. We focused on two of them for further analysis: 1) Glycolysis and 2) TCR signaling. Functional metabolic profiling validated the phosphoproteomic results, confirming the prediction of a lower glycolytic rate in γδ compared to αβ T cells. In concordance with that observation, we found that γδ T cells expressed lower levels of GLUT1 and lower glucose uptate. These differences were not compensated by increased oxidative phosphorylation. In fact, γδ T cells showed significantly lower oxygen consumption and lower mitochondrial mass than αβ T cells, which we interpret as a sign of lower energy demands. Lower energy requirement has been described as an adaptation mechanism used by tumor cells, in which they ‘resign’ functions that have high energy cost, such as protein secretion(67). This lower energy production may be associated with the lower secretion of cytokines observed in γδ T cells, representing an intrinsic trait of this subset. But Vγ9Vδ2 T cells still rely on glycolysis to perform their effector functions, as highlighted by recent studies(41), therefore we sought to design strategies to boost Vγ9Vδ2 CAR-T cell metabolism.
As part of that analysis, TXNIP emerged as a potentially actionable target based on its high expression and mitochondrial localization in γδ T cells. This pleiotropic protein is a well-characterized regulator of redox homeostasis, primarily through its inhibitory interaction with thioredoxin 1 (Trx-1)(68) and thioredoxin 2 (Trx-2) (69) (70), and as inductor of apoptosis (71). In addition, TXNIP is known to reduce GLUT1 expression, inhibit glucose uptake(72), and limit effector T cell proliferation(73); and its loss enhances IFNγ production and tumor cell killing by T cells(48). In addition, it has been described that a transient decrease in TXNIP following T cell stimulation is necessary for CD28-driven metabolic priming(74) in αβ T cells. However, to the best of our knowledge, its specific role in γδ T cells had not been described. We hereby provide evidence that pharmacological inhibition of TXNIP using SRI-37330 (50) during γδ ex-vivo expansion increased their mitochondrial mass, spare respiratory capacity, and CAR expression levels.
Functionally, CAR γδ T cells treated with SRI-37330 exhibited significantly increased production of proinflammatory cytokines, including Granzyme B, TNFα, and IL-2. Together, our results support a role for TXNIP in the metabolic phenotype of γδ T cells and warrant further exploration of the translational potential of TXNIP inhibition as an enhancer of CAR-T cell products.
Another key finding of our study was the differential activation of the canonical TCR signaling driven by CAR engagement. Previous studies have reported a divergence between αβ and γδ TCR signaling(35, 75, 76), which as attributed to structural variations in the TCR complex components, such as the composition and stoichiometry of CD3 chains(35). This model, however, does not explain the differences that we reported, because we employed the exact same CAR construct as the signal trigger in both cell subsets. Therefore, we conclude that γδ T cells present subset-intrinsic traits that condition the response to CAR-induced signaling, including lower activation of NFAT and AP-1 (JUN) signaling. These pathways are of high relevance for T cell proliferation and survival, and JUN signaling deficiencies have been associated with induction of T cell exhaustion (55, 56). With that idea in mind, we tested the effects of supplementing γδ CAR-T cells with a costimulatory signal designed to boost AP-1 signaling. To provide tissue specificity, we made that signal conditional to the presence of RANKL, a cytokine that is highly prevalent in the tumor/bone microenvironment, using a chimeric switch receptor. Two CSR versions were generated, containing either CD27 or 41BB costimulation. While both molecules are expected to induce AP-1 signaling, CD27 was selected based on its known role in survival of γδ T cells through induction of antiapoptotic genes (66) and its lower expression in comparison to αβ T cells, as reported in this manuscript. In turn, 41BB was chosen based on its ability increase oxidative phosphorylation in αβ CAR-T(77). Eventually, we determined that the CD27-containing CSR showed superior RANKL binding, and verified that it induced AP-1 signaling, preserved a central memory phenotype, and promoted CAR-T cell persistence without compromising antitumor efficacy.
These findings reinforce the notion that cellular product design requires tailoring to the lineage context, emphasizing the need for T cell type-specific synthetic receptors optimization. Importantly, our results laid the foundations for the development and translation of novel CAR-T cell products integrating TXNIP inhibition and/or armoring with CSRs. Current and future efforts are focused on the investigational new drug (IND)-enabling studies to further characterize the pharmacological and toxicological profiles of our new products.
MATERIALS AND METHODS
Cell lines
C4–2B (CRL-3155, androgen-independent) cell line was purchased from the American Type Culture Collection and cultured as recommended. PSCA expression was induced by transduction of C4–2B cells with a retroviral vector encoding the codon-optimized cDNA for PSCA. Deidentified Healthy Donor Buffy Coats were obtained from LifeSouth community blood centers or OneBlood (Florida Blood Services, FL). All cell lines were periodically mycoplasma tested (MycoAlert, Lonza).
γδ and αβ T cells activation and expansion from PBMCs
PBMCs were isolated from buffy coats by density-gradient centrifugation. Briefly, buffy coats were diluted 1:2 in PBS (1×, pH 7.4, room temperature). In 50-mL tubes, 30 mL of diluted buffy coat were layered over 15 mL of LSM. Tubes were centrifuged for 20 min at 2500 rpm and 20°C with minimal braking. PBMC monolayers were collected, washed twice with PBS (5 min at 1500 rpm, 20°C), resuspended in 3–4 mL ACK buffer, and incubated for 10 min to lyse red blood cells. After a final PBS wash, cells were resuspended in RPMI (5% FBS, antibiotic, 100 IU/mL IL-2). Following cell counting, γδ T cells were expanded as described(78): PBMCs were resuspended at 1×10^6^ cells/mL with 4 μM Zol and plated in 24-well plates (2 mL/well); media and Zol were refreshed on day 3(79), and cells were used for transduction on day 5. αβ T cells were expanded by resuspending PBMCs at 1×10^6^ cells/mL in X-Vivo (5% human serum, antibiotics, 300 IU/mL IL-2) with 5 μg/mL OKT3, plated in 24-well plates (2 mL/well), and transduced 2 days after stimulation.
T cells retroviral transduction
The CAR (pMSGV1-PSCA-8t-28z) and CSR (pMSGV1-RANK-CD27-P2A-CD34 or pMSGV1-RANK-4–1BB-P2A-CD34) plasmids were encapsulated in a RD114-pseudotyped retrovirus generated by transient transfection of 293GP cells (80–82). γδ and αβ T cell obtained from PBMCs stimulation were transduced two consecutive days in 6-well plates coated with RetroNectin ^®^ and the virus after spinoculation. For greater detail about the transduction protocol and quality control of the cells please refer to the supplementary materials and methods.
Due to the variability of γδ T cell purity within donors, we established a threshold of > 75% γδ T cells out of total CD3^+^ cells to conduct further experiments, unless in the experiment in question we could gate the γδ T cell population by flow cytometry. When necessary, negative selection enrichment of γδ T cells was performed using magnetic columns and the human TCRγ/δ + T Cell Isolation Kit, following Miltenyi recommended protocol.
Cytokine quantification
T cell activity in the presence or absence of target cells was assessed by cytokine quantification. Cocultures were established in U-bottom 96-well plates with 1×10^5^ C4–2B PSCA+ tumor cells/well and a 1:1 ratio of total T cells; single cultures served as negative controls. After overnight incubation, supernatants were collected and frozen for ELISA or ELLA analysis. ELISAs were performed using 2G1 (capture) and B133.5-biotin (detection) anti–IFN-γ antibodies, with HRP-streptavidin, TMBA, and 0.08 M H_2_SO_4_ for detection; absorbance was read at 450/550 nm. Granzyme B, IFN-γ, IL-2, and TNF-α ELLA assays were performed using pre-coated cartridges per manufacturer instructions.
Single-cell cytokine profiling
CAR-T polyfunctionality was assessed using the Bruker Single-Cell Adaptive Immune Secretome assay. Briefly, seven days post-transduction, αβ CAR-T cells were separated into CD4+ and CD8+ subsets, and compared with enriched γδ CAR-T cells from the same donor. Cells were stimulated for 2 h with PMA/ionomycin or left unstimulated. After viability staining, 30,000 cells were loaded per IsoCode chip and incubated for 16 h to monitor cytokine secretion. Polyfunctionality percentages were computed using IsoLight software.
AP-1 transcription factor activation
AP-1 activation was measured by quantifying nuclear JUN-D, c-Fos, and phospho–c-JUN abundance in activated cells. 5×10^6^ UT, CAR, or Switch-CAR γδ T cells from three donors were stimulated for 1 h at 32°C with plate-bound hr-PSCA-Fc (0.5 μg/mL) and soluble hr-RANKL (1 μM). Cells were washed, pelleted, and nuclear proteins extracted using the ActiveMotif Nuclear Extract kit. AP-1 complex components were quantified by colorimetry using the TransAM^®^ AP-1 kit as described by manufacturer.
T cell stimulation with plate bound molecules
When target-cell stimulation was not feasible, T cells were activated using OKT3- or PSCA-Fc-coated plates. Coating was performed by diluting OKT3 (0.5 mg/mL) or PSCA-Fc (0.2 mg/mL) in PBS and incubating plates for 3 h at 37°C, 5% CO_2_, followed by two PBS washes. T cells (1×10^6^ cells/mL in RPMI with 5% FBS, antibiotics, and 100 IU/mL IL-2) were plated at 200 μL (96-well) or 2 mL (24-well) for the indicated stimulation times.
Phenotype characterization
Memory, costimulatory, and exhaustion markers (CD45RA, CD28, CD27, CD69, TIM-3, PD-1) were assessed on unstimulated cells by staining for 20 min at 4°C, followed by flow cytometry on an LSRII (FACSDiva). Data were analyzed using FlowJo.
Phosphoproteomics analysis by stable isotope labeling by amino acids in cell culture (SILAC) MS.
C4–2B PSCA+ cells were cultured in RPMI supplemented with heavy lysine and arginine (^1^^3^C_6_^14^N_4_-arginine, 200 mg/L; ^13^C_6_-lysine, 40 mg/L) until > 95% of proteins were labeled. Twenty million tumor cells per sample were cocultured with T cells at a 1:2 total T-cell ratio or 2.8:1 CAR-T ratio (37°C, 5% CO_2_). Six triplicate conditions were prepared to assess basal phosphorylation in (1) untransduced (UT) γδ or (2) UT αβ T cells after 1 h, CAR-induced basal phosphorylation in (3) γδ or (4) αβ CAR-T cells at 0 h, and antigen-induced phosphorylation in (5) γδ or (6) αβ CAR-T cells after 1 h. After incubation, cells were washed twice with cold PBS, pelleted, and flash-frozen for protein extraction and MS (82). Cells were lysed, protein concentration measured by Bradford assay, and aliquots of 2.8 mg (pY) and 200 mg (global phosphorylation/expression) were prepared. Proteins were reduced, cysteines alkylated, digested with trypsin, and lyophilized prior to pY enrichment.
A 24-mg pooled bulk sample was also prepared. Disulfides were reduced, cysteines alkylated, and samples digested overnight (1:20 trypsin:substrate). Peptides were acidified with 1% TFA, desalted on C18 Sep-Pak cartridges, and lyophilized. pY-enriched samples were labeled with TMT11plex; global phosphorylation samples were quenched with 5% hydroxylamine, pooled by plex, and lyophilized. Phosphopeptides were enriched by IMAC and basic reversed-phase chromatography before MS. Samples were analyzed on a nanoflow UHPLC coupled to an Orbitrap mass spectrometer, acquiring the top 20 MS/MS spectra in data-dependent manner (resolution 45,000; 1E5 AGC target; MaxIT 86 ms for global and 300 ms for pY; isolation window 0.8 with 0.2 offset; first mass m/z 100; NCE 24 and 30). A detailed protocol is provided in the Supplementary Materials and Methods.
Data Analysis and availability
Peptides were identified using the UniProt human database. Threshold criteria are defined in more detail in the supplementary materials and methods. Reporter ion intensities were used for the relative quantification of each peptide in the TMT global pSTY data and pY data. Both data sets were normalized using IRON (iron_generic–proteomics(83)) against the 5x pooled bulk sample channels within each plex. The three injection replicates in different Plex were average together to calculate Log_2_ ratios between conditions by subtracting averaged sample replicates. Welch’s t-test was used to determine statistical significance in the difference between the log_2_ ratio. Data normalization was evaluated with scatterplots, and PCA analysis separated samples by the groups expected on the experimental design. The phosphoproteomics data are available from PRIDE/ProteomeXchange using the dataset identifier, PXD007085, and 10.6019/PXD007085.
Pathway enrichment analysis
Phosphorylation differences between conditions were calculated with log_2_ ratios and p values obtained from Welch’s t test where there were at least two replicates of phosphorylation abundance per phospho-site. Resulting log_2_ ratios, p-values, and gene IDs per phospho-site were then analyzed in the Qiagen Ingenuity Pathway Analysis (IPA) software. Canonical pathways predictions were obtained with an IPA phosphorylation analysis using our own data set as reference, setting cutoffs at p-value ≤ 0.05, and log_2_ ratio ≥ 0.585 or ≤ −0.585.
Western Blot
T cells from three healthy donors were cocultured with C4–2B PSCA+ cells to induce phosphorylation of the CAR CD3ζ domain after 1, 10, 30, or 60 min of activation. Whole-cell lysates were prepared in RIPA buffer with protease and phosphatase inhibitors. CAR constructs were immunoprecipitated using protein-L magnetic beads as described (82). Twenty-five micrograms of protein per lane were resolved by SDS-PAGE and transferred to nitrocellulose using the Bio-Rad turbo protocol. Membranes were blocked in PBS + 5% BSA for 1 h at room temperature and incubated overnight at 4°C with primary antibodies against CAR-CD3ζ and phospho-sites pY72, pY83, and pY142. Detection was performed using TidyBlot and imaged on an Odyssey Fc. Band intensities were quantified in Image Studio (v6.0) and normalized to the total CAR-CD3ζ loading control.
SeaHorse Analysis: Glycolytic Rate and T cell metabolic profiling
T cells were stimulated in OKT3-coated or uncoated plates for 48 h. Seahorse assays (Glycolytic Rate Assay or T Cell Metabolic Profiling Kit) were performed following manufacturer instructions. Briefly, cartridges were hydrated in 200 μL/well molecular-grade water in a 37°C heat-only incubator for 24 h, then water was replaced with Seahorse XF calibrant (200 μL/well) and incubated for 1 h. In parallel, T cells were collected, washed, and resuspended in Seahorse XF RPMI (10 mM glucose, 2 mM glutamine, 1 mM pyruvate) at 2×10^6^ cells/mL. Next, 50 μL/well (1×10^5^ cells) were plated in 96-well sample plates with ≥ 4 technical replicates and four background wells. Plates were centrifuged at 200 g for 2 min, monolayer formation was verified, and plates were incubated for 30 min in the same incubator.
For Glycolytic Rate Assays, 130 μL/well pre-warmed media were added (final 180 μL/well). Rotenone/antimycin A (Rot/AA, 50 μM, 20 μL/port) and 2-Deoxy-D-glucose (2-DG, 500 mM, 22 μL/port) were loaded into ports A and B, respectively. For Metabolic Profiling, 150 μL/well pre-warmed media were added (final 200 μL/well), and ports A, B, and C were loaded with oligomycin A (13.5 μM, 25 μL/port), BAM15 (25 μM, 25 μL/port), and Rot/AA (5.5 μM, 25 μL/port). Assays were run on a Seahorse XF Pro analyzer using Wave Pro software and manufacturer-recommended analysis templates.
Glucose uptake
T cells were stimulated in OKT3-coated or uncoated plates for 24 h, lifted, and washed with warm PBS. Cells were resuspended in glucose-free media at 1×10^6^ cells/mL and incubated for 1 h at 37°C, 5% CO_2_. Glucose-Cy5 (0.2 μM) was then added and incubated for 20 min. Cells were washed with cold PBS, stained for CAR, CD3, TCRVδ2, and TCRαβ or TCRVδ1 for 20 min, and gated using DAPI for viability. Flow cytometry was performed immediately afterward.
Mitochondrial mass quantification
MitoTracker Deep Red (MTDR) was reconstituted per manufacturer instructions and diluted 3:1000 in pre-warmed PBS + 2% BSA. T cells stimulated for 24 h (in OKT3-coated plates) and control were washed with warm PBS + 2% BSA and stained with CD3, TCRαβ, TCRVδ2, CD4, CD8, and MTDR for 20 min at 37°C. Cells were washed and resuspended in DAPI for viability before flow cytometry. PBMCs were stained similarly with additional CD19 and TCRVδ1 antibodies.
Mitochondrial potential quantification
T cells were stimulated in OKT3-coated or uncoated plates for 24 h, then lifted and washed with warm (37°C) PBS. Cells were resuspended in warm RPMI (10% FBS, antibiotics) at 1×10^6^ cells/mL, and each sample was split into two sets. Carbonyl cyanide m-chlorophenylhydrazone (CCCP) was added to one set (50 μM) and incubated for 15 min at 37°C. Tetramethylrhodamine ethyl ester (TMRE, 160 nM) was then added to both sets and incubated for 30 min at 37°C. Cells were washed with cold PBS and kept on ice for viability and surface staining. A PBS-diluted yellow fixable viability dye was added for 10 min, followed by a wash in cold PBS + 2% BSA and staining with CD3, TCRαβ, and TCRVδ2 antibodies for 20 min. After a final wash, samples were analyzed by flow cytometry. ΔMFI was calculated by subtracting each sample’s TMRE MFI from its paired CCCP-treated TMRE MFI.
TXNIP inhibition
PBMCs were stimulated with 4 μM Zol and 5 μM SRI-37330 and plated in 24-well plates (2×10^6^ cells per well as describe above). Media were replaced with fresh Zol and SRI-37330 on day 3. UT T cells were used on day 5 for metabolic analyses. Transductions were performed as usual, maintaining inhibitor-containing media with half-volume replacements every other day.
In vivo mouse model
All procedures were approved by the University of South Florida IACUC (14112R) and following the Guidelines for the Care and Use of Laboratory Animals manual published by the National Institutes of Health. Male 6-week-old NSG mice (Jackson Laboratory, #005557; Bar Harbor, ME) were intra-tibially injected with 5×10^5^ luciferase-expressing C4–2B PSCA+ cells in 20 μL PBS(21). Tumor burden was monitored twice weekly by bioluminescence imaging (IVIS^™^ ILUMINA 200, Perkin Elmer) after D-luciferin injection. Two weeks later, mice were randomized into five groups and treated via retro-orbital injection with PBS, UT, CAR, CAR/RANK-CD27, or CAR/RANK-4–1BB γδ T cells (1.2×10^7^ cells/mouse). Mice received 100 IU IL-2 IP every 48 h for two weeks. At endpoint, spleen, blood, and hind limbs were collected. T cells were isolated from spleen by mechanical dissociation and from tibial bone marrow by centrifugation(84). Cell were cryopreserved for flow cytometry. Additional details are in supplementary materials and methods.
Statistical Methods
Statistical tests are specified in each figure. Phospho-site abundance data show means with individual points from three replicates run in duplicate; Welch’s t tests were applied when ≥ 2 replicates yielded values. Flow cytometry, Seahorse assays, and western blots were analyzed using paired t-tests, with donor samples represented by distinct symbols. Significant differences (p ≤ 0.05) are marked with horizontal lines and asterisks. Data are shown as means ± SD unless indicated. Analyses were performed in GraphPad Prism 9.1.1, except phospho-site analyses, which were conducted by the Moffitt Biostatistics and Bioinformatics Shared Resource.
Supplementary Material
Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Wang CQ, Lim PY, Tan AH. Gamma/delta t cells as cellular vehicles for anti-tumor immunity. Frontiers in immunology. 2023;14:1282758.38274800 10.3389/fimmu.2023.1282758 PMC 10808317 · doi ↗ · pubmed ↗
- 2Garber K. Gammadelta t cells bring unconventional cancer-targeting to the clinic - again. Nat Biotechnol. 2020;38:389–91.
- 3Raverdeau M, Cunningham SP, Harmon C, Lynch L. Gammadelta t cells in cancer: A small population of lymphocytes with big implications. Clin Transl Immunology. 2019;8:e 01080.31624593 10.1002/cti 2.1080 PMC 6787154 · doi ↗ · pubmed ↗
- 4Yazdanifar M, Barbarito G, Bertaina A, Airoldi I. Gammadelta t cells: The ideal tool for cancer immunotherapy. Cells. 2020;9.33375150 10.3390/cells 10010009 PMC 7822128 · doi ↗ · pubmed ↗
- 5Park JH, Lee HK. Function of gammadelta t cells in tumor immunology and their application to cancer therapy. Exp Mol Med. 2021;53:318–27.33707742 10.1038/s 12276-021-00576-0PMC 8080836 · doi ↗ · pubmed ↗
- 6Groh V, Steinle A, Bauer S, Spies T. Recognition of stress-induced mhc molecules by intestinal epithelial gammadelta t cells. Science. 1998;279:1737–40.9497295 10.1126/science.279.5357.1737 · doi ↗ · pubmed ↗
- 7Holmen Olofsson G, Idorn M, Carnaz Simoes AM, Aehnlich P, Skadborg SK, Noessner E, Vgamma 9vdelta 2 t cells concurrently kill cancer cells and cross-present tumor antigens. Frontiers in immunology. 2021;12:645131.34149689 10.3389/fimmu.2021.645131 PMC 8208807 · doi ↗ · pubmed ↗
- 8Brandes M, Willimann K, Moser B. Professional antigen-presentation function by human gammadelta t cells. Science. 2005;309:264–8.15933162 10.1126/science.1110267 · doi ↗ · pubmed ↗