B7-H3–Targeted CAR-Vδ1T Cells Exhibit Potent Broad-Spectrum Activity against Solid Tumors

Cancer poses a substantial global health challenge, driven by increasing incidence and mortality rates (1). Treating solid tumors has consistently posed a significant challenge. Conventional treatments like surgery, radiotherapy, and chemotherapy have limited effectiveness and frequently result in significant side effects, especially for patients with metastatic or recurrent diseases (2). Over the past decade, immunotherapy has emerged as a promising approach for hematologic malignancies. Adoptive cell therapy, particularly chimeric antigen receptor T (CAR-T) cell therapy, has demonstrated remarkable clinical efficacy (3–5). Despite extensive research, CAR-T cell therapy’s clinical outcomes in solid tumors have not matched the success seen in hematologic malignancies. This discrepancy is primarily attributed to challenges related to the tumor immune microenvironment, tumor infiltration, and antigen heterogeneity (6–10). Advancements in solid tumor treatment, including antibody–drug conjugates and immune checkpoint inhibitors, have now expanded the landscape of cell-based therapies. Notably, T-cell receptor T cell, tumor-infiltrating lymphocytes, and other cell therapies are now gaining prominence (11–13).

In recent years, γδT cells have gained attention due to their potential in antitumor immunity (14–15). T cells are divided into two major subsets based on their TCR chains: αβT cells and γδT cells. Human γδT cells are further subgrouped into Vδ1T, Vδ2T, Vδ3T, and Vδ5T cells, with Vδ1T and Vδ2T being the main subgroups. Vδ2T cells are the predominant subset of γδT cells in peripheral blood mononuclear cells (PBMC), whereas Vδ1T cells are less abundant in peripheral blood and are mainly distributed in epithelial and mucosal tissues, such as the skin, intestines, spleen, and liver (16–18). However, Vδ1T cells stand out as promising candidates for solid tumor treatment due to their advantageous biological properties: (i) Vδ1T cells recognize tumor antigens independently of MHC antigen presentation, making them ideal candidates for developing “off-the-shelf” cell therapy products without gene editing. Additionally, they can target and eliminate MHC class I–deficient tumors, bypassing the immune evasion mechanisms often employed by tumors (19–20). (ii) Vδ1T cells are tissue-resident lymphocytes that perform distinct and specialized roles in immune surveillance of human tissues, integrating the fundamental functions of both innate and adaptive immune cells (17). Vδ1T cells exhibit innate tumor homing capabilities, allowing them to efficiently migrate to tumor sites and infiltrate solid tumors compared with other immune cell types (21). (iii) Vδ1T cells have been associated with favorable prognosis in various types of cancer, indicating their potential as a prognostic biomarker and therapeutic target (22–25). (iv) Vδ1T cells integrate pathways of classical TCR signaling and costimulatory signaling, as well as activation mechanisms found in NK cells, such as NKG2D, DNAM-1, NKp46, NKp44, NKp30, and other signaling pathways (14). Therefore, Vδ1T cells, equipped with diverse tumor recognition mechanisms, hold promise for addressing tumor antigen heterogeneity.

Despite the predominant focus on Vδ2T cells in research and clinical trials due to their well-established isolation and activation methods, clinical outcomes in solid tumor therapy have fallen short of expectations (26). Moreover, Vδ1T cells predominantly inhabit tissues, with significantly lower representation in peripheral blood. Their isolation and activation require sophisticated techniques (16, 21, 27–29). Consequently, Vδ1T cells received limited attention from researchers, impacting the pace of their clinical development (26). Although clinical trial data on Vδ1T cells remain limited, recent reports highlight promising clinical outcomes and favorable safety profiles for CD20-CAR-Vδ1T cells in treating relapsed/refractory B-cell malignancies (30). This demonstrates that CAR-modified Vδ1T cells present substantial potential in tumor therapy. This study established a method for preparing CAR-Vδ1T cells and systematically evaluated their effects across various solid tumor models, providing a novel option for cellular therapy in solid tumors.

In our target selection, we focused on B7-H3, a pan-tumor antigen. B7-H3 exhibits high expression in various solid tumor cells while being nearly absent or minimally expressed in normal tissues. Currently, immunotherapeutic strategies targeting B7-H3 are being developed, including mAbs, antibody–drug conjugates, and immune cell therapies involving gene-modified T cells or NK cells (31). B7-H3 is a type I transmembrane protein within the B7 protein family. It consists of 316 amino acids and exhibits both immunosuppressive and protumor functions (32–33). Their roles encompass T-cell inhibition, conversion of macrophage phenotypes from M1 to M2, promotion of angiogenesis, tumor cell survival, proliferation, and drug resistance (34). Additionally, its expression on tumor cells correlates with unfavorable outcomes (35–36).

Therefore, we hypothesize that allogeneic B7-H3–targeting B7-H3-CAR-Vδ1T cells could serve as a novel approach for solid tumor therapy. Our study introduces a GMP-compliant protocol to generate a substantial number of these cells from peripheral blood. We demonstrate their effective cancer surveillance across diverse cancer cell lines and xenograft models, both in vitro and in vivo. Additionally, we enhance their tissue homing and functional persistence by coexpressing IL2.

This study utilized five distinct B7-H3–positive solid tumor cell lines: SK-N-AS (neuroblastoma, RRID: CVCL_1700), SW1990 (pancreatic cancer, RRID: CVCL_1723), A549 (non–small cell lung cancer, RRID: CVCL_0023), HCT-15 (colorectal cancer, RRID: CVCL_0292), and MDA-MB-231 (breast cancer, RRID: CVCL_0062). The cell lines were cultured under specific conditions: SK-N-AS in DMEM supplemented with 10% FBS and 1% Non-Essential Amino Acids; SW1990, HCT-15 and MDA-MB-231 cells in DMEM with 10% FBS; A549 in F12K medium with 10% FBS; and Raji (RRID: CVCL_0511), Raji-B7-H3, and AML-3 (RRID: CVCL_1844) cells in RPMI 1640 medium with 10% FBS. All cells were tested for Mycoplasma before using in the study with Mycoplasma Detection Kit (ExCell Bio, MB000-1591) and were authenticated using short tandem repeat profiling, and the corresponding identification reports were accessible. All cell lines were sourced from Nanjing Cobioer Biosciences, except for SW1990 and Raji, which were obtained from ATCC.

The CAR construct was assembled from several components: the anti–B7-H3 single-chain variable fragment (scFv), an Fc hinge, CD28 and CD3ζ transmembrane domains, and optionally, a sIL2 linked by a T2A self-cleaving sequence. This construct was synthesized and subsequently cloned into the lentiviral plasmid pCDH-CMV-MCS-EF1-CopGFP. The accuracy of the construct was verified through sequencing at GENEWIZ Bio. Lentivirus was then transiently generated in suspension 293T cells using a transfection reagent (polyethyleneimine) and four plasmids: (i) pCDH-CAR; (ii) pRSV-Rev (RRID: Addgene_12253); (iii) pMDLg/pRRE; and (iv) pMD2G. Following this, the viral harvests were gathered and concentrated via ultracentrifugation.

PBMCs were procured from commercially available whole blood leukocyte cones (BOKANG BIOENGINEERING Bio) and isolated using Ficoll-Paque PLUS density gradient media (GE, 17-1440-03) as per the manufacturer’s guidelines. Vδ1T cells were enriched via an Anti-Human TCR Vδ1 Antibody (Miltenyi Biotec, 130-120-440) and Anti-PE MicroBeads (Miltenyi Biotec, 130-048-801), followed by a 48-hour in vitro activation. These activated Vδ1T cells were subsequently transduced with the lentivirus CAR construct. After transduction, the cells were expanded in X-VIVO medium (Lonza, 04-418QCN) supplemented with 100 IU/mL of human IL2 (Novoprotein, GMP-CD66), 20 ng/mL of human IL15 (Novoprotein, GMP-C016), and 30 ng/mL of human IL21 (Novoprotein, GMP-CC45). Irradiated artificial antigen-presenting cells (aAPC) and K562 cells (RRID: CVCL_0004) overexpressing membrane CD64, CD86, CD137L, and IL15-eGFP (Co60/50 Gy; Supplementary Fig. S1A) were introduced during the cell culture cycle at a ratio of 2:1 (aAPC to Vδ1T). The CAR-Vδ1T cell culture was scaled up in vitro using a WAVE Bioreactor (GE). Vδ2T cells were enriched using a PE antihuman Vδ2T antibody (BioLegend, clone: B6) and Anti-PE MicroBeads (Milenyi Biotec, 130-048-801), and the subsequent CAR-Vδ2T cells were prepared identically to the CAR-Vδ1T cells and expanded in culture flasks. Isolated PBMCs were activated with Dynabeads Human T-Activator CD3/CD28 (Gibco, 11131D) for 48 hours and transduced with the lentivirus CAR construct. The culture medium for CAR-αβT was identical to that of CAR-Vδ1T and CAR-Vδ2T, with the exception that aAPC was not included during the culture process.

The purity of Vδ1T cells, Vδ2T cells, and αβT cells was ascertained via flow cytometry using the following antibodies: anti-human CD3 (RRID: AB_10774514, BD, Clone: SK7), anti-human Vδ1T (Invitrogen, clone: TS8.2), anti-human Vδ2T, and anti-human TCR-αβT (BioLegend, clone: IP26). A custom-made anti-CAR idiotype antibody (iCarTab Biotechnology) was utilized against the B7-H3-scFv fragment to determine the positive rate of B7-H3-CAR. The expression of natural cytotoxic receptors in CAR-Vδ1T cells was separately stained with anti-human NKG2D (BioLegend, clone: 1D11), anti-human DNAM1 (BioLegend, clone: 11A8), anti-human NKp30 (BioLegend, clone: P30-15), antihuman NKp44 (BioLegend, clone: P44-8), and anti-human NKp46 (BioLegend, clone: 9E2) antibodies. The B7-H3 expression in tumor cells was detected using the anti-human B7-H3 antibody (Miltenyi Biotec, Clone: REA1094). All aforementioned antibodies were diluted at a 1:100 ratio and applied for staining at room temperature for 20 minutes. Following staining, the cells were rinsed with PBS prior to analysis. All samples were examined using an ACEA/NovoCyte flow cytometer, and a portion of the data was processed using FlowJo software (RRID: SCR_008520).

The real-time inhibitory effect of Vδ1T and CAR-Vδ1T cells on the growth of solid tumor cells was monitored in vitro using a real-time cell analysis instrument (ACEA/xCELLigence). Tumor cells (5,000–10,000 cells in 100 μL volume) were seeded into well plates with chips 1 day before. After 16 to 24 hours of tumor cell growth, varying numbers of effector cells were added based on different effector–target ratios. The final coincubation volume was 200 μL, and the coincubation medium used was the medium for tumor cells without any cytokines. Tumor cell growth was monitored in real-time at 37°C in a 5% CO₂ incubator.

Prior to coincubation with Vδ1T or CAR-Vδ1T cells, target cells were stained with carboxyfluorescein succinimidyl amino ester (CFSE) at a temperature of 37°C for a duration of 15 minutes. The coincubation system was set up with 2 × 10⁵ target cells per well, and varying quantities of effector cells were introduced based on the predetermined effector–target ratios. Following a 24-hour incubation period, the cells were harvested. Subsequently, 5 μL of 7-AAD antibody and 10 μL of Annexin V (sourced from Multi Sciences Biotech) were employed to detect early and late apoptosis in target cells via flow cytometry. In addition, the supernatant from the coincubation was collected for a subsequent cytokine release assay.

The persistence of antitumor efficacy of CAR-Vδ1T cells was assessed via an antigen-repeated stimulation assay. For each antigen stimulation, 2 × 10⁵ CFSE-stained target cells were used, and Vδ1T or CAR-Vδ1T cells were introduced based on the effector–target ratio during the initial stimulation. Following a 24-hour stimulation period with target cells, a subset of coincubated cells was harvested to evaluate tumor cell apoptosis using flow cytometry. The remaining cells were then subjected to an additional 24-hour incubation with fresh CFSE-stained target cells, and this process was reiterated as per the number of stimulations. The coincubation medium employed in this study was RPMI 1640, supplemented with 10% FBS and devoid of any cytokines. The coincubation volume was maintained at 1 mL, and a fraction of the coincubation supernatant was collected prior to each introduction of tumor cells for subsequent cytokine detection.

Cytokines present in the coincubation supernatant were quantified by flow cytometry, employing the cytokine CBA kit (BD Biosciences) in accordance with the manufacturer’s guidelines. The panel of cytokines detected included IL2 (catalog number: 558270), IFNγ (catalog number: 558269), TNFα (catalog number: 558273), IL17A (catalog number: 560383), granzyme B (catalog number: 560304), and IL6 (catalog number: 558479). The analysis was conducted using FCAP Array v1.0 software.

All animals used in the experiment were approved by Soochow University Animal Ethics Committee. For the animal preclinical studies, we used the ARRIVE1 reporting guidelines. All mice in this study were cultured in a specific pathogen-free grade animal laboratory.

The study employed female NPG mice (NOD-Cg. PrkdcSCID IL2Rgcnull/vst), ages 5 to 6 weeks, which are highly immunodeficient models developed by Beijing Weitongda Biotechnology Co., Ltd. For each solid tumor model, specific tumor cells were either subcutaneously or orthotopically injected, followed by a single intravenous infusion of various types of CAR-T cells. To establish the MDA-MB-231 orthotopic tumor model, postdigestion of tumor cells, a cell suspension of 2 × 10⁷/mL was prepared using PBS containing 50% matrix gel. The mice were then anesthetized with 1.25% tribromoethanol, and 50 μL of the cell suspension was injected into the mammary fat pad of the mice. The SK-N-AS neuroblastoma model was inoculated with 3 × 10⁶ tumor cells per mouse. The pancreatic cancer SW1990 model, lung cancer A549 model, and colorectal cancer HCT-15 model each received a transplantation of 1 × 10⁶ tumor cells per mouse. In all mouse models, the number of transfused effector cells ranged from 1 × 10⁷ to 1.5 × 10⁷ cells per mouse. Throughout the experiment, the tumor volume and body weight of all mice were regularly monitored to track their growth status. The tumor volume was calculated using the formula: Volume = (D × d² × π)/6 (D represents the long diameter; d represents the short diameter).

The peripheral pharmacokinetics of CAR-T cells, as well as the infiltration of CAR-T cells into tissues, were assessed by flow cytometry. This was achieved by staining with an anti-human CD3 antibody (sourced from BD Biosciences, clone: SK7) and an anti–B7-H3-scFv antibody (sourced from iCarTab Biotechnology). The phenotypic characterization of CAR-T cells was conducted using anti-human CD45RA (BioLegend, clone: HI100) and CD27 (BioLegend, clone: M-T271) antibodies. Additionally, the expression of PD-1 was determined by staining with an anti-human PD-1 antibody (BioLegend, clone: EH12.2H7).

In the study involving the AML-3 acute myeloid leukemia model, AML-3 cells (1 × 10⁶ tumor cells per mouse) were administered via i.v. injection. Following a 7-day inoculation period, either PBS or CAR-Vδ1T cells were infused intravenously in a single dose. Throughout the entire experiment, regular assessments were conducted, including fluorescence imaging, weight monitoring, and tracking the proliferation of CD3⁺ T cells in peripheral blood. Additionally, survival curves were plotted to monitor the progress of the study.

The collected tissues including the heart, liver, spleen, lung, and kidney from the mice were fixed in 10% paraformaldehyde formalin, embedded in paraffin, sectioned, stained, and examined under a microscope to observe any inflammation or tissue injury. The examination was conducted by Cancercell Bio in Suzhou.

In the antigen-repeated stimulation assay, AG-CAR-Vδ1T and AQ-CAR-IL2-Vδ1T cells were stimulated with Raji-B7-H3 cells four times. Following stimulation, the cell suspension was collected and sent to Singleron Biotechnology for single-cell sequencing analysis. Quality control, dimensionality reduction, and clustering were performed using Seurat v3.1.2 (RRID: SCR_016341). Gene expression profiles were generated from SynEcoSysTM raw reads using CeleScope v1.5.2 (RRID: SCR_023553) with default parameters.

RNA was extracted from two group of CAR-Vδ1T cell suspension with the same number cells (5 × 10⁶) using RNA Isolation Kit (TIANGEN, DP419), and 1 μg RNA was reversely transcribed into cDNA with RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1622). Then PCR system was performed using 2× Color SYBR Green qPCR Master Mix (EZBioscience, A0012-R2), and gene-specific upstream and downstream primers were designed for PCR amplification according to target genes and internal control GAPDH, primers are as follows:

• TNF (F′: GCC​CAT​GTT​GTA​GCA​AAC​CC; R′: TGA​GGT​ACA​GGC​CCT​CTG​AT);

• IFNγ (F′: GAGTGTGGAGA CCATCAAGGA; R′: TGG​ACA​TTC​AAG​TCA​GTT​ACC​GAA);

• PRF1 (F′: GGGCTGATGCCACCATT; R′: GGC ACTTGGGCTCTGGAAT);

• MKI67 (F′: GGA​TCG​TCC​CAG​TGG​AAG​AG; R′: TCT​CGT​GGG​CCA​CAT​TTT​CT);

• CXCR6 (F′: AAG​CAT​CTC​TGC​TGG​TGT​TCA; R′: CCCACCAGACCACAGACAAA);

• CCR5 (F′: AAACTCT CCCCGGGTGGAAC; R′: AGCATGTTGCCCACAAAACC);

• IL2 (F′: TAC​ATG​CCC​AAG​AAG​GCC​AC; R′: TTG​CTG​ATT​AAG​TCC​CTG​GGT);

• GAPDH (F′: GATTCCACCCATGGCAAATTC; R′: CTGGAAGATGGTG ATGGGATT).

Statistical analysis was performed using GraphPad Prism V8.01 (GraphPad Prism, RRID: SCR_002798). Significance between two independent groups was analyzed using an unpaired Student t test, whereas significance among multiple groups was analyzed using either one-way ANOVA or two-way ANOVA. The survival curve was performed using the log-rank (Mantel–Cox) test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, nonsignificant results.

Single-cell RNA sequencing data generated in this study are publicly available in Gene Expression Omnibus at GSE275242. All other raw data generated in this study are available upon request from the corresponding author.

Expanding Vδ1T cells in vitro is challenging compared with Vδ2T cells. Although some protocols have demonstrated successful expansion (16, 27–29), their clinical application lags behind that of Vδ2T cells (26). Here, we present a novel method for expanding Vδ1T cells from peripheral blood. In our study, we combine this approach with lentiviral transduction of a B7-H3-CAR construct, resulting in B7-H3-CAR-Vδ1T cells using a clinical-grade integrated protocol (Fig. 1A). After direct enrichment of Vδ1T cells from healthy donor PBMCs using an anti-human TCR-Vδ1T antibody sorting kit, the cells were activated in vitro for 48 hours. Subsequently, lentivirus-mediated transduction introduced the B7-H3-CAR construct. The transduced cells were then stimulated and amplified with irradiated aAPCs and cytokines (IL2, IL15, and IL21). Finally, enriched Vδ1T cells or CAR-Vδ1T cells were cryopreserved after 19 days of expansion in vitro.

In our study, Vδ1T cells from donor PBMCs constituted less than 1% (Supplementary Fig. S1B), with an enriched yield typically reaching 10⁷ cells (Supplementary Fig. S1C). Following the expansion protocol, Vδ1T cells expanded ∼15,000-fold on average (range: 8,400–28,000) across a cohort of six healthy volunteers (Fig. 1B). The final cell products exhibited more than 95% purity for Vδ1T cells, with residual αβT cells or Vδ2T cells below 1% (Fig. 1C). Medium CAR transduction efficiency was 57% ± 24% (mean ± SD; Fig. 1D; Supplementary Fig. S1D). Additionally, these cells expressed natural cytotoxic receptors such as NKG2D, DNAM-1, and the NKp series receptors (NKp30, NKp44, and NKp46), enabling direct recognition of tumor cell ligands (Fig. 1E). Our results demonstrate successful in vitro generation of an ample supply of high-purity Vδ1T or CAR-Vδ1T cells.

Initially, we designed a second-generation CAR construct called AG-CAR (Fig. 2A), comprising an antigen-binding region (B7-H3–scFv), hinge region (IgG4-Fc), transmembrane domain (CD28), costimulatory domain (CD28), and intracellular domain (CD3ζ). Vδ1T cells were transduced with the CAR construct at a multiplicity of infection of 10, resulting in AG-CAR-Vδ1T cells (Fig. 2B). We evaluated the antitumor specificity of these cells by overexpressing B7-H3 in Raji cells (Raji-B7-H3; Supplementary Fig. S2A). AG-CAR-Vδ1T cells exhibited target-specific cytotoxicity against Raji-B7-H3 cells (Fig. 2C), producing more IFNγ and TNFα during coculture (Fig. 2D). Surprisingly, IL2 was not detected in the supernatant. We tested AG-CAR-Vδ1T cells against four B7-H3–positive tumor cell lines (SK-N-AS, A549, SW1990, and HCT-15) representing diverse cancer types (Supplementary Fig. S2B and S2C). In vitro, AG-CAR-Vδ1T cells effectively inhibited the growth of all tested tumor cell lines, dependent on B7-H3 mean fluorescence intensity (Fig. 2E; Supplementary Fig. S2D). However, IL2 was not produced (Fig. 2F). Concerned about antitumor durability, we evaluated CAR-Vδ1T cell persistence by daily stimulation with Raji-B7-H3 cells. Both AG-CAR-Vδ1T cells and untransduced Vδ1T cells showed decreased cytotoxicity (Fig. 2G). To enhance persistence, we explored IL2 supplementation. Although IL2 was absent in coculture media (Fig. 2D and F), adding IL2 improved functional persistence in vitro (Fig. 2H). In vivo, using a subcutaneous neuroblastoma mouse model (SK-N-AS), i.p. IL2 injections did not yield efficacy (Fig. 2I and J), suggesting insufficient stimulation of CAR-Vδ1T cells.

To investigate factors contributing to long-lasting antitumor activity in Vδ1T cells, we designed CAR construct coexpressing IL2, IL21, IL15, and IL10 (Supplementary Fig. S3A). Four types of CAR-Vδ1T cells, each coexpressing a different cytokine, were prepared (Supplementary Fig. S3B), and their cytokine expression was detected (Supplementary Fig. S3C). Notably, the AQ-CAR-IL2 group exhibited stronger inhibition of A549 cell growth (Supplementary Fig. S3D) and superior persistence in killing B7-H3–positive AML-3 tumor cells (Supplementary Fig. S3E and S3F), with higher cytokine release (Supplementary Fig. S3G). During antigen-repeated stimulation, AQ-CAR-IL2 cells consistently expressed NKG2D and DNAM1 (Supplementary Fig. S3H), emphasizing IL2’s crucial role in achieving durable antitumor activity compared with other examined cytokines. Next, we compared the functional persistence of AQ-CAR-IL2 and AG-CAR group (Fig. 3A) and IL2 was only detected in the AQ-CAR-IL2 culture supernatant without antigen stimulation (Fig. 3B). AQ-CAR-IL2-Vδ1T cells demonstrated increased cytotoxicity after repeated stimulation with Raji-B7-H3 cells (Fig. 3C), proliferating gradually in coincubation (Fig. 3D). In vitro, AQ-CAR-IL2 cells produced higher IFNγ, TNFα, and detectable IL2 compared with AG-CAR cells (Fig. 3E). Their potent antitumor effect extended to all tested tumor cell lines (Fig. 3F), accompanied by specific cytokine release (Fig. 3G).

To assess the antitumor efficacy of AQ-CAR-IL2-Vδ1T cells in vivo, we first created the murine xenograft models of neuroblastoma (SK-N-AS), pancreatic cancer (SW1990), and non–small cell lung cancer (A549), which represent the three cancers with little progress in clinical therapies (Fig. 4A). Prior to Vδ1T cell or AQ-CAR-IL2-Vδ1T cell infusion, mice were grouped to ensure comparable tumor sizes (Supplementary Fig. S4A). The results showed that AQ-CAR-IL2-Vδ1T cells effectively inhibited the growth of all three tumor cells in vivo (Fig. 4B–D). At the point of sacrifice, tumors were no longer detectable in five out of six mice in the neuroblastoma model (SK-N-AS), four out of six mice in the pancreatic cancer model (SW1990) and six out of all six mice in the non–small cell lung cancer model (Fig. 4E–G). Thus, in contrast to AG-CAR-Vδ1T cells, which were unable to control tumor growth (SK-N-AS model; Fig. 2J), coexpression of IL2 by AQ-CAR-IL2-Vδ1T cells enabled significant tumor growth control across these three in vivo tumor models.

In addition to creating three subcutaneous tumor models, we also created an orthotopic tumor model using B7-H3-positive breast cancer cell line MDA-MB-231 to further evaluate the antitumor efficacy of AQ-CAR-IL2-Vδ1T (Supplementary Fig. S4B). As an isotype control, CD19-CAR-IL2-Vδ1T cells targeting CD19 were employed (Fig. 5A; Supplementary Fig. S4C). AQ-CAR-IL2 effectively inhibited the growth of MDA-MB-231 cells in vitro (Supplementary Fig. S4D). Remarkably, in vivo results demonstrated complete tumor elimination in mice treated with AQ-CAR-IL2 cells after 11 days (Supplementary Fig. S4E; Fig. 5B and C). Despite detecting CD3⁺ T cells in tumor tissues of five mice at D21 after administration (Fig. 5D) in the CD19-CAR-IL2 group, no significant efficacy was observed. This suggests that B7-H3 antigen-dependent in vivo cytotoxicity by CAR-Vδ1T cells is crucial. Peripheral CD3⁺ T cells were detected at various time points after administration. The AQ-CAR-IL2 group exhibited a higher CAR-positive rate (Fig. 5E). Additionally, infiltration of CAR-Vδ1T cells into organs was observed at the study endpoint (Fig. 5F). Organ tissues were harvested and weighed, revealing no significant difference in relative organ weight between the CAR group and the control group, except for the liver (Fig. 5G). Further histopathologic examination [hematoxylin and eosin (H&E) staining] of the liver treated with PBS and CAR-Vδ1T cells showed no abnormalities (Fig. 5H).

To assess the infiltration ability of AQ-CAR-IL2-Vδ1T cells in tumor and organ tissues, we established a subcutaneous colorectal cancer model (HCT-15). CAR-Vδ1T cell infiltration was evaluated on days 5, 9, and 12 after treatment (Fig. 6A). Results showed varying percentages of CAR-Vδ1T cells infiltrating tumor tissues at different time points, with significantly increased CAR positivity (Fig. 6B). In the remaining mice (n = 10), AQ-CAR-IL2-Vδ1T cells effectively suppressed HCT-15 tumor growth in vivo (Fig. 6C), leading to complete tumor clearance in seven out of 10 mice by D29 (Fig. 6D). At this time point, all mice in the PBS and Vδ1T groups, as well as five mice with tumor clearance in the AQ-CAR-IL2 group, were euthanized (Supplementary Fig. S5A). Among the remaining five mice in the AQ-CAR-IL2 group, two showed tumor clearance, whereas three with tumor burden were re-infused with AQ-CAR-IL2-Vδ1T cells. After 14 days, tumor volume in the re-infused mice remained stable, and two additional mice maintained tumor-free status (Fig. 6E; Supplementary Fig. S5B). Surprisingly, at the endpoint, a high proportion of CD3⁺ T cells and CAR positivity were detected in the tumor tissues of the three mice with tumor burden (Supplementary Fig. S5C and S5D). We speculate that early euthanasia may have obscured efficacy. Additionally, CD3⁺ T-cell infiltration was observed in mouse organs (heart, liver, spleen, lung, kidney, and bone marrow) at different time points (Fig. 6F). Organ weights and histopathologic examination (H&E staining) at D29 revealed no significant differences between the AQ-CAR-IL2 group and the control group (Fig. 6G and H). These findings demonstrate the safety and organ infiltration capability of AQ-CAR-IL2-Vδ1T cells in mice.

In addition to evaluating the antitumor efficacy of AQ-CAR-IL2-Vδ1T cells across various solid tumor models, we established a mouse blood tumor model using the AML-3 cell line (Supplementary Fig. S6A). AQ-CAR-IL2-Vδ1T cells significantly inhibited tumor growth in vivo (Supplementary Fig. S6B), leading to prolonged survival (Supplementary Fig. S6C). CD3⁺ T cells were detected in peripheral blood at different time points (Supplementary Fig. S6D). Importantly, no GvHD occurred during the experiment, and the mice maintained stable body weight (Supplementary Fig. S6E). Additionally, H&E staining revealed no pathologic abnormalities in tissues from mice treated with AQ-CAR-IL2-Vδ1T cells after 63 days of inoculation (Supplementary Fig. S6F).

Comparing AQ-CAR-IL2-Vδ1T cells with other immune cell–derived therapies, including Vδ2T and αβT cell-based approaches, we evaluated their activities in the HCT-15 xenograft tumor model (Fig. 7A; Supplementary Fig. S7A). Although no statistical difference in immune cell infiltration into tumor tissue was observed among the experimental groups, AQ-CAR-IL2-Vδ1T cells showed slightly increased CD3⁺ T cell infiltration at D5 after administration (Fig. 7B; Supplementary Fig. S7B). The discrepancy in tumor-infiltrating CD3⁺T cells may be due to higher CAR transduction efficiency in AQ-CAR-IL2-Vδ1 cells (Fig. 6), leading to accelerated tumor regression. In mouse organs, CAR-Vδ1T cells exhibited significantly higher infiltration than CAR-Vδ2T and CAR-αβT cells (Fig. 7C), highlighting their natural tissue-resident ability. Pharmacokinetic results confirmed greater expansion of AQ-CAR-IL2-Vδ1T cells in peripheral blood (Fig. 7D). Both AQ-CAR-IL2-Vδ1T and AQ-CAR-IL2-αβT groups achieved complete tumor clearance by D17, whereas AQ-CAR-IL2-Vδ2T efficacy was moderate (Fig. 7E and F). After 21 days, remaining mice were sacrificed (Fig. 7G; Supplementary Fig. S7C). AQ-CAR-IL2-Vδ1T cells predominantly exhibited a naïve phenotype in both peripheral blood and organs (Fig. 7H and I), with lower PD-1 expression than other groups (Fig. 7J). No GvHD disease occurred, and mice maintained stable body weights (Fig. 7K). Enlarged spleens were observed in the AQ-CAR-IL2-αβT group (Supplementary Fig. S7D and S7E), but H&E staining showed no abnormalities (Supplementary Fig. S7F). Cytokine levels in peripheral blood were low across all groups at the experiment’s endpoint (Supplementary Fig. S7G).

To investigate the molecular mechanisms underlying the superior efficacy of IL2-expressing AQ-CAR-IL2-Vδ1T cells compared with AG-CAR-Vδ1T cells, we performed single-cell RNA sequencing on cells after repeated antigen stimulation with Raji-B7-H3 cells in vitro for four consecutive days. Dimensionality reduction analysis revealed five clusters enriched for marker genes related to cytotoxicity, chemotaxis, cell proliferation, and CD8-like Teff cells (Supplementary Fig. S8A). These clusters were named Cytotoxic-gdT, CCR-gdT, MKi67-gdT, MCM-gdT, and CD8 Teff (Fig. 8A). Our focus was on genes relevant to effective cancer immunosurveillance by T cells. The results demonstrated that IL2 expression in AQ-CAR-IL2-Vδ1T cells significantly upregulated genes associated with cytotoxicity, T-helper 1 function, tissue homing, and proliferation compared with AG-CAR-Vδ1T cells (Fig. 8B; Supplementary Fig. S8B). Further validation using RT-qPCR confirmed elevated mRNA levels of key antitumor genes (TNF, IFNG, PRF1, MKI67, CCR5, CXCR6, and IL2) in the AQ-CAR-IL2 group, consistent with the single-cell sequencing data (Fig. 8C). Downstream signaling pathway genes related to IL2 production were also significantly upregulated in AQ-CAR-IL2-Vδ1T cells. Pathways related to T-cell proliferation, activation, killing, chemotaxis, and cytokine production were all upregulated in AQ-CAR-IL2-Vδ1T cells compared with AG-CAR-Vδ1T cells (Fig. 8D). These findings collectively indicate that coexpression of IL2 enhances the effectiveness of AQ-CAR-IL2-Vδ1T cells by modulating genes and signaling pathways associated with cytotoxicity, effector function, tissue homing, and proliferation.

Up to now most preclinical studies and clinical trials involving γδT cells for cancer immunotherapy have utilized unmodified γδT cells, specifically the Vδ2T subset. Unfortunately, Vδ2T cells have shown disappointing clinical efficacy in treating solid tumors (23). In contrast, Vδ1T cells have demonstrated a beneficial role in cancer immunosurveillance, correlating with improved patient outcomes with solid tumors (22–25). Like Vδ2T cells, Vδ1T cells are MHC-unrestricted, reducing the risk of GvHD upon adoptive transfer compared with αβT cells. As a potential “off-the-shelf” universal allogeneic therapy, Vδ1T cells hold promise. However, tumors can still develop despite Vδ1T cells presence. To enhance the immunosurveillance capacity of adoptively transferred Vδ1T cells, we engineered them with a B7-H3–targeting CAR construct. In vitro and in vivo testing using cancer cell lines and murine xenograft models confirmed the efficacy of these modified cells. B7-H3, frequently overexpressed across various cancer cells, complements the universal nature of Vδ1T adoptive cell therapy.

Expanding Vδ1T cells for adoptive cell therapy has historically posed challenges. Among current methods (16, 21, 27–29, 37, 38), Adiect Bio’s strategy has been applied in clinical trials. Our study demonstrates that AQ-CAR-IL2-Vδ1T cells are comparable to Adiect Bio’s CAR-Vδ1T cells in terms of purity, expansion, CAR transfection efficiency, and in vivo/in vitro functions. Although most CAR-T cell products use αβT cells, the optimal CAR construct for Vδ1T cells remains unclear. We optimized a clinical-grade expansion protocol to achieve necessary cell numbers while maintaining purity. Additionally, our AG-CAR-Vδ1T cells exhibited potent in vitro activity against cancer cell lines. By incorporating an IL2 encoding domain (AQ-CAR-IL2), we further enhanced B7-H3 targeting CAR-Vδ1T cell efficacy, particularly in functional persistence and in vivo performance.

In vivo studies demonstrated that AQ-CAR-IL2-Vδ1T cells coexpressing IL2 exhibit robust antitumor efficacy, superior infiltration capabilities, and excellent safety across various solid tumor mouse models. Comparative results revealed that the AQ-CAR-IL2-Vδ1T group outperformed the AQ-CAR-IL2-Vδ2T group and was comparable to the AQ-CAR-IL2-αβT group. Additionally, AQ-CAR-IL2-Vδ1T cells displayed greater infiltration ability and a more youthful phenotype. Notably, CAR-Vδ1T cells exhibited a favorable safety profile without adverse events throughout the experiment. These findings highlight the potential of allogeneic “off-the-shelf” B7-H3-CAR-Vδ1T cells for treating a broad spectrum of solid tumors. However, control groups (Vδ1T or CD19-CAR-IL2-Vδ1T) showed no in vivo efficacy, emphasizing the importance of antigen-specific CAR stimulation. Variability in mice may impact tumor clearance, warranting further investigation into factors influencing CAR-Vδ1T cell efficacy.

Since its approval by the FDA in 1992 for cancer immunotherapy in metastatic renal cell carcinoma, high-dose IL2 has been widely used in the treatment of various types of cancers (39). Currently, IL2 is also often combined with TCRT, tumor-infiltrating lymphocytes, and Vδ2T cells in clinical trials, and the results have shown that this combination provides further enhancement of clinical efficacy (40–42). However, IL2 has been found to be associated with side effects, particularly at high doses, which may cause capillary leakage syndrome in some patients (43–44). In this study, we observed that Vδ1T cells, or CAR-Vδ1T cells, have no detectable IL2 production during their antitumor responses, regardless of whether they were stimulated by tumor antigens. Therefore, we were intrigued by the possibility of coexpressing IL2 in the CAR construct may improve the functional persistence of CAR-Vδ1T cells. Excitingly, the in vitro and in vivo data confirmed that coexpression of IL2 plays a crucial role in the functionality of CAR-Vδ1T cells. However, due to potential safety concerns associated with high doses of IL2 release, we conducted additional studies to assess the level of IL2 release. During in vitro experiments, we compared the IL2 release of AQ-CAR-IL2-Vδ1T and AQ-CAR-IL2-αβT cells (Supplementary Fig. S9A). These results suggest that the level of IL2 released by AQ-CAR-IL2-Vδ1T cells was significantly lower than that of the AQ-CAR-IL2-αβT group when repeatedly stimulated with tumor antigen (Supplementary Fig. S9B). Moreover, we found that the absolute value of IL2 produced by varying numbers of AQ-CAR-IL2-Vδ1T cells varied after coincubation with different tumor cells (Supplementary Fig. S9C). To address the safety concerns about the release of IL2 from AQ-CAR-IL2-Vδ1T cells during future clinical trials, we assumed that AQ-CAR-IL2-Vδ1T cells would expand to the same level as autologous CAR-αβT cells. Based on this assumption, it can be estimated that the maximum amount of IL2 released by AQ-CAR-IL2-Vδ1T cells in the human body is ∼6.5 × 10⁵ IU. This is significantly lower than the current clinical dosage of IL2 (8.8 × 10⁶ IU), indicating the favorable safety profile of AQ-CAR-IL2-Vδ1T cells.

In summary, a novel protocol has been developed to generate CAR-Vδ1T cells from peripheral blood. These cells possess favorable anticancer traits and have the potential for “off-the-shelf” use in targeting a broad spectrum of solid tumors.

No disclosures were reported.

L. Jiang: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. F. You: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. H. Wu: Resources, data curation, validation, investigation, methodology. C. Qi: Conceptualization, supervision, project administration. S. Xiang: Resources, validation, methodology. P. Zhang: Data curation, validation, methodology. H. Meng: Conceptualization, project administration. M. Wang: Conceptualization, project administration. J. Huang: Data curation, validation. Y. Li: Validation, methodology. D. Chen: Validation, methodology. G. An: Supervision, project administration. N. Yang: Project administration, writing–review and editing. B. Zhang: Investigation, project administration. L. Shen: Supervision. L. Yang: Conceptualization, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing.

This work was supported by the National Key R&D Program of China (2022YFC2502700), the Natural Science Foundation of China (grant no. 81872431), Priority Academic Program Development of Jiangsu Higher Education Institutions, the Collaborative Innovation Major Project (grant no. XYXT- 2015304), the Project of State Key Laboratory of Radiation Medicine and Protection, Soochow University (no. GNZ1201803), Major Science and Technology Project of Anhui Province (202203a07020030), Key Research and Development Program of Anhui Province (2023s07020010); the manufacturing industry will prioritize using industry–university–research collaborations to overcome weaknesses and achieve key technological goals in 2022 (JB22107); the National Key R&D Program of China (2023YFC3403700), and the key technical research projects of cell therapy (NCTIB2023XB01003).