Herpes Virus Entry Mediator Costimulation Signaling Enhances CAR T-cell Efficacy Against Solid Tumors Through Metabolic Reprogramming

Adoptive immunotherapy based on chimeric antigen receptor-engineered T (CAR T) cells has demonstrated immense potential for the treatment of hematologic malignancies (1). However, CAR T-cell therapy for solid tumors has met great challenges due to the complexity of the tumor microenvironment (TME). The second-generation CAR is composed of a specific single-chain variable fragment (scFv), a hinge and transmembrane domain, a costimulatory domain (CSD), and the intracellular domain of CD3ζ. The CSD is a key element of a second-generation CAR, with the most widely used CSD derived from 4-1BB or CD28 (2, 3). Previous reports have shown that CAR T cells containing the CD28 CSD (CD28-CAR T) show strong cytotoxicity initially, but poor functional persistence and survival in vivo. Conversely, the initial cytotoxicity of CAR T cells containing the 4-1BB CSD (4-1BB-CAR T) was relatively weak, but they persisted longer in vivo than CD28-CAR T (2). Both CD28-CAR T and 4-1BB-CAR T cells have shown encouraging efficacy in the treatment of hematologic malignancies. However, their efficacy against solid tumors is very limited (4). To improve therapeutic efficacy for solid tumors, CAR T cells containing combined CSDs of 4-1BB and CD28 were tested, but with no significant improvement on treatment efficacy for solid tumors (5–8). Therefore, new and improved costimulatory molecules for the preparation of CAR T cells are important for future solid tumor treatment.

The herpes virus entry mediator (HVEM or CD270) is a costimulatory molecule belonging to the TNF receptor superfamily (9), and was first demonstrated as the cellular entry receptor for HSV-1 (10). It is reported that HVEM can trigger a novel signaling pathway, leading to activation of STAT3 (11). HVEM deficiency reduces CD8⁺ T-cell survival and memory T formation (12), while anti-HVEM antibody promotes T-cell proliferation and memory differentiation (13). In addition, activation of HVEM efficiently activates T cells to clear persistent adeno-associated virus (AAV) in the liver (14). HVEM is also known as a molecular switch, and LIGHT (TNFSF14) and B and T lymphocyte attenuator (BTLA) are classical ligands of HVEM. Initiation of LIGHT-HVEM costimulatory signaling promotes CTL-mediated tumor rejection, allograft rejection, and GVHD (15), whereas BTLA-HVEM coinhibtitory signaling represses activation signal transduction, leading to immunosuppression (16).

In this study, we generated antitumor CAR T cells based on CSD of CD28, 4-1BB, or HVEM. In response to tumor target cells, HVEM-CAR T cells displayed stronger cytokine release and cytotoxicity in vitro. HVEM-CAR T cells additionally showed better antisolid tumor efficacy in vivo than CD28-CAR T or 4-1BB-CAR T cells. HVEM-CAR T cells maintained stable functions in solid tumors, with elevation in both oxidative phosphorylation and glycolysis metabolism. Our findings provide a new CAR T-cell design, based on the CSD of HVEM, for the treatment of solid tumors.

FBS was purchased from Biological Industry (04-010-1A). DMEM (12100046) and RPMI1640 (31800022) were purchased from Invitrogen. X-VIVO 15 medium was purchased from Lonza (04-418Q). Penicillin/streptomycin solution was purchased from Bio-Channel (BC-CE-007). Protease inhibitor cocktail kit was purchased from MCE (HY-K0010). Triton X-100, glycerol, deoxycholate, sodium dodecyl sulfonate (SDS), ethylene diamine tetraacetic acid (EDTA), ethylene diamine tetraacetic acid (EGTA), and other reagents were from Sigma-Aldrich.

Human renal cancer line ACHN (TCHu199) and human colon cancer cell line HCT116 (TCHu99) were purchased from National Collection of Authenticated Cell Cultures (Shanghai, China). Human renal cancer lines, Ketr-3 and OSRC-2, were purchased from Shanghai Tongwei Biotechnology Co., LTD. Ketr-3 and ACHN cells were maintained in DMEM supplemented with 10% FBS, and 1% penicillin/streptomycin. Human renal cancer line OSRC-2 and human colon cancer cell line HCT116 were cultured with RPMI1640 medium, supplemented with 10% FBS and 1% penicillin/streptomycin. OSRC-2 and HCT116 cancer cells were transfected by lentivirus with expression of luciferase, and puromycin was used to collect the drug resistance cells to generate luciferase overexpression cell lines, OSRC-2^Luc+, and HCT116^Luc+. Cell lines were maintained in RPMI1640 medium, supplemented with 10% FBS and 1% penicillin/streptomycin.

T cells (described below) were cultured in X-VIVO 15 medium supplemented with 10% FBS, 1% penicillin/streptomycin, and 200 IU/mL recombinant human IL2 (rhIL2; 101-02, PrimeGene). Cells were incubated at 37°C in a 5% CO₂ humidified atmosphere. All cell lines were used within 20 passages of thawing. Cell line authentication and Mycoplasma testing were performed by the vendor and not reevaluated throughout the course of the studies.

Human peripheral blood mononuclear cells (PBMC) were collected from healthy donors, separated with a Lymphoprep gradient (07851/07861, StemCell Technologies), and stored in liquid nitrogen for CAR T-cell preparation. PBMCs were thawed and activated by CD3/CD28 Dynabeads (11132D, Gibco) at a ratio of 1:1. 24 hours later, T cells were infected with lentiviral constructs encoding carbonic anhydrase IX (CAIX)–targeting CARs with different CSD (4-1BB, CD28, or HVEM) at a multiplicity of infection of 5, and expanded in X-VIVO 15 medium containing 200 IU/mL rhIL2. Dynabeads were removed 5 days after activation. The lentivirus-infected T cells were continuously expanded for another 5 to 7 days in X-VIVO 15 medium supplemented rhIL2, and were used for experiments in vitro or in vivo experiments. The CSDs are derived from intracellular cytoplasmic domain of these costimulatory molecules: 42 amino acids (from 214 to 255, UniPro ID: Q07011) of 4-1BB, 41 amino acids (from 180 to 220, UniPro ID: P10747) of CD28, and 60 amino acids (from 224 to 283, UniPro ID: Q92956) of HVEM. The sequences of scFv against human CAIX was based on murine mAb G250 (17), and scFv against human epithelial cell adhesion molecule (Ep-CAM) was acquired from publication of Shirasu and colleagues (18).

We established a T-cell exhaustion model in vitro as described in previous reports (19, 20). A total of 1 × 10⁶ CAIX-CAR T cells were cocultured with OSRC-2 cells at effector : target (E:T) ratio of 1:5 in 10 cm plates with X-VIVO 15 medium containing rhIL2. After 3 days, defined as one cycle, the cocultures were thoroughly suspended by frequent pipetting, and roughly 2 × 10⁵ of the cell suspension was used for T-cell counting assessment and flow cytometry staining as specified below. The remaining cell suspension was centrifuged. Cells were resuspended and transferred into fresh target cancer cell–coated plates for continuous coculture. This process was repeated for 24 days (eight cycles).

Cytotoxicity of CAR T cells in vitro was determined using the xCELLigence RTCA instrument (ACEA Biosciences). Background impedance was measured by adding 50 μL of target cell culture media to an E-plate (ACEA Biosciences). Target cancer cells were seeded in the E-plate at 5,000 cells/well and allowed to adhere before CAR T-cell addition at the indicated E:T ratio in duplicate. A target cell only control was included as an indicator of cell proliferation. CAR T cytotoxic activity was determined on the basis of the percentage of attached viable target cells (cell index value).

Human CAR T cells were cocultured with target cancer cells for 24 hours at the indicated E:T ratios. Media from the cocultured wells were collected after centrifugation removal of cell debris. Supernatants were diluted into three to 10 times before detection. Cytokines in supernatants were measured using human ELISA kits (Dakewe), RD-TNFa-Hu for TNFα, HIL2-E96.1 for IL2, CT211A for granzyme B (GRZ-B), and RD-IFNg-Hu for IFNγ, respectively. The detection was performed according to the manufacturer's instructions, and the absorbance was detected by Cytation3 (BioTek). Standard curves were illustrated by standards provided by kits. Concentration of cytokines were calculated by multiply dilution ratios.

CAR T cells were cocultured with OSRC-2 cells at E:T ratio of 5:1 in 10 cm plates with X-VIVO 15 medium containing rhIL2 for the indicated time and stained with PerCP against human CD3 (UCHT1, BioLegend), and then harvested by flow cytometry using FACS Aria III (BD). Isolated CAR T cells were lysed in lysis buffer (50 mmol/L Tris·HCl, pH 7.5, 150 mmol/L NaCl, 1% Triton X-100, 0.5% deoxycholate, 0.1% SDS, 10% glycerol, 1 mmol/L EDTA, 1 mmol/L EGTA, 1 × protease inhibitor cocktail), clarified by centrifugation, and the supernatant collected and stored at −80°C. Protein samples were quantified with a BCA kit (P0012, Beyotime), total 5 to 10 ug protein samples were separated by SDS-PAGE, and transferred to nitrocellulose membranes (GE Healthcare). Membranes were blocked, incubated with primary antibodies overnight at 4°C (dilution rate of 1:2,000), followed by secondary antibodies at room temperature for 1 hour (dilution rate of 1:5,000), and visualized by chemiluminescence device (Tanon 4600). Protein expressions were quantified using the ImageJ software. Information for the antibodies used for Western blotting are provided in Supplementary Table S1.

Six-to-8-weeks-old male NPG (NOD-Prkdc^scidIl2rg^null) mice from Vitalstar Biotechnology were maintained in specific pathogen-free animal facilities of the Experimental Animal Center of Xuzhou Medical University (Xuzhou, P.R. China). All animal experiments were approved by the Animal Ethics Committee of Xuzhou Medical University (Xuzhou, P.R. China). All mice were administered intraperitoneally with IL2 (2,000 IU/mouse) daily during T-cell therapy to maintain CAR T-cells survival in NPG mice.

For orthotopic xenograft tumor model, 5 × 10⁵ OSRC-2^Luc+ cells were injected into the subrenal capsule of NPG mice. Mice were randomly assigned into four groups after 14 days: non-transduced T cells (Ctrl-T), CAIX-CAR T cells with 4-1BB, CD28, or HVEM CSD, and injected with 5 × 10⁶ of the respective CAR T cells via tail vein (5 mice per group). Tumor progression was evaluated by luminescence imaging with an IVIS system every week. CAR T cells in blood from caudal vein of NPG mice were analyzed on day 28 by flow cytometry. Overall survival of NPG mice was documented. The experiment was terminated on day 70, and rencal tissues were collected after mouse death.

For metastasis tumor models, 2 × 10⁶ OSRC-2^Luc+ or 1 × 10⁶ HCT116^Luc+ cells were infused into NPG mice via tail vein, respectively. Mice were randomly assigned into four groups when luminescence of tumor could be observed by IVIS system: Ctrl-T cells, CAIX-CAR T cells, or Ep-CAM-CAR T cells with 4-1BB, CD28, or HVEM CSD, and injected with 5 × 10⁶ of the respective CAR T cells via tail vein (8–10 mice per group). Tumor progression was evaluated by luminescence imaging with an IVIS system every 5–7 days. CAR T cells in blood from caudal vein of NPG mice were analyzed on day 14 after CAR T therapy performed by flow cytometry. Overall survival of NPG mice was documented. Lung tissues from metastasis tumor model of OSRC-2 cell were collected and analyzed by hematoxylin and eosin (H&E) staining after mice were sacrificed.

For subcutaneous tumor model, 2 × 10⁶ OSRC-2 cells were injected into the right dorsal flank region of NPG mice. Mice were randomly assigned into three groups on day 14: CAIX-4-1BB, CAIX-CD28, and CAIX-HVEM CAR T cells, and injected with 1 × 10⁷ of the respective CAR T cells via tail vein, 12 mice per group. A total of 4 mice of each group were sacrificed on 1, 2, or 3 weeks after CAR T-cell administration. When effector function and exhaustion of CAR T cells in the tumors were analyzed, tumors were cutted into pieces and added into a digestion solution: 300 U/mL Collagenase I (VIC079, VICMED) and 100 U/mL DNase I (10104159001, Roche). Tumor samples were transferred into gentleMACS C tubes and processed into single-cell suspensions using a Miltenyi gentleMACS dissociator with heaters (37°C). Cell counts and exhaustion of tumor-infiltrating T cells were analyzed by flow cytometry as described previously. Cytotoxicity and cytokine release were analyzed after tumor-infiltrating T cells were sorted by flow cytometry, and metabolic activities of tumor-infiltrating T cells were detected by seahorse (as specified below).

For surface staining, single-cell suspensions prepared from cultured CAR T cells, mice peripheral blood, or tumors harvested from tumor-bearing mice and were stained with antibodies against surface markers, isotype control antibodies were used for gating strategy. Samples were detected on a FACS Canto II cell analyzer (Becton-Dickinson) and analyzed using FlowJo V10. For intratumor CAR T-cell sorting, after staining with viability dye (LIVE/DEAD Fixable Far Red Dead Cell Stain Kit, L34974, Invitrogen) and anti-human CD45 and CD3 antibodies, human CAR T cells were sorted on a BD FACS Aria III cell sorter (Becton-Dickinson) using a 70-mm nozzle and collected into Falcon round-bottom polypropylene tubes (Falcon, USA) containing X-VIVO 15 medium of CAR T cells. For cell counting, 5–10 μL 123count eBeads Counting Beads (Invitrogen) were added into samples during FACS analysis to evaluate T-cell unmbers. The total T-cell number = (counting beads volume × beads density × counted T-cell number ×dilution ratio)/counted beads number. Information for the antibodies used for flow cytometry are provided in Supplementary Table S2.

Total RNA from CAR T cells was extracted using TRIzol reagent (R1100, Solarbio), and the concentrations were measured by Nanodrop. A total of 500 ng RNA samples were used for reverse transcription using PrimeScript RT Master Mix (RR036A, Takara). Gene expression was detected with qRT-PCR using SYBR Green reagent (Q111-02, Vazyme). The detection was performed by LightCycler 96 (Roche). GAPDH was used for normalization, and the relative gene expression was determined using 2^–ΔΔC_T. The experiment was replicated three times. Primer sequences are listed in Supplementary Table S3.

The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were assessed using the Seahorse XFp analyzer (Agilent). Human CAR T cells were sorted from cocultures with OSRC-2 cell after 24 hours in vitro, or from subcutaneous tumor tissues after growth in vivo for 1, 2, or 3 weeks, as described above. Cells were resuspended in 50 μL assay medium (nonbuffered RPMI1640 containing 1 mmol/L sodium pyruvate, 1 mmol/L l-glutamine, and 25 mmol/L glucose; for ECAR, no glucose was added), seeded into XF96 cell culture microplates (Agilent) at 3 × 10⁵ cells/well, cells were incubated for 30 minutes to ensure sedimentation at the bottom of plates, followed by addition with 130 μL assay medium slowly and gently for further 1.5 hours at 37°C without CO₂, and then basal OCR and ECAR were measured for 20 minutes. Cells were subsequently treated with 1.5 μmol/L oligomycin (HY-N6782, MCE), 1 μmol/L carbonyl cyanide p-trifluoromethoxy phenylhydrazone (FCCP; HY-100410, MCE), 40 nmol/L rotenone (HY-B1756, MCE), and 1 μmol/L antimycin A (A8674, Sigma) every 20 minutes to detect the dynamic OCR. Cells were also subsequently treated with 10 mmol/L glucose (1181302, Sigma), 1.5 μmol/L oligomycin, and 50 mmol/L 2-Deoxy-D-glucose (2-DG, HY-13966, MCE) every 20 minutes to measure the glycolytic capacity.

For H&E staining, lung tissues were collected from metastasis tumor model when mice sacrified. The middle lobes of the lung were fixed in 4% paraformaldehyde (VIH130, VICMED), embedded in paraffin (VIH008, VICMED), and sectioned at 5 μm thickness. Slides were stained with H&E. For IHC staining, tumors were collected from the subcutaneous tumor model when mice after CAR-T infusion for 3 weeks. Part of the tumors were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned at 5 μm thickness. Sections were incubated with anti-human CD3 (ab21703, Abcam) or anti-human CAIX (ab15086, Abcam). The immunoreaction was visualized by incubating with diaminobenzidine (HY-15912, MCE). Slides were imaged with an upright light microscope (Olympus Corporation).

Statistical analysis was performed using Graphpad Prism 8.1 software, and data are presented as mean ± SD. Statistical differences between mean values was determined using unpaired Student t test or one-way ANOVA as indicated in figure legends. Survival curves were analyzed by log-rank test. In all cases, a significant result was defined as P < 0.05.

The data generated in this study are available within the article and its Supplementary Data.

CAR T cells targeting the renal cancer antigen, CAIX, were generated to explore the functions of the HVEM CSD in treating solid tumors in comparison with the 4-1BB or CD28 CSDs (Fig. 1A). CAR T cells were generated by lentivirus transduction, and CAR expression was comparable for all constructs (Supplementary Fig. S1A and S1B). The differentiation phenotypes of the three CAR T cells were also similar (Supplementary Fig. S1C and S1D).

The activities of CAIX-CAR T cells were evaluated in vitro with three human renal cancer cell lines OSRC-2 (CAIX-positive), Ketr-3 (CAIX-positive), and ACHN (CAIX-negative) that we have reported previously (21). In response to these tumor cells, ELISA and real-time cellular analysis (RTCA) were used to assess the cytokine release and cytotoxicity, respectively, of the CAR T cells. The ELISA results demonstrated that CAIX-HVEM CAR T cells released higher levels of cytokines, including IL2, IFNγ, TNFα, and GRZ-B, in response to CAIX-positive cancer cells compared with Ctrl-T, CAIX-4-1BB, or CAIX-CD28 CAR T cells at an E:T of 1:5 (Fig. 1B). No significant cytokine release was detected in response to CAIX-negative ACHN cells for all CAR T cells, indicating the specific activation of CAIX-CAR T cells by CAIX-positive cancer cells (Fig. 1B). The RTCA results showed that all CAR T cells had no significant cytotoxicity to CAIX-negative ACHN cells, whereas CAIX-HVEM CAR-T displayed stronger cytotoxicity against CAIX-positive renal cancer cells than CAIX-4-1BB or CAIX-CD28 CAR T cells (Fig. 1C). Furthermore, the superior performance of the HVEM CSD in vitro was also confirmed with a HVEM-CAR T construct targeting the Ep-CAM in the human colon cancer cell line HCT116 (Supplementary Fig. S2). These results demonstrate that the HVEM CSD can enhance cytokine release and cytotoxicity of CAR T cells in vitro.

Orthotopic tumor models have been widely used to simulate the biological features of solid tumor growth and metastasis in patients, and may provide better prediction of potential clinical activity for antitumor therapy (22). To assess the therapeutic efficacy of HVEM-CAR T against orthotopic tumors, NPG mice were injected with 5 × 10⁵ OSRC-2^Luc+ cells under the kidney capsule and treated with 5 × 10⁶ CAR T cells on days 14 (Fig. 2A). In vivo imaging of mice indicate that CAIX-HVEM CAR T cells exhibited more potent antitumor efficacy compared with CAIX-4-1BB or CAIX-CD28 CAR T cells. Furthermore, CAIX-HVEM CAR-T eliminated tumors in three mice on day 67 (Fig. 2B and F). 14 days after CAR-T infusion, CAIX-HVEM CAR-T showed the strongest proliferation in blood of tumor-bearing mice (Fig. 2C and D). The overall survival curves showed that CAIX-HVEM CAR T cells significantly rescued tumor-bearing mice. All mice treated with CAIX-HVEM CAR T cells were still alive at end of the experiment, whereas the mice treated with CAIX-4-1BB or CAIX-CD28 CAR-T showed very limited efficacy in comparison with Ctrl-T treated mice (Fig. 2E). Consistent results were achieved in another repeated experiment (Supplementary Fig. S3). Together, these data demonstrate that CAR T cells containing the HVEM CSD display superior therapeutic efficacy against human renal cancer in an orthotopic xenograft model.

To test the therapeutic efficacy of HVEM-CAR T against solid tumors, we established a pulmonary metastasis model with OSRC-2 cells constitutively expressing luciferase (OSRC-2^Luc+) in NPG mice and performed the treatments as indicated in Fig. 3A. In vivo imaging showed that CAIX-HVEM CAR T cells exhibited superior antitumor efficacy (Fig. 3B) and significantly increased the overall survival of treated mice (Fig. 3C). Gross pathology analyses of lung specimens indicated that tumor burdens were significantly reduced in mice that received CAIX-HVEM CAR-T treatment compared with mice that received Ctrl-T, CAIX-CD28, or CAIX-4-1BB CAR-T (Fig. 3D, bottom right corner). H&E staining of lung tissues indicated that normal pulmonary alveoli were still observed in mice with CAIX-HVEM CAR-T treatment, whereas the lung tissues of other treatment groups were almost completely occupied by metastatic lesions (Fig. 3D). A total of 14 days after CAR-T infusion, CAR T cells in blood were detected by FACS, and CAIX-HVEM CAR T showed the strongest proliferation in blood of tumor-bearing mice (Fig. 3E and F). In another repeated experiment, consistent results were achieved. Furthermore, mice in other groups were terminated at about 40 days, whereas the mice in CAIX-HVEM group were terminated at about 90 days after tumor injection (Supplementary Fig. S4). These data suggest that CAIX-HVEM CAR T cells have better therapeutic efficacy in human renal cancer models with pulmonary metastasis.

To further confirm the therapeutic efficacy of HVEM-CAR T against metastatic tumors, we established a metastatic model of human colorectal cancer with the Ep-CAM–positive cell line HCT116^Luc+, and treated tumor-bearing mice with Ep-CAM-CAR T cells with CD28, 4-1BB, or HVEM CSD (Supplementary Fig. S5A). The results of in vivo imaging showed that Ep-CAM-HVEM CAR T cells more significantly inhibited growth of metastatic tumors than Ep-CAM-4-1BB or Ep-CAM-CD28 CAR T cells (Supplementary Fig. S5B). CAR T-cell detection in blood showed that Ep-CAM-HVEM CAR T cells had the highest proliferation in tumor-bearing mice (Supplementary Fig. S5C and S5D). Furthermore, Ep-CAM-HVEM CAR T cells significantly prolonged the survival of tumor-bearing mice compared with Ep-CAM-4-1BB or Ep-CAM-CD28 CAR T cells (Fig. 3G). These data further indicate that HVEM-CAR T cells exhibit superior therapeutic efficacy against human solid tumors in metastatic tumor models.

The immunosuppressive TME is one of the main reasons for the limited efficacy of CAR T-cell therapy for solid tumors, thus maintaining effector function in tumor tissue is important for CAR-T therapy in solid tumors. To explore the effector functions of HVEM-CAR T cells in tumor tissues, we established a subcutaneous tumor model with 2 × 10⁶ OSRC-2 cells. Tumors were allowed to grow for 2 weeks before CAR-T treatment. The infused CAIX-CAR T cells in tumors were isolated 1, 2, or 3 weeks after treatment, and their cytotoxicity against OSRC-2 cells were evaluated by RTCA ex vivo. All intratumor CAR T cells showed minimal cytotoxicity against target cells at E:T of 1:5 (Fig. 4A), a ratio that all fresh CAR T cells were able to efficiently kill target tumor cells (Fig. 1C). These data suggest that the cytotoxicity of the CAR T cells were attenuated in solid tumor tissues. However, when the E:T ratio was increased to 1:1 or 5:1, CAIX-HVEM CAR T cells sorted from tumors on both weeks 1 and 2 after CAR-T infusion were able to kill target cells more efficiently than CAIX-CD28 or CAIX-4-1BB CAR T cells. The cytotoxicity of the three kinds of CAR T cells were further reduced after being in the tumor for 3 weeks. However, at an E:T of 25:1, CAIX-4-1BB and CAIX-HVEM CAR-T showed stronger cytotoxicity than CAIX-CD28 CAR T cells (Fig. 4A). These data demonstrate that tumor-infiltrating CAIX-HVEM CAR T cells exert more potent cytotoxicity than CAIX-CD28 or CAIX-4-1BB CAR T cells.

To further confirm the effector function of tumor-infiltrating HVEM-CAR T cells, we detected cytokine concentrations (IFNγ, GRZ-B, TNFα) in the cell culture supernant of the RTCA experiments described above by ELISA (E:T = 5:1 for week 1 and week 2, E:T = 25:1 for week 3). As shown in Fig. 4B, on week 1, CAIX-HVEM CAR-T released highest level of cytokines, followed by CAIX-CD28 and CAIX-4-1BB CAR-T orderly; on week 2 and week 3, CAIX-HVEM CAR-T still released the highest level of cytokines. However, the cytokine levels released by CAIX-4-1BB CAR-T were higher than that released by CAIX-CD28 CAR T cells. Together, these results suggest that HVEM-CAR T cells could maintain superior effector function in solid tumors.

To investigate the mechanism for the superior effector function of HVEM-CAR T, we detected the differentiation of tumor-infiltrated CAR T cells after CAR T cells infusion for 1, 2, or 3 weeks by FACS. The results revealed no significant difference in the differentiation state among the three CAR T cells (Supplementary Fig. S6). Therefore, HVEM CSD enhances CAR T effect function not through regulating cell differentiation. Continuous antigen stimulation, hypoxia, and hypoglycemia in the TME can cause T-cell exhaustion (23, 24), and limit their efficacy in the treatment of solid tumors (25). We therefore investigated whether the HVEM costimulatory signal attenuated exhaustion of CAR T cells. The expression of T-cell exhaustion markers, including PD-1, TIM-3, and TIGIT, in tumor-infiltrating CAIX-CAR T cells were analyzed by FACS after CAR-T infusion for 1, 2, or 3 weeks. As shown in Fig. 4C, the PD-1, TIM-3, and TIGIT-positive frequency in CAIX-HVEM CAR T cells were the lowest among the three CAR T cells, mostly followed by CAIX-4-1BB. CAIX-CD28 CAR T cells displayed almost the highest T-cell exhaustion marker expression at every timepoint. These data suggest that HVEM could attenuate exhaustion of CAR T cells in solid tumors.

Expression of target antigen, tumor-infiltrating and proliferative capacity of CAR T cells in tumor tissues determine the therapeutic efficacy of CAR T cells (26). We detected human CAIX antigen expression in tumor tissues after CAR-T treatment for 3 weeks by IHC. The results showed no significant difference in human CAIX antigen expression among the tumor tissues treated by CAIX-HVEM, CAIX-4-1BB, or CAIX-CD28 CAR T cells (Supplementary Fig. S7). Therefore, the difference in the antigen expression was not the reason for the difference in therapeutic efficacy of CAR T cells. To examine the abundance of CAR T cells within tumor tissues, tumor-infiltrating CAR T cells were counted during FACS analysis with counting beads. The results revealed that CAIX-CD28 CAR T cells had the highest abundance in tumor tissues among the three CAR T cells at week 1, whereas the CAIX-HVEM CAR T cells surpassed CAIX-CD28 CAR T cells and maintained highest abundance among the three CAR T cells at week 2 and week 3 (Fig. 4D). We also detected tumor-infiltrating CAR T cells at week 3 by IHC. The results also showed that the abundance of CAIX-HVEM CAR T cells in tumor tissues was much higher than that of CAIX-4-1BB or CAIX-CD28 CAR T cells (Fig. 4E and F). These data demonstrate that the higher tumor-infiltrating and/or proliferative capacity of CAIX-HVEM CAR T cells in tumor tissues could be another reason for its superior therapeutic efficacy compared with CAIX-4-1BB or CAIX-CD28 CAR T cells.

To further investigate whether HVEM-CAR T cells could maintain superior effector function through attenuating exhaustion of CAR T cells, we established a CAR T-cell exhaustion model in vitro. CAR T cells were subjected to serial encounter with new OSRC-2 target cells every 3 days, and the exhaustion state of CAR T cells was determined by PD-1 and LAG-3 expression during target cancer cell stimulations (Fig. 5A). As shown in Fig. 5B, CAIX-HVEM CAR T cells showed lower expression of PD-1 and LAG3 than that in CAIX-4-1BB or CAIX-CD28 CAR T cells. T cell numbers also were counted during FACS analysis with counting beads, and the results revealed that CAIX-CD28 CAR T-cell displayed the most intense proliferation in response to cancer cell stimulation in cycle 1, whereas it declined quickly in cycles 3, 5, and 8. CAIX-HVEM CAR T cells showed a moderate level of cell proliferation in cycle 1, whereas they maintained the highest proliferation among the three CAR T cells in cycles 3, 5, and 8 (Fig. 5C). To examine cytotoxicty of the exhausted CAR T cells with different CSDs, RTCA was performed at an E:T of 5:1 or 1:1. The results indicated CAIX-HVEM CAR T cells displayed the strongest cytotoxicty on cancer cells (Fig. 5D). Futhermore, ELISA analysis of cell culture supernatants from the RTCA experiments revealed that CAIX-HVEM CAR T cells released the most IFNγ, GRZ-B, and TNFα among the three CAR T cells (Fig. 5E). Together, these data further demonstrate that HVEM-CAR T cells could maintain superior effector function through attenuating exhaustion of CAR T cells compared with 4-1BB-CAR T or CD28-CAR T cells.

Metabolic activity critically determines the effector function of T cells (27), and activated effector T cells increase oxidative phosphorylation and aerobic glycolysis to meet the energy demands of cell proliferation and cytokines secretion (28). Because CAR-T with different CSDs displayed distinct effector functions in vitro and in vivo as described above and because metabolic activity of CAR T cells impact antitumor efficacy (29), we sought to detect the metabolic activity of these CAR T cells in tumors. Intratumoral CAIX-CAR T cells were isolated from OSRC-2 tumors 1, 2, or 3 weeks after CAR-T infusion as described above. As shown in Fig. 6, on week 1, the basal OCR, maximal OCR, spare respiration capacity (SRC = maximal OCR − basal OCR), and the ECAR of CAIX-CD28 CAR-T and CAIX-HVEM CAR T cells were higher than that of CAIX-4-1BB CAR T cells; on week 2, basal OCR was similar among all CAR T cells, whereas the maximal OCR and SRC in the CAIX-4-1BB and CAIX-HVEM CAR T cells were significantly higher than those in CAIX-CD28 CAR T cells; on week 3, all metabolic indicators of CAIX-HVEM CAR T cells were the highest among the three CAR T cells. These data indicate that HVEM-CAR T cells maintain their metabolic activity in the tumor tissue.

To further confirm the metabolic patterns regulated by the HVEM CSD in CAR T cells, we detected the metabolic activity of the three CAIX-directed CAR T cells cocultured with OSRC-2 cells for 24 hours, respectively. As shown in Fig. 7A and B, the basal OCR, maximal OCR, SRC, and ECAR of CAIX-HVEM CAR T cells detected by Seahorse were higher compared with that of CAIX-CD28 CAR-T or CAIX-4-1BB CAR T cells. Consistent results were achieved in Ep-CAM-CAR T cells with CD28, 4-1BB, or HVEM CSDs (Fig. 7C and D). These data suggest that the enhanced effector function of HVEM CAR T cells is accompanied by higher metabolic activity.

To gain insights into the mechanisms involved in metabolic reprogramming by distinct costimulatory molecules, we evaluated the signaling pathways involved in glucose metabolism. AKT activation promotes the expression of GLUT1, the main transporter that promotes glucose uptake into T cells (30, 31). Solute carrier family 16 (SLC16A3) is a transporter of glycolysis byproducts, including lactic acid and pyruvate. Thus, AKT, GLUT1, and SLC16A3 are critical positive regulators of glucose metabolism within T cells. The expression of the three regulatory proteins was evaluated in CAR T cells. CAIX-HVEM, CAIX-CD28, or CAIX-4-1BB CAR T cells were cocultured with OSRC-2 cells for 5 minutes. Western blot analysis showed that phosphorylated AKT was higher in HVEM-CAR T cells than CD28-CAR T or 4-1BB-CAR T cells (Fig. 7E and F). The mRNA expression of GLUT1 and SLC16A3 were also significantly higher in HVEM-CAR T cells than CD28-CAR T or 4-1BB-CAR T cells (Fig. 7G). These results are consistent with the in vitro metabolic activities of the three CAR T cells, and demonstrate that HVEM costimulatory signals enhance glucose through an AKT-associated signal pathway.

Addition of 4-1BB or CD28 costimulatory domains to first-generation CARs produced second-generation CARs with significantly improved functions against hematologic malignancies (32), including enhanced proliferation (33) and increased CAR T-cell activation (3). A combination of both 4-1BB and CD28 in third-generation CARs led to further increase in CAR T-cell activation and persistence (5, 6). However, other reports did not show significant differences between second- and third-generation CAR T cells (7, 8, 34). Therefore, all current clinically approved CAR T products, Kymriah, Yescarta, Tecartus, Breyanzi, and Abecma, use a single CSD of either 4-1BB or CD28 (35–40). However, CAR T cells containing the 4-1BB or CD28 CSD have no significant efficacy in the treatment of solid tumors (41).

Strategies utilizing costimulatory molecules to improve the efficacy of CAR T-cell therapy in solid tumors have been intensively investigated (42). Here, we demonstrated that CARs engineered with the novel HVEM costimulatory molecule showed significant improvement for the treatment of solid tumors, including advanced metastases and orthotopic tumors. Using CAR constructs targeting the renal cancer antigen CAIX and the colon cancer antigen Ep-CAM, we demonstrated that HVEM-CAR T cells had stronger activation and killing ability than CD28-CAR T or 4-1BB-CAR T cells in vitro and sustained proliferation and survival in vivo. This indicates that the HVEM CSD has the advantages over either the CD28 or 4-1BB CSD. In addition, HVEM-CAR T cells could attenuate exhaustion, maintain effector functions and metabolic activity stably in solid tumors, which likely contributes to their remarkable therapeutic effect on solid tumors. In addition, the number of CAR T cells in the tumor tissue is an important factor determining their antitumor effect. We found that there was significantly increased abundance of CAR T cells in tumor tissues treated with CAIX-HVEM CAR T cells compared with that treated with CAIX-CD28 or CAIX-4-1BB CAR T cells. However, whether the increased abundance of CAIX-HVEM CAR T cells was caused by the higher tumor-infiltrating ability or proliferation ability in tumor tissues compared with CAIX-CD28 or CAIX-4-1BB CAR-T needs further study.

Compared with 4-1BB or CD28, the HVEM costimulatory signal enabled CAR T cells to maintain the higher oxidative phosphorylation and respiration capacity, which may account for HVEM-CAR T cells being able to maintain prolonged effector functions in tumors in vivo. Metabolic reprogramming during T-cell activation is regulated by costimulatory molecules through their downstream signaling pathways (30). CD28 is a transmembrane protein of the immunoglobulin superfamily, and its intracellular domain contains 41 amino acids (from 180 to 220), including multiple active sites such as YMNM and PYAP (43). It can bind to the P85 regulatory subunit of PI3K (44), growth factor receptor–binding protein 2 (GRB2), and lymphocyte-specific tyrosine protein kinase (45) to activate downstream PI3K/AKT, PKCθ, and other kinase signaling pathways (45, 46). The PI3K/AKT pathway enhances glycolysis to achieve rapid proliferation and cytokines secretion through promotion of glucose uptake by upregulating the expression of GLUT1 (31, 47). CD28 promotes expansion of CAR T cells but induces rapid exhaustion (48). 4-1BB belongs to TNF receptor superfamily, also known as CD137, whose intracellular domain contains 42 amino acids (from 214 to 255), which can recruit TNF receptor–associated factors (TRAF), including TRAF1, TRAF2, and TRAF3. TRAF can activate MAPK and NFκB (49), and p38-MAPK upregulates PGC-1α, resulting in mitochondrial fusion and biogenesis, which substantially increases T-cell oxidative phosphorylation. This outcome increases long-term survival of T cells (50). These reports indicate glycolysis plays a critical role in T-cell effector function, whereas oxidative phosphorylation enhances T-cell survival. HVEM, known as CD270, also belongs to TNF receptor superfamily, and its intracellular domain contains 60 amino acids (from 224 to 283). Herein, our results revealed that both glycolysis and oxidative phosphorylation were elevated in HVEM-based CAR T cells compared with 4-1BB and CD28-CAR T cells, suggesting the excellent metabolic patterns result in the surperior antitumor efficacy and long-term surivival of HVEM-CAR T cells. Studies have reported that HVEM could recruit TRAF2 and TRAF3 to activate the downstream MAPK pathway (51). Meanwhile, our results also illustrated that HVEM could activate the PI3K/AKT pathway, indicating the downstream signal of HVEM CSD promoted the increase of glycolysis and oxidative phosphorylation. These results together reveal the mechanisms involved in the prolonged persistence of functional HVEM-CAR T cells in tumors.

In summary, we developed HVEM costimulatory molecule-based CAR constructs that improve second-generation CAR T cells to treat solid tumors. Compared with CD28 or 4-1BB, HVEM-based CAR T cells show better in vivo effector functions and attenuated exhaustion associated with improved therapeutic activity of solid tumors and metatatasis. These findings will have significant implication for treatment of solid tumors with CAR T cells.

L. Su reports a patent for UNC Ref. 18-0109 licensed to inceptorBio/FastBack Bio; and Inceptor Bio has listed me as a scientific cofounder of FastBack Bio. No disclosures were reported by the other authors.

S. Sun: Resources, data curation, funding acquisition, validation, investigation, writing–original draft. C. Huang: Validation, investigation. M. Lu: Validation, investigation. H. Xu: Software. Y. Yuan: Formal analysis. W. Zhao: Formal analysis. X. Hu: Formal analysis. B. Wang: Investigation. W. Zhang: Investigation. X. Gao: Methodology. J. Zheng: Conceptualization, resources, funding acquisition, project administration, writing–review and editing. L. Su: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing. Q. Zhang: Conceptualization, resources, supervision, funding acquisition, visualization, project administration, writing–review and editing.

This research was supported National Natural Science Foundation of China grant 81773253 (Q. Zhang), 81972242 (X. Gao); Natural Science Foundation of Jiangsu Province grant BK20211057 (Q. Zjang); Natural Science Project of Jiangsu Provincial Education Department grant 19KJB310018 (S. Sun); Research Foundation of Xuzhou Medical University grant D2019023 (S. Sun); Xuzhou Science and Technology Bureau projects grant KC19058 (S. Sun); Youth Science and Technology Innovation team of Xuzhou Medical University grant TD202003 (Q. Zhang); Natural Science Key Project of Jiangsu Provincial Education Department 20KJA320006 (X. Gao); Jiangsu Provincial Key Medical Discipline, The Project of Invigorating Health Care through Science, Technology and Education grant NO.ZDXKA2016014 (J. Zheng); UNC LCCC cancer research fund (L. Su); Qing Lan Project of Jiangsu Province (X. Gao).

The authors thank Qianqian Dai, Cancer Institute, Xuzhou Medical University, for helping with flow cytometry operation and data acquisition.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.