A Costimulatory CAR Improves TCR-based Cancer Immunotherapy

Adoptive cell therapy (ACT) with T cells expressing either native or transgenic T-cell receptors (TCR) is an attractive immunotherapeutic approach because of the large diversity of antigens that can be targeted (1). Ex vivo–expanded, polyclonal T-cell therapies, in which T cells express native (n)TCRs, can be specific for multiple epitopes from multiple tumor antigens, overcoming the problem of tumor heterogeneity. Epstein–Barr virus (EBV)-specific T cells (EBVST) have proven successful against EBV-associated malignancies, as have tumor-infiltrating lymphocytes (TIL) for melanoma and other solid tumors (2–6). Transgenic TCR-based (tTCR) T-cell therapies can efficiently redirect polyclonal T cells to a defined peptide–HLA complex with high reproducibility and defined affinity. These tTCR⁺ T cells have achieved promising clinical results against several tumor antigens expressed in melanoma, myeloma, synovial sarcoma, and acute myeloid leukemia (7–12). However, limited persistence and immune suppression in the tumor microenvironment (TME) are major causes for treatment failure with both treatments (11).

Clinical responses and T-cell persistence are correlated in both nTCR and tTCR therapies (2, 9). Unlike chimeric antigen receptor (CAR) T cells, which target cell surface antigens and are engineered with built-in costimulation, TCRs target immunogenic peptides presented on the cell surface in the context of MHC molecules. Therefore, TCRs rely on naturally occurring, host derived costimulation and cytokine signals (12). In the immunosuppressive TME, many factors impair tumor-specific T-cell function and persistence (11). Most tumors lack costimulatory ligands and inhibit professional antigen-presenting cells that normally present antigen in an immunostimulatory context.

For this study, we engineered CD19-targeting, costimulatory CARs (CoCAR) that lacked the cytotoxic ζ chain of conventional CARs, but still provide T-cell costimulation upon CD19 ligation in the TME of hematologic malignancies (i.e., bone marrow, lymphatic system, blood). CD19 is a well-characterized target for conventional CARs, and for the CoCAR, we selected a single-chain variable fragment (scFv) derived from a CD19-CAR that has been clinically tested (13). In this CD19-CoCAR model system, we confirmed the lack of CoCAR toxicity on target antigen-positive cells by demonstrating the absence of B-cell killing. We created CoCARs with three different costimulatory endodomains to optimize their costimulatory capacity and cytotoxicity was conferred only via the TCR. We validated the CoCAR approach in two model systems: a tTCR targeting an immunodominant epitope of survivin, which is overexpressed in leukemia (14–16), and nTCR-based, polyclonal EBVSTs to target EBV⁺ lymphoma (17). In both model systems, CoCAR⁺ T cells had enhanced persistence and antitumor activity, an effect maintained even in tumors with heterogeneous CoCAR target antigen expression.

Peripheral blood was obtained from HLA-typed (HLA-A2⁺) healthy donors under a Baylor College of Medicine (BCM) Institutional Review Board–approved protocol. Informed written consent was obtained from all donors, and research was guided by the Declaration of Helsinki, the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), Belmont Report, and U.S. Common Rule. Peripheral blood mononuclear cells (PBMC) were separated using Lymphoprep solution (Stemcell Technologies, 07801).

BV173 cells (B-cell acute lymphoblastic leukemia) were obtained from German Cell Culture Collection (DSMZ, #ACC 20, obtained and frozen by C. Arber) and maintained in RPMI1640 medium (Hyclone, SH3002701) supplemented with 20% FBS (Hyclone, 10-082-139) and 2 mmol/L GlutaMAX (Invitrogen 35050-061). A CD19-knockout (KO) BV173 line was generated by CRISPR/Cas9 as described previously (18). In brief, two CD19-specific single-guide RNAs (5′-ttaatacgactcactataGGGCCCCAAGCTGTATGTGTgttttagagctagaaatagc-3′ and 5′-ttaatacgactcactataGGGACCCATGTGCACCCCAAgttttagagctagaaatagc-3′) and a recombinant Cas9 protein (CP01, PNA Bio) were used. 1 μg of each was mixed at room temperature and used to electroporate 0.15 × 10⁶ BV173 cells (three pulses of 1,600 V for 10 ms, Neon Transfection System, Invitrogen). Electroporated cells were expanded in antibiotic-free medium as above, and CD19-negative cells were sorted to greater than 98% purity. A beta-2-microglobulin (B2M)-KO BV173 cell line was generated as previously described and provided by C. Arber (16). Knockout was confirmed by staining with HLA antibodies (Thermo Fisher Scientific, #11-9983-42). BV173 cells stably expressing firefly luciferase-(FFluc) from a retroviral vector were generated as described previously (14) and maintained in RPMI medium with 100 μg/mL geneticin (Gibco, 14072). 293T cells for transfection were obtained from the ATCC (#CRL-3216) and maintained in Iscove's modified Dulbecco's medium (#12440054, Thermo Fisher Scientific) supplemented with 10% FBS and 2 mmol/L GlutaMAX. Cells were expanded for one passage after thawing.

EBV-transformed lymphoblastoid cell lines (LCL) were generated from healthy donor PBMCs by incubation with supernatant from the B95-8 EBV producer cell line (provided by C.M. Rooney), in the presence of 1 μg/mL cyclosporine A (Sandoz), and then maintained in RPMI1640 medium supplemented with 10% FBS and 2 mmol/L GlutaMAX (19). CEM-T2 (TAP transporter-deficient) cells were obtained from the ATCC. All cell lines were Mycoplasma free.

A retroviral vector expressing the survivin T-cell receptor (sTCR) has been described previously (14) and was provided by C. Arber. Briefly, the TCR α and β chains of a T-cell clone specific for the human HLA-A*0201–restricted survivin 95–104 (ELT) epitope were codon optimized by GeneArt (Invitrogen) and cloned into a retroviral vector after replacement of the constant regions with the corresponding murine regions (Fig. 1A). To generate the CD19-CoCAR vector, the CD19-specific FMC63 scFv (20) was cloned in-frame into SFG retroviral vectors [provided by C.M. Rooney (13)] encoding a short Fc hinge and a CD28-derived transmembrane domain with CD28.OX40-, CD28-, or 4-1BB–derived costimulatory endodomains: CD19.28 (CoCAR1), CD19.41BB (CoCAR2), and CD19.28-OX40 (CoCAR3). The CoCAR constructs also contained an IRES sequence, followed by a truncated nerve growth factor receptor (NGFR; ΔCD271) as a detection marker (Fig. 1A). NGFR was introduced into vectors by In-Fusion cloning (Takara Bio In-Fusion HD Cloning Kit, # 639649). Retroviral vector supernatants were produced by transient cotransfection of 293T cells with (i) the plasmids of interest (with LTRs and packaging signals), (ii) the Peg-Pam plasmid encoding MoMLV gag-pol, and (iii) the RDF plasmid encoding the RD114 envelope using GeneJuice transfection reagent (EMD Millipore Corp) as described previously (21). At 48 and 72 hours posttransfection, retroviral supernatant was collected, filtered using a 0.45-μm filter (PALL Life Sciences, #4654), snap frozen, and stored at −80°C.

CD8⁺ T cells were isolated from fresh healthy donor PBMCs of HLA-A2⁺ donors with CD8 magnetic-activated cell sorting (MACS) beads (130-045-201, Miltenyi Biotec GmbH; purity >90%). One million CD8⁺ T cells were stimulated for 3 days in nontissue culture-treated, 24-well plates (Falcon, catalog no. 351147) coated with anti-CD3 produced in-house by the OKT3 hybridoma (ATCC catalog no. CRL-8011, RRID:CVCL_DC77) and anti-CD28 (BD Biosciences, catalog no. 348040, RRID:AB_400367, clone L293), each at 1 μg/mL. ATCs were maintained in T-cell medium consisting of a 1:1 mixture of RPMI1640 and Click's medium (Fujifilm Irvine Scientific, 92705), 10% human AB serum (Valley Biomedical), and 2 mmol/L GlutaMAX (Thermo Fisher Scientific, #35050061). The cells were supplemented with IL7 and IL15 (R&D Systems, 204-IL), both at 10 ng/mL, every 3 days and split as needed.

Dendritic cells (DC) were generated for the initiation of EBVST cultures by isolating monocytes from fresh healthy donor PBMCs with CD14 MACS beads (Miltenyi Biotec; purity >90%). CD14⁺ monocytes were cultured in DC media (CellGenix, 20801-0500), supplemented with 800 U/mL GMCSF (Thermo Fisher Scientific, PHC2011) and 1,000 U/mL of IL4 (R&D Systems, 204-IL) for 5 days, with GMCSF and IL4 replenishment on day 3. DCs were matured on day 5 with 10 ng/mL IL1β, 100 ng/mL of IL6, 10 ng/mL of TNFα (R&D Systems), 1 μg/mL of PGE-2 (Sigma), 400 IU/mL of IL4, and 800 IU/mL GMCSF as described previously (22) and cultured for 2 additional days before harvest.

Fresh healthy donor PBMCs were depleted of naïve CD45RA⁺ T cells using CD45RA MACS beads (130-045-901, Miltenyi Biotec; ref. 23) to isolate memory T cells for EBVST or cytomegalovirus-specific T cell (CMVST) generation. Autologous mature DCs were pulsed at 37°C for 1 hour with 1 ng/mL EBV pepmixes (JPT Peptide Technologies) comprising 15-mer amino acid peptides that overlapped by 11 amino acids and covered the entire protein sequence of IE-1 and pp65 for the CMVST controls and the EBV latent antigens EBNA1, LMP1, and LMP2 for the EBVSTs (17). Memory T cells were then plated with the EBV or CMV pepmix-pulsed DCs at a PBMC:DC ratio of 20:1 in a 24-well tissue culture plate. The stimulated memory T cells were fed with 10 ng/mL of IL7 and IL15, transduced with the retroviral vector on day 3, and supplemented with cytokines every 2 to 3 days throughout the culture. A second stimulation with pepmix-pulsed autologous irradiated PBMCs at a ratio of 1:1 was performed between days 9 and 11. A third stimulation with pepmix-pulsed irradiated PBMCs occurred 7 to 9 days following the second stimulation. Functional assays were performed on the days of each resti-mulation and 7 days after the third stimulation.

ATCs or EBVSTs were transduced 3 days after activation. The day prior to transduction, 24-well, nontissue culture plates were coated with 3.5 μg of RetroNectin (Takara T100B; RetroNectin concentration 7 μg/mL) in 0.5 mL of Dulbecco's PBS (Sigma-Aldrich, D8537) and kept at 4°C overnight. Following RetroNectin removal, 1 mL of retroviral supernatant containing either sTCR or CoCAR encoding retroviral particles was added to each well. For generation of sTCR⁺CoCAR⁺ ATCs or EBVSTs, 1 mL of each retroviral supernatant was mixed and added to each well. Conventional CD19-CAR ATCs were generated by adding 1 mL of retroviral supernatant containing the CD19-CAR encoding retroviral particles (13) to each well. The plates were then centrifuged at 2,000 × g for 1 hour at 4°C to promote viral adhesion to the RetroNectin. Following centrifugation, the supernatant was aspirated and 2 × 10⁵ T cells were added to each well in T-cell medium with 10 ng/mL IL7 and IL15. The plates were incubated at 37°C for the following 2 to 3 days before collection and replating in fresh T-cell medium. ATCs were expanded for 6 to 7 days prior to initiation of functional assays, with supplementation of 10 ng/mL IL7 and IL15 every 3 days. EBVSTs were expanded as described above. For the in vivo experiment, another transduction, as described above, was performed with GFP-firefly luciferase 3 days after the second stimulation.

Chromium release cytotoxicity assays were performed against target B cells. A total of 1 × 10⁶ B cells (LCLs as described above) were incubated with ⁵¹Cr sodium chromate (PerkinElmer) at 37°C for 1 hour and then washed with T-cell medium as above. The following effector cells were used at multiple effector to target cell ratios (20:1, 10:1, 5:1, 2.5:1): nontransduced (NT) ATCs, CoCAR-transduced ATCs, and conventional CD19-CAR ATCs (13). Effector and target cells were cocultured in 96-well plates for 4 hours. Target cells cultured without effector cells or lysed with 0.1% Triton-X (Sigma-Aldrich, #T8787) served as the controls for spontaneous release and maximum release, respectively. Cytotoxicity against B cells was assessed by the amount of chromium released (counts per minute, CPM), as quantified by a gamma counter (PerkinElmer 2470 WIZARD2). Specific lysis percentage was calculated by subtracting the spontaneous release from the measured CPM and then dividing by the difference of the maximum release and the spontaneous release.

We measured transduction efficiencies by flow cytometry and used NT cells as negative controls. Cells were washed in PBS, then the appropriate detection antibodies were added to 2.5 × 10⁵ pelleted cells and incubated for 20 minutes in the dark at room temperature. The cells were then washed in PBS and analyzed on a Gallios flow cytometer (Beckman Coulter 10 color/3 laser). All analysis was conducted with Kaluza Analysis Software (Beckman Coulter). Antigen presenting cell (APC)-conjugated antibody to the constant region of the TCR β-chain was used to detect sTCR (Invitrogen, #14-5961-82, clone H57-597) and PE- or BV421-conjugated CD271 was used for detecting the CoCAR (BD #562562, clone C40-1457).

For coculture assays, ATCs were identified using APC 700-conjugated CD8 antibody (Beckman Coulter, catalog no. A66332, RRID:AB_2750854, clone B9.11), and BV173 target cells were detected with FITC-conjugated anti-CD33 (BD #555626, clone HIM3-4). Cells were phenotyped with BV421-conjugated anti-CCR7 (#353207, clone G043H7) and ECD-conjugated anti-CD45RO (Beckman Coulter #IM2712U, clone UCHL1) to assess memory phenotype. Anti-TIM-3 (FITC; F38-2E2, Thermo Fisher Scientific), anti-LAG3 (#369212, BioLegend), and anti-PD-1 (PC7, #A78885, Beckman Coulter) were used for analysis of markers of T-cell exhaustion.

The initial gate for all analysis was set on lymphocytes using SSC and FSC. A gate was then placed on CD8⁺ cells to characterize CD8⁺ T cells. For coculture assays, an additional gate was set on CD33⁺ cells to identify tumor cells.

We used enzyme-linked immunospot (ELISpot) analysis to quantify the number of cells secreting IFNγ in response to appropriate antigen stimulation. A total of 1 × 10⁵ effector cells were plated in triplicate in a 96-well plate precoated overnight with immobilized anti-IFNγ (MabTech, #3420-3-1000) at 10 μL/mL as described previously (24). T2 cells pulsed with 1 μL of survivin peptide (0.1 mol/L) were used as target cells for experiments using sTCR⁺ and sTCR⁺CoCAR⁺ cells. EBVSTs and CoCAR⁺ EBVSTs were directly pulsed with EBV pepmixes (1 μg/mL) or medium alone. Following overnight incubation at 37°C, cells were removed, and the plates were developed with anti-IFNγ (MabTech, #3420-6-5000 at 1 μL/mL,), dried overnight, and sent to ZellNet Consulting for quantification. To control for background signal, the number of spot-forming cells of the negative control was subtracted from those of the experimental conditions.

ATCs from four different conditions (NT, sTCR alone, CoCAR alone, or sTCR⁺CoCAR⁺) were serially cocultured with tumor cells to measure their antitumor efficacy at an effector to target (E:T) ratio of 1:5. ATCs were plated in 48-well tissue culture plates at 1 × 10⁵ cells per well in multiple replicates. Wild-type (WT), CD19 knockout (CD19-KO), or B2M-KO BV173 cells were added at 5 × 10⁵ cells per well in T-cell medium. No supplemental cytokines were added. Every 3 to 4 days, for up to eight rounds of serial cocultures, a single well was harvested for flow cytometry to quantify remaining BV173 tumor cells and T cells. The same day, 5 × 10⁵ fresh BV173 cells were added to the remaining wells. To determine whether CD19 was required on all target cells, a serial coculture assay was performed using decreasing ratios of WT to CD19-KO BV173 target cells as indicated in figures. The E:T-cell ratio of 1:5 was maintained throughout all rounds of coculture, but the ratio of WT to CD19-KO BV173 cells was varied as specified in the figure legends. Supernatants were collected for IFNγ analysis via ELISA as described below. The expe-rimental endpoint for the cocultures was defined as failure of the T cells to eliminate the tumor cells.

IFNγ ELISAs were performed as directed by the manufacturer's protocol (R&D Systems, #DIF50) to quantify amounts of IFNγ in supernatants from the cocultures described above. Supernatants were harvested 24 hours after the addition of fresh BV173 cells. IFNγ secreted was quantified by the absorbance of anti-IFNγ conjugated to horseradish peroxidase at 450 nm (Tecan InfiniteF50). To convert absorbance to concentration, manufacturer provided IFNγ standards of known concentrations were used to create a standard curve.

WT BV173 cells were cocultured with either sTCR⁺ or sTCR⁺CoCAR3⁺ T cells at an E:T ratio of 1:5 for two rounds of serial coculture as described above. T cells were collected, mixed with BV173 cells at a ratio of 1:5 for 24 hours, and then separated using CD4 and CD8 MACS beads (Miltenyi Biotec, Germany, #130-045-101 and 130-045-201). Mitochondrial function was determined using Agilent Seahorse XFe96 Extracellular Flux Assay Kits (Agilent, catalog no. 102416), and analysis was conducted by the Baylor College of Medicine Mouse Metabolic and Phenotyping Core. Data were analyzed using the Seahorse Wave software (ECAR and OCR/Agilent).

WT BV173 cells were cocultured with either sTCR⁺ or sTCR⁺CoCAR⁺ ATCs at an E:T ratio of 1:5. Four days later, flow cytometry was performed to measure tumor elimination, and 5 × 10⁵ BV173 cells were added back per well of culture. Twenty-four hours later, the cocultured cells were harvested. In conditions in which BV173 cells were still present, CD33 MACS beads (130-045-501, Miltenyi Biotec) were used to deplete residual tumor cells. Total RNA was collected from ATCs using the RNeasy Micro Kit (Qiagen) following kit instructions for extraction from cells. Gene expression analysis (NanoString) was performed by the Baylor College of Medicine Genomic and RNA Profiling Core using the nCounter CAR-T Characterization Panel. Submitted samples were >250 ng and <5 μL. Data were analyzed using the nSolver 3.0 software (NanoString). See statistical section for additional information on NanoString analysis.

Animal experiments were conducted on a protocol approved by BCM's Institutional Animal Care and Use Committee. To test the antitumor activity of sTCR⁺CoCAR⁺ ATCs in vivo, 3 × 10⁶ BV173-FFLuc cells were injected intravenously into the tail vein of female NOD SCID γc^−/− (NSG; Jackson Labs) mice after irradiating mice at 120cGy (Precision XRad320). After the indicated timeframe, ATCs (NT, sTCR⁺, CoCAR⁺, or sTCR⁺CoCAR⁺) at the indicated dose were injected into the tail vein. Tumor signal was measured by bioluminescent imaging (BLI; Xenogen, IVIS, Small Animal Core Facility, Texas Children's Hospital), beginning 1 week after T-cell injection and continuing weekly thereafter. To prepare for BLI, mice were anesthetized with isoflurane (USP, Covetrus, OH #029405) then injected with luciferin (PerkinElmer, #122799) intraperitoneally. After 10 minutes, mice were placed in the IVIS for imaging. Mice were euthanized if they had impaired mobility or had >20% body weight loss. To measure the effect of the CoCAR on EBVSTs, 3 × 10⁶ EBV-LCLs were injected subcutaneously into NSG mice. A total of 5 × 10⁶ FFLuc+ EBVSTs or irrelevant CMVSTs (control) were injected the following day. EBVST bioluminescence signal was measured weekly by BLI, and tumor volume was measured with a digital caliper (Whitworth 0–150 mm) weekly. Mice were euthanized when tumors reached 1.5 mm diameter in the longest axis. Tissue was collected at the experimental endpoint or at times of early euthanasia. All mice were purchased from Jackson Laboratories.

Descriptive statistics with mean and SDs were used to summarize results. Student t test or ANOVA, when appropriate, was used to test for significance between different conditions in each assay. A P value < 0.05 was considered statistically significant. Survival for in vivo experiments was analyzed by the Kaplan–Meier method, and survival differences were compared using the log-rank test. NanoString data were analyzed using the nSolver 3.0 software and the “Advanced Analysis” function with standard normalization; P-value cutoff was set at P = 0.05, and the Benjamini–Hochberg correction was applied.

To identify the optimal signaling domain(s) for our CoCAR strategy, we tested three CoCAR constructs with different costimulatory endodomains: CD19.28 (CoCAR1), CD19.41BB (CoCAR2), and CD19.28-OX40 (CoCAR3). All constructs included truncated NGFR/CD271 (ΔCD271) as a transduction marker (Fig. 1A). To assess their costimulatory capacity in vitro, we coexpressed the CoCARs with a transgenic HLA-A*02:01-restricted TCR targeting the tumor-associated antigen survivin (sTCR; Fig. 1A; ref. 14). NT, single-transduced (sTCR⁺ or CoCAR⁺), and double-transduced (sTCR⁺CoCAR⁺) CD8⁺ T cells were expanded using three antigen-specific stimulations, and autologous, PBMC that contain a fraction of normal CD19⁺ B cells. T-cell expansion, transgene expression, and antigen-specific IFNγ production during expansion was assessed. The dual-transgenic population was enriched in the sTCR⁺CD19.28-OX40⁺ condition from 49% (±18.1) to 77% (±10.4, P < 0.01) between days 9 and 30, whereas the proportion of transgenic cells remained stable in sTCR⁺, sTCR⁺CD19.41BB⁺, and sTCR⁺CD19.28⁺ T cells (Fig. 1B). In contrast, T-cell cultures transduced with any of the three CoCARs alone showed a nonsignificant decrease in transgene-expressing cells over time. T-cell expansion was overall comparable across conditions (Fig. 1C). By the end of the third stimulation, IFNγ secretion in response to survivin peptide was enhanced in sTCR⁺CD19.28-OX40⁺ T cells compared with sTCR⁺ T cells (P < 0.05), as measured by ELISpot. Conversely, IFNγ secretion by sTCR⁺CD19.28⁺ T cells was decreased compared with NT controls at day 30 (Fig. 1D). Corresponding results were obtained using ELISAs measuring total IFNγ in coculture supernatants in response to TCR stimulation (P < 0.01; Fig. 1E). The decrease in IFNγ secretion by sTCR⁺CD19.28⁺ T cells was associated with increased T-cell death between days 20 and 30 resulting in failure to expand (Fig. 1C). Because of poor cytokine release in response to TCR antigen stimulation of CoCAR1-modified sTCR⁺ T cells and increased cell death, further experiments were not performed with this construct. The distribution of effector and memory cells was determined in T cells transduced with CoCAR2 and CoCAR3 constructs and was not significantly different compared with control cells by day 30 (Fig. 1F). To ensure that CoCAR T cells did not kill CD19⁺ target cells, we performed a ⁵¹Cr-release assay with CoCAR⁺ T cells (single-transduced without TCR specificity) compared with conventional CD19-CAR T cells and confirmed lack of cytotoxic activity against CD19⁺ B cells through CoCAR3 (Fig. 1G).

To evaluate whether expression of the CoCAR2 and CoCAR3 enhanced antitumor function of sTCR⁺ T cells, we in vitro challenged engineered T cells or controls up to eight times with BV173 leukemia cells that present both the cognate survivin peptide and CD19 (Fig. 2A and B). T cells were rechallenged every 3 to 4 days until they were no longer able to eliminate the tumor cells. As expected, NT and CoCAR⁺ T cells were unable to kill tumor cells (Fig. 2C), whereas sTCR⁺ T cells killed tumor cells for a median of three challenges, with subsequent outgrowth of tumor cells. In contrast, sTCR⁺CoCAR3⁺ T cells extended antitumor activity by a median of four challenges (P < 0.01); cells from 5/7 donors eliminated tumor cells for all eight rounds of tumor challenges. Sequential tumor killing activity was associated with increased expansion of sTCR⁺CoCAR3⁺ T cells compared with sTCR⁺ T cells (P < 0.05, Fig. 2D). sTCR⁺CoCAR2⁺ T cells did not improve antitumor activity compared with sTCR⁺ T cells (Fig. 2C), and most T cells died over the course of the coculture (Fig. 2D). Thus, incorporation of CoCAR3, but not CoCAR2, improved sTCR⁺ T-cell sequential killing. Therefore, for the remaining experiments, we eliminated CoCAR2 and continued with CoCAR3 in subsequent sections.

To determine whether CoCAR3 also increased cytokine production by stimulated T cells, we harvested coculture supernatants 24 hours after each tumor rechallenge and measured IFNγ by ELISA. IFNγ secretion was consistently higher in sTCR⁺CoCAR3⁺ T cells compared with sTCR⁺ T cells across all timepoints (P = 0.0041; Fig. 2E). CoCAR3 did not induce significant differences in the memory phenotype nor differences in expression of the exhaustion markers TIM-3, LAG3, and PD-1 after the initial two rounds of coculture (Supplementary Fig. S1). In cocultures over time, double-transduced T cells were enriched in the sTCR⁺CoCAR3⁺ T-cell population, with 46.4% (±18.8%) at baseline to 87.8% (±6.2%) by day 14 (P < 0.01, Fig. 2F). These results confirmed that double-transduced T cells had a proliferative advantage during serial tumor challenge.

To validate antigen-specific activity of both sTCR and CoCAR3 constructs, we performed sequential cocultures with CD19-KO (lacking the CoCAR antigen) or B2M-KO (unable to present the survivin peptide) BV173 cells. In cocultures with CD19-KO BV173 cells, no difference in the duration of tumor control of sTCR⁺ and sTCR⁺CoCAR3⁺ T cells was observed, indicating absent or minimal background activity of the CoCAR3 in the absence of CD19 antigen (Fig. 2G). In cocultures with B2M-KO BV173 cells, antitumor activity was completely abrogated (Fig. 2H), confirming that, despite high CD19 expression on the target cells, the CoCAR3 has no effect on tumor killing in the absence of signaling through the class I–restricted sTCR.

To investigate the pathways activated upon CoCAR3 ligation in sTCR⁺CoCAR3⁺ ATCs, we compared the production of T_H1/T_H2 cytokines and cytotoxic granules in coculture supernatants of sTCR⁺CoCAR3⁺ and sTCR⁺ T cells and also performed gene expression analysis on T cells after two challenges with WT BV173 cells (HLA-A*02:01⁺survivin⁺ and CD19⁺). Compared with sTCR⁺ T cells, sTCR⁺CoCAR3⁺ T cells secreted more perforin (3.2-fold increase; P < 0.05), granzyme B (11.1-fold increase; P < 0.05), TNFα (3.8-fold increase; P < 0.005), IFNγ (5.0-fold increase; P < 0.005), and GMCSF (3.7-fold increase; P < 0.005; Fig. 3A). IL2, IL4, IL6, and IL10 were low and comparable between conditions [Fig. 3A (IL2); Supplementary Fig. S2]. Consistent with their increased protein levels, IFNγ (IFNG), granzyme B (GZMB), and perforin (PRF1) gene expression was upregulated (P < 0.05) in sTCR⁺CoCAR3⁺ compared with sTCR⁺ T cells (Fig. 3B). We also observed increased pathway scores for IL signaling (P = 0.0378), T-cell activation (P = 0.0320), costimulatory molecules (P = 0.0288), and mTOR (P = 0.0211) in sTCR⁺CoCAR3⁺ compared with sTCR⁺ T cells (Fig. 3C). Metabolic testing at baseline and after coculture with tumor cells using the seahorse assay showed a trend toward increased spare respiratory capacity in sTCR⁺CoCAR3⁺ over sTCR⁺ T cells after two rounds of coculture (Supplementary Fig. S3), but this difference did not reach statistical significance.

Because CD19 (the CoCAR antigen) might be heterogeneously expressed on tumors or provided by normal bystander cells, we investigated the amount of CD19-positivity required to produce sufficient stimulation through CoCAR3 to enhance antitumor activity. We tested sequential killing ability and cytokine production of sTCR⁺CoCAR3⁺ T cells in a modified coculture assay. By mixing WT with CD19-KO BV173 tumor cells at various ratios, we created tumor target populations with 100%, 75%, 50%, 25%, or 0% CD19⁺ cells (Fig. 4A). With 100%, 75%, and 50% CD19⁺ targets, sTCR⁺CoCAR3⁺ T cells eliminated tumor cells for four additional challenges compared with sTCR⁺ T cells (P < 0.01; Fig. 4B). At 25% CD19-positivity, sTCR⁺CoCAR3⁺ T cells still showed a trend toward prolonged tumor elimination in comparison with sTCR⁺ T cells (median increase of 2.5 cocultures; interquartile range, 1 to 4). Consistent with prior coculture results, sTCR⁺ and sTCR⁺CoCAR3⁺ T cells did not differ in their activity against CD19-negative tumor cells, indicating minimal background signaling of the CoCAR.

As expected, the proportion of CD19⁺ cells in culture did not affect IFNγ production by sTCR⁺ T cells (Fig. 4C, left). sTCR⁺CoCAR3⁺ T cells produced similar amounts of IFNγ when 100%, 75%, or 50% of BV173 cells expressed CD19 (100% CD19⁺: 3,278 ± 1,172 ng/mL, 75%: 2,951 ± 1,297 ng/mL, 50%: 2,915.9 ± 1,185 ng/mL). IFNγ concentration decreased when 25% of tumor cells expressed CD19 (1,534.9 ± 733.0 ng/mL, P not significant vs. 50%, 75% and 100%), but remained 2.4 times higher than in the absence of CD19 (651.3 ± 473.4 ng/mL; P < 0.05).

Next, we evaluated whether CoCAR3 enhanced the in vivo antitumor function of sTCR⁺ T cells in our previously established BV173-FFLuc mouse xenograft model (14, 15). Sublethally irradiated mice were injected intravenously with BV173-FFLuc cells, followed 7 days later by a single intravenous injection of 1 × 10⁷ T cells (Fig. 5A). As expected, NT and CoCAR3⁺ T cells failed to control the leukemia, whereas a significant delay in leukemia progression was observed in mice treated with sTCR⁺CoCAR3⁺ compared with sTCR⁺ T cells, and this delay in tumor progression improved overall survival of mice treated with sTCR⁺CoCAR3⁺ over sTCR⁺ T cell–treated mice by a median of 35 days (Fig. 5B and C; P < 0.01). We confirmed the improved antitumor efficacy in less established tumors using a lower T-cell dose of 4 × 10⁶ cells administered 1 day after tumor cell injection (Supplementary Fig. S4). At this dose, treatment with sTCR⁺ T cells alone did not extend survival over NT control cells, whereas treatment with sTCR⁺CoCAR3⁺ T cells resulted in improved survival by 11 days (P < 0.01).

To investigate the effect of CoCAR3 costimulation on the function of tumor-specific T cells expressing native TCRs with a broad repertoire, we engineered EBVSTs with the CoCAR3. We achieved a mean transduction efficiency of 60.9% (±27.3%), which remained stable during EBVST expansion (Fig. 6A). The transduction procedure briefly slowed EBVST expansion, but fold expansion was similar to NT EBVSTs after the second stimulation (Fig. 6B). The CoCAR3 increased the frequency of EBV-specific T cells within the EBVSTs by a mean of 2.4-fold (90% confidence interval: 2.4 ± 1.3) as measured by an IFNγ ELISpot at the end of the first stimulation cycle (S1) in five of six donors (Fig. 6C). Memory phenotype, determined by CD45RO and CCR7 expression, was unaffected by CoCAR3 expression over two stimulations, and a trend toward a higher proportion of CD8⁺ T cells in CoCAR3⁺ EBVSTs after the second stimulation was not statistically significant (Supplementary Fig. S5). We confirmed lack of bystander killing in a cytotoxicity assay using CoCAR3⁺ EBVSTs as effector cells (NT EBVSTs in the control condition) and an EBV⁺CD19⁺ LCLs as target cells (Supplementary Fig. S6); labeled EBV-negative B cells were not killed, even when mixed with CD19⁺ EBV-LCLs that were targeted and killed by CoCAR3⁺ EBVSTs.

We then evaluated the antitumor activity of engineered CoCAR3⁺ EBVSTs in an autologous EBV-LCL xenograft mouse model. After EBV-LCL tumors were injected subcutaneously and established, NSG mice were injected intravenously with 5 × 10⁶ FFLuc-engineered EBVSTs or CoCAR3⁺ EBVSTs or irrelevant VSTs specific for CMV. T-cell expansion was monitored by BLI (Fig. 6D). CoCAR3⁺ EBVSTs expanded more rapidly than NT-EBVSTs (Fig. 6E and F), which in turn correlated with more rapid tumor control (Fig. 6G). The peak of EBVST expansion, measured by BLI signal at the tumor site, occurred on day 9 after infusion of CoCAR3⁺ EBVSTs and at the time was 7.3-fold higher than for EBVSTs (BLI total flux day 9: 1.7 ± 1.1 vs. 12 ± 1.6 × 10⁹ p/s; P < 0.0001). The increased EBVST expansion preceded a reduction in tumor size, which was 77.1% smaller in CoCAR3⁺ EBVST compared with EBVST-treated mice on day 13 (0.81 ± 0.25 vs. 0.13 ± 0.12 cm³; P = 0.005; Fig. 6G, right).

In this study, we showed that engineered costimulation through a CD19.28-OX40 CoCAR (CoCAR3) enhanced the antitumor activity of T cells with both native and tTCRs in vitro and in vivo in two different xenograft models of leukemia and lymphoma. CoCAR3-mediated costimulation improved the antitumor activity of T cells, even when targeting tumors with heterogeneous CD19 expression. Our approach provides the proof of concept that costimulation with a CoCAR can specifically enhance antitumor function of native and tTCR T cells.

Engineering strategies to provide local costimulation or cytokine signals to T cells within the TME have focused mostly on converting inhibitory immune checkpoint (25) or cytokine signals (26) into immune-stimulatory outputs (recently reviewed in ref. 11) or by direct modification of tTCRs (27). Preclinical evaluation of CoCARs has investigated their ability to mitigate the toxicity of CAR T-cell therapies, in which two different target antigens are recognized simultaneously by two independent CARs, termed a “split CAR” system. In these systems, a first-generation CAR provides CD3ζ signaling and the CoCAR exclusively provides costimulation (28, 29) with the goal to modulate cytotoxicity and potential on-target/off-tumor toxicity of the CAR. However, the immune-enhancing effects of CoCARs in combination with native or tTCRs had not yet been explored.

We initially evaluated three different costimulatory endodomains, (i) CD28, (ii) 41BB, and (iii) CD28 combined with OX40, selected on the basis of their different biology. Although standard CD19.28.ζ CAR T cells expand and contract rapidly after infusion in patients (30), CD19.41BB.ζ CAR T cells are known for slightly slower onset of action and longer-term persistence (31–33). The CD28-OX40 endodomain has previously been validated preclinically in CAR T cells with a range of different antigen specificities (34–39). In a comparison of 12 different costimulatory molecules, OX40 was shown to confer the greatest benefit to 41BBζ.CAR T cells, increasing T-cell prolife-ration, cytotoxicity, and antitumor activity (40). In our experiments, the CD19.28-OX40 CoCAR3 outperformed the CD19.28 and the CD19.41BB CoCAR constructs, with increased proliferation compared with CD19.28 and superior performance in sequential tumor killing assays. Because our CoCAR constructs exerted minimal antigen-independent activity, we hypothesize that minimal baseline signaling of the CD19.28-OX40 CoCAR3 may avoid the effects of tonic signaling that results in T-cell exhaustion, as observed with CD28-OX40ζ in the context of a GD2-CAR (37), while maintaining the beneficial effects of OX40 on T-cell function and persistence (40).

We selected CD19 as an extensively characterized and clinically validated CAR target that could provide the following advantages: (i) CD19 is expressed on both lymphoid malignancies and normal B cells; (ii) normal CD19⁺ lymphocytes are abundant throughout the blood and lymphoid tissues and are easily encountered by adoptively transferred T cells; (iii) B cells can provide additional costimulation as professional antigen-presenting cells if they receive signals from activated T cells; and (iv) activated B cells are often recruited to tumors with proinflammatory environments. Therefore, the CD19-CoCAR may allow for activation of the CoCAR by the tumor itself in cis or in trans by B cells in lymph nodes or the TME.

Our results demonstrate that CoCAR3 enhanced the potency of tumor-specific native and tTCR T cells when encountering CD19 on as few as 25% target cells. Because the CoCAR is not cytotoxic, it should not drive antigen escape, but if antigen escape did occur, or if the CoCAR target antigen was expressed in only a subset of cells, its beneficial effect would therefore not be diminished. Possible explanations for this observation are that (i) CoCAR activation by CD19 in cis leads to sustained CD28-OX40 activation beyond the first target cell recognition and killing; (ii) CoCAR3-engineered T cells produce cytokines, proliferate, and then encounter and kill additional target cells, even in the absence of CD19 expression; and (iii) provision of CD19 in trans may activate CoCAR-mediated signaling and enhance target cell killing through the TCR. However, the spatiotemporal sequence of events could not be fully elucidated in our experimental system.

We investigated our CoCAR3 approach in two different murine xenograft models, one targeting a poorly immunogenic, aggressive acute leukemia using a tTCR, and the other targeting highly immunogenic EBV⁺ LCLs using polyclonal EBVSTs expressing nTCRs. In both models, CoCAR3 enhanced the antitumor activity of the T cells, even though the models differ significantly in several aspects. The tTCR has low affinity, is HLA class I-restricted, and is CD8-dependent. By contrast, polyclonal nTCRs in EBVSTs are, in general, of high affinity, as they are targeting foreign viral antigens (41), and comprise a mixture of CD4⁺ and CD8⁺ T cells. Unlike most EBV⁺ lymphomas and other EBV⁺ cancers, EBV⁺ LCLs are highly immunogenic and express high levels of costimulatory ligands that may mask the full potential of engineered costimulation through the CoCAR3. In our LCL xenograft model, we observed more rapid initial expansion, followed by more rapid contraction of the CoCAR-modified EBVSTs, as tumors were controlled. At the time the T-cell signal decreased, all tumors had resolved by palpation, and none of the mice experienced tumor recurrence, indicating that the lack of EBV antigen (signal 1) and CD19 antigen (signal 2) following rapid tumor clearance resulted in the decline in T-cell numbers.

Finally, we performed gene expression analysis of sTCR⁺CoCAR3⁺ T cells upon tumor challenge to determine changes conferred by the CoCAR in response to targeted tumors. We found upregulation of genes associated with T-cell activation, costimulation, and cytokine signaling, indicating that several pathways contribute to the observed functional changes. The upregulation of these genes was associated with increased granzyme B, perforin, IFNγ, and other cytokines and enzymes, which may confer increased antitumor activity to CoCAR-modified cells. Our group and others have previously investigated the interactions between CAR signaling and TCR signaling in VSTs. The combination of signaling through both the CAR and through the native/tTCR can produce excessive CD3ζ signaling, which can in turn produce activation-induced cell death (AICD; refs. 42, 43). The CoCAR approach avoids this excessive ζ signaling seen with conventional CARs, thus decreasing the risk of AICD.

We consider our proposed combination of TCR targeting with CoCAR costimulation to be a model system that could be adapted to target other cancer types, either by targeting tumors with high B-cell numbers in the TME or by changing the antigen-targeting domain to an alternative antigen relevant to the targeted tumor. The lack of costimulatory ligand expression in the TME is particularly problematic in solid tumors. Accordingly, durable clinical benefit after TIL therapy or tTCR T-cell therapies in solid tumors has only been achieved in a minority of patients treated (6, 7, 9), indicating a major need to develop strategies that deliver immune stimulation for TCR-based ACT. Additional provision of CoCAR-modified T cells with supplemented or engineered cytokine signaling could further enhance their efficacy by completing the three-signal requirement of T-cell activation (44).

In summary, we provide a proof-of-concept study that engineered costimulation with a CoCAR can significantly enhance the antitumor activity of TCR-based cell therapies, with both native and tTCRs, while sparing normal cells that express the CoCAR-targeted antigen. Our novel approach may be extended to other antigens with the goal to provide tumor- and TME-specific costimulation.

B. Omer reports grants from Gabrielle's Angel Foundation, NIH, Cancer Prevention and Research Institute of Texas, and Texas Children's Cancer Center K12 during the conduct of the study. C. DeRenzo reports grants from the Rally Foundation for Childhood Cancer Research, The Truth 365, the Sarcoma Foundation of America, and the NIH grant K12CA0904335 during the conduct of the study. S. Gottschalk reports other support from TESSA Therapeutics, Immatics, Catamaran Bio, TIDAL, and Nektar Therapeutics outside the submitted work; in addition, S. Gottschalk has multiple patents in the field of engineered T-cell therapies pending. C. Arber reports grants from Cancer Prevention & Research Institute of Texas during the conduct of the study; in addition, C. Arber has a patent for Survivin-specific T cell receptor targeting tumor but not T cells issued, a patent for Reprogramming CD4 T cells into cytotoxic CD8 cells by forced expression of CD8ab and class I-restricted T-cell receptors issued, licensed, and with royalties paid from Immatics, and a patent for transgenic c-MPL provides ligand-dependent costimulation and cytokine signals to TCR-engineered T cells pending. C.M. Rooney reports grants from NIH-NCI during the conduct of the study; grants and personal fees from Tessa Therapeutics; personal fees from Allovir, Marker Therapeutics, Bellicum Pharmaceuticals, Allogene, Abintus, Bluebird Bio, Memgen, TScan, Turnstone Biologics, and Walking Fish outside the submitted work. No disclosures were reported by the other authors.

B. Omer: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. M.G. Cardenas: Data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. T. Pfeiffer: Data curation, formal analysis, writing–review and editing. R. Daum: Data curation, writing–review and editing. M. Huynh: Data curation, writing–review and editing. S. Sharma: Data curation, writing–review and editing. N. Nouraee: Data curation, methodology, writing–review and editing. C. Xie: Data curation. C. Tat: Data curation. S. Perconti: Data curation, writing–review and editing. S. Van Pelt: Data curation, supervision, writing–review and editing. L. Scherer: Data curation, writing–review and editing. C. DeRenzo: Methodology, writing–review and editing. T. Shum: Data curation, methodology, writing–review and editing. S. Gottschalk: Conceptualization, methodology, writing–review and editing. C. Arber: Conceptualization, resources, formal analysis, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing. C.M. Rooney: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.

This work is supported by NIH SPORE in lymphoma (5P50CA126752; B. Omer, C.M. Rooney), Gabrielle's Angels Foundation for Cancer Research (B. Omer), Texas Children's Cancer Center K12 (K12CA090433; B. Omer), Cancer Prevention & Research Institute of Texas (CPRIT IIRA RP160345; C. Arber, M.G. Cardenas) and Sharon Levin Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.