Secreted Fas Decoys Enhance the Antitumor Activity of Engineered and Bystander T Cells in Fas Ligand–Expressing Solid Tumors

Chimeric-antigen receptor (CAR)-modified T cells have demonstrated remarkable clinical success in treating hematologic malignancies. However, extension to solid tumors has been more challenging due to many factors including the hostile tumor microenvironment (TME), which is replete with suppressive cells and molecules that limit effector T-cell function (1, 2). One such suppressive molecule is Fas ligand (FasL), which is expressed by a range of TME cells including tumor cells, tumor endothelium, myeloid-derived suppressor cells (MDSC), tumor-associated macrophages (TAM), regulatory T cells (Treg), and cancer-associated fibroblasts (CAF; refs. 3–8). Upon engagement with its endogenous receptor Fas (expressed by activated T cells), FasL induces apoptotic cell death (3–8).

Under normal physiologic conditions, Fas/FasL signaling plays a crucial role in regulating normal immune function, including T-cell homeostasis, and establishing immune privilege. For example, contraction of the expanded population of antigen-specific T cells following resolution of an immune response against a viral infection is partially mediated by FasL-induced apoptosis (9). Similarly, FasL, expressed in tissues of testes, ovary, and cornea, induces apoptotic death of infiltrating T cells and helps maintain the immune privilege of these organs (10). However, some tumors exploit this pathway as a mechanism to evade immune-mediated elimination. In preclinical studies, Zhu and colleagues demonstrated that FasL-expressing polymorphonuclear (PMN)-MDSCs found in TiRP melanoma tumors induced apoptotic cell death of tumor-infiltrating lymphocytes (TIL), an effect that was alleviated by blocking Fas/FasL engagement (6). Similarly, Lakins and colleagues showed that CAFs and other stromal cells recruited to the tumor site in mice engrafted with B16.OVA melanoma tumors could induce OVA-specific CD8⁺ T-cell death via the Fas/FasL signaling pathway (7). Finally, elevated levels of FasL have been detected in various solid tumors including bladder, breast, cervical, colon, gastric, and pancreatic cancers, and sarcoma, and correlated with disease progression, increased metastasis, and poor overall survival (11–14). For instance, Zietz and colleagues reported that FasL was detected in >70% of 40 angiosarcomas analyzed, and high-level expression correlated with decreased numbers of CD8⁺ TILs and inferior patient survival (15).

Given the detrimental impact of tumor-derived FasL on both endogenous and adoptively transferred antitumor immunity, we sought to protect effector T cells by engineering them to express a secreted Fas decoy (FD) molecule. This molecule was designed (i) to be produced upon T-cell engagement with cognate antigen at the tumor site and (ii) to shield both transgenic and bystander (nonmodified) tumor-specific cells, including TILs and NK cells, from FasL-induced cell death. Using pancreatic cancer models, we demonstrate enhanced T-cell expansion and sustained effector function of FD-engineered T cells in vitro and in vivo. Furthermore, incorporating the proproliferative cytokine IL15 to produce an FD and IL15 fusion protein (FD+15) strengthened the antitumor response by simultaneously shielding the apoptotic signal of FasL and promoting T-cell growth and survival.

Our strategy of protecting engineered as well as endogenous bystander effector cells from tumor FasL–mediated immunosuppression could enhance their persistence in the TME, resulting in a more pronounced antitumor response. However, this approach does not address the potential for immune evasion by the tumor due to heterogeneous antigen expression, which is another major hurdle in targeting solid tumors using cell therapies.

Peripheral blood for T-cell generation was obtained from nine healthy volunteers after obtaining written informed consent on a protocol approved by the Baylor College of Medicine Institutional Review Board in accordance with the Declaration of Helsinki, The Belmont Report, and the U.S. Common Rule. Immediately after blood was drawn, peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation using Lymphoprep (Serumwerk Bernburg AG for Alere Technologies AS; catalog no. 1114544). 293T, K562, CAPAN-1, CFPAC-1, ASPC-1, DU145, PC-3, MCF-7, MDA-MB-448, LAN-1, and NK-92 were obtained from the American Type Culture Collection between 2012 and 2014, and passages between 8 and 24 were used in the experiments. 293T, CAPAN-1, CFPAC-1, ASPC-1, DU145, and PC-3 were cultured in complete Iscove's modified Dulbecco's medium (IMDM; Gibco BRL Life Technologies, Inc.; catalog no. 12440-046) supplemented with 10% heat-inactivated fetal bovine serum (FBS; HyClone; catalog no. SH30088.02HI) and 2 mmol/L GlutaMAX (Gibco BRL Life Technologies; catalog no. 35050-061). K562 and LAN-1 were cultured in RPMI-1640 medium (Cytiva-HyClone; catalog no. SH30096.01) supplemented with 10% heat-inactivated FBS and 2 mmol/L GlutaMAX. NK-92 cells were cultured in RPMI-1640 medium supplemented with 10% heat-inactivated FBS, 2 mmol/L GlutaMAX, and 50 IU/mL IL2. MCF-7 and MDA-MB-448 were cultured in DMEM (Cytiva-HyClone; catalog no. SH30081.01) supplemented with 10% heat-inactivated FBS and 2 mmol/L GlutaMAX. All cell lines were maintained in a humidified incubator containing 5% carbon dioxide (CO₂) at 37°C. Mycoplasma testing was performed on cell culture supernatants using the MycoAlert Mycoplasma Detection Kit (Lonza; catalog no. LT07-418), and all cell lines were confirmed to be negative for Mycoplasma. The identity of all cell lines was validated by short tandem repeat profiling, performed by University of Arizona Genetics Core or MD Anderson Cancer Center Cytogenetics and Cell Authentication Core.

We previously synthesized a human, codon-optimized, second-generation CAR with specificity against prostate stem cell antigen (PSCA) by cloning published scFv sequences (clone 2B3; refs. 16, 17) in-frame with the IgG2-CH3 domain (spacer), CD28 costimulatory domain, and the zeta (ζ) chain of the T-cell receptor (TCR) CD3 complex in an SFG retroviral backbone (Supplementary Fig. S1A; ref. 18). The SFG retroviral vector was provided by Dr. Cliona Rooney and was originally received from Dr. Richard Mulligan (19). We also previously constructed a second-generation mucin 1 (MUC1)-specific CAR containing 41BB as a costimulatory domain (Supplementary Fig. S1B; ref. 20).

To generate the FD vector, we used DNA 2.0 to design (ATUM) a codon-optimized sequence encoding the signal peptide and extracellular domain of the human Fas (CD95) protein (amino acids 1–173) with the restriction sites Xho1 and Sph1 incorporated up and downstream, respectively (Supplementary Table S1); this was custom synthesized by Integrated DNA Technologies (IDT). The FD DNA insert was incorporated into an SFG retroviral vector that contained the fluorescent marker mOrange linked by an IRES sequence downstream of the insertion site.

To generate the FD+15 construct, a codon-optimized sequence encoding IL15 was linked to the C-terminus of the FD sequence using a G-S linker. Similarly, to generate the FΔ+15 vector in which the IL15 component was functional, but the decoy was defective through the incorporation of two amino acid substitution (R86S and R87S) mutations previously published to abrogate binding to FasL (21). DNA sequences of all constructs are included in Supplementary Table S1.

Constructs encoding PSCA antigen, click beetle green luciferase CBG99 (22), and a previously described firefly luciferase mutant Akaluc (23) were generated by incorporating synthesized linear DNA fragments (IDT) encoding these proteins into an SFG retroviral vector backbone upstream of GFP tag linked by an IRES sequence.

Retroviral supernatant for all constructs was generated by transfection of 293T cells, as described previously (24). Briefly, 3.5 ×10⁶ 293T cells were plated in 100 mm tissue culture-treated dishes (BD Biosciences) in 10-mL complete IMDM. A day later, cells were transfected with PegPam-e, RDF, and the DNA construct was loaded into GeneJuice Transfection Reagent (Millipore; catalog no. 70967; ref. 25). Retroviral vector for GFP-Firefly luciferase (GFP_FFLuc) has been described previously (26). All virus supernatants were collected at 48 and 72 hours, pooled together, sterile filtered using 0.2-μm filters (Pall Corporation; catalog no. 4652) and stored at −80°C until transduction.

To generate CAR T cells, 1 × 10⁶ PBMCs were plated in each well of a non-tissue culture–treated 24-well plate that had been precoated with the CD3-specific antibody OKT3 (1 mg/mL; Ortho Biotech, Inc.) and CD28 (1 mg/mL; Becton Dickinson & Co.; catalog no. 555725). Cells were cultured in complete media (RPMI-1640 containing 45% Clicks medium (Irvine Scientific, Inc.; catalog no. 9195), 10% FBS and 2 mmol/L L-GlutaMAX), which was supplemented with recombinant human IL2 (50 IU/mL, NIH, Bethesda, MD) on day 1. On day 3 after OKT3/CD28 T-blast generation, 1 mL of retroviral supernatant was added to a 24-well non–tissue culture–treated plate precoated with recombinant fibronectin fragment (FN CH-296; Retronectin; Takara Shuzo; catalog no. T100B) and centrifuged at 2,000 × g for 90 minutes. OKT3/CD28-activated T cells (0.2 ×10⁶ per mL) were resuspended in complete media supplemented with IL2 (100 IU/mL) and then added to the wells and centrifuged at 400 × g for 5 minutes. To generate cells expressing both a CAR and FD, FΔ+15 or FD+15, activated T cells were transduced sequentially, first with the CAR construct (on day 3) and then with either FD, FΔ+15, or FD+15 on day 4. T cells used in the in vitro experiments involving live-cell bioluminescence imaging or in vivo experiments underwent a third transduction on day 5 with either CBG99, Akaluc, or GFP-FFLuc retroviral supernatant. Transduction efficiency was measured 3 days after the last transduction by flow cytometry.

We generated CAPAN-1 cell lines that expressed transgenic PSCA to ensure homogeneous antigen expression. To do this, PSCA-GFP retroviral supernatant was plated in a non–tissue culture–treated 24-well plate (1 mL/well), which was precoated with FN CH-296. CAPAN-1 cells (0.2 × 10⁶ per well) were added to the plates and then transferred to a 37°C, 5% CO₂ incubator. Transgene expression was analyzed by flow cytometry 1-week after transduction. Cells were subsequently sorted based on GFP expression to exclude nontransduced tumor cells using a MoFlo flow cytometer (Cytomation). Sorted cells were cultured for 2 weeks in order to obtain sufficient numbers to prepare a cryopreserved bank for all experiments. GFP expression was assessed by flow cytometer and confirmed to be 100% on the day of cryopreservation. K562 cells expressing PSCA (K562-PSCA) were also generated using the same procedure. 293T cells engineered to produce FD or FD+15 (293T-FD and 293T-FD+15 cells) were generated by transducing 293T cells with the FD or FD+15 vectors, and subsequent flow-cytometric sorting was based on mOrange expression to exclude nontransduced cells. Cells were cultured and cyrobanked after sorting as done for CAPAN-1-PSCA and K562-PSCA cells.

The following antibodies were used for T-cell phenotyping: CD3-APC (Beckman Coulter; catalog no. IM2467U), CD3-FITC (Beckman Coulter; catalog no. 6604625), CD4-Krome Orange (Beckman Coulter; catalog no. A96417), CD8-Pacific Blue (Beckman Coulter; catalog no. A82791), CD69-ECD (Beckman Coulter; catalog no. 6607110), CD45RO-APC (BD Biosciences; catalog no. 340438), CCR7-FITC (BD Biosciences; catalog no. 561271), CD25-FITC (BD Biosciences; catalog no. 347643), CD28-PC7 (BD Biosciences; catalog no. 561791), CD95-PC5 (BD Biosciences; catalog no. 561977). PSCA antigen and FasL expression on tumor cells was measured using anti-PSCA (Santa Cruz Biotechnology. Inc.; catalog no. sc-80654) and anti-human CD178-APC (BD Biosciences; catalog no. 564262), respectively. CAR molecules were detected using Goat anti-human F(ab’)2 antibody conjugated with AlexaFluor647 (Jackson ImmunoResearch Laboratories, Inc.; catalog no. 109-606-097). Cells were stained with saturating amounts of antibody (∼5 μL) for 20 minutes at 4°C, washed (PBS, Sigma-Aldrich; catalog no. D8537), and then acquired on Gallios flow cytometer (Beckman Coulter Inc). Analysis was performed using Kaluza Flow Analysis Software (Beckman Coulter Inc.) version 2.1. The general gating strategy used for the analysis of T-cell phenotype is illustrated in Supplementary Fig. S2A.

1 × 10⁶ cells for each cell line were seeded in a 24-well plate in 2-mL culture media containing 10 μg/mL of recombinant human IFNγ (PeproTech US; catalog no. 300-02). After 24-hour incubation, cells were collected for flow cytometry to assess the expression of Fas and FasL. Cells were incubated with the anti-human Fas and anti-human FasL mentioned above, and samples were acquired using Gallios flow cytometer (Beckman Coulter Inc.). Analysis was performed using Kaluza Flow Analysis Software (Beckman Coulter Inc.) version 2.1.

The cytotoxicity and specificity of engineered T cells were evaluated in a standard 4–6 hours ⁵¹Cr-release assay, using E:T ratios of 40:1, 20:1, 10:1, and 5:1. Effector T cells were coincubated in triplicate with target cells (293T, CAPAN-1, and CFPAC-1) labeled with ⁵¹Cr (PerkinElmer; catalog no. NEZ030005MC) in a V-bottomed 96-well plate. At the end of the incubation period at 37°C and 5% CO₂, supernatants were harvested, and radioactivity counted using a PerkinElmer 2470 Automatic Gamma Counter. The percentage of specific lysis was calculated as follows: % specific cytotoxicity = [experimental release (cpm) – spontaneous release (cpm)]/[maximum release (cpm) – spontaneous release (cpm)] × 100.

To measure T-cell expansion upon antigen stimulation, 1×10⁶ T cells were cultured with 1×10⁶ irradiated K562-PSCA or CAPAN-1-PSCA in the absence of exogenous cytokines for six days. Tumor cells were irradiated (100 Gy) to halt their expansion using Rad Source RS2000 Biological X-Ray Irradiator (Rad Source Technologies) before placing them in culture with T cells. Wild-type K562 cells, which do not express PSCA, were used as control (nonstimulated condition). T-cell counts were obtained by flow cytometer–based quantification—all cells in the wells were collected, labeled with a CD3-APC antibody to distinguish T cells from tumor cells, and mixed with CountBright Absolute Counting Beads (Invitrogen; catalog no. C36950) and 7-AAD (BD Biosciences; catalog no. 51-68981E) to exclude dead cells. Absolute T-cell counts were calculated according to CountBright beads manufacturer's recommended protocol. Sample acquisition was done using the Gallios flow cytometer and was halted at 2,500 beads. Data analysis was performed in Kaluza Flow Analysis Software version 2.1.

T cells were stimulated for 5 hours by coculturing with irradiated CAPAN-1-PSCA tumor cells at an E:T of 1:4 (1 × 10⁵ T cells and 4 × 10⁵ tumor cells). Stimulated T cells were collected and used for IsoPlexis Single-Cell Secretome analysis using a kit and protocol from the manufacturer (IsoPlexis United States; catalog no. 1000-4 PANEL-1001-4). Polyfunctionality was determined by calculating the frequency of T cells that secreted two or more cytokines included in the panel within the fraction of T cells that secreted at least one cytokine upon stimulation.

Transgenic Fas decoy production was measured by Fas ELISA (R&D Systems; catalog no. DY326) according to the manufacturer's instructions. Briefly, supernatant samples harvested from T cells activated in 24-well non-TC plates coated with recombinant PSCA antigen (CUSABIO; catalog no. CSB-YP018840HU) at indicated concentrations or irradiated K562-PSCA cells at indicated E:T. To measure the production of FD+15 fusion molecule, samples were used in soluble Fas and IL15 ELISAs (R&D Systems, catalog no. DY247) to measure the two components separately. To measure the production of effector cytokines by T cells, supernatant harvested at 48 hours after activation by K562-PSCA tumor cells at an E:T of 1:1 (1 × 10⁶ T cells and 1 × 10⁶ tumor cells) was used to perform a 13-plex multiplex assay (Millipore Sigma; catalog no. HSTCMAG28SPMX13) according to the manufacturer's instructions.

Interaction of FasL with the FD was measured using an ELISA-based assay. 200 ng/mL recombinant FasL (BioLegend; catalog no. 589406) was first incubated in different media conditions (fresh IMDM, conditioned media obtained from 293T cells, or conditioned media obtained from 293T-FD cells) at 37°C for 1 hour and subsequently used in a FasL ELISA (R&D Systems; catalog no. DY126). Reduced detection of FasL in 293T-FD conditioned media compared with fresh IMDM or 293T conditioned media was considered neutralization due to FD binding.

T-cell viability after overnight incubation in different media conditions containing recombinant FasL was used as a functional measure of FasL neutralization. In this assay, recombinant Fas ligand (200 ng/mL) was added to CAR PSCA T cells resuspended in 293T conditioned medium or 293T-FD conditioned medium. 293T conditioned medium supplemented with the human FasL-blocking antibody NOK-2 (BD Biosciences, catalog no. 556375) at a concentration of 1 μg/mL was used as an additional control. After overnight (18 hours) culture at 37°C, T cells were labeled with Annexin V–APC (BD Biosciences; catalog no. 550474) and 7AAD according to the manufacturer's protocol, and cell viability was monitored using a Gallios flow cytometer and assessed using Kaluza Flow Analysis Software version 2.1.

T-cell:tumor coculture experiments were performed to assess the in vitro antitumor activity of T cells. CAPAN-1-PSCA tumor cells (0.25×10⁶ cells) were plated in a 6-well tissue culture-treated plates in 3 mL of culture medium. A day later, 2.5 ×10⁴ CAR, CAR.FD, CAR.FΔ+15, or CAR.FD+15-modified T cells were added to tumor cells. For the coculture experiments involving Fas blockade, the human Fas antibody ZB4 (Millipore Sigma, catalog no. 05-338) was added at 10 μmol/L final concentration immediately after adding T cells to the tumor wells. Antitumor activity was monitored on days 3 and/or 6 using a flow cytometer to quantify the number of cells. Briefly, all cells in the wells were collected, labeled with a CD3-APC antibody to distinguish T cells from tumor cells, and mixed with CountBright Absolute Counting Beads and 7-AAD to exclude dead cells. Sample acquisition was done using the Gallios flow cytometer and was halted at 2,500 beads. Absolute numbers for residual viable tumor and T cells were calculated according to the CountBright beads manufacturer's recommended protocol. Fold changes in tumor and T-cell numbers were then calculated using initial (day 0) absolute counts as reference for all subsequent time points. Data analysis was performed in Kaluza Flow Analysis Software version 2.1. The general gating strategy used to analyze the coculture samples is illustrated in Supplementary Fig. S2B.

To assess the cytotoxicity of effector cells against luciferase-labeled tumor targets, 5×10³ tumor cells were seeded in 96-well white flat bottom tissue culture-treated plates (Corning) in 100 μL T-cell medium. 1 × 10³ effector cells (CAR T cells or NK-92 cells) were added, resuspended in 100 μL T-cell medium, for an E:T of 1:5 and a final volume of 200 μL per well. Three days later, D-Luciferin (PerkinElmer; catalog no. 122799) was added to each well (final concentration of 30 μg/mL), plates were incubated for 10 minutes at 37°C, and luminescence signals from residual tumor cells were captured using CLARIOstar microplate reader (BMG Labtech Inc). Luminescence signals from nontransduced T cells or nontransduced NK-92 cells were normalized to 100% and used as control conditions to compare other treatment groups.

To assess the bystander activity of the decoy and IL15 combination (FD+15) on T cells, a transwell-based assay was utilized in which 5 × 10⁵ CAR MUC1 T cells were added to the lower chamber of the transwell, and in the upper chamber, equal numbers of either CAR PSCA or CAR.FD T cells were stimulated with CAPAN1 tumor cells at E:T of 1:1. Transwell inserts (upper chambers) with 0.4 μm pores were used to prevent cell migration while allowing secreted proteins to pass through. To monitor if the FD+15 produced by stimulated CAR.FD+15 cells promoted the expansion of CAR MUC1 T cells, T cells in the lower chambers were quantified by Trypan Blue (Gibco, catalog no. 15250061) exclusion after three days of culture.

The bystander effects of FD+15 were measured on NK-92 cells by culturing 1×10⁶ NK-92 cells in RPMI medium, supernatant obtained from unmodified 293T cell cultures, or FD+15-containing supernatant collected from 293T.FD+15 cell cultures in the absence of exogenous cytokines in a 24-well plate. Four days later, cells were quantified by Trypan Blue exclusion.

To assess the in vivo activity of engineered T cells, 6- to 8-week-old female NSG mice (NOD.Cg-Prkdcscid IL2rgtm1Wjl/SzJ, Jackson ImmunoResearch Laboratories, Inc.) were injected with 5 × 10⁶ CAPAN-1-PSCA cells suspended in 1× PBS subcutaneously (s.c.) in the left flank. Tumor cells were engineered to overexpress PSCA to prevent early tumor-relapse due to the outgrowth of antigen-negative tumor cells in order to allow proper follow-up of antitumor T-cell response and persistence. Once the tumor reached a size of approximately 140 mm³ (∼4 weeks), animals were injected intravenously (i.v.) with 2 × 10⁶ GFP-FFLuc⁺ CAR PSCA, CAR.FD, CAR.FΔ+15 or CAR.FD+15 T cells. Tumor volume was monitored by caliper measurement and calculated using the formula: length × width × width/2. T-cell expansion and persistence were monitored using the IVIS Lumina In vivo Imaging system (Caliper Life Sciences) 10 minutes after injection (i.p.) with 100 μL of D-luciferin (15 mg/mL), and the images were analyzed using Living Image software (Caliper Life Sciences). Mice were euthanized once the tumor volume reached the protocol limit (1,500 mm³) or in the event of tumor ulcerations that grew >2 mm in diameter despite treatment or recurred. Mice were monitored for manifestations of toxicities by research personnel as well as BCM veterinary staff by measuring body weight and visual symptoms of illness (e.g., ruffled coat, hair-loss, and/or skin redness).

To assess the bystander effects of the decoy-producing T cells, 6- to 8-week-old female NSG mice (NOD.Cg-Prkdcscid IL2rgtm1Wjl/SzJ, Jackson ImmunoResearch Laboratories, Inc.) were injected with 5 × 10⁶ wild-type CFPAC-1 cells suspended in 1× PBS s.c. into the left flank. Once the tumor reached a size of approximately 200 mm³ (∼4 weeks), animals were injected i.v. with 2 ×10⁶ CBG99⁺ CAR MUC1 T cells together with 2 × 10⁶ Akaluc⁺ CAR PSCA, or CAR.FD, or CAR.FD+15 T cells. Tumor size was measured by weekly caliper measurement, and tumor volume (mm³) was calculated by calipers (length × width × width/2). T-cell expansion and persistence were monitored using the IVIS Lumina In vivo Imaging 10 minutes after i.p. injection with 100 μL of D-luciferin (15 mg/mL) to capture luminescence produced by the bystander (CAR MUC1) T cells. Four hours later, imaging was repeated with an i.p. injection of 100 μL solution of 5 mmol/L Akalumine-HCL (Tokeoni; Millipore Sigma; 808350) to capture the luminescence of CAR PSCA, CAR.FD, or CAR.FD+15 T cells.

All in vivo analysis was performed using Living Image software (Caliper Life Sciences, Inc.). Experiments were performed according to a protocol approved by the Baylor College of Medicine Institutional Animal Care and Use Committee.

Results are reported as mean ± SEM unless stated otherwise. All statistical analyses were performed using GraphPad Prism 8.3.0. Statistical significance between/among groups was determined using one-way ANOVA, two-way ANOVA, or unpaired t tests as indicated in the figure legends. P < 0.05 were considered statistically significant. P > 0.05 = nonsignificant (ns); *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.

The data generated in this study are available within the manuscript and its supplementary files or from the corresponding author upon reasonable request.

To target solid tumors expressing PSCA, we had previously engineered a second-generation CAR specific for human PSCA (Supplementary Fig. S1A; ref. 18). In this study, we efficiently transduced T cells from healthy donor PBMCs with gamma retroviral vectors expressing this CAR (Fig. 1A). These cells produced specific ⁵¹Cr release from PSCA-expressing target cells in short-term (5 hours) isotope release assays; as shown in Fig. 1B, antitumor activity against the PSCA⁺ pancreatic cancer cell lines CAPAN-1 and CFPAC-1 correlating with their PSCA expression levels (Supplementary Fig. S3) was observed with minimal activity against PSCA^– 293T cells.

To better understand the potential impact of the Fas/FasL pathway on CAR T cells, we first explored the dynamics of Fas expression during their manufacture. At baseline, approximately half of the unstimulated T cells in freshly isolated donor PBMC expressed Fas (Fig. 1C and D; Supplementary Fig. S4). Within 24 hours of activation with anti-CD3 and anti-CD28, virtually all T cells expressed Fas, and this was maintained for at least 15 days after retroviral transduction (Fig. 1C and D; Supplementary Fig. S4). Fas upregulation conferred upon T cells sensitivity to FasL-mediated apoptosis, as illustrated in Fig. 1E, where CAR T-cell viability (assessed by Annexin V and 7-AAD staining) decreased upon exposure to increasing concentrations of recombinant FasL. To understand the susceptibility of T cells to FasL expressed by tumor cells, we measured FasL on a range of tumor cell lines including pancreatic (CAPAN-1 and CFPAC-1), prostate, and breast cancer cell lines and found that at baseline, almost all expressed FasL, and this was further upregulated upon IFNγ exposure (Fig. 1F; Supplementary Fig. S5A). Fas expression on these cell lines was minimal/absent under the same conditions (Supplementary Fig. S5B). To test the impact of tumor-expressed FasL on CAR T cells, we performed cocultures of CAPAN-1-PSCA tumor cells with CAR PSCA T cells in the absence or presence of a Fas-blocking antibody (clone: ZB4, 10 μmol/L). Quantification of residual tumor cells 3 days after coculture revealed improved antitumor activity of CAR PSCA T cells in the presence of the blocking antibody (Fig. 1G). Taken together, these data suggest that interactions between Fas on CAR T cells and FasL expressed by tumor cells can suppress effector T cells, resulting in suboptimal tumor control (Fig. 1H).

To protect T cells from FasL-mediated suppression, we developed an FD molecule to prevent Fas engagement (Fig. 2A). We designed the FD as a secretable molecule so that it could also potentially block Fas signaling in bystander tumor-specific cells. Our retroviral vector encodes the extracellular portion of Fas linked with mOrange via an IRES sequence to facilitate transgene detection (Fig. 2B) and can be coexpressed with the CAR PSCA at high efficiency (Fig. 2B and C). To evaluate the impact of transgenic FD expression on CAR T cells, we analyzed our cell populations by flow cytometry. This revealed that CAR PSCA and CAR.FD T cells were similar with respect to CD4:CD8 distribution as well as expression of activation markers (e.g., CD69, CD27, and CD28) and memory markers (naïve/undifferentiated, CCR7⁺CD45RO^–, central memory, CCR7⁺CD45RO⁺, effector memory, CCR7^–CD45RO⁺, and effector/terminally differentiated, CCR7^–CD45RO^–; Fig. 2D).

To assess if FD expression affected cell expansion upon antigen exposure, we stimulated CAR and CAR.FD T cells with irradiated K562 cells engineered to express PSCA (K562-PSCA). We observed no significant difference (P = 0.96) in T-cell growth over the course of 6 days (Fig. 2E). The two T-cell populations also secreted comparable amounts of effector cytokines IFNγ and TNFα upon stimulation with K562-PSCA cells (Fig. 2F). In addition, CAR-mediated cytolytic activity to PSCA⁺ tumor targets CAPAN-1 and CFPAC-1 was not affected by FD expression in a ⁵¹Cr-release assay (Fig. 2G). Finally, to confirm that CAR.FD cells produced the decoy, T cells were stimulated with plate-bound recombinant PSCA, and 48 hours later, the supernatant was harvested for Fas quantification by ELISA.

As shown in Fig. 2H, CAR.FD T cells produced FD at levels that correlated with the magnitude (Supplementary Fig. S6A and S6B) and duration of stimulation (Supplementary Fig. S7), indicating that FD is likely to be secreted locally at the tumor site upon antigen stimulation. Although this stimulation dependency is unusual for transgenes, it was unsurprising based on our prior experience with the vector, which requires T-cell activation for the expression of secreted proteins (e.g., cytokines and chemokines; ref. 27). We also tested the stability of FD at physiologic temperature (37°C) by measuring FD in medium obtained from FD-transduced 293T (293T-FD), which suggested that the half-life of this soluble decoy molecule is approximately 12 days (Supplementary Fig. S8). Taken together, these results demonstrate the feasibility of engineering CAR T cells to secrete FD locally in the TME.

To evaluate the functional characteristics of the decoy and its impact on transgenic T cells, we first examined the FasL-neutralizing properties of FD using an ELISA-based neutralization assay. As shown in Fig. 3A, after 1-hour incubation, detection of recombinant FasL incubated in the 293T-FD conditioned medium was significantly reduced compared with incubation in control (293T) supernatant or fresh medium, suggesting sequestration of FasL by the decoy.

To investigate if the neutralizing effect of FD was sufficient to protect T cells from FasL-induced apoptotic death, CAR PSCA T cells were cultured with recombinant FasL and either control (293T) or FD-containing supernatant (293T-FD). Control supernatant supplemented with FasL-blocking antibody (NOK-2) was used as an additional control. After 24 hours of exposure, cell viability was assessed by Annexin V and 7-AAD staining, which showed that culture with the decoy-containing supernatant was associated with superior T-cell viability, similar to NOK-2 containing supernatant and supernatant lacking recombinant FasL, whereas viability was markedly reduced in control supernatant containing FasL (Fig. 3B and C). This improved T-cell survival translated to superior antitumor activity, as measured in a 6-day coculture experiment in which CAPAN-1 tumor cells were mixed with either CAR PSCA or CAR.FD cells at a low E:T (1:10) to recreate the challenging conditions of the TME. As shown in Fig. 3D, tumor cells expanded rapidly when left untreated and, although CAR PSCA T cells transiently controlled tumor growth between days 0 and 3, tumor cells eventually grew rapidly by day 6. In contrast, CAR.FD T cells exhibited enhanced antitumor activity, as indicated by fewer residual tumor cells on day 6. This improved the cytolytic activity of CAR.FD cells correlated with their enhanced survival/expansion compared with CAR PSCA T cells (Fig. 3E).

As an additional measure of T-cell effector function, we compared the polyfunctionality of CAR PSCA and CAR.FD T cells when stimulated by irradiated CAPAN-1-PSCA tumor cells using an IsoPlexis Single-Cell Secretome assay (Supplementary Fig. S9). Stimulated CAR.FD T cells demonstrated a higher degree of polyfunctionality, as indicated by the increased frequency of T cells secreting two or more effector cytokines within the population that secreted at least one cytokine following tumor stimulation (Fig. 3F). In summary, these results from our in vitro functional assessment of FD demonstrated that the decoy interacts with and abrogates FasL-induced apoptosis, resulting in improved T-cell survival and antitumor effects upon encounter with FasL-expressing tumors.

To investigate whether incorporating a growth-promoting cytokine signal could further augment the antitumor effects of FD-modified CAR T cells, we developed a new retroviral construct encoding FD and human IL15 cytokine fusion molecule (FD+15; Fig. 4A). CAR PSCA T cells were transduced to express coexpress FD+15 at levels comparable to FD alone (Fig. 4B and C). To assess the effects of FD+15 expression on T cells, we evaluated the phenotype of CAR.FD+15 cells by flow cytometry and found that it was similar to that of T cells expressing CAR only or CAR.FD (Supplementary Fig. S10). Effector cytokine production capacity of CAR.FD+15 was also similar to that of CAR PSCA and CAR.FD T cells (Supplementary Fig. Supplementary S11).

To confirm that the fusion molecule was secreted, we collected supernatant from CAR.FD+15 T cells cultured with or without recombinant PSCA antigen for 48 hours and measured both FD and IL15 by ELISA. As shown in Fig. 4D, T cells secreted both FD and IL15 upon activation, whereas minimal quantities were detected in the absence of activation. When stimulated with K562-PSCA cells, CAR.FD+15 T cells demonstrated superior expansion compared with CAR PSCA or CAR.FD cells (Fig. 4E), likely due to the presence of the IL15 fusion protein.

To assess if coexpression of FD+15 with the CAR PSCA further enhanced T-cell effector function upon exposure to FasL⁺ tumor cells, we compared the antitumor activity and expansion of CAR PSCA only, CAR.FD, and CAR.FD+15 engineered cells following coculture with CAPAN-1 cells (E:T of 1:10). As shown in Fig. 4F, tumor cell numbers increased rapidly when left untreated, and CAR PSCA T cells only transiently slowed tumor growth. Although CAR.FD T cells had improved antitumor effects compared with CAR PSCA cells, CAR.FD+15 T cells exhibited the greatest antitumor activity of all treatment conditions. This enhanced tumor-killing activity correlated with superior survival and expansion of CAR.FD+15 T cells compared with both PSCA-CAR and CAR.FD T cells (Fig. 4G).

To confirm the efficacy of FD+15 in an additional FasL⁺ tumor model, we assessed the antitumor activity of our engineered T cells against ASPC-1 pancreatic cancer cells using a luciferase-based cytotoxicity assay. As seen with CAPAN-1, FD and FD+15 expression enhanced the ability of CAR PSCA T cells to kill ASPC-1 cells during a 3-day coculture at E:T of 1:5 (Supplementary Fig. S12A). We also coexpressed the decoys with a CAR MUC1 in T cells and tested their activity against ASPC-1 and MDA-MB-468 (breast cancer) cells using the luciferase cytotoxicity assay. As shown in Supplementary Fig. S12B and S12C, CAR MUC1 T cells coexpressing FD eliminated more tumor cells than cells expressing only the CAR MUC1, and this was further improved with FD+15 coexpression.

Lastly, to evaluate if FD was effective in effector cells other than T cells, we expressed FD and FD+15 in NK-92 cells, which have been extensively evaluated in preclinical and clinical studies because of their antitumor properties and their potential as an off-the-shelf NK-cell therapy for cancer. As indicated by mOrange expression (Supplementary Fig. S13A), we were able to transduce these cells to express FD and FD+15. When cocultured with ASPC-1 and MDA-MB-468 cells, NK-92 cells expressing FD showed enhanced tumor-killing capacity, which was further strengthened by the expression of FD+15 when tested in the luciferase-based cytotoxicity assay (Supplementary Fig. S13B and S13C). Taken together, these results indicate that the FD+15 fusion molecule further potentiates the antitumor activity and survival/expansion of T cells exposed to FasL⁺ tumors.

To assess the in vivo activity of FD-engineered T cells, we utilized a xenograft model of pancreatic cancer in which NSG mice were engrafted with 5 ×10⁶ CAPAN-1-PSCA cells subcutaneously (s.c.) in the left flank followed by an intravenous injection (i.v.; single dose) of 2 ×10⁶ GFP-FFLuc–labeled T cells once the tumor reached approximately 140 mm³. An additional cytokine-only control for this experiment included T cells modified with a construct encoding FΔ+15, which is a fusion protein composed of mutant Fas incapable of binding to FasL and functional IL15. Cell expansion and persistence were measured by bioluminescence imaging whereas tumor volume was monitored by caliper measurement (Fig. 5A). To assess the safety profile of FD-engineered T cells in vivo, we monitored mouse weight over the course of the experiment and additionally routinely examined animals for evidence of ruffled coat, hair loss, and/or skin redness. As shown in Fig. 5B, mouse weight remained stable in all groups, irrespective of treatment and even during peak T-cell expansion (3–4 weeks after T-cell injection). No other manifestations of toxic side effects indicated by ruffled coat, hair-loss, and/or skin redness were noted, indicating the safety of the decoy. T cells localized to the tumor within 7 days of infusion in all animals except recipients of nontransduced (NT) T cells, as assessed by bioluminescence imaging (Fig. 5C). Compared with the CAR PSCA group, the CAR.FD and CAR.FΔ+15 groups showed improved antitumor activity (Fig. 5D and E). CAR.FD+15 T cells, which were protected from FasL-induced apoptosis and also received IL15 cytokine support, exhibited superior expansion and persistence compared with CAR PSCA and CAR.FD T cells (Fig. 5F), and this resulted in the more robust and longer lasting antitumor effects (Fig. 5D and E). In summary, results from our in vivo study indicate that FD-engineered CAR T cells localize to the tumor and mediate their cytotoxic potential at disease sites without evidence of systemic toxicity. Protection from the tumor FasL via the decoy combined with the growth-promoting effects of IL15 provided by the FD+15 fusion molecule resulted in greater T-cell proliferation and functional persistence at the tumor, leading to the most robust antitumor activity observed in the study.

To this point, our efforts focused on assessing the impact of FD expression on the transduced CAR T cells. However, the decoy molecule is a secreted compound that may also affect nonengineered bystander cells. To address this potential “bystander benefit,” we first evaluated if FD+15 secreted by CAR.FD+15 T cells could stimulate the growth of CAR MUC1 T cells in a transwell assay in which the CAR MUC1 T cells were in the lower chamber of the transwell and either CAR PSCA T cells or CAR.FD T cells stimulated with CAPAN-1 tumor cells were in the upper chamber (Supplementary Fig. S14A). Cell counts of bystander cells from two different donors obtained three days later demonstrated a significantly greater expansion in the wells containing CAR.FD+15 cells stimulated by CAPAN-1 tumor cells in the upper chambers (green bars), compared with the condition with activated CAR-only T cells in the upper chamber (Supplementary Fig. S14B and S14C). We also tested the bystander effects of FD+15 on NK-92 cells by culturing them in RPMI media without any supplements, supernatant obtained from nontransduced 293T cells, or supernatant obtained from 293T-FD+15 cells. As shown in Supplementary Fig. S15, after a 4-day culture period, a significant increase in the number of NK-92 cells was observed when cultured in 293T-FD+15 supernatant, compared with 293T supernatant, or media, indicating “bystander” stimulation of NK-92 cells by IL15 in the FD+15.

To evaluate the in vivo “bystander effect” of the decoys, we engrafted NSG mice with the pancreatic cancer cell line CFPAC-1 and subsequently treated these animals with a 1:1 mix (2×10⁶ of each) of CAR MUC1 T cells (which served as bystander cells) and either CAR PSCA, CAR.FD, or CAR.FD+15 T cells. To track the fate of each T-cell type in vivo, CAR MUC1 cells were cotransduced with CBG99 and CAR PSCA T cells were labeled with Akaluc, and tumor volume was measured by calipers (Fig. 6A). T-cell localization at the tumor was seen in all treated animals, with greater overall expansion of CAR PSCA/CAR.FD/CAR.FD+15-expressing cells (right) compared with bystander cells (left; Fig. 6B). The antitumor activity CAR MUC1in the animals receiving coadministered CAR.FD surpassed that seen in animals coinjected with CAR PSCA only T cells and was further improved when the CAR MUC1 T cells were delivered with CAR.FD+15 T cells (Fig. 6C). This enhanced antitumor response when coadministering CAR MUC1 T cells with CAR.FD or CAR.FD+15 T cells correlated with the expansion of CAR MUC1 T cells in the tumor (Fig. 6B and D). These data indicate that bystander cells were protected, which provided enhanced antitumor effects in vivo (Fig. 6C and E). We ascribe this benefit to enhanced CAR MUC1 T-cell activity since tumors resected 5 weeks after T-cell infusion had fewer residual MUC1⁺ tumor cells in mice treated with CAR MUC1+CAR.FD and CAR MUC1+CAR.FD+15 treatments versus those treated with CAR MUC1+CAR PSCA or left untreated (Supplementary Fig. S16). Taken together, these results support the tumor-localized bystander benefit of the decoy.

Redirecting the cellular immune response against cancer by engineering T cells to express CARs with specificities against tumor-associated antigens (TAA) has resulted in remarkable clinical success, particularly in patients with CD19⁺ malignancies. Unfortunately, solid tumors have proven more challenging to treat with CAR-engineered T cells due in part to inhibitory mechanisms deployed by the tumor to restrict T-cell persistence and limit T-cell cytolytic function. These mechanisms include upregulation of checkpoint molecules (e.g., PD-L1), production of suppressive cytokines (e.g., IL4, and TGFβ), and expression of death receptor ligands such as FasL, which is an immunomodulatory molecule that induces apoptotic cell death upon engagement with its receptor Fas, expressed by activated T cells (11, 28–31). In this study, we have shown the benefit of protecting T cells from tumor-induced FasL-mediated apoptosis using an FD. Using CAR PSCA T cells in a FasL⁺ pancreatic cancer model system, we demonstrate that coexpression of both the CAR and the decoy produces superior T-cell expansion and antitumor effects both in vitro and in vivo. Additionally, the provision of cytokine support from a bifunctional fusion molecule of FD and IL15 cytokine further improved T-cell proliferation and persistence, and potentiated antitumor activity in a FasL-rich TME. Finally, we showcase the paracrine immune-enhancing properties of the FD+15 decoy molecule in a second pancreatic tumor (CFPAC-1) model. In this model, we also demonstrate the beneficial effects conferred by the FD + IL15 to bystander tumor-specific effector T cells.

Given the detrimental impact of tumor-expressed FasL on adoptively transferred T cells, our group and others have previously explored genetic engineering strategies to render T cells resistant to both tumor FasL^– and activation-induced cell death (AICD) by ablating Fas in T cells. For example, Dotti and colleagues used siRNA to knockdown Fas to render EBV-specific T cells resistant to EBV-associated lymphoma-expressed FasL and demonstrated enhanced T-cell persistence and antitumor activity (32). Similarly, Ren and colleagues applied CRISPR/Cas9 gene editing to ablate Fas expression by CD19-targeted CAR T cells, resulting in reduced AICD death following exposure to CD19⁺ K562 chronic myelogenous leukemia tumor cells (33). However, the major issue with both these strategies is the risk associated with permanently eliminating Fas/FasL signaling in T cells, particularly given the crucial role FasL plays in maintaining T-cell homeostasis (9). An alternative to ablation is the use of a nonsignaling dominant-negative receptor, which acts as a sink for FasL. Indeed, Yamamoto and colleagues explored such a strategy and reported improved persistence and effector function of TCR- and CAR-modified T cells in FasL⁺ murine melanoma and human xenograft B-cell acute lymphoblastic leukemia models, respectively (30). Oda and colleagues expanded upon this approach by developing a chimeric “switch” receptor with the incorporation of the 41BB endodomain, thereby allowing T cells to harness and invert the apoptotic FasL signal to one that was costimulatory. This produced prosurvival signaling and metabolic changes in T cells, resulting in enhanced proliferation and antitumor activity in leukemia and pancreatic cancer models (34). Although these strategies effectively augmented T-cell activity in the TME, each was designed to benefit only the engineered cells and confer no protective advantage to adjacent effector cells. In contrast, our secreted FD was designed to provide not only autocrine protection but also paracrine protection (as illustrated in our CFPAC-1 xenograft model), thereby potentiating the benefit of the strategy.

The TME is replete with cells including TILs, NK cells, and NKT cells whose inherent effector potential is inhibited due to tumor-imposed immunosuppression or exhaustion (35, 36). Such phenomena have been reported by several groups including Woroniecka and colleagues, who documented high-level expression of molecules synonymous with T-cell exhaustion, including PD-1, LAG-3, TIGIT, and CD39, as well as diminished production of effector cytokines such as IFNγ, IL2, and TNFα by human glioblastoma–resident TILs versus matched peripheral blood T cells (37). TME modulation with recombinant immunostimulatory cytokines and checkpoint inhibitors has been explored to rejuvenate the effector potential of TILs (38–40). For instance, Klapper and colleagues demonstrated that administration of high-dose IL2 produced lymphocytosis in metastatic renal cell carcinoma patients with subsequent durable complete as well as partial responses (41). Similarly, checkpoint inhibitors that block suppressive molecules such as PD-1 and CTLA-4 have also been associated with clinical benefit (42, 43). For instance, Tumeh and colleagues reported increased T-cell infiltration and proliferation in metastatic melanoma patients treated with anti–PD-1 therapy, which directly correlated with reduction in tumor size (44). Although effective in producing clinical responses in patients with certain tumor types (e.g., metastatic melanoma, renal carcinoma, and non–small cell lung cancer), systemic therapies such as high-dose IL2 and checkpoint inhibitors often exhibit nontumor tropic drug-related adverse effects including thrombocytopenia, liver toxicity, pneumonitis, and other autoimmune-like phenomena that can result in patient mortality (45, 46). Thus, considering the pivotal role endogenous effector cells can play in mediating antitumor responses, we sought to design our decoy to be soluble in order to shield both engineered and bystander cells from FasL-mediated apoptosis. Furthermore, because the soluble decoy molecule is secreted only upon CAR engagement with its target on tumor cells, its benefit is concentrated in the TME, thereby mitigating off-tumor effects. Finally, to augment the proliferation and persistence of both engineered and bystander cells, we linked our decoy to IL15, which is a cytokine known to promote the growth and maintenance of effector cells such as memory CD8⁺ T cells and NK cells (47, 48). Indeed, the additive benefit of both shielding (from FasL) and supporting the proliferation of both engineered and bystander cells can be appreciated in the CFPAC-1 model, where we observed superior antitumor activity with CAR PSCA T cells engineered with FD+15 administered in combination with CAR MUC1 T cells. Although the animal models used in the study allowed us to assess the effects of the decoys in the context of human FasL⁺ tumors and human CAR T cells, one limitation is the lack of complete TME components. This incomplete xenograft TME prevented a full assessment of the broader effects our decoys could have on additional sources of FasL (e.g., CAFs and TAMs) as well as endogenous tumor-infiltrating effectors (e.g., NK cells and TILs).

The current study focused on the Fas/FasL axis using CAR-engineered cells directed to pancreatic cancer-expressed TAAs. However, our platform has the potential to be customized for different types of malignancies and extended to other suppressive molecules in the TME (e.g., PD-L1, TGFβ, and IL10). This approach could also be adapted to include other stimulatory cytokines like IL2, IL7, and IL21. Additionally, as all adoptively transferred cellular therapies are prone to tumor-driven immunosuppression, our approach could be applied to strengthen the clinical efficacy of a range of cellular therapies beyond CAR T cells such as TIL, cancer/testis antigen (e.g., NY-ESO-1, MAGE, and PRAME)-specific T cells, transgenic TCR T cells, and NK-cell therapies. Finally, one of the pitfalls of adoptively transferred CAR- and TCR-modified T cells is that the engineered cells usually target a single TAA, leading to the risk of tumor immune escape due to antigen loss/modulation, as has been reported in CD19-CAR T-cell recipients, in which >20% of relapses were due to loss of CD19 (49, 50). Due to its bystander effects, we hypothesize that our decoy strategy can indirectly mitigate the risk of such immune evasion by harnessing the antitumor activity of endogenous effector cells in the TME that can potentially recognize multiple tumor-expressed antigens beyond those recognized by the adoptively transferred cells. However, our strategy is not intended to and does not directly address the challenges posed by the heterogeneous nature of TAA expression in solid tumors, particularly immune evasion.

In nature, the Fas/FasL axis is designed to maintain homeostasis of immune effectors such as antigen-specific T cells by limiting hyperactivation and uncontrolled expansion (51). Thus, any engineering approach to modulate this pathway requires careful consideration, given the risk of systemic toxicity as a result of autonomous T-cell proliferation. Indeed, naturally occurring mutations of FAS have been shown to result in autoimmune lymphoproliferative syndrome (ALPS), which is characterized by uncontrolled lymphoproliferation leading to disorders including hemolytic anemia, thrombocytopenia, and neutropenia (52). However, in our in vitro as well as in vivo assessment of the FD-engineered T cells, we saw no evidence of autonomous T-cell proliferation. Indeed, both T-cell proliferation and persistence were dependent on the presence of target antigen and upon removal. Moreover, in our in vivo analysis, T cells declined in number gradually with decreasing antigen availability as the tumor shrank. Although our preclinical assessment suggests a positive safety profile, transient expression of the FD+IL15 via transfection or incorporation of a suicide switch (e.g., inducible caspases) to eliminate engineered T cells in the event of unexpected toxicity could be considered during clinical translation (53).

To summarize, this study demonstrates the feasibility of engineering T cells to express a secreted FD that can enhance CAR T-cell expansion/persistence and antitumor activity when targeting FasL⁺ tumors. The potency of the engineered CAR T cells was further increased by incorporating IL15 into the FD, resulting in a Fas decoy and IL15 fusion molecule, without adversely affecting T-cell phenotype, antigen-specificity, and dependency on the target antigen to sustain T-cell activity. In addition, our soluble decoy strategy exhibits bystander effects and, thus, bears the potential to support survival as well as tumor-killing activity of endogenous T and other effector (e.g., NK, NKT, and γδ) cells in the TME.

P. Bajgain reports grants from NIH, the American Cancer Society, Cancer Prevention and Research Institute of Texas Scholar Award, V Foundation for Cancer Research, Elsa U. Pardee Foundation, and Adrienne Helis Malvin Medical Research Foundation during the conduct of the study; in addition, P. Bajgain has a patent application pending to Baylor College of Medicine. P. Lulla reports personal fees from Karyopharm outside the submitted work. H.E. Heslop reports grants from CA125123 during the conduct of the study; other support from AlloVir, Marker Therapeutics, grants and personal fees from Tessa Therapeutics, grants from Athenex, other support from Fresh Wind Biotherapies, personal fees from GSK and Kiadis outside the submitted work. J. Vera reports grants from NIH, American Cancer Society, Cancer Prevention and Research Institute of Texas, V Foundation for Cancer Research, Elsa U. Pardee Foundation, and Adrienne Helis Malvin Medical Research Foundation during the conduct of the study; personal fees from Marker Therapeutics, Marker Therapeutics, AlloVir, and Marker Therapeutics outside the submitted work; in addition, J. Vera has a patent application pending to Baylor College of Medicine. A.M. Leen reports grants and other support from AlloVir and other support from Marker Therapeutics outside the submitted work; in addition, A.M. Leen has a patent for submitted patents related to novel molecules and methods of manufacture of T cells, all owned by Baylor College of Medicine pending. No disclosures were reported by the other authors.

P. Bajgain: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. A.G. Torres Chavez: Data curation, formal analysis, investigation. K. Balasubramanian: Data curation, investigation. L. Fleckenstein: Data curation, investigation. P. Lulla: Supervision. H.E. Heslop: Resources, funding acquisition. J. Vera: Conceptualization, resources, supervision, funding acquisition, methodology. A.M. Leen: Conceptualization, resources, supervision, funding acquisition, methodology, writing–review and editing.

This study was supported by NIH grant P01 CA094237, NIH grant P50 CA126752, NIH grant P50 CA186784, NIH grant P30 CA125123, American Cancer Society Mentored Research Scholars grant MRSG-14-197-01-LIB, Cancer Prevention and Research Institute of Texas Scholar Award RR170024, V Foundation for Cancer Research grant T2016-006, Elsa U. Pardee Foundation, and the Adrienne Helis Malvin Medical Research Foundation. We are thankful to Dr. Malcolm Brenner (project support, scientific advice, and manuscript revision), Dr. Cliona Rooney (project support, scientific advice, and manuscript revision), Dr. Norihiro Watanabe (scientific advice and flow cytometry technical assistance), Walter Mejia (illustrations/schematics and formatting of figures), and Dan Fick (manuscript revision). We also thank the Texas Children's Hospital Small Animal Imaging Facility, Texas Children's Hospital Flow Cytometry Core Laboratory, and the shared resources support from the Dan L. Duncan Comprehensive Cancer Center (P30CA125123).

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.