Constitutive Signaling from an Engineered IL7 Receptor Promotes Durable Tumor Elimination by Tumor-Redirected T Cells

Adoptive immunotherapy using T cells modified with chimeric antigen receptors (CAR) has achieved remarkable clinical efficacy against refractory leukemia (1) and lymphoma, but challenges remain in translating these successes to solid tumors. Substantial expansion and persistence of adoptively transferred T cells are necessary for durable antitumor efficacy (2). Of the 3 signals required for optimal T-cell activation and expansion (3), CAR activation can recapitulate signal 1 [T-cell receptor (TCR) activation] and signal 2 (costimulation) but cannot sustain a positive signal 3 derived from immunostimulatory cytokines that are scarce in tumor microenvironments. In xenograft tumor models, signal 3 has been supplemented with injections of cytokines such as IL2 to augment antitumor activity, without major adverse effects (4). However, systemic administration of cytokines to patients with cancer has caused significant toxicity (5–7). Alternative approaches such as genetic modification of T cells to secrete or trans-present cytokines (8) carry a risk of severe adverse events, including neurotoxicity and cytokine release syndrome from systemic accumulation of secreted cytokine (9), whereas T cells that overexpress cytokine receptors do not eliminate the need for exogenous cytokine (10). Therefore, a strategy to safely deliver cytokine signals to CAR T cells remains elusive.

Here, we present a strategy to selectively provide signal 3 to T cells with a constitutively active IL7 cytokine receptor (C7R). This novel chimeric receptor provides signal 3 without the requirement for exogenous agents or the nonspecific bystander T-cell activation caused by forced expression of transgenic cytokines. The growth and survival of C7R-expressing CAR T cells remain antigen dependent, but in the presence of tumor, these cells have superior antitumor activity in multiple model systems.

We chose to focus our attention on IL7, because this cytokine bolsters the persistence of tumor-specific T cells (11), and T cells genetically modified to either secrete IL7 or overexpress the IL7 receptor (in conjunction with administered IL7) display enhanced antitumor efficacy in preclinical models (10, 12). Recently, constitutively active IL7 receptors have been reported to transmit IL7 signaling without the need for ligand or the common receptor gamma chain (γ_c), as a result of IL7Rα homodimerization due to cysteine and/or proline insertions in the transmembrane domain (13, 14). Once the homodimer is formed, cross-phosphorylation of JAK1/JAK1 activates STAT5 (15), a core signaling node downstream of IL7. To discover whether this class of receptors could produce a consistent signal 3 to complete the three-signal requirement for optimal CAR T-cell activity, we selected a constitutively active IL7 receptor variant (IL7R*) that has significant STAT5 activation (16). To avoid additional activation of the receptor by external ligand and provide a means of detecting transduced cells, we replaced the native extracellular domain of the receptor with ectodomains derived from CD34. To learn whether ectodomain size factored into the efficiency of protein expression and function, we used ectodomains from Q8 (65 amino acids; ref. 17) and CD34 (259 amino acids). The Q8 ectodomain consists of a CD34 epitope mounted on top of a CD8 spacer, allowing detection by the anti-CD34 antibody clone QBEND10. Retroviral-mediated expression of the CD34–IL7R* and Q8–IL7R* constructs in healthy donor T cells revealed poor expression of the Q8–IL7R* fusion protein and suboptimal STAT5 activation (Supplementary Fig. S1). In contrast, CD34–IL7R* was robustly expressed in T cells and was functionally active. Therefore, we used CD34–IL7R*, henceforth referred to as C7R, for all subsequent experiments (Fig. 1A).

To determine the relative effects of C7R in CD4 and CD8 T cells, we separated the two subpopulations using antibody-coated magnetic beads, activated and transduced them, and cultured the T-cell subsets separately from each other. We found that C7R was readily expressed by both CD4 and CD8 T cells (Fig. 1B and C and Supplementary Fig. S2) and produced greater constitutive activation of STAT5 in T cells than a control construct consisting of a truncated CD34 (Δ34) molecule (ref. 18; Fig. 1D–G). Importantly, C7R did not promote antigen-independent expansion of CD4 and CD8 T cells in vitro (Fig. 1H and I). Although C7R-transduced cells persisted significantly longer in antigen- and cytokine-depleted conditions than control cells in vitro, the C7R population began to contract by 14 to 21 days, with all cells dying by day 70 after initiation of the persistence assay. This confirmed that C7R does not sustain autonomous T-cell expansion, an important property for safety.

To evaluate whether C7R could increase antitumor efficacy of CAR T cells, we treated GD2⁺ neuroblastoma cells with T cells expressing a GD2–CAR comprising a 14g2a scFv linked to a CD8α stalk and transmembrane domain, and a 41BB.ζ signaling endodomain (Supplementary Fig. S3A). 14g2a-based GD2–CAR T cells have shown a safe profile in clinical trials treating patients with neuroblastoma (19, 20), and although complete remissions have been achieved in selected patients, higher efficacy remains desirable. In comparing T cells expressing either the GD2–CAR alone or a bicistronic construct containing the GD2–CAR and C7R (GD2–CAR.C7R), we found that C7R did not induce significant differences in the memory subset composition or the CD4/CD8 percentages of GD2–CAR T cells (Supplementary Fig. S3B–S3D). Autonomous expansion of GD2–CAR.C7R T cells was also absent (Supplementary Fig. S4). Although C7R increased secretion of IFNγ and TNFα in GD2–CAR T cells after stimulation with LAN-1 tumors (Fig. 2A), this was not associated with any increase in the potency of T-cell killing during a 4-hour cytotoxicity assay (Fig. 2B). However, GD2–CAR.C7R T cells significantly outperformed GD2–CAR T cells when we measured their ability to maintain cytotoxicity and expansion after repeated encounters with tumors during in vitro sequential coculture killing assays (Fig. 2C). We found that GD2–CAR T cells failed by the third challenge, losing their ability both to expand and to eliminate tumor cells (Fig. 2D and E). In contrast, GD2–CAR T cells expressing C7R responded to all 3 sequential tumor challenges. To determine the relative contributions of increased proliferation versus reduced apoptosis to the improved cell expansion of GD2–CAR.C7R T cells, we used Cell Trace Violet labeling after the first coculture. Upon subsequent restimulation with tumor cells, we found that GD2–CAR.C7R T cells showed greater cell division than T cells expressing only the GD2–CAR (Fig. 2F and G). To assess whether C7R also reduced T-cell apoptosis, we used Annexin V and 7-AAD staining following the second tumor restimulation. Flow-cytometric analyses showed larger populations of Annexin V⁺/7-AAD⁺ GD2–CAR T cells compared with GD2–CAR.C7R T cells (Fig. 2H), demonstrating increased viability generated by C7R despite sequential tumor challenges. To further understand the molecular basis for these results, we used NanoString technology to perform gene expression analysis of GD2–CAR and GD2–CAR.C7R T cells after the second tumor restimulation (Fig. 2I and Supplementary Table S1). BCL2, which mediates the antiapoptotic effects of IL7 (21, 22), was one of the top genes upregulated by C7R in GD2–CAR T cells. We also found upregulation of cytolytic GZMA and downregulation of proapoptotic FAS and CASP8, which are involved in cellular apoptosis and activation-induced cell death (AICD; ref. 23). Therefore, C7R augments both proliferation and survival of GD2–CAR T cells to enhance their performance during sequential encounters with tumor cells.

We next tested the ability of C7R-enhanced GD2–CAR T cells to eradicate metastatic neuroblastoma in a xenograft model. We engrafted neuroblastoma cells in nonobese diabetic (NOD) severe/combined immunodeficient (SCID) Il2rg^−/− (NSG) mice by intravenous injection of the multidrug-resistant, NMYC nonamplified neuroblastoma cell line CHLA-255 modified to express firefly luciferase (CHLA-255 FFluc; ref. 24). Treatment with a low dose of GD2–CAR T cells 1 week after tumor engraftment increased median survival by 1 week, compared with control mice treated with T cells expressing an irrelevant CAR. Mice receiving T cells expressing C7R and a nonfunctional GD2–CAR with a truncated endodomain (GD2–CARΔ.C7R) had identical survival to the control mice (Fig. 3A and B). In contrast, disease was eliminated in mice infused with GD2–CAR.C7R T cells. In a parallel experiment in which T cells rather than CHLA-255 cells were GFP–FFluc transduced, we saw no expansion of GD2–CAR T cells with the limited dose that we infused but observed robust expansion of GD2–CAR.C7R T cells with accumulation of T-cell signal in the liver, a site of extensive neuroblastoma metastasis (Fig. 3C and D). These results demonstrated that GD2–CAR T cells could not persist against tumors in vivo whereas GD2–CAR T cells expressing C7R could proliferate and survive to mediate metastatic tumor clearance.

To investigate whether C7R could augment the performance of other CAR T cells, we coexpressed the molecule with an EPHA2–CAR intended to treat glioblastoma (25). U373 glioblastoma cells genetically modified with GFP–FFluc were injected intracranially into SCID mice. Seven days later, we administered intratumorally 10⁴ EPHA2–CAR T cells, a cell dose at which gliomas could not be eradicated based on our previous experience (unpublished data). Glioblastoma bioluminescence increased rapidly in mice treated with control T cells coexpressing C7R and a nonfunctional EPHA2–CAR (EPHA2–CARΔ.C7R), and no significant improvement in antitumor control was seen in mice receiving EPHA2–CAR T cells (Fig. 3E and F). In contrast, tumors were completely eliminated in mice infused with the low “stress” dose of EPHA2–CAR T cells when they coexpressed C7R (EPHA2–CAR.C7R), and these mice remained disease free at the conclusion of the experiment. When we repeated the experiment using EPHA2–CAR T cells also transduced with GFP–FFluc, bioluminescent signal from T cells expressing the CAR alone had largely dissipated 4 to 6 days after infusion. In comparison, although EPHA2–CAR.C7R T cells lacked the significantly greater expansion observed in the (extracranial) neuroblastoma models, there was a trend toward greater T-cell persistence as determined by area under the curve (AUC) comparison between EPHA2–CAR and EPHA2–CAR.C7R T cells (Supplementary Fig. S5).

Although C7R-expressing CAR T cells do not show sustained autonomous/antigen-independent growth and survival in vitro (Fig. 1H and I), the transgene is modified from a class of constitutively active IL7R variants expressed in 10% of pre–T-cell acute leukemias (26), potentially increasing the risk of unwanted expansion. However, expression of this mutant receptor per se is not oncogenic, and transformation occurs only in association with multiple other driver mutations (27). Moreover, these mutational events are oncogenic only in pre–T-cell leukemia and have not been reported in the mature, post-thymic T cells that are the target cell population for most adoptively transferred T-cell products, including our current approach. As an additional safety measure, however, we generated T cells coexpressing a clinically validated inducible caspase 9 (iC9) suicide gene (28) that can electively eliminate T cells. After double transduction with iC9 and GD2–CAR.C7R, T cells remained sensitive to iC9 signaling and underwent apoptosis in vitro within 24 hours of exposure to the chemical inducer of dimerization AP20187 (CID; Fig. 4A). We next asked if iC9 could efficiently remove GD2–CAR.C7R T cells in vivo after tumor regression. In order to evaluate T-cell activity and tumor growth simultaneously, we used a subcutaneous LAN-1 neuroblastoma model. Tumor cells were injected in the left dorsal flanks of NSG mice, and 8 days later we intravenously injected 1 × 10⁶ GD2–CAR.C7R T cells alone or doubly transduced with iC9. Control mice received GD2–CARΔ.C7R T cells. All T cells were cotransduced with GFP–FFluc for in vivo visualization. After 3 weeks, LAN-1 tumors outgrew in control mice, which were euthanized. We found that T cells coexpressing the GD2–CAR.C7R T cells and iC9 vectors demonstrated similar antitumor efficacy and in vivo expansion as GD2–CAR.C7R T cells alone (Fig. 4B and C). To model the T-cell deletion required to control toxicity after immunotherapy, we administered 3 doses of CID to mice beginning at 28 days after T-cell infusion within the same experimental approach. We observed a loss of T-cell bioluminescence signal (mean 93%) immediately after CID administration (Fig. 4D), and the signal remained at baseline 2 weeks later without tumor recurrence, at which time the experiment was terminated. These results confirmed that CAR T cells coexpressing C7R could be used together with iC9 if needed, without detriment to efficacy and permitting elective T-cell deletion.

Our results demonstrate that signal 3 in T-cell activation is essential for sustaining the activity of CAR T cells against solid tumors, and that constitutively active IL7 receptors can be used to provide signal 3 for the enhancement of adoptive immunotherapy. Under the stress of repeated antigen exposure that T cells will likely encounter within solid malignancies, we found that only GD2–CAR T cells coexpressing C7R were able to undergo multiple rounds of expansion and retain antitumor activity. The superiority of C7R-enhanced CAR T cells was shown for two different CAR and tumor models in vivo in which functional CAR T cells administered at ineffective doses could, if combined with C7R, sustain the ability to eradicate established tumors. In our metastatic neuroblastoma model, although we cannot detect tumors by bioluminescence at 1 week after inoculation, we and others have extensively confirmed that CHLA-255 xenograft tumors in immunodeficient mice are very aggressive, with 100% tumor engraftment (24, 29–32). Therefore, the continued lack of CHLA-255 bioluminescence denotes tumor clearance rather than failure of engraftment.

Our gene expression analysis revealed that C7R-promoted survival of GD2–CAR T cells during repeated tumor cell challenges is correlated with an increase in BCL2 transcription and reduced expression of FAS and CASP8. This would suggest that C7R exerts a broad antiapoptotic influence within CAR T cells to decrease their susceptibility to AICD and is consistent with a previous study that highlighted the role of IL7 in reducing the contraction of antigen-expanded T cells (33).

Although other studies have shown that the addition of signal 3 support will enhance CAR T-cell efficacy (12, 34), the delivery of IL7 signaling to T cells by C7R confers two distinct advantages. As a constitutively active cytokine receptor that does not respond to external IL7, C7R is uniquely designed to provide signal 3 only to modified T cells without affecting bystander immune cells. This avoids a potential mechanism for treatment-related toxicity associated with cytokine-secretion strategies for bolstering signal 3; for example, a recent clinical trial using CAR T cells that inducibly secreted IL12 was halted after a correlation was demonstrated between serum IL12 escalation and manifestation of adverse effects (9). Secondly, because IL7 signaling primarily promotes survival of T cells (35, 36) within lymphoreplete environments, synergistic activation from the CAR is required to drive proliferation and overall expansion. Consequently, C7R would be expected to focus CAR T-cell expansion at tumor sites in an antigen-dependent manner, and minimal proliferation should occur in tissues where the CAR is not activated. Other potential benefits from C7R include resistance to immunosuppressive agents such as TGFβ within the tumor microenvironment (37), as we have identified downregulation of TGFβRII in GD2–CAR.C7R T cells relative to GD2–CAR T cells alone (Supplementary Table S1).

The low toxicity seen in a patient with glioblastoma successfully treated with intraventricular IL13Rα2 CAR T-cell infusions (38), together with our observation that C7R functionally enhanced EPHA2–CAR T cells without substantially increasing expansion in our orthotopic glioblastoma model, suggests a low risk for adverse events if the CAR T-cell and C7R combination strategy was used for glioblastoma treatment. We found no evidence that C7R could induce antigen-independent proliferation, although the NOD/SCID and NSG models have obvious limitations for assessing the long-term fate of human T cells in vivo (39). The occurrence of xenogeneic graft-versus-host disease (xenoGVHD) was dependent on the dose of human cells and on the donor used. Because even incipient xenoGVHD may distort the apparent potency of a human cell therapy in mice, in this study we selected donors who produced a low incidence of xenoGVHD even at high T-cell doses (10⁷/mouse) and administered just ≤10⁶ T cells. Thus, we saw no xenoGVHD even if T cells expanded post injection; a similar threshold effect was seen after allogeneic hematopoietic stem cell transplantation in humans in which low-dose donor lymphocyte infusions caused less GVHD than high doses (40, 41). As added protection against potential concerns, however, we found that inclusion of a dimerizable iC9-mediated safety switch would allow deletion of CAR T cells expressing C7R.

Given the effective application of C7R to GD2–CAR T cells and EPHA2–CAR T cells against metastatic neuroblastoma and orthotopic glioblastoma, we believe our C7R molecule has the potential to enhance many other CAR T-cell candidates in testing today. It will be of interest to discover whether this same approach can increase the antitumor activity of other adoptive cell therapies for cancer, including those based on the specificities of native (42) or transgenic (43) TCRs as well as those exploiting the properties of natural killer (NK) cells and NK T cells (44), because all may respond to IL7 supplementation.

A cDNA encoding a mutant IL7Rα with a TTGTCCCAC insertion between base pairs 731 and 732 (IL7R*; ref. 15) was synthesized (Genscript). pSFG.C7R was generated by IN-Fusion (Takara) cloning using a XhoI and MluI-digested SFG vector backbone, the IL7R* cDNA, and the entire extracellular domain of CD34 (ΔCD34), which was available in our laboratory.

A cDNA encoding an N-terminal leader peptide, the GD2-specific 14g2a single-chain variable fragment (scFv), a CD8 stalk and transmembrane domain, and a 41BB.ζ endodomain was synthesized (Biobasic) and cloned by IN-Fusion (Takara) cloning into an SFG retroviral vector upstream of an internal ribosomal entry site (IRES) and truncated NGFR. For pSFG.GD2–CAR-2A-C7R, the GD2–CAR was subcloned into an SFG vector upstream of a 2A sequence and C7R. For pSFG.GD2Δ-CAR-2A-C7R, the 2A-C7R was cloned downstream of a nonfunctional GD2–CAR available in the laboratory composed of the 14g2a scFv, a short IgG1 exodomain spacer, a CD28 transmembrane domain, and a truncated CD28 endodomain (RSKRSRLL).

The generation of pSFG.EPHA2-CAR-2A-CD19t encoding an EPHA2-specific CAR consisting of the EPHA2-specific 4H5 scFv (25) and a 41BB.ζ endodomain, a 2A sequence, and CD19t is described elsewhere (unpublished data). pSFG.EPHA2-CAR-2A-C7R was generated by IN-Fusion (Takara) cloning replacing 2A-CD19t with 2A-C7R. For pSFG.EPHA2-CARΔ-2A-C7R, 2A-C7R was cloned downstream of a nonfunctional EPHA2–CAR (unpublished data).

The vector was generated as previously described (28).

All restriction enzymes were purchased from New England Biolabs, and the sequence of all cloned constructs was verified by Seqwright.

Peripheral blood mononuclear cells (PBMC) from healthy donors were obtained under a Baylor College of Medicine Institutional Review Board–approved protocol with informed consent obtained in accordance with the Declaration of Helsinki. When CD4 and CD8 T cells were individually evaluated, PBMCs were labeled with CD4 or CD8 magnetic selection beads (Miltenyi) and positively selected following the manufacturer's instructions. For T-cell activation on day 0, bulk or selected T cells were suspended in complete medium consisting of 90% RPMI-1640 (Hyclone), 10% FBS (Hyclone), and 1% glutamax (Gibco), and cultured in wells coated with OKT3 (CRL-8001; ATCC) and CD28 antibodies (BD Biosciences). IL15 and IL7 (Peprotech) were added 1 day after activation, and cells were retrovirally transduced on day 2 (45). T cells were used for experiments beginning at 9 to 12 days after OKT3 and CD28 activation. High retroviral transduction rates were achieved using retroviral supernatants collected 48 hours after 293T cell transfection, spindown of 1 mL of retrovirus/well in a 24-well retronectin coated plate followed by the addition of 0.1 million T cells. Viral supernatants generated by transient transfection were titered on human activated T cells by serial dilutions (Supplementary Fig. S6). Average infectious units (iu)/mL ranged from 0.6 to 2 × 10⁶/mL, with a resulting multiplicity of infection (MOI) of 6 to 21 when 1 mL of viral supernatants was used to transduce 1 × 10⁵ T cells.

Fluorochrome-conjugated antibodies were purchased from BioLegend (CCR7, CD45RO, NGFR), Abnova (CD34), Thermo Fisher (Life Technologies, CD8), eBioscience (CD4), Beckman Coulter (CD3), and BD Biosciences [CD8, CD4, CD3, CD34, STAT5 (pY694), Annexin V, 7-AAD]. For surface staining, cells were incubated with antibodies for 15 minutes at 4°C. Cells were acquired on a Beckman Coulter Gallos (10,000 events), and analysis was performed using FlowJo 10.0.7r2 (TreeStar). Proliferation analysis was performed using Flowjo 9.3.2 (TreeStar).

A 4-hour luciferase-based cytotoxicity assay was performed using the LAN-1 neuroblastoma cell line expressing GFP–Firefly luciferase (GFP–FFluc) based on a previously described protocol with minor modifications (46). Briefly, 2 × 10⁴ LAN-1 neuroblastoma cells were plated per well in a 96-well black plate (Corning). Twenty-four hours later, CAR T cells were added in varying effector-to-target (E:T) ratios. The viable number of LAN-1 cells per well was determined using a standard curve generated by serial dilution of LAN-1 cells. The formula used to calculate the percent cytotoxicity is as follows: (Cell number in untreated well − Cell number in assay well)/(Cell number in untreated well).

LAN-1 cells (0.5 × 10⁶) and T cells (1 × 10⁶) transduced with GD2–CAR or GD2–CAR.C7R were cocultured in a 24-well plate using fresh culture media without IL15 and IL7. Seven days later, cells were harvested for either FACS analysis or T-cell quantification by trypan-blue exclusion. CAR T cells were then replated at a 2:1 E:T ratio with fresh LAN-1 cells in fresh cell culture media to start the second and third tumor cocultures. At the conclusion of the third coculture, T cells were counted, and the coculture was analyzed by FACS.

To count antibody-stained cells, following a PBS wash 25 μL of counting beads (Life Technologies) and 2 μL of 7-AAD were added (for dead cell exclusion), and cells were immediately analyzed. Acquisition of events was based on collection of 3,000 counting beads.

T cells expressing GD2–CAR or GD2–CAR.C7R were cultured with LAN-1 cells using a 1:4 E:T ratio in a 24-well plate in complete culture medium without cytokines. Twenty-four hours later, supernatants were harvested. IFNγ and TNFα release was quantitated using ELISA kits (R&D Systems).

Transduced T cells were harvested and resuspended at 0.5 × 10⁶ cells/mL of complete medium without cytokines, then plated at 0.5 × 10⁶ cells per well in a 48-well tissue cultured plate. Twenty-four to 72 hours later, cells were harvested into a FACS tube and washed in cold flow buffer (PBS containing 5% FBS). Fix & Perm Reagent A (100 μL; Life Technologies) was added to the cells, gently vortexed, and incubated for 3 minutes at room temperature before 3 mL of ice-cold methanol was slowly added to the tube with constant vortexing. The tubes were then incubated for 10 minutes at 4°C. Afterward, the tubes were centrifuged, and the methanol was discarded, followed by another wash step with cold flow buffer. Fix & Perm Reagent B (100 μL; Life Technologies) and 5 μL of anti-STAT5 antibody were then added to the cells. The cells were gently vortexed, then incubated in the dark for 30 minutes at room temperature. Afterward, the cells were washed one more time with cold flow buffer and then immediately analyzed.

After a single coculture with LAN-1 tumor cells, GD2–CAR T cells and GD2–CAR.C7R T cells were labeled with Cell Trace Violet using a kit purchased from Thermo Fisher in accordance with the manufacturer's instructions. T cells were then rechallenged with tumor cells for 1 week before analysis. 7-AAD was added to exclude dead cells.

Cells were incubated with Annexin V antibody and 7-AAD and analyzed by flow cytometry. For experiments with iC9, the CID AP20187 was purchased from Takara Clontech.

LAN-1 and U373 cells were purchased from ATCC and used to generate LAN-1 GFP–FFluc and U373 GFP–FFluc. CHLA-255 and CHLA-255 FFluc were established and maintained as previously described (31). Routine Mycoplasma surveillance was performed using an enzyme-based assay (Lonza), and cells were authenticated within a year of the experiments described using STR profiling.

Total RNA was collected using the QIAzol reagent and the miRNeasy Micro Kit (Qiagen). Gene expression analysis used the Immunology Panel version 2 (NanoString) at the Baylor College of Medicine Genomic and RNA Profiling Core using the nCounter Analysis System. Data were analyzed using nSolver 3.0 software (NanoString). The GEO accession number for the dataset is GSE102567.

All animal experiments followed a protocol approved by the Baylor College of Medicine Institutional Animal Care and Use Committee.

Ten- to 14-week-old female NSG mice were implanted subcutaneously in the dorsal left flank with 2 million LAN-1 neuroblastoma cells in 100 μL of basement membrane Matrigel (Corning). Eight days later, mice were divided into groups based on tumor sizes such that the group tumor means and variances were similar. They were then injected intravenously with 1 million GD2–CAR T cells (10–12 days after PBMC isolation). Tumor sizes were measured twice a week with calipers, and the mice were imaged for bioluminescence signal from T cells at the same time points using the IVIS system (Xenogen Corporation) 10–15 minutes after 150 mg/kg d-luciferin (Xenogen) per mouse was injected intraperitoneally. The mice were euthanized when the tumor diameter was equal to or greater than 15 mm, or when the tumor exceeded 10% of the mouse body weight.

Ten- to 14-week-old female NSG mice were intravenously injected with 1 million Firefly-luciferase expressing CHLA-255 cells (CHLA-255 FFluc). Seven days later, mice were injected intravenously with 1 million GD2–CAR T cells (10–12 days after PBMC isolation). In parallel experiments, tumor growth or T-cell expansion was indirectly assessed by weekly bioluminescent imaging as described above. In the experiments where tumor growth was tracked, mice groups were not standardized for tumor burden because CHLA-255 FFluc luminescence was not detectable at the time of T-cell injection.

U373 glioblastoma cells (10⁵) were established intracranially in 8-week-old male ICR-SCID mice as previously described (25). Seven days after tumor engraftment, 10⁴ T cells were injected intracranially directly into tumors. Tumor growth or T-cell expansion was assessed by weekly bioluminescent imaging as described above. Mice in tumor growth experiments were standardized for tumor burden but not variances.

Graphs and statistics were generated using Prism 5.0 software for Windows (Graphpad Software Inc.). Measurement data are presented as mean ± SEM. The differences between means were tested using the paired two-tailed t test. For the mouse experiments, changes in tumor radiance from baseline at each time point were calculated and compared between groups using a two-tailed paired t test or Welch t test, when appropriate. One-way ANOVA and Bartlett's test for equal variances were used when appropriate to ensure similar tumor means and variances between groups. Survival determined from the time of tumor cell injection was analyzed by the Kaplan–Meier method, and differences in survival between groups were compared by the log-rank test.

T. Shum is co-author on a provisional patent for constitutively active cytokine receptors for cell therapy. B. Omer has ownership interest in a provisional patent for constitutively active cytokine receptors for cell therapy. Z. Yi has ownership interest in a patent on chemotherapy-resistant immune cells. M.K. Brenner has ownership interest in Marker Therapeutics and a patent, and is a consultant/advisory board member for Bluebird Bio. S. Gottschalk has ownership interest in patent applications in the fields of T-cell and/or gene therapy for cancer, and is a consultant/advisory board member for Merrimack. C.M. Rooney has ownership interest in Bluebird Bio, Marker, Viracyte, and a provisional patent for constitutively active cytokine receptors for cell therapy. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Shum, B. Omer, M.K. Brenner, C.M. Rooney

Development of methodology: T. Shum, H. Tashiro, R.L. Kruse, D.L. Wagner, K. Parikh, D. Liu, R. Parihar, P. Castillo, L.S. Metelitsa, C.M. Rooney

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Shum, B. Omer, H. Tashiro, D.L. Wagner, K. Parikh, Z. Yi, T. Sauer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Shum, H. Tashiro, D.L. Wagner, K. Parikh, Z. Yi, T. Sauer, D. Liu, P. Castillo, H. Liu, S. Gottschalk, C.M. Rooney

Writing, review, and/or revision of the manuscript: T. Shum, B. Omer, R.L. Kruse, T. Sauer, R. Parihar, P. Castillo, M.K. Brenner, L.S. Metelitsa, S. Gottschalk, C.M. Rooney

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Shum, S. Gottschalk

Study supervision: M.K. Brenner, C.M. Rooney

We would like to acknowledge David Rowley and Farrah Kheradmand for helpful discussion. We would also like to acknowledge Gianpiettro Dotti and Barbara Savoldo for generously providing constructs and for helpful discussion.

This work was supported by P01CA094237, 1RO1CA173750, an Alex's Lemonade Stand Foundation Reach grant as well as T32DK060445 and HL092332 that supported Thomas Shum. This work was funded in part by the Howard Hughes Medical Institutes Med into Grad Initiative. It was also supported by the Cytometry and Cell Sorting Core at Baylor College of Medicine with funding from the NIH (P30 AI036211, P30 CA125123, and S10 RR024574) and the expert assistance of Joel M. Sederstrom.

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.