Glypican-3–Specific CAR T Cells Coexpressing IL15 and IL21 Have Superior Expansion and Antitumor Activity against Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is the fourth most common cause of cancer-related death in the world (1). Lack of curative therapies for unresectable and/or metastatic disease, which occurs in the majority of newly diagnosed cases, results in dismal prognoses (1). Chimeric antigen receptor (CAR)–expressing T cells show clinical successes for the treatment of CD19-positive hematologic malignancies (2–6). In contrast, CAR T cells demonstrate only modest antitumor activity in patients with solid tumors, including HCC, in part due to their limited expansion and persistence (7–12). As the overall therapeutic efficacy of CAR T cells strongly correlates with their expansion and persistence in patients with CD19-positive malignancies (2, 4), translational approaches to enhance these properties may improve the antitumor efficacy of CAR T therapy in patients with HCC.

Human interleukin 15 (IL15) and IL21 are required for optimal T-cell activation, expansion, differentiation, and function (13, 14). These cytokines are absent in the HCC microenvironment, depriving T cells of survival signals (13, 14). In preclinical models of CD19⁺ malignancies, neuroblastoma or gliomas, CAR T cells coexpressing either IL15 or IL21 controlled tumors significantly better than CAR T cells alone (15–18). Additionally, IL15 and IL21 can synergistically promote antigen-dependent T-cell expansion and cytolytic function (19, 20). However, whether IL15, IL21, or their combined expression enhance the antitumor effector function of CAR T cells against HCC remains unknown. We previously systematically evaluated the in vitro and in vivo activity of T cells expressing CARs targeting glypican-3 (GPC3), an antigen expressed in over 70% of HCCs but not in nonmalignant tissues (21–27). We selected the GPC3-CAR containing a 4-1BB costimulatory endodomain—“GBBz”—for further clinical development, as this receptor induces favorable T_H1-polarized effector cytokine release upon tumor cell engagement and produced superior expansion and antitumor activity (27). Here, to determine the impact of IL15 and IL21 on T-cell survival, persistence, and antitumor activity in preclinical models of HCC, we coexpressed IL15, IL21, or both with the GBBz GPC3-CAR in T cells.

We demonstrated that GBBz GPC3-CAR T cells coexpressing IL15 and/or IL21 specifically and effectively killed GPC3-positive tumor cells, including HCC, in an antigen-dependent manner. Our results also indicated that constitutive transgenic expression of both IL15 and IL21 together enriched for less-differentiated T cells, which were better protected from apoptosis during repeated exposures to tumor cells. We showed that combined IL15 and IL21 expression maintained the expression of T-cell factor-1 (TCF-1), a transcription factor critical for T-cell development and survival. CRISPR-Cas9–mediated knockout of TCF7 (the gene encoding TCF-1) in IL15 and IL21 coexpressing GPC3-CAR T cells eliminated the improvement in proliferative capacity. Finally, GPC3-CAR T cells coexpressing IL15 and IL21 exhibited the most robust peak expansion and sustained persistence in vivo and mediate superior tumor control and survival of HCC tumor–bearing mice compared with either cytokine alone or controls. These results provided a strong rationale to evaluate IL15 and IL21 coexpressing GBBz CAR in patients with liver tumors (NCT02932956 and NCT02905188).

The HCC cell line Hep3B, rhabdoid tumor cell line G401, lung carcinoma cell line A549, and human embryonic kidney cell line 293T were obtained from the American Type Culture Collection and were cultured in Dulbecco's modified Eagle media with 10% fetal bovine serum (FBS) except for 293T cells, which were cultured in Iscove modified Dulbecco media with 10% FBS. Cells were kept in culture for less than 3 weeks for any given experiment. Cells were not reauthenticated in the past year. Cells were regularly tested and were negative for Mycoplasma contamination. The HCC cell line Huh-7 was a kind gift from Dr. Xiao-Tong Song (Baylor College of Medicine, Houston, TX) and was maintained in Dulbecco's modified Eagle medium, and its identity was confirmed at the Characterized Cell Line Core Facility at The University of Texas MD Anderson Cancer Center (Houston, TX) by short tandem repeat method. Cell lines were obtained between 2014 and 2015. A549-GPC3 cells were generated by transducing A549 cells with a retroviral vector encoding GPC3; Huh-7 firefly luciferase (Ffluc) cells were similarly generated using an eGFP.Ffluc construct (27).

The codon-optimized minigene encoding cytokines IL21, IL15, and the GPC3-CAR “GBBz” (27) linked with a T2A sequence and flanked by NcoI and MluI restriction enzyme sites was synthesized using the GeneArt system (Thermo Fisher Scientific). This fragment was subcloned into the pSFG retroviral vector yielding the 21.15.GBBz retroviral construct (Fig. 1A). 21.GBBz was generated from 21.15.GBBz by PCR amplifying genes encoding IL21 (F: ATCCTCTAGACTGCCATGGAACGGATC/R: CGTCCCTCGGCTCTAGAATCTTCG) and GBBz (F: TAGAGCCGAGGGACGGGGCT/R: ATGATGACGCGTTAATCATCTGGGGGG), followed by In-Fusion cloning (cat. #638920, Takara Bio) into the pSFG retroviral vector backbone according to the manufacturer's manual (Fig. 1A). 15.GBBz was similarly generated by amplifying the gene fragment encoding both IL15 and GBBz (F: ATCCTCTAGACTGCCATGAGAATCAGCAAGCCCC / R: ATGATGACGCGTTAATCATCTGGGGGG) and subsequent In-Fusion cloning (Fig. 1A). The 21.15 construct was produced by PCR amplifying the fragment encoding both IL21 and IL15 from 21.15.GBBz (F: ATCCTCTAGACTGCCATGGAACGGATC/R: CAGTGCGGCCGCTCAGGCCCTGCTGGTGTT) and in the process generating NcoI and NotI restriction enzyme sites in the introduced 5′ and 3′ overhangs. This fragment was cloned into a pSFG plasmid bearing the orange monomeric derivative of DsRed fluorescent protein (28). Sequencing was performed following each cloning step to confirm the correct alterations were made (Epoch Life Science). A CD19-specific CAR (FMC63 scFv, CD28, and 4-1BB costimulatory domains, CD3ζ signaling endodomain) was used as negative control (29).

Transient transfection of HEK 293T cells with plasmids for GPC3-CAR constructs, RDF plasmid encoding the RD114 envelope, and PegPam3 plasmid encoding the MoMLV gag-pol was used to generate retroviral supernatants as previously described (30). OKT-3/CD28 mAb–coated plates were used to stimulate human peripheral blood mononuclear cells isolated from healthy volunteer donors (Gulf Coast Regional Blood Center). After 48 hours, cells were washed and replated in complete RPMI medium and maintained in incubators at 37°C with 5% CO₂ (27).

Cytotoxicity of GPC3-CAR T cells was assessed as described previously (27) using a standard 4-hour chromium 51 (⁵¹Cr) release assay. Briefly, target cells were labeled with ⁵¹Cr for 1 hour followed by incubation with effector cells for 4 hours at 37°C using multiple effector-to-target (E:T) ratios. Cell culture supernatants were collected, and radioactivity was measured in a gamma counter (PerkinElmer).

The Human IL15/IL21 ELISA MAX Deluxe kit (BioLegend) was used to measure IL15 and IL21 concentrations according to the manufacturer's instructions. Briefly, 0.5 × 10⁶ resting CAR T cells were cultured in the presence or absence of Huh-7 cells at a 1:1 ratio. Cell culture supernatants were collected at 72 hours, centrifuged, and frozen until the time of assay. Multiplex cytokine/chemokine immunoassays were performed as previously described using the MILLIPLEX MAP human cytokine/chemokine magnetic bead kit (EMD Millipore, cat. #HCYTOMAG-60K) according to the manufacturer's instructions (27).

To assess the production of IL1b and IL6, monocytes negatively selected from human PBMCs by Pan monocyte isolation kit (Miltenyi Bio 130-096-537 kit) were plated at 2.5:2.5:1 tumor, CAR T-cell, and monocyte ratio, respectively. Monocyte purity was confirmed by FACS and was above 95% for each coculture experiments. Supernatants were collected at 20 hours and evaluated with IL1b and IL6 ELISA kits (BioLegend, cat. #437005 and R&D Systems, cat. #D6050, respectively) according to the manufacturer's manual.

GPC3-CAR expression was detected using the anti-F(ab)₂ Alexa Fluor 647-conjugated antibody (Jackson ImmunoResearch) and anti-goat IgG₁ isotype control (Jackson ImmunoResearch). The following antibodies were used for T-cell phenotyping analyses: anti-CD4-APC/Fire 750 (BioLegend), anti-CD8-V500 (BD Biosciences), anti-CD45RO-PE/Cy7 (BioLegend), anti-CD62L-AF488 (BioLegend), anti-CD19-PerCP/Cy5.5 (CCR1; BioLegend), anti-CD3-PE (BD Biosciences), anti-CD279-PerCP/Cy5.5 (PD-1; BioLegend), anti-CD223-PE/Cy7 (LAG-3; BioLegend), and anti-CD366-BV421 (TIM-3; BD Biosciences). Anti-bovine IgG (Sigma-Aldrich) was used to block nonspecific binding of other murine antibodies following CAR staining. Flow cytometry assessment was performed on either an LSR-II (BD Biosciences) or iQue Screener PLUS (Intellicyt Corporation). Results were analyzed using FlowJo software (FlowJo). To detect intracellular TCF-1 expression, CAR T cells were first stained for surface expression of CAR, CD4, and CD8 using the antibodies listed above, followed by staining with anti-TCF1-PE (TCF7, BioLegend) used in conjunction with the True-Nuclear Transcription Factor Buffer Set (BioLegend) according to the manufacturer's instructions.

In vitro expansion and persistence were assessed by repeatedly coculturing CAR T cells with fresh Huh-7 tumor cells every 3 to 4 days at a 1:1 ratio. At the end of each coculture interval, CAR T-cell counts were measured by flow cytometry staining for the CAR and using CountBright beads (ThermoFischer; cat. #C36950) according to the manufacturer's instructions. Annexin V staining was performed to assess CAR T-cell viability 2 days after each stimulation using the ApoScreen Annexin V Apoptosis Kit (SouthernBiotech; cat. #10010-09) according to the manufacturer's instructions.

CAR T cells were sorted using a Sony SH800Z instrument (Sony Biotechnology) and expanded for 1 week in complete RPMI supplemented with 100 units/mL penicillin, 100 μg/mL streptomycin, and 0.25 μg/mL amphotericin B (Thermo Fisher Scientific). Sorted CAR T cells were cocultured with tumor cells at a 2:1 (E:T) ratio for 3 days. To confirm the absence of tumor cells prior to RNA extraction, cocultured cells were analyzed by flow cytometry and the absence of residual tumor cells was confirmed. RNA was extracted using the RNeasy Mini Kit (Qiagen; cat. #74104) as per the manufacturer's protocol and measured at the Genomic and RNA Profiling Core at Baylor College of Medicine (BCM) using the nCounter Analysis System (NanoString Technologies) and the predefined nCounter Human Immunology V2 panel (cat. #XT-CSO-HIM2-12). Gene-expression data were normalized and analyzed using nSolver software (NanoString Technologies). Benjamini–Hochberg correction was used for multiple comparisons.

We screened 6 TCF7-specific sgRNAs and based on their efficiency to knock out TCF7 (resulting in diminished expression of TCF1, the protein encoded by the gene) measured by FACS selected 2 sgRNAs targeting the TCF7 gene locus (target sequences GGGGTCCACTTACCAGCGGGGG and GAGCTCGGCCAGGTCGCGCTCGG) for subsequent experiments. sgRNA design and TCF7 gene disruption in primary activated T cells were performed as previously described (31).

All mice used in this study were maintained at the Small Animal Core Facility of Texas Children's Hospital and handled under protocols approved by the BCM Institutional Biosafety Committee and Institutional Animal Care and Use Committee. Tumor cells or CAR T cells were diluted in 100 μL normal saline and were injected via indicated routes.

In vivo antitumor activity of infused CAR T cells and mouse survival were evaluated in murine HCC xenograft models as described previously (27) with modifications. Briefly, 12-week-old female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG, The Jackson Laboratory) mice were injected intraperitoneally (i.p.) with 2 × 10⁶ Ffluc⁺ Huh-7 cells or 5 × 10⁶ Ffluc⁺ G401 cells followed by 0.5 × 10⁶ or 2 × 10⁶ CAR T cells in 100 μL normal saline intravenously (i.v.) via tail vein 1 or 2 weeks later, as indicated. Mice were assessed daily, and tumor bioluminescence was measured after i.v. injection of luciferin using the IVIS Lumina III imaging system (PerkinElmer).

To evaluate CAR T cell in vivo proliferation and persistence, mice were injected with 2 × 10⁶ Huh-7 cells i.p. followed 2 weeks later by i.v. injection of 2 × 10⁶ CAR T cells coexpressing an optimized Ffluc (32). Mice were imaged every other day following CAR T-cell injection to monitor expansion. Blood and spleens were collected on days 15 and 18, respectively, and evaluated for the presence of CAR T cells by flow cytometry. All animals were sacrificed according to institution standards prior to harvesting spleens. Cells were stained for mouse CD45 using anti-mouse CD45-PE or PerCP/Cy5.5 (both BioLegend) and human CD4, human CD8, and the GPC3-CAR as described above. IL15 and IL21 in blood plasma were measured using the MILLIPLEX MAP human cytokine/chemokine magnetic bead kit (EMD Millipore).

To measure the ability of inducible caspase-9 (iC9) to eliminate IL15- and IL21-producing cells, the construct iC9.21.15.NGFR was generated by infusion cloning. In brief, the fragment of IL21 and IL15 was amplified from 21.15.GBBz and fused in frame with the amplicons of iC9 (33) and the truncated Neural Growth Factor Receptor (NGFR; ref. 34). T cells were cotransduced with GBBz, iC9.21.15.NGFR, and Ffluc to track their in vivo expansion and persistence with bioluminescence imaging. The chemical inducer of dimerization AP20187 (Clontech Laboratories; cat. #635069) was dissolved according to the manufacturer's instructions given at 50 μg/mouse dose i.p. in 100 μL solution on days 8 and 10 after adoptive transfer.

Data were summarized using descriptive statistics. One-way or two-way ANOVAs adjusted for donor effect followed if applicable by pairwise comparisons between groups were carried out. Response variables were log-transformed, if necessary, to achieve normality. Analysis was performed using SAS version 9.4. P values < 0.05 were considered statistically significant.

We generated a set of CAR constructs based on our previously optimized GBBz GPC3-CAR (27) with additional sequence(s) for human IL21 and/or IL15 (Fig. 1A) using the clinically validated Moloney murine leukemia virus–derived SFG retroviral vector backbone. After transduction, all constructs were stably expressed by human peripheral blood T cells, with constructs containing IL21 (21.GBBz and 21.15.GBBz) demonstrating slightly lower overall transduction efficiency compared with the GBBz construct (P < 0.001; Fig. 1B).

To measure IL15 and IL21 production, supernatants were collected from GPC3-CAR T cells cultured with and without GPC3-positive tumor cells and evaluated by ELISA. IL15 and IL21 were indeed secreted by CAR T cells engineered to express the corresponding genes at baseline (Fig. 1C and D). Following CAR stimulation by GPC3-positive HCC cells, IL15 and IL21 production increased significantly from T cells coexpressing the corresponding transgenes (P < 0.001), and IL21 concentrations remained significantly higher than IL15 (P < 0.001). Whereas T cells expressing 15.GBBz produced significantly more IL15 than 21.15.GBBz T cells at baseline (P < 0.01) and after stimulation (P < 0.001), IL21 production did not differ between 21.GBBz and 21.15.GBBz groups in either condition. Given that IL15 and/or IL21 could potentially induce antigen-independent T-cell proliferation, we evaluated the ability of each T-cell group to maintain autonomous growth. Whereas all groups underwent an initial burst of proliferation after transduction, in the absence of antigen stimulation, no viable CAR T cells remained in any group after 50 days (Supplementary Fig. S1).

In summary, transduced T cells stably expressed GPC3-CAR constructs and produced significant quantities of one or both cytokines, as appropriate, without evidence of antigen-independent autonomous growth.

We next explored whether IL21 and/or IL15 coexpression affects the efficacy and/or specificity of GPC3-CAR–mediated tumor cell killing using a chromium-51 release assay (27). T cells expressing any of the four GPC3-CAR constructs specifically and effectively lysed GPC3-positive tumor cells (Huh-7, Hep3B, G401, A549-GPC3) in an antigen-dependent manner regardless of IL21/IL15 coexpression (Fig. 2A).

To evaluate an additional measure of T-cell activation, we determined the T_H1 and T_H2 effector cytokine production profiles of GPC3-CAR T cells following coculture with GPC3-positive or -negative target cells. GPC3-negative A549 cells did not induce significant effector cytokine production, suggesting that the GPC3-CARs in this study do not trigger consequential tonic signaling (Supplementary Fig. S2). CAR engagement by GPC3-positive Huh-7 cells specifically induced cytokine production by GPC3-CAR T cells but not by control groups (Fig. 2B and Supplementary Fig. S2). IL15 coexpression caused a significant decrease in IL13 production compared with T cells expressing constructs without IL15 (P < 0.001; Fig. 2B). A similar decline was observed for GM-CSF in 15.GBBz but not in 21.15.GBBz T cells (P = 0.0052). IL21 coexpression significantly enhanced IL2 production compared with groups lacking IL21 (21.15.GBBz vs. GBBz, P = 0.0047; 21.GBBz vs. GBBz, P = 0.0016).

Cytokine release syndrome (CRS) and neurotoxicity are serious and potentially lethal side effects of CAR T-cell therapies and are primarily mediated by IL1 and IL6 produced by monocyte/macrophages (35). These myeloid cells can be influenced by TNFα, IFNγ, and GM-CSF, which are predictors of severe CRS (36). In coculture systems, we detected no difference in IL1 and IL6 produced by monocyte/macrophages cells between GPC3-CAR T-cell groups, suggesting that the coexpression of IL15 or IL21 was unlikely to increase the risk of CRS or neurotoxicity (Supplementary Fig. S3).

CD4⁺ and CD8⁺ T cells can produce different effector cytokines (37), and CD8⁺ T-cell homeostasis is supported by IL15, which is expressed in 15.GBBz and 21.15.GBBz T cells. To examine if the differences in effector cytokine production were related to differences in CD4/CD8 composition, we evaluated this parameter at baseline and following tumor cell engagement (Fig. 2C). We found that combined expression of IL15 and IL21 increased the CD8⁺ GPC3-CAR T-cell population (with a corresponding decrease in the CD4 population) versus GBBz alone at baseline (P = 0.0436). Following two rounds of stimulation, the CD8⁺ subset was enriched in 15.GBBz and 21.15.GBBz T-cell groups (P = 0.0478 and P = 0.0078, respectively; Fig. 2C and Supplementary Fig. S4A).

Overall, GPC3-CAR T cells demonstrated effective GPC3-specific short-term cytotoxic activity in vitro regardless of cytokine coexpression. The cells underwent IL15- and/or IL21-specific changes in both cytokine production profile and CD4/CD8 T-cell phenotype distribution that could have benefited their antitumor efficacy.

Limited in vivo expansion is a major barrier for effective immunotherapy against solid tumors. To test the proliferative capacity of GPC3-CAR T cells, we repeatedly exposed them to fresh tumor cells in vitro every 3 to 4 days in the absence of exogenous cytokines. After the second round of stimulation with fresh HCC cells (day 7), T cells expressing GBBz, 15.GBBz, and 21.15.GBBz began to expand to higher numbers than 21.GBBz T cells and control groups (21.15.GBBz/15.GBBz vs. 21.GBBz, P < 0.001; GBBz vs. 21.GBBz, P = 0.0486; Fig. 3A). By the end of the third stimulation (day 10), 21.15.GBBz and 15.GBBz T cells expanded significantly more than the GBBz group (15.GBBz vs. GBBz, P = 0.0013; 21.15.GBBz vs. 15.GBBz, P = 0.0243), suggesting a critical role for IL15 in enhancing expansion. After the fourth stimulation (day 14), only 21.15.GBBz T cells continued to proliferate, yielding significantly higher cell numbers than all other groups (P < 0.001).

To explore which factors may have contributed to the increased proliferative capacity of 21.15.GBBz T cells, we measured the subset composition among the 4 CAR T cell groups. Compared with GBBz T cells, both 21.GBBz and 21.15.GBBz T cells displayed a significantly higher proportion of CD4⁺ central memory cells (T_cm, CD45RO⁺/CD62L⁺; P = 0.0035 and P < 0.001, respectively) and of CD8⁺ stem cell memory/naïve cells (T_scm/T_n, CD45RO⁻/CD62L⁺; P = 0.014 and P = 0.012, respectively; Fig. 3B). After stimulation with tumor cells, these differences were no longer detected (Supplementary Fig. S4B and S4C). Although CD8⁺ 21.GBBz T cells had a higher proportion of less-differentiated cells, these cells also expressed higher amounts of TIM-3 (P = 0.0012) at baseline and significantly higher expression of LAG-3, TIM-3, and PD-1 (P < 0.001, P = 0.001, and P = 0.0138, respectively) after stimulation with HCC compared with GBBz cells, indicating an exhausted phenotype (Supplementary Fig. S5).

Next, we examined whether the observed differences in proliferation were related to differences in rates of apoptosis using annexin V staining. We found that coexpression of IL15 alone or in combination with IL21 decreased the rate of apoptosis in T cells resulting in more live cells compared with GBBz alone following three rounds of stimulation with GPC3-positive tumor cells (P = 0.0015, as measured on day 9; Fig. 3C). This finding corresponded to the superior expansion of 21.15.GBBz T cells observed in Fig. 3A. Thus, the superior in vitro expansion of GPC3-CAR T cells coexpressing IL15 and IL21 was associated with a lower apoptosis rate and increased T_scm/T_n and T_cm populations.

To explore the mechanisms driving differences in proliferation between GPC3-CAR T-cell groups, we examined gene-expression profiles before and after exposure to GPC3-positive HCC cells. We detected significant differences in overall expression patterns at baseline (after manufacturing) in 21.GBBz and 21.15.GBBz T cells compared with GBBz T cells (day 0; Supplementary Figs. S6A–S6C and Fig. S7). Following stimulation with HCC cells (day 3), the gene-expression profiles of 15.GBBz, 21.GBBz, and 21.15.GBBz T cells were different from that of GBBz T cells (Fig. 4A and Supplementary Fig. S6D–S6F) with global differences included genes related to cytotoxicity (GZMA, GNLY, PRF1), chemotaxis (CCR1, CCR2, CCR5), and apoptosis/survival (BCL-2, TCF7/TCF-1).

Because IL15 and IL21 coexpressing CAR T cells had improved proliferation/survival, we focused on genes related to apoptosis and proliferation and found that BCL-2 and TCF7 were overexpressed in 15.GBBz and 21.15.GBBz compared with GBBz T cells (P = 0.023 and 0.0025, respectively; Supplementary Fig. S6D–S6F). No difference was detected at the protein level for BCL-2 (Supplementary Fig. S8), suggesting that it does not play a key role in the enhanced survival of 21.15.GBBz T cells. The TCF-1 protein, encoded by TCF7, is a critical transcription factor for T-cell development, expansion, and survival (38, 39). Prior to stimulation (day 0), the TCF7 gene was expressed similarly in all the GPC3-CAR T-cell groups (Supplementary Fig. S6A–S6C). After stimulation with HCC cells, the gene expression of TCF7 was significantly increased in 21.GBBz (2.2-fold increase, P = 0.014), 15.GBBz (2.3-fold increase, P = 0.028), and to an even greater extent in 21.15.GBBz T cells (4.1-fold increase, P = 0.0003) compared with in GBBz T cells (Fig. 4A). The proportion of TCF-1 (protein)–positive cells at baseline was significantly higher in 21.15.GBBz T cells compared with other CAR T-cell groups (21.15.GBBz vs. 15.GBBz in CD4 subset: P = 0.0223; in CD8 subset: P < 0.001; Fig. 4B and C). The expression of either IL15, IL21, or the combination improved TCF-1 protein expression in both CD4⁺ and CD8⁺ CAR T cells (Fig. 4B and C). After two consecutive stimulations with tumor cells, the percentage of TCF-1–positive cells was the highest in 21.15.GBBz T-cell group for the CD4 subset (21.GBBz vs. 21.15.GBBz, P = 0.002; 15.GBBz vs. 21.15.GBBz, P = 0.0296) whereas in the CD8 subset, the 15.GBBz and 21.15.GBBz T cells had the highest percentage of TCF-1–positive cells (15.GBBz vs. 21.GBBz P = 0.0124; 21.GBBz vs. 21.15.GBBz P < 0.001; Fig. 4C). TCF-1 expression in the CD8 subset remained highest after two stimulations in the 21.15.GBBz group compared with all other groups (21.GBBz vs. 21.15.GBBz, P = 0.0044; Fig. 4D).

To determine whether TCF-1 was required for the enhanced proliferative capacity of 21.15.GBBz T cells, we knocked out TCF7 with CRISPR-Cas9 and measured CAR T-cell expansion upon repeated engagement with GPC3-positive tumor cells (Supplementary Fig. S9). We found that TCF7-KO eliminated the proliferative benefit of 21.15.GBBz T cells compared with GBBz T cells (Fig. 4E). These results suggested that IL21- and IL15-mediated maintenance of TCF-1–positive cells was responsible for the increased proliferative capacity of 21.15.GBBz T cells.

To evaluate the in vivo expansion and persistence of GPC3-CAR T cells, we established HCC xenografts in NSG mice and injected T cells cotransduced with the individual GPC3-CAR constructs and an eGFP.Ffluc construct optimized for tracking small numbers of cells in vivo via bioluminescence imaging (Fig. 5A; ref. 32). As observed in our previous study (27), GBBz T cells expanded effectively for 8 days, after which the population contracted and disappeared entirely by 15 days after injection (Fig. 5B and C; ref. 27). 21.GBBz and 15.GBBz T cells had a similar timeline of peak expansion compared with GBBz T cells but persisted longer in vivo before their numbers began to decline (day 12, P < 0.001). Coexpression of IL15 and IL21 induced the highest expansion and sustained persistence of GPC3-CAR T cells in vivo (day 15, 21.15.GBBz vs. 21.GBBz P = 0.0012). Peripheral blood analysis 15 days after adoptive transfer showed increased frequency of CD8⁺ 15.GBBz and 21.15.GBBz T cells compared with GBBz T cells (P = 0.0076 and P < 0.0002, respectively; Supplementary Fig. S10). In the spleen, the frequency of CD4⁺ 21.GBBz and 21.15.GBBz CAR T cells was significantly elevated compared with GBBz T cells (P = 0.0386 and P = 0.003, respectively). The frequency of CD8⁺ 21.15.GBBz cells was significantly higher in the spleen compared with other groups including 15.GBBz (P = 0.0083; Fig. 5D).

As a safety assessment, serum concentrations of IL15 and IL21 were measured in all therapeutic groups (day 15). IL21 and IL15 serum concentrations in mice treated with cytokine-containing CAR T cells were similar to those of control and GBBz T-cell–infused mice at the peak of expansion (Supplementary Fig. S11A and S11B). As an additional safety measure, we evaluated if the coexpression of the clinically validated inducible caspase-9 could prevent the proliferation of 21.15.GBBz T cells in vivo. We found that after administering AP20187, the chemical inducer of dimerization intravenously, 21.15.GBBz can be effectively eliminated in tumor-bearing mice. Thus, iC9 suicide switch may be used, if necessary, in the clinical setting for unexpected toxicities (Supplementary Fig. S11C–S11E).

Finally, to evaluate the effects of IL15 and IL21 coexpression on the antitumor activity of GPC3-CAR T cells, we adoptively transferred these cells into mice bearing Ffluc-labeled tumor xenografts and monitored tumor growth weekly. In a relatively slow-growing GPC3⁺ malignant rhabdoid tumor model, coexpression of IL15 and/or IL21 significantly enhanced the antitumor responses after injection of GPC3-CAR T cells (week 5: 21.15.GBBz / 15.GBBz vs. GBBz, P < 0.001; 21.GBBz vs. GBBz, P = 0.0011; Supplementary Fig. S12), and IL15 coexpressing CAR T-cell groups induced a more rapid antitumor effect (week 5: 21.15.GBBz/15.GBBz vs. 21.GBBz, P < 0.001). Next, we examined GPC3-CAR T-cell antitumor responses in a rapidly growing HCC xenograft model injecting 2 × 10⁶ CAR T cells. We found that 15.GBBz and 21.15.GBBz T cells mediated superior antitumor activity compared with GBBz or 21.GBBz T cells and control groups (P < 0.001; Supplementary Fig. S13). In this model, 21.15.GBBz T cells eliminated tumors more rapidly than 15.GBBz T cells (week 4, P < 0.001). Lastly, to further stress the antitumor potential of GPC3-CAR T cells, we injected a low dose of 5 × 10⁵ CAR T cells in mice engrafted with rapidly growing HCC xenografts. At this dose, only 21.15.GBBz T cells, but not 15.GBBz T cells, maintained antitumor activity, which translated into significant survival advantage (15.GBBz vs. 21.15.GBBz, P < 0.001, Fig. 5E–G). These results demonstrated that 21.15.GBBz T cells had superior expansion, persistence, and antitumor activity against HCC in vivo.

Here, we showed preclinical evidence that human GPC3-CAR T cells coexpressing IL15 and IL21 had superior expansion and antitumor activity in preclinical models of HCC. Gene-expression analysis identified TCF-1 as a key transcription factor associated with the increased proliferative capacity of 21.15.GBBz T cells.

We found that coexpression of IL21 and/or IL15 with GBBz did not affect the potent, specific, short-term in vitro cytolytic activity of T cells against HCC tumors cells; in contrast, we did detect significant differences in effector cytokine production polarization when comparing GBBz T cells with those coexpressing one or both cytokines. As in our previous study, GBBz T cells secreted a T_H1-polarized cytokine profile (high IFNγ and GM-CSF; low IL10 and IL4; ref. 27). This overall trend was recapitulated in the current study, with CAR T-cell groups showing GBBz-mediated T_H1 polarization regardless of cytokine coexpression. However, we found a striking decrease in IL13 production in CAR T cells coexpressing IL15 (15.GBBz and 21.15.GBBz T cells). IL13 is a T_H2-cytokine primarily produced by CD4⁺ T cells that generates many of the same biological effects as IL4, including decreasing the antitumor function of T cells and promoting tumor cell proliferation (40, 41). IL13 also plays an important role in homeostasis of myeloid-derived suppressor cells (MDSC), which can dampen the efficacy of immunotherapies, increase metastasis formation, cancer progression and inhibit CAR T-cell activity (41–45); therefore, limiting the amount of IL13 in the tumor microenvironment may enhance the therapeutic potential of CAR T cells (45, 46).

Because expansion and persistence of CAR T cells are strong predictors of clinical activity, a key objective of this study was to improve these parameters. We demonstrated that coexpression of IL15 and IL21 in GPC3-CAR T cells increased the proportion of naïve/stem cell memory and central memory GPC3-CAR T cells after manufacture. These less-differentiated T cells have greater proliferative capacity than more mature cells (47–49), providing a potential proliferative advantage for GPC3-CAR T cells coexpressing IL21. Given that all experimental groups were manufactured under the same culture conditions including supplementation with IL15 and IL21, this finding was unexpected. Continuous production of IL21 via transgenic expression from the GPC3-CAR throughout the culturing process likely influenced the T-cell phenotype.

We determined that compared with other experimental groups, coexpression of IL15 alone or in combination with IL21 decreased the proportion of transduced T cells undergoing apoptosis after multiple stimulations with HCC tumor cells; thus, proportionately increasing surviving CAR T cells. Finally, we found that the TCF7 gene encoding TCF-1, a key transcription factor in T-cell development, expansion, memory formation, and survival (38, 39, 50, 51), was more highly expressed in the CD8⁺ subsets of GPC3-CAR T cells expressing IL15, IL21, or both compared with in CD8⁺ GBBz T cells. Following two rounds of stimulation with HCC cells, TCF-1 expression was highly maintained in 21.15.GBBz T cells and was associated with enrichment for and continued expansion of CD8⁺ CAR T cells. CRISPR-Cas9 knockout of TCF7 eliminated the proliferative advantage of 21.15.GBBz T cells, suggesting that TCF-1 plays a role in enhancing the expansion and survival of IL15- and IL21 coexpressing GPC3-CAR T cells.

In cancer patients, the effector-to-target ratio of CAR T cells to cancer cells greatly favors cancer; thus, its complete elimination requires CAR T cells to kill repeatedly and expand. In sequential killing assays, in which GPC3-CAR T cells are repeatedly exposed to fresh tumor cells in vitro, we found that combined expression of IL15 and IL21 resulted in the most expansion. Additionally, in an aggressive xenograft model of HCC treated with the lowest dose of GPC3-CAR T cells, thereby stressing their functional capacity, cells transduced with 21.15.GBBz resulted in significant survival advantage. These findings suggest that IL15 and IL21 should provide the most potent antitumor activity in the clinical setting.

Safety remains a central requirement for all cancer treatments. Elevated concentrations of IL15 and IL21 can cause side effects as described in early-phase studies of subcutaneous or intravenous recombinant IL21 and IL15 administration in patients with cancer (52). In these studies, at the maximum tolerated dose (MTD), peripheral blood peak concentrations of IL15 and IL21 were 1,608 pg/mL and 141 ng/mL, respectively (53, 54). In mice infused with GPC3-CAR T cells expressing IL15 and/or IL21, cytokine concentrations at the peak of T-cell expansion remained 100- to1000-fold below the peak corresponding MTDs measured in humans. Therefore, we do not anticipate systemic toxicities in patients treated with GPC3-CAR T cells coexpressing IL15 and/or IL21. Nevertheless, we show that the use of inducible caspase-9 can effectively eliminate IL15 and IL21 expressing GPC3-CAR T cells in vivo; thus, this “suicide switch” system may be used during clinical evaluation (33).

In conclusion, GPC3-CAR T cells coexpressing IL15 and IL21 were effective at treating HCC in preclinical models. Our findings address a major barrier by enhancing the expansion and persistence of therapeutic cells, resulting robust antitumor responses. We provided mechanistic insight and preclinical evidence to substantiate testing GPC3-CAR T cells coexpressing IL15 and IL21 in patients, an avenue that will be explored in two ongoing clinical trials (NCT02932956 and NCT02905188).

S.A. Batra, P. Rathi, L.S. Metelitsa, and A. Heczey have pending patent applications related to GPC3-CARs and/or cytokine coexpression. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.A. Batra, P. Rathi, A. Heczey

Development of methodology: S.A. Batra, P. Rathi, T.P. Nguyen, M. Mamonkin, L.S. Metelitsa, A. Heczey

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Batra, L. Guo, J. Fleurence, R.S. Shaik, T.P. Nguyen, A. Heczey

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. Batra, P. Rathi, A.N. Courtney, J. Fleurence, J. Balzeau, R.S. Shaik, M.-F. Wu, S. Bulsara, A. Heczey

Writing, review, and/or revision of the manuscript: S.A. Batra, P. Rathi, A.N. Courtney, J. Fleurence, R.S. Shaik, M.-F. Wu, S. Bulsara, L.S. Metelitsa, A. Heczey

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.A. Batra, P. Rathi, L. Guo, A.N. Courtney, J. Fleurence, J. Balzeau, R.S. Shaik, T.P. Nguyen, M.-F. Wu, S. Bulsara, M. Mamonkin, L.S. Metelitsa, A. Heczey

Study supervision: A. Heczey

We thank Dr. Malcolm Brenner (Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX) for helpful discussions. We also thank Elise Shen for help with animal experiments, Erica Di Piero (Texas Children's Cancer Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX), and Catherine Gillespie (Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX) for helpful edits of this manuscript. This study was supported by National Institute of Health S10 OD020066 to the Texas Children's Flow Cytometry Core and American Cancer Society Mentored Scholar Award to A. Heczey.

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.