Engineering genetic devices for in vivo control of therapeutic T cell activity triggered by the dietary molecule resveratrol

Significance

Chimeric antigen receptor (CAR)–engineered T cell therapies have shown tremendous success in the clinic, but excessive cytotoxic activity and poor control over engineered T cells limit the application of CAR-T therapies. Here we have developed resveratrol (RES)–triggered regulation devices (on/off) that could be installed into CAR-T cells, which allow precise control over T cell activity through adjustment of RES dosage. We further demonstrated RES-inducible/repressible CAR expression and reversible control over T cell activation via a RES-titratable mechanism. Our results reveal that RES_ind-CAR T cells can be dose-dependently activated by RES with strong anticancer cytotoxicity. Our RES-controlled systems establish proof of concept for strategies to control cancer immunotherapies based on the RES-regulated repression/induction of therapeutic immune cells.

Abstract

Chimeric antigen receptor (CAR)–engineered T cell therapies have been recognized as powerful strategies in cancer immunotherapy; however, the clinical application of CAR-T is currently constrained by severe adverse effects in patients, caused by excessive cytotoxic activity and poor T cell control. Herein, we harnessed a dietary molecule resveratrol (RES)–responsive transactivator and a transrepressor to develop a repressible transgene expression (RES_rep) device and an inducible transgene expression (RES_ind) device, respectively. After optimization, these tools enabled the control of CAR expression and CAR-mediated antitumor function in engineered human cells. We demonstrated that a resveratrol-repressible CAR expression (RES_rep-CAR) device can effectively inhibit T cell activation upon resveratrol administration in primary T cells and a xenograft tumor mouse model. Additionally, we exhibit how a resveratrol-inducible CAR expression (RES_ind-CAR) device can achieve fine-tuned and reversible control over T cell activation via a resveratrol-titratable mechanism. Furthermore, our results revealed that the presence of RES can activate RES_ind-CAR T cells with strong anticancer cytotoxicity against cells in vitro and in vivo. Our study demonstrates the utility of RES_rep and RES_ind devices as effective tools for transgene expression and illustrates the potential of RES_rep-CAR and RES_ind-CAR devices to enhance patient safety in precision cancer immunotherapies.

T cell therapy with tumor-targeted chimeric antigen receptor (CAR)–engineered T cells has proven to be a transformative cancer treatment for a range of target indications (1, 2). CARs are constructed via a surface-displayed single-chain antibody variable domain (scFv) coupled to intracellular signaling components, to endow patient-derived T cells with the ability to recognize antigens and subsequently eliminate tumor cells (3). Ever since the approval of CD19-specific CAR T cells for the treatment of B cell leukemias and lymphomas, dramatic clinical successes have been achieved and numerous CAR T cell products are emerging (4–7).

Despite initial encouraging results, several challenges remain for CD19 CAR T cells, one main challenge being a lack of control over these engineered cells (8, 9). Rapid responses after large cell infusion doses can induce several immune-mediated side effects, such as cytokine release syndrome (CRS), tumor lysis syndrome, and neurotoxicity (10, 11). However, achieving potent therapeutic effects requires the robust expansion and sequential activation of a sufficient number of CAR T cells. Accordingly, it is a significant challenge to administrate appropriate doses of this “living drug” in order to balance in vivo efficacy and toxicity, further complicated by the highly divergent responses of patients (12–14).

Synthetic devices have been shown to improve the safety and feasibility of CAR T cell therapy by regulating CAR expression and controlling T cell activation (15, 16). To mitigate immunotherapy toxicity, several safety devices have been developed to control engineered T cells in a “function-off” or “function-on” manner. Previous studies have reported several user-controlled function-off devices to eliminate infused T cells from patients exhibiting severe toxicity, including suicide devices (17, 18), antibody-based elimination devices (9, 19, 20), small molecule-controlled STOP-CAR devices (21), and photoswitchable CAR devices (22). However, function-off devices may be insufficient to ensure safety against toxicities induced immediately upon cell infusion.

Complementary strategies using a function-on approach include several inducible devices with tunable CAR T cell function control: small molecule-inducible CAR devices (23–27), intermediary antibody or protein-based CAR devices (28–30), as well as light/ultrasound-responsive noninvasive devices (31–33). Synthetic Notch (synNotch) receptor-based circuits were also designed to enhance the precision of tumor recognition (34–38). Although these devices theoretically enable safer CAR T cell therapies, synthetic modules controlled by even safer natural molecules capable of diffusing throughout the entire body would offer a more versatile mechanism of control. Further development of orthogonally regulated devices that enable facile control of CAR T cell functions to ensure safer device control in immunotherapy is needed.

The stilbenoid natural product resveratrol (RES, 3,4′,5-trihydroxystilbene), notably present in red wine, grapes, peanuts, and berries, has been shown to be a safe inducer compound for genetic devices designed for therapeutic applications (39). Beyond its demonstrated utility in genetic engineering applications, this molecule possesses wide-ranging beneficial effects, including clinical evidence for activities against aging, cancer, cardiovascular diseases, and inflammation (40–43).

To enable safer strategies for cancer immunotherapy with both a function-off and function-on modality, we designed a resveratrol-repressible transgene expression (RES_rep) device and a resveratrol-inducible transgene expression (RES_ind) device based on resveratrol-responsive regulatory elements. We demonstrate that our RES_rep and RES_ind devices can reversibly and flexibly control transgene expression in vitro and in vivo in a highly tunable manner. We further apply these devices to control CAR expression and CAR-mediated cell activities by developing a resveratrol-repressible CAR expression (RES_rep-CAR) device and a resveratrol-inducible transgene expression (RES_ind-CAR) device. We then demonstrate that the RES_rep-CAR device can repress CAR expression and CAR-mediated cell activation in primary human T cells and a xenograft mouse model of leukemia, all with triggerable T cell inactivity in case of severe toxicity. We further confirm the titratable and reversible capabilities of the RES_ind-CAR device for both CAR expression and T cell activity in vitro and in vivo; this is a feature that could allow physicians to precisely control T cell activity by adjusting the dosage of resveratrol according to the clinical responses of individual patients. These resveratrol-controlled transgene expression devices extend synthetic biology-inspired tools to clinically relevant applications and provide potential strategies for improving the safety of T cell immunotherapies by preventing fatal toxicity in clinical settings.

Results

Design, Construction, and Characterization of an RES_rep Device in Mammalian Cells and Mice.

Capitalizing on resveratrol-responsive TtgR, a specific transcriptional repressor of efflux pump (TtgABC) in Pseudomonas putida (44), we engineered a RES_rep device in mammalian cells (Fig. 1A). Our designed RES_rep device consists of a resveratrol-dependent transactivator ResA₃ (TtgR-VPR) that we fused to a synthetic activator VPR (comprising three transcriptional activators: VP64, p65, and Rta) (45) via the C terminus of TtgR. The chimeric transactivator can bind to the resveratrol-dependent promotor P_ResA1, which contains a ResA₃ binding site (O_TRC1, reverse complement and site-directional mutation of O_TtgR) positioned in front of a minimal human cytomegalovirus immediate-early promoter (P_hCMVmin). Transactivator binding results in recruitment of the transcription initiation complex to trigger transgene expression; subsequently, resveratrol releases ResA₃ from P_ResA1, thereby silencing transgene expression (Fig. 1A).

Fig. 1.

Design and characterization of the resveratrol-repressible transgene expression device (RES_rep) in mammalian cells and mice. (A) Scheme for the RES_rep device. The resveratrol-dependent transactivator ResA₃ (TtgR-VPR) module is formed by fusing a tripartite activator composed of VP64, p65, and Rta in tandem (VPR) to the C terminus of TtgR. In the absence of resveratrol, constitutively expressed ResA₃ binds a chimeric target promoter P_ResA1 (O_TRC1-P_hCMVmin) containing a tandem ResA₃-specific operator site O_TRC1 5′ of a minimal human cytomegalovirus immediate-early promoter (P_hCMVmin), and activates SEAP production. The presence of resveratrol results in the release of ResA₃ from P_ResA1, thereby silencing SEAP production. (B) Resveratrol-repressible SEAP expression in different mammalian cell lines (HEK293, hMSC-TERT, HEK293FT, HeLa, or Hana3A) were cotransfected with pLF36 (P_hCMV-ResA₃-pA) and pLF31 (P_ResA1-SEAP-pA) and cultivated for 48 h in the presence or absence of resveratrol. (C) Selection of stably transgenic resveratrol-repressible cell lines (HEK_RES-rep-SEAP). HEK293 cells were cotransfected with pLF51 (ITR-O_TRC1-P_hCMVmin-SEAP-pA::P_hCMV-TtgR-VPR-P2A-PuroR-pA-ITR) and the transposase expression vector pCMV-T7-SB100 (P_hCMV-SB100X-pA). The selected cell clones were profiled for their resveratrol-repressible SEAP regulation performance. The blue frame marks the best-in-class clone chosen for the following experiments. (D) Resveratrol dose-dependent SEAP production in HEK_RES-rep-SEAP cells. Different color bars represent different time periods for profiling SEAP production. (E) Long-term SEAP production kinetics of HEK_RES-rep-SEAP stable cells. (F) Reversibility of HEK_RES-rep-SEAP-mediated SEAP production. HEK_RES-rep-SEAP cells were cultivated for 6 d while altering resveratrol concentrations from 0 to 20 μM. Cell density was adjusted to 5 × 10⁴ every 48 h, and SEAP production was profiled every 12 h. (G) Dose-dependent resveratrol-repressible SEAP production in mice. C57BL/6J mice implanted with 2 × 10⁶ microencapsulated HEK_RES-rep-SEAP cells were intraperitoneally injected with resveratrol once a day (0 to 500 mg/kg/d). SEAP production in blood was profiled after 48 h. (H) Long-term resveratrol-repressible SEAP production kinetics in mice. SEAP production was profiled at day 2, 4, 6, and 15. The data for B–F are mean ± SD, n ≥ 3. The animal data of G and H are mean ± SEM. P values were calculated with Student’s t test (n = 6 mice). *P < 0.05; **P < 0.01; ***P < 0.001 versus control. N.S., not statistically significant.

We initially optimized the synthetic transactivator ResA by fusing TtgR to different transactivation domains (VP16, VP64, and VPR) driven by P_SV40 (simian virus 40 promoter) or P_hCMV (human cytomegalovirus promoter; SI Appendix, Fig. S1 A–H) as well as optimizing the promoter P_ResA with various iterations of O_TRC1 tandem repeats (SI Appendix, Fig. S1 I–N). Ultimately, we found that a combination of P_hCMV-driven ResA₃ and P_ResA1-driven secreted human placental alkaline phosphatase (SEAP) demonstrated the best induction performance; this version of the RES_rep device also achieved significant induction performance in five human cell lines (Fig. 1B). Importantly, assessment of resveratrol-mediated toxicity in mammalian cells showed that resveratrol had no negative effect on cell viability (SI Appendix, Fig. S2), using resveratrol concentrations of 0 to 50 μM.

We next explored the potential of the RES_rep device for cell/gene therapy, by stably integrating the RES_rep device driving SEAP into HEK293 cells. Although the device achieved success across a series of clonal cell lines, the best inhibition (∼15.5-fold) was in a monoclonal HEK_RES-rep-SEAP cell line (Fig. 1C). The HEK_RES-rep-SEAP cell line showed excellent resveratrol dose- and time-dependent transgene expression (Fig. 1D). Moreover, this cell line exhibited sustained output over a 14-d culture (Fig. 1E) and displayed fully reversible transgene expression kinetics (Fig. 1F). For in vivo performance analysis of the resveratrol-controlled RES_rep device, HEK_RES-rep-SEAP cells were microencapsulated in coherent alginate-poly-(l-lysine)-alginate beads and implanted into mice. After mice were given resveratrol (intraperitoneal [i.p.] injection, ranging from 0 to 500 mg/kg daily), we observed that SEAP production from the RES_rep device was controlled in a resveratrol dose-dependent manner (Fig. 1G) and observed sustained repression for 15 d in mice (Fig. 1H). These results confirm that RES_rep can precisely control transgene inhibition in mammalian cells and in mice.

Characterization of Resveratrol-Repressible CAR Expression (RES_rep-CAR) and CAR-Mediated T Cell Activity.

Given previous studies reporting that resveratrol exerts potent antitumor effects against some cancers (46, 47) and may enhance human T cell activity (48, 49), we profiled the effects of resveratrol on the Jurkat T cells that we planned to use as the recipient cells for our devices to exclude any unanticipated impacts from the induction/repression agent (resveratrol). A total of 30 μM resveratrol had no significant effect on CAR expression or CD19 antigen-induced activation of unmodified Jurkat T cells (negative control) or conventional CAR-engineered Jurkat T cells (positive control; SI Appendix, Fig. S3).

We next developed a RES_rep-CAR device to inhibit CAR expression in engineered Jurkat T cells and primary T cells; this device was capable of selectively inactivating its therapeutic effects in the presence of resveratrol (Fig. 2A). First, we optimized the RES_rep-CAR device in Jurkat T cells (Jurkat_RES-rep-CAR) and found that a combination of P_mPGK (mouse phosphoglycerate kinase gene promoter)-driven ResA₃ and P_ResA1-driven CAR-P2A-EGFP achieved the best resveratrol-mediated performance in Jurkat T cells (SI Appendix, Fig. S4). These Jurkat_RES-rep-CAR cells, which were transduced with the optimal RES_rep-CAR device, exhibited dose-dependent resveratrol-repressible CAR expression (Fig. 2B). These responses include CD69 expression (a general surface activation marker; Fig. 2C) and hIL-2 production (a general readout for effector T cell activation; Fig. 2D). Moreover, the RES_rep-CAR device enables fully reversible CAR expression in Jurkat T cells (Fig. 2E). We also found that the presence of resveratrol reduced the extent that the CD19 antigen stimulated the activation of the Jurkat_RES-rep-CAR. These results demonstrate that the RES_rep device can fine tune the repression of both CAR expression levels and the CAR-mediated cell activation of Jurkat T cells.

Fig. 2.

Resveratrol-titratable T cell inactivation in Jurkat T and primary T cells engineered with RES_rep-CAR. (A) Scheme of the RES_rep-CAR device. Jurkat T cells or human primary T cells were modified utilizing lentivirus for resveratrol-repressible CAR expression. In the absence of resveratrol, RES_rep-CAR T cells exhibited sustained CAR expression, and these cells were activated upon antigen stimulation. In the presence of resveratrol, CAR expression was inhibited and the cells were insensitive to antigen stimulation. (B) Dose-dependent resveratrol-repressible CAR expression in Jurkat T cells (Jurkat_RES-rep-CAR). Jurkat T cells engineered with lentiviral pLF78 (LTR-P_mPGK-ResA₃-LTR) and pLF100 (LTR-P_ResA1-CAR-LTR) were cultivated in medium containing 0 to 30 μM resveratrol, and CAR expression was quantified by flow cytometry after 48 h. (C and D) Resveratrol and antigen-triggered Jurkat_RES-rep-CAR cell activation. Cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with CD19⁻ K562_EGFP cells (white squares) or CD19⁺ K562_mCherry cells (red squares) at an E:T ratio of 2:1. Surface expression of early T cell activation marker CD69 (C) and hIL-2 secretion (D), quantified after an overnight incubation. (E) Reversibility of RES_rep-CAR device-mediated CAR expression in Jurkat T cells. Cells transduced with lentiviral pLF363 (LTR-P_mPGK-ResA₄-LTR) and pLF100 were cultivated without (gray rectangles) or with 20 μM resveratrol (blue rectangles). The relative mRNA level of CAR was quantified by qPCR every 12 h and normalized to the control group incubated without resveratrol. Cell density was adjusted to 1 × 10⁶ cells/mL every 48 h. (F) Dose-dependent resveratrol-repressible CAR surface expression in human primary T cells (RES_rep-CAR T cells). RES_rep-CAR T cells engineered with optimized the RES_rep-CAR device were transduced with lentiviral pLF363 and pLF100, and then cells were cultivated in medium containing 0 to 15 μM resveratrol. CAR expression was quantified by flow cytometry after 48 h. (G and H) RES_rep-CAR T cell-mediated cytokine release. RES_rep-CAR T cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with CD19⁻ K562_EGFP cells (white squares) or CD19⁺ K562_mCherry cells (red squares) at an E:T ratio of 5:1. hIL-2 (G) and hIFN-γ (H) were quantified after 18-h incubation. (I) RES_rep-CAR T cell-mediated targeted cytotoxicity. RES_rep-CAR T cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with a mixture of target cells containing CD19⁻ K562_EGFP cells and CD19⁺ K562_mCherry cells at an E:T ratio of 10:1. After 18 h, cells were collected for flow cytometry analysis and cytotoxicity values were calculated based on the ratios of CD19⁺ K562_mCherry cells to CD19⁻ K562_EGFP cells. The data in B–I are mean ± SD, n = 3. N.D., not detectable.

We subsequently optimized the RES_rep-CAR device in primary human T cells, which required the development of a particularly resveratrol-sensitive TtgR mutant (H67A) (50). The combination of P_mPGK-driven ResA₄ (TtgR^H67A-VPR) and P_ResA1-driven CAR exhibited excellent transgene inhibition upon incubation with 15 μM resveratrol in primary human T cells (SI Appendix, Fig. S5). These RES_rep-CAR T cells derived from primary human T cells exhibited titratable resveratrol-repressed CAR expression (Fig. 2F) as well as excellent CAR-regulated cell deactivation upon CD19 antigen stimulation in the presence of increased concentration of resveratrol, assessed via hIL-2 (Fig. 2G), and hIFN-γ (Fig. 2H) expression.

To test the cytotoxic performance of the engineered primary human T cells, we labeled CD19⁺ K562 cells with mCherry (CD19⁺ K562_mCherry) and CD19⁻ K562 cells with EGFP (CD19⁻ K562_EGFP) and then mixed them as a population of two target cell types that could be identified and measured by flow cytometry (SI Appendix, Fig. S6). The cytotoxic activity of the engineered primary T cells expressing CARs or the RES_rep-CARs were quantified as the selective reduction of CD19⁺ K562_mCherry cells. RES_rep-CAR T cells showed ∼25% in vitro cytotoxic efficacy, tunable based on resveratrol concentration (Fig. 2I). Moreover, CAR-mediated T cell activation could be repressed by the RES_rep-CAR device in a time-dependent manner (SI Appendix, Fig. S7). Importantly, we confirmed that 15 μM resveratrol alone had no significant effects on either target cell lethality or cytokine production of non-RES_rep regulated primary T cells modified with conventional CAR (SI Appendix, Fig. S8). In summary, we successfully constructed a RES_rep-CAR device for fine tuning CAR expression and applied it in Jurkat T cells and primary T cells.

Anticancer Performance of RES_rep-CAR T Cells in NCG Mice.

To investigate whether the cytotoxic activity of RES_rep-CAR T cells could be regulated by resveratrol in mice, we evaluated target cell lethality and cytokine release in a rapid xenograft mouse tumor model of B cell leukemia (23) (Fig. 3). In this model, we implanted a mixture of CD19⁺ K562_mCherry and CD19⁻ K562_EGFP cells into the peritoneal cavity of NCG mice (NOD-Prkdc^em26Cd52Il2rg^em26Cd22/NjuCrl: triple immunodeficiency and lacking functional/mature T, B, and natural killer [NK] cells, along with reduced macrophage and dendritic cell function). After 10 h, we injected i.p. unmodified human primary T cells (negative control) or engineered human T cells equipped with RES_rep-CAR into the mice. Subsequently, we administered various doses of resveratrol (50, 100, or 200 mg/kg/d) or vehicle control by i.p. injection at the indicated times. Mice were killed after 48 h for peritoneal lavage collection, from which recovered cells were analyzed by flow cytometry; serum cytokine release analysis occurred at 24 h and 48 h (Fig. 3A).

Fig. 3.

Resveratrol-repressible activation of primary human T cells engineered with RES_rep-CAR in NCG mice. (A) Experimental design for in vivo assessment of resveratrol-controlled RES_rep-CAR T cell-mediated cytotoxicity against K562 cells in a rapid xenograft mouse tumor model. NCG mice inoculated with 3 × 10⁷ mixed K562 cells (CD19⁻ K562_EGFP cells:CD19⁺ K562_mCherry cells = 1:1) were injected intraperitoneally with 2 × 10⁷ unmodified primary human T cells (control) or RES_rep-CAR T cells, followed by the administration of various resveratrol doses (50, 100, or 200 mg/kg/d) or vehicle control at the indicated times. After 48 h, blood was drawn for cytokine examination and peritoneal cells were quantified by flow cytometry to measure in vivo cytotoxicity. (B) Representative flow cytometry data show the resveratrol-repressible depletion of the CD19⁺ K562_mCherry cell population. (C) Quantified target cytotoxicity results for all experimental groups. Cytotoxicity values were calculated from the ratios of CD19⁺ K562_mCherry cells to CD19⁻ K562_EGFP cells. n = 4 mice. (D–G) In vivo hIL-2 (D and F) and hTNF (E and G) release at 24 h and 48 h. The data in C–G are mean ± SEM. P values were calculated with Student’s t test (n = 4 mice). *P < 0.05; **P < 0.01; ***P < 0.001 versus control.

Mice injected with RES_rep-CAR T cells showed resveratrol-repressible depletion of the CD19⁺ K562_mCherry cell population upon treatment with different resveratrol dosages; no depletion was observed for mice injected with control T cells (Fig. 3B). Further, comparison of the two dosages showed that serum cytokine levels (hIL-2 and hTNF) and extent of target cytotoxicity were tunable based on in vivo resveratrol concentration (Fig. 3 C–G). We confirmed that the high dosage of resveratrol (200 mg/kg/d) had no significant effect on the cytotoxicity capability in the positive control CAR T cells transplanted into mice (SI Appendix, Fig. S9). In summation, RES_rep-CAR T cells were able to achieve tunable cancer cell lethality in a mouse leukemia model.

Design, Construction, and Characterization of a RES_ind Transgene Expression Device in Mammalian Cells and Mice.

Our RES_rep-CAR T cells enable a strategy for negatively regulating T cell function, an attractive approach when the clinical response from a constitutively activated cell type requires inactivation; nevertheless, a RES_ind-CAR expression device should enable fine tuning therapeutic activity in a function-on manner. Pursuing this, we engineered a resveratrol-inducible on device (RES_ind) for controlling transgene expression in mammalian cells. Our designed RES_ind device comprises a resveratrol-dependent transrepressor ResR₁ (TtgR-KRAB), based on fusing a Krüppel-associated box protein (KRAB) to the C terminus of TtgR; this can bind to the synthetic resveratrol-inducible promotor (P_ResR), harboring a constitutive expression promoter embedded with transrepressor ResR₁ binding sites (O_TRC1, a variant of O_TtgR), thereby repressing transgene expression. Resveratrol releases transrepressor ResR₁ from P_ResR and induces transgene expression (Fig. 4A).

Fig. 4.

Design and characterization of a resveratrol-inducible transgene expression device (RES_ind) in mammalian cells and mice. (A) Schematic showing RES_ind device design and the experimental procedure for resveratrol-controlled transgene expression assessment in mice. TtgR was fused to a transsilencing human Krüeppel-associated box (KRAB) domain to obtain the resveratrol-dependent transrepressor ResR₁ (TtgR-KRAB), which is driven by the human cytomegalovirus immediate early promoter P_hCMV. In the absence of resveratrol, ResR₁ binds to a chimeric target promoter P_ResR12 [O_TRC1-P_hCMV-(O_TRC1)₂], containing the ResR₁-specific DNA binding site O_TRC1 positioned at the 3′ and 5′ ends of P_hCMV; this represses P_hCMV-driven SEAP production. In the presence of resveratrol, ResR₁ is released from P_ResR12 and initiates SEAP production. (B) Resveratrol-inducible SEAP production in various mammalian cell lines. HEK293, hMSC-TERT, HEK293FT, HeLa, and Hana3A cells cotransfected with pLF25 (P_hCMV-ResR₁-pA) and pLF18 (P_ResR12-SEAP-pA) were cultivated for 48 h in the presence or absence of 20 μM resveratrol. (C) Selection of stably transgenic resveratrol-inducible cell lines (HEK_RES-ind-SEAP). HEK293 cells were cotransfected with pLF50 (ITR-P_ResR12-SEAP-pA::P_hCMV-TtgR-KRAB-P2A-PuroR-pA-ITR) and the transposase expression vector pCMV-T7-SB100 (P_hCMV-SB100X-pA). The selected cell clones were profiled for their resveratrol-inducible SEAP regulation performance. The blue frame marks the best-in-class clone chosen for the following experiments. (D) Resveratrol dose-dependent SEAP production in HEK_RES-ind-SEAP cells. Different color bars represent different time periods for profiling SEAP production. (E) Long-term SEAP production kinetics in HEK_RES-ind-SEAP stable cells. (F) Reversibility of HEK_RES-ind-SEAP-mediated SEAP production. HEK_RES-ind-SEAP cells were cultivated for 6 d while altering the resveratrol concentrations from 0 to 20 μM. Cell density was adjusted to 5 × 10⁴ every 48 h and SEAP production was profiled every 12 h. (G) Dose-dependent resveratrol-inducible SEAP production in mice. C57BL/6J mice implanted with 2 × 10⁶ microencapsulated HEK_RES-ind-SEAP cells were intraperitoneally injected with resveratrol once a day (0 to 250 mg/kg/d). SEAP production in blood was profiled after 48 h. (H) Long-term resveratrol-inducible SEAP production kinetics in mice. SEAP production was profiled on days 2, 4, 6, and 15. The data in B–F are mean ± SD, n ≥ 3 independent experiments. The animal data G and H are mean ± SEM. P values were calculated with Student’s t test (n = 6 mice). **P < 0.01; ***P < 0.001 versus control.

To obtain the envisioned RES_ind device with minimal leakage in the absence of resveratrol and strong resveratrol-induced transgene expression, we optimized the RES_ind device by testing different O_TtgR variants, different variants of resveratrol-responsive promoters, and different strengths of constitutive promoters for driving TtgR expression (SI Appendix, Figs. S10 and S11). Due to its superior induction performance, the variant O_TRC1 (reverse complement and site-directional mutation of O_TtgR) was initially chosen as the optimal binding sequence for the subsequent optimization of resveratrol-responsive promoters. Ultimately, we determined that a combination of P_hCMV-drived TtgR with the resveratrol-responsive promoter P_ResR12 [O_TRC1-P_hCMV-(O_TRC1)₂] exhibited optimal induction performance in HEK293 cells (SI Appendix, Fig. S11O), and this version of the RES_ind device also achieved significant induction performance across five human cell lines (Fig. 4B) and strong time-course–dependent performance (SI Appendix, Fig. S12).

As in our characterization of the RES_rep device in assessing suitability for cell/gene therapy, we evaluated the best monoclonal HEK_RES-ind-SEAP cell line (378-fold induction; Fig. 4C), from the examined clones and found that it showed excellent resveratrol dose- and time-dependent transgene expression (Fig. 4D). Moreover, this cell line exhibited sustained resveratrol-induced transgene output throughout a 14-d culture (Fig. 4E) and fully resveratrol-reversible transgene expression kinetics (Fig. 4F).

We then microencapsulated HEK_RES-ind-SEAP cells and implanted them into mice to assess resveratrol-induced in vivo transgene expression from the RES_ind device. After mice were intraperitoneally administrated with resveratrol at doses ranging from 0 to 250 mg/kg daily, we observed that resveratrol could control SEAP production in a dose-dependent manner (Fig. 4G), over an induction window of 15 d (Fig. 4H). These results confirm that the RES_ind device can fine tune transgene induction in mammalian cells and in mice.

Seeking to decrease the amount of RES needed to trigger the RES_ind device, and ultimately aiming to achieve controlled transgene expression based on the oral administration of resveratrol, we attempted to further improve the sensitivity of our RES_ind device by testing various mutants of the TtgR protein (SI Appendix, Fig. S13). The RES_ind-mut device proved to be substantially more sensitive to resveratrol than our initial device. The combination of P_hCMV-driven ResR₄ (TtgR^H67A-KRAB) and P_ResR12-driven SEAP exhibited the best induction performance in HEK293 cells with dose- and time-course–dependent induction kinetics, especially at lower resveratrol concentrations (0 to 5 µM; SI Appendix, Fig. S13 A and B). We next constructed and selected a highly sensitive and inducible isogenic cell clone that allowed for the RES-inducible transgene expression of SEAP (HEK_{RES-ind-mut-SEAP}). The RES_ind-mus device continued to show superior transgene induction at low resveratrol concentrations (0 to 6 µM; SI Appendix, Fig. S13C). When the microencapsulated HEK_{RES-ind-mut-SEAP} cells were implanted into mice, regardless of delivery method (intraperitoneal injection or oral intake of resveratrol), resveratrol controlled transgene expression in a dose-dependent manner (SI Appendix, Fig. S13 D–G). These results demonstrate that the RES_ind-mus device is successful in being more sensitive to resveratrol.

Tunability and Kinetics of RES_ind-CAR Expression in Jurkat T Cells.

We next developed a RES_ind-CAR expression device to positively regulate CAR expression in engineered Jurkat T cells and primary T cells; this device was capable of selectively activating therapeutic effects in the presence of resveratrol and upon CD19 antigen stimulation (Fig. 5A).

Fig. 5.

Resveratrol-titratable activation of Jurkat T and primary T cells engineered with RES_ind-CAR. (A) Schematic of the RES_ind-CAR device. Jurkat T cells or human primary T cells were engineered with lentivirus to express RES_ind-CAR. In the absence of resveratrol, CAR expression was repressed and T cells were insensitive to antigen-mediated activation. The presence of resveratrol induced CAR expression, and these CAR⁺ T cells were activated by antigen-positive (CD19⁺) K562 target cells (assessed based on cytokine production, CD69 expression, and cell killing assays). (B) Dose-dependent resveratrol-inducible CAR and EGFP expression in Jurkat T cells (Jurkat_{RES-ind-CAR-EGFP}). Jurkat_{RES-ind-CAR-EGFP} cells engineered with lentiviral pLF63 (LTR-P_mPGK- ResR₁-LTR) and pLF64 (LTR-P_ResR11-CAR-P2A-EGFP-LTR) were cultivated with different concentrations of resveratrol for 48 h. Expression levels of CAR and EGFP were quantified by flow cytometry. (C) Long-term Jurkat_{RES-ind-CAR-EGFP}-mediated EGFP expression. (D and E) Reversibility of Jurkat_{RES-ind-CAR-EGFP}-mediated EGFP expression. Cells were cultivated without (gray rectangles) or with 20 μM resveratrol (blue rectangles) periodically; EGFP expression was profiled every 12 h. Cell density was adjusted to 1 × 10⁶ cells/mL every 48 h. (F and G) Resveratrol and antigen-triggered Jurkat_{RES-ind-CAR-EGFP} cell activation. Cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with CD19⁻ K562_EGFP cells (white squares) or CD19⁺ K562_mCherry cells (red squares) at an E:T ratio of 2:1. After overnight incubation, the surface expression of the T cell activation markers CD69 (F) and hIL-2 secretion (G) were quantified. (H) Dose-dependent CAR surface expression of the optimized RES_ind-CAR device in human primary T cells (RES_ind-CAR T cells). RES_ind-CAR T cells engineered with lentiviral pLF355 (LTR-P_ResR15-CAR-LTR) and pLF361 (LTR-P_mPGK-ResR₄-LTR) were cultivated in medium containing 0 to 15 μM resveratrol, and CAR expression was quantified by flow cytometry after 48 h. (I and J) Resveratrol-controlled cytokine release from RES_ind-CAR T cells. RES_ind-CAR T cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with CD19⁻ K562_EGFP cells (white squares) or CD19⁺ K562_mCherry cells (red squares) at an E:T ratio of 5:1. hIL-2 (I) and hIFN-γ (J) production were quantified after overnight incubation. (K) Resveratrol-inducible target cytotoxicity of RES_ind-CAR T cells. Cells were cultivated with various concentrations of resveratrol for 48 h and then incubated with a mixture of target cells containing CD19⁻ K562_EGFP cells and CD19⁺ K562_mCherry cells at an E:T ratio of 10:1. After 18 h, cells were collected for flow cytometry and cytotoxicity values were calculated based on ratios of CD19⁺ K562_mCherry cells to CD19⁻ K562_EGFP cells. The data in B–K are mean ± SD, n = 3. N.D., not detectable.

To characterize the performance of the RES_ind-CAR device in T cells and obtain an optimal RES_ind-CAR device, with low transgene expression leakage in the absence of resveratrol and high induction rates in the presence of resveratrol, we used the RES_ind-CAR device to control the transcriptional activation of a transgene cassette, comprising both enhanced green fluorescent protein (EGFP) and CAR, and optimized the resveratrol-inducible CAR-P2A-EGFP expression device in Jurkat T cells (Jurkat_{RES-ind-CAR-EGFP}) by measuring the EGFP expression level. Guided by a previous study (51), we optimized the resveratrol-inducible promotor and the constitutive promoters for driving ResR₁ expression with P_mPGK and P_hEF1α (human extension factor α gene promoter). We found that a combination of P_mPGK-driven ResR₁ and P_ResR11 [O_TRC1-P_hEF1a-(O_TRC1)₂]–driven CAR-P2A-EGFP showed the best induction performance in Jurkat T cells (SI Appendix, Fig. S14).

Jurkat_{RES-ind-CAR-EGFP} cells, which were transduced with the optimal RES_ind-CAR-EGFP device, exhibited excellent resveratrol dose-dependent control of CAR and EGFP expression, with extremely strong transgene induction at a low (15 μM) resveratrol concentration (Fig. 5B). Moreover, by observing changes in EGFP expression via flow cytometry, we found that Jurkat_{RES-ind-CAR-EGFP} cells possessed excellent long-time induction performance with negligible leakage and nonsaturating increases in EGFP expression over 14 d (Fig. 5C and SI Appendix, Fig. S15). Moreover, experiments in which we periodically added then removed resveratrol demonstrated that the device enables fully reversible transgene expression (Fig. 5 D and E and SI Appendix, Fig. S16). We also verified that the induced Jurkat_{RES-ind-CAR-EGFP} cells, which express CAR in a resveratrol dose-dependent manner, could be activated by CD19-positive K562 cells (CD19⁺ K562_mCherry), as assessed by CD69 induction (Fig. 5F) and hIL-2 production (Fig. 5G). We also observed that the RES_ind-CAR device showed CD19 antigen density-dependent response in Jurkat T cells (SI Appendix, Fig. S17). These results demonstrate that the RES_ind device can fine tune CAR expression and CAR-mediated cell activation in Jurkat T cells.

Resveratrol-Inducible CAR Expression in Primary Human T Cells.

Next, we further optimized the RES_ind-CAR device in primary human T cells with different configurations of resveratrol-inducible promotors and various mutants of the TtgR protein (SI Appendix, Figs. S18 and S19) guided by data from previously reported studies (44, 50, 52). A combination of P_hEF1α-driven ResR₄ (TtgR^H67A-KRAB) and P_ResR15 (O_TRC1-P_hEF1a-O_TRC1)–driven CAR showed the best induction performance and was used for further studies in primary T cells.

Primary T cells engineered with the optimized RES_ind-CAR device exhibited titratable CAR expression controlled by resveratrol (Fig. 5H) and resveratrol-controlled CAR-related cell activation upon antigen stimulation, as assessed by hIL-2 production (Fig. 5I) and hIFN-γ production (Fig. 5J). Cytotoxicity assays demonstrated that the cytotoxic efficacy of RES_ind-CAR T cells against CD19⁺ K562_mCherry is excellent and is tunable on resveratrol dosage (Fig. 5K). Moreover, CAR-induced T cell activation can be mediated by the RES_ind-CAR device in a time-dependent manner as well (SI Appendix, Fig. S20). In summary, we generated a RES_ind-CAR device to control transgene induction, CAR T cell activation, and cancer cell lethality in T cells.

Resveratrol-Controlled Cancer Cell Lethality Mediated by RES_ind-CAR T Cells in NCG Mice.

To evaluate antitumor efficacy mediated by the RES_ind-CAR device in vivo, we established two xenograft tumor mouse models. We first investigated whether the activity of RES_ind-CAR T cells could be regulated by resveratrol in a dose-dependent manner in a rapid xenograft mouse model of B cell leukemia (Fig. 6). In this model, unmodified T cells (Ctrl) or RES_ind-CAR T cells were injected i.p. after NCG mice were implanted with a mixture of CD19⁺ K562_mCherry and CD19⁻ K562_EGFP cells. These mice were then administrated with various doses of resveratrol (5 to 100 mg/kg/d) or vehicle control. Serum cytokine release was analyzed every 24 h and the cytotoxicity of RES_ind-CAR T cells was quantified by flow cytometry at 48 h (Fig. 6A). Mice injected with RES_ind-CAR T cells showed resveratrol-dependent depletion of the CD19⁺ K562_mCherry cell population (Fig. 6B). RES_ind-CAR T cells killed target cells in a resveratrol-inducible manner, with the strongest cytotoxicity observed with a resveratrol dosage of 10 mg/kg/d (Fig. 6C). Consistently, monitoring in vivo hIL-2 and hTNF cytokine release demonstrated that the activity of RES_ind-CAR T cells was tunable to the resveratrol concentration (Fig. 6 D–G).

Fig. 6.

Targeted cytotoxicity by primary human T cells engineered with RES_ind-CAR in NCG mice. (A) Experimental design for in vivo assessment of resveratrol-inducible cancer cell lethality by RES_ind-CAR T cells in a rapid xenograft tumor mouse model. NCG mice inoculated with CD19⁻ K562_EGFP cells and CD19⁺ K562_mCherry cells were injected intraperitoneally with 2 × 10⁷ unmodified primary human T cells (control) or RES_ind-CAR T cells and then administered with resveratrol or vehicle control at the indicated times. After 48 h, blood was drawn for cytokine analysis and peritoneal cells were quantified by flow cytometry to measure in vivo cytotoxicity. (B) Representative flow cytometry data show the resveratrol-inducible depletion of the CD19⁺ K562_mCherry cell population. (C) Quantified flow cytometry results of targeted cytotoxicity by RES_ind-CAR T cells. (D–G) In vivo hIL-2 (D and F) and hTNF (E and G) release of RES_ind-CAR T cells controlled by resveratrol at 24 h and 48 h. The data in B–G are mean ± SEM. P values were calculated with Student’s t test (n = 4 to 5 mice). *P < 0.05; **P < 0.01; ***P < 0.001 versus control. N.S., not statistically significant.

We further assessed antitumor efficacy mediated by RES_ind-CAR T cells in an aggressive leukemia xenograft model in NSG mice (Fig. 7). We engrafted mice with Renilla luciferase (Rluc)–expressing CD19⁺ K562 cells intravenously. After 5 d, mice received control unmodified T cells (Ctrl), conventional CAR T cells (Conv. CAR), or RES_ind-CAR T cells intravenously, followed by daily intraperitoneal (50 mg/kg) or oral (250 mg/kg) resveratrol (Fig. 7A). Serum cytokines were quantified after T cell infusion on day 3 and significant secretion of hIL-2 and hTNF was found in mice infused with RES_ind-CAR T cells and resveratrol (Fig. 7 B and C). Additionally, tumor burden was demonstrated by bioluminescence imaging, and the results demonstrated that tumor growth was significantly suppressed in mice that received RES_ind-CAR T cells with resveratrol treatment for 28 d (Fig. 7 D–F). A similar suppression was seen in mice that received conventional CAR T cells. The observed tumor flux in mice that received the RES_ind-CAR T cells in the presence of resveratrol was significantly lower than those that received the RES_ind-CAR T cells in the absence of resveratrol (Fig. 7F). In contrast, tumor flux in mice that received RES_ind-CAR T cells in the absence of resveratrol increased similarly to mice that received unmodified T cells. Furthermore, we observed prolonged survival in mice infused with RES_ind-CAR T cells and resveratrol (Fig. 7G).

Fig. 7.

In vivo efficacy of RES_ind-CAR T cells in CD19⁺ K562 xenograft mouse models. (A) Experimental design for xenograft mouse model in vivo studies. NSG mice were injected intravenously (i.v.) with Rluc-expressing CD19⁺ K562 cells (1 × 10⁶). After 6 d, mice were infused with unmodified T cells (Ctrl), conventional CAR T cells (Conv. CAR), or resveratrol-inducible CAR expression T cells (RES_ind-CAR), followed by daily intraperitoneal (50 mg/kg) or oral (250 mg/kg) administration of resveratrol. (B and C) Human cytokine hIL-2 (B) and hTNF (C) were measured at 3 d after T cell infusion. (D) Tumor burden images obtained by BLI of groups treated with: unmodified T cells (Ctrl), conventional CAR T cells (Conv. CAR), RES_ind-CAR without resveratrol, RES_ind-CAR with resveratrol by injection, and RES_ind-CAR with resveratrol by oral administration at different days as indicated. (E) Quantification of tumor burden on day 28 was plotted based on BLI total flux (photons per second) of each mouse shown in D using IVIS imaging. (F) The tumor burden of each group was monitored for up to 42 d after T cell infusion. Tumor burden was quantified as the total flux from the Rluc activity of each mouse using IVIS imaging. The data in B, C, E, and F are mean ± SEM. P values were calculated with Student’s t test (n = 4 to 5 mice). **P < 0.01; ***P < 0.001 versus control. (G) Kaplan–Meier curve representing the survival percentage of the experimental groups. P values were determined using two-sided log-rank test. N.S., not statistically significant.

Collectively, these results demonstrate that the cytotoxic activity of RES_ind-CAR T cells against tumor cells can be fine tuned by resveratrol in mice, suggesting potential applications for controlling CAR T cell activity in patients.

Discussion

Here, we report the development of resveratrol-controlled transgene expression devices (RES_rep and RES_ind) based on the TtgR regulatory device to preciously control transgene expression in mammalian cells and mice. We further demonstrate that the RES_rep and RES_ind devices are effective in controlling CAR expression in T cells to guide the recognition and eradication of tumor cells, both in vitro and in vivo. These resveratrol-controlled devices establish a proof of concept for cancer immunotherapy strategies based on the resveratrol-regulated repression or induction of immune cells.

In recent years, synthetic biology-inspired therapeutic programs have achieved favorable outcomes in the treatment of various diseases, including diabetes (53–55), insulin resistance (56), liver disorders (57), bacterial infections (58), and cancer (59–62); several of these treatments require effective, long-term therapeutic outputs, achieved by the sustained delivery of trigger molecules. The natural dietary molecule resveratrol, an antioxidant in red wine, has been a part of the human diet for thousands of years and is generally recognized as safe; resveratrol has already seen use in cosmetics and some healthcare products. This study demonstrates that RES_ind and RES_rep devices can achieve titratable, reversible, and long-term control of low concentration resveratrol-regulated transgene expression in mammalian cells and mice.

CAR T cell therapy is a paradigm-shifting therapeutic approach for cancer treatment that is achieving tremendous success in the clinic; however, its increasing utilization is revealing many unpredictable and/or uncontrollable side effects from on-target/off-tumor toxicity, including severe CRS (sCRS) and neurotoxicity (13), tumor lysis syndrome, and the eradication of normal cell types expressing low levels of an antigen (63). Our RES_rep-CAR expression device can be used to block the activity of infused CAR T cells, potentially being deployed to prevent toxicity-associated morbidity and mortality when patients are considered at a high risk for clinical complications in early-stage CAR T therapy. We verified that T cells engineered with RES_rep-CAR provide effective and titratable inhibition over T cell activation, while retaining their inducible therapeutic efficacy upon resveratrol removal. The pattern and extent of functional inhibition achieved by resveratrol in our study highlights the clinical potential of RES_rep-CAR.

We envision the RES_ind-CAR device to be implemented with patients who are in danger of developing severe responses to cell infusion. Our results demonstrate that T cells engineered with RES_ind-CAR can fine tune CAR expression levels and associated CAR-mediated cell activation, in vitro and in vivo, by adjusting administered resveratrol concentrations. In practice, the use of RES_ind-CAR in clinical trials could be coincident with monitoring cell infusion toxicity responses and real-time CAR T cells efficacy, enabling ongoing resveratrol-mediated CAR T cell activity adjustment to achieve highly precise, personalized, therapeutic outcomes.

Despite the excellent regulation performance of resveratrol-controlled transgene expression devices in vitro and in vivo, there is still room for improvement from a translational perspective. For example, given the low bioavailability of resveratrol by oral administration, the response sensitivity of resveratrol-controlled devices will require further engineering and optimization. We anticipate that more sensitive devices could be obtained by optimizing the operator sequences and directed evolution approaches, or via rational protein design to alter the transactivator and/or transrepressor properties. Furthermore, combination with our previously engineered and orthogonal systems (53) could enable multilayered control and even safer therapies.

With its broad potential uses, promising therapeutic impact for numerous diseases, and nutraceutical properties, resveratrol is a particularly suitable trigger molecule for clinical applications. Our work shows that resveratrol-controlled genetic devices can expand the toolbox of cybergenetics and highlights them as a safe, robust, and convenient strategy for the dynamic control of therapeutic outputs for future gene- and cell-based precision medicine. Considering that the resveratrol-responsive TtgR, a transcriptional repressor from P. putida, might present potential immunogenicity, the potential immunogenicity of RES_rep-CAR or RES_ind-CAR needs to be further evaluated before clinical trials. In conclusion, our proof of concept and initial in vivo application of two resveratrol-sensitive devices for fine tuning the activity of transgene expression and CAR T cell therapeutic activity illustrates how safe dietary molecules can usher in a new era of precision oncoimmunotherapies.

Materials and Methods

Generation of Stable Cell Lines.

The HEK_RES-rep-SEAP cell line, transgenic for resveratrol-repressed SEAP expression, was constructed by cotransfecting 5 × 10⁴ HEK293 cells with 190 ng of pLF51 (ITR-O_TRC1-P_hCMVmin-SEAP-pA:: P_hCMV-TtgR-VPR-P2A-PuroR-pA-ITR) and 10 ng of Sleeping Beauty transposase expression vector pCMV-T7-SB100 (P_hCMV-SB100X-pA). The HEK_RES-ind-SEAP cell line, transgenic for resveratrol-triggered SEAP expression, was constructed by cotransfecting 5 × 10⁴ HEK293 cells with 190 ng of pLF50 (ITR-P_ResR12-SEAP-pA::P_hCMV-TtgR-KRAB-P2A-PuroR-pA-ITR) and 10 ng of Sleeping Beauty transposase expression vector pCMV-T7-SB100. The HEK_{RES-ind-mut-SEAP} cell line, transgenic for resveratrol-triggered SEAP expression, was constructed by cotransfecting 5 × 10⁴ HEK293 cells with 190 ng of pLF365 (P_ResR12-SEAP-pA::P_hCMV-TtgR^H67A-KRAB-P2A-PuroR-pA-ITR) and 10 ng of Sleeping Beauty transposase expression vector pCMV-T7-SB100. After cultivation with 1 µg/mL puromycin for 2 wk, the monoclonal cell lines were selected and identified.

Lentiviral Transduction and Selection of Stable Cell Lines in Jurkat T and K562 Cells.

The Jurkat_{RES-ind-CAR-EGFP} cell line, transgenic for resveratrol-triggered CAR and EGFP expression, was constructed by cotransducing 1 × 10⁵ Jurkat T cells with two lentiviruses: pLF63 (LTR-P_mPGK-ResR₁-LTR) and pLF64 (LTR-P_ResR11-CAR-P2A-EGFP-LTR). Infected Jurkat T cells were cultured for 5 d and then incubated with 10 μM resveratrol for 24 h. EGFP-positive cells were selected based on EGFP fluorescence intensity by FACS-mediated cell sorting using a Becton Dickinson FACSAria Cell Sorter (BD Biosciences). After the EGFP selection, the transduced Jurkat T cells were expanded for further experiments. CD19-negtive K562 target cells, labeled with the EGFP (CD19⁻ K562_EGFP) cell line or CD19-positive K562 target cells labeled with the mCherry (CD19⁺ K562_mCherry) cell line, were constructed by transducing 1 × 10⁵ K562 cells with lentiviral pLF354 (LTR-P_hEF1α-EGFP-LTR) or pLF352 (LTR-P_hEF1α-CD19-mCherry-LTR), respectively. Infected K562 cells were expanded and selected based on EGFP or mCherry fluorescence intensities by FACS-mediated cell sorting using a Becton Dickinson FACSAria Cell Sorter (Becton Dickinson). A Rluc-expressing CD19⁺ K562 cell line was constructed by transducing CD19⁺ K562_mCherry cells with lentiviral pLF367 (LTR-P_hEF1a-Rluc-P2A-PuroR-LTR); the Rluc-expressing cell clones were selected and identified after incubation with 0.5 µg/mL puromycin for 2 wk.

Isolation, Culture, and Transduction of Primary Human CD4⁺ and CD8⁺ T Cells.

Human peripheral blood mononuclear cells (PBMCs) were isolated from the leukocyte concentrate of the blood from deidentified healthy donors (a byproduct from blood banks in the manufacturing of red blood cell and thrombocyte concentrates from anticoagulated whole blood; Shanghai Blood Center) using Ficoll gradients (Cedarlane Laboratories), as approved by the Medical Ethics Committee of Yangpu Hospital affiliated with Tongji University (approved ID: LL-2018-SCI-003). Primary human CD4⁺ and CD8⁺ T cells were then isolated from PBMCs using a CD4⁺ and CD8⁺ T cell isolation kit (Miltenyi, 130-045-201 and 130-045-101). T cells were cryopreserved in 90% Roswell Park Memorial Institute (RPMI) 1640 medium and 10% dimethyl sulfoxide (DMSO). Before lentiviral transduction, isolated T cells were thawed, cultured in serum-free hematopoietic cell culture medium X-Vivo 15 (Lonza, 04-418Q), and subsequently activated with anti-CD3/CD28 antibodies (Stem Cell Technologies, 10971) and 50 units/mL recombinant human IL-2 (Peprotech, AF-200-02) for 2 d. Activated primary T cells were resuspended with cell culture medium supplemented with 8 μg/mL polybrene and then transduced with concentrated lentivirus mixture with a multiplicity of infection (MOI) of ∼10 for each virus and spun for 90 min at 1,800 × g. Infected primary T cells were maintained at a density of ∼1 × 10⁶ cells/mL in X-Vivo 15 medium supplemented with 50 units/mL hIL-2 for 9 d, after which, they were rested and could be used in assays.

Quantification of CAR Expression on T Cells.

Jurkat T or primary human T cells were washed, stained with biotin conjugated anti-mouse CD19-specific monoclonal scFv antibodies (FMC63, Bioswan, R19PB-100), followed by streptavidin-phycoerythrin (PE) staining (Bioswan, R19PB-100), and finally stained with a live/dead viability stain. Stained cells were washed three times and then analyzed by flow cytometry. PE fluorescence was quantified by excitation at 561 nm and emission at 586 nm. The percentage of gated cells and their median fluorescence intensity (MFI) were analyzed by the FlowJo software. Weighted CAR expression was evaluated as the percentage of gated cells multiplied by their median fluorescence. Unmodified cells were used as controls.

Quantification of hIL-2 or hIFN-γ Production.

Cytokine release assays were performed using hIL-2 or hIFN-γ enzyme-linked immunosorbent assay (ELISA) kits. Engineered Jurkat T cells or human primary T cells were incubated with CD19⁺-K562_mCherry or CD19⁻-K562_EGFP target cells at an effector-to-target (E:T) ratio of 5:1 (5 × 10⁵ cells:1 × 10⁵ cells) in duplicate wells in 96-well flat-bottom plates. After incubation for 18 h, the supernatants were collected and analyzed for human hIL-2 or hIFN-γ by ELISA (BioLegend, 431804 or 430101).

Cytotoxicity Assay of RES_rep-CAR or RES_ind-CAR Devices in Engineered Primary Human T Cells.

CD19⁺ K562_mCherry cells and CD19⁻ K562_EGFP target cells were mixed at a 1:1 ratio and then incubated with primary human T cells expressing RES_ind-CAR or RES_rep-CAR at an E:T ratio of 10:1 (1 × 10⁶ cells:1 × 10⁵ cells) in 48-well plates. After overnight incubation, cells were washed in cell staining buffer and stained with a Fixable Viability Dye eFluor 660. Stained cells were washed three times and analyzed by BD LSRFortessa flow cytometry. Singlets were gated using forward versus side scatter and followed by gating based on forward scatter area (FSC-H) versus forward scatter height (FSC-A) characteristics. Live cells were gated via the allophycocyanin (APC) channel, and K562 target cells were gated by excluding T cells via the TexasRed channel (mCherry) versus the fluorescein isothiocyanate (FITC) channel (EGFP). Cytotoxicity was calculated by normalization to unmodified T cells (Ctrl). The normalized cytotoxicity was calculated by the following formula:

$$\mathrm{Cytotoxicity}\left(\%\right)=\left(1-\frac{\%\mathrm{CD}{19}^{-}-\mathrm{K}{562}_{\mathrm{EGFP}}\mathrm{cells}/\%\mathrm{CD}{19}^{+}-\mathrm{K}{562}_{\mathrm{mCherry}}\mathrm{cells}}{\%\mathrm{CD}{19}^{-}-\mathrm{K}{562}_{\mathrm{EGFP}}\mathrm{cells}\mathrm{in}\mathrm{control}\mathrm{wells}/\%\mathrm{CD}{19}^{+}-\mathrm{K}{562}_{\mathrm{mCherry}}\mathrm{cells}\mathrm{in}\mathrm{control}\mathrm{wells}}\right)\times 100$$

Animal Experiments.

SEAP production in mice.

The HEK_RES-rep-SEAP and HEK_RES-ind-SEAP stable cell lines engineered for RES-controlled SEAP repression and expression were microencapsulated into coherent alginate-poly-(L-lysine)-alginate beads (400 μm diameter; 200 cells/capsule) using a B-395 Pro Encapsulator (BÜCHI Labortechnik AG) according to the manufacturer’s instructions. Microcapsules were generated with the following parameters: 0.2-mm nozzle with a vibration frequency of 1,300 Hz, 25-mL syringe at a flow rate of 405 units, and 1.10-kV for bead dispersion. Ten-week-old male wild-type C57BL/6J mice (East China Normal University [ECNU] Laboratory Animal Center) were intraperitoneally implanted with 2 × 10⁶ microencapsulated cells suspended in 800 μL fetal bovine serum (FBS)-free Dulbecco’s modified Eagle’s medium (DMEM). The mice were then daily intraperitoneally injected with resveratrol at doses ranging from 0 to 500 mg/kg. Blood samples were collected, and SEAP levels were quantified in the serum by using a chemiluminescence-based assay kit (Roche Diagnostics GmbH, 11779842001).

Rapid xenograft tumor model in NCG mice.

Female immunodeficient NOD-Prkdc^scid Il2rg^em26/NjuCrl (NCG) mice, 6 wk of age, were purchased from Nanjing Biomedical Research Institute of Nanjing University (T001475). NCG mice were initially injected with 3 × 10⁷ target K562 cells (including 1.5 × 10⁷ CD19⁺ K562_mCherry cells and 1.5 × 10⁷ CD19⁻ K562_EGFP cells) intraperitoneally. Ten hours after cell implantation, engineered primary T cells (conventional CAR, RES_rep-CAR, or RES_ind-CAR) or unmodified T cells were intraperitoneally injected into mice. The mice were then administrated with resveratrol at doses ranging from 0 to 200 mg/kg/d by intraperitoneal injection, twice each day. Control mice were injected with the vehicle alone. Two days after resveratrol induction, blood samples were collected to determine in vivo cytokine release level using a Cytometric Bead Array (CBA) kit (BD Biosciences 558270 and 560112). Mice were killed by carbon dioxide asphyxiation followed by cervical dislocation. Peritoneal cells were collected based on a procedure described previously (64). Collected cells were centrifuged at 400 × g for 8 min and resuspended in 1 mL of 1× red blood cell lysis solution (Thermo Scientific, 00-4333) for 5 min at room temperature, centrifuged again at 400 × g for 8 min, stained with eFluor 660 conjugated live/dead viability stain, fixed in 4% paraformaldehyde, and finally washed twice and resuspended with cell staining buffer. Samples were analyzed with a BD LSRFortessa cytometer. Data analysis and target cell quantification were performed using FlowJo software (TreeStar), as described for in vitro cell killing experiments. Data plotting and statistical analysis (Student’s t test) were performed using Prism software (GraphPad).

CD19⁺ K562 xenograft tumor model in NSG mice.

Briefly, female immunodeficient NSG mice (6 wk old) were inoculated with 1 × 10⁶ Rluc-expressing CD19⁺ K562 cells intravenously. After 5 d, mice were randomly divided into five groups and each mouse was injected with 1 × 10⁷ unmodified T cells (Ctrl), conventional CAR T cells (Conv. CAR), or RES_ind-CAR T cells, intravenously. Resveratrol was administered intraperitoneally at a dose of 50 mg/kg or by oral administration at a dose of 250 mg/kg. Murine blood was drawn 3 d after T cell infusion, and the secretion of human cytokines in serum was measured using a CBA kit (BD Biosciences, 558270 and 560112). Tumor progression was regularly monitored by bioluminescent imaging (BLI) using an IVIS Lumina imager (PerkinElmer) and quantified as the photon flux over the entire mouse body. For in vivo imaging, each mouse was intraperitoneally injected with 6 mg/kg coelenterazine substrate solution (Yeasen Biotech, Chemical Abstracts Service [CAS] no. 55779-48-1] immediately before imaging.

Ethics.

All experiments involving animals were performed in accordance with the protocol approved by the ECNU Animal Care and Use Committee and in direct accord with the Ministry of Science and Technology of the People’s Republic of China on Animal Care Guidelines. The protocol used in this study was approved by the ECNU Animal Care and Use Committee (protocol ID: m20160305).

Statistical Analysis.

All in vitro data are represented as the mean ± SD of three independent experiments (n = 3). For the animal experiments, each treatment group consisted of randomly selected mice (n = 4 to 6). Blood sample analysis was blinded. Comparisons between groups were performed using Student’s t test, and the results are expressed as means ± SEM. Differences were considered statistically significant at P < 0.05. Prism 6 software (GraphPad Software, Inc.) was used for statistical analysis. n and P values are indicated in the figure legends.

Data Availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or SI Appendix, Supplementary Methods. All genetic components related to this paper are available with a material transfer agreement and can be requested from H.Y. (hfye@bio.ecnu.edu.cn).