Enhancing the safety of CAR-T cell therapy: Synthetic genetic switch for spatiotemporal control

Abstract

Chimeric antigen receptor T (CAR-T) cell therapy is a promising and precise targeted therapy for cancer that has demonstrated notable potential in clinical applications. However, severe adverse effects limit the clinical application of this therapy and are mainly caused by uncontrollable activation of CAR-T cells, including excessive immune response activation due to unregulated CAR-T cell action time, as well as toxicity resulting from improper spatial localization. Therefore, to enhance controllability and safety, a control module for CAR-T cells is proposed. Synthetic biology based on genetic engineering techniques is being used to construct artificial cells or organisms for specific purposes. This approach has been explored in recent years as a means of achieving controllability in CAR-T cell therapy. In this review, we summarize the recent advances in synthetic biology methods used to address the major adverse effects of CAR-T cell therapy in both the temporal and spatial dimensions.

Synthetic biology contributes to the realization of intelligent CAR-T cell therapies with enhanced safety and controllability.

INTRODUCTION

Adoptive T cell therapy is a prominent therapeutic method that demands precise T cell editing and signal regulation. This process involves ex vivo expansion of tumor-specific T cells followed by their reintroduction into patients to swiftly restore host immunity (1). Recent clinical trials have demonstrated that engineered T cells are highly active tumor effector cells. Among these, chimeric antigen receptor T (CAR-T) cell therapy has shown promising outcomes in the management of diverse malignancies, particularly hematological cancers. The activation intensity of CAR-T cells plays a crucial role in determining their effectiveness and is closely linked to CAR design. The initial generation of CAR exhibited suboptimal activation due to the absence of a second signal in the intracellular segment, which is vital for T cell activation. Subsequent modifications involved the incorporation of one or more costimulatory molecules, such as CD28 or 4-1BB, which substantially promote activation. However, the adverse effects were also more pronounced.

The safety concerns associated with CAR-T cells can be divided into the following three types: cytokine release syndrome (CRS), immune effector cell–associated neurotoxicity syndrome (ICANS), and on-target off-tumor (OTOT) toxicity. These adverse effects primarily result from unregulated activation and can be attributed to two factors, excessive activation of the immune system due to the uncontrolled timing and extent of CAR-T cell effects and toxicity to normal cells due to uncontrolled off-tumor phenomenon in space. Furthermore, toxicity is also associated with treatment-related factors, such as CAR-T cell dosage and drug conditioning chemotherapy, as well as factors related to patient heterogeneity, including tumor burden, the presence of underlying disease, and the levels of relevant biomarkers. At present, the use of the interleukin-6 (IL-6) receptor antagonist tocilizumab (2) and the IL-1 receptor antagonist anakinra (3) is the main clinical approach used to alleviate CRS and ICANS. However, while these medications inhibit the overactivation of CAR-T cells, they also suppress the normal immune function of the host’s endogenous T cells, potentially leading to diminished immune capability and increased susceptibility to infections. For OTOT toxicity, the ideal solution is to select strict tumor-specific antigens (TSAs) as targets in CAR design. However, the identification of such TSAs poses challenges, and most of the targets that are currently used in clinical practice still carry the risk of OTOT toxicity. Consequently, there is a pressing need for approaches to enhance the controllability of CAR-T cells while limiting their cytotoxic effects.

Synthetic biology is a burgeoning discipline that uses engineering principles to reengineer and modify natural biological systems while constructing novel standardized biological components and systems. Characterized by its systematic, modular, and standardized engineering features, synthetic biology offers theoretical guidance and technical support for engineered cell therapies. Given the intricate nature and controllability of the T cell activation signaling pathway, T cells and associated signaling systems serve as ideal platforms for synthetic engineering (4). Consequently, CAR-T cells represent a primary application of synthetic biology within the realm of immunology. In this review, we summarize the latest advances in the application of synthetic biology methods for controlling CAR-T cells both temporally and spatially to reduce adverse effects. In addition, we address the persisting challenges and outline future prospects in this field.

THE TOXICITY OF CAR-T CELL THERAPY

The common cause of adverse effects in patients receiving CAR-T cell therapy is uncontrolled activation. The effects can be divided into two groups based on temporal and spatial dimensions (Fig. 1).

Fig. 1. Overview of the toxicity of CAR-T cell therapy.

In the temporal dimension, CAR-T cells are overactivated to produce excessive amounts of cytokines, resulting in CRS. Subsequently, immune cells and cytokines from the periphery infiltrate the central nervous system, leading to ICANS. In the spatial dimension, CAR-T cells may identify target antigens on normal cells, resulting in OTOT toxicity. Figure created with BioRender.com.

In the temporal dimension, prolonged or excessive CAR-T cell activity can lead to an excessive release of cytokines and overactivation of the immune system. CRS is the prevailing toxic reaction (5, 6), and the incidence of grade 3 or above CRS has reached 30% in some clinical trials (Table 1). CRS arises because of high-level immune activation and is characterized by elevated levels of interferon-γ (IFN-γ), granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-2, IL-8, IL-5, IL-6, IL-10, or tumor necrosis factor–α (TNF-α) (7). Symptoms of CRS include fever, hypoxia, terminal organ dysfunction, cytopenia, hypotension, coagulation dysfunction, and hemophagocytic lymphohistiocytosis (8). Increased cytokine levels in the cerebrospinal fluid and destruction of the blood-brain barrier (BBB) are the causes of ICANS (9, 10). The reported incidence of ICANS is approximately 0 to 73%, and ICANS occurs simultaneously with or after CRS in most patients (Table 1) (11). Symptoms of ICANS include encephalopathy, cognitive deficits, speech disorders, seizures, and brain edema. Early alterations in cytokine profiles (IFN-γ, soluble gp130, IL-6, and soluble IL-6 receptor) or serum biochemical marker levels (ferritin and C-reactive protein) can be used to evaluate CRS and ICANS severity (12). Teachey et al. (13) developed a model to predict the occurrence of CRS by monitoring the levels of IFN-γ, IL-1, and MIP72α. ICANS biomarkers mainly include proinflammatory activation markers (IL-1 and IL-6), pro-proliferative markers (IL-15 and GM-CSF), immunomodulators (IL-10 and IFN-γ), and baseline systemic factors (14). Interventions that can reduce or completely inhibit CAR-T cell function when relevant biomarkers are detected are needed to avoid cytokine storms.

Table 1. Incidences of CRS and ICANS of CAR-T cell therapy overviewing most recent research (2022–2023).

GPRC5D, G protein–coupled receptor class C group 5 member D; BCMA, B cell maturation antigen; CD19, cluster of differentiation 19; CD22, cluster of differentiation 22; CD7, cluster of differentiation 7; CD20, cluster of differentiation 20; PSMA, prostate-specific membrane antigen; CLDN18, claudin-18; CLDN6, claudin 6; MET, mesenchymal epithelial transition factor; EpCAM, epithelial cell adhesion molecule.

Disease	Target	Incidence of CRS (%)	Incidence of ≥grade 3 CRS (%)	Incidence of ICANS (%)	Incidence of ≥grade 3 ICANS (%)	Clinical trial identifier	Refs.
Multiple myeloma	GPRC5D	100	0	0	0	NCT05016778	(115)
	GPRC5D	76	0	9	3	ChiCTR2100048888	(116)
	BCMA	88	5	15	3	NCT03651128	(117)
	BCMA	56	2	14	0	NCT04093596	(118)
	BCMA	94	10	3	0	NCT03815383NCT03751293NCT04295018NCT04322292	(119)
	BCMA	92	10	1	0	NCT03090659	(120)
	BCMA	35	6	0	0	NCT03093168	(121)
	BCMA	100	8	17	8	NCT04155749	(122)
	BCMA	76	1	5	0	NCT04181827	(123)
	BCMA	90	0	0	0	NCT04720313	(124)
	BCMA	94	28	39	11	NCT03502577	(125)
	BCMA	84	4	18	4	NCT03361748	(126)
	BCMA	76	7	37	2	NCT02658929	(127)
	CD19/CD22	27	0	7	0	NCT03919526	(128)
	CD19/CD22	50	15	5	5	NCT03448393	(129)
	CD19	96	24	28	28	NCT03825718	(130)
	CD19	50	17	25	8	NCT03573700	(131)
Acute lymphoblastic leukemia with central nervous system lymphoma	CD19	90	19	38	23	NCT02782351	(132)
Primary central nervous system lymphoma	CD19	58	0	50	8	NCT02445248	(133)
Acute lymphoblastic leukemia	CD7	83	0	0	0	NCT04538599	(134)
	CD7	100	0	67	0	ISRCTN15323014	(135)
	CD19/CD22	88	28	21	4	ChiCTR2000032211	(3)
	CD19/CD22	99	19	37	5	NCT04340154	(136)
Large B cell lymphoma	CD19	38	2	31	5	NCT03483103	(137)
	CD19	50	1	12	4	NCT03575351	(138)
	CD19	50	0	10	0	NCT03484702	(139)
	CD19	100	8	73	23	NCT03761056	(140)
	CD19	74	6	19	10	NCT04148430	(3)
	CD19	85	10	50	25	NCT03954106	(141)
	CD19	59	8	8	0	NCT03630159	(142)
	CD19	48	5	20	3	NCT04089215	(143)
B cell non-Hodgkin lymphoma	CD19/CD20	70	10	17	2	NCT03097770	(144)
	CD19	18	5	23	0	NCT03344367	(145)
B cell malignancies	CD19	21	0	0	0	NCT02807883	(146)
	CD19	43	0	29	0	NCT04206943	(147)
T cell malignancies	CD7	95	5	10	0	NCT04572308	(148)
	CD7	100	13	0	0	NCT04004637	(149)
Neuromyelitis optica spectrum disorder	BCMA	100	0	0	0	NCT04561557	(150)
Gastric cancer	CLDN18	95	0	0	0	NCT03874897	(151)
Metastatic castration-resistant prostate cancer	PSMA	60	30	10	10	NCT03089203	(152)
Solid tumors	CLDN6	46	5	5	0	NCT04503278	(153)
	MET	14	0	0	0	NCT03060356	(154)
Epithelial tumors	EpCAM	0	0	17	0	NCT02915445	(155)
Myasthenia gravis	BCMA	0	0	0	0	NCT04146051	(156)

In the spatial dimension, CAR-T cells may be localized and uncontrollably activated toward normal cells that express target antigens, resulting in OTOT toxicity. The ideal antigen should be expressed only on tumor cells (called TSAs, such as neoantigens), but few TSAs are currently available (15). Thus, tumor-associated antigens (TAAs), including human epithelial growth factor receptor-2 (HER2), epidermal growth factor receptor (EGFR), B7 homolog 3 (B7-H3), carbonic anhydrase IX (CAIX), mesothelin, and ganglioside 2 (GD2), are chosen as targets; however, all of these antigens exhibit a certain degree of expression in normal tissues (16, 17). Although the amount of TAAs expressed in normal cells is substantially lower than that in tumor cells, CAR-T cells can respond to low-expression antigens due to their high sensitivity. When exposed to antigens, infused CAR-T cells can undergo activation and proliferation by several orders of magnitude, potentially leading to severe toxicity. Table 2 presents a comprehensive list of representative clinical cases of OTOT toxicity during treatment. For instance, CAIX-targeted CAR-T cells recognize normal epithelial cells of the inner wall of the bile duct, thereby causing discrete cholangitis (18, 19); HER2-targeted CAR-T cells can lead to OTOT toxicity, such as mild skin itching and severe upper gastrointestinal bleeding (20); and EGFR-targeted CAR-T cells can cause mucosal and skin side effects (21–23). Therefore, the pressing challenge lies in achieving precise spatial localization of CAR-T cells within tumor tissues and mitigating adverse effects.

Table 2. Typical cases of OTOT toxicity of CAR-T cell therapy overviewing recent research.

HER2, human epidermal growth factor receptor 2; CEACAM5, carcinoembryonic antigen cell adhesion molecule 5 gene; CAIX, carbonic anhydrase IX; EGFR, epidermal growth factor receptor; CD133, cluster of differentiation 133.

Disease	Target	Off-tumor site	Result	Clinical trials identifier	Refs.
Colon cancer	HER2	Lungs and liver	High doses of CAR-T cells targeting HER2 on normal lung cells cause fatal cytokine storms	NCI-09-C-0041	(157)
Biliary tract cancers and pancreatic cancers	HER2	Gastrointestinal mucosa	Upper gastrointestinal hemorrhage	NCT01935843	(20)
Gastrointestinal malignancies	CEACAM5	Lungs	Respiratory toxicity	NCT01212887	(158)
Renal cell carcinoma	CAIX	Bile duct epithelium	Liver toxicity	DDHK97-29/P00.0040C	(18, 19, 159)
Biliary tract cancer	EGFR	Epidermis	Skin rashes	NCT01869166	(22)
Cholangiocarcinoma	CD133	Epithelium and vascular endothelium	Severe dermatologic, oral mucosal, and gastrointestinal toxicities	NCT02541370	(160)
B cell lymphoma and acute lymphoblastic leukemia	CD19	Normal B cells	B cell depletion, hypogammaglobulinemia	NCT00924326	(161)
Gastric cancer	CLDN18	Gastric mucosa	Mucosal erosion	NCT03874897	(151)

THE SECURITY GUARANTEES OF CAR-T CELLS IN THE TEMPORAL DIMENSION

T cell immunity in the human body can be regulated by activation or inactivation through various mechanisms, thereby enabling the balance of activity and safety. Consequently, the incorporation of a regulatory module into CAR-T cells could be used to achieve regulated activation, substantially enhancing the safety of this therapy while maintaining its effectiveness. The characteristics and advantages of synthetic biology include the combination of biology and engineering, which allows for the modularization, hierarchy, standardization, and quantification of biological functions. By using the logic-gating principles of electric circuits to combine standardized regulatory elements such as promoters, repressors, enhancers, and regulated genes, gene circuits can be created. The integration of these genetic circuits into T cells or the modification of chassis T cells can be used to effectively address safety concerns (Fig. 2).

Fig. 2. Synthetic biology methods can effectively improve the safety and controllability of CAR-T cells.

In the temporal dimension, the synthetic biology approaches for guaranteeing safety include killing switches, adapter switches, small-molecule drug switches, and chassis cell modification. In the spatial dimension, methods include Boolean logic gates (AND gates, NAND gates) constructed from synthetic receptors, redirected migration of T cells, and targeting the tumor microenvironment. Dual control of the temporal and spatial dimensions can be achieved by light-operated switches and ultrasound-operated switches. Figure created with BioRender.com.

Killing switches

The activation of killing switches is induced by specific small molecules, which initiate the programmed cell death pathway that results in irreversible termination of the immune response. Incorporating killing switches is among the straightforward and highly effective approaches for enhancing the controllability of CAR-T cells.

On the basis of their diverse underlying mechanisms, killing switches can be classified into three categories: inducible caspase-9 suicide gene system (iCas9) switches, ganciclovir-activated switches, and antibody-dependent cell-mediated cytotoxicity (ADCC) switches. Fusion proteins with modified human caspase-9 and human FK506-binding proteins (FKBPs) are the main components of iCas9 switches. Chemical inducers of dimerization (CIDs), such as AP1903/rimiducid or AP20187, can mediate FKBP12-F36V homodimerization, after which caspases send an apoptotic signal to the cell. iCas9 switches have shown high effectiveness and have been the subject of numerous clinical trials (Table 3). According to clinical data, CID can eliminate CAR-T cells in the peripheral blood and central nervous system, resulting in the prompt alleviation of adverse events such as graft-versus-host disease (GVHD) and CRS (24, 25). Furthermore, this control mechanism remains effective even when CAR-T cells undergo a sharp decrease after drug administration and subsequent reamplification over time (26). Ganciclovir-activated switches turn ganciclovir into a toxic metabolite using the herpes simplex virus-1 thymidine kinase (HSV-TK) protein, which leads to cell death. The administration of ganciclovir effectively eliminated CAR-T cells and suppressed their functionality in a xenograft model (27). Since HSV-TK is an exogenous virus-derived protein, the use of ganciclovir-activated switches may pose a safety concern regarding immunogenicity in contrast to iCas9 (28). Moreover, the activation of HSV-TK by ganciclovir in vitro requires a period of 3 days, which is considered relatively slow (29). HSV-TK can also be used as an enzyme-based reporter gene for positron emission tomography (PET) imaging (30), and it has been successfully used in a phase 1 clinical trial to visualize CAR-T cells within the bodies of seven glioblastoma patients (31). The combination of the HSV-TK reporter with synthetic Notch (synNotch) enables the imaging of CAR-T cells that specifically bind to their target antigens (32). A truncated protein such as a truncated human EGFR polypeptide (huEGFRt) that is recognized by monoclonal antibodies (mAbs) is coexpressed and helps make up ADCC switches. After the provision of mAbs, CAR-T cells can be eliminated by ADCC or complement-dependent cytotoxicity (CDC). JCAR017 is a CAR-T cell product containing huEGFRt that can be recognized by cetuximab and rendered ineffective; this product showed superior safety in the TRANSCEND NHL 001 trial (NCT02631044) (33). The ADCC switches often take longer to work than the killing switches discussed above. In addition, the ability of cetuximab to penetrate the BBB is highly restricted (34), which could restrict the ability of ADCC switches to alleviate CRS and ICANS. Moreover, lymphodepletion is administered to patients before CAR-T cell therapy in current clinical practice, aiming to reduce the circulating immune cell count and enhance the antitumor efficacy of CAR-T cell therapy. Therefore, this switch is not applicable in such circumstances (35).

Table 3. Examples of applying synthetic biology methods to enhance the safety of CAR-T cell therapy for spatiotemporal regulation.

Dimension	Category	Type	Phase	Evidence of mitigating CRS/ICANS/OTOT toxicity	Refs.
Security guarantee in the temporal dimension	Killing switches	iCas9 switch	Phase 1/2 NCT03373097	Following the administration of rimiducid, a precipitous decline in the peripheral blood CAR-T cell count was observed, and this reduced level was sustained after the subsequent dose. After 6 weeks, CAR-T cells began to reamplified and could still be controlled by rimiducid.	(26)
			Phase 1/2 NCT03016377	After administering rimiducid, there was a notable improvement in ICANS symptoms, coupled with the removal of >60% of the CAR-T cells from the circulation within 4 hours and >90% within 24 hours.	(162)
			Phase 1 NCT01494103	Approximately 85–95% of CD3+CD19+ T cells in the bloodstream and over 90% in the cerebrospinal fluid were eradicated subsequent to the administration of CID. GVHD-related CRS were relieved within 2 hours.	(24)
			Phase 1 NCT03170141	The safety of CAR-T cells is considered acceptable. iCas9 switch was armored but not used.	(163)
			Phase 1 NCT04196413	The safety of CAR-T cells is considered acceptable. iCas9 switch was armored but not used.	(164)
			Phase 1 NCT01822652	The safety of CAR-T cells is considered acceptable. iCas9 switch was armored but not used.	(165)
			Phase 1/2 NCT02414269	The safety of CAR-T cells is considered acceptable. iCas9 switch was armored but not used.	(166)
		ADCC switch	Phase 1 NCT02631044	The safety of CAR-T cells is considered acceptable. ADCC switches were armored but not used.	(33)
			Phase 1 NCT02311621	The safety of CAR-T cells is considered acceptable. ADCC switches were armored but not used.	(167)
			Phase 1 NCT03618381	The results of the clinical trial have not been disclosed yet.	/
			Phase 1 NCT02498912	The results of the clinical trial have not been disclosed yet.	(168)
		Ganciclovir activated switch	Phase 1BB-IND 9149	An immune response against CAR-T cells emerged, limiting the efficacy and persistence of subsequent infusions.	(169)
			Phase 1 NCT00730613	The safety of CAR-T cells is considered acceptable. Hy-TK killing switch was armored but not used.	(170)
			Phase 1/2 NCT04097301 terminated	The clinical trial has been terminated. HSV-TK killing switch was armored but not used.	(171)
			Preclinical study	The administration of ganciclovir can effectively eliminate CAR-T cells in vivo and suppresses their functionality in mice.	(27)
	Adapter switches	FITC-folate adapter switch	Phase 1 NCT05312411	The results of the clinical trial have not been disclosed yet.	/
			Preclinical study	It can quickly terminate CRS-like toxicity within 3 hours and preemptively prevent the occurrence of CRS-like toxicity.	(39)
		CD123-specific targeting module	Phase 1 NCT04230265	The clinical trial results were unpublished. TM123 UniCAR-T exhibited reversible toxicity in mice, unlike CD123 CAR-T with notable hematological toxicity.	(52)
	Small-molecule switches	On switch	Phase 1/2 NCT02744287 Suspended	The CAR-T cells equipped with the iMC on switch exhibited rimiducid-dependent activity. Nevertheless, clinical trials were halted owing to pronounced adverse reactions.	(72)
			Preclinical study	The activity of CD19-DARIC T cells in vivo is completely dependent on rapamycin-mediated control.	(61)
		Off switch	Preclinical study	Precise modulation of CAR expression levels using small molecules enables reversible control of CAR-T cell activity in vivo.	(77)
	Chassis cell modification	Metabolic engineering modification	Preclinical study	Two weeks following administration of CAR-T cells in the absence of exogenous uridine, cell counts decreased substantially in vivo.	(83)
Security guarantee in the spatial dimensions	Boolean logic gates constructed by synthetic receptors	AND gate (synNotch)	Phase 1 NCT05617755	The findings from the clinical trial have not been published. However, in preclinical investigations, it has demonstrated robust antitumor efficacy and exhibited a high degree of specificity.	(172)
		A AND NOT B gate (CAR + iCAR)	Phase 1/2 NCT05736731	The findings from the clinical trial have not been published. Preclinical study showed that engineered T cells selectively kill HLA-A*02–negative cells but not HLA-A*02–expressing cells in vitro and in vivo.	(94)
	Redirected migration of T cells	Cytokine-operated	Phase 1 NCT04153799	The results of the clinical trial have not been disclosed. The in vitro preparation method for CAR-T cells incorporating CXCR5 appears to be feasible.	(173)
			Preclinical study	The T cell infiltration elicited by CXCR2-expressing T cell therapy exhibited a more than twofold increase compared to the control T cell therapy.	(98)
		Orthogonal receptor-drug pair	Preclinical study	T lymphocytes can be specifically localized and last for at least 7 days in vivo.	(100)
		Light-operated	Preclinical study	The infiltration of T lymphocytes at the tumor site was substantially enhanced upon local light stimulation.	(101)
	Targeting tumor microenvironment	Trending toward growth-inhibitory acidic TME	Phase 1/2 NCT03393936	No dose-limiting toxicities have been observed to date with no indications of OTOT toxicity attributed to either product.	(174)
		Dual oxygen-sensing switch	Preclinical study	A preclinical study had shown that this hypoxia-regulated CAR expression system substantially improves its antitumor specificity.	(104)
Dual security guarantee in the temporal and spatial dimensions	Light-operated switches	UV-sensitive photolyzable switch	Preclinical study	A preclinical study had shown that the activity of CAR-T cells is in a dose-dependent manner by the mediator and time-spatial controlled by UV.	(108)
	Ultrasound-operated switches	Focused ultrasound switch	Preclinical study	The effectiveness of FUS-CAR-T cells to non–UV-irradiated tissues (expressing target antigens) was only 9.8%, while that of the standard CAR-T cells was 99.8%.	(112)

The killing switch, which is a well-established synthetic biology approach, has been implemented in numerous clinical trials (NCT00710892, NCT03373097, NCT02631044, etc.). However, there are still critical issues that need to be addressed. First, the inducer (AP1903 or AP20187) for the iCas9 switches used in clinical trials has not been approved for sale. This limitation restricts the clinical application of the suicide gene system, making the selection of commercially available drugs a more practical choice. The RapaCaspase9 suicide gene uses the readily available marketed drug rapamycin and has demonstrated favorable outcomes in preclinical experiments, indicating promising potential for future clinical applications (36). Second, upon activation, these switches initiate an irreversible death mechanism, posing considerable challenges for both the high manufacturing costs and the prolonged persistence of CAR-T cells after treatment. In addition, suicide genes originating from nonhuman sequences may increase the immunogenicity risk of CAR-T cells, inducing both humoral and cellular immune responses (37). This intensifies therapeutic risks and the regulatory burden for post-administration patient monitoring. Compared to virus-derived HSV-TK, iCas9, which is of human origin, carries a reduced risk of immunogenicity. Clinical reports indicate that T cells harboring the iCas9 suicide gene in the human body persist for an extended period (38).

Adapter switches

The adapter molecule functions as a bridging molecule, binding one end to the universal CAR on CAR-T cells and the other end to the antigen on tumor cells. The “on” or “off” state and the degree of the effect are regulated by adjusting the presence and quantity of adapter molecules. Lee et al. (39) demonstrated that by regulating the concentration and dosing of the adapter molecule, CRS-like toxicity can be quickly terminated or prevented. Furthermore, adapter switches enable swift target switching, and the associated CAR-T cells demonstrate a high level of universality, thereby improving programmability and reducing therapeutic costs.

Biological orthogonal pairs are the foundation of adapter switches and provide the adapter molecule with a high degree of specificity for binding to CARs. The split, universal, and programmable (SUPRA) CAR serves as an exemplary adapter switch system. The zipFv functions as an adapter molecule that consists of a single-chain variable fragment (scFv) that specifically binds to tumor cells and a leucine zipper that interacts with the zipCAR. This enables simultaneous binding to both tumor cells and engineered T cells expressing zipCAR, which leads to the activation of cytotoxic functions (40). A major advantage of this switch is that when a higher-affinity zipFv is added, it can compete with the original zipFv, thereby converting the switch to off (41). Another classic example of a biomolecular orthogonal pair is the fluorescein isothiocyanate (FITC)–anti-FITC system, which achieves specific recognition of tumor antigens and anti-FITC CAR by coupling FITC with antibodies (42–46). The FITC-folate adapter switch is a relatively mature design and has undergone evaluation in a phase 1 trial for osteosarcoma (NCT05312411). In addition, the biotin-avidin system can specifically bind biotin-labeled antibodies to CARs equipped with avidin (47, 48). The PNE-mAb 52SR4 system links the peptide neoepitope (PNE), derived from the GCN4 peptide sequence, with a specific antibody. This enables the system to identify CARs that carry the mouse mAb 52SR4 in an scFv format (46, 49, 50). The bacterial toxin-antitoxin barnase-barstar system is a biomolecular orthogonal pair that can be applied in CAR-T cell therapy. The engineered chimeric receptor, which incorporates barstar, is designed to specifically bind to the tumor-targeted barnase (51).

Adapter switches demonstrate promising prospects for practical applications. The CD123-specific targeting module (TM123) is an adaptor that connects the universal CAR-T cell platform and CD123-positive leukemia cells (52). Meyer et al. developed a TM123 variant (TM123-4-1BBL) that can specifically bind to CD123-positive acute myeloid leukemia (AML) cell lines, universal CAR, and the 4-1BB costimulatory receptor, which can enhance persistence and effector function. This switch improves the controllability of the system, which may misrecognize normal hematopoietic progenitor cells and endothelial cells expressing CD123 (52, 53). The receptor tyrosine kinase–like orphan receptor 1 (ROR1) has up-regulated expression in bone marrow stromal cells after chemotherapy and is prominently expressed in the lungs. Consequently, conventional CAR-T cells targeting ROR1 have been found to induce lethal bone marrow failure and pulmonary toxicity in mice (54, 55). Switchable CAR-T (sCAR-T) systems provide a systematic and precisely controllable system for CAR-T cell targeting of ROR1 through the use of the GCN4 peptide–tagged (recognizing mouse mAb 52SR4 in scFv format) antigen-binding fragment (targeting tumor antigens) adapter molecule (56).

On the basis of the adapter switches, logic gates can also be introduced. For instance, the “AND” gate ensures that CAR-T cells are exclusively activated when both externally administered adapters are present, and simultaneous recognition of the two target antigens occurs (40). In the case of the A AND (NOT B) gate, the affinity of adapter A for adapter B surpasses its affinity for the receptor on T cells. Consequently, when both antigens A and B exist on normal cells, adapter B binds to adapter A, thereby impeding its ability to activate T cells (40). Cho et al. (57) developed a SUPRA CAR system capable of executing a three-input (A AND B) AND NOT C logic gate, which demonstrated exceptional controllability by conducting intricate biocomputation. The latching orthogonal cage–key protein (LOCKR) switches use the binding between the “cage protein” and the “key protein” to adopt an active conformation. This conformation, in turn, binds with CAR, allowing for the implementation of three-input (A AND B) AND NOT C logic (58). However, notably, a system constructed of exogenous substances and multitarget antigens may cause immunogenicity and a greater risk of off-target toxicity.

Small-molecule switches

By harnessing the advantages of genetic engineering and orthogonal chemical tools, researchers have developed small molecule–based safety switches. These safety switches are designed using small-molecule drugs, including proteolytic targeted chimeric (PROTAC) compounds (59), rapamycin (60–62), lenalidomide (63), dasatinib (64, 65), and grazoprevir (55). These switches can be divided into three categories: on, off, and “on/off” dual switches. Small-molecule switches not only allow flexible control over the activation and inactivation states of CAR-T cells but also have the potential to facilitate rest periods and prevent or reverse T cell exhaustion induced by tonic CAR signaling (55, 66). Moreover, through the control of drug dosage, targeting tumor cells characterized by high antigen expression while preserving healthy tissues with lower antigen expression becomes possible (55). Through comprehensive exploration of small-molecule switches, leakage activity tends to decrease when the system is turned off, accompanied by an expanding dynamic range between the on and off states.

The classical on switch system is the chemically induced dimerization (CID) system. CAR is inactive because of the dispersion between subunits, and units can dimerize to form a complete CAR. When both the recognition of the target antigen and the presence of dimeric drugs are satisfied, CAR-T cells can achieve killing. Other systems include the rapamycin-FKPB12-FRB system (60, 61), the Al120-RS3-hRBP4 system (67), and the lenalidomide-CRBN-IKZF3 system (63). Notably, rapamycin has potent immunosuppressive properties (68, 69), which may partially diminish the effectiveness of CAR-T cell therapy. Currently, there are several structurally modified rapamycin compounds aimed at reducing immunosuppressive activity (70). However, as an mTORC1 inhibitor, rapamycin has been reported to promote the accumulation of CXCR4 and potentially promote the infiltration of CAR-T cells (71). GoCAR-T cells, which target prostate stem cell antigen (PSCA) and are armed with a rimiducid-inducible MyD88/CD40 costimulation switch, demonstrated rimiducid-dependent activity (72). However, clinical trials were halted because of severe adverse reactions. Furthermore, the tet-tetracycline system serves as an on switch. After addition, tetracycline binds to the DNA binding domain rtetR, causing rtetR to recognize its binding sequence tetO and leading to activation of the promoter for the initiation of CAR expression (73, 74). In addition, Yamada et al. (75) achieved dual control of blue light and tetracycline through the Cry2-CIB1 light-induced binding switch. Sahillioglu et al. (76) developed a universal on switch for antigen receptors that has immunoreceptor tyrosine-based activation motif (ITAMs) without the need for covalent modification.

Off switches often use small molecules to degrade and inactivate the CAR. Richman et al. (77) fused CAR with the ligand-induced degradation (LID) domain, which contains a hidden degron that is exposed when specific small-molecule ligands are administered, resulting in proteasomal degradation of CAR-LID fusion proteins and inhibition of CAR-T cell activity. Juillerat et al. developed a framework controlled by proteases and protease inhibitors in which the protease target site is cleaved and CAR is preserved and functional in the absence of the protease inhibitor asunaprevir (ASN). In contrast, treatment with ASN leads to CAR degradation via a degron-mediated proteolysis pathway due to the inhibition of hepatitis C virus (HCV) nonstructural protein 3 (NS3) protease cleavage activity (78). Similarly, lenalidomide (63) or PROTAC (59) can be used to induce CAR protein hydrolysis. Moreover, other off switches, such as the tyrosine kinase inhibitor dasatinib, can reduce the cytotoxicity and proliferation capacity of CAR-T cells (64, 65). Because of its short half-life, the transition between the on and off states can be more controllable (79). In addition, Mitchell’s team found that the in situ coupling of polyethylene glycol (PEG) to the surface of CAR-T cells (via PEGylation) resulted in the formation of a polymeric spacer, which inhibited cell-to-cell interactions. This process mitigated CRS and neurotoxicity (80).

In addition, dual on/off switches can be constructed. Duong et al. (81) designed the dual-switch system using a rimiducid-induced activation switch and a rapamycin-induced apoptosis switch. This drug-responsive dual-switch system led to more controlled regulation compared to that of the single-switch system, as the dual-switch system enables the simultaneous maintenance of the “activated” state of CAR-T cells, facilitating robust proliferation and prolonged presence while also enabling the “inactivated” state, which effectively attenuates adverse effects through the integration of safety switches. However, a clinical trial of BPX-603, which was equipped with a rimiducid-inducible MyD88 and CD40 (iMC) on switch and an iCas9 suicide switch, was suspended because of toxicity issues (Table 3). In addition, Khalil and colleagues (82) developed a toolkit containing 11 programmable synthetic transcription factors. These transcription factors can be selectively activated on demand using US Food and Drug Administration (FDA)–approved small-molecule inducers to trigger specific cellular programs such as proliferation and antitumor activity. This selective activation allows for precise temporal control of CAR-T cell behavior.

Chassis cell modification

Chassis cells are the “hardware” and cornerstone of synthetic biology. Engineering the metabolic network of chassis cells allows them to achieve the desired functions. Wiebking et al. knocked out the uridine monophosphate synthase (UMPS) gene in T cells, rendering their proliferation dependent on exogenous uridine. By modulating the availability of uridine, the proliferation of T cells was effectively regulated, as evidenced by experimental findings both in vitro and in vivo (83). Furthermore, future work will involve the introduction of synthetic biology gene networks to construct chassis cells with multi-stability, which will allow CAR-T cells to mutually switch and self-regulate between various stable states (84).

THE SECURITY GUARANTEES OF CAR-T CELLS IN THE SPATIAL DIMENSION

Boolean logic gates constructed by synthetic receptors

The function of synthetic receptors depends on the outer membrane domain (determining the target cell type) and the intracellular domain (determining the signal type, positive or negative, and signal strength). Combining synthetic receptors with different functions can create Boolean logic gates such as the AND gate and “NAND” gate, thereby enabling precise tumor tissue targeting and reducing OTOT toxicity (Fig. 2).

The AND gate serves as an activation mechanism only when two distinct target antigens are recognized simultaneously. The classic AND gate is the synNotch receptor system. Unlike CAR receptors, which directly initiate T cell activation upon antigen binding, synNotch receptors recognize target antigens via extracellular recognition domains and undergo cleavage and the subsequent release of transcriptional activation domains to induce CAR expression, thereby targeting the second antigen. This system can selectively remove target cells containing both antigens to mitigate the risk of OTOT toxicity (85). Furthermore, research has shown that synNotch CAR circuits can inhibit CAR-mediated tonic signaling, thereby promoting the maintenance of long-lived memory and a nonexhausted phenotype (86). Zhu et al. developed a class of synthetic intramembrane proteolysis receptors that can optimize the sensitivity and strength of CAR-T cells, which provided modular design strategies for synNotch design. This receptor has the advantages of a compact structure, high tunability, and a low risk of immunogenicity due to the use of human proteins (87). Furthermore, the synNotch receptor can also drive the production of IL-2, thereby minimizing systemic IL-2 toxicity as much as possible (88). The potential risk of OTOT toxicity persists when healthy tissues expressing the antigens targeted by CAR are located in close proximity to tumor cells (54). Constructing synNotch CARs with high and low affinity that target the same antigen can enhance the discrimination between normal and tumor cells based on differences in antigen expression levels (89). Furthermore, a SPLIT CAR can be generated by combining a low-affinity CAR with a chimeric costimulatory receptor (CCR). In this construct, the CAR includes only the CD3ζ chain, while the CCR exclusively contains the costimulatory domain. Both receptors necessitate the simultaneous recognition of the target antigen to elicit a robust T cell response at full intensity (90). AND gates for multiple input signals have also been developed. Sukumaran et al. (91) identified three chimeric receptors: PSCA, IL-4, and transforming growth factor–β (TGF-β). Drawing on a comprehensive understanding of the complex T cell signaling network, Tousley et al. devised a Boolean logic AND-gated CAR-T cell system by combining LAT and SLP-76. This groundbreaking design represents an inaugural approach that is capable of promptly, directly, and reversibly modulating activity, and this approach is strictly confined by the recognition of dual antigens (92).

The NAND gate, also known as the A AND (NOT B) gate, is designed to activate when it identifies target cells expressing the A antigen but not the B antigen. This gate is composed of a CAR and an inhibitory CAR (iCAR) that incorporates CTLA-4 or PD-1. CAR-triggered T cell responses can be effectively inhibited after the iCAR recognizes the target antigen, and this inhibition is dynamically reversible. When the iCAR dissociates from its ligand, the activation signal of the CAR is restored, causing the T cell response to restart (93). For example, CAR targeting carcinoembryonic antigen (CEA) (94) or mesothelin (MSLN) (95) and iCAR targeting human leukocyte antigen–A2 (HLA-A2) are used to target tumor cells with down-regulated major histocompatibility complex (MHC) expression and the loss of heterozygosity (LOH) (tumor cells without A02). Witte and colleagues (96) engineered a dual-inhibitory domain CAR (DiCAR) by incorporating two immune cell inhibitory signaling domains. The purpose of this design was to finely regulate the cytotoxicity of CAR-T cells, thereby enhancing the inhibitory efficacy of CAR-T cells in comparison to that of an iCAR equipped with a single PD-1 domain. Notably, the affinity of iCAR needs to be within an appropriate range. If the affinity of the iCAR is too high and if there is a small amount of expression in tumor cells, there is a possibility of inadvertent inhibition of tumor cells as well.

CAR-T cells composed of multiple synthetic receptor Boolean logic gates can more accurately localize to tumor tissues and reduce the effect of OTOT toxicity. However, this approach also faces several challenges. First, these systems require the expression of various genes, and larger genome sizes can lead to a notable reduction in the efficiency of transgenic integration (97). Second, the potential immunogenicity of non–human-derived components, such as viral transcription factors, is a major clinical hurdle. Increasing the number of recognized antigens increases the probability of immune escape, and tumors need to change only one of the expression patterns of multiple antigens to prevent killing. Attaining an efficacious immune response against tumors entails a delicate equilibrium between safety and efficacy, necessitating the careful consideration of both factors when addressing safety concerns.

Redirected migration of T cells

In addition to the aforementioned approaches, synthetic biology techniques can also be used to genetically engineer CAR-T cells, thereby allowing for their redirected migration toward tumor sites (Fig. 2). Peng et al. (98) introduced the natural chemokine receptor C-X-C motif chemokine receptor 2 (CXCR2) into engineered T cells to make them tend to melanoma cells that expressed chemokines C-X-C motif chemokine ligand 1 (CXCL1) and C-X-C motif chemokine ligand 8 (CXCL8). Moon et al. (99) introduced the chemokine receptor C-C chemokine receptor type 2b (CCR2b) into CAR-T cells to make them tend to malignant pleural mesotheliomas (MPMs) that express the chemokine C-C motif ligand 2 (CCL2). Native receptors for natural ligands may have the problem of insufficient tumor cell specificity, which can be solved by orthogonal receptor-drug pairs composed of synthetic tropism receptors and small-molecule ligands. The synthetic tropism receptors developed by Park et al. (100) tend to bind to the bioinert small-molecule clozapine-N-oxide. Xu et al. (101) developed a photoactivatable-chemokine receptor (PA-CXCR4) to mediate light-induced chemokine signaling activation and guide cell migration to light-stimulated sites. Furthermore, Vincent et al. (102) used tumor-colonizing multifunctional bacteria (probiotics) to generate synthetic antigens within the tumor and release chemotactic factors. This approach directs CAR-T cells specifically toward the tumor site.

Targeting the tumor microenvironment

As the tumor microenvironment is characterized by hypoxia, modifying CAR-T cells to function only in a hypoxic microenvironment can substantially minimize the potential for OTOT toxicity and broaden the range of target antigens exploitable for therapeutic applications (Fig. 2). CAR-T cells can sense the hypoxic environment through hypoxia-inducible factor 1α (HIF1α) and subsequently induce CAR expression to recognize the antigen. CAR expression stops rapidly when the hypoxic environment signal is removed (103). In addition, HypoxiCAR is a dual oxygen sensing system based on the HIF1α-mediated CAR expression system in which the oxygen-dependent degradation domain (ODD) of the HIF1α is added to the CAR to achieve CAR degradation under normoxia and CAR expression under hypoxia (104). However, certain organs in the body may share similar hypoxic characteristics with the tumor microenvironment, such as the intestine (105) and kidneys (106), making it particularly important to consider these tissues when selecting targets. In addition, tumors frequently secrete proteases to facilitate invasion and promote tumor development. Incorporating a linker sensitive to tumor-specific proteases, along with a masking peptide, can hold the CAR in an inactive state. CAR-T cells are activated only when they are within the tumor microenvironment containing these specific proteases, leading to the destruction of tumor cells (107). Moreover, the products CCT301-38 and CCT301-59 transform the growth-inhibitory acidic tumor microenvironment into an activating signal, thereby reducing the risk of OTOT toxicity. These interventions are currently being investigated in the clinical trial NCT03393936 (Table 3).

THE DUAL SECURITY GUARANTEES OF CAR-T CELLS IN THE TEMPORAL AND SPATIAL DIMENSIONS

There are also methods for guaranteeing dual security in temporal and spatial dimensions, including light-operated switches and ultrasonic-operated switches (Fig. 2). Light-responsive switches offer precise control over the activation, localization, and inactivation of CAR-T cells through the regulation of the presence, distribution, and intensity of light. Zhang et al. developed an ultraviolet-sensitive photolyzable molecule (FITC-O-folate) adapter system to realize dual control by an adapter molecule and light, in which the FITC and folic acid groups are connected by a photocleavable o-nitrobenzyl ester structure. The FITC structure can bind to T cells loaded with anti-FITC receptors, and the folate structure can bind to the folate receptors of tumor cells. Exposure to 365-nm light results in the termination of CAR-T cell activation due to disruption of the photocleavable o-nitrophenyl ester structure (108). Moreover, a cleavable off switch adapter can be used to facilitate the cleavage of biotin upon exposure to ultraviolet light, thereby impeding its recognition by CAR-T cells (109). Huang et al. (110) developed a light-inducible nuclear translocation and dimerization system to achieve spatiotemporal control of CAR-T cells. Furthermore, the heat generated from light can also drive CAR expression through the utilization of a photothermal switch (111). One of the greatest challenges in using light-operated switches is the poor penetration ability of light, which may limit their practical application in deep-seated tumors or tissues. Furthermore, the restricted functionality of light-operated switches to primary or local sites may limit their application, especially considering that most patients undergoing CAR-T cell therapy typically present with metastatic diseases involving multiple and hidden metastatic sites.

Ultrasound-operated switches have similar advantages as light-operated switches, but they have the added benefit of deeper penetration. Ultrasound waves have a greater ability to penetrate tissues and reach deeper parts of the body, which limits light wave penetration. Wu et al. activated the expression of heat shock proteins through heat energy conversion by ultrasound, thereby activating the expression of CAR through the heat shock protein promoter. These switches can be directly controlled by magnetic resonance imaging (MRI)–guided focused ultrasound (FUS) without any exogenous cofactors. Short-pulse FUS stimulation enables precise and spatiotemporal control of CAR-T cells at specific locations and desired time points (112).

CONCLUSION AND FUTURE PROSPECTS

The rapid advancement of synthetic biology has facilitated the development of numerous sensors and controllable gene circuits for engineering and enhancing immune cell functions beyond the capabilities of the natural immune system. By manipulating molecular components, synthetic biology enables the robust and modular control of the cellular behavior of CAR-T cells, thereby offering innovative solutions for achieving specific immune cell functions inspired by natural systems. Overall, these functions can be regulated in a self-controlled and user-controlled paradigm, which is realized by automatic cell execution behavior (such as automatic activation upon recognition of target cells) and human manipulation through external factors (such as the utilization of small molecules and light). On the basis of engineering, modular, and systematic characteristics, synthetic biology is undoubtedly of great value in enhancing safety. Recent efforts have concentrated on applying synthetic biology methods to address safety concerns in CAR-T cell therapy across the temporal and spatial dimensions (Fig. 3). Numerous synthetic biology approaches involving CAR-T cells have demonstrated promising outcomes in clinical trials. As synthetic biology continues to develop and technology matures, we anticipate broader clinical applications in the coming years.

Fig. 3. The toxicity of CAR-T cell therapy hinders its further application.

In this review, we categorize the adverse effects associated with CAR-T cell therapy into two domains and provide a comprehensive overview of recent advancements in using synthetic biology approaches to address safety concerns. Figure created with BioRender.com.

Several challenges remain in the application of synthetic biology to enhance the safety of CAR-T cells. First, the immunogenicity caused by nonhuman substances, such as nonhuman transcription factors or proteins with specific functions derived from viruses, bacteria, or other animals and plants, is still worthy of vigilance. Vigilance is needed to explore and construct human-derived synthetic systems that minimize the risk of immune rejection. Second, there is a risk of expression leakage in modified CAR-T cell gene circuits, which is a common concern in synthetic biology–based gene expression regulation models. In CAR-T cell design, gene circuits are constructed by applying different biological elements, and design-build-test-learning is performed to reduce the leakage of expressed genes. In addition, ensuring that the introduced gene circuit is biologically orthogonal to the endogenous T cell response is crucial. Last, the delivery and integration of complex gene circuits in synthetic biology modifications face technical limitations, thereby imposing increased burdens and technical requirements on industrial production. Currently, gene delivery systems for CAR-T cell production use both viral and nonviral delivery systems. Viral delivery systems have limitations in terms of their gene payload and pose a risk of insertional mutagenesis. Nonviral delivery systems, such as piggyBac, Sleeping Beauty, and the Tol2 transposon system, offer lower production costs, a reduced risk of insertional mutagenesis, and larger gene loads. However, these methods demonstrate lower gene transfer efficiency and require longer in vitro cultivation. Future efforts should concentrate on exploring new methods to increase gene delivery/transduction efficiency and load to facilitate the industrial production of CAR-T cells equipped with synthetic biology switches.

In the future, CAR-T cells are expected to advance toward more sophisticated engineered cells, a goal that can be achieved with the wide variety of components, logic gates, and genetic circuits provided by synthetic biology approaches. In addition, the development of universal CAR-T cell therapy is a future research focus. Compared with conventional CAR-T cells, universal CAR-T cells may cause more serious adverse reactions, such as GVHD and immunological rejection. Therefore, the forthcoming stages of research will prioritize the modification of universal chassis cells to enhance safety. Moreover, thorough investigations of the adverse effects of CAR-T cell therapy are urgently needed. Clinical trials of BPX-601 and BPX-603, which included additional switches, were stopped because of severe adverse effects. The FDA has also announced additional inquiries into the risk of secondary T cell malignancies linked to CAR-T cell therapy (113). Studies further indicate that CAR-T cell therapy may activate dormant viruses, which can lead to severe complications (114). These findings underscore the persistent need for a comprehensive assessment of safety concerns related to CAR-T cell therapy. In summary, as a highly promising antitumor cell therapy, CAR-T cell therapy exhibits remarkable potential for treating solid tumors and hematological malignancies. We anticipate the emergence of safer, more controllable, more intelligent, and universal CAR-T cell therapy in future clinical treatments.