Insights into next-generation immunotherapy designs and tools: molecular mechanisms and therapeutic prospects

Abstract

Immunotherapy has revolutionized the oncology treatment paradigm, and CAR-T cell therapy in particular represents a significant milestone in treating hematological malignancies. Nevertheless, tumor resistance due to target heterogeneity or mutation remains a Gordian knot for immunotherapy. This review elucidates molecular mechanisms and therapeutic potential of next-generation immunotherapeutic tools spanning genetically engineered immune cells, multi-specific antibodies, and cell engagers, emphasizing multi-targeting strategies to enhance personalized immunotherapy efficacy. Development of logic gate modulation-based circuits, adapter-mediated CARs, multi-specific antibodies, and cell engagers could minimize adverse effects while recognizing tumor signals. Ultimately, we highlight gene delivery, gene editing, and other technologies facilitating tailored immunotherapy, and discuss the promising prospects of artificial intelligence in gene-edited immune cells.

Introduction

Immunotherapy has inaugurated a novel era in the fight against malignancies. Among the forefront of these advancements, gene-edited immune cell therapies, particularly Chimeric Antigen Receptor (CAR)-T cell therapy, have exhibited remarkable efficacy in the battle against hematological malignancies. CAR-T cell therapy enables the specific targeting of genetically modified T cells into the tumor cells, circumventing the reliance on MHC molecules. The procedure involves the redirection editing of T cells in vitro, which are then reinjected into the patient (Fig. 1). Although most patients exhibit a favorable response and overall remission rate to CAR-T cell therapy, tumor recurrence remains a great challenge [1]. The essential contributor to treatment failure is the loss of CD19 or BCMA [2, 3] and the emergence of antigen-negative clones [4]. Furthermore, the controllability of CAR-T cell-related encephalopathy syndrome (CRES) and cytokine release syndrome (CRS) represents major obstacles to scaling up the clinical application [5, 6]. Although CAR-T cell therapy has been investigated for the potential in treating solid tumors and autoimmune diseases [7, 8], expanding CAR-T cell therapies remains challenging due to the safety, efficacy, and effectiveness concerns. The inherent limitations of autologous cell therapies, including protracted manufacturing timelines [9], prohibitive costs [10], and donor dependency [11, 12], have catalyzed a paradigm shift toward developing universal allogeneic CAR-immune cell platforms. As an alternative therapeutic strategy, allogeneic CAR-immune cell-based therapy demonstrates superior cost-effectiveness and significantly reduces patient waiting periods [13] (Fig. 1). Moreover, this approach circumvents the limitations associated with patient donor cell quality and insufficient quantity through the utilization of cells derived from healthy donor sources, including peripheral blood mononuclear cells (PBMC) [14], hematopoietic stem/progenitor cells (HSPC) [15], and induced pluripotent stem cells (iPSC) [15, 16].

Fig. 1

CAR production. A. Autologous CAR-immune cells. The process includes isolating immune cell populations from the patient, followed by transduction of these cells with viral or non-viral vectors to express the CAR. After ex vivo expansion, the engineered immune cells are infused back into the patient. B. CAR generation in vivo. Injecting nanoparticles containing mRNA that encodes the CAR could achieve CAR expression in vivo. C. Allogeneic CAR-immune cells. Immune cells are isolated from healthy donors, undergo genetic editing to reduce immunogenicity and minimize immune rejection, and are then produced at a large scale with rigorous quality control before being applied in clinical settings

Beyond CAR-T cells, next-generation immunotherapy tools encompass a diverse array of gene-edited immune cells, including CAR-natural killer (CAR-NK) cells, CAR-Macrophage (CAR-M) cells and CAR-unconventional T cells. These alternative approaches possess unique advantages in augmenting the anti-cancer activity and minimizing off-target effects, offering novel avenues for deciphering the complex code of clinical translation. Moreover, emerging cutting-edge technologies could construct more sophisticated cellular programming systems to realize the precise regulation of the multi-antigen targeting profiles, thereby yielding novel optimization strategies for the next-generation gene-edited immune cells [17]. To illustrate, the engineered CARs precisely modulate the targeting and activity of gene-edited immune cells. Meanwhile, adapter molecules could augment their adaptability and facilitate high-specificity engagement with a diverse array of cell types [18]. We summarize the timeline of important events in multi-targeting strategies in Fig. 2.

Fig. 2

Timeline of key events. [1] Eshhar, Z., Waks, T., Gross, G. & Schindler, D. G. Specific activation and targeting of cytotoxic lymphocytes through chimeric single chains consisting of antibody-binding domains and the gamma or zeta subunits of the immunoglobulin and T-cell receptors. Proc Natl Acad Sci U S A 90, 720–724 (1993). [2] NCT00924326 [3] Hegde, M. et al. Combinational targeting offsets antigen escape and enhances effector functions of adoptively transferred T cells in glioblastoma. Mol Ther 21, 2087–2101 (2013). [4] Fedorov, V. D., Themeli, M. & Sadelain, M. PD-1- and CTLA-4-based inhibitory chimeric antigen receptors (iCARs) divert off-target immunotherapy responses. Sci Transl Med 5, 215ra172 (2013). [5] https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-blinatumomab-consolidation-cd19-positive-philadelphia-chromosome-negative-b-cell [6] Hegde, M. et al. Tandem CAR T cells targeting HER2 and IL13Ralpha2 mitigate tumor antigen escape. [7] Ruella, M. et al. Dual CD19 and CD123 targeting prevents antigen-loss relapses after CD19-directed immunotherapies. J Clin Invest. 126, 3814–3826 (2016). [8] Roybal, K. T. et al. Engineering T Cells with Customized Therapeutic Response Programs Using Synthetic Notch Receptors. Cell 167, 419–432 e416 (2016). [9] Bielamowicz, K. et al. Trivalent CAR T cells overcome interpatient antigenic variability in glioblastoma. [10] Sukumaran, S. et al. Enhancing the Potency and Specificity of Engineered T Cells for Cancer Treatment. Cancer Discov 8, 972–987 (2018). [11] Pan, J. et al. Sequential CD19-22 CAR T therapy induces sustained remission in children with r/r B-ALL. Blood 135, 387–391 (2020). [12] Liu, E. et al. Use of CAR-Transduced Natural Killer Cells in CD19-Positive Lymphoid Tumors. N Engl J Med 382, 545–553 (2020). [13] Au, K. M., Park, S. I. & Wang, A. Z. Trispecific natural killer cell nanoengagers for targeted chemoimmunotherapy. Sci Adv 6, eaba8564 (2020). [14] NCT04660929 [15] NCT04606433 [16] A first-in-human study of CD123 NK cell engager SAR443579 in relapsed or refractory acute myeloid leukemia, B-cell acute lymphoblastic leukemia, or high-risk myelodysplasia.—ASCO [17] Wang, X. et al. Allogeneic CD19-targeted CAR-T therapy in patients with severe myositis and systemic sclerosis. Cell 187, 4890–4904 e4899 (2024). [18] Wang, X. et al. Allogeneic CD19-targeted CAR-T therapy in patients with severe myositis and systemic sclerosis. Cell 187, 4890–4904 e9 (2024). [19] Zhou, J.-e. et al. Lipid nanoparticles produce chimeric antigen receptor macrophages (CAR-M) in situ for the treatment of solid tumors. Nano Today 61, 102,610 (2025)

Leveraging single-target strategies to enhance immunotherapy has proven insufficient in confronting immunological challenges, indicating that next-generation immunotherapy might necessitate multi-targeting approaches. To propel the evolution of immunotherapy, this review evaluates the merits and drawbacks of various gene-edited immune cell therapies and discusses the design and function of multi-specific antibodies and cell engagers, focusing on strategies aimed at achieving multi-targeting profiles. The review will also explore how gene delivery/editing technologies and artificial intelligence (AI) can aid the development of personalized immunotherapy.

Gene-edited immune cells

CAR-T cell

CAR-T cells, autologous/allogeneic T cells genetically transduced to express antigen-specific CAR proteins by retrovirus/lentivirus/non-viral vectors, could specifically target tumor antigens and exert potent anti-tumor effects [17, 19]. The main components of the CAR construct, including the extracellular target-binding domain (single-chain fragment variable, scFv), a hinge domain, a transmembrane domain, and an intracellular signal domain, usually determine the function of the CAR-T cell (Fig. 3A). Affinity tuning of the binder domain in CAR construct design is crucial, as it promotes optimal CAR performance in low-level targeted antigen-expressing tumors, whereas overly low or high affinities may increase off-target cytotoxicity or the risk of tumor escape [20]. The single-chain nanobody has the potential to be widely used in CAR production due to its small size, high specificity, and selective stability [21]. In contrast to classical CAR structural domains, the nanobodies exhibit reduced immunogenicity and contribute to overcoming the immunogenicity problem of linker [22]. The scFv structural domains within the nanobody-based nanosome structure maintain their affinity without exhibiting a loss, whereas the potential for cross-pairing between the VH and VL domains of two different scFv molecules in the classical CAR structure leads to decreased affinity [23]. The hinge and transmembrane domains could respectively be involved in binding target epitopes and maintaining CAR stability [24, 25]. The intracellular signal domain comprises the costimulatory domain (cluster of differentiation (CD) 28, inducible co-stimulator (ICOS), 4-1BB (CD137), OX40 (CD134), and CD27) [26] and activation domain (CD3ζ or Fc receptor FcεRIγ) [27], is another key factor in CAR engineering. Containing three immunoreceptor tyrosine­based activation motifs (ITAMs), the activation domain is involved in the transduction of CAR-T cell activation. In the second and third-generation CAR [28, 29], the costimulatory domain and additional domains are generally introduced to enhance T-cell activation [30]. The engineered CAR-T cells could recognize the tumor antigen directly and specifically through scFv, thereby secreting granzymes, perforin, and inflammatory cytokines to exert oncolytic impact.

Fig. 3

Overview of CAR. A. Structure of CAR The structure of a CAR includes the extracellular domain, transmembrane domain, and intracellular domain. The extracellular domain comprises the single-chain variable fragment (scFv), which is composed of VH, VL, and a flexible linker, along with the hinge region. The intracellular domain includes co-stimulatory domains and signaling domains. B. The CAR evolution The evolution of CAR structures primarily involves modifications to the intracellular domain. First-generation CARs contain a simple CD3ε domain. The second-generation and third-generation CARs introduce one and two co-stimulatory domains, respectively. Fourth-generation CARs incorporate additional functional domains, such as those promoting cytokine release. The fifth-generation CARs integrate JAK/STAT signaling to enhance immune cell persistence and sensitivity to immunosuppressive microenvironments

The fourth/fifth generation CAR-T cells have been developed to reinforce CAR-T cell potency. Fourth-generation CAR-T cells, referred to as T-cells redirected for universal cytokine-mediated killing (TRUCKs), have been further engineered to secrete specific cytokines such as IL-12, IL-15, IL-18, CCL19, and IL-7 [31] (Fig. 3B), thereby recruiting and activating other innate immune cells to enhance anti-tumor effects. Incorporating additional co-stimulatory molecules and signaling cascade pathways, such as the JAK/STAT pathway, fifth-generation CAR-T cells are engineered to effectively combat the tumor microenvironment (TME), with enhanced persistence and cytotoxic efficiency [32] (Fig. 3B). Meanwhile, another defined fifth-generation CAR-T cells, "off-the-shelf" products designed with allogeneic CAR-T cells, show potent anti-tumor activity, markedly reducing the temporal and economic burdens associated with CAR-T cell production and expediting the commercialization pathway [33].

CAR-NK cell

NK cells, as innate lymphocyte subsets, have been an attractive strategy for immune cell therapy. Multiple mechanisms could trigger the activation of NK cell-mediated tumor eradication, such as recognizing tumors with downregulated expression of MHC class I molecule [34, 35], binding to overexpressed NK cell activating ligands on tumor cell surface [34, 36], and antibody-dependent cell cytotoxicity (ADCC) [37]. Generally, immunoglobulin-like receptors on NK cells, killer-cell immunoglobulin-like receptors (KIR), could suppress NK cell function by binding to MHC molecules on normal cells [38]. Some tumors downregulate MHC expression to evade T cell-mediated cytolysis, permitting NK cells to exert unimpeded anti-tumor activity in the absence of inhibitory signals from “self” MHC molecules [34]. It is recognized that the “missing-self” condition constitutes the distinctive oncolytic potential of NK cells [39]. Based on the “inducing self” hypothesis, tumor cells upregulate stimulatory ligands associated with NK cell activation, which provides sufficient activating signals for NK cells to override inhibitory signals by binding to multiple activating receptors—natural killer group 2, member C/natural killer group 2, member D (NKG2C/NKG2D) [40]. Additionally, NK cells express CD16 that binds to the constant region (Fc) of the antibody to exert killing effects via ADCC [41].

CAR-NK cells could be redirected to specific antigens and induce enhanced tumor-antagonizing immunity [42]. CAR activation and endogenous receptor-mediated NK cell activation kill target cells via diverse mechanisms, such as releasing perforin and granzyme for cytolysis [43], upregulating death ligands—FAS cell surface death receptor (FAS ligand) and TNF-related apoptosis-inducing ligand (TRAIL) [36], and secreting interferon-gamma (IFN-γ) for activating other immune cells [44]. Most current CAR constructs, including CD3ζ and co-stimulatory domain for T cells, employed in CAR-NK cells are identical to those used in CAR-T cells. Several studies showed that cooperative activation may occur through the interaction of CAR with activated receptors on NK cells [45, 46]. Given the unique signaling domains intrinsic to NK cells [47], including DNAX-activation protein 12 (DAP12), DAP10, and CD244, designing optimal and specific CAR constructs for NK cells may optimize their therapeutic potential and potentiate their intrinsic cytotoxic capabilities [48, 49]. NK cell-tailored CARs have shown higher cytotoxicity against tumors and improved proliferation [50,51,52]. Li et al. engineered NK cells derived from iPSC to express the CAR construct, comprising the transmembrane domain of NKG2D, the 2B4 co-stimulatory domain, and the CD3ζ signaling domain [53], which endowed NK cells with potent anti-tumor activity and prolonged survival time in vivo. Notably, the replacement of scFv with NK cell activating receptors or nanobodies exhibited effective tumor-killing ability [54].

CAR Macrophage Cell

Solid tumors often suppress immune cell function and promote cell exhaustion by the inhibitory TME. In contrast to the poor infiltration of CAR-T/NK cells into tumors, macrophages represent the predominant immune cell population in the TME and possess distinctive properties of tumor infiltration [55]. Macrophages could remodel the immunosuppressive TME. For instance, macrophages could degrade fibrotic matrix in tumors by secreting matrix metalloproteinases (MMP), thereby creating a microenvironment conducive to immune cell infiltration [56, 57]. CARs for macrophages share the same overall structure as those constructed for CAR-T/NK cells, but CAR-M cells harness more appropriate intracellular signal transduction domains for their activation, such as CD147, FcRγ, multiple epidermal growth factor-like domains protein 10, adhesion G protein-coupled receptor Bai1, and myeloid-epithelial-reproductive tyrosine kinase. Introducing CARs with these specific intracellular structures could redirect macrophages, whilst enhancing phagocytosis function, facilitating infiltration ability, and counteracting the immunosuppressive effects of the TME [58,59,60,61].

Given the remarkable plasticity of tumor-associated macrophage (TAM), CAR-M cell therapy significantly enhances the viability and resilience of immune cells within the TME, potentiating anti-neoplastic effects and mitigating immunosuppressive challenges [62, 63]. Klichinsky et al. reported that introducing CARs into macrophages via chimeric adenoviral vectors enabled CAR-M cells to maintain the M1 phenotype [61]. The CAR-M cells could also repolarise M2 macrophages to M1 macrophages in the TME by secreting cytokines and chemokines [61]. Moreover, CAR-M cells possess diverse killing capacities, such as antigen-specific phagocytosis and antigen-independent phagocytosis [64, 65]. Further research may concentrate not only on the particular phagocytosis of CAR-M cells but also on utilizing the antigen-presenting and immunomodulatory functions to enhance tumor immunotherapy.

The comparison of CAR-T cells, CAR-NK cells, and CAR-M cells is summarized in Table 1.

Table 1 Comparison between different CAR-immune cells

CAR-Unconventional T cells

Conventional T cells used to introduce CAR constructs are αβ T cells, whereas non-conventional T cells are a special class of T lymphocyte subsets including natural killer T (NKT) cells, mucosa-associated invariant T (MAIT) cells and γδ T cells [66]. Unconventional T cells diverge from the MHC molecule-mediated antigen recognition paradigm and exhibit unique antigen recognition and immune functions, providing a promising platform for CAR-immunotherapy.

NKT cells co-express TCR and NK cell markers (CD161) and could recognize lipid antigens via CD1d molecules [67]. CAR-NKT cells with dual antigen recognition mediated by CAR and CD1 d molecules could rapidly secrete cytokines such as IFN-γ and IL-4 upon stimulation [68], which mobilize and activate immune cells and inhibit TAMs, thus modulating the TME functionality [69]. Rotolo et al. have demonstrated that when targeting CD19, CAR-NKT cells demonstrate enhanced lymphoma eradication efficacy compared to CAR-T cells both in vivo and in vitro [70]. Moreover, CAR-NKT could traverse the blood–brain barrier, facilitating the clearance of cerebral lymphomas—a feat unattainable with CAR-T cell therapy [70]. The in vivo investigation of CAR-NKT cells has substantiated their proficiency in eradicating M2 macrophages that express CD1 d molecules, while simultaneously instigating epitope spreading to potentiate tumor suppression [71]. In hematological malignancies, CD33-specific allogeneic CAR-NKT cells displayed superior tumor eradication efficacy relative to CAR-T cells, including elimination of CD33-negative/low-expressing leukemia stem and progenitor cells – a capability not documented in CAR33-T cell interventions [72]. Additionally, NKT cells exhibit minimal propensity to induce GvHD, and even attenuate GvHD, which potentially through the secretion of IL-4 to modulate post-transplant immune responses [73], thereby presenting a more favorable safety profile. CRISPR/Cas9-mediated disruption of HLA class I/II complexes generated universal CAR-NKT cells that showed superior resistance to allogeneic immune rejection while maintaining enhanced antitumor efficacy compared to conventional CAR-T cells and peripheral blood mononuclear cell (PBMC)-derived NKT cell products [74].

γδ T cells harbor γδ TCRs that mediate direct recognition of tumor-associated antigens (TAAs), including phosphoantigens and major histocompatibility complex class I-related chain molecules A/B, while exhibiting innate cancer-fighting effector through cytokine secretion and cytotoxicity [75]. The cytotoxic activity of γδ T cells is independent of MHC molecular recognition. It is mediated through potent multifaceted mechanisms, including the secretion of IFN-γ, the release of perforin and granzyme, as well as the activation of TRAIL and Fas/FasL pathways [76, 77]. γδ T cells possess an enhanced homing proficiency compared to αβ T cells and are predominantly localized in epithelial tissues, such as the skin and intestinal mucosa [78], thus inherently providing a strategic advantage in combating epithelial-originated neoplasms, including gastric and pulmonary carcinomas. Rscher et al. pioneered the engineering of CAR-γδ T cells using retroviral vectors, substantiating their ability to effectively eliminate lymphoma cells in vitro [79]. Subsequent developed advancements, including transposon systems and mRNA electroporation [80], have further enhanced their oncolytic potency and targeting precision in animal models. Notably, γδ T cells don’t elicit GvHD, and CAR-γδ T cells have manifested a promising safety profile in preclinical studies, underscoring their potential as a next-generation allogeneic CAR-T therapy [81, 82].

MAIT cells function as innate immune effector cells, recognize microbial metabolites presented by MR1 molecules through TCR-mediated interactions [83], and are typically enriched in mucosal tissues [84]. The metabolite-sensing capability of MAIT cells, which enables the detection of abnormal metabolites within TME, provides a distinctive targeting edge for the engineering of CAR constructs. In addition, the localization of MAIT cells in the mucosal barrier confers a more natural penetration advantage of CAR-MAIT in treating solid tumors. Nevertheless, the restricted extracellular expression of MR1 and the diminished abundance of MAIT cells within tumor tissues pose formidable obstacles for the clinical translation of CAR-MAIT therapy [85]. Furthermore, beyond their antitumor functions, MAIT cells exhibit oncogenic potential in certain malignancies [86, 87], plausibly inducing T-cell exhaustion through CTLA-4/PD-1/TIM-3 pathways or recruiting immunosuppressive cells via IL-8 secretion. Thus, research on CAR-MAIT cells is still in its early stages, but their contribution to mucosal immunity and antimicrobial defense suggests therapeutic promise for treating infection-associated cancers or solid tumors [88,89,90].

Collectively, CAR-unconventional T cells exhibit reduced classical MHC dependency compared to CAR-T/NK/M cells, rendering them more promising candidates for the development of allogeneic universal CAR cell therapies [66]. In addition, CAR-unconventional T cells are more resistant to the immunosuppressive TME while maintaining natural killing abilities. Collectively, CAR-unconventional T cells present highly promising therapeutic potential. Nevertheless, these unconventional T cells confront significant challenges, including limited source and difficulty in vitro expansion, which impede their clinical translation [91]. For instance, NKT cells constitute approximately 5% of peripheral blood, while γδ T cells account for less than 1% of PBMC [92], and both are challenging to expand in vitro to achieve therapeutically effective quantities. Several strategies have aimed to differentiate genetically engineered hematopoietic stem and progenitor cells into NKT cells for in vitro expansion, bolstering their feasibility for clinical translation potential [93, 94].

Preclinical and clinical applications of gene-edited immune cells

CAR-T cell

Since 2017, seven CAR-T cell products have been approved by the U.S. Food and Drug Administration (FDA), including 5 targeting CD19 and 2 targeting BCMA (Fig. 4). The seventh FDA-approved CD19-targeted CAR-T cell product, Aucatzyl (obe-cel), employs a rapid antigen-binding dissociation mechanism to minimize T-cell hyperactivation, thereby significantly reducing the incidence of grade ≥ 3 CRS (2.4%) and immune effector cell-associated neurotoxicity syndrome (ICANS, 7.1%) [95]. Notably, Aucatzyl ushered in a new era of enhanced CAR-T cell therapy risk control by integrating mechanistic innovation with rigorous clinical validation, establishing it as the first FDA-approved CAR-T cell therapy to achieve exemption from the Risk Evaluation and Mitigation Strategy [96]. This regulatory milestone reflects a paradigm shift in cellular immunotherapy and demonstrates that engineered T-cell therapies concurrently optimize safety thresholds and clinical accessibility. Recent studies have compared the efficacy of BCMA-targeted CAR-T cells (cilta-cel/ide-cel) with standard therapy in patients with relapsed/refractory (R/R) multiple myeloma. It was demonstrated that the infusion of cilta-cel had a higher overall response (84.6%, n = 208), complete response (CR) (73.1%, n = 208), negative frequency of minimal residual disease (MRD) (60.6%, n = 208), and a lower risk of disease progression (progression-free survival at 12 months: 75.9%, n = 208) [97]. Meanwhile, the early use of cilta-cel may reduce CAR-T cell-induced adverse events [97]. Another study, evaluating the efficacy of ide-cel in 254 patients with R/R multiple myeloma, reported that 71% of patients in the ide-cel group responded, with a CR rate of 39%, and substantially prolonged the median progression-free survival (13.3 months) [98]. A phase II clinical trial, evaluating CAR-T cell efficacy for high-risk R/R large B-cell lymphoma (LBCL) as a part of first-line therapy, demonstrated that axi-cel, an anti-CD19 CAR-T cell product, is both safe and efficacious [99]. Axi-cel also possessed a durable clinical benefit in elderly patients (≥ 65 years) with R/R LBCL, showing manageable toxicity, and improved clinical outcomes [100]. The therapeutic potential of CAR-T cells targeting other targets is being further explored in pre-clinical and clinical trials (Table 2), such as CD7 [101], CD22 [102], GPRC5D [103], NKG2D [104], CD123 [105], SLAMF7 [106], and Siglec-6 [107].

Fig. 4

FDA-approved CAR-T products

Table 2 Clinical trials of CAR-T cells

Although CAR-T cell products for solid tumors have not been approved, numerous clinical trials are underway and show promising results, including targeting HER2 [108], IL-13Rα2 [109], GD2 [110], ROR1 [111], EGFR [112], CEA [113], mesothelin (MSLN) [114], and CD133 [115]. In the phase I clinical trial of 37 patients with gastrointestinal tumors treated with claudin18.2-specific CAR-T cells, the overall response rate was 48.6%, the disease control rate was 73.0%, and a 6-month duration response rate was 44.8%, with tolerable adverse events [116]. Similarly, a phase I/II clinical trial of 27 children with R/R neuroblastoma treated with GD2-targeted CAR-T cells showed a higher response rate and improved 3-year overall survival [110]. Vitanza et al. reported repeatedly dosed intracranial B7-H3 CAR-T cells in three children with R/R diffuse intrinsic pontine glioma (DIPG) [117]. Among them, a patient demonstrated sustained clinical and radiographic improvement through 12 months, with evidence of local immune activation [117], suggesting the feasibility and preliminary tolerability of repeated administration. Frustratingly, most research on CAR-T cells in solid tumors still lacks adequate overall response rates and possesses serious toxicities. With ongoing clinical trials exploring novel targets, CAR-T cell therapy is poised to make further breakthroughs in the treatment paradigms for solid tumors.

CAR-NK cell

Clinical trials of CAR-NK cells have shown encouraging results. Table 3A summarizes representative CAR-NK clinical trials. IL-15 armored anti-CD19 CAR-NK cells with inducible caspase-9 safety switch had remarkable and durable cytostatic effects on cancer and tolerable adverse events in treating patients with R/R CD19-positive lymphoma [118]. Among 11 patients, objective responses occurred in 8 of them, including 7 patients who achieved CR [118]. In a recent I/II R/R B-lymphoid malignancies trial, similar IL-15 CAR-NK cells endowed NK cells with comparable efficacy to autologous anti-CD19 CAR-T cells, and exhibited lower ICANS [119]. Anti-CD70 CAR-NK cells secreting IL-15 against CD19-negative B cell lymphomas were constructed in a pre-clinical mouse model of a xenograft lymphoma line. Repetitive administration of the CAR-NK cells induced sustained tumor remission and enabled long-term survival of CAR-NK cells in mice [120]. Thus, pre-clinical trials of CAR-NK cells have illuminated a ray of hope in treating various solid tumors [121, 122].

Table 3 Clinical trials of CAR-immune cells beyond CAR-T cells

CAR-M cell

Compared to other gene-edited immune cells, research in CAR-M cells is still in its infancy (Table 3B). Pre-clinical trials have developed CAR-M cells for solid cancer treatment, such as targeting human epidermal growth factor receptor 2 (HER2) [123,124,125], CD133 [126], mesothelin [127], GD2 [128], glypican-3 (GPC3) [129], anaplastic lymphoma kinase (ALK) [130], mucin 1 (MUC1), and CD19 [127]. Zhang et al. constructed a CAR-147-M targeting HER2, with a CD147 intracellular signal transduction domain, dramatically upregulated multiple MMP expression and promoted the degradation of tumor mechanical barrier, thereby suppressing tumor growth in vivo [60]. Furthermore, levels of inflammatory cytokines (IL-1, TNFα), paramount mediators of CRS, were reduced in mice treated with CAR-M cell therapy, whereas levels of anti-tumor factors (IL12, IFNγ levels) increased [60]. iPSC-derived CAR-M cells could also expand in vivo and maintain their anti-tumor efficacy [127]. The HER2-targeted CAR-M cell therapy, CT-0508, has reported preliminary outcomes from a Phase I clinical trial in HER2-overexpressing tumors, predominantly breast and esophageal cancers, demonstrating favorable tolerability, early antitumor activity, and remodeling of the intratumoral immune microenvironment [131]. Nevertheless, clinical trials of CAR-M cells have not reported final results [59].

CAR-Unconventional T cells

Up to now, clinical studies of CAR-NKT cells remain limited, with only five trials registered on ClinicalTrials.gov (Table 3C). A phase I clinical trial evaluating CAR-NKT cells co-expressing GD2 and IL-15 reported an ORR of 25% (3/12) and showed an enhanced safety profile in pediatric patients with neuroblastoma [132]. CAR-γδ T cells manifest a relatively advanced status in the research progression of unconventional T cells, particularly in hematological oncology, where early encouraging signals have been detected (Table 3D). A subset of clinical trials has advanced to phase I/II, revealing superior safety profiles and significant objective tumor regression. ADI-001, an allogeneic CAR-γδ T cell therapy targeting CD20, was evaluated in patients with relapsed/refractory B-cell lymphoma [133]. The Phase I clinical trial revealed an ORR of 67% (4/6) and a CR of 67% (4/6), while exhibiting a favorable safety profile. Allogeneic B7H3-targeted CAR-γδT cells demonstrated favorable tolerability in a clinical trial involving 7 patients with high-grade recurrent GBM, with no instances of grade 3 or higher CRS, ICANS, or GvHD, achieving an objective response rate (ORR) of 42.9% (3/7). Despite the considerable anticipation surrounding CAR-MAIT cell application in mucosa-associated solid tumors (such as colorectal, hepatocellular, and lung cancers), CAR-MAIT cells are still confined to the preclinical phase, with no research having progressed to clinical trials to date. The advancement of CAR-MAIT technology depends on additional experimental substantiation of its safety and practicability.

New challenges for immune cell therapy: multi-antigen targets gene editing

Challenges in target selection

Extending CAR immune cell therapy, the frontrunner in hematological malignancies, to the vast majority of tumors (solid tumors) represents a significant challenge for immunotherapy. A primary factor contributing to the difficulty of treating solid tumors with conventional CAR-T therapies (autologous, single-target) is the restricted and inconsistent expression of tumor-specific antigens (TSAs) [134, 135]. Targets in hematological malignancies, such as CD19, are frequently characterized by highly specific expression in malignant B cells. In contrast, solid tumors predominantly express TAAs, which are also present at low levels in normal tissues [136]. The utilization of TAAs as targets for CAR-T cell therapies frequently leads to off-target toxicity. For instance, CAR-T cell targeting HER2, CEA, and MSLN has documented instances of pulmonary toxicity and, in some cases, fatal outcomes [137,138,139]. Moreover, the tumor heterogeneity inherent in solid tumors presents another substantial challenge for CAR target selection. Diverse tumor antigens are frequently expressed heterogeneously within solid tumors and across different patients. Consequently, single-target CAR-T cell therapy may only eliminate subpopulations of tumor cells expressing the target antigen, while residual antigen-negative or low antigen-expressing cells can persist and drive tumor progression [140]. The TME specific to solid tumors further complicates CAR-T cell therapy [141]. Physically, the TME impedes CAR-T cell infiltration into the tumor [142, 143]. Additionally, the inhibitory components within the TME can induce CAR-T cell exhaustion or functional inhibition, thereby further constraining the efficacy of CAR-T therapy [142, 144]. Notably, preclinical models could not fully replicate human tumor complexities, and this discrepancy significantly complicates clinical translation. CNS toxicity observed in GD2 CAR-T cell-treated neuroblastoma xenograft mice [145] has not been reported in clinical trials [146, 147]. In contrast, gastrointestinal toxicity associated with CLDN18.2 CAR-T cell therapy, as reported in clinical trials, was not detected in preclinical models [116]. Therefore, target selection for solid tumors necessitates a balance between efficacy and safety, driving the identification of TSA targets. The discovery of promising TSAs relies on the advancement of innovative technologies, such as robust high-throughput genome-wide CRISPR screening platforms, which may facilitate the identification of more suitable and effective antigenic targets for CAR-T cell therapy [148, 149].

Challenges of tumor recurrence

Even in the promising hematological malignancies, CAR-T cell therapies are not entirely trouble-free. The glaring problem is tumor recurrence after treatment, which represents a critical bottleneck limiting long-term efficacy. One of the important reasons for tumor recurrence lies in the antigenic loss/downregulation of tumors, a phenomenon that tumor cells have evolved a set of avoidance mechanisms to evade immune recognition in response to pressure from CAR-T cells [150, 151]. Potential mechanisms underlying this evasion include mutations in genes encoding antigens (e.g., CD19 [150, 152], BCMA [153, 154], lineage switching of tumor cells [155], and epigenetic modifications of gene expression [156], among others. In addition, tumor heterogeneity serves as a critical factor driving tumor recurrence [157, 158]. Tumor cells from antigen-negative clones are resistant to conventional single-target CAR-T cell therapies and may proliferate after treatment, thus leading to tumor recurrence. Equally important, factors intrinsic to CAR-T cells themselves, independent of target antigen expression, significantly influence therapeutic efficacy. The limited persistence of CAR-T cells in vivo, along with immunosuppressive signals from prolonged antigenic stimulation and the TME, could lead to CAR-T cell exhaustion, resulting in diminished efficacy of CAR-T cells in tumor suppression [159, 160].

Insights from previous therapeutic challenges highlight a critical realization: the long-term efficacy of single-target CAR-T therapies is inherently limited. Future CAR-T therapies must be designed to engage multiple targets. On one hand, targeting multiple tumor antigens can address the challenges posed by tumor heterogeneity and antigen loss. On the other hand, incorporating additional targets, such as co-stimulatory molecules or immune checkpoints, can enable functional signaling regulation of CAR-T cells, activating proliferative pathways or blocking inhibitory pathways. This approach can effectively mitigate cellular exhaustion and enhance the durability of the therapy.

Multi-antigen targeting cell strategies

The multi-targeting strategy represents a novel immunotherapeutic approach based on genetically engineered immune cells, aiming to enhance the durability and effectiveness of immunotherapy while avoiding tumor escape and recurrence. These engineered immune cells could target multiple tumor antigens through pooled CAR-T cells, introducing multiple scFvs and CAR constructs. Pooled combination of CAR-T cells entails mixing CAR-T cells with distinct targets. Another approach involves modifying CAR structural designs: (1) to fuse multiple scFvs in a single chimeric protein; (2) to introduce multiple CAR constructs targeting different targets in a single immune cell. The comparisons of different immunotherapy designs are summarized in Table 4.

Table 4 Comparison between different immunotherapy designs

Biological logic gates

OR-gate

OR-gate strategy could extend the CAR-T cell targeting antigen library to eradicate heterogeneous tumor cells. Conjointly targeting CD19 and CD123 augments the possibility of eliminating subclones that may acquire selective advantages under solely targeting CD19, thus preventing the recurrence of CD19-negative tumors [161]. Of note, the simultaneous administration of two CAR-T cell lines exerts considerable immunological pressure on tumor cells, potentially leading to the concurrent escape of both targeted antigens. Pooled CAR-T cell therapy, namely concurrently administering different CAR-T products at a specific ratio [162], and cocktail therapy, which administters different CAR-T cell products sequentially at specific time intervals, both enhance tumor remission rates and reduce relapse risk [163]. The OR-gate strategy enables the engineered cells, such as tandem CAR-T cells, dual CAR-T cells, and trivalent CAR-T cells, to recognize two or more targeted antigens, and recognition of any one of these targets can result in their activation (Fig. 5A).

Fig. 5

Overview of logic gates. A. OR gate Dual CAR (with complete co-stimulatory and activation domains per receptor), Tandem CAR and Trivalent CAR belong to the OR-gate design, where activation of CAR-immune cells can be triggered by either one of the targeted antigens. B. AND gate Dual CAR (with separated co-stimulatory and activation domains), On-Switch and SynNotch CAR belong to the AND gate design, requiring the simultaneous presence of two targeted antigens to activate CAR-immune cells. ON switch: The structure of CAR is divided into two parts, one carrying the intracellular structural domain of FKBP12 and the other carrying the structural domain of FRB. As an inducer, rapamycin bridges FKBP12 and FRB to assemble a functional CAR that is inactivated (OFF) in the absence of rapamycin and activated (ON) upon addition to achieve spatiotemporal regulation of toxic effects. C. IF-BETTER gate The presence of Tumor-Associated Antigen (TAA) 2 facilitates the activation of CAR-immune cells in the absence of TAA1. However, it is not a necessary condition for activation when TAA1 is present in sufficient amounts. D. NOT gate Binding of TAA2 to the CAR induces the activation of immune cells, whereas this activation does not occur in the presence of TAA1

Dual CAR-T cells could express two independent CAR constructs, each specific for cognate antigens. Approximately 30% of patients with R/R B-ALL experience CD19-negative relapses. OR-gate dual CAR-T cells targeting CD19 and CD123, developed by Ruella et al., could prevent tumor recurrence due to antigen loss after CD19 targeting therapy [161]. Collectively, the dual CAR-T cells demonstrated superior efficacy over single antigen-targeted CAR-T cells or pooled CAR-T cells.

CAR constructs of tandem CAR-T cells contain two tandem-specific scFv structural domains. Tandem CAR-T cells demonstrate more durable activity compared to dual CAR-T cells or pooled CAR-T cells and do not exhibit increased exhaustion [23]. When simultaneously binding two antigens, tandem scFvs manifest synergistic effects, forming superior and stable immunological synapses that induce an exponentially more efficient response in tandem CAR-T cells [164]. BCMA/CD19-targeting tandem CAR-T cells for treating multiple myeloma (MM) achieved high therapeutic responsiveness while exhibiting good tolerability and safety control [165]. Tandem scFv structural domains recognizing CD19/CD20 have been designed to prevent antigen escape, which could be efficiently activated by either target antigen [166]. Likewise, OR-gate BCMA/CD319-targeted CAR-T cells for MM treatment also demonstrated superior efficacy compared to pooled CAR-T cells and dual CAR-T cells [167]. The study comparing the efficacy of tandem CD19/CD22 dual CAR-T cells and sequential CD19/CD22 CAR-T cells reported similar immune responses for both, but both were superior to single CD19 CAR-T cells [168]. Targeting CD19 and CD22 could elicit long-term MRD-negative remission in patients with R/R B-ALL [169]. Notably, the larger size of tandem CAR constructs poses a significant concern, as this could exceed the cargo capacity of viral vectors and impact transduction efficiency [170]. And incorrect pairing and hydrophobic interactions between VH and VL domains could result in scFv aggregation, potentially leading to CAR-T cell exhaustion [171, 172]. Additionally, murine-derived scFvs elevate immunogenicity risk in tandem CAR-T cells [173]. Pleasantly, developing fully human heavy chain variable antibodies and nanobodies offers promise for addressing these challenges [21, 174].

Interestingly, CAR-T cells binding three different receptors have been reported to minimize the potential for tumor antigen escape and effectively kill multiple tumor cells [175, 176]. In patients with glioblastoma (GBM), CAR-T cells targeting three antigens, including human epidermal growth factor receptor 2 (HER2), interleukin-13 receptor subunit alpha-2 (IL13Rα2), and ephrin-A2 (EphA2), have demonstrated the capacity to overcome antigenic heterogeneity, and achieved nearly complete tumor cell clearance [176]. The superior activity against GBM may be attributed to trivalent CAR-T cell-enhanced antigen-binding capacity, signaling, and immune synapse formation.

AND-gate

OR-gate could augment CAR-T cells' antigen recognition capacity, thereby arousing stronger tumor-antagonizing immunity. Nevertheless, OR-gate may lead to on-target, off-tumor toxicity. To enhance the existence and functional maintenance of genetically engineered cells in healthy tissues, AND-gate represents a seminal advance for multi-targeting strategy development [177].

The ON-switch CAR incorporates a split-receptor design, where the scFvs from traditional CARs are divided into two separate parts that could assemble in a small molecule-dependent manner, effectively functioning as an ON-switch for CAR-T cell activation (Fig. 5B). This design, representing the earliest forms of AND-gate, enables precise control over the timing, spatial distribution, and activity levels of CAR-T cell activation, thereby reducing off-target risks. Wu and colleagues demonstrated the controllability and efficacy of modified rapamycin by harnessing ON-switch to form functional CD19-targeted CAR-T cells to reduce tumor load in leukemia mice [178]. The activation of AND-gate CAR design necessitates two separate target recognitions. In the AND logic gate, T-cell lysogenic cytotoxicity could be triggered only when both CARs specifically bind to the corresponding target antigen [179] (Fig. 5B). Normal tissues may express one type of tumor-associated antigen (TAA) alone at low levels, but rarely both. Thus, the combined activation signal input improves CAR-T cells’ ability to discriminate between tumor and normal cells.

Similar to split ON-switch receptors, AND-gated dual CARs typically split the signaling domain, with one CAR providing a co-stimulatory signal (CD28/4-1BB) and the other providing an activating signal (CD3ζ), each targeting a different antigen. Dual CAR-T cells can be partially activated when recognizing a single target antigen with a low level of cytokine secretion. Conversely, when two target antigens are present simultaneously, dual CAR-T cells could be fully activated and exhibit anti-tumor activity. AND-gated trivalent CAR-T cells could target tumor site-specific expression patterns, such as prostate stem cell antigen (PSCA), transforming growth factor-β (TGF-β), and IL-4, and their effector functions could be confined to the pancreatic cancer site [180]. CAR-T cells targeting tumor site-specific expression patterns exhibit resilience to suppressive immune factors (TGF-β and IL-4), and this targeting approach can be extrapolated to other suppressors within TME [180]. For instance, AND-gate-regulated CAR-T cells could target suppressive factors in TME, including chemokine-receptor network, tumor vasculature [181, 182], extracellular matrix [183, 184], hypoxia [185], and immunosuppressive cytokines [186, 187], to enhance CAR-T cell survival and tumor infiltration, thereby providing a promising avenue for solid tumor treatment. Hou et al. developed bispecific CAR-T cells targeting IL-13Rα2 and TGF-β, which could counteract TGF-β-mediated immune suppression and reshape the TME, significantly driving tumor clearance in glioblastoma multiforme [187]. It was reported that IL-15-modified CAR-T cells could additionally target myeloid-derived suppressor cells to reverse their immunosuppressive effects, enhancing antitumor activity [188].

Synthetic Notch (SynNotch) receptor represents an enhanced iteration of the ON-switch, namely the IF–THEN gate. SynNotch receptor system can limit CAR expression through temporal modulation, thereby reducing non-specific attacks on healthy tissues and enhancing the safety of CAR-T cell therapy [189,190,191]. The specific binding of CARs targeting antigen A and SynNotch receptors could induce the release of transcriptional regulatory factors [189]. Transcription factor is translocated to the nucleus and binds to its activation sequence, which regulates transcription and expression of antigen B-specific CAR. This process results in transient expression of antigen B-specific CAR, thereby enabling CAR-T cells to express specifically only at tumor sites and ensure highly specific killing [189]. Cortese et al. describe a HER2 SynNotch-driven CEA-CAR-NK cell in which CEA-CAR expression is induced only upon amplified HER2, providing an innovative and safe strategy for treating HER2-amplified colorectal cancer (CRC) resistance [192]. In GBM, IF–THEN gate-regulated CAR-T cells achieve high specificity and more complete killing of heterogeneous tumors by employing the highly tumor-specific neoantigen EGFRvIII or the tissue-specific antigen MOG as initiating signals through a multi-targeting strategy [193]. In addition, SynNotch receptors could enhance CAR-T cell efficiency by inducing local cytokine interleukin production, thereby counteracting the immunosuppressive TME [194]. It has been shown that CAR-T cells regulated by the SynNotch system better control tumor load via circumventing the premature differentiation and depletion of CAR-T cells induced by antigen-independent tonic signaling [193, 195]. However, murine-derived SynNotch systems and synthetic transcription factors may elicit immunogenic responses, and the transgene size may pose an additional challenge in this context. Zhu et al. developed a humanized SynNotch-like system for precise regulation of gene-edited T cells by expressing customizable synthetic intramembrane proteolysis receptors (SNIPRs) [189]. Transmembrane and near-membrane structural domains of this receptor could be tuned to the sensitivity of SNIPR by altering the amino acid sequence, providing greater flexibility in customization. Interestingly, the SNIPR system could trigger recognition signals at tumor sites exclusively, which in turn drives the release and expression of ligand-dependent transcription factors. This confines CAR construction to tumor sites, reducing T-cell depletion caused by tonic signals. Additionally, SNIPRs have a compact structure and are compatible with humanized transcription factors, significantly lowering the risk of immunogenic responses.

The IF-BETTER gate could recognize both target antigens. When antigen A is present at high levels, T-cell activation does not require antigen B, as sufficient levels of antigen A alone could trigger T-cell activation [196] (Fig. 5C). Nevertheless, when the expression level of antigen A is low, antigen B contributes to reaching the activation threshold for signaling following antigen A-specific CAR-mediated recognition [196]. The regulatory logic gate implies that CAR-T cells will exhibit a preference for antigen B. Currently, the IF-BETTER gate is only applied in pre-clinical models of hematological malignancies. Through the quantitative analysis of antigen distribution and density in both acute myeloid leukemia (AML) and normal hematopoietic cells, Haubner et al. engineered a sensitivity-optimized adhesion G protein-coupled receptor E2 (ADGRE2)-CAR co-expressed with a C-type lectin domain family 12 member A (CLEC12 A)-targeted chimeric costimulatory receptor (CCR), which preferentially engages ADGRE2^positive/CLEC12 A^positive leukemic stem cells, rather than ADGRE2^low/CLEC12 A^negative, thereby minimizing off-target toxicity and effectively overcoming antigen escape mechanisms in AML [197]. Given the low expression of TSAs in solid tumors, IF-BETTER may exert far-reaching immune activity in treating solid tumors. Furthermore, the particular identification of antigen B in the IF-BETTER gate could rely on either a chimeric costimulatory receptor or a cell-surface scFv, thus circumventing the second CAR-induced, off-tumor toxicity observed in OR-gate configurations [196].

NOT-gate

NOT-gate engineered T cells could specifically recognize two target antigens (e.g., A and B). In contrast to conventional T cell activation by CAR-specific recognition, NOT-gate employs an inhibitory CAR (iCAR). The inhibitory signaling domains contained in iCARs are frequently designed to target antigen B that present solely in normal tissues to suppress cell activation (Fig. 5D). Under the regulation of NOT-gate, T cells can be activated exclusively when CAR recognizes antigen A and iCAR does not recognize antigen B. Programmed death (PD)−1/cytotoxic T-lymphocyte-associated protein (CTLA)−4-based iCAR demonstrated the superior ability to discriminate between target and off-target cells in vivo and in vitro [198]. Equivalently, Bangayan and colleagues generated iCARs with two inhibitory signaling structural domains (PD1-PD1, PD1-sialic acid binding Ig-like lectin 9, PD1-leukocyte associated immunoglobulin-like receptor 1) to regulate the delicate balance of conducting signals between CAR and iCAR, significantly improving the inhibitory efficiency of iCARs [199]. By employing iCARs to target the frequent HLA loss in tumor cells, the mechanism of tumor immune escape has been ingeniously exploited to provide a more precise and safe, yet effective strategy for CAR recognition [200,201,202]. Bassan et al. introduced an iCAR containing the leukocyte immunoglobulin-like receptor B1 (LIR1) inhibitory domains to target HLA-A*02 (“A2”) loss of heterozygosity, thereby avoiding on-target, off-tumor toxicity in normal tissues [201]. Equivalently, Hwang et al. developed a NOT-gate-based allele-sensing CAR to target tumor cells with frequent HLA loss, and the CAR could provoke potent anti-tumor immunity in vivo and in vitro models [200]. Emerging studies are exploring potential NOT-gate-based therapeutic targets to combat tumors, emphasizing that NOT-gate could facilitate the safer and more efficacious therapeutic target development [203].

Adding the suicide switch to genetically engineered cells is a traditional strategy to improve safety [204]. The suicide switch activation leads to the complete destruction of the engineered cells, which irreversibly terminates complex therapies and renders treatments costly [205]. In contrast, the NOT-gate mechanism involving iCAR-mediated transient and reversible functional inhibition avoids these drawbacks [198]. Furthermore, after separation from the inhibitory antigen, NOT-gate CAR-T cells still maintain their reactivity to the target antigen. Therefore, NOT-gate could expand targeted tumor antigen profiles, promoting the application of tumor antigens that are highly expressed in tumor tissues but are also expressed to some extent in healthy tissues. The combined OR-NOT logical gating brings a novel approach to multi-targeting strategies. Frankel et al. introduced NOT-gating-based endomucin, a specific target for hematopoietic stem cells, to develop OR-gating-based FLT3/CD33 targeting CAR-NK cells [206]. The OR-NOT three-input logic gate CAR-NK cells can specifically target tumor cells while protecting normal hematopoietic stem cells, thereby providing a wider therapeutic window for hematological malignancy treatment.

The split, universal, and programmable (SUPRA) CAR

Adapter-mediated CAR-immune cells have the potential to overcome the limitations of tumor heterogeneity and traditional targeting paradigms and to modulate therapy in a temporal and spatial manner [207,208,209]. The split, universal, and programmable (SUPRA) CAR system with a modular design can theoretically target almost any antigen. THE SUPRA CAR system consists of two components [209] (Fig. 6A): (1) a universal receptor containing a leucine zip adapter (zipCAR) that connects to intracellular signaling structural units; (2) a separate switch molecule containing a homologous leucine zip with antigen-specific scFv (zipFv). SUPRA CAR-T cells containing zipCAR remain functionally inactive and can only exert anti-tumor activity when a zipFv containing a homologous leucine zipper is exogenously added and combines with zipCAR to reconstitute a functional CAR. Based on this platform, the SUPRA CAR could switch targeting antigens by altering the exogenous switch molecule's scFv without designing a new genetically engineered cell [210]. Moreover, through modulating the zipFv, such as by increasing the concentration of switch molecules or introducing competitive binding of scFv, SUPRA CAR-T cells could regulate anti-tumor response strength. This not only broadens the spectrum of targeted antigens but also enhances the safety of genetically engineered cells for in vivo applications [210]. Designing dual or trivalent CAR systems often necessitates precisely regulating receptor expression levels and fine-tuning scFv affinities to achieve an optimal balance of signaling strength against different antigen combinations [211]. In contrast, the SUPRA CAR system modulates receptor signaling strength by altering the zipFv composition or adjusting the zipFv quantity [207].

Fig. 6

SUPRA CAR, multi-specific antibody, and multi-specific cell engager. A. SUPRA CAR a. Structure of SUPRA The structure of SUPRA involves a zipCAR and a zipFv, with the scFv targeting TAAs. b. SUPRA CAR based on OR gate Any zipFv with scFvs targeting different TAAs can bind to the zipCAR, mediating the activation of gene-edited immune cells. c. SUPRA CAR based on AND gate The zipFv with scFvs targeting different TAAs must simultaneously bind to the corresponding zipCAR to promote the activation of gene-edited immune cells. d. SUPRA CAR based on NOT gate Introducing zipFvs targeting two different TAAs, where TAA2 is absent on tumor cells (expressed only on normal cells), allows gene-edited immune cells to target and kill tumor cells. In normal tissues, the presence of TAA2 results in dimerization of the zipFv targeting TAA1 with the zipFv targeting TAA2, preventing the activation of gene-edited immune cells and reducing toxicity to normal tissues. e. OFF switch Introducing inhibitory zipFvs that competitively bind to the zipCAR with the zipFv targeting TAA can inhibit the activation of gene-edited immune cells. B. Multi-specific antibody a. BsAb The Fab segments of a BsAb target CD3 and a TAA, respectively, linking immune cells and tumor cells to promote tumor killing. b. TsAb TsAb introduces an additional anti-CD28 fragment, which, while linking immune cells, also binds to co-stimulatory domains, enhancing the oncolytic effect. c. MicAbody Based on natural killer group 2 member D (NKG2D), MicAbody is a specific type of bispecific antibody. NKG2D triggers the activation and proliferation of immune cells by recognizing and binding to activating ligands expressed on immune cells, thereby promoting the killing of target cells. C. Multi-specific cell engager a. Bispecific cell engager One arm of the bispecific antibody targets the CD3ε/CD16 chain to activate T/NK cells, while the other arm targets a TAA, promoting the formation of a cytolytic synapse to mediate tumor cell killing. b. TriKE TriKE contains scFvs that specifically bind to the activating receptor CD16 on NK cells and TAA, linked to the cytokine IL-15. IL-15 induces the activation and expansion of NK cells while avoiding off-target toxicity

The combination of SUPRA CAR-T cells and logic gates sheds novel inspiration on evoking robust anti-tumor efficacy and biosafety. Adding different combinations of zipFvs (α-Axl zipFv, α-Her2 zipFv, or both) to mixtures of HER2/Axl-expressing cancer cells and RR zipCAR-expressing T cells effectively activated T cells in response to any of the combination antigens [207] (Fig. 6A). SUPRA CAR-T cells incorporating NOT-gates remarkably lessen the toxicities associated with tumor-antagonizing immunity. Binding of an A-specific zipFv to the zipCAR could be prevented by administering zipFvs that specifically recognize antigens A and B, respectively, where a antigen B-specific zipFv could dimerize with an antigen A-specific zipFv [212]. The SUPRA CAR-T cells based on NOT gates can prevent the reconstruction of an antigen A-specific CAR when antigen B's safety signal is recognized, thereby circumventing off-target killing issues (Fig. 6A). Cho et al. found that the generation of an inhibitory zipCAR via the co-suppressor receptor BTLA led to a robust inhibition of IFN-γ secretion from T cells [208]. A ZipFv containing the B and T lymphocyte-associated inhibitory domain endows NOT logic function in NK cells, reducing cytotoxicity in mouse xenograft models, and establishes a bridge between CAR-immunotherapy and augmented safety [208]. Notably, the administration of SUPRA CAR requires periodic injection of zipFv protein, and the introduction of exogenous pharmaceutical agents presents several challenges, such as small molecule penetration and immunogenicity. Recently, Tousley et al. [213] put forth a novel adapter-mediated CAR, the LINK CAR, comprising fully humanized components and a modular design. The elegant LINK CAR design could address the aforementioned limitations with high efficacy while preventing off-target toxicity [213].

Multi-specific antibodies

BsAb

Bispecific antibody (BsAb), an engineered antibody binding to two distinct antigens or epitopes, shows broader therapeutic potential than conventional antibodies that possess a Y-shaped structure with two identical antigen-binding sites (Fig. 6B). BsAb represents an additional strategy for redirecting T cells. BsAb could bind to diverse epitopes or antigens, diverging from conventional bivalent, monospecific antibodies, which have been developed for the application of drug delivery, specific blockade, redirecting effector cells, and multi-targeting. BsAb could remodel T cells by targeting both TAAs and the CD3 complex, thereby broadening the spectrum of antigens targeted by T cells and allowing T cell-mediated cytotoxic activity through the establishment of a cytolytic synapse. Endowing BsAbs with antigenic targeting and co-stimulatory structural domains (4-1BB/CD28) agonism could transcend the limitations of liver toxicity or poor efficacy caused by co-stimulatory signaling agonism alone [214, 215]. Transformable CARs were generated by constructing bispecific antibodies (MicAbodies) to bind to the inactive NKG2D domain, the extracellular portion of the CAR, allowing for flexible targeting of CAR-T cells to bulk masses of TAAs via MicAbodies [215, 216]. Some researchers extended this platform beyond tumors, demonstrating its potential for killing HIV-infected CD4⁺ T cells [217] (Fig. 6B). In mouse and Macaca fascicularis models, CLDN18.2 × 4-1BB-targeted BsAbs could evoke robust anti-tumor efficacy and durable immune memory responses via activating tumor-local 4-1BB signaling without inducing CRS or hepatotoxicity [215]. The combination of immune checkpoint blockade and 4-1BB agonism offered a renewed direction for enhancing immunotherapy efficacy. Specifically, tetravalent PD-L1 × 4-1BB BsAbs could convert “cold tumors” to “hot tumors”, prominently reactivating exhausted T cells to exert anti-tumor activity [218]. The phase I/II clinical trial reported that CD3 × GD2-targeted BsAb-armed T cells, in combination with IL-2 and GM-CSF, could enhance efficacy in treating neuroblastoma, exhibiting massively enhanced tumor-antagonizing activity and a favorable safety profile [219].

TsAb

Building on the BsAb foundation, tri-specific antibodies (TsAbs) further enhance the tumor-suppression capacity of immune cells by targeting multiple antigens simultaneously and introducing co-stimulatory signals [220] (Fig. 6B). The CD38 × CD3 × CD28-targeting TsAb in myeloma mice could induce vigorous immune responses, efficaciously facilitating T-cell activation. The co-stimulatory structural domain CD28 served as an additional target for myeloma cells, thus strengthening the specific recognition and eradication of tumor cells [220]. Additionally, tri-specific antibody construction introduces an Fc mutation for ablating FcγR binding, reducing adverse effects related to non-specific cytokine release. Inspiringly, IMT030122, a tri-specific antibody, could promote CD8⁺ T cell activation via targeting EpCAM, 4-1BB, and CD3, executing far-reaching tumor suppression [221]. Ye and colleagues validated that tri-specific nano-antibodies targeting PD-L1, 4-1BB, and NKG2A (or TIGIT) could effectively trigger NK/T cell activation, while engaging co-stimulatory and co-inhibitory receptors to initiate innate and adaptive immunity, demonstrating greater efficiency in promoting immune cell proliferation and activation compared to BsAbs that regulate a single immune target [222]. Likewise, GNC-038, a tetra-specific antibody was demonstrated to be a highly proficient nexus, further suppressing the CD19⁺ tumor progression [223]. A pioneering platform has recently been conceived that employs monoclonal antibodies to induce NK cells to exert ADCC function. This innovative approach, which does not entail genetic modification, allows for the production of “off-the-shelf” CAR-like NK cells and facilitates the facile targeting of various TAAs [224].

Cell engagers

Cell engagers serve as a bridge between TAAs and immune cells, facilitating the formation of a lytic immune synapse that promotes cytolytic protein secretion and enables the sequential killing of tumor cells even at low cell engager concentrations [225, 226]. Notably, the effector cell recruitment by cell engagers typically follows the targeting of TAAs, thereby ensuring tumor-specific immune cell activation. Predominantly, cell engagers are comprised of distinct scFvs, whereas those retaining the Fc domain are often engineered to abrogate its functional activity, thus mitigating the risk of off-target T cell activation or lysis and better unleashing the potential of immunotherapy [227].

Bispecific cell engagers

Bispecific T cell engagers (BiTEs) have paramount implications for battling cancer cells. Representing a class of tandem scFv BsAbs, BiTEs could specifically activate T cells by recruiting them through a binding domain targeting the CD3ε chain, and another arm targets TAAs (Fig. 6C). Given the diminutive dimensions of BiTEs, T cells and tumor cells maintain a proximate position following their conjugation via this construct, thereby facilitating the formation of a cytolytic synapse that enables the release of granzyme B or perforin from T cells [228]. Notably, this BITE-mediated formation of cytolytic synapses doesn’t depend on MHC antigen presentation and TCR recognition, implying that BiTE may counteract the immune escape dilemma and reinforce CAR-T cell recognition. Furthermore, BiTEs display exceptional properties compared to conventional BsAbs, including eliciting efficacious cell killing at minimal concentrations without the necessity of prior T-cell activation. Tarlatamab (AMG757), a BiTE product targeting DLL3, remarkably retarded cancer growth for small-cell lung cancer (SCLC) treatment [229, 230], and the FDA is currently reviewing the biologics license applications submitted for this product. By employing a similar strategy to BiTEs, another study designed mesothelin-targeting CAR-T cells capable of secreting a FAP-specific T-cell engaging molecule (TEAM), which remodels the tumor stroma and mobilizes T-cell activity to enhance targeting efficacy against pancreatic ductal adenocarcinoma (PDAC) [114].

Bispecific killer cell engagers (NKCE) could recruit NK cells and target tumor sites by replacing the anti-CD3ε structural domain in a BiTE with CD16, possessing excellent anti-tumor activity and inducing deeper cancer regression in pre-clinical studies. AFM13, a prominent representative of NKCE, could recruit NK cells to CD30⁺ tumor regions by binding CD16 on the NK cell surface [231]. Combining AFM13 with in vitro-induced NK cell expansion may tackle the hurdles posed by tumor recognition and NK cell activation, offering a novel tactic for treating CD30⁺ hematological malignancies [231]. An innovative alkaline phosphatase (ALP)-responsive and transformable bispecific conjugate (Supra-BiCE) has been validated to mobilize NK/T cells by self-assembling into nanoribbons and transforming into long nanoprotofibers in regions of elevated ALP expression, thereby enhancing the retention and accumulation of NK/T cells within the TME [232].

Cytotoxic cell engagers are developed by bispecific antibodies that are specific for the non-ligand-binding site of the high-affinity immunoglobulin G receptor (FcγRI). FcγRI, expressed in cytotoxic immune cells, includes monocytes, macrophages, dendritic cells, and interferon-gamma (IFN-γ)-stimulated polymorphonuclear neutrophils, and could recruit cytotoxic immune cells to the tumor site for cytotoxicity killing. In preclinical studies, targets of HER2/neu [233] and epithelial cell adhesion molecule (EpCAM) [234] boosted the anti-tumor responses and slowed tumor occurrence and progression.

Multispecific cell engagers

Multi-specific cell engagers targeting multiple TAAs significantly improve the toxicity profile of eliciting anti-cancer responses, exhibiting pleiotropic activation and synergistic effects of immune cells and achieving broader functionality compared to bispecific cell engagers. Tri-specific T cell engagers (TriTEs) integrating co-stimulatory signals into BiTEs could potentiate T cell activation and augment tumor-antagonizing responses, while introducing additional TAA targets mitigates the risk of immune escape due to antigen loss. Tapia-Galisteo and colleagues devised a TriTE for CRC treatment that contains CD3, EGFR, and EpCAM-specific recognition structural domains, which could specifically eliminate EGFR⁺/EpCAM⁺ tumor cells, whilst not exerting cytolysis on double-negative cells or inducing T-cell activation alone [235]. TriTE targeting human serum albumin (HSA), CD3, and Claudin 18.2 (CLDN18.2) manifests considerable tumor suppression of gastric and pancreatic cancers in xenograft mouse models, emphasizing that multi-targeting fueled the anti-cancer activity [236]. Endowing immune cells with resistance to inhibitory signals via TriTE could effectively circumvent the obstacles posed by immunosuppression in solid tumors [237]. The dual inhibitory immune checkpoint (ICP)-targeted nanobody-based TriTE, designed for HLA-G × PD-L1 × CD3, exhibited enhanced tumor suppression and prolonged survival in a human-derived non-small cell lung cancer (NSCLC) xenograft model utilizing immunodeficient mice [238]. Additionally, checkpoint inhibitory T cell-engaging antibody (CiTE) mitigates immune escape risk induced by PD-L1 upregulation and reduces the incidence of immune-related adverse events [237, 238]. Herrmann and colleagues have validated that αCD3 × αCD33-targeted CiTE has high selectivity for CD33⁺/PD-L1⁺ cells in vitro and in mouse models. Interestingly, CiTE did not target solely PD-L1⁺ cells, highlighting CiTE's satisfactory safety profile in addition to markedly reinforced anti-tumor efficacy [239]. Ding et al. evaluated the functional activity of nanoantibody-based TriTEs in a xenograft mouse model, demonstrating that TriTEs possessed a superior capacity for T-cell infiltration compared to BiTEs while eliciting no toxic responses [240]. Moreover, combining multispecific T-cell engagers with oncolytic virotherapy also offers an interesting perspective for addressing the challenge of tumor heterogeneity and T-cell infiltration [241].

Various studies have endeavored to develop tri-specific Killer Engagers (TriKEs) that enhance BiKEs functionality [242,243,244] (Fig. 6C). Cross-linking of IL-15 to an antibody manifests to significantly improve NK cell survival in vivo [245]. Felices et al. designed a novel tri-specific killer engager comprising scFvs that specifically bind to CD19 and the NK cell activation receptor CD16, with both scFvs linked to IL-15 [246]. IL-15 TriKEs could restore NK cell functional deficits, such as impaired cytokine secretion and exhaustion, in patients with chronic lymphocytic leukemia (CLL), offering an optimized strategy for better battling tumors. The IL-15 portion linked to scFvs specifically acts on NK cells, inducing NK cell activation and proliferation whilst avoiding off-target cytotoxicity [246]. TriKEs could also target C-type lectin domain family 12 member A (CLEC12A) in AML, showing superior efficacy compared to targeting CD33 [247]. Engineering T cells to express CARs targeting GPC2 and simultaneously secreting GD2 × CD16a bispecific innate immune cell engagers (BiCEs) could fuel the anti-cancer activity of immunotherapy [248]. GPC2 and GD2 are both targets of neuroblastoma, whereas CD16a is expressed on innate immune cells, namely NK cells or macrophages, within the TME. This approach, by simultaneously targeting two TAAs (GPC2/GD2) and secreting BiCEs to activate bystander innate immune cells, offers a potent and multifaceted strategy for treating neuroblastoma. In a nutshell, multispecific cell engagers offer a more comprehensive range of antigen-targeting possibilities and more sophisticated signaling capabilities, which collectively enhance anti-tumor capabilities. Furthermore, development of this multispecific platform will address the limitations of conventional bispecific cell engager signaling and demonstrate a more efficacious avenue for activating and targeting immune effector cells.

The application of multi-antigen targeting cell strategies

OR-gate emphasizes broadening tumor recognition by selecting combinations of multiple TAAs with low co-expression levels in critical normal tissues. This approach aims to address the challenges of tumor heterogeneity and immune escape. Specifically, in hematological malignancies, well-validated targets such as CD19 and BCMA are frequently employed. Multi-targeting strategies often combine these established targets with other antigens to enhance tumor coverage and reduce the likelihood of recurrence. Examples include combinations such as CD19 × CD123 [249], CD19 × CD70 [250], and CD19 × CD20 × CD22 to prevent CD19-negative relapse [251]. The complex immunosuppressive TME in solid tumors presents more sophisticated challenges for target selection [252]. Combining TAAs that target different resistance mechanisms, such as pairing growth factor receptor targets with stromal targets or immune checkpoint targets, can overcome resistance mediated by the TME or immune suppression [253]. Combinations of TAAs targeting different resistance mechanisms, such as combining growth factor receptor targets with stromal or immune checkpoint targets, can overcome TME or immunosuppression-mediated resistance [253], e.g., PD-L1 × CD19 [254], PD-L1 × HER2 [255], PD-L1 × GPC3 [256]. Introducing chemokine receptors that match the chemokines produced by the tumor conferred specific homing ability of CAR-T cells to the tumor, thereby increasing immune infiltration. Constructed EGFR/CXCR5 CAR-T cells demonstrated enhanced migration to CXCL13-positive tumors in preclinical models [257]. An early stage I clinical study is evaluating the therapeutic efficacy of EGFR/CXCR5 CAR-T cells (NCT05060796) in patients with NSCLC (often with high levels of secreted CXCL13). In AND-gate design, the combination of two TAAs with characteristic expression in tumors and limited expression in normal tissues could effectively improve the specificity and safety of treatment, such as CEA × MSLN [258]. Some critical components of the TME in solid tumors often serve as limiting factors for targeting, such as CAFs. Combinations of the target FAP, which targets CAFs, with other TAAs have been explored, including FAP × MSL [114, 259], FAP × CLDN18 [260], and others. For NOT-gate design, iCAR targets should be selected based on their restricted expression in key normal tissues that need to be protected from off-target effects. For instance, in hematological malignancy treatment, CD93 is predominantly absent from non-hematopoietic tissues, and CD93-targeted CAR-T cells demonstrated efficacy in clearing AML in preclinical models [261]. However, human endothelial cells exhibit high expression of CD93, and the incorporation of iCAR to recognize endothelial cell-expressed CD19 significantly enhances the precision of CAR-T cell discrimination [261]. A promising and versatile strategy involves targeting HLA molecules via iCAR, which addresses the loss of heterozygosity phenomenon prevalent in tumor cells. This approach has been employed in several studies in combination with various TAAs [200, 202, 262, 263].

Multi-specific antibodies and cell engagers offer similar advantages as OR logic-gated CARs, reducing the potential for tumor escape by engaging multiple TAAs, but each has its own functional focus [264]. Multi-specific antibodies emphasize the incorporation of additional functional modules, and the selection of their targeted TAAs is designed to leverage the intrinsic functional properties of the antibody itself to enhance therapeutic efficacy. A notable example is the FDA-approved bispecific antibody, Amivantamab (EGFR/cMET BsAb), which is used for treating non-small cell lung cancer [265]. Amivantamab blocks EGFR-cMET receptor activation, facilitates receptor degradation, and triggers Fc effector-mediated ADCC [265]. Clinical trials in NSCLC are currently assessing the efficacy of a bispecific antibody designed for simultaneous dual immune checkpoint blockade (TIM-3/PD-1) [266] (NCT04931654). Additionally, the tri-specific antibody CS2009, targeting PD-1/VEGFA/CTLA-4, is undergoing Phase I clinical studies (NCT06741644). Cell engagers, conversely, are more inclined towards recruiting and activating endogenous immune cells to facilitate endogenous killing. Their TAA combinations are primarily focused on achieving efficient immune cell engagement and activation. Blinatumomab, the CD19 × CD3 BiTE approved by the FDA for treating ALL, is a prime example. Clinical studies of TriKEs targeting HER2 × CD16 × NKG2D, which activates both NK cells and T cells, in patients with advanced solid tumors initially demonstrated good tolerability and promising anti-tumor activity [267].

Optimal TAA combinations for multi-targeting

Searching clinical trials on “ClinicalTrials.gov” reveals a growing trend of multi-targeting gene-edited cells in CAR therapy over the past 5 years. More than 50 studies related to “dual CAR”, “multi-specific CAR”, “BsAbs”, and “cell engagers” have been conducted, encompassing both hematological and solid tumors. These studies converge on optimizing TAA combinations, including CD19, BCMA, HER2, CLDN18.2, EpCAM, PSMA, and other relevant candidates, which are overexpressed in various cancers such as hepatocellular carcinoma, breast cancer, pancreatic cancer, and prostate cancer. Multi-targeting strategies are geared towards bypassing the essential roadblocks associated with single-target approaches, such as antigen loss, tumor heterogeneity, and resistance, thereby potentially enhancing the efficacy and durability of immunotherapy. While preliminary clinical outcomes provide a new dawn for immunotherapy, additional research is imperative to comprehensively investigate and adress safety and long-term efficacy concerns.

Dual targets

In designing dual-targeted gene-edited immune cells, CD19, CD20, and BCMA remain the primary targets to attack hematological malignancy. For solid tumors, HER2, PSMA, EGFR, VEGF, and EpCAM are the prominent focuses, often combined with inhibitory immune checkpoints in multi-target designs. Numerous studies have integrated inhibitory immune checkpoints into multi-target designs, emphasizing the importance of blocking inhibitory signaling pathways like PD-1/PD-L1 to fuel exhausted T cells and neutralize the immunosuppressive effects within the TME [268, 269]. The NK cells targeting PD-L1 and MICA/B exhibited a synergistic enhancement in cytotoxic activity within lung cancer models, leading to a substantial inhibition of tumor progression, potentially through the extensive pyroptosis mechanism induced by CAR-NK cells [270]. Developing armored CARs integrating cytokine receptors constitutes an alternative strategy to counteract T-cell exhaustion. The localized delivery of CD44/CD133 dual CAR-T cells, engineered with IL7Rα armoring to specifically target glioma stem cells, elicited sustained tumor regression in GBM and effectively mitigated T-cell exhaustion [271]. This therapeutic approach is currently under investigation in Phase I clinical trials (NCT05577091). The Clinical trials concerning dual-target combinations are summarized in Table 5. Notably, substantial concerns have been voiced about the potential risks linked to unpredictable cell proliferation and toxicity. Ouyang et al. engineered PD-1 knockdown CAR-T cells by introducing anti-PD-1 short hairpin RNA (shRNA) into T cells, potentially alleviating cell exhaustion while concurrently mitigating safety concerns [272].

Table 5 Dual targets combination of CAR-immune cells

Multi-targets

Preliminary clinical data of multi-targeting strategies demonstrated improved specificity and affinity of CAR-immune cell therapy while reducing toxicity to normal tissues, showing encouraging results. Applying multi-targeting strategies to CAR could provide cancer patients with more accurate, efficacious, robust, and secure therapeutic options. A recent study conducted by Bubb et al. demonstrated the therapeutic potential of targeting hematopoietic cytokine receptors in AML, revealing a marked delay in tumor progression and extended survival in NSG mice engrafted with mixed HCR⁺ leukemia cells through the development of trivalent CAR-T cells targeting KIT, MPL, and FLT3 [273]. Furthermore, Bubb and colleagues engineered a BiTE incorporating SCF, TPO, and FLT3LG triple ligands along with an anti-CD3 scFv as an alternative therapeutic strategy, which conferred a dose-dependent tumor suppressor effect on T cells in in vitro experiments [273]. Several ongoing studies are enrolling volunteers to investigate multi-specific CAR-T cell therapies, including target combinations such as CD19 × CD20 × CD22, mesothelin × GPC3 × GUCY2C, and CLDN6 × GPC3 × mesothelin × AXL (Table 6). Moreover, over 10 studies have explored multispecific antibodies and cell engagers targeting T cells and NK cells (Table 6). For example, a clinical trial is underway to evaluate the anti-tumor efficacy and bio-safety of SENTI-202, a three-input OR-NOT logic-gated product, in treating AML, CD33⁺ and FLT3⁺ hematological malignancies (NCT06325748).

Table 6 Multi-target combination of CAR-immune cells

Developing multi-targeting immune cell therapies has secured significant milestones in hematological malignancy treatment, including ALL, AML, non-Hodgkin lymphoma (NHL) and diffuse large B-cell lymphoma (DLBCL), resulting in slowed tumor occurrence and prolonged patient survival. Clinical trials are currently being conducted for solid tumors, including pancreatic, prostate, breast, liver, ovarian, and cervical cancers (Table 7), with some preliminary positive outcomes. Besides, the multi-targeting strategy is not solely confined to treating neoplastic diseases but has also been extended to managing autoimmune disorders and viral infections, with related clinical trials currently underway or pending (NCT06548620, NCT06451159, NCT04648046). Crucially, multi-targeting strategies still present diverse challenges, including antigen selection accuracy, CAR design optimization, and strategy development to maintain efficacy while reducing treatment-related adverse effects.

Table 7 Current clinical applications in disease

CAR delivery methods: Viral and non-viral approaches

Transient modification

Transient expression of CAR

To completely eradicate all cancer cells, it is essential to ensure stable and persistent CAR expression in engineered immune cells that circulate into the body [274]. Stable CAR expression allows sufficient time for immune cells to specifically target and eliminate all tumor cells, while durable CAR expression provides a reliable guarantee against tumor recurrence. Conversely, in diseases beyond cancer, such as fibrosis, autoimmune disorders, metabolic diseases, and age-related diseases, CAR-T cell therapy could achieve a desired therapeutic effect by reducing the number of pathologically defective cells below the pathological threshold, rather than requiring complete eradication [118]. Transient CAR expression confers a more favorable therapeutic profile in the context of extratumoral diseases, especially enabling improved safety. While the conventional approach to generating CAR-T cells typically involves the permanent integration of a transgene through viral vector transduction, advances in RNA delivery technology now permit the transient expression of specific CARs by transfecting T cells with mRNA [275]. As mRNA doesn't integrate into the host genome, this approach avoids oncogenicity risk associated with random insertion of viral vectors [275, 276]. Moreover, transient CAR expression enjoys the characteristic of specific T cell targeting for a limited duration, thereby mitigating adverse reaction risks concerning sustained CAR expression [276]. Transient CAR expression allows for more precise and flexible control during the therapeutic process, enabling medical personnel to adjust therapeutic dosages on a case-by-case basis to manage potential toxicities. Advantageously, transient CAR expression reduces non-tumor-targeted toxicity and enhances biosafety [277]. Even if immune cells are redirected to healthy tissue via CAR-mediated targeting, the risk of off-target damage can be mitigated by the short duration of CAR expression [278]. In addition, programming T cells in vivo to generate CAR-T cells does not lead to cytokine release beyond physiological levels, as observed in ex vivo cultures, which could be attributed to the unaltered physiological environment of T cells. Furthermore, tumor-antagonizing immunity could be exerted without necessitating multiple rounds of in vitro expansion, a characteristic absent in isolated cultured CAR-T cells, which simplifies technical requirements and reduces production costs [277]. Nevertheless, transient CAR expression may attenuate the anti-cancer responses of T cells, thus necessitating repeated administrations to sustain stable therapeutic outcomes [278]. In a nutshell, transient CAR expression offers a safer and more flexible approach to CAR-T cell therapy, presenting clear advantages in risk reduction and enhanced treatment control, despite the repeated administration challenge.

Multiple strategies could achieve transient CAR expression in vivo, including delivery of mRNA encoding CAR, introduction of non-integrating vectors, and inducible CAR systems. By delivering lipid nanoparticles (LNPs) containing mRNA encoding CAR into cells, the translation of mRNA prior to its degradation enables CAR expression [42, 279]. Owing to the lack of CAR genetic information integration into the genome, effector cells lose CAR expression as a consequence of mRNA degradation (Fig. 7A). Selecting a delivery vector is crucial for determining the extent of transient expression in the mRNA-based strategy. Generally, vectors that do not integrate into the genome could permit transient expression of the inserted fragments, making certain viral vectors, such as adenoviruses and adeno-associated viruses, eligible candidates. One strategy facilitates transient CAR-T cell production in vivo following subcutaneous implantation by incorporating retroviral particles encoding CAR into scaffolds for T cell engineering, showing superior persistence compared to conventional CAR-T cells [280]. Non-viral vectors such as liposomes and nanoparticles exhibit comparable behavior in facilitating transient CAR expression. Injecting CAR-T cell-programmed nanoparticles represents a promising alternative to conventional CAR-T cell therapy, offering substantial benefits such as reduced costs, the capacity for large-scale production, and the potential to treat a broader patient population. Additionally, the in vivo programming approach obviates the need for invasive pre-treatment procedures, further enhancing patient convenience and safety. Transient CAR expression could be achieved by introducing inducible elements regulated by specific conditions, an approach known as “inducible CAR systems”, which has been extensively employed in myriad studies to investigate various aspects of CAR dynamics and functionality. Using the CRISPR/Cas9 system to integrate the genes of CAR constructs into the downstream region of T-cell activation-dependent promoters correlates CAR expression with T-cell activation and generates a positive feedback loop that enables the control of CAR expression by modulating the cell activation state [281]. The drug-controlled VIPER CAR system devised by Li and colleagues, comprising both inducible ON and OFF switches, could tailor T cell activation levels in real-time and limit CRS, thereby obtaining a favorable safety profile [282]. Moreover, researchers illustrated the manipulability of the VIPER CAR system in xenotumor-transplanted mice, showing enhanced functionality when contrasted with alternative drug-gated systems.

Fig. 7

Emerging novel strategies. A. Lipid nanoparticle (LNP)-mediated expression of CAR. LNPs enter cells via endocytosis and then degrade, releasing mRNA encoding the CAR, which is subsequently translated to promote CAR expression. Immune cells expressing CARs exert CAR-mediated anti-tumor effects, but this function gradually diminishes as the mRNA degrades. B. Application of gene editing technology in improving the function of immune cells C. Strategies for gene delivery a. Viral vectors. Lentivirus: Single-stranded RNA genome, enveloped, conical core capsid, capable of host genome integration. Adenovirus: Double-stranded linear DNA genome, non-enveloped, icosahedral capsid, episomal persistence without integration. Retrovirus: Single-stranded RNA genome, enveloped, spherical capsid, capable of host genome integration. b. Electroporation-mediated CAR mRNA delivery Electric field-induced membrane permeabilization enables intracellular uptake of CAR-encoding mRNA, followed by ribosomal translation to produce CAR proteins. c. Transposon-based genomic integration CAR-containing donor DNA and transposase are co-delivered into immune cells. The transposase recognizes specific sequences on the donor DNA, excises CAR sequences, and then mediates their site-specific integration into the host genome via sequence recognition. d. Nanoparticle-mediated plasmid delivery Nanoparticles serve as carriers to encapsulate CAR-encoding plasmid DNA

Transient expression of co-stimulatory factor

Artificial gene circuits enable the transient expression of accessory molecules in an inducible manner, fueling the anti-tumor activity and reinforcing the potencies of CAR-T cell therapy [189, 283]. The Uni-Vect system provides a novel paradigm for achieving transient expression of soluble cofactors [284]. The conventional approach to achieving transient expression entails employing a dual viral vector system, wherein a constitutively expressed CAR construct and an inducibly expressed cofactor are expressed, respectively. Nevertheless, the dual viral vectors engender the formation of heterogeneous cell populations, which in turn increases the cost of cell sorting and limits technique applicability. In contrast, the single-vector Uni-Vect system optimizes the dual-vector system via incorporating the NFAT gene, which is a transcription factor dependent on TCR activation [284]. The Uni-Vect system initiates the expression of downstream auxiliary molecules upon TCR-mediated tumor signal recognition, consequently activating the NFAT pathway while simultaneously reducing the adverse effects associated with the constitutive expression of adjuvant molecules [284]. Smole et al. further fortified CAR-T cell functionality by loading diverse adjuvant molecules into the Uni-Vect system, emphasizing that the system could augment tumor-antagonizing responses through IL-12, mitigate CRS via IL-6α, and stimulate CAR-T cell proliferation through FOXO1 [284].

Gene editing technology

Iterative updates in gene editing technology facilitated the next generation of advanced engineered cells. Traditional genetic engineering involves introducing donor DNA with flanking sequences homologous to the target site to achieve homologous recombination (HR) [285]. The wider application of HR is constrained by its inefficiency in many organisms. Programmable nucleases in genome editing recognize and cleave specific DNA sequences, potentially producing double-strand breaks (DSBs), which are subsequently repaired by non-homologous end joining (NHEJ) or homology-directed repair (HDR), depending on the availability of a donor DNA template [286]. NHEJ-mediated repair requires no template and connections are random, whilst random insertions, introductions or nucleotide deletions at the break sites may result in the loss of function of corresponding proteins [287]. Consequently, NHEJ could alter specific genes to achieve gene knockouts, whereas HDR could integrate donor DNA into DSB sites with greater precision, thereby enabling more accurate gene editing [288].

Gene editing endonucleases, namely DSB inducers, are partitioned into meganucleases (MN), zinc finger nucleases (ZFN), transcription activator-like effector nucleases (TALEN), and clusters of regularly interspaced short palindromic repeats (CRISPR)-associated nucleases (Cas) based on structure. Genetic engineering modifies the specificity of MN, an endonuclease, by generating proteins that incorporate MN structural domains or by modifying protein residues within the DNA-binding domain. However, the paucity of natural MNs and the complexity and inefficiency of modifying MN specificity constrain MN application [289]. ZFN, a DNA-binding protein comprising at least three zinc finger domains that each recognize three bases and bind to DNA via a helix, could target unique DNA sites within a cell by designing two ZFNs to recognize complementary strands of the same double-stranded DNA [290]. Similar to ZFN, TALEN employs a comparable methodology and serves as a competitive alternative, which is a fusion protein consisting of a transcription activator-like effector (TALE) and a FokI endonuclease [291, 292]. TALE contains multiple conserved repeat units with residues at positions 12 and 13 responsible for nucleotide recognition [292]. The ability to redirect TALE specificity by redesigning these residues is more straightforward than modifying ZFN, thereby making TALEN a more competitive candidate for gene editing applications [293]. Frustratingly, all these approaches rely on intricate protein engineering to target specific DNA sequences, and delivering the proteins into cells for multiplexed editing poses additional obstacles, which constrain their extensive use in genome manipulation [293].

In contrast to programmable nucleases that recognize target sequences through protein-DNA interactions, the CRISPR/Cas9 system in gene editing relies on RNA-guided nucleases. The CRISPR/Cas9 system has ushered in a significant advancement in biological research, characterized by reduced editing costs, enhanced editing flexibility, and efficiency. The CRISPR system generates pre-crRNA and tracrRNA, which form a complex with Cas9, and the spacer sequence within the crRNA guides Cas9 to cleave the target DNA, while modifying the target sequence simply requires adjusting the sgRNA sequence [294, 295]. PAM is a short sequence motif (approximately 2–5 bp) positioned adjacent to the crRNA target sequence on invading DNA, and plays a pivotal role in sgRNA-guided recognition and cleavage of target sequences by Cas9 [295]. The superior efficiency and accuracy of CRISPR/Cas9 targeting, coupled with its extensive range of targeting genes and versatility, offer robust technical support for achieving complicated gene editing.

The CRISPR/Cas technology shows bright therapeutic prospects in multifaceted disciplines, including the development of antimicrobial agents to target highly toxic and drug-resistant bacteria [296], correcting CFTR gene defects in cystic fibrosis patients [297], preventing muscular dystrophy [298], and treating viral infections [299, 300]. Employing gene editing technology in cancer treatment has increasingly become intertwined with immunotherapy, particularly in CAR-T cell production (Fig. 7B). Notably, the utilization of autologous T cells, currently the predominant method for CAR-T cell production, faces numerous challenges, such as a lengthy production process, the need for personalized customization, high treatment costs, and the lack of sufficient quality and quantity of autologous immune cells in many late-stage cancer patients due to prior radiation and chemotherapy [301]. Reassuringly, allogeneic donor T cells, an off-the-shelf product, serve like a precision scalpel, adeptly slicing through the intricate knots of autologous CAR-T cell production, providing an efficacious and generalized therapeutic strategy for patients [302]. However, allogeneic T cells expressing endogenous TCRs may recognize alloantigens in the recipient, leading to GvHD, and the host's immune system may also mount a rapid immune rejection against allogeneic T cells via potent immune responses to HLA molecules on their surface [303, 304]. Utilizing ZFN and TALEN technologies to knock out endogenous TCRs in T cells has been demonstrated to effectively prevent the occurrence of GvHD [305, 306], while the additional elimination of HLA-A molecules on allogeneic T cells could significantly mitigate host immune rejection [307]. TALEN technology has also been reported to eliminate TRAC, CD52, and deoxycytidine kinase from CAR-T cells to generate universal CAR-T cells with an efficacious tumor-inhibitory effect [308, 309]. Utilizing CRISPR/Cas9 to target and eliminate B2M and TRAC genes generates CAR-T cells with enhanced lytic activity and reduced adverse effects, providing a safer, more stable, and durable therapeutic option [310]. Nevertheless, the independent processes of silencing endogenous genes and integrating the CAR transgene lead to a heterogeneous population of CAR-T cells, characterized by variable levels of gene modification and CAR expression. Georgiadis et al. integrated CAR expression alongside CRISPR/Cas9-mediated gene editing via embedding sgRNA elements into the ΔU3 3′ long terminal repeat (LTR) of a CAR-encoding lentiviral vector, and the subsequent cell selection process, facilitated by magnetic beads, could effectively generate high levels of TCR⁻/CAR⁺ edited cell populations [311]. Additionally, the random integration of CAR transgenes into T cells poses potential risks, including oncogenic transformation, transcriptional silencing, and clonal expansion. To tackle these challenges, Eyquem et al. strategically directed the insertion of a CAR transgene to the T-cell receptor α constant (TRAC) locus, a genomic site that governs the expression of endogenous TCRs, thereby attaining consistent CAR expression while selectively diminishing TCR levels. This innovative strategy amplifies the efficacy of tumor rejection, curtails tonic CAR signaling, and postpones T-cell differentiation and exhaustion [312, 313]. The implementation of a multifaceted gene editing strategy, involving the incorporation of multiple gRNA expression cassettes into the lentiviral vectors of a single CAR, facilitates the generation of universal CAR-T cells with disruption of the inhibitory FAS, PD-1, and CTLA-4 pathways, hence augmenting their durability and effectiveness in combating tumors [314]. Furthermore, overexpression of HLA-E [315] and CD47 [316] has been demonstrated to attenuate host NK cell and macrophage-mediated immune responses.

In addition to generating universal CAR-T cells, a vast array of studies indicated that the targeted knockdown of inhibitory receptors or signaling molecules, including PD-1, CTLA-4, Fas, and others, could significantly augment the therapeutic efficacy and persistence of these cells. This strategy effectively enhanced the performance of immune cells in counteracting the immunosuppressive TME, thereby enhancing the overall efficacy of immunotherapy [317,318,319]. Further, CRISPR/Cas9 technology could edit up to five genes for efficient multiple-gene editing. The CRISPR/Cas9 technology illuminates the viability of comprehensive genome engineering in mice, unveiling new frontiers for exploring entire signaling pathways and genes with redundant functions [320]. Inspiringly, the first human experiment testing the safety of genome editing with the multiple CRISPR/Cas9 system showed the initial safety and feasibility of autologous TCR-T cells after simultaneous knockdown of TRAC, TRBC, and PD-1 in treating patients with advanced refractory tumors [321]. Notably, the removal of inhibitory molecules is a double-edged sword in immunotherapy. Accordingly, it is essential to examine whether the ablation of particular inhibitory signals may culminate in uncontrolled cell proliferation or severe autoimmune responses.

Vector-mediated gene delivery

Currently approved commercial CAR-T products deliver CAR transgenes using lentiviral or retroviral vectors (Fig. 7C), an approach that, while generally effective, presents certain non-negligible limitations: 1) Viral vectors could be constrained in their capacity to accommodate cargo of significant size (5–15 kbp) and exhibit notable cytotoxicity [194]; 2) Introducing genes from viruses into the human genome poses substantial safety concerns [322, 323]; 3) The need to accommodate an ocean of gene cargoes necessitates sophisticated vector engineering, whilst their manufacturing process is inherently costly. Various studies concentrate on exploring non-viral gene delivery methods to overcome safety and financial challenges associated with viral vectors. Electroporation utilizes high-voltage electric fields to generate transmembrane pores, thereby facilitating the permeation of macromolecular carriers into the cellular interior [324] (Fig. 7C). However, the potency of electroporation is contingent upon the transient expression of CAR, which persists for only a few weeks. Furthermore, robust electric fields have been demonstrated to precipitate cellular dysfunction, including lipid peroxidation, protein denaturation, the generation of reactive oxygen species, and DNA damage, which ultimately culminates in a reduction in cell viability [325]. The relative simplicity, low cost, and suitability for mass production of transposon vectors have led multiple researchers to endeavor to investigate transposon-mediated gene delivery platforms such as Sleeping Beauty and PiggyBac systems [326, 327] (Fig. 7C). Despite their relatively diminished transduction efficiency compared to viral systems, they are perceived as a safer alternative. Nevertheless, malignant transformation of CAR-T cells was reported in two patients with B-cell malignancy treated with CD19-specific CAR-T cells generated via the PiggyBac transposon platform, indicating that the transposon system also possesses potential risks [328].

The feasibility and high efficiency of the CRISPR/Cas9 gene delivery strategy have been empirically validated in human cells, thereby establishing a crucial scientific foundation for accelerating the advancement of gene therapy and precision medicine. The CRISPR/Cas9 genome-targeting system could successfully correct ILR2A mutations in patients with autoimmune diseases [329], restoring cell signaling functions [329]. Additionally, the CRISPR/Cas9 genome-targeting system could replace endogenous TCRs with novel TCRs, thereby generating engineered TCR-T cells that exhibit precise retargeting to TAAs and display highly efficacious anti-tumor functionalities. However, knock-in targeted by the CRISPR/Cas9 system presents potential toxicity of the DNA template used, and evidenced less efficiency than AAV-mediated gene delivery [330]. Donor DNA delivery based on CRISPR/Cas9 technology, by incorporating CD19/TAG-72 scFvs into CD3ε to generate T cells expressing fusion proteins, could enhance tumor-targeting capacity and even demonstrate more efficacious oncolytic potential than CAR-T cells, in Raji cell xenograft models [331].

Using biochemical carriers, such as detergents and pore-forming toxins, enables the penetration of cell membranes to form pores that facilitate cargo passage [332] (Fig. 7C). Nevertheless, biochemical carriers also encounter significant obstacles, including incompatibility of cargo materials with carriers, which could be influenced by factors such as charge, size, and hydrophobicity; limited compatibility of carrier materials with target cells, which may hinder efficient endocytosis; and inherent potential toxicity.

Nanoparticles, which belong to a magnetic category of delivery vehicles, offer a panoply of substantial merits, including enhanced safety, efficiency, and controlled pharmacokinetic properties (Fig. 7C). The distinctive physical, chemical, and biological characteristics of nanoparticles have rendered nanoparticles a subject of interest across diverse disciplines, involving medicine, electronics, and the environment. Nanoparticles could load reagents and facilitate their targeted delivery to specific sites, navigating restrictions imposed by physiological mechanisms. Therefore, nanoparticle utilization in CAR-T cell production portrays an alternative, off-the-shelf product that is distinct from universal CAR-T cells, specifically enabling in vivo CAR-T cell generation [333]. In contrast to other delivery technologies, nanoparticle-mediated delivery doesn’t require mechanical permeabilization of the cell membrane and could bind to receptors on target cells to initiate endocytosis for cargo delivery. Following administration, nanoparticles could continuously program effector cells to generate sufficient quantities of CAR-T cells, which elicit vigorous immunity responses comparable to those of commercial CAR-T cell products [334].

Compared to the labor-intensive and complex ex vivo production employing traditional viral vectors, the nanoparticle-mediated generation of CAR-T cells depicts a novel strategy. The nanoparticle-mediated generation of CAR-T cells necessitates only a few straightforward steps and could be manufactured at multiple centers, thus offering significant economic benefits. Notably, generating CAR-T cells in vivo is particularly challenging because it requires reprogramming senescent and depleted T cells, which are often compromised in cancer patients. Such impairment, frequently a result of extensive pre-treatment, can significantly undermine the efficiency of the reprogramming process. By loading nanocarriers with TCR-stimulating signals or pro-cell growth factors, transduced T cells exhibit boosted survival and proliferation [335]. The co-delivery of anti-programmed cell death-1 (anti-PD1) and anti-tumor necrosis factor receptor superfamily member 4 (anti-OX40) via dual immunotherapy nanoparticles has been demonstrated to induce a more pronounced T cell activation alongside an enhanced immunological memory, consequently exhibiting markedly superior therapeutic outcomes [336]. Lipid nanocarriers epitomize a gold standard for non-viral gene delivery vectors. In the heart failure mouse models, CD5-targeted lipid nanoparticles could effectively deliver modified mRNA to T cells, thereby achieving transient and efficient expression of CD19-CAR [337]. By devising activated lipid nanoparticles (aLNPs) that mimic antigen-presenting cells (APCs), Metzloff et al. achieved one-step activation and transfection of primary human T cells [338]. This reduces the complexity of producing CAR-T cells [338]. Additionally, injecting aLNPs effectively decreased tumor size in the murine xenograft model of leukemia, showing their potential as a fast and efficient platform for making mRNA CAR-T cells [338].

Perspective and outlook

Considering the prevailing complexities associated with immunotherapy, the selection of strategies necessitates a meticulous evaluation of existing impediments, particularly in the context of solid tumor treatment, wherein tumor heterogeneity and off-target toxicity emerge as predominant considerations.

The AND-gate endeavors to partially emulate the targeting of TSAs through integrating diverse TAAs, thereby augmenting precise antigen-specific recognition. Additionally, precise tumor targeting offers the potential benefit of reducing non-specific immune activation, thereby reducing the risk of exhaustion. Nevertheless, this strategy introduces a dual-edged effect: while it enhances the safety profile of targeting mechanisms, the AND-gate imposes more stringent requirements on the co-expression levels of antigens on the tumor surface, thereby posing a potential risk of CAR-immune cell activation failure due to the downregulation or complete loss of expression of a single antigen. OR-gate aims to effectively broaden the coverage of tumor heterogeneity by integrating diverse TAA targets. However, this strategy faces the challenge of rigorously selecting target combinations to balance therapeutic efficacy with potential off-target risks. The ultimate efficacy of the OR-gate is contingent upon the quality of the selected TAA combinations. In the absence of rigorous preclinical validation and systematic design, indiscriminate stacking of multiple TAAs may exacerbate the risks of off-target effects and non-tumor toxicity [339, 340]. Moreover, broader tumor targeting of the OR-gate might induce heightened immune stress, potentially increasing the risk of tumor immune escape and susceptibility to self-exhaustion [339]. NOT-gate improves safety by excluding certain normal tissues from targeting, which is valuable in specific situations where there are well-defined ‘safety antigens’ in key normal tissues. Simultaneously, NOT-gate expands the targetable antigen range. The challenge, however, resides in identifying and validating suitable and broadly applicable safety antigens, coupled with the necessity to meticulously balance activation and inhibition signaling to ensure optimal functional efficacy. SUPRA CAR presents advantages with a dose-adjustable safety profile, offering precise modulation of efficacy and safety, although its clinical value awaits further validation through additional data.

Multi-specific antibodies and cell engagers embody a distinct therapeutic paradigm compared to gene-edited immune cells. Multi-specific antibodies and cell engagers, with several FDA-approved products currently available [264], signify notable advancements in immunotherapy, particularly in the short term and for specific tumor types. The efficacy of a multi-specific antibody/cell engager, whether functioning as a bridge between tumor cells and immune cells or blocking inhibitory signals, is closely associated with the functional status of the patient's intrinsic immune system. In contrast, gene-edited immune cells function as "living drugs" engineered to exert a more proactive tumoricidal effect, potentially eliciting distinct therapeutic responses and exploring the possibility of achieving profound, durable remissions. Although gene-edited immune cells and multi-specific antibodies/cell engagers have demonstrated substantial efficacy in treating hematological malignancies, within the more complex domain of solid tumor therapy, they encounter shared challenges of the TME. BsAb/BiTE were designed to target specific components/immune checkpoints within the TME to counteract the immunosuppressive effects of the TME and enhance therapeutic efficacy [341,342,343]. Conversely, gene-edited immune cells tend to address the complexity of the TME by reprogramming cellular functions. For instance, CAR-M cells aim to infiltrate and remodel the TME. In addition, CAR-T cells are being engineered with logic gates (e.g., targeting the TME cytokine—TGFβ/IL-4) and other modifications specifically designed to counteract the TME. In efforts to augment persistence and mitigate exhaustion, CAR-immune cells have undergone genetic modifications to enhance their intrinsic durability, metabolic adaptability, and resistance to exhaustion [344, 345]. Whereas the action duration of multi-specific antibodies/cell engagers is mainly influenced by pharmacokinetics [346]. However, by exploring strategies such as targeting costimulatory receptors, there is potential to endow multi-specific antibodies/cell engagers with enhanced and prolonged immune responses [347]. Finally, in terms of production and application, multi-specific antibodies/cell engagers often benefit from relatively mature manufacturing processes and their "off-the-shelf" potential. In contrast, CAR-immune cells face complex production and high cost limitations. However, their prospects are substantial, with advancements like allogeneic CAR therapies, in vivo CAR-T cell generation, and non-viral methods reducing complexity and costs, improving future application convenience.

Future development of multi-targeting designs relies on further optimization of CRA/antibody structure, balance of multi-gene co-expression, screening for better targets, and facilitating clinical translation (automated production, efficacy prediction). AI empowers the development of these strategies. The advancements in artificial intelligence, a focal point of the 2024 Nobel Prize, could drive CAR-immune cell development in multiple ways. In clinical translation, AI analyzes patients'baseline information, genomic, and proteomic data. The synergy of AI with Internet of Things (IoT) sensors streamlines the complex process of CAR-immune cell production and monitors the conditions during transportation and storage (Fig. 8).

Fig. 8

Application of Artificial Intelligence (AI) in Immunotherapy. A. AI plays a pivotal role in the production, storage, and logistics of CAR-immune cells by leveraging IoT sensors. B. AI enhances the precise identification of CAR targets through integration with CRISPR/Cas9 technology. C. AI continuously monitors immunotherapy patients via the detection of biochemical indicators. D. AI is capable of analyzing baseline information, as well as genomic and proteomic data from patients. E. AI facilitates the discovery of novel targets and the optimization of CAR construct structures. F. AI can screen suitable candidates for various gene-edited cell therapies, develop personalized treatment plans, and predict treatment outcomes

By analyzing patients'immune status, tumor characteristics, and genetic information, AI could screen suitable candidates for various gene-edited cell therapies and personalize treatment plans while providing predictions of treatment outcomes [348,349,350]. Ongoing surveillance of patients'physiological parameters and relevant biomarkers using AI helps in the early identification of potential adverse reactions, thereby enabling timely intervention and enhancing the safety of immunotherapy. Additionally, by integrating large-scale data analysis from scientific literature, AI could drive the structural optimization of CAR constructs and the discovery of new targets for immunotherapy [350]. The harmonious combination of AI with CRISPR/Cas9 technology enables precise identification of genetic targets, predicts, the prediction and minimization of off-target effects, and the optimization of guide RNA designs, thereby enhancing the specificity of CRISPR/Cas9 technology [351].

The determination of antibody-antigen recognition specificity is critical for immunotherapy, whereas conventional approaches frequently encounter limitations, including high costs and long experimental durations. Current computational models exhibit significant dependence on 3D structural data, yet technical barriers in data acquisition constrain their biological applications [352, 353]. Yan Huang et al. developed AbAgIntPre, an innovative deep learning framework that exclusively utilizes amino acid sequence data for antibody-antigen interaction prediction. The model was trained on high-quality antibody-antigen complexes obtained from SAbDab (a database that collects the structures of antibody and antigen complexes) and CoV-AbDab (a database that specializes in collecting data on antibodies that bind to coronaviruses), enabling broad-spectrum antigen interaction prediction [354]. Expectedly, this model has considerable potential for CAR target screening by sequence prediction of antibody-antigen binding, allowing rapid screening of high-affinity antibody fragments for CAR construction. This computational framework demonstrates significant potential for CAR target screening through sequence-based prediction of antibody-antigen binding affinities, enabling systematic identification of high-affinity antibody fragments critical for chimeric antigen receptor engineering. In addition, the SARS-CoV model constructed by Huang et al. validated the ability to predict mutant antigens [354]. This feature might be used for CAR design of new epitopes or mutant proteins targeting tumor antigens, bridging the bottleneck of the lack of structural data for tumor neoantigens. Though modelling patient-specific antigen binding, AbAgIntPre is expected to support individualized CAR design, thereby reducing the risk of off-target effects.

In CAR-T cell therapy, structural optimization of scFv constitutes a pivotal step in augmenting therapeutic effectiveness. Martarelli et al. implemented AI-driven computational platforms to perform molecular docking and targeted molecular dynamics analysis of disparate anti-CD30 monoclonal antibodies, enabling systematic screening of optimal scFvs for the construction of CARs. This approach exemplifies the potential of computational modeling platforms to catalyze the engineering of novel CAR constructs, while concomitantly minimizing financial expenditures and conserving resources [355]. CAR-Toner, an AI-based tool, could predict and optimize the tonic signaling of CAR-T cells, thereby improving their persistence and promoting expansion [356]. By integrating protein databases, structural biology, and advanced deep learning models, the CAR-Toner platform has been pre-trained on over 60 million protein sequences, which could calculate the positive charge patch (PCP) score of proteins and provide strategies to optimize the PCP score through induced point mutations. In the CLL1 antibody refinement, CAR-Toner analysis revealed that a series of CLL1 antibody variants generated by systematically reducing the PCP score eliminate tumors more efficiently compared to wild-type CARs [356]. Additionally, AttABseq represents a sequence-based computational platform that predicts alterations in antigen-antibody binding affinity, occasioned by antibody mutations, thus offering a novel strategy for antibody optimization [357]. Jin and colleagues employed AttABseq to determine optimization trajectories for SARS-CoV-2 VHH and Ebola virus antibodies, demonstrating strong concordance between computational predictions and experimental measurements that validated the platform's reliability [357]. This achievement provides a valuable lesson for CAR scFv affinity optimization. However, implementation necessitates integration of supplementary datasets (including tumor antigen-specific mutant binding profiles) and experimental validation in oncology field.

Furthermore, AI has demonstrated transformative potential in CAR antigen prediction and target optimization. Mason et al. developed an integrated framework combining CRISPR-Cas9-mediated antibody mutant library and training convolutional neural network-based antigen specificity prediction models, enabling high-throughput identification of functionally optimized therapeutic antibody variants from a multitude of sequences [358]. Additionally, machine learning models could predict and validate optimal metabolic pathways, thereby optimizing the metabolic processes of CAR-T cells and enhancing their efficacy against solid tumors, leading to metabolically enhanced CAR-T cell therapy products [359]. Employing nanogrid-based time-lapse imaging combined with visual artificial intelligence, Rezvan et al. identified a subpopulation of CD8-fit T cells with high migratory and continuous killing capabilities, which are primed to boost cancer treatment outcomes in CAR-T therapy [360].

Notably, in the early stages of AI applications in immunotherapy, the lack of appropriate datasets poses significant challenges for training initial machine learning algorithms. The complexity of various biological structures and patient information further complicates these efforts, which necessitates continued exploration. Nevertheless, the comprehensive AI platform development, from antigen selection to CAR-T cell structure optimization, production, clinical trials, and patient prediction, through the integrated synergy of multi-omics data such as genomics, proteomics, and imaging, will contribute to more precise gene-edited immune cell therapies [350].

Conclusions

While gene-edited immune cell therapies, particularly CAR-T cells, have achieved remarkable success in hematological malignancies, realizing their full potential as next-generation immunotherapy tools, especially in solid tumors, requires overcoming significant challenges. To surmount the limitations of CAR-T cells, extensive studies prioritize developing alternative gene-edited immune cells such as CAR-NK and CAR-M cells. Furthermore, harnessing multi-targeting strategies havs emerged as a critical direction to broaden targetable antigens, counteract tumor heterogeneity, and improve clinical efficacy. Innovative technologies like Boolean logic gates, SUPRA CARs, multi-specific antibodies, and cell engagers are accelerating the development of multi-targeting immunotherapies by providing enhanced flexibility, improved specificity, and simplified design frameworks for diverse targeting applications.

Progression in next-generation immunotherapy tools is being propelled by emerging technologies in gene editing and gene delivery. Synthetic mRNA and transient modification methods, CRISPR/Cas9-mediated cell engineering, and novel synthetic materials for gene delivery provide a boost to enhance tumor susceptibility to immunotherapies and achieve sustained specific-targeting capabilities of engineered immune cells. Looking ahead, the integration of AI holds immense promise in further optimizing these next-generation immunotherapy tools. AI-driven approaches can revolutionize target identification, CAR design, and personalized treatment strategies, paving the way for more effective and safer immunotherapies.

The future of next-generation immunotherapy is bright, with the convergence of tumor analysis, synthetic biology, and artificial intelligence promising to unlock even more sophisticated and effective therapeutic strategies [339]. Harnessing multi-targeting strategies in conjunction with these advancements holds the key to realizing the full therapeutic prospects of gene-edited immune cells and transforming cancer treatment. Strategically oriented towards the future, continued development and combined application of gene editing technologies, gene delivery strategies, and biomaterials will offer robust backing to support and optimize the breadth and security of targeting profiles. Diverse disciplines could combine fully exploit the therapeutic potential of engineered T cells, sparking activation and amplifying their potency. The integration of metabolic engineering and CAR-T cells represents a novel synergistic approach to regulate CAR-T cell persistence through epigenetic and phenotypic modifications, which may emerge as a promising avenue for future research. The synergistic alliance of oncolytic viruses and CAR-T cells unveils a strategic approach for confronting the dilemmas of solid tumors (NCT05393440), encompassing oncolytic viruses encoding BiTEs, cytokines, and immune checkpoint inhibitors [134, 361], as well as oncolytic viruses encoding CAR targets [362, 363]. Moreover, as data sources from human trials continue to be refine, machine learning may assume a crucial role in the future advancement of novel therapies, facilitating the prediction of optimal parameters for critical treatment factors. The convergence of tumor analysis and synthetic biology may spur sophisticated paradigm innovation for tumor antigen identification, empowering more effective responses to the complex challenges inherent in tumor therapy.

Box 1. Key Terms

1. MHC: Cell-surface proteins required for antigen presentation to T cells. CAR-T cell therapies bypass MHC dependency to directly target tumors

2. CD19/BCMA: CD19, B-cell surface marker targeted in leukemia/lymphoma therapies. BCMA: Plasma cell marker in multiple myeloma. In CAR-T cell therapy, CD19/BCMA are important targets, and their loss is an important cause of tumor recurrence

3. Antigen-negative clone: Tumor subpopulations evade CAR-T cell recognition by downregulating target antigens, resulting in poor therapeutic efficacy

4. CRES: Possible neurological side effects of CAR-T therapy, such as confusion or seizures

5. CRS: A systemic inflammatory response triggered by the rapid release of cytokines from activated immune cells is a common side effect of CAR-T therapy

6. Solid tumor: Cancers that form solid masses, such as lung or breast cancer, as distinguished from hematologic malignancies

7. Autologous/Allogeneic cell therapies: Patient/healthy donor-derived cells (e.g., T cells) engineered ex vivo and reinfused into the same individual/patients

8. Universal allogeneic CAR-immune cell platforms: Standardized CAR-immune cell therapies developed from healthy donors are available for multiple patients

9. Off-target toxicity: CAR-T cell therapy targets healthy cells/tissues and causes damage

10. Adaptor: Molecules that act as connecting bridges to enhance the adaptability and specificity of CAR-immune cells

11. Multitargeting: Enables immune cells to attack multiple antigens simultaneously, distinguishing it from single-targeting strategies and reducing the risk of cancer escape due to antigen loss