The first-generation (1G) CARs, consisting of a scFv fused to CD3ζ or FcγR, showed promising results in early preclinical trials. The adoptive transfer of ERBB2-specific CAR-T cells, which contain a scFv fused to the CD3ζ signaling domain, in BALB/c nude mice demonstrated their ability to recognize and infiltrate ERBB2-expressing tumor cells while slowing tumor growth (20). Similarly, studies have shown that tumor-infiltrating lymphocytes (TILs) can be engineered to express a CAR targeting the MOv gamma (MOvγ) antigen. This is achieved using a scFv derived from the MOv18 monoclonal antibody, which specifically binds to folate receptor alpha (FRα), a tumor-associated antigen overexpressed in many ovarian carcinoma cells. The MOvγ-specific CAR incorporates an FcγR signaling domain that can trigger functional responses upon recognition of the MOvγ antigen both in vitro and in murine models. In these in vivo experiments, mice received exogenous IL-2 following the transfer of engineered T cells (20, 21). These findings demonstrate the ability to redirect T cell specificity and effector functions toward a target surface antigen. However, translating this approach into clinical success was challenging, primarily due to the absence of co-stimulatory signaling domains in the 1G CARs. Unlike natural TCRs, which require co-stimulatory signals for full T cell activation and sustained function, 1G CAR-T cells exhibited poor in vivo persistence and failed to maintain a long-term antitumor response, resulting in reduced therapeutic efficacy (9, 22). Moreover, early clinical trials highlighted the risk of severe side effects, such as CRS, demanding adjustments to CAR design and safety mechanisms (17, 23).

The development of 1G CARs was a pioneering step in cancer immunotherapy. Despite several limitations, the lessons learned from these early designs were important in understanding the potential of targeting CARs against tumor antigens and guiding the development of more advanced generations of CARs. Ongoing advancements continue to enhance the clinical efficacy and safety of CAR-T cell therapy, bringing new hope to patients, especially those with resistant disease.

To address the lack of co-stimulatory signals in 1G CARs, 2G CARs were engineered to include a co-stimulatory signaling domain in addition to CD3ζ ( Figure 2A ). This design is based on the fact that T cells require two main signals to be fully activated: 1) antigen recognition, which occurs when the TCR expressed on T cells binds to a peptide/MHC complex presented by antigen-presenting cells (APCs), and 2) a co-stimulatory signal, which is provided by the trans interaction between the co-stimulatory receptors, such as CD28 on T cells, with their corresponding ligands, CD80/CD86, on APCs (24). As a result, one of the most significant advances in 2G CARs is the incorporation of a signaling domain capable of providing both activation and the necessary co-stimulatory signals in cis within the same receptor (25). CD28 and 4-1BB/CD137 are the most commonly used co-stimulatory domains in 2G CARs; delivering these secondary signals is critical for promoting T cell survival, proliferation, and cytokine production, thereby enhancing the antitumor efficacy of CAR-T cells (26). However, the type of co-stimulatory domain, whether CD28 or 4-1BB, has been shown to differ in several key aspects, leading to distinct functional outcomes of CAR-T cell therapy (27). CAR-T cells with CD28 co-stimulatory signals, referred to as CD28ζ CAR-T cells, undergo metabolic reprogramming predominantly toward aerobic glycolysis, leading to the early dominance of effector T cells and the development of effector memory (14). It has been observed that infusing CD28ζ CAR-T cells promotes rapid accumulation of T cells and cytokine release. This results in potent immune responses against tumor cells, which is particularly beneficial in rapidly growing cancers. However, in clinical settings, this has been associated with reduced CAR-T cell persistence and a higher incidence of CRS and immune effector cell-associated neurotoxicity syndrome (ICANS) compared to 4-1BB-based CARs (5). Incorporating the 4-1BB co-stimulatory signaling domain in CAR-T cell products has demonstrated metabolic reprogramming of T cells toward oxidative phosphorylation, promoting the development of central memory T cells, which is crucial for long-term immunity. 4-1BB signaling has also been associated with increased mitochondrial biogenesis, as well as improved CAR-T cell survival and persistence, which is consistent with clinical observations (14). Furthermore, 4-1BB signaling promotes a more regulated and gradual activation, which may help alleviate exhaustion and reduce the severity of CRS compared to CD28-based CARs (26, 28). Although clinical trials in ALL patients treated with CD28- or 4-1BB-co-stimulated CD19-specific CAR-T cells revealed comparable initial response rates, CAR-T cells with 4-1BB co-stimulatory domains appear to outperform those with CD28 in terms of clinical efficacy in CLL trials (29, 30).

2G CARs have demonstrated remarkable success in clinical studies, marking a significant step forward in the progress of CAR-T cell therapy, particularly for hematological malignancies. The clinical success of 2G CARs has been primarily driven by their ability to target CD19, a surface antigen expressed on the majority of B cell malignancies. This specificity has enabled CAR-T cells to effectively eradicate malignant B cells; however, it also results in the depletion of healthy mature B cells due to their shared CD19 expression (17). Landmark clinical trials using CD19-targeted CAR-T cells for B cell malignancies, such as ALL and CLL, have shown favorable outcomes, with many patients achieving complete remission (CR). In 2017, the FDA approved the first CAR-T cell therapies, Kymriah (tisagenlecleucel) and Yescarta (axicabtagene ciloleucel), for the treatment of B cell acute lymphoblastic leukemia (B-ALL) and diffuse large B cell lymphoma (DLBCL), respectively. These therapies have resulted in high response rates and durable remissions in patients with refractory or relapsed (r/r) ALL ( Table 1 ) (4).

Kymriah (tisagenlecleucel), an autologous CD19-targeted CAR-T cell product, was the first CAR-T cell therapy approved by the FDA. This approval followed a Phase 2 multicenter study that evaluated the safety and efficacy of this treatment in pediatric and young adult patients (up to 25 years old) with r/r B-ALL. This trial revealed an 81% overall remission rate with long-term CAR-T cell persistence following a single infusion, despite the occurrence of reversible high-grade toxic effects such as CRS and neurologic events; all responders tested negative for minimal residual disease (MRD) (18). This was an extension of the primary clinical trial called ELIANA (NCT02435849), the first Novartis-sponsored global CAR-T cell therapy registration trial. In 2018, it was approved for adult patients with r/r large B cell lymphoma who had failed at least two lines of systemic treatment, including DLBCL. This approval was based on the pivotal clinical trial called JULIET, which demonstrated an overall response rate of 52%, with CR achieved in 40% of patients with r/r DLBCL (31).

Yescarta (axicabtagene ciloleucel), an autologous CD19-targeted CAR T-cell product, was the second CAR-T cell therapy approved by the FDA in 2017 for treating patients with r/r B cell lymphoma who had failed at least two rounds of systemic chemotherapy. Yescarta was initially approved based on a Phase 1/2 clinical trial called ZUMA-1, which involved patients with DLBCL, with a minority having primary mediastinal large B cell lymphoma (PMLBCL) (32). This study found that 50% of the treated patients survived overall, and 41% remained progression-free for up to two years. Following this, a Phase 3 clinical trial was conducted to evaluate the safety and efficacy of Yescarta as a second-line treatment for patients with large B-cell lymphoma who did not respond to first-line chemoimmunotherapy or experienced a disease relapse within 12 months of their initial treatment (33). Yescarta offered an important new therapeutic option for patients with limited treatment options whose cancers had not responded to multiple lines of treatment, such as those with DLBCL, PMLBCL, and DLBCL arising from follicular lymphoma (FL).

Following the FDA’s initial approvals of CAR-T cell therapies in 2017, including Kymriah and Yescarta, several other CAR-T cell products have also received regulatory approval. Table 1 outlines all FDA-approved CAR-T cell therapies based on the 2G CAR structure.

Although the introduction of co-stimulatory domains, such as CD28 or 4-1BB, into 2G CARs improved CAR-T cell activation and persistence, there is still room for improvement in terms of efficacy, response durability, and side effect management. As previously mentioned, co-stimulatory molecules have distinctive features that can be utilized synergistically to enhance the performance of CAR-T cells. Third-generation (3G) CAR-T cells were developed to further improve the therapeutic potential of CARs. Instead of a single domain as in 2G CARs, 3G CARs incorporate two co-stimulatory domains, such as CD28 and 4-1BB/CD137 or OX40, into the same construct ( Figure 2B ) (34). This dual co-stimulatory design leverages the complementary signaling pathways of each molecule, aiming to boost T cell activation, persistence, and antitumor efficacy while also reducing immune exhaustion and improving safety profiles. Early studies suggest that 3G CAR-T cells may have greater potency and durability compared to their 2G counterparts, yet further clinical validation is needed (34). Clinical evaluation of 3G CAR-T cells has provided important insights into their therapeutic potential. A side-by-side comparison study of 2G and 3G CARs was conducted to assess their in vivo behavior and therapeutic efficacy in patients with r/r B-cell non-Hodgkin’s lymphoma (NHL) using CD19-specific CAR-T cells. In this study, the 2G CAR construct utilized CD28 as a co-stimulatory domain, whereas the 3G CAR construct included both CD28 and 4-1BB co-stimulatory domains. In the Phase 1 cohort of this ongoing SAGAN clinical trial (NCT01853631), 16 patients (11 with active disease and 5 with no measurable disease) received simultaneous infusions of both products. The 3G CD19 CAR-T cells exhibited superior expansion and persistence in vivo compared to their 2G counterparts, yielding clinical responses in 6/11 patients with active disease (3 CRs and 3 partial responses (PRs)). Additionally, 4/5 patients who were already in remission experienced durable and continued CR (35). The 3G CARs have demonstrated remarkable efficacy in patients with no measurable disease, particularly those in remission following autologous stem cell transplantation, indicating a potential benefit in preventing relapse when the disease burden is low. Despite these encouraging findings, the absence of antigen spreading observed in some patients may limit the durability of responses (defined as the duration from the onset of the initial response to disease progression or death for any reason) and increase the risk of tumor escape mechanisms. Antigen spreading refers to the phenomenon in which an immune response initially triggered by a targeted tumor antigen expands to recognize additional non-targeted tumor antigens (36). This highlights the need for strategies that can enhance the breadth and sustainability of the immune response in CAR-T cell therapies. The SAGAN trial is still active and recruiting additional patients with B cell lymphoma, ALL, and CLL to further compare the safety and efficacy of 2G versus 3G CAR-T cells. Similar results were reported in a Phase I/IIa clinical trial evaluating 3G CD19 CAR-T cells in patients with B cell leukemia or lymphoma. This study found that 3G CAR-T cells were well-tolerated; however, 4 patients required hospitalization due to adverse effects. Although 6 out of 16 treated patients experienced CRs, there were no significant differences in durability compared to 2G CAR-T cells (32, 37). In another study by Schubert et al., 8 patients (2 adults with r/r ALL, 2 r/r CLL, 2 DLBCL, 1 transformed FL, and 1 mantle cell lymphoma (MCL)) received CD19-targeted CAR-T cells at doses of either 1x10⁶ or 5×10⁶ transduced cells/m² as part of a dose-escalation protocol targeting up to 20x10⁶ cells/m². This study reported clinical responses even with low numbers of transferred cells, a safety profile characterized by no grade >2 of CRS or ICANS, and CAR-T cell detectability for over three months post-infusion. Additionally, CAR-T cells were observed to migrate to different regions, including the cerebrospinal fluid (CSF), in cases of central nervous system (CNS) involvement (38).

Beyond CD19, 3G CAR-T cells targeting other markers such as CD20, CD22, and HER2 have been studied (39). However, 3G CAR-T cells face numerous challenges, including high costs, manufacturing difficulties, limited efficacy in some patient populations, and tumor resistance, particularly in solid tumors with an immunosuppressive milieu. These constraints emphasize the need to rationalize the development of next-generation CAR-T cells and expand their therapeutic applications beyond hematological malignancies.

Fourth-generation (4G) CAR-T cells, also known as “T cells redirected for universal cytokine-mediated killing” (TRUCKs), mark a leap forward in CAR-T cell therapy. These CAR-T cells are designed to recognize specific tumor antigens and secrete transgenic cytokines upon activation, unlike previous generations that relied on the two-signal model for T cell activation, which includes TCR (signal 1) and co-stimulatory signals (signal 2) ( Figure 2B ) (40). Cytokine supplementation can actively remodel the tumor microenvironment (TME) and provide a third signal required for sustained T cell responses (40, 41). TRUCKs can be engineered to secrete one or more cytokines that improve CAR-T cell function or reprogram the TME (42). Efforts to enhance T cell activity and persistence have centered on integrating cytokines such as IL-2, IL-7, IL-12, IL-15, IL-18, IL-21, and IL-23, each offering unique advantages (40, 42). Proliferative cytokines, including IL-2 and IL-15, have been shown to enhance antitumor responses, with IL-15 playing a crucial role in T cell survival and homeostasis, yielding promising results in both preclinical and clinical settings (43). CAR-T cells engineered to co-express IL-15 have demonstrated enhanced expansion, persistence, and antitumor activity across different tumor models. In B cell malignancies, CD19-targeted CAR-T cells incorporating membrane-bound IL-15 achieved CR with incomplete blood count recovery and MRD negativity (CRi MRD^-) lasting for five months in patients with r/r B-ALL. This treatment resulted in the greatest in vivo CAR-T cell expansion and persistence observed in those patients, although relapse with CD19-negative disease eventually occurred (44). In solid tumors that express disialoganglioside GD2, such as glioblastoma, neuroblastoma, and lung cancer, T cells co-expressing GD2-specific CAR and IL-15 have demonstrated promising preclinical efficacy. In glioblastoma, constitutive IL-15 expression enhanced CAR-T cell proliferation and tumor clearance in an orthotopic xenograft model, supporting the translational feasibility of this approach (45). In neuroblastoma, GD2-specific CAR-T cells co-expressing IL-15 outperformed their IL-15-negative counterparts, exhibiting reduced exhaustion, improved in vivo persistence, complete tumor eradication, and protection upon tumor re-challenge in a metastatic xenograft model (46). Similarly, in lung cancer models, GD2-specific CAR-T cells expressing IL-15 and incorporating an inducible caspase-9 (iC9) safety switch demonstrated sustained persistence and robust antitumor activity, consistent with findings in other GD2-expressing tumor models (47). However, this enhanced efficacy is frequently associated with a higher incidence of CRS, as observed in Phase 1 clinical trials comparing GPC3-specific CAR-T cells with and without IL-15 expression (48). Utilizing conditional IL-2 expression in CAR-T cells could reduce the toxicity associated with systemic IL-2 administration while maintaining its effectiveness (49). Additionally, IL-7 and IL-21 promote T cell persistence and memory formation, whereas the co-expression of IL-21 and IL-15 has proven effective at controlling tumor growth in hepatocellular carcinoma models (50–52). Cytokine effects vary; for example, IL-23 enhances T cell survival and proliferation via autocrine signaling. In contrast, IL-7 and CCL19 increase T cell infiltration and trafficking into tumor sites, which is often impaired in the TME, as demonstrated in solid tumor and lymphoma models (53, 54). Other cytokines that have shown potential in overcoming tumor-associated challenges include IL-10, IL-12, IL-18, and IL-36. IL-10 has demonstrated its ability to restore the metabolism of exhausted T cells, increasing their proliferation and cytotoxic activity. IL-12 has demonstrated substantial antitumor activity by reprogramming the TME into an immunologically active state, which activates infiltrating lymphocytes, polarizes macrophages toward a pro-inflammatory (M1) state, and reduces regulatory T cell (Treg)-mediated suppression (55, 56). Despite encouraging preclinical results, clinical application of IL-12 has been hindered by severe systemic toxicity, necessitating innovative engineering approaches to maximize benefits while minimizing risks. In response, utilizing an inducible system such as Tet-regulated IL-12 expression to fine-tune IL-12 production has demonstrated the ability to stimulate tumor immunity in vivo without the typical side effects associated with IL-12, warranting further translation of this concept into clinical applications (57). Both preclinical and clinical investigations have revealed that IL-18 can reshape the TME by reducing immunosuppressive cells and boosting pro-inflammatory activity (58). Moreover, IL-36, a recently discovered member of the IL-1 superfamily, has demonstrated potent antitumor activity in mouse models. CAR-T cells expressing IL-36γ can stimulate endogenous immunity by promoting the activation and maturation of APCs, such as dendritic cells (DCs), through increased cytokine production (such as IL-6) and upregulation of co-stimulatory and MHC molecules on their surface (59). Consequently, antigen cross-presentation is enhanced, leading to bystander T cell activation that targets tumor antigens beyond the original CAR target. This process promotes antigen spreading and amplifies the overall antitumor immune response while complementing CAR-T cell activity, suggesting that IL-36 could be a potential candidate for targeting antigen-loss tumors (59, 60). To counteract tumor heterogeneity, Flt3L-secreting CAR-T cells have been designed to enhance endogenous DC activity and promote antigen epitope spreading, especially when combined with immune agonists such as poly(I:C) and anti-4-1BB antibodies (61). Together, these advancements emphasize the importance of integrating cytokine engineering and immune-modulating approaches to improve CAR-T cell performance, particularly against solid tumors. However, TRUCKs carry an increased risk of systemic toxicity due to constitutive cytokine release and inadequate control over cytokine secretion, which can lead to on-target, off-tumor toxicity (OTOT). It is worth noting that preclinical models may not fully replicate human toxicity, raising concerns about the practical applicability of TRUCK-based therapies. These limitations indicate the need to address safety concerns, enhance the therapeutic profile, and optimize the clinical efficacy of 4G CARs. Potential strategies include using inducible cytokine expression systems, integrating synthetic cytokine receptors, and implementing safety switches to modulate T cell activity while controlling treatment-related toxicities.

Building on the advancements of earlier CAR generations, fifth-generation (5G) CARs represent the most advanced generation of CAR-T cell therapy, engineered to overcome previous limitations while improving therapeutic potential and safety. These CARs are based on 2G CAR constructs but also contain a cytokine-inducible receptor that stimulates the JAK-STAT pathway. The truncated cytoplasmic IL-2 receptor β-chain is fused with a STAT3/STAT5 binding motif, allowing for antigen-dependent activation of the JAK-STAT pathway cascade ( Figure 2B ). This design stimulates three signaling pathways simultaneously: TCR activation (via the CD3ζ domain), co-stimulation (via the CD28 domain), and cytokine signaling (via the JAK-STAT3/5 pathway). This leads to sustained cytokine signaling without the need for exogenous cytokine support (62). By mimicking physiological T cell activation and natural cytokine signaling, 5G CARs can enhance the proliferation of less differentiated T cell populations and reduce exhaustion, a key limitation of earlier CAR-T cell therapies. Unlike 4G CAR-T cells, which enhance antitumor activity by secreting cytokines upon activation, 5G CARs incorporate cytokine signaling directly into their intracellular domains, allowing for continuous stimulation and improved resistance to exhaustion (34). Preclinical studies have shown promising results; for instance, Kagoya et al. reported that JAK-STAT-enhanced CAR-T cells surpassed earlier generation CAR-T cells in terms of tumor control, as well as expansion and persistence in CD19⁺ leukemia xenograft models (62). These cells also showed increased T cell proliferation and reduced terminal differentiation, contributing to sustained efficacy over time. In A375-CD19 melanoma, 5G CAR-T cells increased tumor infiltration and expansion, addressing a major hurdle in CAR-T cell therapy for solid tumors. The study also revealed that 5G CAR-T cells exhibited a less differentiated, stem cell-like memory phenotype and reduced apoptosis, resulting in enhanced persistence and antitumor activity (62). The increased release of effector cytokines such as IL-2, IFNγ, and TNFα improved their cytotoxic function and contributed to overcoming the immunosuppressive TME. Despite these promising findings, further research is needed to refine the safety profile, including monitoring and mitigating CRS, optimizing dosing regimens, and expanding therapeutic applications to a broader range of malignancies.

The CAR structural design can significantly influence the toxicity and efficacy of CAR-T cells. Important components that impact this design include the antigen-binding domain, co-stimulatory domains, transmembrane domains, hinge region, and the length of the endodomain (63).

The antigen-binding domain, typically a scFv, is the extracellular component of a CAR that defines its specificity for the target antigen expressed in cancer cells. A scFv is a fusion protein that usually consists of variable heavy (VH) and light (VL) chains linked by a flexible linker peptide (64). Key considerations in scFv design include its affinity for the target antigen, specificity, and origin (e.g., murine, humanized, or fully human). The specificity and affinity of the scFv for its target antigen play important roles in T cell activation (65). Although high-affinity scFvs can induce potent T cell responses and efficient killing of target cells, the activation threshold is inversely correlated with scFv affinity such that further increases in affinity do not necessarily enhance CAR-T cell activity compared to lower-affinity immunoreceptors (65, 66). Moreover, the scFv affinity level also affects the ability of CAR-T cells to distinguish between different antigen densities, with higher-affinity CARs tend to show reduced discrimination between target cells with high or low antigen expression. Studies suggest that an optimally tuned affinity enables CAR-T cells to preferentially target tumor cells with high antigen density, thus reducing the adverse effects associated with off-target activity (64, 67). In addition, murine-derived scFvs tend to induce immune responses in patients, including humoral responses (e.g., human anti-mouse antibodies; HAMA) and cellular anti-CAR immunity. These responses can lead to complications such as anaphylaxis and reduced CAR-T cell efficacy (68). The humanization of scFvs has been shown to significantly reduce immunogenicity while improving the persistence and safety of CAR-T cells compared to their murine-based counterparts. In clinical studies, humanized CAR-T cells have demonstrated superior therapeutic efficacy and durable antitumor effects in patients who previously relapsed following murine CAR-T cell treatment (69, 70). Therefore, scFv humanization is a crucial consideration in the design of CAR-T cell therapies.

The choice of co-stimulatory domains can significantly influence how CAR-T cells respond upon activation by target antigens. CD28-based CARs can trigger more robust and rapid activation, along with increased cytokine release. This rapid response may lead to faster tumor killing, which can be advantageous, especially in tumors with low antigen density. However, this is also associated with a higher risk of severe CRS and ICANS. In contrast, 4-1BB-based CARs provide slower and more sustained activation, which is important for long-term efficacy when sustained T cell persistence is required for complete or prolonged antitumor responses. They typically have a lower incidence of severe CRS and ICANS compared to CD28-based CARs. Therefore, optimizing the co-stimulatory domains based on tumor burden, antigen density, and the affinity of the antigen-binding domain is important. For example, 4-1BB may work better with high tumor burden or high antigen density, whereas CD28 can provide sufficient T cell activation when the CAR’s antigen-binding domain has low affinity or the surface antigen is expressed at low density (71). Several other co-stimulatory domains, including OX40, ICOS, CD27, CD40L, and TLR2, have been explored in preclinical and clinical settings for their potential to improve the safety and efficacy of CAR-T cells. However, only CD28 and 4-1BB remain the most commonly used and FDA-approved (72). The transmembrane domain anchors the CAR to the T cell membrane and plays a crucial role in transferring the activation signal to the intracellular domains following antigen recognition. CARs can be engineered with different transmembrane moieties, mainly derived from type I single-spanning proteins, such as CD3ζ, CD4, CD8α, or CD28. Each of these has been shown to influence the overall stability and functionality of the CAR. It has been demonstrated that CD19 CARs featuring CD8α-derived transmembrane domains tend to provide more controlled activation with lower toxicity and improved persistence compared to CD28-derived transmembrane domains while achieving comparable efficacy in eradicating tumors in preclinical studies (73).

Another component of the CAR structure is the hinge domain, also known as the spacer. This extracellular region connects the scFv to the transmembrane domain, providing flexibility and enabling the CAR to access antigen epitopes (74). The length, flexibility, and composition of the hinge domain can significantly influence CAR-T cell performance and safety (73, 75). Studies reveal that the optimal spacer length depends on the location of the epitope (76). Shorter spacers tend to be more effective for membrane-distal targets, whereas longer spacers offer increased flexibility, facilitating access to membrane-proximal or complex glycosylated epitopes (76, 77). Spacers are classified into two main categories: 1) IgG-based spacers, derived from the Fc region of IgG (commonly IgG1, IgG2, or IgG4), specifically the hinge plus two Ig-like domains including CH2 and CH3; 2) non-IgG-based spacers, derived from the extracellular stalks of natural T cell proteins or other receptor proteins, most commonly the CD8α or CD28 hinge regions (78) ( Figure 3 ). Although IgG-based spacers are widely used, they can bind FcγRs, leading to off-target activation of FcγR-expressing cells and reduced CAR-T cell persistence. To mitigate these effects, modifications such as mutating or deleting the FcγR-binding domain (e.g., CH2 deletion) have been shown to enhance CAR-T cell persistence and antitumor efficacy (79–81). Moreover, the CH2-CH3 spacer has also been replaced with the extracellular domain from the nerve growth factor receptor (NGFR), serving as a modular spacer. This NGFR spacer does not interfere with the CAR’s targeting or signaling domains, allowing for effective tracking, enrichment via anti-NGFR antibodies, and functional assessment of CAR-T cells (82). Conversely, non-IgG-based spacers avoid FcγR interactions and minimize off-target activation, as demonstrated in FDA-approved CAR-T cell products (e.g., Yescarta uses a CD28 hinge; Kymriah uses a CD8α hinge) (83). However, CD28-derived spacers can increase cytokine production, T cell exhaustion, and reactivity against low antigen density compared to CD8α-derived spacers (71). This highlights the importance of spacer design in mitigating adverse effects such as CRS and ICANS by adjusting flexibility and selecting the appropriate spacer composition.

CAR-T cells can be genetically modified to incorporate safety switches that enable dynamic regulation of the adoptively transferred CAR-T cells in vivo, aiming to enhance the safety and efficacy of cellular therapies. These safety switches have primarily been utilized to mitigate life-threatening toxicities associated with CAR-T cell therapy, such as CRS, ICANS, and OTOT effects. Different classes of safety switches have been developed to control the activity of therapeutic T cells.

The concept of logic gating is inspired by computational principles known as “Boolean logic,” which includes gates such as AND, OR, and NOT. These can be applied to CAR-T cell design and function as decision-making circuits within T cells, thereby enhancing the safety, efficacy, and precision of CAR-T cell therapies. CARs implemented with these Boolean logic gate-engineered systems enable T cells to process multiple tumor-specific antigen signals, allowing them to make more refined decisions, similar to how native T cells process multiple inputs before initiating an immune response (97). By improving tumor specificity and minimizing OTOT toxicity, logic-gated CAR-T cells offer a more precise and effective strategy for reducing the likelihood of tumor antigen escape and addressing the complex and heterogeneous tumor microenvironments characteristic of solid tumors.

Beyond genomic changes, emerging methods such as epigenome editing and RNA editing offer transient and reversible therapies. These strategies enable engineered CAR-T cells to dynamically adjust their fate and function while preserving the underlying genome.

Viral vectors, including LVs and AAV, provide high transduction efficiency and durable CAR expression, making them ideal for in vivo CAR-T cell engineering (146, 147). However, concerns remain about LV-associated insertional mutagenesis and AAV-associated pre-existing immunity, indicating the need for alternative delivery methods (147). Non-viral delivery platforms, such as LNPs and polymer-based nanoparticles, have emerged as safer and transient options that eliminate the risk of genomic integrations (144, 148).

Inspired by mRNA vaccine technology, CAR delivery via LNP-based mRNA—the same platform used in the first two COVID-19 vaccines granted Emergency Use Authorization—has demonstrated rapid CAR-T cell induction, enhanced T cell migration, and modulation of T cell phenotype. However, due to the transient nature of RNA-mediated gene expression, repeated dosing may be necessary to maintain efficacy (149). Preclinical success has driven early-phase clinical trials to assess the feasibility of in vivo CAR-T cell therapies in both blood and solid tumors. Nevertheless, several challenges remain, especially the need to achieve precise gene delivery to T cells while avoiding off-target transduction in Tregs or tumor cells, which could compromise antitumor immunity (147). Furthermore, the mRNA cargo delivered by LNPs can pose toxicity risks, including innate immune activation, protein overexpression, and mRNA degradation, which may trigger unintended immune responses and systemic inflammation. For example, as mRNA is inherently unstable and prone to degradation, the resulting RNA fragments can act as danger-associated molecular patterns (DAMPs), which potentially induce cytokine release and tissue-specific toxicities, such as myocarditis. To mitigate these risks, there is a growing demand for incorporating safety switches that can dynamically regulate CAR expression in real-time (150). To pave the way for broader clinical use of in vivo CAR-T cell therapies, several major challenges still need to be addressed. These include enhancing target specificity using ligand-directed targeting vectors while avoiding off-target effects in non-target cells. Strategies such as cell type-restricted gene expression and the use of synthetic biology tools to finely control immune responses are key to moving forward. Additionally, utilizing drug-inducible CAR systems and self-regulating immune circuits may also help improve T cell activity and persistence while reducing toxicity (146). With continuing advances in gene editing and synthetic immunology, in vivo CAR-T cell therapies hold the potential to expand access to cellular immunotherapies, making personalized cancer treatment more scalable and globally accessible.

CAR-NK cells build on the core framework of CAR-T cell technology but are engineered with features tailored to their innate immune functions. Unlike T cells, NK cells express unique surface receptors but lack antigen-specific TCRs. Instead, they rely on innate immune signaling mediated by a diverse array of activating and inhibitory receptors for their activation and function. For example, the activating receptor NKG2D recognizes stress-induced ligands such as MICA and MICB proteins, which are upregulated on virus-infected or cancer cells, triggering rapid NK cell cytotoxicity (151). The net balance of activating and inhibitory inputs ultimately determines the activation of NK cells. Thus, when inhibitory receptors (e.g., KIRs) are unable to detect self-MHC class I (MHC-I) on a target cell due to loss or reduced expression of MHC-I—a phenomenon called “missing-self” recognition—the unopposed activating signals license NK cells to activate and eliminate aberrant cells (152). To address this difference, CAR-NK constructs often incorporate NK-specific signaling domains such as NKG2D, DAP10, and DAP12, which play essential roles in activating NK cell cytotoxic functions (153). These specialized components help improve both the longevity and effectiveness of CAR-NK cell therapies by aligning with the unique activation mechanisms of NK cells. However, one primary challenge with NK cell therapies is their shorter lifespan compared to T cells, which can limit their effectiveness in the body (154). To overcome this issue, next-generation CAR-NK cells have been engineered to produce supportive cytokines such as IL-2 and IL-15 to enhance their ability to survive and expand post-infusion (155). Another advantage is that NK cells naturally lack the TCR, unlike allogeneic CAR-T cell products, which typically require gene editing to remove endogenous TCRs and prevent GVHD. This makes CAR-NK cells particularly appealing as ready-to-use, off-the-shelf therapies (156). Early clinical trials, including those for hematological cancers, have shown encouraging results with improved response rates and fewer severe side effects, including CRS and ICANS. Nevertheless, challenges such as lower transduction efficiency continue to drive ongoing investigations into improving gene delivery methods, including feeder cell-based expansion systems and advanced gene editing approaches to enhance their in vivo persistence (157).

Macrophages are naturally capable of infiltrating solid tumors, reshaping the TME, and maintaining a sustained antitumor response. Building upon these innate properties, CAR-Ms have emerged as an exciting new approach to overcoming key limitations of CAR-T cell therapies, particularly in the treatment of solid tumors. Although CD3ζ is not macrophage-specific and is expressed only in a minor subset of pro-inflammatory (CD3⁺) macrophages, studies have shown that CAR-Ms incorporating CD3ζ can mediate antigen-dependent phagocytosis and direct antitumor activity in macrophages (158). While structurally similar to CAR-T cell constructs, first-generation CAR-Ms utilize the CD3ζ signaling domain, canonical to T cells, to trigger a phagocytic response upon antigen engagement. Furthermore, engineering induced pluripotent stem cell-derived macrophages (iMACs) with dual-signaling CARs containing both CD3ζ and Toll/Interleukin-1 receptor (TIR) domains has been shown to significantly enhance antitumor immunity, promote antigen-dependent phagocytosis, and drive M1-like polarization (159). CAR-M therapies not only improve tumor clearance but also reprogram the TME toward a pro-inflammatory state by secreting mediators such as metalloproteinases (MMPs), reactive oxygen species (ROS), and serine proteases, which collectively diminish M2-type immunosuppression and facilitate immune cell infiltration (159, 160). Unlike CAR-T cells, which primarily mediate direct cytotoxicity, CAR-Ms not only induce antigen-specific phagocytosis but also concurrently cross-present tumor antigens and deliver co-stimulatory signals, thereby stimulating T cells and initiating adaptive immune responses (158, 161). Preclinical studies have demonstrated that CAR-Ms selectively phagocytose tumors, reduce tumor burden, and extend overall survival (158). Moreover, CAR-Ms can be engineered or combined with anti-CD47 antibodies to block the CD47-SIRPα (“do not eat me”) signaling pathway, thereby resisting immunosuppression in the TME. Similar to CAR-NK cells, CAR-Ms pose a minimal risk of GVHD due to their rapid extravasation from the bloodstream and limited in vivo expansion. However, because macrophages lack the proliferative capacity of T cells, enhancing CAR-M scalability and persistence will likely rely on strategies such as off-the-shelf cell therapies (e.g., CAR-iMacs) or combination therapies with T cell-based approaches (162, 163).