Next-generation CAR-T and CAR-NK cell therapies in hematologic malignancies: engineering for persistence, specificity, and safety

Abstract

Chimeric antigen receptor (CAR) T-cell therapy has profoundly reshaped the therapeutic landscape for hematologic malignancies, achieving remarkable response rates in relapsed/refractory B-cell leukemias and lymphomas. Despite this success, significant challenges persist regarding long-term CAR-T cell persistence, precise tumor specificity, and the management of treatment-related toxicities such as cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). Concurrently, CAR-Natural Killer (CAR-NK) cell therapy is emerging as a compelling alternative, offering inherent safety advantages, including a reduced risk of CRS and graft-versus-host disease (GvHD), alongside a promising “off-the-shelf” potential. This review examines the current state of CAR-T and CAR-NK cell therapies in hematologic malignancies, detailing advanced engineering strategies aimed at enhancing their persistence, improving tumor-specific targeting, and bolstering safety. It delves into innovations such as optimized CAR designs, cytokine and metabolic engineering, multi-antigen targeting, and the implementation of sophisticated safety switches. Future directions emphasize the integration of cutting-edge technologies, including advanced gene editing, non-viral delivery systems, and artificial intelligence (AI)-driven CAR design, alongside rational combination therapies, to achieve more durable, precise, and widely accessible treatments for hematologic cancers.

Introduction

The advent of chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment paradigm for hematologic malignancies, particularly B-cell lymphomas and leukemias [1]. This innovative cellular immunotherapy has demonstrated unprecedented clinical success, leading to multiple FDA approvals for anti-CD19 products that have profoundly transformed the management of B-cell malignancies [2]. Patients with relapsed or refractory B-cell acute lymphoblastic leukemia (ALL) and diffuse large B-cell lymphoma (DLBCL) have experienced significant clinical responses, including high overall response rates (ORR) and complete remission (CR) [3, 4]. This therapeutic breakthrough stems from the ability of CARs to redirect T-cell specificity in an MHC-independent manner, enabling direct recognition and elimination of tumor cells [5].

The initial successes of CAR-T cell therapy were achieved with first-generation CARs, which primarily utilized a CD3ζ signaling domain [6]. While these constructs demonstrated tumor-specific killing, their in vivo efficacy was often limited by suboptimal T-cell proliferation and persistence. This observation spurred the evolution of CAR design, leading to second and third-generation CARs that incorporate additional costimulatory domains, such as CD28, 4-1BB (CD137), or OX40 [7]. These enhancements significantly improved CAR-T cell expansion, persistence, and anti-tumor activity. However, this increased potency has also been accompanied by severe, potentially life-threatening toxicities, most notably cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) [7, 8]. Achieving deep and durable clinical responses necessitates a nuanced approach that controls T-cell activation to mitigate systemic inflammation without compromising anti-tumor function.

In parallel with the advancements in CAR-T cell therapy, CAR-Natural Killer (CAR-NK) cell therapy has emerged as a compelling alternative, offering several distinct advantages [9]. Unlike autologous CAR-T cell therapy, which is often hindered by lengthy and costly manufacturing processes and the potential for insufficient or dysfunctional patient-derived T cells due to prior treatments, CAR-NK cells inherently possess “off-the-shelf” potential [9]. Furthermore, CAR-NK cells exhibit a favorable safety profile, characterized by a significantly reduced risk of severe CRS, ICANS, and GvHD compared to CAR-T cells [10]. This inherent safety and logistical advantage positions CAR-NK cells as a transformative solution for democratizing access to cellular immunotherapies, particularly for patients with aggressive hematologic malignancies where rapid intervention is critical [11].

Despite the remarkable progress, both CAR-T and CAR-NK cell therapies continue to face significant barriers that limit their broader clinical translation and the achievement of durable remissions [12]. These challenges include the pervasive issue of antigen escape, where tumor cells lose or downregulate the targeted antigen, leading to relapse [12]. The efficacy, persistence, and safety of CAR-T and CAR-NK cell therapies are deeply intertwined. Furthermore, antigen escape directly compromises long-term efficacy, underscoring the need for strategies that anticipate and counter tumor adaptive mechanisms [13]. Therefore, the continued advancement of CAR-T and CAR-NK cell therapies in hematologic malignancies necessitates a concerted focus on sophisticated engineering strategies that holistically address these interconnected challenges, aiming for enhanced persistence, improved specificity, and superior safety to achieve lasting therapeutic benefits.

Enhancing persistence of CAR-T and CAR-NK cells

The long-term persistence of adoptively transferred CAR-T and CAR-NK cells is a critical determinant of durable clinical responses in hematologic malignancies. Achieving sustained anti-tumor activity requires overcoming intrinsic limitations in cell survival and proliferation, as well as extrinsic suppressive signals within the tumor microenvironment (TME).

Optimizing CAR design for longevity

The structural design of the CAR, particularly the choice of its intracellular costimulatory domains, profoundly influences the persistence and anti-tumor efficacy of CAR-T cells [14]. While second-generation CARs incorporating CD28 or 4-1BB costimulatory domains established the current clinical standard, their functional profiles differ significantly; CD28 promotes rapid effector function but limited persistence, whereas 4-1BB enhances metabolic fitness and memory formation [15–17]. To transcend these binary limitations, next-generation engineering is moving toward third-generation CARs (combining multiple costimulatory domains like CD28-4-1BB-CD3ζ) and novel costimulatory domains such as ICOS, CD27, or OX40 to fine-tune T-cell physiology [18]. Furthermore, ‘armored’ CARs constitutively secreting cytokines (e.g., IL-18, IL-12) or expressing dominant-negative receptors are being deployed to actively remodel the local immune landscape rather than passively relying on intrinsic signaling alone [9, 19–22].

Cytokine engineering and metabolic fitness

The functional longevity and anti-tumor efficacy of CAR-T and CAR-NK cells are profoundly influenced by the cytokine milieu and their metabolic fitness within the hostile tumor microenvironment (TME). CAR-T/NK cell activity is highly dependent on robust cytokine signaling for optimal expansion and function [23]. To address this, genetic engineering strategies are being employed to augment cytokine signaling in CAR cells [24]. This includes the transgenic expression of various cytokines, such as interleukin-2 (IL-2), IL-7, IL-12, IL-15, IL-18, IL-21, and IL-23, or the engineering of cytokine receptors to enhance signal 3 of T-cell activation pathways [25, 26]. For instance, autocrine expression of IL-15 in CAR-NK cells has demonstrated the capacity to sustain their anti-AML functionality [27]. However, uncontrolled or constitutive cytokine expression can lead to systemic toxicities, including severe cytokine release syndrome.

The TME presents a metabolically challenging environment, characterized by hypoxia, acidosis, and nutrient deprivation, which significantly compromises the function and persistence of CAR-T and CAR-NK cells [23].

Promoting memory phenotypes

The differentiation state of T cells used for CAR manufacturing is a significant factor influencing their proliferative capacity, survival, and long-term persistence [28]. Less differentiated T-cell subsets, particularly stem cell memory (T_scm) and central memory (T_cm) T cells, have demonstrated superior persistence and anti-tumor immunity compared to more differentiated effector memory (T_em) and effector T (T_eff) cells [29]. These memory subsets possess enhanced self-renewal capacity and a greater ability to engraft and persist in vivo, which directly correlates with improved clinical outcomes in patients[30].

Strategies to promote the formation and enrichment of these desirable memory phenotypes are actively being investigated [31]. These approaches include optimizing ex vivo cell culture conditions, manipulating specific signaling pathways involved in T-cell differentiation, and enforcing epigenetic changes that favor the memory cell fate [29] (Table 1).

Table 1.

Key engineering strategies for CAR cell persistence

Engineering approach	Specific modification	Mechanism of action	Functional outcome	Refs.
Costimulation (T Cells)	4-1BB (CD137) domain	Promotes mitochondrial biogenesis and oxidative metabolism.	Enhances memory formation; prevents early exhaustion.	[32]
Costimulation (NK Cells)	CD28 domain	Activates PI3K/AKT signaling pathway.	Increases cytotoxicity and rapid effector function.	[22]
Structural Optimization	Tuned spacers/hinges	Modulates distance between CAR and target cell.	Reduces tonic signaling; optimizes synapse formation.	[17]
Cytokine Armoring	IL-15/IL-21 secretion	Provides autocrine “Signal 3” stimulation.	Sustains survival in low-cytokine TME; boosts expansion.	[26]
Metabolic Engineering	PGC-1α overexpression	Regulates mitochondrial respiration genes.	Maintains fitness in hypoxic/hypoglycemic TME.	[33]
Memory Enrichment	T < sub> SCM/T < sub> CM culture	Epigenetic modulation (e.g., AKT inhibition) during manufacturing.	Increases self-renewal capacity; prolongs in vivo persistence.	[29]

Strategies for enhanced specificity and reduced off-target toxicity

A significant challenge in CAR-T and CAR-NK cell therapies is ensuring precise targeting of malignant cells while sparing healthy tissues, thereby minimizing on-target/off-tumor toxicities. This is particularly critical in hematologic malignancies where target antigens may be expressed on normal hematopoietic cells.

Multi-antigen targeting approaches

One of the most persistent challenges in single-antigen CAR-T therapies is antigen escape, a phenomenon where tumor cells lose or downregulate the targeted antigen, leading to disease relapse [34]. This is a direct consequence of the inherent plasticity and evolutionary capacity of cancer cells, which can adapt to strong selective pressures imposed by single-target therapies. Tumors are heterogeneous, comprising diverse cell populations that can undergo phenotypic shifts or genetic alterations to evade immune recognition [35]. To counteract this evolutionary arms race, multi-target strategies have been developed to simultaneously engage several antigens, thereby reducing the probability of antigen escape and promoting more durable responses. These approaches include:

• Pooled CARs This involves co-administration or co-transduction of T cells with different CAR vectors, each targeting a distinct antigen. This method provides broad coverage against heterogeneous tumor populations [36].

• Bicistronic/Dual CARs Here, a single vector encodes two independent CAR molecules separated by a ribosomal skip sequence, allowing for the expression of two distinct CARs within the same cell [37].

• Tandem CARs (TanCARs) This design involves a single CAR construct that contains two scFvs, enabling simultaneous targeting of multiple antigens within one chimeric protein [38].

An example of the effectiveness of multi-antigen targeting is the dual targeting of CD70 and CD33 in acute myeloid leukemia (AML) [39]. This approach has demonstrated superior cytotoxicity against AML cells, even when CD70 expression was downregulated, by effectively recruiting untransduced bystander T cells through a bispecific T-cell engager molecule (TEAM) [39, 40]. The future of CAR therapy is inherently multi-modal, both in terms of targets and potentially in combination with other therapies.

Affinity tuning of antigen-binding domains (scFv optimization)

The affinity of the single-chain variable fragment (scFv), which serves as the antigen-binding domain of the CAR, is a critical parameter that influences both efficacy and safety [41]. Conventionally, it was assumed that higher affinity scFVs would lead to better CAR performance. However, high-affinity scFVs can paradoxically lead to on-target/off-tumor toxicities by binding to healthy cells that express the target antigen at low levels [38]. Moreover, persistently high-affinity binding can contribute to CAR-T cell exhaustion due to chronic stimulation, thereby limiting long-term persistence [42].

Conversely, an exceedingly low-affinity binder may not sufficiently engage the target antigen to activate the CAR-T cells, leading to suboptimal anti-tumor responses [43]. Recent studies suggest that moderate-affinity scFVs may offer a more favorable balance, leading to better efficacy and reduced off-target toxicity, particularly in solid tumors [44]. For AML, specifically, low-affinity CD123 CAR-NK cells have demonstrated superior antigen discrimination, prolonged persistence, and enhanced tumor control compared to their high-affinity counterparts [45]. This indicates that a fine-tuned affinity allows for better discrimination between tumor cells (which typically express high antigen density) and healthy cells (which may express lower antigen density), thereby expanding the therapeutic window [46]. This emphasis on rational, data-driven CAR design, moving beyond empirical approaches, underscores the importance of computational modeling and single-cell analysis for predicting optimal affinity ranges for specific antigen-expression profiles in both malignant and healthy tissues, ultimately leading to truly personalized CAR therapies (Fig. 1; Table 1).

Fig. 1.

This graphic details engineering strategies to optimize CAR T-cell therapy by addressing antigen escape and binding affinity. To ensure broad tumor coverage and prevent resistance, it highlights multi-antigen approaches including pooled CARs (separate cells), bicistronic CARs (two distinct receptors), and tandem CARs (two domains on one receptor). Furthermore, the right panel demonstrates that affinity tuning of the antigen-binding domain is critical for clinical success; while high affinity often leads to T-cell exhaustion and on-target/off-tumor toxicity, moderate affinity provides an optimal balance that enhances efficacy, reduces toxicity, and promotes prolonged T-cell persistence

Logic-gated CAR designs

To further enhance specificity and minimize off-target toxicities, advanced CAR designs incorporating “logic gates” are being developed. These innovative CARs, including “AND,” “OR,” “NOT,” and “IF-THEN” logic-gated systems, require combinatorial antigen expression to precisely trigger or inhibit CAR function.

• “AND” logic-gated CARs (e.g., bispecific CARs, dual CARs, synNotch CARs) are designed to activate only when two or more specific antigens are simultaneously present on the target cell. This approach significantly reduces on-target/off-tumor toxicities by ensuring that CAR activity is confined to cells uniquely defined by a combination of markers, which are typically found only on tumor cells [47].

• “NOT” logic-gated CARs send an inhibitory signal in the presence of a specific antigen that is expressed on normal tissues but absent from tumor cells. This mechanism actively prevents CAR function in healthy tissues, thereby safeguarding against unintended damage [47].

• “OR” logic-gated CARs (e.g., tandem CARs) are designed to activate if any one of multiple specified antigens is present. While less restrictive than “AND” gates, they are valuable for addressing tumor heterogeneity and antigen escape by ensuring that the CAR cell remains active as long as at least one target antigen is expressed [47].

These logic-gated designs represent a significant leap forward in CAR engineering, enabling a level of precision that was previously unattainable. They allow for the creation of CAR cells that can make more complex “decisions” about their targets, moving beyond simple antigen recognition to a more sophisticated, context-dependent activation profile.

Targeting cancer stem cell-specific antigens in hematologic malignancies

Cancer stem cells (CSCs) represent a small subpopulation of tumor cells characterized by self-renewal capacity, differentiation potential, and high tumorigenicity [48]. They are considered the “roots” of aggressive tumors, responsible for tumor initiation, progression, metastasis, and, critically, resistance to conventional therapies and subsequent disease recurrence [48]. Therefore, specifically targeting CSCs holds immense promise for achieving lasting remissions.

CAR-T and CAR-NK cells can be engineered to target CSC-specific surface markers, which are often distinct from those on differentiated tumor cells. Examples of such targets in hematologic malignancies and other cancers include:

• CD133 A glycoprotein expressed on primitive hematopoietic stem cells and enriched in specific leukemia subtypes (e.g., MLL-rearranged leukemias), though its expression on normal stem cells limits its utility as a sole target [49].

• CD44 A transmembrane protein involved in cell adhesion and migration; it is overexpressed in AML CSCs and plays a critical role in their homing and retention within the bone marrow protective niche [50].

• CD123 (IL-3Rα) Highly expressed on CD34 + CD38- AML leukemic stem cells (LSCs) while exhibiting low expression on normal hematopoietic stem cells (HSCs), offering a superior therapeutic window for LSC eradication compared to other myeloid markers [51, 52].

• CD33 A pan-myeloid marker widely expressed on bulk AML blasts. While effective for debulking, it is also expressed on normal myeloid progenitors, necessitating careful safety switches or bridging to transplant strategies [53].

• CLL-1 (CLEC12A) Another promising target enriched on AML LSCs and blast cells but absent on normal HSCs, allowing for potent targeting with reduced risk of prolonged myelosuppression [54].

Targeting LSCs requires overcoming biological barriers beyond simple antigen recognition. LSCs often reside in the bone marrow vascular niche, which provides a sanctuary from systemic therapies and immune surveillance. Furthermore, LSCs exhibit cellular quiescence (G0 phase), rendering them resistant to therapies that target dividing cells, and they possess intrinsic immune evasion mechanisms, such as the downregulation of MHC molecules (Ref C). Therefore, next-generation CAR strategies must not only target surface antigens but also penetrate the marrow niche and break LSC dormancy [55] (Fig. 2).

Fig. 2.

This diagram outlines advanced CAR T-cell strategies designed to enhance precision and durability by utilizing logic-gated designs and targeting cancer stem cells (CSCs)

Logic-gated CAR designs

To improve tumor specificity and safety, three primary “logic gates” are employed:

• “AND” Gate Requires the presence of both Antigen A and Antigen B for T-cell activation, ensuring specific tumor killing while sparing healthy cells that express only one antigen.

• “NOT” Gate Incorporates an inhibitory CAR (iCAR) that recognizes a healthy cell antigen (Antigen B); if encountered, it overrides the activation signal from Antigen A, protecting normal tissue.

• “OR” Gate Utilizes tandem CARs (e.g., scFv-A and scFv-B) to trigger activation if either antigen is present, effectively addressing tumor heterogeneity and preventing antigen escape.

Targeting cancer stem cells (CSCs)

The right panel emphasizes that while conventional therapies often kill bulk tumor cells, the “roots” of the disease—quiescent and resistant CSCs—often survive, leading to relapse (Table 2).

Table 2.

Multi-antigen targeting strategies for CAR cells

Strategy	Design architecture	Biological advantage	Refs.
Pooled products	Co-administration of two distinct cell lines (e.g., anti-CD19 + anti-CD22).	Mitigates antigen escape via broad coverage; distinct kinetics per population.	[36]
Bicistronic/Dual CAR	Single vector encoding two CARs separated by 2 A peptide.	Simultaneous targeting; ensures all transduced cells attack both antigens.	[56]
Tandem CAR (TanCAR)	Single CAR construct with two scFv domains (loop structure).	Synergistic activation (OR gate logic); highly efficient dual-synapse formation.	[57]
Modular Universal	Split CARs (e.g., SUPRA, biotin-binding) with soluble adapters.	Target flexibility; antigen specificity can be switched post-infusion.	[58]

• CSC-specific antigens Strategic targets include CD133, CD44, CD123, CD33, and CLL-1 to selectively eliminate the cells responsible for tumor initiation and metastasis.

• Next-gen strategies To overcome biological barriers, modern CAR T-cells are engineered to penetrate the bone marrow vascular niche, break CSC dormancy, and target surface antigens even in environments characterized by low MHC expression and immune evasion.

Ensuring safety: mitigating on-target/off-tumor effects and systemic Toxicities

The remarkable efficacy of CAR-T cell therapy in hematologic malignancies has been accompanied by unique and potentially severe toxicities, necessitating robust strategies to ensure patient safety.

Management of Cytokine Release Syndrome (CRS) and Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS)

Cytokine Release Syndrome (CRS) and Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS) are the most common and potentially life-threatening adverse events associated with CAR-T cell therapy [59]. CRS is a systemic inflammatory response characterized by a surge of inflammatory cytokines, including interleukin-6 (IL-6), IL-8, IL-10, interferon-gamma (IFN-γ), and granulocyte-macrophage colony-stimulating factor (GM-CSF), released by activated CAR-T cells and endogenous immune cells such as macrophages and dendritic cells [47, 59, 60].

ICANS, a spectrum of neurotoxicities, can manifest independently of CRS but is often observed concurrently [61].

Management strategies for moderate to severe CRS primarily involve the administration of tocilizumab, an anti-IL-6 receptor monoclonal antibody that can rapidly reverse symptoms [62, 63]. Corticosteroids are typically reserved for refractory CRS or for the management of ICANS, although their excessive or premature use may potentially impact CAR-T cell efficacy [62].

Implementation of inducible safety switches

To enhance the safety profile of CAR-T and CAR-NK cell therapies, the incorporation of inducible safety switches, also known as “suicide genes,” has become a critical engineering strategy [64]. These mechanisms allow for the rapid and selective elimination of infused CAR cells in the event of severe toxicities, without causing widespread damage to endogenous cells or compromising the broader immune system.

Key safety switch systems include:

• Inducible caspase-9 (iCasp9) This system involves a modified human caspase-9 fused to a human FK506 binding protein. Upon administration of a small molecule dimerizer (e.g., AP1903/Rimiducid), the iCasp9 construct dimerizes and activates, leading to rapid apoptosis of the transduced CAR cells [65]. A significant advantage of iCasp9 is its dose-dependent control, allowing for titratable elimination of CAR cells, which can help manage toxicities while potentially preserving some anti-tumor activity [66]. This represents a shift from a binary “kill switch” to a more nuanced “rheostat” of CAR cell activity, aiming to optimize the therapeutic window.

• Truncated human epidermal growth factor receptor (EGFRt/EGFRopt) This functionally inert cell surface marker is not expressed on hematopoietic cells and allows for in vivo tracking and selective ablation of CAR cells using an anti-EGFR monoclonal antibody (e.g., cetuximab) [67]. Optimized variants (EGFRopt) enhance surface expression and susceptibility to antibody-dependent cellular cytotoxicity (ADCC), leading to more rapid and robust CAR cell elimination [67, 68].

• Herpes simplex virus thymidine kinase (HSV-TK) This classic suicide gene converts a non-toxic prodrug, ganciclovir (GCV), into a cytotoxic compound that leads to cell death. While effective, HSV-TK can be immunogenic due to its viral origin and may have bystander effects on neighboring cells [66].

The evolution of safety switches reflects a growing maturity in the field, moving beyond initial concerns to optimize the therapeutic window. Future safety mechanisms will likely integrate feedback loops and precise dosage control to fine-tune CAR cell activity in vivo, potentially through drug-activated systems that can be titrated, thereby maximizing therapeutic benefit while minimizing systemic adverse effects (Fig. 3).

Fig. 3.

This graphic outlines strategies for managing CAR T-cell toxicities and implementing molecular safety switches. It describes the mechanisms of Cytokine Release Syndrome (CRS) and ICANS (neurotoxicity), which are driven by systemic cytokine release (e.g., IL-6, IFN-$\gamma$) and managed clinically with Tocilizumab and corticosteroids. To provide an ultimate safety layer, the diagram highlights inducible “kill switches” such as iCasp9, EGFRt, and HSV-TK, which allow for the rapid elimination of CAR T-cells via dimerizers, antibodies, or prodrugs. These interventions, alongside emerging titratable feedback loops, are critical for mitigating life-threatening side effects and improving the overall safety profile of cellular therapies

Strategies for allogeneic CAR-T and CAR-NK cell safety

Allogeneic CAR-T and CAR-NK cell therapies, derived from healthy donors, offer significant advantages, including immediate “off-the-shelf” availability, standardized manufacturing, and scalability, which can address the logistical and cost limitations of autologous approaches [69]. However, these therapies introduce new immunological challenges: the risk of graft-versus-host disease (GvHD) and rapid host immune rejection. GvHD occurs when allogeneic donor T cells recognize host healthy tissues as foreign, leading to severe systemic inflammation [70]. Host immune rejection, conversely, involves the recipient’s immune system recognizing and eliminating the infused allogeneic CAR cells.

To overcome these hurdles, sophisticated genetic engineering strategies are being employed:

• TCR alpha constant (TRAC) knockout CRISPR/Cas9 gene editing is used to disrupt the endogenous T-cell receptor (TCR) alpha constant (TRAC) locus in donor T cells. This prevents TCR-mediated recognition of host alloantigens, thereby abrogating GvHD while preserving CAR-mediated anti-tumor activity [71].

• HLA knockout/camouflage To prevent host T-cell mediated rejection, major histocompatibility complex (MHC) class I molecules (HLA-A, -B, -C) can be knocked out, often by disrupting the β2-microglobulin (B2M) gene, which is essential for HLA class I surface expression. However, complete loss of HLA class I can trigger “missing-self” responses from host NK cells, leading to their rapid elimination. To balance this, more refined strategies include:

  • Selective HLA-C retention Deleting HLA-A and -B alleles while retaining HLA-C expression, as HLA-C can engage inhibitory KIRs (Killer-cell Immunoglobulin-like Receptors) on host NK cells, thereby preserving inhibitory signaling and avoiding NK-mediated clearance [72].
  • Expression of immunomodulatory non-classical HLA molecules Engineering CAR cells to express non-classical HLA molecules like HLA-E or HLA-G, which engage inhibitory receptors (e.g., NKG2A, KIR2DL4) on host NK and T cells, can provide a “don’t eat me” signal and prevent immune recognition [72].

• iPSC-derived cells Induced pluripotent stem cells (iPSCs) represent a transformative solution for generating allogeneic CAR-T/NK cells[73]. iPSCs offer an unlimited, standardized, and genetically modifiable source, enabling multiplex gene editing to incorporate both GvHD prevention and enhanced persistence/functionality features from a single, well-characterized master cell bank [73]. This approach allows for the creation of highly engineered, immune-evasive, and potent “off-the-shelf” products on an industrial scale, significantly improving accessibility and affordability of these advanced therapies. In addition to iPSC-derived products, hematopoietic stem and progenitor cell (HSPC)-derived CAR-engineered immune cells are emerging as a promising next-generation modality. HSPCs can be genetically modified to express CARs and then differentiated into CAR-T, CAR-NK, or CAR-NKT cells, offering strong antitumor activity, long-term persistence through engraftment in the bone marrow, and suitability for allogeneic off-the-shelf therapy. These cells benefit from HSPCs’ ability to self-renew and differentiate, potentially providing a continuous supply of effector cells in vivo, while avoiding GvHD through TCR knockout. Preclinical studies have shown enhanced efficacy against glioblastoma and other malignancies, with improved safety profiles compared to mature cell-derived CARs [74].

CAR-NK cells inherently offer a lower risk of GvHD compared to CAR-T cells due to their distinct recognition mechanisms and lack of TCR, making them particularly attractive for allogeneic applications [75, 76]. The complex immune balancing act involved in developing allogeneic CAR cells underscores the intricate immunological hurdles that must be overcome to achieve truly universal and safe “off-the-shelf” immunotherapies [14]. Success hinges on a deep understanding of both adaptive and innate immune recognition, requiring multi-gene editing strategies to create products that are both immune-evasive and therapeutically potent (Table 3).

Table 3.

Safety mechanisms and their clinical application in CAR cell therapies

Safety mechanism	Trigger/mechanism	Key advantages	Limitations/challenges	Refs.
iCasp9	Small molecule dimerizer (e.g., rimiducid) triggers apoptosis.	Rapid onset (< 30 min); non-immunogenic (human origin).	Reliance on drug supply; cost of dimerizer agent.	[77]
EGFRt/EGFRopt	Anti-EGFR Ab (cetuximab) induces ADCC-mediated lysis.	Dual utility: allows cell sorting/tracking and ablation.	Slower kinetics than iCasp9; requires functional host immune system.	[78]
HSV-TK	Ganciclovir phosphorylation disrupts DNA synthesis.	Well-established clinical history.	Immunogenic (viral protein); slow mechanism of action; restricts anti-viral drug use.	[79]
TRAC Knockout	CRISPR/TALEN disruption of endogenous TCRα chain.	Prevents GvHD in allogeneic products.	Risk of translocations; requires high editing efficiency.	[80]
HLA Knockout	Disruption of B2M (MHC Class I).	Prevents host T-cell rejection (stealth).	Triggers host NK cell “missing-self” cytotoxicity.	[81]

Overcoming tumor microenvironment (TME) barriers in hematologic malignancies

The tumor microenvironment (TME) presents a formidable barrier to the efficacy of CAR-T and CAR-NK cell therapies in hematologic malignancies, as it is often profoundly immunosuppressive, metabolically hostile, and physically restrictive. This complex ecosystem actively supports tumor growth, metastasis, and resistance to therapy.

Addressing immunosuppression and metabolic stress

The hematologic TME actively suppresses CAR function through cellular effectors like Tumor-Associated Macrophages (TAMs) and Myeloid-Derived Suppressor Cells (MDSCs), alongside metabolic barriers such as hypoxia and nutrient deprivation. To counteract cellular suppression, strategies include inhibiting CSF-1R signaling to deplete TAMs or targeting YTHDF2 to overcome MDSC-mediated inhibition [82, 83]. Simultaneously, to address metabolic exhaustion, next-generation CARs are being engineered for enhanced metabolic fitness—for example, by overexpressing PGC-1α to support mitochondrial biogenesis—or are combined with HIF1α inhibitors (e.g., echinomycin) to eradicate hypoxia-resistant CSCs [23, 84].

• Targeting TAMs Tumor-associated macrophages (TAMs) often adopt a pro-tumorigenic M2 phenotype that promotes tumor growth, angiogenesis, and immunosuppression [83]. Therapeutic approaches include inhibiting colony-stimulating factor-1 receptor (CSF-1R) signaling (e.g., with small molecule tyrosine kinase antagonists like PLX3397 or antibodies like RG7155) to deplete TAMs or reprogram them towards an anti-tumor M1 phenotype [83].

• Targeting MDSCs Myeloid-derived suppressor cells (MDSCs) actively suppress T-cell function and promote cancer stemness [85]. Pharmacological inhibition of YTHDF2, an m6A “reader” protein, has been shown to overcome MDSC-induced immunosuppression and improve combined radiotherapy and/or anti-PD-L1 treatment efficacy [82].

• Addressing hypoxia and metabolic stress Hypoxia, a hallmark of the TME, promotes cancer stem cell (CSC) resistance, epithelial-mesenchymal transition (EMT), and metabolic shifts that favor tumor survival [86]. Targeting hypoxia-inducible factor 1 alpha (HIF1α) can eliminate CSCs in hematologic malignancies by abrogating their colony-forming unit activity and repressing a negative feedback loop in the Notch pathway [84]. Furthermore, the high metabolic demands of tumor cells lead to nutrient competition (e.g., for glucose, glutamine, amino acids) and metabolic stress, which compromise CAR-T/NK cell function. Metabolic interventions, such as enhancing the metabolic fitness of CAR cells or co-targeting tumor metabolism, are crucial to create a more favorable environment for immune cell activity [23] (Table 4).

Table 4.

TME barriers and corresponding CAR cell engineering solutions

TME Barrier	Biological Impact	Engineering solution	Refs.
Suppressive Cells	TAMs/MDSCs inhibit T-cell activation.	CSF-1R inhibitors; CAR-macrophages.	[99]
Hypoxia	Impairs mitochondrial function; induces exhaustion.	HIF1α stabilization; metabolic armoring.	[99]
Physical Barrier	Dense ECM prevents tumor infiltration.	CARs targeting FAP (fibroblasts); Heparanase secretion.	[100]
Soluble Factors	TGF-β/IL-10 dampen immune response.	Dominant-negative TGF-β receptors; Cytokine switch receptors.	[91]

Modulating the physical and soluble components of the TME

Physical barriers within lymphoid tissues, such as the extracellular matrix (ECM), impede CAR cell infiltration. Therapeutic solutions involve ECM remodeling using antifibrotic agents (e.g., Halofuginone) or engineering CARs to secrete ECM-degrading enzymes like heparanase [87]. Furthermore, reprogramming Cancer-Associated Fibroblasts (CAFs) via FAP-targeted CARs or TGF-β inhibition can reduce desmoplasia and improve T-cell penetration [88]. Finally, the immunosuppressive cytokine milieu is being addressed through ‘armored’ CARs that secrete pro-inflammatory cytokines (IL-12, IL-18) or express dominant-negative receptors to shield against TGF-β signaling [89].

• Reducing fibrosis Approaches like using Halofuginone to inhibit type I collagen synthesis, or PEGylated hyaluronidase and 4-methylumbelliferone to inhibit hyaluronic acid secretion, can reduce desmoplasia and interstitial pressure, improving drug uptake and immune cell infiltration [87].

• Nanocarriers Nano-drug delivery systems (Nano-DDS) offer a promising method for improved drug penetration and targeted delivery within the TME, overcoming physical barriers and enhancing accumulation in tumor tissues [90].

• Reprogramming cancer-associated fibroblasts (CAFs) Strategies to reprogram or deplete CAFs (e.g., FAP-targeted CAR-T cells, TGF-β inhibition, chemokine blockade) can reduce ECM remodeling and alleviate immunosuppression [88].

The TME is also rich in soluble factors, including cytokines, chemokines, and exosomes, which promote tumor growth, invasion, metastasis, and immunosuppression [91]. Modulating these factors (e.g., anti-TGF-β antibodies, IL-8 neutralization, targeting specific chemokines like CXCR4) can disrupt the TME’s supportive role for tumor cells [89].

Strategies to target both tumor cells and the TME

Next-generation CAR platforms are increasingly designed to integrate multi-targeting, microenvironment modulation, and combinatorial approaches to enhance translational relevance. Multi-targeting involves engineering CARs to recognize both tumor-specific antigens (e.g., CD19/CD22) and TME components (e.g., fibroblast activation protein (FAP) on stromal cells or PD-L1 on immunosuppressive cells), enabling simultaneous elimination of malignant cells and disruption of supportive niches. Microenvironment modulation can be achieved through armored CARs that secrete immunomodulatory cytokines (e.g., IL-12 to reprogram TAMs from M2 to M1 phenotype) or express chemokine receptors (e.g., CXCR4) to improve trafficking into bone marrow niches. Combinatorial designs pair CAR cells with small-molecule inhibitors (e.g., CSF-1R antagonists to deplete MDSCs) or checkpoint blockers (e.g., anti-PD-1), synergistically remodeling the TME to favor anti-tumor immunity. These strategies have shown promise in preclinical models, improving persistence and efficacy while reducing relapse rates [92].

Exploiting cancer stem cell plasticity and niche interactions

Cancer stem cells (CSCs) are maintained by their specific microenvironment, known as the CSC niche, which provides crucial signals for their self-renewal, differentiation, and protection from therapies [55]. A critical aspect of tumor progression and therapy resistance is the remarkable plasticity of CSCs, including the ability of non-CSCs to dedifferentiate into CSC-like cells and for CSCs to switch phenotypes [55]. This dynamic nature implies that simply eliminating existing CSCs may be insufficient if new ones can arise or if existing ones alter their surface markers [93]. The TME acts as a “sanctuary” and “educator” for CSCs, actively inducing these plastic changes through factors like hypoxia, inflammatory cytokines, and stromal cells. This highlights that effective CSC targeting requires not only eliminating existing CSCs but also disrupting the TME’s capacity to generate new ones or protect them, necessitating combination therapies that simultaneously target cancer cells and modulate the TME [94].

Epigenetic reprogramming emerges as a central mechanism driving CSC plasticity, stemness, and immune evasion [95]. This suggests that epigenetic mechanisms are not merely consequences but fundamental drivers of malignant traits, offering reversible targets for therapy. For instance, the epigenetic regulator KDM1B promotes IFN-I-induced stemness and immune escape 140, while FTO inhibitors suppress CSC self-renewal and immune evasion [96]. Therefore, epigenetic therapies, either alone or in combination with CAR cells, hold significant promise for overcoming resistance by reversing aberrant gene expression patterns that drive stemness and immune evasion [97].

The bidirectional crosstalk between TME components (immune cells, stromal cells, ECM, soluble factors) and CSCs creates a self-reinforcing tumor-promoting ecosystem [98].

Future directions and emerging technologies

The rapid evolution of CAR-T and CAR-NK cell therapies continues to push the boundaries of cancer treatment, with significant advancements anticipated from the integration of cutting-edge technologies and innovative therapeutic strategies.

Advanced gene editing and non-viral delivery systems

The precision and efficiency of gene editing tools, particularly CRISPR/Cas9 technology, are revolutionizing the development of next-generation CAR-T and CAR-NK cells. CRISPR/Cas9 enables multiplex genomic engineering [101], allowing for simultaneous knockout of multiple inhibitory molecules (e.g., TCR, HLA class I, PD-1) in CAR cells, which is crucial for generating universal donor cells and enhancing anti-tumor activity [102]. This level of precision allows for fine-tuning of CAR cell properties, such as persistence, specificity, and resistance to TME immunosuppression [103, 104].

Concurrently, there is a growing emphasis on developing non-viral gene delivery methods, such as electroporation and lipid nanoparticles (LNPs), as safer and more scalable alternatives to traditional viral vectors [105]. These non-viral approaches address concerns related to the immunogenicity, cost, and manufacturing complexities associated with viral vectors [105]. A particularly promising area is in vivo CAR cell generation, where targeted lipid nanoparticles (tLNPs) are used for direct messenger RNA (mRNA) delivery to specific T-cell subsets within the patient[106]. This innovative approach bypasses the need for ex vivo manufacturing, significantly simplifying the treatment process, reducing costs, and increasing patient accessibility. The ability to generate CAR cells directly in vivo represents a paradigm shift, potentially making these advanced therapies much more widely available.

In vivo CAR engineering

Given the rapidly growing interest in in vivo CAR engineering, this approach offers a transformative alternative to traditional ex vivo methods by directly reprogramming endogenous immune cells within the patient. Key strategies include nanoparticle-based delivery (e.g., LNPs encapsulating mRNA or DNA for transient or stable CAR expression), viral vectors (e.g., lentiviral or AAV for targeted transduction of T or NK cells), and bioinstructive materials (e.g., scaffolds that recruit and activate cells locally). These methods eliminate manufacturing delays, reduce costs, and enable “off-the-shelf” scalability, with early clinical trials showing promising safety and efficacy in hematologic malignancies. However, challenges such as off-target delivery, immune clearance of vectors, and ensuring durable expression must be addressed. Recent reviews highlight the potential for in vivo engineering to treat not only cancer but also autoimmune diseases through precise immune modulation [107–109].

Next-generation gene editing: beyond CRISPR-Cas9

While CRISPR-Cas9 has revolutionized CAR engineering, double-strand breaks (DSBs) pose risks of translocations and genotoxicity. Base editing and prime editing are emerging as safer alternatives for multiplex engineering. These technologies allow for the precise disruption of inhibitory genes (e.g., PDCD1, TRAC, B2M) or the insertion of CAR constructs without inducing DSBs, significantly preserving genomic integrity in ‘off-the-shelf’ allogeneic products [110].

Logic-gated and adapter CAR platforms

To further refine specificity, ‘AND’ gate circuits (requiring simultaneous binding of two antigens, e.g., CD19 AND CD22) are moving from theory to practice to prevent antigen escape. Additionally, universal adapter CAR systems (e.g., biotin-binding immune receptors or split, universal, and programmable (SUPRA) CARs) allow clinicians to switch targeted antigens post-infusion simply by administering a different soluble adapter antibody, providing real-time control over specificity and toxicity management [111].

AI-driven CAR design and personalized approaches

Artificial intelligence (AI) and machine learning (ML) are rapidly emerging as transformative tools in the design, optimization, and clinical application of CAR-based therapies. AI algorithms can analyze vast datasets, including single-cell transcriptomics data from patient tumors, to identify optimal antigen circuits and predict the efficacy and safety of novel CAR designs.

Beyond CAR design, AI can significantly enhance precision medicine by assisting in various aspects of cancer management. This includes advanced tumor diagnosis, comprehensive evaluation of the TME characteristics (e.g., immune landscape, metabolic profiles), and real-time patient follow-up to predict treatment response[47]. The ability of AI to integrate complex immunogenomic profiling and predict optimal gene editing strategies allows for the tailoring of CAR-NK cell therapies to patient-specific HLA/KIR/SIRPα contexts[72]. Ultimately, AI-driven approaches are paving the way for individualized rational CAR design, promising unprecedented efficacy and safety through truly personalized treatment strategies.

Combination therapies for synergistic effects

The complex and multifaceted nature of cancer, particularly the dynamic interplay within the TME and the plasticity of cancer stem cells, necessitates combination therapeutic approaches to achieve durable remissions. Future strategies will increasingly integrate CAR-T/NK cell therapies with other modalities to achieve synergistic anti-tumor effects:

• Immune checkpoint inhibitors (ICIs) Combining CAR cells with ICIs (e.g., anti-PD-1/PD-L1, anti-CTLA-4 antibodies) can amplify anti-tumor effects, overcome TME-induced immunosuppression, and improve CAR cell persistence [112]. This combination aims to unleash the full potential of CAR cells by disarming the tumor’s immune evasion mechanisms.

• Conventional therapies Integrating CAR cells with chemotherapy or radiotherapy can enhance treatment efficacy. Chemotherapy can reduce tumor burden, address immunosuppressive cells, and potentially expose new antigens, while radiotherapy can improve CAR cell infiltration and promote immune responses [113, 114].

• Small molecule inhibitors Combining CAR cells with small molecule inhibitors targeting specific oncogenic pathways or TME components can disrupt CSC self-renewal, sensitize tumors to CAR cell attack, and improve CAR cell function in hostile microenvironments [94]. Examples include inhibitors of Wnt, Notch, Hedgehog, PI3K/AKT, NF-κB, JAK/STAT, TGF-β, MAPK, and Hippo pathways, as well as epigenetic regulators like m6A demethylases (e.g., FTO, METTL3 inhibitors) and KDM1B inhibitors [94, 115].

• Clinical Translation and Remaining Challenges.

Despite the remarkable progress, the widespread clinical translation of next-generation CAR-T and CAR-NK cell therapies still faces significant challenges. These include the need for improved clinical trial designs that can accurately assess the efficacy of CSC-targeted therapies, especially given their rare populations and dynamic plasticity [35]. Further characterization of CSCs and the TME, particularly at the single-cell level, is crucial to identify universal markers and understand the complex interplay that drives resistance [116]. Overcoming manufacturing complexities, high costs, and ensuring scalability remain critical for broader patient accessibility. The long-term safety profiles, particularly concerning potential off-target effects and secondary malignancies, require continued rigorous monitoring [47].

Conclusion

Next-generation CAR-T and CAR-NK cell therapies represent a transformative frontier in the treatment of hematologic malignancies. While initial CAR-T successes have revolutionized patient outcomes, the field is actively addressing critical limitations related to cell persistence, tumor specificity, and safety.

The path forward is characterized by sophisticated engineering. Enhancing metabolic fitness and employing cytokine engineering, ideally through inducible and controllable systems, are crucial for CAR cells to overcome the harsh, nutrient-deprived, and immunosuppressive tumor microenvironment.

Precision in targeting is being advanced through multi-antigen approaches (pooled, bicistronic, and tandem CARs) that combat tumor heterogeneity and antigen escape, as well as through meticulous affinity tuning of antigen-binding domains to balance efficacy with reduced off-target toxicity. The development of logic-gated CAR designs further refines specificity, enabling more intelligent and context-dependent cell activation. Critically, targeting cancer stem cell-specific antigens, while acknowledging their dynamic plasticity and ability to dedifferentiate, remains a key strategy to prevent disease recurrence.

Ensuring safety is paramount, with continuous advancements in managing cytokine release syndrome and neurotoxicity. The integration of inducible safety switches offers a crucial mechanism for controlling CAR cell activity in vivo. For allogeneic therapies, genetic engineering strategies such as TCR and HLA knockout/camouflage, coupled with the potential of iPSC-derived cellular products, are vital for preventing GvHD and host rejection, paving the way for truly universal CAR cell availability.

Ultimately, the future of CAR-T and CAR-NK cell therapies in hematologic malignancies lies in integrated, multi-modal, and adaptive approaches. This involves a synergistic combination of advanced gene editing and non-viral delivery systems, AI-driven personalized CAR design, and rational combination therapies with immune checkpoint inhibitors, conventional treatments, and small molecule inhibitors that modulate key signaling pathways and reshape the tumor microenvironment.

Author contributions

Mutaz Jamal Al-khreisat, Waleed K. Abdulsahib, Ihsan Khudhair Jasim, H. Malathi, Priya Priyadarshini Nayak, D. Alex Anand, Gunjan Mukherjee, Aashna Sinha, and Norbek Kholboyev contributed to the conception, design, and drafting of the manuscript. All authors read and approved the final version of the manuscript.

Funding

Not applicable.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.