The common γ-chain (γ_c) family of cytokines comprise IL-2, -4, -7, -9, -15, and -21, which all signal through receptors containing a shared γ_c subunit, and play crucial roles in lymphocyte proliferation, differentiation, and homeostasis (25). As such, this family of cytokines has received significant attention as potential armoring payloads for CAR T-cells.

Among these, IL-15 has emerged as a particularly promising candidate. IL-15-armored CAR T-cells have been demonstrated to exhibit enhanced proliferation, improved anti-tumor efficacy, sustained killing upon repeated tumor challenges in vitro and in vivo, prolonged survival in tumor-bearing mice, and promotion of central memory or stem cell memory-like phenotypes (26–30). In one study using a syngeneic melanoma mouse model, IL-15-armored CAR T-cells not only improved intrinsic T-cell function but also beneficially remodeled the TME, enhancing natural killer (NK) cell activation and reducing immunosuppressive ‘M2’ macrophage abundance (29). Another study found that CAR T-cells modified to fuse IL-15 to the scFv portion of the CAR could deplete immunosuppressive MDSCs in murine glioblastoma models, based on the observation that MDSCs in these models as well as human glioblastoma samples express the α subunit of the IL-15 receptor. This MDSC depletion was associated with reduced tumor growth and improved survival of the mice (31).

Despite these encouraging findings, concerns have been raised about the clinical translation of IL-15 armoring due to toxicities observed in trials of recombinant human IL-15, as well as some studies of IL-15-armored CAR T-cells in mice, including cytokine release syndrome (CRS), thrombocytopenia, and liver toxicity (32–36). To address safety concerns, several groups have developed regulatory mechanisms to try to prevent these toxicities. For example, Chen et al. incorporated an inducible caspase-9 (iC9) suicide gene into their IL-15-armored CAR construct. They demonstrated that administration of a chemical inducer of dimerization (CID) drug, which causes the iC9 protein to dimerize and activate apoptosis, enabled rapid elimination of the CAR T-cells (26).

IL-15-armored CAR T-cells, including those with the iC9 suicide gene, have now been evaluated in phase I trials. In one study, 4 out of 12 patients with glypican-3 (GPC3)-positive solid tumors achieved a partial response following treatment with IL-15-armored GPC3-targeting CAR T-cells, whereas none of the patients treated with unarmored CAR T-cells experienced an objective response (37). Notably, IL-15-CAR T-cell treatment was associated with a higher incidence of CRS that necessitated intervention with IL-1 or IL-6 blockade. In three patients with refractory CRS, activation of the iC9 switch successfully reduced circulating CAR T-cells and alleviated symptoms. Interestingly, this trial also reported an increase in effector memory and a decrease in central memory T-cell subsets among IL-15-armored CAR T-cells (37), an observation that contrasts with prior preclinical findings.

Another approach to circumvent the toxicities associated with constitutive IL-15 expression is to place IL-15 expression downstream of antigen-dependent T-cell activation. Ma et al. engineered CAR T-cells with IL-15 expression controlled by the interferon (IFN)-γ gene promoter, which activated upon TCR signaling. This antigen-dependent expression system led to selective IL-15 expression following tumor recognition and improved tumor control in a xenograft mouse model of gastric cancer (28).

The combination of two γ_c cytokines, IL-15 and IL-21, has also been tested in the preclinical setting, based on their known synergistic effects on cytotoxic lymphocytes (38, 39). CAR T-cells armored with both cytokines demonstrated superior anti-tumor activity in in vivo models compared to unarmored or single cytokine-armored CAR T-cells, and showed enhanced proliferation, increased frequencies of stem cell memory and central memory populations (40), and reduced exhaustion following repeated antigen stimulation (41).

The IL-12 family comprises heterodimeric cytokines with pleiotropic roles in immune regulation. Among these, IL-12 and IL-23, members of the family which are considered to have predominantly pro-inflammatory effects (42), have garnered particular interest as potential armoring payloads for CAR T-cells.

IL-12 consists of two subunits, a p35 and a p40, and is primarily produced by antigen-presenting cells (APCs) (42). It promotes IFN-γ secretion from T-cells and NK cells, which in turn can further stimulate IL-12 release by macrophages and other APCs (42, 43). As an armoring payload, IL-12 has been demonstrated to enhance CAR T-cell effector functions and reprogram endogenous immune cells in the TME toward a pro-inflammatory, anti-tumor state (23, 42, 44). Although early efforts to armor CAR T-cells with IL-12 showed improved anti-tumor activity, constitutive or systemic expression of IL-12 has been associated with significant toxicities (45–48). In a phase I trial of an IL-12-secreting CAR T-cell therapy, two-thirds of patients who also received lymphodepleting chemotherapy developed dose-limiting hemophagocytic lymphohistiocytosis or macrophage activation-like syndrome (48).

More recent efforts to develop IL-12-armored CAR T-cells have therefore also endeavored to improve their safety profile by restricting IL-12 expression to the TME. One solution has been to bind IL-12 to the membrane of CAR T-cells (49, 50), which Hu et al. combined with fusing the IL-12 to a peptide targeting cell surface vimentin, a protein overexpressed by various solid tumors (49). These studies demonstrated that, compared to unarmored CAR T-cells, membrane-bound IL-12-CAR T-cells released more IFN-γ, and demonstrated superior anti-tumor efficacy in murine models of solid tumors, including large established tumors, without inducing systemic toxicity. These effects were accompanied by changes in immune cell populations in the TME, including enhanced dendritic cell (DC) maturation (49, 50). Lee et al. confirmed these findings in a model of ovarian cancer peritoneal metastasis, where systemic IL-12 administration caused toxicity, but membrane-bound IL-12-armored CAR T-cells did not (50).

Hombach et al. inserted IL-12 into the extracellular domain of the CAR. This configuration modulated CAR T-cells toward an NK cell-like phenotype, enabling them to kill both antigen-positive and antigen-negative tumor cells. This was demonstrated both in vitro and in a mouse model of CEA-negative ovarian cancer, where IL-12-armored CAR T-cells were effective, while conventional unarmored CEA-targeted CAR T-cells were not (51).

Other methods of achieving tumor-restricted IL-12 expression include: fusing IL-12 to a collagen-binding domain to localize it to exposed collagen in tumor vasculature and stroma (52); fusing it to a tumor-specific scFv (53); and placing IL-12 expression under control of a nuclear factor of activated T-cells (NFAT)-responsive promoter to link cytokine secretion to CAR activation (54, 55).

IL-23 is a heterodimeric cytokine that shares the same p40 subunit as IL-12, but pairs this with a p19 subunit. Ma et al. discovered that due to upregulation of p19 expression in T-cells upon activation, transducing T-cells with the p40 subunit was sufficient to induce IL-23 secretion from T-cells. They demonstrated that IL-23-secreting CAR T-cells exhibited improved tumor killing and reduced functional exhaustion compared to unarmored CAR T-cells in syngeneic mouse models of solid tumors. Notably, IL-23 acted via autocrine signaling to enhance CAR T-cell function. When compared to other armoring strategies, IL-23-armored CAR T-cells did not induce weight loss in immunodeficient mice, suggesting a superior safety profile relative to IL-15 and IL-18-armoring (56).

In contrast, another study found that PSMA-targeted CAR T-cells engineered to co-express an IL-23-targeting monoclonal antibody (mAb) led to eradication of prostate cancer in a murine model, suggesting that IL-23 blockade, not supplementation, may be beneficial in some contexts (57). These findings illustrate that the benefits of armoring can be context-dependent, influenced by tumor type, the immune microenvironment, and the preclinical model being used. Moreover, IL-23 is a multifaceted cytokine with reported pro-tumoral and anti-tumor effects, which may reflect differences in cytokine concentration (56, 58). These observations underscore the importance of considering not only the CAR T-cell dose, but also the ‘dose’ of the armoring payload when evaluating therapeutic efficacy and safety, which are critical factors for clinical translation.

Apart from engineering CAR T-cells to secrete or express cytokines, another strategy to exploit cytokine signaling is to modify the cytokine receptors. This approach has been applied particularly to cytokines of the γ_c family and circumvents the off-target effects that can result from transgenic cytokine expression, which may activate signaling on bystander cells in addition to the CAR T-cells themselves. Shum et al. developed a constitutively signaling IL-7 receptor, termed C7R, which activated the IL-7 signaling axis in CAR T-cells but was unresponsive to extracellular IL-7. They demonstrated that C7R-expressing CAR T-cells were more resilient to repeated tumor challenges, and capable of controlling or clearing tumors in several in vivo models (68). A phase I clinical trial of C7R-CAR T-cells demonstrated tolerability, with mostly grade 1 inflammation-associated toxicities, apart from one case of grade 4 CRS, and reported partial responses in 2 out of 7 patients with diffuse midline glioma (69). Another approach to harnessing IL-7 signaling has been to incorporate a portion of the IL-7 receptor-α intracellular domain into the CAR structure itself, which was demonstrated to enhance T-cell proliferation and improve tumor control (70, 71).

An additional strategy to avoid off-target effects is the design of mutated, or orthogonal, cytokine and receptor pairs that interact exclusively with each other, and not with their native counterparts. This concept was first demonstrated using an orthogonal IL-2 and IL-2 receptor pair (72). Following on from this, Kalbasi et al. engineered T-cells to express a synthetic cytokine receptor ‘o9R’, which combines the extracellular domain of the orthogonal IL-2 receptor with the intracellular domain of the IL-9 receptor. Upon stimulation with orthogonal IL-2, o9R signaling induced features consistent with a stem cell memory phenotype in the T-cells, an attribute potentially beneficial for CAR T-cell therapy. CAR T-cells expressing o9R and receiving orthogonal IL-2 also exhibited greater anti-tumor activity in syngeneic mouse models compared to CAR T-cells expressing the orthogonal IL-2 receptor (73).

More recently, Bell et al. replaced the extracellular domains of several heterodimeric cytokine receptors, including those of γ_c cytokines, and IL-10 and IL-12 receptors, with heterodimerizing leucine zipper motifs. This modification led to constitutive signaling through these receptors. CAR T-cells expressing these leucine zipper-modified cytokine receptors showed improved cytotoxicity in response to multiple tumor challenges in vitro, and CAR T-cells with a modified IL-2 receptor demonstrated superior anti-tumor activity in vivo in mouse xenograft models of lung cancer and sarcoma (74).

Designing synthetic circuits to restrict cytokine production specifically within the tumor is another strategy aimed at preventing systemic cytokine release and thereby reducing toxicity. Synthetic Notch (synNotch) receptors have been a pioneering example of synthetic gene circuits in the CAR field. These receptors consist of a synthetic antigen-recognizing extracellular domain, a transmembrane domain derived from the Notch signaling receptor, and a synthetic intracellular transcription factor. When the extracellular domain binds its cognate antigen, the synNotch receptor undergoes transmembrane cleavage, releasing the transcription factor to activate expression of chosen target genes placed downstream of a synthetic promoter (75). While the downstream payload can, in principle, be any gene of interest, synNotch-controlled cytokine release provides a compelling example of the utility of this system. Allen et al. demonstrated that CAR T-cells co-expressing a synNotch receptor designed to drive IL-2 production were able to eradicate solid tumors without inducing significant toxicity (76).

Although TGF-β is a cytokine with pleiotropic effects, it has drawn particular interest in the CAR T-cell field for its tumor-promoting and immune-inhibitory roles in the TME of many solid tumors, especially its suppression of T-cell effector functions (77, 78). Kloss et al. aimed to counteract this by armoring CAR T-cells with a ‘dominant-negative’ TGF-β receptor II (dnTGF-βRII), a truncated receptor lacking the intracellular signaling domain, which acts as a decoy to sequester TGF-β and prevent signaling in both CAR T-cells and neighboring immune cells. These dnTGF-βRII-armored CAR T-cells were resistant to TGF-β-induced suppression and exhibited enhanced proliferation and improved anti-tumor efficacy in a xenograft mouse model of prostate cancer (79). Enhanced efficacy with dnTGF-βRII armoring has since been reported in preclinical models of prostate cancer (80), ovarian cancer (81, 82), pancreatic cancer (83, 84), esophagogastric cancers (84), and glioblastoma (85). Notably, Tran et al. showed that their ROR1-targeting CAR T-cells induced TGF-β expression in tumor cells that did not initially overexpress this cytokine, but that this was mitigated by dnTGF-βRII armoring, which restored anti-tumor activity (83). A further advanced version of this strategy replaces the intracellular domain of TGF-βRII with that of another receptor, such as IL-15 receptor-α, thereby simultaneously blocking TGF-β and inducing IL-15 signaling. This dual-function receptor, named TB15, was shown to enhance CAR T-cell function beyond dnTGF-βRII or IL-15 armoring alone (86).

A dnTGF-βRII-armored PSMA-targeting CAR T-cell has since been evaluated in a phase I trial against metastatic castration-resistant prostate cancer. Grade 2 or higher CRS occurred in 5 out of 13 patients and was mostly manageable with immunosuppressive therapy, although one patient developed grade 4 CRS and concurrent sepsis and subsequently died. Four patients experienced ≥30% reductions in prostate-specific antigen levels, although the best radiographic response was stable disease (87). In a second phase I trial, two of nine patients experienced fatal immune-mediated toxicities, leading to early trial termination (88, 89). Retrospective analyses could not conclusively explain these severe toxicities, although the investigators speculated that modifying the co-stimulatory domain by replacing 4-1BB might help mitigate cytokine release without impairing efficacy (89). These findings underscore the gap between promising preclinical outcomes and complex clinical realities, possibly due to the widespread use of immunodeficient mouse models, which fail to fully capture human immune responses and associated toxicities.

An alternative strategy for blocking TGF-β signaling involves engineering CAR T-cells with an extracellular TGF-β-binding scFv. One group found that this approach paradoxically converted TGF-β into a stimulatory signal that enhanced CAR T-cell proliferation and activated immunostimulatory pathways (90). This group subsequently developed a bispecific CAR, targeting both IL-13Rα2 and TGF-β, which improved survival in glioblastoma mouse models (91). Other approaches for inhibiting TGF-β signaling include SMAD7-expressing CAR T-cells (92, 93), and knockout of the endogenous TGF-βRII in CAR T-cells using CRISPR/Cas9 technology (94), both of which improved anti-tumor efficacy in vivo.

A major factor limiting CAR T-cell efficacy within the TME is exhaustion, which can involve multiple signaling pathways, including activation of immune checkpoints that suppress CAR T-cell proliferation, cytotoxicity and other effector functions (95, 96).

PD-1, a cell surface receptor on T-cells, is one such inhibitory immune checkpoint; its sustained expression or upregulation is widely regarded as a hallmark of exhaustion (97). An armoring strategy that has gained significant interest involves engineering CAR T-cells that secrete anti-PD-1 scFvs, thereby blocking the PD-1/PD-L1 axis within the TME (95). This approach has been demonstrated to enhance CD107a expression (a marker of degranulation), as well as secretion of effector molecules, such as IFN-γ and granzyme B, by both CAR T-cells and endogenous tumor-specific T-cells, suggesting a broader restoration of anti-tumor immunity (98–100). Anti-PD-1 scFv-secreting CAR T-cells have demonstrated improved anti-tumor activity in preclinical in vivo models of non-small cell lung cancer (NSCLC) (98), breast cancer (101), ovarian cancer (99), and hepatocellular carcinoma (HCC) (100). Similarly, CAR T-cells that secrete antibodies that block PD-L1, the ligand for PD-1, have demonstrated enhanced efficacy in models of pancreatic cancer (102) and renal cell carcinoma (RCC) (103). In a humanized RCC mouse model, Wang et al. reported that tumors treated with anti-PD-L1 antibody-secreting CAR T-cells exhibited increased endogenous T-cell infiltration and a reduction in M2-like macrophages, indicating TME modulation by the secreted payload (103). Based on such promising preclinical results, several anti-PD-1 or anti-PD-L1-armored CAR T-cells have entered phase I clinical trials. One report described CAR T-cells with IFN-γ-induced secretion of anti-PD-1 nanobodies achieving partial responses in 6 out of 11 patients with mesothelin and PD-L1-positive mesotheliomas, as well as one complete response, and toxicities that were manageable with supportive care (104). Further reports from early phase clinical studies are awaited.

Combinations of PD-1 or PD-L1 blockade with additional mechanisms to enhance CAR T-cell function or modulate the TME have also been explored. For example, one group armored CAR T-cells to secrete a bispecific scFv targeting both PD-1 and TREM2 (105), an immunoglobulin receptor expressed on immunosuppressive TAMs and MDSCs (105, 106). These dual-targeting CAR T-cells outperformed those secreting scFvs against either PD-1 or TREM2 alone in vivo, and were associated with decreased M2 macrophages and MDSCs, and increased proportion of CD8⁺ T-cells (105). Another strategy combined PD-1 and TGF-β pathway inhibition via secretion of a bifunctional ‘trap’, a fusion of an anti-PD-1 scFv with the TGF-βRII ectodomain. In a xenograft model of prostate cancer, these dual-inhibition CAR T-cells suppressed tumor growth more effectively than unarmored or anti-PD-1-only armored CAR T-cells (107).

Another immune checkpoint, often referred to as the ‘don’t eat me’ signal, involves CD47, a molecule frequently overexpressed on tumor cells. CD47 binds to the signal regulatory protein α (SIRPα) receptor on APCs, thereby inhibiting their phagocytic function (108). One strategy to counter this has been to engineer CAR T-cells to secrete a fusion protein, SIRPα-Fc, which blocks CD47. Chen et al. provided in vitro evidence that SIRPα-Fc enhanced macrophage phagocytic activity. In vivo, SIRPα-Fc-armored CAR T-cells exhibited improved anti-tumor efficacy and prolonged survival across several mouse models of solid tumors. This armoring also modulated immune cell phenotypes, promoting a T-cell central memory phenotype and increasing M1 macrophages and CD11c⁺ DCs in the TME (109).

Alternative strategies to target this axis have also been reported. Martins et al. armored CAR T-cells with a signal regulatory protein γ (SIRPγ)-related protein that binds to CD47 with high affinity, while Xie et al. engineered CAR T-cells to secrete CD47-specific nanobodies. In both cases, blocking the CD47 checkpoint enhanced CAR T-cell function and anti-tumor activity, including in models with tumor antigen heterogeneity (110, 111).

Interestingly, another group observed that concomitantly administering an anti-CD47 mAb with CAR T-cells depleted the CAR T-cells themselves secondary to macrophage-mediated phagocytosis, revealing a significant potential shortcoming of utilizing CD47 blockade as an armoring strategy (112). This issue was overcome by engineering CAR T-cells to co-express a variant of CD47 (CD47_E) that retained SIRPα interaction but was resistant to blockade by anti-CD47 antibodies. Co-administering CD47_E-CAR T-cells and a CD47-blocking mAb thereby protected the CAR T-cells from macrophage-mediated phagocytosis, whilst unleashing phagocytosis against tumor cells. In multiple in vivo models, CD47_E-CAR T-cells administered alongside a CD47-blocking mAb exhibited increased persistence, enhanced macrophage infiltration into the TME, and improved anti-tumor activity (112).

The solid TME is characterized by hypoxia, metabolic stress, and high cellular turnover, which leads to the accumulation of extracellular adenosine triphosphate (ATP). This ATP is converted to adenosine by CD39 and CD73, enzymes which are expressed on tumor cells and immunosuppressive cells within the TME. Activation of adenosine receptor 2A (A_2AR) by adenosine represents another immune checkpoint that impairs anti-tumor immune responses, including limiting CAR T-cell effector functions (113–116).

Adenosine deaminase (ADA) is an enzyme that catabolizes adenosine into inosine, thereby preventing accumulation of adenosine. Qu et al. found that armoring CAR T-cells to secrete ADA increased the resistance of CAR T-cells to exhaustion, promoted intratumoral CAR T-cell expansion, and improved tumor control (117). Similarly, Hu et al. explored the benefits of armoring CAR T-cells with ADA, hypothesizing that this strategy would not only relieve adenosine-mediated immunosuppression, but also provide inosine as an alternative fuel to support CAR T-cell proliferation and function. To avoid supplying inosine to tumor cells, they engineered ADA-secreting CAR T-cells with the secreted ADA anchored to the CAR T-cell. This strategy resulted in enhanced CAR T-cell proliferation, higher inosine concentrations, greater T-cell infiltration into the TME, and improved anti-tumor activity (118).

Other strategies to combat adenosine-induced immunosuppression have also been explored, including knocking-down or deleting A_2AR expression in CAR T-cells using short hairpin RNA (shRNA) (119) or CRISPR/Cas9 technology (120, 121). These approaches have been demonstrated to enhance CAR T-cell effector functions. As gene editing technologies like CRISPR/Cas9 continue to improve in safety and efficacy, the concept of armoring CAR T-cells may evolve beyond supplying one or two transgenic payloads to include more complex modifications to the CAR T-cell genome (122, 123).

Apart from immune-inhibitory signals from tumor and immune cells, the activities of other stromal cells in the TME, such as CAFs, can also contribute to the suppression of CAR T-cell effector functions. CAFs secrete soluble factors, including immunosuppressive cytokines, and play a key role in producing the collagen-rich extracellular matrix (ECM) which create a physical barrier to CAR T-cell infiltration (18, 19, 124). One strategy to counteract CAFs is to target fibroblast activation protein (FAP), a pan-CAF marker, which some groups have achieved by engineering FAP-targeting CAR T-cells (125–127). One study found that treating pancreatic tumors, characterized by their dense desmoplastic stroma, in mice with FAP-targeting CAR T-cells reduced the integrity of the desmoplastic matrix, thereby making the tumors more susceptible to infiltration by CAR T-cells targeted against mesothelin, a TAA (126). Wehrli et al. also aimed to address CAF-associated immunosuppression by armoring mesothelin-targeting CAR T-cells to secrete a bispecific molecule targeting both CD3 and FAP, thereby inducing CAR-mediated killing of antigen-expressing tumor cells and engaging T-cells to kill CAFs. These armored CAR T-cells were shown to be effective against multiple models of pancreatic cancer (128). Notably, the FAP-targeted scFv used in this study did not cross-react with murine FAP, a challenge that the authors navigated by using several preclinical models, including patient-derived organoids with patient-matched CAFs, to characterize the armored CAR T-cells (128). This highlights one challenge that affects the study of armored CAR T-cells, which is that of potential cross-species barriers between the armoring payload and host cells in preclinical mouse models.

Another approach to combating the tumor stroma is to target the ECM, which is composed of macromolecules including collagens, proteoglycans, and glycoproteins (129). One strategy is to armor CAR T-cells with enzymes that degrade ECM components. Both heparanase-expressing CAR T-cells (130) and hyaluronidase-expressing CAR T-cells (131) have demonstrated increased tumor infiltration and enhanced anti-tumor activity, including when hyaluronidase and IL-7 armoring were combined (132). Moreover, Zheng et al. engineered CAR T-cells to express a synNotch receptor that drives the expression of three ECM-degrading enzymes, matrix metalloproteinase (MMP) 9, MMP12, and heparanase. They demonstrated that these synNotch CAR T-cells were more resistant to exhaustion and apoptosis after tumor challenge and exhibited enhanced tumor infiltration and anti-tumor activity in vivo (133).

Bispecific T-cell engagers (BiTEs) are fusion proteins specific for two targets, one of which is an anti-CD3 scFv that engages T-cells, thereby enabling T-cell-mediated killing of cells expressing the second target (152). While BiTEs are being developed as standalone immunotherapeutics, engineering CAR T-cells to secrete BiTEs has also emerged as a strategy to overcome antigen heterogeneity. For example, Choi et al. engineered EGFRvIII-targeting CAR T-cells that secreted BiTEs against EGFR, and demonstrated in vitro that these armored CAR T-cells could recruit bystander T-cells to kill tumor cells. In orthotopic mouse models of glioblastoma with heterogeneous or negative EGFRvIII expression, these BiTE-armored CAR T-cells induced tumor regression (153). This strategy has since been tested in a first-in-human trial involving intraventricular delivery to three patients with recurrent glioblastoma, where it led to rapid tumor shrinkage. However, for two of the patients, the response was transient, underscoring the need for further optimization to enhance durability (154).

Similar strategies have been explored in other tumor types. These include: Muc16-targeting CAR T-cells armored with BiTEs against a peptide derived from the intracellular TAA Wilms tumor 1 (WT1), which demonstrated improved efficacy in ovarian cancer models with low Muc16 expression (155); and GPC3-targeting CAR T-cells that secrete BiTEs against B7-H3, which showed enhanced cytotoxicity against HCC cell lines with heterogeneous expression of GPC3 and B7-H3, although this was demonstrated only in vitro (156).

The repertoire of bispecific molecules that can be secreted by CAR T-cells continues to expand. Based on their findings of abundant CD16a-expressing innate immune cell infiltration in high-risk neuroblastoma TMEs, Pascual-Pasto et al. decided to engineer anti-glypican-2 (GPC2) CAR T-cells that secreted a bispecific innate immune cell engager (BiCE) targeting GD2 and CD16a, the latter serving to engage CD16a⁺ innate immune cells, including NK cells. They demonstrated that BiCE-armored CAR T-cells exhibited CAR-mediated cytotoxicity against GPC2-positive tumor cells and harnessed bystander NK cell activity against GD2-positive cells. This dual activity led to enhanced intratumoral NK cell retention and improved anti-tumor efficacy in patient-derived xenograft models of neuroblastoma in mice (157).

Damage-associated molecular patterns (DAMPs) and pathogen-associated molecular patterns (PAMPs) are molecular motifs recognized by pattern recognition receptors (PRRs), which can activate innate immune responses in an antigen-independent manner (158). As such, they offer a potential strategy for circumventing tumor antigen heterogeneity.

Toll-like receptors (TLRs) are an important class of PRRs, each receptor recognizing distinct PAMPs in a specific subcellular compartment (158). TLR5, for example, recognizes flagellin, a protein subunit of bacterial flagella. Flagellin can initiate pro-inflammatory signaling pathways via TLR5, which is expressed on several innate immune cell types including macrophages and DCs, and has therefore attracted attention for its potential use as a vaccine adjuvant (159). Several groups have also explored its use as an armoring payload in CAR T-cells. More recently, Niu et al. engineered CAR T-cells that secreted flagellin under the control of an NFAT-responsive promoter. While flagellin armoring did not enhance the cytotoxicity of CAR T-cells in vitro, it significantly prolonged tumor control and survival in immunocompetent mouse models of solid tumors, suggesting that its benefit lies in activating endogenous immune responses. Supporting this, the authors observed increased intratumoral abundance of CD86⁺ macrophages (indicative of M1 macrophage polarization) and CD103⁺ DCs (associated with DC activation). Additionally, in a model of antigen-heterogeneous melanoma, flagellin-armored CAR T-cells suppressed tumor growth more effectively than their unarmored counterparts (160).

Another bacterial protein explored as a CAR T-cell armoring payload is neutrophil-activating protein (NAP) from the bacterium Helicobacter pylori, which has been shown to exert various immunomodulatory effects. These include acting as a neutrophil chemoattractant and promoting IL-12 and IL-23 expression via TLR2 on neutrophils and monocytes, thereby supporting Th1 responses (161, 162). Jin et al. demonstrated that armoring CAR T-cells with NFAT promoter-inducible NAP improved anti-tumor activity in multiple syngeneic tumor models, including those with antigen-heterogeneity. NAP-armored CAR T-cells increased tumor infiltration of neutrophils, M1 macrophages and NK cells, and promoted ‘epitope spreading’ among endogenous CD8⁺ T-cells, that is, activation of CD8⁺ T-cells recognizing antigens other than the CAR-targeted TAA, suggesting robust bystander immune activation (163).

Beyond protein-based PRR agonists, nucleic acid-sensing pathways have also been leveraged. The cytosolic PRR RIG-I is activated by RNA and induces type I IFNs, which play broad roles in innate and adaptive immunity (164). Johnson et al. engineered CAR T-cells to secrete RN7SL1, a non-coding RNA that activates RIG-I, via extracellular vesicles that are preferentially taken up by immune rather than tumor cells. They found that delivery of RN7SL1 by the CAR T-cells promoted pro-inflammatory features and restricted immunosuppressive features among endogenous myeloid cells, and promoted the development of endogenous CD8⁺ T-cells with an effector memory-like phenotype. When combined with immune checkpoint blockade, RN7SL1-CAR T-cells eradicated antigen-heterogeneous melanomas in immunocompetent mice and prevented tumor outgrowth upon rechallenge with antigen-negative tumor cells, suggesting the generation of CAR antigen-independent memory (165).

Finally, PRR agonists as immune adjuvants have been combined with CAR T-cells secreting immune cell growth factors as another method of harnessing the endogenous immune system. Lai et al. armored CAR T-cells with Fms-like tyrosine kinase 3 ligand (Flt3L), a hematopoietic growth factor that is essential for steady-state DC development (166, 167), and demonstrated associated intratumoral expansion of conventional type 1 DCs, which play an important role in cross-presenting antigens to and activating cytotoxic CD8⁺ T-cell responses. When combined with poly(I:C), an immunostimulatory TLR3 agonist, Flt3L-secreting CAR T-cells significantly suppressed tumor growth in both antigen-homogeneous and heterogeneous tumor models. Treated mice also rejected antigen-negative tumor rechallenges, again indicating induction of epitope spreading and a durable anti-tumor immune response (168).

Trogocytosis is a phenomenon that has received increased attention in recent years within the field of immunology. It refers to the transfer of fragments of cell membrane, along with associated proteins and other molecules, between cells that are in direct contact with each other (169). Trogocytosis has been observed between tumor cells and CAR T-cells, leading to the transfer of TAAs from the tumor cells to the CAR T-cells. This can contribute to antigen escape and even CAR T-cell fratricide, where CAR T-cells inadvertently recognize and kill one another (170). In some cases, CAR molecules themselves can be transferred from CAR T-cells to tumor cells, further reducing anti-tumor efficacy (171). Lu et al. identified cholesterol metabolism as a key pathway involved in trogocytosis, potentially due to cholesterol’s role in modulating the biophysical properties of lipid membranes. They found that ATF3, a stress-responsive transcription factor, is upregulated in cytotoxic T lymphocytes (CTLs) exposed to tumor-derived signals. ATF3 suppresses expression of the enzyme cholesterol 25-hydroxylase (CH25H), which in turn promotes trogocytosis between CTLs and tumor cells, leading to T-cell dysfunction and tumor progression. By armoring CAR T-cells to co-express CH25H, they were able to reduce the extent of trogocytosis, preserve antigen expression on tumor cells, and enhance CAR T-cell infiltration and anti-tumor efficacy (172). Of note, other recent work has also examined how incorporation of different CAR transmembrane domains affect the probability of trogocytosis occurring, and how this can be used to modulate the effects of CAR T-cell therapy (173).

CXCL8, also known as IL-8, is a chemokine associated with poor prognosis in numerous solid tumor types and with the recruitment of immunosuppressive immune cells to the TME. Several groups have sought to target this by armoring CAR T-cells with CXCR1 or CXCR2, both of which are CXCL8 receptors, with the aim of exploiting tumor CXCL8 expression to improve CAR T-cell migration to tumors (12, 176, 177). Notably, CXCL8 expression has been shown to be upregulated by irradiation of some tumors, suggesting the benefits of combining radiotherapy with CAR T-cell therapy in specific contexts (176, 177). CXCR1/CXCR2-armoring has been demonstrated to improve CAR T-cell infiltration, persistence, and tumor control in vivo across several studies (176–180). However, findings by Talbot et al. suggested that the benefits of CXCR2-armoring were context-dependent, as this did not enhance tumor control in a rapidly growing primary orthotopic xenograft model of osteosarcoma, but did confer benefits in a metastatic model (179). Interestingly, multiple groups have reported potentially improved safety profiles of CXCR-armored CAR T-cells compared to unarmored counterparts, possibly due to more targeted trafficking to tumors and thus fewer off-tumor toxicities (178, 179). On the other hand, one group found that while CXCR2 armoring enhanced trafficking into the TME, CAR T-cell persistence, and therefore anti-tumor efficacy, remained limited by the immunosuppressive TME. They demonstrated that co-armoring with the pro-inflammatory cytokines IL-15 or IL-18 could overcome this limitation, resulting in significantly improved in vivo anti-tumor activity (181).

Lesch et al. found CXCL16 to be highly expressed by both tumor and myeloid stromal cells in murine pancreatic cancer models and human specimens. Armoring CAR T-cells with CXCR6, the receptor for CXCL16, led to increased anti-tumor efficacy and prolonged the survival of treated mice (182). They also noted elevated expression of very late antigen-4 (VLA-4), an integrin that facilitates transendothelial migration of lymphocytes, in CXCR6-transduced T-cells, suggesting that harnessing the CXCL16-CXCR6 axis may offer multiple benefits for improving CAR T-cell trafficking. Talbot et al., who also tested CXCR6-armored CAR T-cells against osteosarcoma, found similar benefits in a metastatic xenograft model, and again reported an improved safety profile compared to unarmored CAR T-cells (179).

The CCL2-CCR2 axis has also drawn attention, as some solid tumors, including pleural mesothelioma and NSCLC, can secrete high levels of CCL2 (183, 184). CAR T-cells transduced with CCR2b, the dominant isoform of CCR2, showed improved tumor infiltration and tumor control in mouse models, including successfully crossing the blood-brain barrier to control brain metastases from NSCLC (183–185). On a similar theme, Li et al. armored CAR T-cells with CXCR5, the receptor for CXCL13, based on the expression of CXCL13 in a high proportion of NSCLC samples. These CAR T-cells exhibited enhanced migration toward CXCL13-positive targets and improved in vivo anti-tumor activity (186). Moreover, the combination of chemokine receptor (including CCR2b and CXCR5) and IL-7 armoring has been demonstrated to boost CAR T-cell expansion and anti-tumor activity in immunodeficient mouse tumor models (187, 188).

Tian et al. identified the CCL4/CCL5-CCR5 axis as one that is significantly correlated with CD8⁺ T-cell infiltration in solid tumors. They engineered CAR T-cells to express both CCR5 and IL-12, which enhanced T-cell trafficking and reversed TAM-induced immunosuppression (189). Conversely, the CCL1-CCR8 axis has been identified as having a role in the recruitment of immunosuppressive Tregs into the TME. Cadilha et al. armored CAR T-cells with both CCR8, to drive recruitment of CAR T-cells into the TME, and dnTGF-βRII, reasoning that this would counteract any Treg-mediated immunosuppression. They demonstrated that these dual-armored CAR T-cells significantly improved tumor control and survival in vivo (190).

Finally, chemokine armoring has also been found to be beneficial for recruiting other immune cells to the TME. Dual armoring with IL-7 and CCL19 has received particular focus, as these cytokines are essential for organizing the T-cell zone in lymphoid organs, and for recruiting T-cells and DCs expressing CCR7, the receptor for CCL19 (191, 192). Adachi et al. applied this mechanism to facilitate T-cell and DC recruitment to tumors (193), a strategy that has since also been studied by other groups. IL-7 and CCL19 dual armoring has been found to result in enhanced proliferation of the CAR T-cells, reduced expression of exhaustion markers, development of a central memory phenotype, remodeling of the TME, and improved in vivo control of tumors and survival in preclinical mouse models, including survival superior to armoring with either cytokine alone (193–196). Similar findings have been reported with dual armoring with IL-7 and CCL21, another CCR7 ligand, which also promoted T-cell and DC recruitment, increased central memory phenotype T-cells, and improved tumor control in an antigen-heterogeneous tumor model (197). Early phase clinical trials of dual IL-7 and CCL19-armored CAR T-cell have reported preliminary evidence of anti-tumor activity and no grade >2 toxicities (198, 199), although further studies will be needed to confirm clinical benefit.

Taken together, this wide assortment of studies have demonstrated the utility of exploiting chemokine signaling pathways to improve recruitment of CAR T-cells and other immune cells to solid tumors. However, the benefits of individual strategies are likely to be context-dependent and must be tailored according to the chemokine/chemokine receptor expression profile of the specific tumor being targeted. While chemokine receptor armoring alone has yielded enhanced anti-tumor effects in some instances, other studies suggest that additional payloads, such as pro-inflammatory cytokines, may be necessary to overcome the immunosuppressive TME and fully realize therapeutic potential.

An armoring payload that potentially provides the benefits of both chemokine and pro-inflammatory signaling is LIGHT, also known as tumor necrosis factor superfamily member 14 (TNFSF14). Using RNA sequencing data from The Cancer Genome Atlas (TCGA) database, Zhang et al. identified a strong correlation between LIGHT expression and gene signatures of tertiary lymphoid structures (TLS) (200), structures in tumors that play key roles in recruiting immune cells and shaping immune responses (201). They demonstrated that CAR T-cells armored with LIGHT upregulated expression of chemokines CCL19, CCL21, and CXCL13 in both stromal cells and tumor cells, enhancing the ability of the stromal cells to recruit T-cells in migration assays. In vivo, LIGHT-armored CAR T-cells improved tumor infiltration, accelerated tumor regression, and prolonged survival in mice compared to their unarmored counterparts (200).

Separately, another group characterized the benefits of LIGHT-armoring from the perspective of overcoming tumor antigen heterogeneity (202). LIGHT binds to two receptors: lymphotoxin-β receptor (LTβR), which is broadly expressed on non-lymphoid cells, and herpesvirus entry mediator (HVEM), expressed on various immune cells. LIGHT–HVEM signaling has been shown to promote pro-inflammatory responses, including induction of Th1-type immunity and enhancement of NK and CD8⁺ T-cell activity (203, 204). In contrast, LIGHT–LTβR interactions can induce apoptosis in LTβR-expressing tumor cells, suggesting a mechanism for direct, antigen-independent tumor cell killing (205). Cai et al. demonstrated that LIGHT-armoring enhanced CAR T-cell proliferation and cytotoxic function in vitro, including against tumor cell lines with heterogeneous antigen expression. In vivo, LIGHT-armored CAR T-cells showed increased tumor infiltration and improved tumor control in immunodeficient mouse models, with no detectable toxicities observed in immunocompetent mice (202).