How can we balance risk and benefit of interleukin-18 armored T cell therapies?

Impact statement
Although successful in the treatment of specific blood cancers, CAR T cell therapy has shown limited efficacy against solid tumors. A key barrier in this regard is the highly immunosuppressive nature of the solid tumor microenvironment (TME). One proposed means to address this entails the co-engineering of CAR T cells to produce interleukin (IL)-18, an approach that is currently being investigated in clinical trials. This minireview provides an overview of published clinical and preclinical studies of IL-18 armoring. We conclude that IL-18 consistently improves the anti-tumor efficacy of CAR T cells, but may elicit toxicities that arise from its pro-inflammatory properties. We also describe a number of strategies that set out to harness this cytokine in a more tumor-targeted and ultimately safer manner.

Introduction
Chimeric antigen receptors (CARs) are synthetic transmembrane proteins that redirect the MHC-independent activation of immune effector cells (notably T cells) when they encounter native cell surface antigens [1]. As of December 2025, the United States Food and Drug Administration had approved seven autologous CAR T cell therapies targeting either CD19 in B-cell malignancies or B cell maturation antigen in multiple myeloma [2–4]. Recent years have seen an increase in global efforts to extend CAR T cell therapy to solid tumors, with over 300 registered clinical trials collectively targeting 46 distinct solid tumor-associated antigens [5].
To achieve clinical efficacy, particularly against solid cancers, CAR T cells need to avoid attenuation by immunosuppressive cells residing in the tumor microenvironment (TME) [6]. Regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs) and M2 polarized macrophages can all inhibit T cells through inhibitory immune checkpoints or immunosuppressive factors, such as transforming growth factor-beta (TGF-β) and interleukin (IL-)10 [7].
Fourth generation CAR T cells, also known as armored CARs or TRUCKs (T cells Redirected for Universal Cytokine Killing), are designed to counteract the immunosuppressive TME by secreting pro-inflammatory cytokines [8, 9]. Commonly used examples include IL-7, IL-12, IL-15 or IL-18 [10–13]. In this minireview, we focus specifically on IL-18 and on strategies to maximize the therapeutic index of this approach.
Historically, IL-18 was known as interferon (IFN)-γ inducing factor owing to its ability to enhance IFN-γ secretion by CD4+ T cells, CD8+ T cells and Natural Killer (NK) cells [14–16]. Interleukin 18 is produced mainly by macrophages and dendritic cells (DCs) and is released as an inert precursor known as pro-IL-18. Canonical activation of pro-IL-18 is mediated by caspase-1 cleavage in the inflammasome, after which biologically active IL-18 is secreted through Gasdermin-D plasma membrane pores [17–19]. Upon release, IL-18 either binds to IL-18 receptor (IL-18R)α on the surface of T cells and NK cells or is neutralized by a soluble decoy receptor, IL-18 binding-protein (IL-18BP) [20, 21]. The interaction between IL-18Rα and IL-18 is stabilized by the accessory receptor IL-18Rβ, which facilitates activation of the transcription factor, NF-κB through the adaptor proteins, TRAM (TRIF-related adaptor molecule) and MyD88 (myeloid differentiation primary response protein 88) [22, 23].
Interleukin 18 exerts pleiotropic proinflammatory actions that include NK cell activation, DC maturation and context-dependent activation of either Th1 or Th2 responses [24–26]. The anti-tumor effects of IL-18 alone or in combination with IL-12 were reported in mouse models as early as 1998 [27–29]. Subsequent clinical trials of recombinant human IL-18 confirmed that it had a reasonable safety profile (albeit linked to some grade 3 and one grade 4 adverse reactions) [30, 31]. However, anti-tumor efficacy was insufficient [30, 31], prompting research into IL-18-based combination therapies including CAR T cell approaches.

IL-18 armoring improves anti-tumor activity of engineered T cells in pre-clinical models
Armoring/supplementation of either CAR- and T cell receptor (TCR)-engineered T cells with IL-18 has been evaluated in several independent preclinical studies, employing both xenograft and syngeneic mouse models (Table 1). In all but two cases, IL-18 armoring was achieved by stable viral transduction. By contrast, Huang et al. [40] administered mature IL-18 protein by intraperitoneal injection, while Olivera et al. [43] transiently electroporated T cells with IL-18 mRNA, IL-12 mRNA, or both. In 13 of these studies, IL-18 was shown to boost anti-tumor efficacy. Exemplifying this, Chmielewski et al. [37] and Ng et al. [36] both showed that IL-18 facilitated CAR-mediated tumor clearance in the setting of low target antigen expression. Efficacy was further improved by fusing biologically active IL-18 to a leader peptide to direct release of constitutively active IL-18 via the secretory pathway [33]. Additionally, decoy-resistant versions of IL-18 have been generated by mutagenesis, obviating the natural antagonistic effects of IL-18BP on this cytokine [43, 47].
In contrast, two studies reported that IL-18 armoring alone was not sufficient to boost CAR T cell anti-tumor activity. Ma et al. found that IL-18-armored anti-GD2 CAR T did not prolong the survival of mice engrafted with a CHLA-255 human neuroblastoma xenograft, instead inducing toxicity manifested as weight loss [41]. Such toxicity was also reported by Fisher-Riepe et al. in a similar model [39]. Additionally, Olivera et al. found that Pmel-1 TCR transgenic T cells and anti gp75 CAR T-cells achieved improved tumor control only if engineered to transiently co-express IL-12 and IL-18 mRNA, but not IL-18 alone [43]. This finding is in line with an earlier report that IL-12 upregulates IL-18Rβ and thus synergizes with IL-18 in inducing IFN-γ expression [48]. Contrasting with this however, Chmielewski et al. found that the combination of IL-12 and IL-18 did not improve anti-tumor activity beyond that observed with IL-18 alone, highlighting the importance of context in the biological actions of these cytokines [37]. The combination of IL-12 and IL-18 has also proven to be highly toxic in some studies [38].

IL-18 has several mechanisms of anti-tumor action
The aforementioned pre-clinical studies have provided useful insights into mechanisms by which IL-18 can enhance tumor control. A key effect is autocrine stimulation via binding to T cell-associated IL-18R, amplifying the production of IFN-γ. In keeping with this, engineering of CAR T cells to produce membrane-bound IL-18 allowed engagement of IL-18R in cis and enhanced in vitro anti-tumor activity [49]. Single cell RNA sequencing analysis revealed that IL-18 armoring was linked to enhanced NF-κB signaling and gene expression associated with the cell cycle, T cell activation, interferon stimulation and antigen-presentation [36]. Further confirming the importance of autocrine stimulation, genetic knock out of IL-18R in IL-18 producing CAR or TCR-engineered T cells decreased efficacy in several models [32–34].
By binding IL-18R in cis, IL-18-armoring supports the proliferation of CD4+ CAR T cells, which, in turn, enabled the expansion of CD8+ CAR T cells [32]. Indeed, expansion of CD8+ T cells, both adoptively transferred and of host origin, has been a widely reported action of IL-18 by several groups [32–35, 38, 39]. Within the CD8+ T cell compartment, IL-18 has been reported to promote a CCR7- effector memory and T-bet+ FoxO1low terminal effector phenotype [32, 37]. In agreement, a recent clinical showed that most IL-18 armoring of anti-CD327 CAR T cells resulted in an amplification of CD8+ effector T cells [50]. Additionally, RNAseq data indicate that IL-18 armoring downregulates naïve T cell markers (CD27, CD127, CD62L) [36]. However, three other preclinical studies showed that IL-18 instead promoted the CD8+ CCR7+ CD62L+ central memory phenotype [33, 35, 46]. These discrepancies may be context-dependent, relating for example, to the specific CAR T cell and tumor model under study.
Although IL-18 drives T cell activation and differentiation, it has also been shown to decrease exhaustion [33, 35, 44], manifested also as a reduction in PD-1, TIM-3 and LAG-3 triple positive T cells [34]. Metabolic impact of IL-18 armoring was shown most convincingly in CAR-expressing γδ T cells, indicated by increased mitochondrial mass accompanied by upregulation of both the glucose transporter, GLUT1 and amino acid transporter, CD98 [45].
Importantly, effects of IL-18 armoring are not limited to autocrine actions. Avanzi et al. found that IL-18 secreted by CAR T cells also improved the anti-cancer activity of endogenous host T cells. Thus, IL-18-armored anti-CD19 CAR T cells increased the survival of mice engrafted with a mixture of CD19+ and CD19NEG EL4 tumors. Splenocytes isolated from these mice that lacked CAR expression had an increased cytolytic and IFN-γ producing capacity when co-cultured with CD19NEG tumors, unlike control splenocytes from mice treated with non-armored CAR T cells [33]. Nonetheless, when tested in the pmel-1 TCR transgenic model, impact on anti-tumor efficacy was less prominent if IL-18R expression was abrogated in host (rather than CAR T) cells only [34].
Interleukin 18 also has multiple effects on the myeloid compartment, albeit variable across different models. Several studies have reported macrophage re-polarization from an anti-inflammatory M2 (CD206+/MHC-IIlo) to a pro-inflammatory M1 (MHCII+) phenotype [33–37, 45]. Additionally, the frequency of splenic and intra-tumoral DCs were increased, accompanied by a more mature and activated phenotype (CD11c+MHC-II+) [33–36]. Armoring with IL-18 has also been shown to reduce immunosuppressive intratumoral M2 macrophages alone [37], or in addition to both monocytic (CD11b+, Ly6C+) and granulocytic (CD11b+, Ly6G+) MDSCs [34]. Notably however, Kunert et al. found little impact of IL-18 on tumor-infiltrating myeloid cell numbers [38] while Jaspers et al. found that IL-18 upregulated PD-L1 on F4/80+ macrophages and DCs [35], an undesirable effect linked to IFN-γ production [51]. In line with this, combination therapy with IL-18-armored CAR T cells and a PD-L1 blocking antibody led to improved anti-tumor activity [35].
In addition to the autocrine and paracrine mechanisms described above, Huang et al. presented evidence that IL-18 may also act via an IL-18R independent mechanism [40]. Accordingly, when IL-18Rα was knocked out in both the host and the infused CAR T cells, recombinant IL-18 could still improve anti-tumor activity. Since IL-18Rβ cannot bind IL-18 with meaningful affinity alone, authors speculated that an additional unknown receptor may also be operative.
Finally, it should be noted that in some circumstances, dysregulated IL-18 activity has been linked with undesirable tumor-promoting effects [52–54]. However, this may reflect the actions of chronic low-level inflammation driven by this cytokine, contrasting with its effects when released acutely and at high-level by an IL-18 armored T cell.

Toxicity associated with IL-18
Armoring with IL-18 is generally considered to be safer than with IL-12. Illustrating this, Drakes et al., showed that IL-12- but not IL-18-armored T cells caused fatal toxicity in sublethally irradiated mice [34]. Moreover, IL-12 has proven highly toxic in man when administered as a cytokine or in the context of armored TIL cells (Qi and Maher, manuscript under review), in contrast to the more modest side effects of IL-18 therapy [30, 31].
Nonetheless, there are a number of indicators to suggest that excessive IL-18 activity could also impose the risk of increased toxicity. First, several pre-clinical studies have demonstrated the ability of IL-18-armored T cells to induce severe and sometimes lethal toxicity [32, 36, 45]. Moreover, clinical evidence supports the important pro-inflammatory role of IL-18. Chronically elevated serum IL-18 is associated with pro-inflammatory diseases, such as hemophagocytic lymphohistiocytosis/macrophage activation syndrome (HLH/MAS) [55, 56]. These “IL-18opathies” are also characterised by CD8+ T cell expansion and macrophage hyperactivation [16, 57]. Driorio et al. found that patients experiencing CD19 CAR T cell-induced severe cytokine release syndrome (CRS) had a similar serum proteomic signature to that of HLH patients [58]. Evidence has also been presented that IFN-γ signaling promotes CRS [59]. Moreover, IL-18 has emerged as a biomarker associated with immune effector cell-associated neurotoxicity syndrome (ICANS) [58]. Consequently, by raising serum IFN-γ levels, it is logical that IL-18 armoring could potentially contribute increase risk and severity of both CRS and ICANS.

Clinical experience with IL-18 armored T cells
Recently, the first clinical trial of IL-18-armored CD19 CAR T cells was reported in patients with B cell lymphoma (NCT04684563) [13]. Although 20 of 21 subjects had failed prior CAR T cell therapy, 11 achieved complete remission of disease by 3 months with a further 6 partial responses noted, giving a median duration of response of 9.6 months at median follow up of 17.5 months. Responses appeared to be more frequent if prior CAR T cell therapy had been with a CD28- rather than 4-1BB-containing product. These impressive data may also have been contributed to by the use of a shortened (3 days) manufacturing process, known to enhance T cell fitness. Authors presented evidence that active IL-18 was buffered effectively by IL-18BP in treated patients, mitigating risk of excessive toxicity. Cytokine release syndrome occurred in 13 subjects of which 3 reached grade 3 (correlated with higher CAR T-cell expansion), while neurotoxicity occurred in 3 patients (all grade 1–2) and there were no cases of hemophagocytic syndrome. While this would generally be considered an acceptable safety profile, it should be noted however that one subject developed tocilizumab/corticosteroid-refractory CRS and was ultimately treated with IL-18 binding protein. Elsewhere it is reported that one (perhaps the same) subject developed transient pulmonary edema in the context of grade 3 CRS, which was deemed a dose-limiting toxicity (DLT) that required expansion of the 3 × 107 cell dose level to 6 subjects.
In a second clinical trial, 5 acute myeloid leukemia patients were treated with IL-18-armored anti-CD371 CAR T cells (NCT06017258) [50]. Three achieved minimal residual disease negative disease status, also confirming the therapeutic activity of this experimental approach. All 5 treated patients developed CRS. Onset of symptoms correlated with a peak in serum IL-18 and IFN-γ levels, as well as NK cell expansion and activation suggesting that biologically active IL-18 was present. Both patients who received the highest planned dose of 3 × 105 cells/kg experienced DLTs, namely, prolonged cytopenias and grade 4 CRS respectively. The latter DLT was resistant to two doses of tocilizumab and ultimately was successfully treated with the IFN-γ-blocking antibody, emapalumab. In short, these clinical data provide strong clinical support for the ability of IL-18 CAR armoring to boost efficacy, but highlight the fact that this approach may accentuate the risk of inflammatory toxicity in some cases.

Strategies to mitigate IL-18 mediated toxicity
Given the aforementioned considerations, efforts have been made to restrict the functional impact of IL-18 armoring systems to the TME. Conceptually, this is particularly appropriate for solid tumors given their propensity to originate from and metastasize to parenchymal organs - meaning that effective technologies would minimize unwanted IL-18 activity in the circulation. One commonly used system entails placing the IL-18 cDNA under the transcriptional control of a Nuclear Factor of Activated T cells (NFAT)-based promoter. Since NFAT upregulation is coupled to activation of CAR by its target antigen, IL-18 is preferentially produced in the TME [60, 61]. Chmielewski et al. demonstrated the safety and efficacy of T cells armored with IL-18 under the control of NFAT/IL2 minimal promoter [37]. Using this dual vector approach, they observed no toxicity and no increase in serum IFN-γ. However, serum levels of IL-18, IL-6 and IL-27 were elevated indicating that this system may not have been completely stringent. In keeping with this, NFAT-regulated IL12 constructs proved lethal in mice [62] and also caused toxicity in clinical trials [63], likely due to non-specific upregulation of NFAT by signals not related to binding of target antigen. Providing reassurance, the same system was used to express IL-18 in TCR-engineered T cells without evident toxicity or detectable levels of circulating IFN-γ or tumor necrosis factor (TNF) α [38]. This suggests that a degree of leakiness of the system may be tolerable for IL-18 armoring in light of the lower toxicity seen with this cytokine compared to IL-12. More recently, a similar NFAT plus synthetic TATA box regulated IL-18 expression system has been incorporated into a single lentiviral vector system together with a GD2 specific CAR [39]. When tested in preclinical xenograft model of GD2-expressing malignancy, superior anti-tumor efficacy was once again demonstrated. This system has now been advanced to early phase clinical testing in patients with GD2-expressing malignancies (EU CT 2022– 501725–21–00). Finally, Hu et al. have also independently described an NFAT-regulated IL-18 armoring technology as a device to improve the safety of this approach [32].
As an alternative approach, Justicia-Lirio et al. developed a doxycycline-inducible IL-18 technology [46]. Anti-CD19 CAR T cells armored with this system showed excellent safety and efficacy in xenograft mouse models. This strategy was also used to improve safety of IL-12 armoring, providing a testament to its stringency [62].
A distinct strategy to restrict IL-18 activity to the TME was developed by Hull et al. [45] Since T cells lack caspase 1 activity, the caspase 1 proteolytic cleavage site in pro-IL18 was modified to one favored by granzyme B (GzB-IL18). The resulting GzB-IL18 propeptide is constitutively released in an inactive state by the armored CAR T cells. However, it selectively acquires biological activity when CAR T cell degranulation occurs, owing to co-localization with released granzyme B. In a syngeneic mouse model of head and neck squamous cell carcinoma, mice treated with Gzb-IL18-armored panErbB-specific CAR T cells had an execllent safety profile and improved survival due to enhanced tumor control. In contrast, armoring of panErbB CAR T cells with constitutively active IL-18 was lethal in this model. Toxicity was associated with increased serum levels of several cytokines, including IFN-γ.

Discussion
IL-18 armoring improves the efficacy of CAR and TCR-engineered T cell therapies, as shown in numerous pre-clinical studies. Moreover, recent clinical experience supports the utility of this approach in hematological malignancies. However, these studies also suggest the potential for uncontrolled IL-18 activity to aggravate CAR T cell-mediated toxicities such as CRS and ICANs, especially if coupled to a CAR that already has significant toxic potential. This provides a strong rationale for the use of engineering strategies that can improve safety of IL-18 armoring, while preserving or even improving its efficacy. Three such strategies have been discussed here, namely,: NFAT- or doxycycline-controlled IL-18 transcription or modification of the cleavage site within pro-IL-18 to one favored by granzyme B. The first of these technologies, NFAT-inducible IL-18, is already undergoing clinical evaluation. Interleukin 18-armored CAR T cells are currently being tested in a small number of clinical trials (Table 2). Ultimately, only the results of these and additional studies will convincingly show whether IL-18 significantly contributes to anti-tumor efficacy without causing toxicities beyond tolerable limit. In the meantime, additional pre-clinical studies will provide further insights into the various mechanisms by which secreted IL-18 boosts the anti-tumor activity of both engineered T cells and the endogenous immune system.