Single-day non-activated IL-18-armed CAR T-cells establish a durable, stem-like state with enhanced persistence

Abstract

Chimeric antigen receptor (CAR) T-cell therapies have transformed the treatment of B-cell malignancies, yet challenges including manufacturing delays, T-cell exhaustion, and limited persistence impede broader clinical success. Here, we report the single day production of non-activated CAR T-cells engineered to secrete interleukin-18 (IL-18), a pro-inflammatory cytokine that enhances T-cell function. These non-activated CART-IL18 cells exhibit robust anti-tumor efficacy across xenograft models of lymphoma, leukemia, and pancreatic cancer. IL-18 expression enhances the functional advantages of naïve-like non-activated CAR T-cells, resulting in improved persistence, metabolic fitness, and resistance to exhaustion. Single-cell transcriptomic analysis revealed upregulation of IL7R, KLF2, and MCL1, alongside suppression of inhibitory checkpoint genes such as PDCD1, TOX, and HAVCR2. Metabolomic profiling demonstrated enhanced mitochondrial bioenergetics, with increased spare respiratory capacity and accumulation of α-ketoglutarate, malate, and spermine. Functional in vitro and in vivo profiling demonstrated enhanced per-cell cytotoxicity and in vivo durability. We complemented these studies with single-cell transcriptomic and metabolomic analyses to define CAR T-cell biological states beyond what is captured by xenograft tumor clearance. This IL-18-enhanced, activation-free CAR T product offers a clinically actionable platform with the potential to reduce vein-to-vein time while improving product potency and persistence, providing a rationale for clinical testing in patients with tumors refractory to standard CAR T.

Introduction

Chimeric antigen receptor (CAR) T-cells are effective ‘living drugs’ for relapsed or refractory (R/R) B-cell malignancies, yet many patients ultimately relapse, underscoring the need for improved therapies. ^{1, 2} Strategies to enhance CAR T-cell persistence and function are critical, especially in the R/R setting where T-cell exhaustion is common.^{1, 2} Preclinical studies show that CAR T efficacy can be boosted by co-expressing stimulatory cytokines or by optimizing manufacturing to preserve less-differentiated T-cell states, but few of these approaches have been translated into the post-CAR T failure setting.^3–7

One promising strategy involves engineering CAR T-cells to secrete interleukin-18 (IL-18), a pro-inflammatory cytokine that promotes T-cell persistence and anti-tumor immunity.^8–12 In a first-in-human study, IL-18–secreting anti-CD19 CAR T-cells (CART19-IL18) achieved an 81% overall response rate in patients with R/R B-cell lymphoma, most of whom had progressed after conventional CAR T therapy.¹³ The median progression-free survival (PFS) was 8.7 months, with durable complete remissions observed in several patients beyond 20 months.¹³ Notably, this trial used a rapid 3-day manufacturing process that relies on TCR-mediated activation; however, the median vein-to-vein time (V2VT) remained prolonged at 67 days (range: 26–137), due to both manufacturing limitations and protocol-driven safety staggers.¹³

Shorter V2VT has been associated with improved response rates and survival. In relapsed/refractory large B cell lymphoma (LBCL), reducing V2VT from 54 days to 24 days led to an estimated 3.2-year gain in life expectancy.¹⁴ Nonetheless, prolonged V2VT (>40 days) remains common in both CAR T trials and in routine clinical practice,¹⁵ representing a substantial barrier to timely and effective CAR T therapy.

We previously developed a CAR T protocol that enable lentiviral transduction of T-cells in the absence of TCR activation, producing highly functional CAR T-cells within 24 hours of leukapheresis.¹⁶ By avoiding ex vivo TCR-mediated activation, this approach mitigates the progressive differentiation and exhaustion programs induced by TCR signaling, leading to more durable anti-leukemic activity in mice.¹⁶ Here, we evaluate whether single-day, non-activated manufacturing defines a distinct CAR T-cell biological state and whether this state can be effectively combined with IL-18 armoring across hematologic and solid tumor contexts. The present study defines the transcriptional, metabolic, and functional phenotypes of non-activated CAR T-cells. This supports the clinical development of a potent CAR T therapy in which CAR signaling occurs in the absence of prior TCR activation, enabling rapid manufacture while preserving favorable T-cell biology, for patients with advanced disease refractory to conventional CAR T therapy.

Methods

Lentiviral Vectors

Replication-defective lentivirus was produced using a third-generation system. Transfer plasmids encoded an anti-CD19-BBζ CAR,¹⁷ anti-CD19-huIL18 CAR,¹³ anti-mesothelin CAR, or CARTmeso-huIL18 CAR. The CARs contained an scFv targeting CD19 (FMC63 clone)¹⁸ or human mesothelin (M5 clone), linked to CD8 hinge and transmembrane domains, and 4–1BB and CD3ζ signaling domains. The cytokine and CAR transgenes were expressed from a single transcript using a T2A element under the control of an EF-1α promoter. Vectors were produced by transient transfection of HEK293T cells.

T-cell Isolation, Transduction, and Cell Culture

Peripheral blood leukocytes were obtained from healthy donors with informed consent under IRB-approved protocols at the University of Pennsylvania. Healthy donor T-cells were purified by negative selection. Patient-derived T-cells were isolated by CD4/CD8 positive selection. For activated CAR T-cells, enriched T-cells (1×10⁶/mL) were cultured in X-VIVO 15 with 5% human AB serum (Valley Biomedical, Winchester, VA), L-glutamine (Cambrex), HEPES (Cambrex), IL-7 (10 ng/mL, Miltenyi Biotec), and IL-15 (10 ng/mL, Miltenyi Biotec). Cells were activated with anti-CD3/CD28 beads (3:1 ratio) for 24 hours before lentiviral transduction (MOI=3) and cultured for 1 to 3 days, then cryopreserved until use.¹⁹ For non-activated CAR T-cells, T-cells (1×10⁷/mL) were resuspended in supplemented X-VIVO 15 with IL-7/IL-15 and transduced with lentivirus (MOI=3–5). Cells were harvested within 24–36 hours for infusion. All cell lines (NALM-6, SY5Y, Jeko1, AsPC1 and HEK293T) were obtained from ATCC, maintained in standard conditions, and authenticated by short tandem repeat profiling.

Flow cytometry

T-cell differentiation was assessed using the following antibodies: anti-CCR7–FITC (150503, BD Pharmingen); anti-CD45RO–PE (UCHL1), anti-CD8–H7APC (SK1, BD Biosciences); anti-CD4–BV510 (OKT4), anti-CD3–BV605 (OKT3), anti-PD1–PerCP (EH12.2H7), anti-Lag3–BV711 (11C3C65) and anti-Tim3–PeCy5 (F38–2E2). The anti-CAR19 idiotype was provided by Novartis (Basel, Switzerland). Flow cytometry was performed on BD LSR Fortessa. Analysis was performed using Flowjo (Tree Star Inc. version 10.1).

Single-cell cytokine release assay

Resting T-cells were transduced with lentivirus and cultured with IL-7 and IL-15 for 5 days. Non-activated CAR T-cells were then stimulated with CD19- coated beads (Acrobiosystems) at the ratio of 1:2 for 16 h. Cytokine production was analyzed using the IsoSpark Duo platform.

Cytotoxicity assays

CAR T-cell killing was measured using a real-time impedance-based assay (Maestro TrayZ, Axion Biosystems). CD19-expressing SY5Y target cells (1×10⁴/well) were plated, and CAR T-cells were added in at indicated effector-to-target ratios. Cell index (relative cell impedance) was monitored hourly. Data were normalized to the maximum total integrated intensity or cell index value following effector-cell plating. Shaded lines reflect the mean of replicate wells ± SD.

Single-cell RNA sequencing and analysis

Briefly, human T-cells were isolated from mouse peripheral blood by CD4/CD8 MicroBead (Miltenyi Biotec) positive selection, washed with PBS containing 0.04% BSA, and filtered through a 40 μm cell strainer to obtain a single-cell suspension. Recovered CD4/8⁺ T-cells were analyzed as a bulk population. ScRNA-seq libraries were generated using a Chromium Single-Cell 3’ Library and Gel Bead Kit (10x Genomics, v3). Data were processed using the Scanpy (v1.11.0) and anndata (v0.11.4) packages in Python.²⁰ Cells were filtered for quality based on thresholds for gene expression and total counts, normalized to total transcript counts of 10,000 per cell, and analyzed using UMAP and Leiden clustering. T-cell subsets were identified by canonical markers, and GSEA was performed using gseapy (v1.1.8) and the GO_Biological_Process_2023 gene ontology database.²¹

Liquid Chromatography-High Resolution Mass Spectrometry

Serum metabolites were measured by liquid chromatography-high resolution mass spectrometry adapted from previously published approaches.²² Metabolomic analyses were performed at matched tumor volumes to minimize confounding effects of tumor burden. Samples were quenched with pre-chilled −80°C 80:20 methanol:water, vortexed, incubated at −80°C, centrifuged, and the supernatants were evaporated to dryness under nitrogen. Dried extract were reconstituted and analyzed on a ZIC-pHILIC column using a Vanquish Duo UHPLC System (Thermo Fisher Scientific), coupled to a Q Exactive Plus mass spectrometer operating in polarity-switching mode (70–1000 m/z). Data were acquired using XCalibur (Thermo Fisher Scientific, version 4.1) and analyzed on Tracefinder (Thermo Fisher Scientific, version 5.1) using a 5ppm window from the predominant M-H negative ion. For isotope tracing, isotopologue enrichment was calculated using FluxFix.

In vivo models

All animal experiments were approved by the University of Pennsylvania IACUC. Six- to 8-week-old NOD-SCID γc−/− (NSG) mice were used. For leukemia/lymphoma models, mice were injected IV with 1×10⁶ NALM6 or JeKo-1 cells expressing click beetle green luciferase and eGFP, in 0.1 mL sterile PBS.^{17, 23} CAR T-cells were injected IV 4–5 days later at indicated doses.

Activated CAR T-cell groups were dosed based on CAR⁺ cell number. Non-activated CAR T-cell groups were dosed based on total T-cell number to account for delayed expression and with an estimated ~8% CAR transduction efficiency based on prior validation.

Tumor progression was monitored each week via bioluminescence using a Xenogen IVIS Spectrum system (Caliper Life Science). Total flux was quantified using Living Image 4.4 (PerkinElmer). Concentrations of human cytokines in mouse serum were measured by a Luminex bead array platform (Life Technologies) according to the manufacturer’s instructions.¹

For the pancreatic cancer model, 2×10⁶ AsPC-1 cells were injected subcutaneously in the right flank. When tumors reached 300 mm³, mice were treated with CAR T-cells. Tumor progression was monitored by bioluminescence imaging or caliper measurement. Tumor volumes were calculated as V = (L * Ŵ2)/2, where L is length (longest axis), and W is width (shortest axis). T-cell engraftment was monitored by flow cytometry of peripheral blood, stained using anti-human CD45, CD4, CD8, CCR7 and CD45RO antibody in Trucount tubes (BD Biosciences).

Quantification and statistical analysis

Analyses were performed using GraphPad Prism (version 10). Comparisons of more than two groups used one-way ANOVA with Tukey’s post-hoc test. Error bars represent ± SEM unless stated otherwise.

Results

Non-activated IL-18-expressing CAR T-cells control lymphoma and leukemia xenografts

We first examined a non-activated, single-day manufacturing platform for CART19-IL18 cells in a JeKo-1 lymphoma model (Fig. 1A-B). NSG mice with JeKo-1 tumors were treated with activated (Day 3, D3, 3.3e5 CAR+ cells) or non-activated (Day 1, D1, 2e6 total T-cells) CART19 cells, with or without IL-18 expression (Fig. 1C). By day 24, mice treated with non-activated CART19 cells showed lower tumor bioluminescence imaging (BLI) compared to activated cells (Fig 1D-E). While adding IL-18 to activated CART19 cells enhanced tumor control, its effect in non-activated CART-cells was to reduce response variability and the incidence of late tumor progression (Fig 1D-E). Notably, only the non-activated CART19-IL18 product significantly reduced tumor burden relative to baseline (Fig. 1F-G).

Figure 1. IL-18 expression enhances CART function in a JeKo-1 xenograft model.

A) Lentiviral vectors encoding anti-CD19 CAR and co-expressing human IL-18 or GFP control. B) Schematic of CAR T-cell manufacturing illustrating the transduction of healthy human T-cells with CART19 lentivirus either without prior activation (D1) or following CD3/CD28 activation and a three-day expansion (D3). C) NSG mice were engrafted with 1×10⁶ JeKo-1 lymphoma cells and treated five days later with either activated (D3, 3.3e5 CAR+ cells) or non-activated (D1, 2e6 total T-cells) CART19 cells, with IL-18 expression or GFP control; (n=8 mice per group). Tumor burden was assessed over time using BLI. D–E) Comparison of JeKo-1 tumor BLI between non-activated (D1) CART19 and activated (D3) cells on day 24 and day 34. Data are mean ± SEM. Statistical significance assessed by one-way ANOVA with multiple comparisons or paired tests as indicated. *p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant. F–G) Estimation plots for t-tests comparing tumor BLI at day 34 to baseline, showing reduced response variability in IL-18-expressing non-activated CART19 cells.

We next examined the anti-tumor function of non-activated CART19-IL18 cells in a Nalm6 xenograft model (Fig. 2A), using both patient (Fig. 2B) and healthy donor (Fig. 2E) leukapheresis products as starting material. NSG mice were engrafted with 1×10⁶ luciferase-expressing Nalm6 cells and treated with either non-activated (D1) or activated (D1) CAR19 T-cells, with or without IL18 expression. In mice treated with patient-derived cells, IL-18 expression significantly reduced tumor BLI by day 25 compared to the GFP control (Fig. 2B) and significantly increased the concentration of circulating human T-cells, a measure of persistence (Fig. 2C), without altering the CD4:CD8 ratio (Fig. 2D).

Figure 2. IL-18 expression enhances non-activated CART persistence and anti-tumor function in a Nalm6 xenograft model.

A) Schematic of in vivo experiment: NSG mice were engrafted with 1×10⁶ luciferase-expressing Nalm6 leukemia cells and treated on day 5 with patient- or healthy donor-derived CART19 cells that were either activated or non-activated, with or without IL-18 expression. Tumor burden was assessed by serial bioluminescence imaging (BLI). B) Tumor BLI in mice treated with non-activated CART19 cells (D1, 2e6 total T-cells), with or without IL-18 expression, derived from patient lymphocytes; (n=7 mice per group). C-D) Peripheral blood T-cell concentrations (C) and CD4:CD8 ratios (D) at day 25 with non-activated (D1) CART19 cells with or without IL-18 expression. E) Tumor BLI in mice treated with healthy donor-derived CART19 cells (non-activated D1: 2e6 total T-cells; activated D1: 2e6 CAR⁺ T-cells); (n=5 mice per group). F) IL-18-expressing CART19 cells exhibited improved persistence in peripheral blood. G) The TCM:TEM in non-activated compared to activated CART19 products. H–J) Blood IL-18 and IL-10 concentrations in mice treated with non-activated or activated IL-18 CART19. Data are mean ± SEM. For analyses of more than two groups, statistical significance was assessed by one-way ANOVA with post-hoc t-tests and Tukey’s correction for multiple comparisons. For comparisons of two groups, simple t-tests were performed (*p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant). Error bars depict mean ± SEM.

Using healthy donor T-cells, we compared non-activated (D1) and activated (D1) products to isolate the impact of IL-18. IL-18 expression again improved tumor control with non-activated CART19 cells (Fig. 2E), while there was no significant difference in tumor control between activated and non-activated CART19-IL18 cells (Fig. 2E). IL-18 expression also increased T-cell persistence (Fig. 2F). Phenotypically, non-activated products had a higher ratio of central memory (TCM) to effector memory (TEM) cells (Fig. 2G). Peripheral blood analysis showed higher IL-18 and lower IL-10 concentrations in mice receiving non-activated cells (Fig. 2H-J).

Non-activated CART19-IL18 cells have distinct transcriptional and metabolic profiles

Having established robust antitumor efficacy, we next examined whether TCR activation during ex vivo manufacturing imprints durable transcriptional and metabolic features not captured by tumor burden alone. To define the metabolic and transcriptional correlates of non-activated CART19-IL18 functionality, we performed single-cell RNA sequencing (scRNA-seq) and serum metabolomics on samples collected at day 38 from the Nalm6 model (Fig. S1), a timepoint characterized by effective tumor control following extensive antigen-driven expansion.

ScRNA-seq revealed that T-cells recovered from mice treated with non-activated CART19-IL18 cells upregulated genes linked to persistence and survival, including IL7R, KLF2, and MCL1 (Fig. 3A-D).^24–26 The immune checkpoints PDCD1, HAVCR2, LAG3, and TIGIT had low normalized transcriptional expression in all conditions (Fig. 3C-D), consistent with the expected phenotype of recently activated memory T-cells.^{27, 28} Within the CD8⁺ population, non-activated cells had markedly reduced KLRG1, suggesting less terminal differentiation,²⁹ and increased expression of the trafficking regulator FUT7 and co-stimulatory receptor TNFRSF18 (Fig. 3B-C).³⁰ CD4⁺ non-activated cells showed increased expression of the co-stimulatory receptor CD27 and mitochondrial respiratory chain components MT-ND4L and MT-ATP8 (Fig. 3C-D).³¹ Gene set enrichment analysis (GSEA) confirmed that non-activated cells were enriched for pathways related to mitochondrial ATP synthesis, translation, telomere maintenance, biosynthesis, and survival, whereas activated cells were enriched for antigen presentation and T-cell activation pathways (Fig. 3E–F, Tables S1-2).

Figure 3. Non-activated CART19-IL18 cells display distinct metabolic signatures, enhanced biosynthetic activity, and reduced terminal differentiation.

A) UMAP plot from scRNA-seq experiment of peripheral blood T cells collected from mice treated with activated and non-activated CART19-IL18 cells. B) Normalized IL7R and KLRG1 expression superimposed on UMAP plots. C) Violin plots showing key gene expression markers in CD4⁺ and CD8⁺ T-cells from activated and non-activated CART19-IL18 groups. Transcriptomic analyses were performed on recovered human T-cells, irrespective of CAR expression. D) Heatmaps showing normalized gene expression of the top differentially expressed genes in CD4⁺ and CD8⁺ subsets across conditions. E–F) Gene set enrichment analysis of CD4⁺ (C) and CD8⁺ (D) CART19-IL18 cells from non-activated vs activated groups. Red bars indicate pathways enriched in non-activated cells; blue bars indicate those enriched in activated cells. G) Targeted serum metabolomics reveals higher α-ketoglutaric acid, malate, and spermine in non-activated groups. Samples collected at timepoints with comparable tumor burden between groups. Data shown as mean ± SEM. *p < 0.05 (simple unpaired t-test). H) Dot plot of normalized gene expression for key metabolic genes across CD4⁺ and CD8⁺ T-cell subsets in activated and non-activated conditions. Circle size reflects proportion of cells; color intensity reflects mean gene signature score.

Metabolomic profiling showed that mice treated with non-activated cells had elevated serum levels of α-ketoglutarate, malate, and spermine—metabolites linked to mitochondrial function and antioxidant activity (Fig. 3G).^32–35 To correlate these differences to cellular pathways, we examined normalized mRNA expression levels of a targeted panel of metabolic enzymes. Non-activated cells had increased spermidine/spermine N(1)-acetyltransferase 1 (SAT1), glutaminase (GLS), sodium-coupled neutral amino acid transporter 1 (SLC38A1), and cytoplasmic malate dehydrogenase (MDH1) transcript levels, suggesting an enhanced capacity to use glutamine to fuel oxidative phosphorylation (Fig. 3H).

Non-activated CAR T-cells preserve a naïve-like, exhaustion-resistant phenotype with high metabolic fitness

To further characterize the functional differences between activated and non-activated IL-18–expressing CAR T-cells, we performed in vitro metabolic, phenotypic, and cytolytic profiling. Seahorse metabolic flux analysis revealed that non-activated CAR T-cells had lower basal oxygen consumption rates (OCR) and proton efflux rates (PER), consistent with quiescence, but a significantly higher spare respiratory capacity (SRC), a key indicator of mitochondrial fitness and persistence (Fig. 4A–B).³⁶ Peak rates of OCR (O2 max) were higher in non-activated CAR T-cells, implying an enhanced ability to support bursts of proliferation and differentiation across groups (Fig. 4A–B). Glycolytic activity, as measured by extracellular acidification rate (ECAR), was also significantly lower in the non-activated group, both at baseline and following oligomycin treatment (Fig. 4B). A composite bioenergetic profile demonstrated that non-activated CAR T-cells relied predominantly on mitochondrial ATP production during stress responses, whereas activated CAR T-cells depended more on glycolysis (Fig. 4C).

Figure 4. In vitro profiling reveals that non-activated CART19-IL18 cells retain cytotoxic function and display improved metabolic fitness compared to activated counterparts.

A) Seahorse XF analysis of oxygen consumption rate (OCR) and proton efflux rate (PER) showing mitochondrial respiration and glycolysis in activated versus non-activated CART19-IL18 T-cells (n=3 healthy donors). B) Quantification of spare respiratory capacity (SRC), basal extracellular acidification rate (ECAR), and oligomycin-induced ECAR (SRC: n=5 technical replicates; ECAR: n=3 technical replicates per condition, from n=3 healthy donors). C) Bioenergetic profile showing relative contributions of mitochondrial and glycolytic ATP production under basal and stressed conditions in both T-cell populations (n=3 healthy donors). D) Representative flow cytometry gating strategy for assessing T-cell differentiation (CD45RO, CCR7) and exhaustion phenotypes (PD1, LAG3, TIM3) in activated and non-activated CART19-IL18 T-cells (n=3 healthy donors). E) Real-time cytotoxicity assay using impedance measurements to assess tumor killing of SY5Y-19 neuroblastoma cells by non-transduced T-cells, activated, and non-activated CART19-IL18 T-cells (n=3 healthy donors) at 2:1 E:T ratio. F) Quantification of differentiation states (Naïve, TCM, TEFF, TEM) 72 hours after antigen exposure (n=3 healthy donors). The dotted line indicates the introduction of effector cells G) Quantification of exhaustion phenotype expression (PD1, LAG3, TIM3) before and after antigen exposure (n=3 healthy donors). Data are shown stratified by CAR⁺ and CAR⁻ populations. Data are mean ± SEM. For analyses of more than two groups, statistical significance was assessed by one-way ANOVA with post-hoc t-tests and Tukey’s correction for multiple comparisons. For comparisons of two groups, simple t-tests were performed (*p < 0.05, **p < 0.01, ***p < 0.001). Error bars depict mean ± SEM.

Phenotypic analysis by flow cytometry revealed that non-activated CART19-IL18 cells retained a significantly higher proportion of naïve T-cells compared to their activated counterparts, which were predominantly TEM and TCM cells (Fig. 4D, F and S2). This less-differentiated, naïve-like state persisted even after 3 days of continuous antigen stimulation (Fig. 4F). Expression of exhaustion markers PD-1 and TIM-3 was also significantly lower in the non-activated population (Fig. 4D, G and S2).

The presence of polyfunctional T-cells (those that produce ≥2 cytokines upon stimulation with antigen in vitro) is associated with higher clinical response rates.³⁷ Based on single-cell multiplex cytokine profiling, IL-18-expressing non-activated CAR T-cells had a 2.03-fold increase in polyfunctional strength index (PSI) compared to CAR T-cells expressing GFP as a control (Fig. S3A). This increased PSI was due primarily to increased expression of effector cytokines (especially IL-12 and IL-2) and chemokines (Fig. S3 A-E).

To determine whether these differences translated into functional potency, we performed real-time impedance-based cytotoxicity assays using SY5Y neuroblastoma target cells expressing CD19 (SY5Y-19). Despite their naïve-like phenotype, non-activated cells displayed short-term cytolytic activity comparable to activated cells over a 48-hour period (Fig. 4E). Importantly, IL-18 alone did not induce cytotoxicity, and its effects were observed only in the context of CAR-mediated antigen recognition, indicating augmentation of CAR-dependent effector function rather than antigen-independent activity (Fig. 4E and Fig. S4). These findings indicate that non-activated CAR T-cells expressing IL-18 preserve a potent short-term functional killing capacity while retaining a metabolically and phenotypically naïve-like state that may promote their observed long-term persistence and in vivo efficacy.

IL-18 enhances non-activated CAR T efficacy in a pancreatic cancer xenograft model

To extend our findings to solid tumors, we tested anti-mesothelin (M5 scFv) CAR T-cells (CARTmeso) in a pancreatic cancer (AsPC-1) xenograft model (Fig. 5A-B). We compared multiple doses of non-activated CAR T-cells to activated cells (D3), both with and without IL-18 expression (Fig. 5A-B). These panels represent the same experiment but separated for visual clarity due to the number of treatment groups. By day 27, the IL-18-expressing non-activated CAR T-cells led to the highest blood concentration of human CD45+ cells (dose of 1×10⁶ Fig. 5C). The non-activated CARTmeso-IL18 cells also significantly reduced tumor volume compared to non-transduced control T-cells (Fig. 5D). The mean tumor volume was not significantly different between the non-activated CARTmeso-IL18 and the most comparable dose of activated CARTmeso-IL18 cells (1×10⁵ CAR+ cells, Fig. 5D). Importantly, because non-activated products were dosed based on total T-cell number, the estimated CAR+ dose in these groups is substantially lower than the activated CAR+ based dosing groups, representing a conservative comparison that is consistent with similar per-cell anti-tumor potency.

Figure 5. IL-18 enhances CART persistence and regulates exhaustion, activation, and metabolic function in a pancreatic cancer xenograft model.

A–B) NSG mice were subcutaneously injected with 2×10⁶ AsPC-1 pancreatic tumor cells. After 3 weeks, mice were treated with healthy donor-derived, non-activated or activated anti-mesothelin CAR T-cells (CARTmeso) with or without IL-18 expression, at the indicated doses. Tumor volumes were monitored over time; (n=5 mice per group). Panels A and B represent groups from the same experiment and are displayed separately for visualization clarity. C) Blood concentrations of human CD45⁺ cells at day 27. D) Tumor volumes at day 33 for each treatment condition. confirmed that IL-18-expressing, non-activated CAR T-cells significantly reduced tumor burden relative to non-transduced (NTD) controls. E) UMAP plot of CD4+ and CD8+ subsets in scRNA-seq of peripheral blood T cells collected from mice treated with IL-18-secreting or control non-activated CARTmeso cells. F) Normalized expression of IL7R and TOX superimposed on UMAP plots. G–I) Differential gene expression analysis in CD8+ and CD4+ non-activated T-cell subsets comparing IL-18 to control groups. J–K) Gene set enrichment analysis (GSEA) in CD8+ and CD4+ subsets revealed IL-18-induced enrichment of pathways related to protein translation, biosynthesis, and cellular respiration. CD8+ cells with IL-18 also upregulated gene sets related to telomerase activity and DNA repair. Control (without IL-18 expression) groups showed enrichment of MAPK/PI3K signaling, IL-8 signaling, and immune activation pathways. Transcriptomic analyses were performed on recovered human T-cells, irrespective of CAR expression. Statistical significance assessed by one-way ANOVA and post-hoc t-tests with Tukey’s correction for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Single-cell RNA-sequencing (scRNA-seq) of peripheral blood T-cells at day 90 (Fig. S5) showed that, in both CD4+ and CD8+ non-activated subsets, IL-18 was associated with increased expression of the interleukin-7 receptor (IL7R) and KLF2, a marker of T-cell quiescence, and reduced expression of the exhaustion markers PDCD1, HAVCR2 (TIM-3), TOX, and TIGIT (Fig. 5E-G). For CD4+ cells, IL-18 also increased expression of LEF1 and the anti-apoptotic MCL1 (Fig. 5H). IL-18 was also associated with reduced levels of BTLA, ENTPD1, and SLAMF6 (exhaustion markers) in CD4+ cells.

Differential gene expression analysis for the CD8+ population confirmed that IL7R was among the top up-regulated genes in IL-18-expressing cells (Fig. 5I). TXNIP (thioredoxin-interacting protein) was the most upregulated gene in this population, with known functions in restraining late T-cell expansion and promoting memory T-cell differentiation (Fig. 5I).^{38, 39} Multiple regulators of T-cell activation, including TSC22D3, TGFBR3 and TNFAIP3, were also increased in the IL-18-expressing group.^40–42 In non-activated CARTmeso cells, IL-18 reduced CHI3L2, a factor that can induce T-cell apoptosis and exhaustion,⁴³ and reduced expression of the exhaustion markers HAVCR2, TOX, and TIGIT (Fig. 5I).

GSEA identified pathways associated with IL-18 transgenic expression in non-activated CAR T products (Fig. 5J-K, Tables S3-4). In both CD4+ and CD8+ subsets, IL-18 expression was associated with upregulation of pathways associated with protein translation and the macromolecule biosynthetic process (lead genes including EEF1A1, RPL13A, RPL3, RPL5, and RPL6, Fig. 5J-K). In CD8+ cells, IL-18 was associated with a gene signature associated with telomerase activity and DNA repair (HSP90AB1, HSP90AA1, MAP3K4, ATM, MAPK1, XRCC5, TNIP1, and CCT4, Fig. 5J-K). Conversely, CARTmeso cells without IL-18 expression were associated with increased PI3K and MAPK pathway activity, positive regulation of IL-8 signaling, and increased NK and T-cell activation signatures (Fig. 5J-K). Together, these findings demonstrate that IL-18 expression augments persistence and biosynthetic capacity, while retaining the naïve-like differentiation profile of non-activated CAR T-cells.

Discussion

In this study, we demonstrate that IL-18–expressing CAR T-cells manufactured in a single day without TCR activation represent a biologically distinct therapeutic state and maintain potent anti-tumor function against hematologic and solid tumors. By avoiding ex vivo activation, our method preserves a less-differentiated, metabolically fit T-cell phenotype characterized by durable engraftment, high bioenergetic reserve, and exhaustion resistance. Despite lower predicted CAR+ cell doses due to reduced lentiviral transduction efficiency,¹⁶ the non-activated CART-IL18 cells mediated robust tumor control, suggesting superior per-cell potency. This approach also offers a solution to reduce V2VT, potentially expediting the treatment of patients with advanced, unstable disease refractory to currently approved CAR T products.

Our findings build on a recent clinical trial where activated CART19-IL18 cells manufactured using a 3-day protocol were effective in patients who had relapsed after standard CAR T therapy.¹³ We show that IL-18 expression enhances the functional advantages conferred by the preserved naïve-like phenotype of non-activated CAR T-cells. This cooperative effect is reflected transcriptionally by the upregulation of genes critical for T-cell persistence and survival, IL7R, KLF2, and MCL1, along with the suppression of exhaustion-related genes like TOX and PDCD1. GSEA confirmed that these non-activated CART19-IL18 cells are programmed for longevity, with enriched pathways for protein synthesis, mitochondrial metabolism, and DNA repair.

Boosting T-cell metabolic fitness through cytokine engineering may be particularly beneficial in patients with impaired T-cell health due to age, malignancy, or prior treatment (including failure to respond to standard CAR T therapy).^44–49 IL-18 has multiple effects on tumor immunoreactivity, including recruiting T-cells, increasing the activation of native tumor-infiltrating lymphocytes, increasing CD206- M1 macrophages, and decreasing Tregs, suppressive CD103+ DCs, and M2 macrophages, and counteracting the inhibitory IL-18 binding protein (IL-18BP) that is expressed in many cancers.^{8, 12, 50, 51} This inherent flexibility may enable IL-18-expressing cells to sustain effector function under the stress of advanced tumor burdens and reduced T-cell functional competence from prior treatment.

Metabolically, non-activated CART-IL18 cells displayed increased SRC, an important correlate of T-cell longevity and functionality.³⁶ This was accompanied by elevated levels of glutamine-derived intermediates such as α-ketoglutarate and malate, as well as polyamines like spermine—metabolites that support oxidative phosphorylation, biosynthesis, and redox homeostasis.^52–54 Their abundance in serum, coupled with increased expression of associated metabolic enzymes (e.g., SAT1, GLS, SLC38A1), highlights the flexibility in metabolic fuel usage in non-activated CAR T-cells, in part by using amino acids such as glutamine as metabolic substrates. Strategies to enhance metabolic fitness and optimize differentiation status prior to adoptive transfer have previously been limited to activated T-cells and their progeny,^55–58 but our results suggest that limiting ex vivo activation may have beneficial effects on CAR T-cell mitochondrial function. Compared to their activated counterparts, non-activated cells retained a less differentiated, naïve-like phenotype, even following antigen stimulation. As a conceptual model, we suggest that prolonged ex vivo T-cell stimulation partially ‘wastes’ the therapeutic potential and metabolic reserves of the activated cellular products, because the T-cell uses its initial, developmentally important cytotoxic burst against the activating substrate instead of the tumor. In contrast, the non-activated platform allows the full hierarchy of differentiated T-cell states to engage with the tumor. Treatment with IL-18 is complementary with non-activated CAR T-cells, because it substantially enhances the persistence and in vivo responses of these cells while preserving their naïve-like differentiation. Despite their less differentiated state, the short term cytolytic function in IL-18-expressing non-activated CAR T-cells remained intact, suggesting that effector differentiation is not required for initial rapid tumor control. Functionally, despite reduced CAR transduction efficiency and lower absolute numbers of CAR⁺ cells, non-activated IL-18–expressing CAR T-cells achieved comparable or superior tumor control relative to their activated counterparts across multiple xenograft models, including B-cell lymphoma, leukemia, and pancreatic adenocarcinoma. This suggests enhanced per-cell potency, likely attributable to preserved stemness, reduced checkpoint expression, and efficient metabolic programming. Notably, non-transduced T-cells did not mediate measurable tumor control in any in vivo or in vitro assay presented here, arguing against a target-independent alloresponse contributing to the observed effects. In addition, the ability to manufacture and infuse CAR T-cells within 24h of leukapheresis opens the possibility of treating patients with rapidly progressive malignancies who may otherwise not survive conventional manufacturing delays. This fast-turnaround capability, which will require parallel advances in release testing and clinical workflows, may be particularly impactful in diseases such as high-grade lymphomas, where response kinetics are tightly linked to V2VT.¹⁴

Beyond biological advantages, the single day, non-activated platform provides meaningful practical and economic benefits. By eliminating the need for extended culture, the protocol reduces reagent usage, labor, and infrastructure demands, offering a leaner and more cost-effective manufacturing footprint. The shortened timeline also opens the possibility for fresh, same-day infusion protocols, potentially eliminating the need for cryopreservation and reducing release testing delays. While current FDA guidelines are optimized for cryopreserved products, fresh CAR T-cell infusion is being piloted at select sites. Broader adoption of fresh product workflows will require updated regulatory frameworks and manufacturing standards.

Limitations of this work include the need to estimate the CAR+ cell percentage in non-activated products at the time of infusion. Because non-activated T-cells exhibit delayed lentiviral post-entry kinetics, including reverse transcription and integration, detectable CAR protein expression is delayed.¹⁶ Consequently, non-activated products were dosed based on total T-cell numbers with an estimated 8–10% CAR expression rate,¹⁶ meaning the exact CAR+ dose is estimated rather than directly measured. Based on these efficiencies, the predicted CAR+ dose in non-activated groups was approximately 10-fold lower than in the activated groups, representing a conservative comparison. Additionally, some correlative analyses in this study, specifically the scRNA sequencing and serum metabolomics, were not restricted to the CAR+ T-cells. Transcriptomic analyses were performed on recovered human CD4+ and CD8+ T-cells irrespective of CAR expression to capture the integrated transcriptional state of the in vivo–expanded T-cell compartment. Therefore, these findings highlight broader transcriptional and metabolic differences in the T-cell population and overall immune milieu in response to the CAR T-cell therapy rather than representing isolated CAR+ cell biology.

Together, these findings define absence of TCR activation as a biologically meaningful determinant of CAR T-cell fate that can be productively integrated with cytokine armoring for translational application. Our data provides a mechanistic and functional definition of a distinct non-activated CAR T-cell product state and demonstrate the feasibility of IL-18 engineering within a single day platform to improve therapeutic access, manufacturing efficiency, and clinical relevance. Our data support further translational development of non-activated CART-IL18 cells, particularly in the post CAR T relapse setting.

Key Points.

• CART-IL18 cells manufactured within 24 hours without TCR activation retain a naïve-like, exhaustion-resistant phenotype with increased per-cell potency, with the potential to shorten vein-to-vein time.

• IL-18 expression enhances the biological advantages of the non-activated state, yielding improved T-cell persistence, metabolic flexibility, and functional durability.

Data Availability Statement

All data generated or analyzed during this study are included in this published article.