Systemic pre-conditioning favors effector over exhausted CD8 T-cell subsets following Sup2-IL33 armored CAR T-cell therapy

Abstract

Background

Systemic pre-conditioning prior to chimeric antigen receptor (CAR) T-cell infusion enhances engraftment and increases infiltration and activity of adoptively transferred cells within the tumor. Despite promising responses in some patients, additional strategies are needed to enhance overall efficacy of CAR T cells in patients with solid tumors. Armoring CAR T cells to express immunomodulatory cytokines can enhance treatment efficacy. In this study, we assess how systemic pre-conditioning impacts CAR T cells alone and in combination with interleukin (IL)-2 superkine (Sup2) and IL-33 cytokine armoring in the context of B16F10 melanoma.

Methods

Local and systemic pre-conditioning regimens were assessed in combination with TRP1-specific CAR T cells with or without Sup2 and IL-33 armoring in immunocompetent mice with B16F10 melanoma. CAR T-cell expansion kinetics and phenotype were evaluated with luminescence imaging, flow cytometry, and single-cell RNA sequencing.

Results

Systemically preconditioned mice treated with Sup2-IL33 CAR T cells elicited greater tumor control and extended overall survival compared with non-conditioned or locally preconditioned mice. Depletion of Sup2-IL33 CAR T cells in preconditioned mice resulted in loss of durable tumor control. While total body irradiation (TBI) increased unarmored CAR T-cell engraftment, Sup2-IL33 armoring increased tumor infiltration of effector cells over TOX+exhausted cells. RNA velocity trajectory analysis predicted divergent differentiation of tumor-infiltrating progenitor cells into either effector or exhausted T cells.

Conclusions

Systemic preconditioning increased the expansion and persistence of transferred CAR T cells.

TBI preconditioning in combination with cytokine armoring promoted the emergence of diverse effector phenotypes that correlated with enhanced efficacy of Sup2-IL33 CAR T cells in the context of B16F10 melanoma. Durable responsiveness and extended survival following combined therapy required continued CAR T-cell persistence despite the ability of Sup2-IL33 armoring to induce endogenous tumor immunity.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Systemic preconditioning with total body irradiation has been reported to enhance chimeric antigen receptor (CAR) T-cell engraftment in the host as well as increase infiltration and immune cell activation in the tumor. To improve efficacy of CAR T cells against solid malignancies, they can be armed with immunomodulating cytokines. While pre-conditioning and cytokine arming are common treatment modalities in CAR T-cell therapy, few studies have specifically explored interactions between pre-conditioning regimens with cytokine arming on CAR T-cell activity and efficacy.

WHAT THIS STUDY ADDS

• Total body irradiation pre-conditioning and Sup2-IL33 armoring synergistically enhanced the engraftment, cytotoxic activity, and persistence of CAR T cells. Persistence of Sup2-IL33 CAR T cells was required for durable tumor control in pre-conditioned mice. Following combination treatment, trajectory analysis predicted distinct T-cell differentiation into effector over exhausted cell states, which required cell-intrinsic ST2 but not IFNAR signaling.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Pre-conditioning regimens given prior to CAR T-cell therapy are largely based on patient-derived factors. This study highlights that pre-conditioning regimens also have the capacity to improve the functionality of CAR T cells. The combined effects of pre-conditioning treatment and engineered CAR T cells can be leveraged to improve patient responses to CAR T-cell therapy.

Introduction

Melanoma is the deadliest form of skin cancer with a 5-year survival rate of 30% in patients with advanced disease.¹ Current treatment strategies include chemotherapy, BRAF/Mek inhibitors, and immune checkpoint blockade therapies.¹ Adoptive cell therapies (ACT) for advanced-stage melanoma have limited efficacy due to tumor heterogeneity, antigen loss, disrupted trafficking, and immunosuppression in the tumor microenvironment.2,6 Engineering adoptively transferred T cells to express immunomodulatory cytokines can improve efficacy by directly supporting T-cell function or by reprogramming suppressive cell populations in the tumor.^{7 8} We previously reported that chimeric antigen receptor (CAR) T cells armored with the interleukin (IL)-2 superkine (Sup2), which has enhanced binding affinity to CD122, and the alarmin IL-33 synergistically delayed B16F10 melanoma tumor growth.^{9 10} Sup2 and IL-33 armoring transiently improved CAR T-cell engraftment and increased tumor infiltration and activation of various endogenous immune cell populations. However, antitumor responses quickly declined as CAR T-cell levels decreased.⁹ CAR T-cell expansion was extended by pre-conditioning hosts through transient depletion of inhibitory CD4 T cells. Moreover, CD4 T-cell depletion allowed CAR and endogenous CD8 T cells to differentiate into memory T cells that delayed B16F10 growth after rechallenge.¹¹ Here, we hypothesize that clinically-relevant pre-conditioning regimens, which are more broadly lymphodepleting and immunomodulatory, will sustain CAR T cell-dependent antitumor efficacy.

Systemic chemotherapy or radiation as pre-conditioning lymphodepleting regimens is a critical component of ACT that directly stimulates immunological responses, reduces tumor burden, and increases T-cell engraftment.^{2 12 13} Lymphodepletion augments T-cell engraftment by reducing endogenous immune populations, increasing the availability of homeostatic cytokines and creating space for transferred cells to proliferate.14,16 Chemotherapy and radiation also promote normalization of tumor vasculature and induce CXCL9 and CXCL10 expression, which promotes infiltration and retention of CXCR3+ T cells in the tumor microenvironment.^{17 18} Additionally, direct tumor cell death leads to the release of inflammatory cytokines, including type I interferons (IFNs).^{13 19} Type I IFNs indirectly enhance CD8 T-cell activation through augmenting dendritic cell maturation and cross-presentation. Type I IFNs have also been reported to act directly on CD8 T cells, serving as “signal three” during T-cell activation, highlighting a potential role for radiation-induced type I IFNs in supporting T-cell activity following adoptive transfer.20,25

Here we demonstrate that pre-conditioning with total body irradiation (TBI) prior to Sup2-IL33 CAR T-cell therapy significantly enhanced B16F10 tumor control and extended survival. CAR T-cell engraftment was universally improved by irradiation; however, enhanced tumor control was limited to mice treated with CAR T cells armored with Sup2 and IL-33. In contrast to non-conditioned hosts, in which Sup2-IL33 armored CAR T cells did not persist, TBI pre-conditioning sustained CAR T-cell levels that were required for long-term antitumor responses. CAR T-cell intrinsic IL-33 but not type I IFN signaling was required to expand multiple tumor-infiltrating CD8 T-cell effector subsets in pre-conditioned mice treated with Sup2-IL33 CAR T cells. This study highlights a synergistic relationship between systemic pre-conditioning with Sup2 and IL-33 armoring.

Materials and methods

Cell lines and culture conditions

B16F10 (CRL-6475) and HEK 293T (CRL-1168) cell lines were purchased from American Type Culture Collection (ATCC). B16F10 cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 (Cytiva SH30027) supplemented with 7.5% heat inactivated fetal bovine serum (FBS) (Cytiva SH3002702) 100 U/mL penicillin, and 100 mg/mL streptomycin (Gibco 15140122). HEK 293T cells were cultured in high glucose Dulbecco’s Modified Eagle Media (Cytiva SH30022.01) supplemented with 10% FBS, 1% Minimum Essential Media (MEM) non-essential amino acids (Gibco 11140050), 1% sodium pyruvate (Gibco 11360070), 1% N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES) (Gibco 15630080). Primary T cells were cultured in RPMI-1640 supplemented with 10% FBS, 1% MEM non-essential amino acids, 1% sodium pyruvate, 1% HEPES, 50 µM of 2-mercaptoethanol (Gibco 21985023), and 25 U/mL human IL-2 (NCI-Frederick BRB Repository).

Pre-conditioning protocol

Pre-conditioning regimens were given 7 days post tumor inoculation and 24 hours prior to ACT. For TBI, mice were given one dose of 5 Gy (500 cGy) using an X-Rad 320. For local irradiation, an adjustable collimator was used to define a 2×2 mm field of exposure surrounding the tumor. Mice treated with chemotherapy received 4 mg of cyclophosphamide (Thermo Scientific 203960010) and 4 mg of fludarabine (Caymen Chemical 14251) per mouse injected intraperitoneally.

CAR and cytokine constructs

TA99 CAR, Sup2-IL33, and PpyRE9 luciferase retroviral constructs have been previously described in Brog et al.⁹ The TA99 CAR construct is comprised of an extracellular single-chain variable fragment (scFv) specific for TRP1 (sequence gifted from Dr Darrell Irvine, Massachusetts Institute of Technology), the CD8 transmembrane domain, and intracellular CD3 zeta and CD28 signaling domains. The cytokine construct contains Super2 and mouse IL-33, separated by a T2A self-cleaving peptide, followed by a (GGS)₂ linker, an internal ribosomal entry sequence (IRES), and GFP. Luciferase expressing construct contains PpyRE9 coding sequence, IRES, and the fluorescent protein, Clover.

Retroviral packaging and viral transfection

To package retroviruses, 2.6 million 293T HEK cells were seeded in 10 cm dishes 18–24 hours prior to transfection with the calcium chloride method. In brief, 2×HEPES buffered saline was added dropwise to a mixture containing 10 µg of pCl-Eco, 10 µg of CAR, cytokine, or luciferase construct, and 51 µL of 2.5M calcium chloride. After a 15-minute incubation at room temperature, the plasmid mixture was added to the 293T HEK cells in a dropwise fashion. Media was changed 7 hours after transfection. Virus was collected 48 and 72 hours post transfection, filtered through a 45 µM filter, and stored at -80°C until use.

Generation of CAR T cells

Splenocytes harvested from 8–12 weeks old donor mice were homogenized using the plunger of a 5 mL syringe, filtered through a 70 µM cell strainer, lysed with 2 mL per spleen of 1×red blood cell (RBC) lysis buffer (BioLegend 420301), and passed through a second 70 µM filter. Splenocytes were cultured with 1 µg/mL of concanavalin A (Sigma C5275) for 48 hours. T cells were plated at 8 million cells per well in a 24-well plate with retroviral supernatant containing 10 µg/mL of polybrene (Sigma 107689) and spinoculated for 90 min at 1,500×g at 37°C. Cells were cultured at 0.5–1 million cells/mL. Transductions were performed sequentially every 24 hours starting with Sup2-IL33, TA99 CAR, and finally luciferase expressing vectors. Transduction efficiency was assessed by flow cytometry. Expression of fluorescent proteins or G4S linker (Cell Signaling Technology 71645) were used to assess transduction efficiency.

In vivo mouse experiments

Female C57BL/6, CD45.1-congenic or Thy1.1-congenic mice were ordered from Charles River or Jackson Laboratories. IFNAR KO (IFNAR KO, Jax stock # 028288) and ST2 KO (ST2 KO, Jax stock # 039982) donor mice were ordered from Jackson Laboratories. All mice were housed in specific-pathogen-free facilities. Mice were 8–12 weeks of age at the onset of experiments. All experiments were approved by the Institutional Animal Care and Use Committee at Dartmouth College. ARRIVE1 reporting guidelines were used for animal studies.²⁶ Mice were inoculated intradermally with 300,000 B16F10 cells in 50 µL of HBSS. Tumor volume was calculated by multiplying length, width, and depth, determined with calipers. Four to six mice were assigned so that average tumor volume was similar across groups prior to therapy. Mice displaying moribund signs or reaching a maximum tumor burden of 15 mm in diameter were euthanized. 7 million CAR T cells in 100 µL of HBSS were transferred intravenously through the tail vein on days 6–8 post tumor inoculation. For CAR T-cell depletion study, 250 µg of anti-Thy1.1 (CD90.1) (Bio X Cell BE0214) was given intraperitoneally on days 17 and 24 following ACT. For luminescence imaging, mice were intraperitoneally injected with 3 mg luciferin, incubated for 8 min, and then imaged for 5 min on a Xenogen VivoVision IVIS imager (PerkinElmer). Luminescence was quantified with the Living Image V.4.7.3 (PerkinElmer).

Tissue processing and isolation

Tumors, tumor-draining lymph nodes (tdLNs), and spleens were harvested and filtered through a 70 µM cell strainer using the back of a 5 mL syringe. RBCs in splenocytes were lysed, and then splenocytes were refiltered through 70 µM cell strainer prior to analysis. To isolate tumor-infiltrating leukocytes (TILs), tumors were resuspended in 80% Percoll, underlaid with 40% Percoll, and spun at 500×g for 20 min at 2,000 rpm with no brake. The buffy coat between 40% and 80% was collected and washed with media prior to analysis.

Cell staining

Single-cell suspensions were stained with viability dye for 10 min at room termperature (RT). Cells were then incubated with Fc block before staining for 30–45 min on ice with antibodies described in online supplemental tables 1 and 2. For intracellular cytokine staining, cells were stimulated with a Phorbol 12-myristate 13-acetate (PMA), ionomycin, and Brefeldin A cocktail (BioLegend 423303) for 3 hours in 37°C CO₂ incubator prior to staining for surface markers as previously described, then fixed, permeabilized, and stained with antibodies for cytokines (BioLegend 426803).

Multiplex Immunohistochemistry (IHC)

Tumor were fixed in 10% neutral balanced formalin, embedded in paraffin blocks, and cut into 4 µm sections. Staining with antibodies listed in online supplemental table 3 was done using an Opal 6-Plex Manual Detection Kit (Akoya Biosciences, NEL811001KT) following manufacturer’s guidelines and mounted using VECTASHIELD Vibrance Antifade Mounting Medium (Vector Laboratories: H-1700). Whole slide scans and Multi-spectral imaging (MSI) regions were acquired using a Vectra V.3.0 microscope. Spectral libraries were obtained using unstained, 4’,6-diamidino-2-phenylindole (DAPI) stained, and single-color control slides for each Opal using CD45 as a primary antibody. Images were unmixed using InForm V.2.60 software and analyzed with HALO (Indica Labs) module Highplex FL V.4.0.4. ImageJ software was used to add scale bars and adjust brightness and contrast.

Cytotoxicity analysis

CAR T cells were co-cultured in vitro with B16F10 tumors expressing luciferase at the indicated effector to target ratios for 18–24 hours at 37°C. Luciferin (200 µg/mL) was then added and incubated for 30 min. Bioluminescence was measured on a SpectraMax i3x. For ex vivo experiment, CD45.2+CAR and CD45.1+host CD8 T cells were sort purified from tumors excised from mice treated with TBI and Sup2-IL-33 armored CAR T cells 6–7 days post transfer.

scRNA library construction, sequencing, and analysis

Single-cell RNA sequencing (scRNA) sequencing was performed on CD45+cells from tumor and tdLNs using the 10x Genomics platform according to manufacturer’s directions. Cells were loaded onto a Chromium Next GEM chip targeting 10,000 cells for capture and processed using the 10x Genomics 3’ v3.1 chemistry (PN-1000268). Libraries were prepared per manufacturer’s instructions and sequenced using NextSeq 2000 paired-end 100 cycling kits. 10x Genomics pre-processing was performed via Cell Ranger count (V.8.0.0) to produce gene expression matrices. Transcriptional profiling was performed using Seurat (V.5.1.0). The MULTIseqDemux() function was used to annotate biological replicates marked by the hashtags. Cells expressing <200 or >5,500 unique genes and >3.5% or 5% mitochondrial genes (sample dependent) were excluded from analysis. Doublets and cells with no hashtag were eliminated. Genes expressed by less than 3 cells were excluded. Sample integration was performed using Pearson residuals as in satijalab.org/Seurat/archive/V.4.3/sctransform_v2_vignette. Cells from the tumor samples were subsetted from the integrated object containing both tissues. Principal component analysis dimension reduction was performed on the tumor dataset. Cells were then clustered using the first 20 principal components (PCs). The cluster resolution of 0.1 was determined by clustree(). Clusters were visualized with Uniform Manifold Approximation and Projection Analysis (UMAP). The FindMarkers() function was used to find top differentially expressed genes for each cluster compared with all other cells. Top differentially expressed genes were used to identify clusters.

Projection and Gene Set Enrichment Analysis of CD8 T-cell clusters

Cells expressing Cd8a were subsetted and independently normalized and transformed using SCT transform. Nearest neighboring unsupervised clustering was performed on integrated data with a resolution of 0.9. Clusters were visualized with both UMAP and factor analysis (FA). The FindMarkers() function was used to generate differentially expressed gene (DEG) lists for each CD8 cluster. CD8 T-cell clusters were identified using DEGs and canonical cell marker markers. Gene Set Enrichment Analysis (GSEA) was performed with the fgsea package (V.1.34.4). DEGs were ranked by average log2FC and CD8 subset gene signatures from Giles et al.²⁷ Clusters were projected onto reference map of tumor infiltrating T cells using the ProjecTILs package (V.3.5.1).²⁸

RNA velocity-based trajectory analysis

Individual BAM files from cell ranger output were used to generate.loom files containing spliced and unspliced RNA count matrices using the velocyto.py pipeline (V.0.17.17).²⁹ The integrated Seurat object was converted to an anndata object using modified code from https://smorabit.github.io/blog/2021/velocyto/. Subsequent analysis was run using the SCANPY and scVelo pipelines.³⁰ Nearest neighbor graph calculations were performed using the first 50 principal components imported from the Seurat object. A force-directed graph layout was computed with PAGA initialization to improve velocity stream visualization.³¹ Separate anndata objects were generated for each treatment group by merging corresponding.loom files with the original anndata object. Each data matrix was normalized prior to calculation of first and second order moments for velocity estimation among nearest neighbors using the first 50 principal components and 30 neighbors with the UMAP method. Velocity was calculated using the “stochastic” model. Calculated velocities were projected onto the force-directed graph generated from the initial merged anndata object containing all samples as described above.

Statistical analysis

In vivo experiments were repeated at least two times with a total of 8–10 mice per group. Statistical analysis was performed on combined data from all experiments using one-way or two-way analysis of variance with post hoc test as specified in the figure legend using GraphPad Prism. Data are plotted as the mean+SD. Survival was plotted using Kaplan-Meier plots, and the log-rank Mantel-Cox test was used to assess statistical differences. P values: ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. The luminescence fold change was calculated by dividing the raw flux value with the average flux 4 days post CAR T cell. Fold change was used to normalize luminescence between experiments. Correlation between flux and survival was quantified with Pearson’s correlation coefficient in GraphPad Prism.

Results

TBI increases CAR T-cell expansion in the tumor and enhances tumor control by Sup2-IL33 CAR T cells

We previously reported that without pre-conditioning, TRP1-specific CAR T cells have no effect on B16F10 melanoma growth; however, armoring CAR T cells with Sup2 and IL-33 cytokines reduced tumor growth, largely through induced activation of endogenous antitumor immunity.^{9 11} Pre-conditioning is typically required both as a “bridging therapy” to reduce tumor burden prior to CAR T-cell infusion and to enhance the engraftment of adoptively transferred CAR T cells.^{7 15} To determine how pre-conditioning alters CAR T-cell efficacy against melanoma, we treated mice with a sublethal dose of 5 Gy TBI (figure 1A). Tumor growth and overall survival was modestly and similarly delayed by TBI pre-conditioning alone, Sup2-IL33 CAR T-cell therapy alone, or TBI pre-conditioning and unarmored CAR T cells (figure 1B–D). In comparison, TBI pre-conditioning and Sup2-IL33 armored CAR T-cell therapy substantially delayed tumor growth and extended overall survival (figure 1B–D). CAR T-cell expression of either Sup2 or IL-33 alone was not sufficient to improve overall survival in combination with TBI pre-conditioning, but IL-33 armoring could mediate moderate tumor control, indicating that both cytokines are required for maximal activity (online supplemental figure 1A and B). A second commonly used pre-conditioning regimen, cyclophosphamide and fludarabine (Cy/Flu) chemotherapies, also delayed tumor growth in combination with Sup2-IL33 CAR T cells but did not extend survival to the same extent as TBI (online supplemental figure 1C and D). As expected, high-dose local radiation transiently reduced B16 tumor burden, but it did not extend survival compared with TBI and Sup2-IL33 CAR T-cell therapy (online supplemental figure 1D and E). This suggests that persistent B16F10 tumor control requires systemic TBI pre-conditioning and Sup2-IL33 armored CAR T cells.

Figure 1. TBI enhances CAR T-cell engraftment and tumor control by Sup2-IL33 CAR T cells. (A) Experimental schematic. C57BL/6 mice inoculated with 3×10⁵ B16F10 i.d. (day −8) were treated with 5Gy TBI (day −1) followed by luciferase-expressing CAR T cells (7×10⁶ i.v.) (day 0). In vivo kinetics of T cells were tracked by luminescence imaging starting 2 days post transfer. (B) Average tumor volume in each treatment group post transfer. Data are plotted as mean+SD and analyzed by two-way ANOVA with Tukey’s post hoc test. (C) Kaplan-Meier curve showing survival post CAR T-cell transfer assessed with log-rank (Mantel Cox) test. (D) Tumor weights at 15 days post-CAR T-cell transfer. (E) Representative images of luciferase-expressing CAR T cells with or without TBI or Sup2-IL33 armoring. (F) (Left) Representative image depicting tumor and total regions of interest used to quantify bioluminescence. (Right) Bioluminescence on indicated day after CAR T-cell transfer. Data are plotted as mean+SD and assessed with a mixed-effects model using Tukey’s multiple comparison test. (A–F) Data are from three independent experiments with at least 10 biological replicates per treatment group. ANOVA, analysis of variance; CAR, chimeric antigen receptor; i.d., intradermal; IL, interleukin; i.v., intravenous; Sup2, superkine; TBI, total body irradiation.

Systemic pre-conditioning regimens globally deplete host immune populations, liberating homeostatic cytokines to support the engraftment of adoptively transferred T cells.³² To assess CAR T-cell engraftment, we engineered CAR T cells to express PpyRE9 luciferase and conducted longitudinal imaging to track their abundance and localization. CAR T cells were present at similar levels 6 days post transfer across treatment groups (figure 1E,F). As previously reported without pre-conditioning, Sup2-IL33 armored CAR T cells exhibited peak expansion at 6–8 days but were lost 12 days post transfer (figure 1E,F).^{9 11} In line with observed tumor control, singly armored Sup2 or IL-33 CAR T cells displayed low or intermediate expansion, respectively, in comparison to dual cytokine armored cells (online supplemental figure 1F). In general, TBI pre-conditioning led to greater CAR T-cell persistence independent of armoring. However, TBI pre-conditioning combined with Sup2-IL33 armoring resulted in the most robust and prolonged expansion of CAR T cells in the tumor and total mouse compared with other groups (figure 1E,F). Pre-conditioning with chemotherapy similarly increased CAR T-cell engraftment, while local radiation had no effect on CAR T-cell kinetics (online supplemental figure 1G). Collectively, these data show that systemic pre-conditioning leads to durable CAR T-cell engraftment and augments and prolongs tumor control by Sup2-IL33 CAR T cells.

CAR T-cell persistence is required for durable tumor control in TBI+S233 CAR T-cell treated mice

While all mice treated with TBI and Sup2-IL33 CAR T cells had improved tumor control, duration of tumor control varied among mice. We stratified treated mice as either “Short-term” responders (STR) or “Long-term” responders (LTR) based on their tumor growth and survival kinetics (figure 2A,B). Tumor growth kinetics in STR and LTR diverged 20–30 days post CAR T-cell transfer. Comparison of relative CAR T-cell luminescence revealed that loss of tumor control correlated with reduced CAR T-cell signal in STR (online supplemental figure 2A). To assess this trend across STR and LTR, we plotted CAR luminescence at day 14–16 or 26–28 against overall survival. At both time points, higher luminescence positively correlated with extended survival (figure 2C). This indicates that CAR T cells are likely responsible for continued tumor control. To test this hypothesis, we generated CAR T cells from Thy1.1+ donor mice and then used anti-Thy1.1 antibody to deplete CAR T cells starting 17 days post-transfer to model the observed loss of CAR T cells in short-term responders, verified by luminescent imaging (figure 2D, online supplemental figure 2B). Depletion of CAR T cells resulted in similar loss of tumor control and reduced overall survival observed in STR (figure 2E). These data show that continued persistence of Sup2-IL33 armored CAR T cells is necessary for prolonging tumor control. To determine whether cytokine expression by transferred T cells is sufficient for long-term tumor responses, we evaluated tumor control and survival following treatment with Sup2-IL33 armored T cells without expression of the tumor specific CAR. Sup2-IL33 armored T cells exhibited decreased tumor control and survival in the absence of CAR expression (figure 2F). Bioluminescence tracking indicated that Sup2-IL33 armored T cells systemically expanded at similar levels with or without CAR expression. However, armored T cells lacking CAR expression did not accumulate well in the tumor, suggesting that CAR expression facilitates tumor recruitment or retention of transferred T cells (online supplemental figure 2C). Collectively, these results demonstrate that long-term tumor control requires persistence of T cells expressing both Sup2-IL33 cytokines and a tumor specific CAR.

Figure 2. CAR T-cell persistence is required for tumor control by Sup2-IL33 CAR T cells in TBI pre-conditioned mice. (A)(Left) Individual tumor growth in mice after treatment with TBI and Sup2-IL33 CAR T cells. (Right) Kaplan-Meier curve showing survival in short- and long-term responder groups after CAR T-cell transfer. Data are from two independent experiments with a total of nine biological replicates. (B) Tumor weight stratified by short and long-term responders 41 days post CAR T-cell transfer. Data from two independent experiments analyzed with Mann-Whitney test. (C) Correlation between quantified luminescence at specified time points and survival time post CAR T-cell transfer. Total flux is plotted as a fold change from the average flux 4 days after CAR T-cell transfer. Correlation was calculated using Pearson’s correlation coefficient. Data collected from four independent experiments with 15–20 total biological replicates per treatment. (D) Experimental schematic. Luciferase-expressing CAR T cells generated from Thy1.1 donor mice were transferred into Thy1.2 hosts. Mice received 250 µg of depleting Thy1.1 antibody i.p. on days 17 and 25 post CAR T-cell transfer. (E) Tumor growth and survival after CAR T-cell depletion. Gray vertical lines indicate treatment with Thy1.1 antibody. Dashed red and blue lines represent average tumor growth and survival of short-term and long-term responders depicted in A. Data shown from one to two independent experiments with six to eight mice per treatment. (F) (Left) Average tumor volume in each treatment group. Data are plotted as mean+SD and analyzed by two-way ANOVA with Tukey’s post hoc test. (Right) Kaplan-Meier curve depicting survival post CAR T-cell transfer assessed with log-rank (Mantel-Cox) test. Data from two independent experiments with 10 biological replicates total. Data are representative of two independent experiments with 10 total biological replicates. ANOVA, analysis of variance; CAR, chimeric antigen receptor; GzmA, granzyme A; GzmB, granzyme B; IL, interleukin; i.p., intraperitoneal; i.v., intravenous; Sup2, superkine; TBI, total body irradiation; TIM3, T-cell immunoglobulin and mucin-domain containing-3.

TBI increases activation and cytotoxic functionality of Sup2-IL33 armored CAR and host endogenous CD8 T cells in the tumor-draining lymph node

Given that TBI and cytokine armoring of CAR T cells can independently modulate both CAR and host responses, we sought to examine differences in CD8 T-cell activation and effector activity across treatments. CD45.1⁺ CAR T cells and CD45.2⁺ host immune cells in the tdLN and spleen were analyzed 1 week post CAR T-cell transfer (figure 3A, online supplemental figure 3A). Non-armored and Sup2-IL33 armored CAR T cells comprised >50% of CD8 T cells in TBI pre-conditioned mice, while Sup2-IL33 CAR T cells represented only 1–8% of CD8 T cells in non-irradiated mice, demonstrating that TBI dramatically increases engraftment of CAR T cells within the tdLN and spleen (figure 3B, online supplemental figure 3B).

Figure 3. TBI enhances CAR T-cell engraftment in the tdLN. (A) Experimental schematic. CD45.1+CAR T cells were transferred into CD45.2 hosts. tdLN cells were analyzed by flow cytometry 7 days after CAR T-cell transfer. (B) Representative gating and proportions of CD45.1+CAR T and CD45.2+host CD8 T cells. (C) Frequency of CD44+CAR and host CD8 T cells. (D) Representative histograms and MFI of CD44 expression. (E) Representative gating and frequency of PD-1 and TIM3 expression. (F) Representative gating and proportion of CD8 T cells expressing granzymes A and B. (A–E) Data are combined from two independent experiments with four to five biological replicates per experiment and plotted as mean+SD. (F) Data are from one experiment with four to five biological replicates. (A–F) Ordinary one-way analysis of variance with Tukey’s multiple comparisons test. CAR, chimeric antigen receptor; IL, interleukin; i.v., intravenous; MFI, mean fluorescent intensity; PD-1, programmed cell death protein 1; Sup2, superkine; TBI, total body irradiation; tdLN, tumor-draining lymph node.

To assess T-cell activation, we first examined CD44 expression levels by flow cytometry. TBI pre-conditioning significantly increased the frequency of CD44+CD8 CAR T cells compared with non-conditioned mice. Sup2-IL33 armoring further increased the frequency and mean fluorescent intensity (MFI) of CD44 on CD8 CAR T cells in TBI-pre-conditioned mice in both tdLN and spleen (figure 3C,D, online supplemental figure 3C). In the tdLN, TBI pre-conditioning also increased the frequency and MFI of CD44 on host CD8 T cells, with the highest CD44 MFI observed in the combination treatment group (figure 3C,D). Conversely, host CD8 T cells in the spleen did not increase in frequency and intensity of CD44 following treatment with TBI and Sup2-IL33 CAR T cells (online supplemental figure 3C and D). These data indicate that TBI pre-conditioning and Sup2-IL33 armoring of CAR T cells increases the proportions of activated CD8 CAR and host CD8 T cells in the tdLN and increased activation of CD8 CAR but not host CD8 T cells in the spleen.

Expression of programmed cell death protein 1 (PD-1) and T-cell immunoglobulin and mucin-domain containing-3 (TIM3) checkpoint inhibitory receptors were used to further assess activation and exhaustion-associated markers on CD44+CD8 T cells. As previously reported, Sup2-IL33 armoring expanded proportions of PD-1+TIM3- activated CAR T cells.^{9 11} TBI pre-conditioning further increased frequencies of PD-1+TIM3- cells and induced PD-1+TIM3+ Sup2-IL33 CAR T cells in both tdLN and spleen (figure 3E, online supplemental figure 3E). Conversely, TBI did not increase PD1+TIM3- unarmored CAR T cells in either tissue (figure 3E, online supplemental figure 3E). TBI pre-conditioning or Sup2-IL33 armoring resulted in similar PD-1 expression on host T cells, while combination treatment further increased PD-1+TIM3- host T cells and also resulted in PD-1+TIM3+ cells in the tdLN (figure 3E). In contrast, PD-1 and TIM3 expression levels on host CD8 T cells were similar in pre-conditioned mice treated with either non-armored or Sup2-IL33 armored CAR T cells and elevated in non-conditioned mice treated with Sup2-IL33 CAR T cells in the spleen (online supplemental figure 3D).

Given the differential expression of activation and exhaustion markers on both CAR and host CD8 T cells in the tdLN, we next evaluated the expression of cytotoxic enzymes, granzymes A and B. Expression of one or both granzymes was induced by Sup2 and IL-33 armoring alone, but their expression was further augmented by TBI pre-conditioning (figure 3F). Granzyme A and B expression by host CD8 T cells was absent except in combined TBI and Sup2-IL33 CAR T-cell treated mice. This indicates that combination treatment increased expression of cytotoxic enzymes despite increased expression of checkpoint inhibitory molecules. Taken together these data demonstrate that TBI in combination with Sup2-IL33 armoring enhances CAR T-cell engraftment and induces greater activation of both CAR T and host CD8 T cells in the tdLN.

TBI and Sup2-IL33 armoring of CAR T cells increases activation and cytotoxic capacity of CD8 CAR T cells in the tumor

We next examined tumor infiltrating CD45.1+CAR and CD45.2+host cells 1 week post CAR T-cell transfer (figure 4A). Only 20% of CD8 T cells in the tumors of non-irradiated mice were Sup2-IL33 armored CAR T cells. Non-armored CAR T cells represented 50% and Sup2-IL33 armored CAR T cells represented 80% of CD8+T cells in the tumors of TBI pre-conditioned mice (figure 4B). Thus, in contrast to the tdLN, where TBI resulted in similar CAR T-cell engraftment independent of cytokine armoring, the highest tumor-infiltration of CD8 CAR T cells required both TBI and Sup2-IL33 armoring.

Figure 4. Frequency and activation phenotype of CAR T cells in the tumor are increased by total body irradiation and Sup2 and IL-33 armoring. (A) Experimental schematic. CD45.1+CAR T cells were transferred into CD45.2 hosts. Tumor-infiltrating CD45+cells were analyzed 7 days after ACT. (B) Representative gating and frequency of CD45.1+CAR T and host CD45.2+CD8 T cells. (C) Frequency of CD44+CD8 cells. (D) Representative histograms and MFI of CD44 expression. (E) Representative gating and proportions of PD-1-expressing and TIM3-expressing CD8 T cells. (F) Representative histograms and MFI of PD-1 expression. (G) Ex vivo cytotoxicity of B16-luciferase tumor cells following 18–24 hours co-culture with either CAR or host CD8 T cells sorted from TBI and Sup2-IL33 CAR T-cell treated tumors at specified effector to target ratios. (A–G) Data are combined from two independent experiments with four to five biological replicates per experiment and plotted as mean+SD. Ordinary one-way analysis of variance with Tukey’s multiple comparisons test. ACT, adoptive cell therapies; CAR, chimeric antigen receptor; IL, interleukin; i.v., intravenous; MFI, mean fluorescent intensity; PD-1, programmed cell death protein 1; Sup2, superkine; TBI, total body irradiation; TIL, tumor infiltrating leukocyte; TIM3, T-cell immunoglobulin and mucin-domain containing-3.

Although the frequency of tumor-infiltrating CD44+CAR and host CD8 T cells were similar across treatments in the tumor, CD44 MFI was considerably higher in CAR T cells after combined treatment (figure 4C,D). Proportions of PD-1 and TIM3 expressing cells were similar in unarmored CAR T cells with TBI and Sup2-IL33 armored CAR T cells without TBI pre-conditioning (figure 4E). However, combination treatment resulted in expanded proportions of PD-1^INT or PD-1^HITIM3- and PD-1^HITIM3+ CAR T cells (figure 4E). Tumor infiltrating host CD8 T cells were predominately PD-1^HITIM3+ after TBI pre-conditioning, irrespective of cytokine armoring (figure 4E). Similarly, PD-1 MFI followed the same trend and was highest in cytokine armored CAR CD8 T cells with TBI pre-conditioning (figure 4F). To evaluate CD8 T-cell functionality, we first conducted immunofluorescence staining of tumors to assess and quantify granzyme B expression by CAR and host CD8 T cells (online supplemental figure 4A). TBI pre-conditioning significantly increased granzyme B expression by CAR and host CD8 T cells after Sup2-IL33 CAR T-cell treatment (online supplemental figure 4B). To examine if increased granzyme expression by CD8 CAR T cells correlates with increased tumor killing ex vivo, we co-cultured B16-luciferase tumor cells with CAR and host CD8 T cells sorted from tumors of mice treated with TBI and Sup2-IL33 CAR T cells (figure 4G). Tumor killing can be quantified by measuring a reduction in live B16F10-dependent luminescence following T-cell co-culture. While host CD8 T cells displayed minimal tumor cell killing across effector to target ratios, CD8 CAR T cells killed tumor cells in a dose-dependent manner with 66% and 33% reduction in luminescence at the 2:1 and 0.25:1 effector to target ratios, respectively. Collectively, these results indicate that while TBI pre-conditioning and Sup2-IL33 armoring can induce activation markers on host CD8 T cells, it is particularly effective at inducing tumor infiltration, activation, and cytotoxic effector potential of CD8 CAR T cells.

TBI pre-conditioning combined with Sup2-IL33 CAR T cells promote multiple effector CD8 T-cell subsets in the tumor

To further profile tumor-infiltrating CD8 T cells, we performed scRNAseq analysis. Two proliferating, a naïve-like, a progenitor, three effector, and four exhausted CD8 T-cell clusters were identified (figure 5A). Cell states were inferred based on expression of canonical cell markers and top DEGs (figure 5B, online supplemental figure 5A and B).^{27 33 34} Proliferating clusters highly expressed cell cycle genes (Mki67, Birc5, Stmn1). The naïve-like and progenitor cluster highly expressed naïve and central memory cell associated genes (Tcf7, IL7r) and the transcription factor Klf2 (figure 5B).^{33 34} Klf2 suppresses expression of the exhaustion-associated transcription factor Tox and enables expression of the effector-associated transcription factor, T-bet.³⁵ Non-treated (NT) tumors were enriched in naïve-like and progenitor subsets (figure 5C).

Figure 5. TBI and Sup2-IL33 CAR T-cell therapy increases tumor-infiltrating CD8 effector subsets. (A) UMAP of CD8 T-cell clusters. (B) Dot plot displaying selected genes used for cluster identification. (C) Distribution of clusters across conditions. (D) Gene Set Enrichment Analysis scores of identified clusters compared with CD8 T-cell subsets identified in Giles et al. (E) Contour projection of indicated clusters onto a UMAP depicting T-cell subsets from ProjecTILs reference dataset. (F) FA plot depicting CD8 T-cell clusters. (G) RNA velocity fields projected onto FA plots for indicated treatments. Direction of arrows indicates predicted differentiation paths. Arrow length indicates the rate of predicted differentiation. (H) Summary of predicted differentiation pathways. CAR, chimeric antigen receptor; Eff-Mem, effector-memory cluster; Eff-Term, effector-terminal cluster; Exh-Term, exhausted-terminal subset; Exh-Tox, exhausted-Tox; FA, factor analysis; IFN, interferon; IL, interleukin; i.v., intravenous; NK, natural killer; NT, non-treated; Sup2, superkine; TBI, total body irradiation; Treg, regulatory T cell; UMAP, Uniform Manifold Approximation and Projection.

The effector-memory cluster (Eff-Mem) expressed genes associated with stemness, effector activity, and exhaustion (Slamf6, Prf1, Pdcd1). The effector-terminal cluster (Eff-Term) expressed the highest expression of cytotoxic effectors and terminal differentiation marker KLRG1 (Gzmb, Gzma, Klrg1). All three effector clusters expressed genes associated with response to type I IFNs (Ly6c2, Irf7, Isg15), with the highest expression in the effector-IFN cluster (Eff-IFN) (figure 5C).^{33 34 36} Effector subsets were highly enriched in response to combination treatment (figure 5C).

Four exhausted clusters were identified by Tox expression.³⁷ This included two subsets expressing either high or low Tox (Exh-Tox^HI, Exh-Tox^Lo), an exhausted-terminal subset (Exh-Term) expressing some cytotoxic effectors (Prf1, Ifng), and a cluster expressing natural killer (NK) cell-associated killer cell lectin-like receptors (KLRs) (Exh-NK) (figure 5B).³⁴ Tumors treated with TBI alone or Sup2-IL33 CAR T cells alone were enriched in exhausted CD8 T-cell subsets. Tumors treated with TBI alone were predominantly comprised of the Exh-Tox^HI subset while tumors treated with Sup2-IL33 CAR T cells alone were enriched in Exh-Term and Exh-NK subsets (figure 5C).

We next generated GSEA scores for each cluster in relation to T-cell states responding to chronic Lymphocytic Choriomeningitis Virus (LCMV) infection reported by Giles et al (figure 5D).²⁷ Naïve-like, proliferating and Eff-IFN clusters had high enrichment scores for “naïve”, “proliferating”, and “IFN stimulated” T-cell states, respectively (figure 5D). The Eff-Term population had strong enrichment with multiple effector and exhausted cell states (figure 5D). Exh-Tox^HI and Exh-NK clusters were weakly enriched for the “progenitor” and “KLR exhausted states”, respectively. The progenitor, Eff-Mem, and Exh-Term clusters were not significantly enriched for any T-cell states in Giles et al (figure 5D). Seeking to better characterize these clusters, we sought a tumor specific reference. The ProjecTILs reference atlas encompasses T-cell states found in non-treated (NT) B16 and MC38 tumors.²⁸ We projected our clusters as contour plots onto color-coded T-cell states defined in the ProjecTILs reference atlas (figure 5E). The naïve-like, Eff-Term, and Exh-Term clusters associated with corresponding “CD8-Naïve”, “CD8-Effector/Memory”, and “CD8-exhausted” reference states. The Eff-Mem and Eff-IFN clusters projected onto both “CD8-Effector/Memory” and “CD8-Exhausted” reference states (figure 5E). Finally, the progenitor cluster projected onto both “CD8-Early Activated” and “CD8-Effector/Memory” reference states (figure 5E). Except for the naive-like and Exh-Term clusters, other clusters did not map clearly to a functionally comparable state, suggesting that T cells in TBI and Sup2-IL33 armored CAR T-cell treated tumors diverge from cell states found in NT B16 and MC38 tumors.

We used RNA velocity trajectory analysis to predict transitions between cell clusters.²⁹ RNA velocity projections were displayed on an FA plot to better decipher linear relationships between clusters (figure 5F). RNA velocity predicted that progenitor cells give rise to either the Tox^Low or Tox^HI exhausted subsets in NT tumors or the three effector populations in the combination treated tumors. The Tox^Low cluster was also predicted to give rise to both Exh-Term and Exh-NK subsets (figure 5G). These results suggest that progenitors transition into exhausted subsets in NT tumors. However, combination treatment with TBI and Sup2-IL33 CAR T cells skews the differentiation of tumor infiltrating T cells towards effector states with increased expression of genes associated with IFN response and cytotoxic effector activity (figure 5H). This prediction aligns with prior work demonstrating that CD8 progenitors differentiate into either effector or exhausted subsets, in response to environmental cues.³⁸ Our data suggest that signals important in effector differentiation are provided by TBI pre-conditioning combined with Sup2-IL33 armoring. The absence of either treatment results in differentiation towards an exhausted state.

CAR T-cell intrinsic IL-33 but not type I IFN signaling is necessary for increased CD8 CAR T-cell activation and tumor control in combination treated mice.

Radiation is known to induce expression of type I IFNs,¹⁹ which can act as a vital third activation signal, promoting CD8 T-cell expression of cytolytic enzymes, including granzyme B.²⁵ Given that the three effector clusters expressed high levels of type I IFN stimulated genes, we determined whether type I IFNs skew CAR T cells toward effector phenotypes. We treated pre-conditioned mice with Sup2-IL33 CAR T cells generated from IFNAR-deficient donor mice. However, neither tumor control nor expression of CD8 CAR T-cell activation markers was impaired in the absence of type I IFN signaling (online supplemental figure 6A-E). These results indicate that type I IFN-induced ISG expression in effector subsets may be a byproduct of pre-conditioning, rather than a requirement for CD8 effector fate.

TBI pre-conditioning also resulted in significant expansion and persistence of CAR T cells (figure 1E,F). But despite expanding, non-armored CAR T cells failed to maintain long-term tumor growth. We therefore examined the alternative hypothesis that effector T-cell fate is induced by persistent Sup2 and/or IL-33 expression following systemic pre-conditioning. Because IL-33 but not Sup2 was sufficient to mediate moderate tumor control (online supplemental figure 1A), we investigated a cell-intrinsic requirement for IL-33 signaling in promoting effector T-cell fate. Sup2-IL33 CAR T cells were generated from ST2-deficient (ST2 KO) mice, which are unresponsive to IL-33 signaling. Despite dual cytokine armoring, ST2-deficient CAR T cells had reduced tumor control compared with WT CAR T cells in vivo (online supplemental figure 7A). However, WT and ST2-deficient CAR T cells displayed similar in vitro killing of tumor cells, ruling out an intrinsic defect in T-cell cytolytic activity (online supplemental figure 7B). Consistent with reduced tumor control, tumor-infiltrating ST2-deficient CAR T cells had reduced proportions of CD44+, PD-1+TIM3, and PD-1+TIM3+ cells (figure 6A,B). In contrast, host CD8 T cells were phenotypically similar between Sup2-IL33 armored WT and ST2-deficient CAR T cells (figure 6A,B). PD-1 MFI was also significantly reduced in CD8 CAR but not host CD8 T cells in tumors treated with ST2-deficient CARs (figure 6C). In the tdLN, the frequency of PD-1+CAR T cells and PD-1 MFI were significantly decreased in ST2-deficient CD8 CAR T cells compared with WT CD8 CAR T cells, while no changes were observed in PD-1 expression by host CD8 T cells between treatments (figure 6D,E). Despite expression of activation markers on host CD8 T cells, tumor control was impaired following treatment with Sup2-IL-33 ST2-deficient CAR T cells, indicating that therapeutic efficacy is dependent on CD8 CAR T-cell activity. These results also demonstrate an essential cell intrinsic role for IL-33 signaling in promoting CD8 CAR T-cell activation in the context of this therapy.

Figure 6. ST2-deficient Sup2-IL33 armored CD8 CAR T cells have reduced expression of activation markers and display weaker tumor control. (A–D) Expression of activation markers by CD8 T cells in the tumor. Frequency of CD44+ (A) CD8 CAR and host CD8 T cells in the tumor. (B) Representative gating and frequencies of PD-1 and TIM3 expression by CD8 CAR T and host cells. (C) Representative histograms and MFI of PD-1. (D) (Left) Representative gating and frequency of PD-1+CD8 CAR T cells. (Right) Representative histograms, frequency and MFI of PD-1 by CD8 CAR T cells in the tdLN. (E) (Left) Representative gating, frequency and MFI of PD-1 by host CD8 T cells. (Right) Representative histograms and PD-1 MFI of host CD8 T cells in the tdLN. (A–F) Data plotted from two independent experiments with 9–10 biological replicates total and assessed with one-way ANOVA using Tukey’s multiple comparisons test. ANOVA, analysis of variance; CAR, chimeric antigen receptor; IL, interleukin; MFI, mean fluorescent intensity; PD-1, programmed cell death protein 1; Sup2, superkine; tdLN, tumor-draining lymph node; TIM3, T-cell immunoglobulin and mucin-domain containing-3.

Discussion

Patients routinely receive varying types and intensities of systemic pre-conditioning therapy prior to ACT depending on the stage and type of cancer.¹³ This creates immunological space by decreasing competing lymphocytes to increase engraftment of transferred T cells. Systemic pre-conditioning can also promote vital secondary effects, including stimulating host immunity, improving immune trafficking to the tumor, eliminating immunosuppressive cell populations, and reducing tumor burden.19,21 Bridging therapies, including high-dose localized radiation, administered in the weeks leading up to ACT can have similar secondary effects.^{18 39} In our study, CAR T-cell engraftment and tumor control were similar between hosts pre-conditioned with systemic total body irradiation or Cy/Flu chemotherapy, but localized pre-treatment with high-dose radiation was not sufficient for CAR T-cell engraftment. This emphasizes the importance of systemic pre-conditioning in creating immunological space despite its potential toxicities.

Systemic pre-conditioning alone delayed B16F10 melanoma growth and increased CAR T-cell engraftment; however, it did not further improve tumor control by unarmored CAR T cells. We previously reported Sup2 and IL-33 cytokine armoring is sufficient to induce potent antitumor responses in multiple solid tumor models.^{9 11} This therapy led to increased proportions of endogenous tumor-specific CD8 T cells, NK cells, and M1-like macrophages.⁹ Here we demonstrate that the combination of systemic pre-conditioning and Sup2-IL33 armoring increased CAR T-cell engraftment, leading to prolonged tumor control. In contrast to non-conditioned hosts, Sup2-IL33 armored CAR T cells were essential to achieve prolonged tumor control in TBI pre-conditioned hosts. Strategies that favor CAR T cell-mediated tumor protection over host-dependent responses may be more clinically reliable in patients who have already received multiple lines of immunodepleting therapies. CAR T cells can also be better tuned by engineering inducible regulators, logic gates, or kill-switches to improve specificity and safety.⁴⁰

Chronic tumor antigen stimulation drives dysfunctional CD8 T-cell states that can often be rescued by blocking PD-1 on progenitor exhausted cells.⁴¹ We identified a distinct TCF1+ progenitor population lacking expression of classic exhaustion markers. Tox+ exhausted cells were enriched in tumors from NT or singly treated mice while Tox^neg effector subsets were enriched in combination treated mice. RNA velocity trajectory analysis predicted that the TCF1+ progenitor cells give rise to both effector and exhausted subsets but through distinct differentiation pathways. This prediction aligns with Zander et al, which describes progenitor T cells diverging into effector or exhausted cells based on IL-21 production by CD4 Th cells in the context of chronic viral infection.³⁸ While CD4 Th cells and IL-21 expression were not observed in combination-treated tumors, synergy between IL-2 and IL-33 has previously been reported to expand a synthetic Tox^NEG effector phenotype in transferred OT-I TCR tg T cells.⁴² Specifically, Corria-Osario et al showed that an IL-15-like IL-2 variant, IL2v, promotes the expansion of TCF1+ progenitors and proposed that IL-33 acts cell extrinsically to induce differentiation of progenitors into Tox^neg PD-1+ effectors with enhanced antitumor efficacy.⁴² We have extended these findings by demonstrating a CD8 T-cell intrinsic requirement for IL-33 signaling in driving antitumor CD8 T-cell effector responses. Future studies are needed to determine the mechanism for how ST2 deficiency limits effector function and whether effector cells subsequently differentiate into exhausted cells.

Despite robust initial responses, only a fraction of combination-treated mice had prolonged tumor control. In vivo CAR T-cell tracking revealed that CAR T-cell persistence directly correlated with durable tumor control. Multiple factors could contribute to CAR T-cell loss. Trajectory analysis predicted that the progenitor subset, which expressed Tcf7 and Klf2, transcription factors that maintain stem-like properties or prevent exhaustion and support continued effector function,^{35 43} forms the root of both exhausted and effector subsets. Loss of progenitor cells due to excessive differentiation into effector subsets prior to tumor elimination may result in loss of tumor control. It is also possible that the three Tox^neg effector CD8 T-cell subsets identified have distinct activity and persistence within the tumor. The effector memory subset, which did not correlate strongly with any one reference cell state,^{27 28 33 44} expressed genes associated with both stemness and checkpoint receptors but lacked granzyme expression. In contrast, the effector terminal cluster, enriched in Gzmb and Gzma transcripts, correlated with the “effector memory” state within ProjectTILs. Our effector terminal cluster expressed Klf2, a transcription factor that prevents exhaustion and supports continued effector function. Future studies are needed to assess how CD8 T-cell subsets contribute to or are required for tumor control. STR to combination treatment may selectively lose or never develop critical effector subsets. Factors that induce progenitor maintenance and selective differentiation into specific effector subsets could inform ACT therapies that resist exhaustion.

Our study highlights the importance of evaluating ACTs in the context of pre-conditioning regimens. While Sup2-IL33 armored CAR T cells mediated tumor control in both non-conditioned and pre-conditioned hosts, the mechanism, durability, and extent of tumor control differed greatly. Pre-conditioning increased engraftment and in combination with Sup2-IL33 armoring increased effector function of adoptively transferred CAR T cells. Future studies that determine how pre-conditioning supports effector T-cell differentiation will be important to fully understand mechanisms by which combination therapy synergizes to extend antitumor immunity. We identified a T cell-intrinsic requirement for IL-33 but not type I IFN signaling in sustaining CAR T cell-mediated tumor control against B16F10 melanoma. While irradiation is well recognized to induce type I IFN release, it also promotes broad systemic changes in inflammatory cytokines, which may impact both host and transferred T-cell responses.

A limitation of our study is the inability to definitively distinguish CAR T from host CD8 T cells in the scRNAseq dataset. However, flow cytometric analysis indicated that the majority of tumor-infiltrating CD8 T cells in non-conditioned and pre-conditioned mice were host and CAR T cells, respectively. CAR T-cell phenotype may be impacted by in vitro culture conditions, including continuous expression and response to IL-2 and IL-33.^{45 46} Therefore, the observed differences in trajectory could be a result of differences in the source and environment of CD8 T cells. These experiments were also performed exclusively with TRP1-specific CAR T cells in the context of B16F10 melanoma. Future studies that assess the efficacy in other tumor models may help elucidate common changes in CD8 CAR T-cell phenotype necessary for enhanced tumor control.

Data availability statement

Data are available upon reasonable request.