TBK1 Targeting is Identified as a Therapeutic Strategy to Enhance CAR T-Cell Efficacy Using Patient-Derived Organotypic Tumor Spheroids

Abstract

Novel therapeutic strategies are needed to improve the efficacy of chimeric antigen receptor (CAR) T cells as a treatment of solid tumors. Multiple tumor microenvironmental factors are thought to contribute to resistance to CAR T-cell therapy in solid tumors, and appropriate model systems to identify and examine these factors using clinically relevant biospecimens are limited. Here, we examined the activity of B7-H3–directed CAR T-cells (B7-H3.CAR-T) using 3D microfluidic cultures of patient-derived organotypic tumor spheroids (PDOTS) and then confirmed the activity of B7-H3.CAR T-cells in PDOTS. While B7-H3 expression in PDOTS was associated with B7-H3.CAR-T sensitivity, mechanistic studies revealed dynamic upregulation of co-inhibitory receptors on CAR T-cells following target cell encounter that led to CAR T-cell dysfunction and limited efficacy against B7-H3–expressing tumors. PD-1 blockade restored CAR T-cell activity in monotypic and organotypic tumor spheroids with improved tumor control and upregulation of effector cytokines. Given the emerging role of TANK-binding kinase 1 (TBK1) as an immune evasion gene, we examined the effect of TBK1 inhibition on CAR T-cell efficacy. Similar to PD-1 blockade, TBK1 inhibition restored CAR T-cell activity in monotypic and organotypic tumor spheroids, prevented CAR T-cell dysfunction, and enhanced CAR T-cell proliferation. Inhibition or deletion of TBK1 also enhanced sensitivity of cancer cells to immune-mediated killing. Taken together, our results demonstrate the feasibility and utility of ex vivo profiling of CAR T cells using PDOTS and suggest that targeting TBK1 could be used to enhance CAR T-cell efficacy by overcoming tumor-intrinsic and -extrinsic resistance mechanisms.

INTRODUCTION

Chimeric antigen receptor (CAR) T-cell therapy has improved clinical outcomes in select hematological malignancies, providing durable remissions in a subset of patients (1,2). The use of CAR T-cell therapy for solid tumors is a field of huge potential from which a few promising results are beginning to emerge (3,4). However, the majority of solid tumor–directed CAR T-cell therapy clinical trials in recent years have failed to demonstrate clinical efficacy despite promising in vitro and/or in vivo data (5). While resistance to CAR T-cell therapy is understood in select cases (e.g., immune escape due to loss of tumor antigen expression) (6), emerging data suggests a more substantial barrier to CAR T-cell success is the immunosuppressive tumor microenvironment (TME), which can limit CAR T-cell infiltration and effector function (7–9). To date, the specific factors in the TME that limit the function of CAR T cells in solid tumors remain poorly understood, representing a major barrier to the successful deployment of CAR T cells in the treatment of solid tumors.

CAR T cells targeting B7-H3 (B7-H3.CAR-T) have been assessed preclinically both in vitro and in vivo against several solid tumor models (10–12). These findings have resulted in the initiation of several clinical trials, demonstrating the feasibility of targeting this antigen in humans (13). Results thus far however have not provided compelling evidence that CAR T cells targeting B7-H3 fare any better than other solid tumor–directed CAR T cells as they likely succumb to T-cell dysfunction. Understanding these limitations in CAR T-cell function and finding mechanisms to overcome them is expected to improve this therapeutic approach for patients with solid tumors.

Given the need to understand the contributions of TME and patient-specific factors to the heterogeneity of clinical responses to CAR T-cell therapy, there has been renewed interest in developing methods to study tumor–immune dynamics in a patient-specific manner (14,15). While several 3D patient-derived tumor models have been described (16–18), available model systems that recapitulate key features of the TME are more limited. Patient-derived organotypic tumor spheroids (PDOTS) contain tumor cells and autologous immune and stromal cells, and recapitulate key features of the TME, thereby enabling the study of tumor–immune dynamics and providing a useful model to study the effects of the TME on T-cell dysfunction (19–21). Based on our previous experience utilizing PDOTS to assess immunotherapies, and our observations that modulation of the TME within this system can affect the functionality of said therapies (19,20), we sought to utilize this methodology to explore the activity of solid tumor–directed CAR T cells.

Here, we demonstrate antitumor activity of B7-H3.CAR-T using PDOTS derived from solid tumor explants. We confirm the importance of target antigen (B7-H3) expression for CAR T-cell efficacy and identify dynamic alterations in CAR T cells in the model TME. We show the utility of PDOTS in examining combinatorial immunotherapeutic strategies to enhance CAR T-cell functionality and show that TANK-binding kinase 1 (TBK1) inhibition can enhance CAR T-cell efficacy by preventing CAR T-cell dysfunction and by lowering the cytotoxicity threshold of cancer cells to effector cytokines.

MATERIALS AND METHODS

CAR T-cell generation.

B7-H3.CAR-T were generated as described previously (11). Briefly, the scFv from the B7-H3 376.96 mAb was introduced into a CAR construct containing a human CD8α hinge and transmembrane domain, a CD28 intracellular costimulatory domain, and a CD3ζ intracellular signaling domain. The B7-H3.CAR cassette was cloned into the retroviral vector SFG and provided to us by Dr. Gianpiettro Dotti (University of North Carolina at Chapel Hill, Chapel Hill, NC). Retroviral supernatant was generated by transfecting HEK293T cells (RRID:CVCL_0063) with a plasmid mixture (the B7-H3.CAR retroviral vector, the Peg-Pam-e plasmid encoding MoMLV gag-pol, and the RDF plasmid encoding the RD114 envelope, which were kindly provided by Dr. Dotti) using the GeneJuice transfection reagent (Merck Millipore). Viral supernatant was collected 48 and 72 hours after transfection. Peripheral blood mononuclear cells (PBMCs) were obtained from normal healthy adult donors from Research Blood Components, LLC (Watertown, MA). IRB-approved written consent was obtained (by Research Blood Components, LLC) from each donor. All studies were conducted in accordance with the Declaration of Helsinki. PBMCs were isolated from unpurified buffy coats using density gradient centrifugation (Lymphoprep, STEMCELL, #07851) and were subsequently plated in a 24-well nontreated tissue culture plate (Thermo Fisher, catalog #142475) precoated with anti-CD3 (1 μg/mL, Miltenyi, #130–093-377) and anti-CD28 (1 μg/mL, BD Bioscience, Cat# 555725, RRID: AB_396068) to induce T-cell activation. Activated T cells were transduced with the retroviral supernatant using retronectin-coated plates (Takara Bio Inc). T cells were harvested after three days and expanded in complete medium (45% RPMI-1640 (Corning) 45% Click’s medium (Irvine Scientific), 10% FBS (Gemini Bio BenchMark, Cat. No. 100–106-500), 2 mM GlutaMAX (Gibco, 252030–081), 100 unit/mL of Penicillin and 100 mg/mL of streptomycin (Gibco, 15140–122) with IL-7 (10 ng/mL; PeproTech) and IL-15 (5 ng/mL; PeproTech). CAR transduction efficiency was assessed by flow cytometry using the rhB7-H3-Fc chimera protein (R&D Systems, Cat. No. 1027-B3), 1:20 dilution and goat-anti-human IgG-APC antibodies (Jackson ImmunoResearch Labs Cat# 115–136-146, RRID:AB_2338651) 1:100 dilution. CD19.CAR were generated in the same manner using a CD19 scFV, as previously described (22).

Human cancer cell lines.

Human intrahepatic cholangiocarcinoma cell line ECC4 (previously identified as ICC3; DepMap ID ACH-001843, RRID:CVCL_A1ZB) was generated from a patient sample collected by Dr. Cristina R. Ferrone in collaboration with Dr. Nabeel Bardeesy at Massachusetts General Hospital (IRB protocol DF/HCC 02–240), as previously described (23). Human triple negative breast cancer cell line SUM159 (RRID:CVCL_5423) was a gift from Dr. Donald J. Buchsbaum (University of Alabama at Birmingham) (24). Cell lines were stably transfected with a lentiviral vector encoding luciferase (Luc) and green fluorescent protein (GFP). ECC4-GFP cells were cultured in RPMI1640 medium (Corning) supplemented with 10% FBS (Gemini Bio BenchMark, Cat. No. 100–106-500) and 1% pen/strep (glico). SUM159-GFP cells were cultured in Ham’s F-12 medium (Gibco) supplemented with 10% FBS. B7-H3 knockout of SUM159 was performed using a CRISPR/Cas9 CD276 Gene Knockout Kit v2 (Synthego, CA), sgRNA sequences: C*U*C*ACAGGAAGAUGCUGCGU; G*C*A*CUGUGGUUCUGCCUCAC; G*G*G*ACAGUGAUUGUGGCAGU. 10101 (RRID: CVCL_E3EB) and M160 (RRID: CVCL_E3EC) cell lines were generated from melanoma patient samples at MGH (IRB protocol DF/HCC 11–181). Cell line authentication was performed by SNaPshot sequencing (25). Melanoma cell line M160 was confirmed to have mutations in ATM, NRAS, CDKN2A, TERT and TP53, and melanoma cell line 10101 was confirmed to have mutations in TP63, TP53, ERBB3, MPL, ATM, NRAS, MET, CDK4, BRAF, APC, TERT and TSC2. The cells were cultured in RPMI1640 medium containing 10% FBS and 1% Pen/Strep. All cells were cultured at 37°C in a 5% CO₂ atmosphere. Time in culture for all cell lines was 4–6 weeks. Lentiviral vector-encoding green fluorescent protein (GFP) was introduced, as previously described (12), by transduction with viral supernatant produced by transfection of HEK293T cells with the pMD2.G plasmid (envelope; RRID: Addgene_12259), the psPAX2 plasmid (gag-pol; RRID: Addgene_12260) and the lentiviral vector encoding the GFP construct. All cell lines were authenticated and tested negative for mycoplasma and interspecies contamination using the Mycoplasma detection kit (Lonza, #LT07–318).

2D cytotoxicity assay.

Tumor cell lines were plated in 96-well plates at 3×10⁴ cells/well. CAR T cells were added at the indicated effector:target (E:T) ratios and incubated for between 24 to 96 hours at 37°C in a 5% CO₂ atmosphere. Residual tumor cells in the co-culture were assessed by thiazolyl blue tetrazolium bromide (MTT) assay (Sigma).

Patient specimens.

Patient tumor samples (n=29 total) were collected and analyzed according to Dana-Farber/Harvard Cancer Center (DF/HCC) Institutional Review Board (IRB)-approved protocols. Written informed consent was obtained from all subjects. A cohort of patients (Supplemental Tables S1-3) treated at Massachusetts General Hospital was assembled for PDOTS profiling. Samples were collected between January 2021 and March 2023. Explanted tumors were placed in 50-mL conical tubes in complete DMEM (chilled) and stored at 4°C. All samples processed within 24 hours of surgical resection. These studies were conducted according to the Declaration of Helsinki and approved by the DF/HCC IRB.

Flow Cytometric Assessment of Human Cell Lines and Patient Tumors.

Single-cell suspensions from tumor cell lines or patient samples were stained with Zombie NIR (Biolegend, Cat# 423105) in 1x PBS at 1:500 for 30 min on ice in dark. Following one wash with 1x PBS, the cells were stained with human Fc blocker (BioLegend Cat# 422302, RRID:AB_2818986) at 1:100 and multiple fluorescent antibodies in different panels at 1:100 dilution in 100 μL of flow buffer (0.4% FBS in PBS) at 4°C 40 min in dark. Cells were washed once with flow buffer, then fixed and permeabilized with Foxp3/Transcription Factor staining set (Invitrogen, 00–5523-00) following the manufacturer’s instructions. After fixing, the cells were washed once with 1x permeabilization buffer and further stained with anti-Foxp3-PE (Thermo Fisher Scientific Cat# 12–4777-42, RRID:AB_1944444), and/or anti-TCF7-Alexa647 (BioLegend Cat# 655204, RRID:AB_2566620) in 1x permeabilization buffer 1:100 dilution at 4°C 30 min in dark. Cells were then washed once with flow buffer and assessed with Cytek™ Northern Lights flow cytometry (Cytek). The data were analyzed with FlowJo software v10.10 (RRID:SCR_008520). The following antibodies were used in the experiments: anti-B7-H3-APC (BioLegend Cat# 135607, RRID:AB_2566062), anti-CD45-AF700 (BioLegend Cat# 368514, RRID:AB_2566374), anti-CD8-BV785 (BioLegend Cat# 301045, RRID:AB_11219195), anti-CD4-BV605 (BioLegend Cat# 317437, RRID:AB_11204077), anti-CD25-BV650 (BioLegend Cat# 302633, RRID:AB_11203536), anti-CD11b-PerCP (BioLegend Cat# 101230, RRID:AB_2129374), anti-CD39-APC-Cy7 (BioLegend Cat# 328226, RRID:AB_2571981), anti-TIM3-FITC (BioLegend Cat# 345021, RRID:AB_2563936), anti-CD14-PE-Cy5 (BioLegend Cat# 301863, RRID:AB_2860766), anti-CTLA-4-PE-Dazzle594 (BioLegend Cat# 349922, RRID:AB_2566198), anti-LAG-3-BV510 (BioLegend Cat# 369318, RRID:AB_2715781), anti-PD-1-BV421 (BioLegend Cat# 329919, RRID:AB_10900818), anti-CD3-Pacific Blue (BD Biosciences Cat# 558117, RRID:AB_397038), anti-PD-L1-PE (BD Biosciences Cat# 568722, RRID:AB_3572255), anti-CD66B-PerCP-Cy5.5 (BD Biosciences Cat# 562254, RRID:AB_11154419); and anti-CD15-PE-Cy5.5 (Biorbyt Cat# orb910738, RRID:AB_3572256).

Multivariate linear regressions analysis.

A multivariate linear regression was performed with R Project for Statistical Computing (v4.3.2, RRID:SCR_001905). Features used in the model included select immune cell proportions from flow cytometry panels for each sample as well as B7-H3 tumor cell proportions from a flow cytometry panel and B7-H3 and PD-L1 mean fluorescence intensity (MFI) values for each sample. All variables included in the model are shown in the horizontal axis of the model results visualization. A multiple linear regression was fitted to predict PDOTS response (% change from control) to following CAR T-cell therapy. Visualization of the slope estimates and p-values for the model was performed with Python Programming Language (v3.8.16, RRID:SCR_008394), MatPlotLib (v3.5.2, RRID:SCR_008624) and seaborn (v0.12.0, RRID:SCR_018132). 0.05 was the p-value threshold used to determine significance.

Spheroid Preparation and Microfluidic Culture with CAR T cells.

Cancer cell line–derived monotypic tumor spheroids (MTS) and PDOTS were prepared and cultured, as previously described (19,20). MTS/PDOTS were prepared in hydrogels composed of rat tail type I collagen (Corning, Cat#:35426) at a final concentration of 1.7mg/mL. Spheroid–collagen mixtures (10 μL) were loaded into the center gel region of the 3D microfluidic culture device (Dax-01, AIM Biotech), and after incubation (30 minutes at 37°C, 5% CO₂) in sterile humidity chambers. B7-H3.CAR-T were added into one of the side channels in the device, then collagen hydrogels containing tumor spheroids/PDOTS were hydrated with media (RPMI supplemented with 10% FBS and 1% penicillin-streptomycin), in presence or absence of the indicated treatments. For 1–3 × 10⁴ target cells (MTS/PDOTS) per device 1.7–5.1 × 10⁴ CAR T-cells were added (accounting for ~70% transduction efficiency) into one of the side channels in the microfluidic device. Drug treatments included pembrolizumab (250 μg/mL) (Merck Sharp&Dobine, NDC 0006–3026-01), TBK1i (Compound 1; SelleckChem, S8922 (19)), TBK1 PROTAC 3i (Bio-techchne/TOCRIS, 7259), and ruxolitinib (MedChemExpress, HY-50856). For cytokine/neutralization studies, the following antibodies were used: anti-human IFNγ (BioXcell, cat#: BE0235), anti-human TNFα (BioXcell, cat#: SIM0001), anti-human CCL2 (BioXcell, cat#: BE0185), anti-human CXCL10 (R&D Systems, cat#: MAB266), anti-human IL-8 (R&D Systems, cat#: MAB208), anti-human CCL3 (R&D Systems, cat#: AF270-NA), and anti-human CCL4 (R&D Systems, cat#: AF-271-NA). Antibodies (10 μg/mL) were added 30 minutes before MTS co-cultures with CAR T cells.

Viability assessment of tumor spheroids and PDOTS.

Dual label fluorescence live/dead staining was performed using Hoechst/propidium iodide (Ho/PI) staining solution (Nexcelom, CSK-V0005) or acridine orange/propidium iodide (AO/PI) Staining Solution (Nexcelom, CS2–0106) (19,20). Images were obtained following incubation with the stain solution (30 min, 37°C, 5% CO₂), image capture and analysis were performed using a Nikon Eclipse NiE fluorescence microscope equipped with motorized stage (ProScan) and ZYLA4.2 Plus USB3 Camera (Andor) and NIS-Elements AR software package (Nikon). Live and dead cells were quantified by measuring total raw cell area for each dye. Percent change and log2FC (L2FC) data were generated using raw fluorescence data (live) for given treatments relative to control conditions. For GFP-labeled tumor cells, tumor-specific GFP signals were quantified along with Ho/PI imaging.

Cytokine Profiling.

Multiplexed analysis of secreted cytokines was performed by a bead-based immunoassay approach using the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (Millipore Sigma, Cat# HCYTMAG-60K-PX30). Conditioned media (25 μL) from cell line-derived tumor spheroids and PDOTS were assayed neat. The assay plate was run on MAGPIX with xPONENT software (Millipore). Concentration levels (pg/mL) of each protein were derived from 5-parameter curve fitting models. Fold changes relative to control samples were calculated and plotted as L2FC. Lower and upper limits of quantitation (LLOQ/ULOQ) were imputed from standard curves for cytokines above or below detection.

Normalized growth rate inhibition measurements.

Studies were performed as previously described (21). Briefly, cell lines were seeded with RPMI complete medium (RPMI with 10% FBS and 1% penicillin-streptomycin) in 384-well Cell Carrier plates (Perkin Elmer) and cultured for 24 h before treatment. 10101 and M160 cells were plated at 1,000 cells/well and treated with a half-log, single agent, dilution series of TNFɑ (Cat#: 10291-TA-100, R&D Systems) or with a half-log dilution series of TNFα (R&D Systems) 0.005 – 500 ng/mL plus IFNγ (Cat#: 285-IF100, R&D Systems) 0.001–125 ng/mL in combination. After 72h, cells were stained with LIVE/DEAD Far Red Dead Cell Stain (LDR, 1:5,000) (Thermo Fisher Scientific) and Hoechst 1 μg/mL (33342, Sigma-Aldrich). Cells were then fixed with 4% formaldehyde (Sigma-Aldrich) and imaged with a 10× objective on an ImageXpress confocal microscope (Molecular Devices). MetaXpress (RRID:SCR_016654) software was used to segment nuclei on the basis of their Hoechst signal, and the LDR intensity was used to classify cells as total and dead. Live-cell counts were normalized to DMSO-controls (26). Experiments were performed in triplicate. Cells were treated with TNFα and IFNγ prepared in PBS containing 0.05% Tween-20 (for dispensing aqueous solutions) and dispensed using an HP D300e Digital Dispenser (HP, Palo Alto, CA) (the final concentration of Tween-20 was < .0008%). TBK1i (SelleckChem, S8922 (19)) was prepared as a 10 mM stock in DMSO and dispensed in half-log dilution series using the HP D300e dispenser 2 h before the addition of cytokines.

Analysis of B7-H3 expression in cancer cells.

For Supplementary Figure S1A, cell line annotations and transcript per million (TPM)-normalized bulk RNA sequencing data (21Q4 v3) from The Cancer Dependency Map (DepMap) (27) were downloaded from the DepMap portal (https://depmap.org/portal/download/all/) (RRID:SCR_017655). TPM values were log2 transformed with an offset of 1. Cell lines were grouped according to ‘primary disease type’ annotation. Skin cancer was subgrouped into melanoma, Merkel cell carcinoma and squamous cell carcinoma. Groups with < 3 cell lines in the dataset were filtered out. Boxplots were generated using ggplot2 (RRID:SCR_014601) (28). Data handling and visualization was performed in R version 4.2.3. For Supplementary Figures S1B-D, the publicly available single-cell genomics dataset provided by Hodis et al. (32), available on the Broad Single Cell Portal website (https://singlecell.broadinstitute.org/single_cell) detailing melanocytes engineered to express cancer-associated mutations was analyzed. Briefly, mutations were introduced to primary human epidermal melanocytes in vitro by repetitive cycles of electroporation and selection. Cells grown in vitro were processed with 10x Genomics Single Cell 3’ v3. For Supplementary Figures S1E-G, the publicly available single-cell genomics dataset provided by Jerby-Arnon et al. (33) available on the Broad Single Cell Portal website (https://singlecell.broadinstitute.org/single_cell) in which malignant cells from resected tumors from patients with metastatic melanoma (n=14) were analyzed.

Incucyte tumor killing assays.

ECC4-GFP cells were seeded into an ultra-low attachment (ULA) dish (Corning, Cat#: 3471) (6×10⁶ cells/dish) for 24 hours (37°C 5% CO₂). Monotypic tumor spheroids (MTS, 40–100 μm) were isolated by sequential filtration, pelleted, and resuspended in 1.7 mg/mL collagen Type I (Corning, Cat#: 35426) in 1x PBS. MTS were seeded in 96-well plates (1×10⁵/well in 50 μL). After incubation (30 min at 37°C, 5% CO₂), CAR T cells in RPMI complete medium (2×10⁵/well in 150 μL) were added with Incucyte Cytotox NIR (1:1000) (SARTORIUS, Cat#: 4846), plus or minus pembrolizumab (250 μg/mL) or TBK1 inhibitor (1 μM). 10101 melanoma cells (5 X 10⁴ cells/well of 96-well plate) were co-cultured with 5 × 10⁴ B7-H3.CAR-T per well with or without pembrolizumab (250 μg/mL) or TBK1i (Compound 1, 1μM). The CAR T cells were pre-stained with 0.5 μM CellTracker Green CMFDA (Invitrogen, Cat#: C7025) in RPMI medium without serum at 37°C, 5% CO₂ incubator for 30 minutes, then the CellTracker stain solution was removed by 2 washings and resuspended in complete RPMI medium. The co-cultured cells were monitored every hour with a 20x objective using the Incucyte Zoom system (Essen Bioscience), housed in a cell culture incubator at 37°C, 5% CO₂, and set to take 9 images per well at each time point for 3 days. The ECC4-GFP or pre-stained CAR T cells were quantified by Zoom software (Essen Bioscience) and analyzed using GraphPad Prism (RRID:SCR_002798). The end-point data were statistically analyzed with one-way ANOVA.

B7-H3.CAR-T growth and viability assays.

Human melanoma cells (10101) were plated in a 96-well plate (8,000 cells/well) in 10% FBS/RPMI supplemented with 1% penicillin-streptomycin and cultured for 24h. Then, supernatants were taken from each well and B7-H3.CAR-T were added into the 10101 containing wells and clear wells (10,000 cells per well in 200μL complete RPMI media) with or without anti-PD-1 (pembrolizumab 250 μg/mL) or TBK1i 1μM). After 3 days incubation at 37^oC, 5% CO_2, the CAR T-cell containing supernatants were transferred into a new plate and an equal amount of reagent of CellTiter-Glo kit (Promega, Cat#: G7570) was added. The plates were read on a Cytation 5 plate reader (BioTek) for luminescence after 15 minutes incubation in the dark.

Nucleofection with Synthetic sgRNA/Cas9 Protein CRISPR Complexes to generate TBK1 Knockout CAR T cells.

Synthetic guide RNAs (sgRNA) for TBK1 (UCUCUCAUUUGAACAUCCAC) or control sgRNA (GUAUUACUGAUAUUGGUGGG) (30 pmol) (Synthego, CA) with recombinant Cas9 protein (15 pmol) Gene Knockout Kit v2 (Synthego, CA) in a total volume of 10 μL OptiMEM (Gibco, Cat#:31985–062) were incubated at room temperature for 10–15 minutes to allow ribonucleoprotein (RNP) complex formation. Cells (5–10 × 10⁵) in 10 μL OptiMEM were mixed with 10 μL of guide sgRNA/Cas9 RNP complexes. Mixtures were transferred to a 16-well Nucleofector Cuvette Holder and nucleofection was performed using the Lonza 4D Nucleofector System (Lonza, MA). Following nucleofection, 400 μL of cell culture medium was added to each well and the samples were transferred to 6-well plates. Finally, 4 mL of complete medium was added to each well and cells were incubated for 72 hours (37°C, 5% CO₂) to allow gene editing. The same protocol was used for generating TBK1-KO CAR T cells and 10101 TBK1 KO melanoma cells.

Western blot analysis.

Human melanoma 10101-TBK1 KO cells (1×10⁶) or B7-H3.CAR-T.TBK1 KO cells (1×10⁶) and the parental cells were plated in six-well plates and cultured for 24h. Cells were lysed in 1x RIPA lysis buffer (Millipore, Cat#:02–188) supplemented with Halt Protease Inhibitor/Phosphatase Inhibitor (Thermo Fisher, Cat#:1861281). The total protein concentration of the lysates was determined using a Pierce BCA Protein Assay Kit (Thermo Fisher), and proteins (20 μg/well) were run on a 6–12% Bis-Tris gel (Invitrogen, NP0335BOX). Proteins were then transferred from the gel to a PVDF membrane with iBlot2 system (Invitrogen) and blotted the membrane with primary anti-NAK/TBK1 antibody (1:1,000) (EP611Y) (Abcam Cat# ab40676, RRID:AB_776632) for overnight incubation, then stained the membrane with anti-rabbit fluorescent antibody (1:10,000) (LI-COR Biosciences Cat# 926–68021, RRID:AB_10706309). The membranes were imaged using an ODYSSEY Imaging system (LI-COR).

Single-cell RNA sequencing (scRNAseq) and analysis of CAR T cells and PDOTS.

For scRNAseq analysis of PDOTS +/− CAR T-cell challenge, B7-H3.CAR-T and cryopreserved PDOTS (melanoma, 10214) were thawed and rested in media to serve as baseline (0 hour) controls. Remaining PDOTS were prepared in collagen for 3D culture. B7-H3.CAR-T (E:T 1:1) were added with or without TBK1i (1 μM), 250 μg/mL pembrolizumab, or the combination, compared to untreated control for 24 hours before PDOTS +/− CAR T cells were collected for processing. Droplet-based isolation of single cells and subsequent library generation was performed with the Chromium Controller (10X Genomics) according to the manufacturer’s specifications using the 10x Chromium Next GEM single cell 5’ kit v2 (1000263). Characterization of the sequencing library was performed with the Bioanalyzer High Sensitivity DNA kit (Agilent; 5067–4626) and Qubit (ThermoFisher) instruments. Each pair of sample libraries were sequenced using Illumina NextSeq 500 with two paired-end reads (read1= 26; read2= 46). Read alignment and quantification was performed using CellRanger (v8.0.0) and a custom GRCh38 (Ensembl 112) human genome reference with the full CAR plasmid construct included. All samples were deemed high quality, based on CellRanger output metrics including estimated number of cells, mean reads per cell, mean genes per cell, and reads mapped confidently to transcriptome. Samples were aggregated, filtered, and normalized using Cumulus (v1.4.3) (29), removing barcodes with genes less than 150 or greater than 5000 genes. Barcodes containing > 20% mitochondrial genes were also filtered. Counts were normalized to 10,000 per cell and then log transformed. Downstream processing was performed using Pegasus (v1.9.1). 2,000 highly variable genes were identified after excluding genes expressed in less than .05% of cells and were used for Principal component analysis (PCA). Subsequently, the top 50 principal components (PCs) were used to compute the k-nearest neighbors graph (k = 30). PCs and the affinity matrices from the computed kNN graphs were used for Leiden clustering. A cluster purity analysis was performed to further clean up data, removing a total of 1.4% of cells that had <5 neighbors belonging to the same cluster. Cells were subsequently re-clustered (resolution = .2) and manually annotated into 9 distinct clusters based on differentially expressed marker genes. Uniform manifold approximation and projection (UMAP) embeddings were calculated to visualize clusters. Further subclustering was performed in major cellular compartments (e.g. CAR T, myeloid, tumor) and annotated based on differentially expressed marker genes. The CAR T-cell cluster was characterized by assessing the expression levels of the CAR plasmid. The expression threshold was established by analyzing the bimodal distribution of CAR plasmid, setting the threshold at the lower end of the second peak. To compare the CAR T cells at baseline (0 hr) vs 24 hr after PDOTS co-culture, GSEA (v4.3.2) was run using a pre-ranked list of genes, where genes were ranked by the AUROC metric resulting from the differential expression analysis (via Pegasus) between the two conditions. To characterize pathway level changes within the tumor population, GSEA was similarly run using a pre-ranked list of genes, ranked by AUROC, comparing the M2 vs M1 tumor sub-populations as well as the untreated control (0hr) vs 24 hr conditions within the tumor compartment.

Statistical Methods, Data Analysis, and Software.

All graphs with error bars report mean ± s.e.m. values except where indicated. Statistical tests, number of replicates, and independent experiments are listed in the text and figure legends. GraphPad Prism (v10.1.0, RRID:SCR_002798) was used for basic statistical analysis and plotting (http://www.graphpad.com). Schematics were generated with BioRender (biorender.com) (RRID:SCR_018361) using a paid license.

Data availability.

The datasets generated and analyzed in this study are available upon request. Raw sequencing data are available through the NCBI database of Genotypes and Phenotypes (dbGap) (RRID:SCR_002709) with the accession number XXXX[Number to be provided before publishing online] and processed data are accessible through Gene Expression Omnibus (RRID:SCR_005012) with the accession number GSE277569.

RESULTS

Monotypic and organotypic tumor spheroids are sensitive to B7-H3.CAR-T

B7-H3 (which is encoded by CD276) is a tumor-associated antigen overexpressed in many cancer types (30,31). Examination of B7-H3 expression in cancer cell lines confirmed high expression in melanoma, bile duct cancer, pancreatic cancer, breast cancer, and colon cancer, with significantly lower expression in leukemia, myeloma, and lymphoma (Supplementary Fig. S1A). To evaluate upregulation of B7-H3 in malignant cells, we examined engineered melanoma cell lines in which various oncogenic mutations were introduced by step-wise CRISPR editing (32). Increased B7-H3 expression was observed after the introduction of oncogenic BRAF V600E mutation and was preserved with introduction of subsequent mutations, but it was barely detectable in normal melanocytes and melanocytes with loss of CDKN2A (Supplementary Fig. S1B–D). Examination of CD276 (B7-H3) expression by scRNAseq (33) confirmed broad expression across patients in melanoma (Supplementary Fig. S1E–F). Therefore, we sought to assess the functionality of B7-H3–targeted CAR T cells in solid tumors, including melanoma, bile duct cancer, and breast cancer.

Using a previously validated CAR construct (11), we prepared second-generation B7-H3–specific CAR T cells (B7-H3.CAR-T) containing a CD28 costimulatory domain using healthy donor CD3⁺ T cells. We observed that there were roughly equal proportions of CD4⁺ and CD8⁺ T cells, and CAR transduction efficiency was consistently over 60% (Supplementary Fig. S2A–D). We next confirmed in vitro activity of B7-H3.CAR-T against B7-H3–expressing and B7-H3–null cancer cells (Supplementary Fig. S2E) in conventional 2D co-culture assays, observing both time and dose (E:T ratio) dependent killing of ECC4-GFP (intrahepatic cholangiocarcinoma) and SUM159-GFP (triple negative breast cancer) cell lines (Supplementary Fig. 2F–G).

To confirm activity of solid tumor–directed CAR T cells in 3D microfluidic culture, we next examined the efficacy of B7-H3.CAR-T using cell line–derived monotypic tumor spheroids (MTS) derived from ECC4-GFP and SUM159-GFP cells (Fig. 1A). B7-H3.CAR-T added to the side (media) channel of microfluidic devices containing ECC4-GFP or SUM159-GFP tumor spheroids were able to infiltrate through the collagen matrix and eliminate tumor cells in a time-dependent manner, with maximal activity observed at 72–96 hours and with an E:T ratio of 1:1 (Supplementary Fig. S2H–I). Loss of tumor specific GFP expression corresponded with overall loss of tumor cell viability following B7-H3.CAR-T treatment (Supplementary Fig. S2H–I). Dose-dependent killing of B7-H3–expressing tumor spheroids was also observed, with an E:T ratio as low as 1:10 and peak activity at E:T 1:1 (Supplementary Fig. S2J). Further, killing of tumor spheroids was antigen specific, as substituting B7-H3.CAR-T for control CAR-T cells targeting the B cell–lineage antigen CD19 (CD19.CAR-T) did not impact tumor cell viability (Fig. 1B-C), and SUM159 tumor cells lacking B7-H3 expression (B7-H3–null) (Supplementary Fig. S2E) were insensitive to B7-H3.CAR-T (Fig. 1C). These results demonstrate target-specific, and time- and dose-dependent, elimination of B7-H3–expressing tumor cells/spheroids by B7-H3.CAR-T in 3D microfluidic culture.

Figure 1 |. Ex vivo profiling of CAR T cells using monotypic and organotypic tumor spheroids.

A, Schematic of co-culture of CAR T cells with cell line-derived (monotypic) tumor spheroids in 3D microfluidic culture with B7-H3.CAR T cells added to the side channel with tumor spheroids grown in collagen hydrogels in the center/gel region. B, Viability assessment of ECC4-GFP tumor spheroids in 3D microfluidic culture treated with B7-H3.CAR-T and CD19.CAR-T (E:T 1:1) compared controls (n=9, 3 independent experiments). Mean values (bars) and individual values (open circles) are shown. One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ***P<0.001, ****P<0.0001, ns = not significant). C, Viability assessment of SUM159-GFP or B7-H3-null (B7-H3.ko) SUM159 tumor spheroids in 3D microfluidic culture treated with B7-H3.CAR-T and CD19.CAR-T (E:T 1:1) compared to untreated controls (n=6, 2 independent experiments). Mean values (bars) and individual values (open circles) are shown. One-way ANOVA with Tukey’s multiple comparisons test (****P<0.0001, ns = not significant). D, Schematic of PDOTS processing, culture, and treatment. E, Waterfall plot for PDOTS (n=25, indicated tumor types) treated with B7-H3.CAR-T (E:T 1:1) compared to untreated control PDOTS. Mean values (bars) are shown. F-H, PDOTS viability assessment from patients with (F) colorectal carcinoma (CRC), (G) pancreatic neuroendocrine tumor (PNET), and (H) pancreatic adenocarcinoma (PDAC). Mean values (bars) and individual values (open circles) are shown. (n=3 biological replicates per condition, one-way ANOVA with Tukey’s multiple comparisons test; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns = not significant). I, Multiple linear regression summary for the association of PDOTS with ex vivo response to B7.H3-CAR-T with baseline tumor and immune features.

We next sought to determine the activity of B7-H3.CAR-T using PDOTS from primary tumor explants. PDOTS contain both tumor cells and autologous immune cells (19,20) providing a model system to evaluate the contribution of patient- and tumor-specific heterogeneity on the response to a given therapy in a model TME (21,34) (Fig. 1D, Supplemental Table S1). The PDOTS exhibited variable content and composition of immune cells, consistent with prior reports (19) (Supplementary Fig. S3A–B). PDOTS treated with B7-H3.CAR-T (E:T 1:1) exhibited reduced viability compared to untreated control, with 60% (15/25) of PDOTS exhibiting clear objective response (Fig. 1E). For select PDOTS, we examined multiple E:T ratios and included control CAR T cells (CD19.CAR-T) or non-targeting T cells (NTT) for comparison. We observed dose-dependent elimination of PDOTS by B7-H3.CAR-T, but no significant effect of CD19.CAR-T or NTT on PDOTS viability (Fig. 1F-H).

Post hoc multivariate analysis of clinical or pathologic features, including tumor mutational profiles, failed to identify specific features explaining the patient-specific variability in the sensitivity of PDOTS to B7-H3.CAR-T (Supplementary Fig. S3C). We next performed multivariate analysis of PDOTS response data and multispectral immunophenotyping flow cytometry data to determine if PDOTS composition influenced B7-H3.CAR-T sensitivity. The only statistically significant association between PDOTS sensitivity to CAR T cells (indicated by negative slope) was % B7-H3–expressing tumor cells (CD45^–) (Fig. 1I). The proportion of baseline CD4⁺ and CD8⁺ cells was associated with response (negative slope), but was not statistically significant (negative log₁₀-p value <1.301). Enrichment of PD-L1⁺CD45⁺ cells, monocytes/macrophages (CD14⁺), and increased B7-H3 MFI were associated with lack of response to B7-H3.CAR-T cells, but also did not reach statistical significance. Univariate analysis confirmed elevated B7-H3 expression in PDOTS responsive (R; n=16) to B7-H3.CAR-T compared to those non-responsive (NR; n=9) (Supplementary Fig. S3D). The extent to which B7-H3 expression correlated with B7-H3.CAR-T efficacy was modest (R² = 0.3748) but statistically significant (P=0.0011) (Supplementary Fig. S3E). No significant difference in B7-H3 MFI was observed between responders and non-responders (Supplementary Fig. S3F). These results demonstrate that B7-H3.CAR-T are active in 3D microfluidic culture using monotypic and organotypic tumor spheroids and confirm the importance of target antigen expression in CAR T-cell efficacy, whilst also suggesting that factors other than target antigen expression contribute patient-specific variations in CAR T-cell efficacy.

PD-1 blockade enhances CAR T-cell efficacy in monotypic and organotypic tumor spheroids

Similar to tumor-infiltrating T cells, CAR T cells can become dysfunctional, or exhausted, leading to diminished efficacy (35,36). Immune checkpoint blockade with anti-PD-1 or PD-1 (PDCD1) deletion enhances CAR T-cell function in pre-clinical studies (37), and has been shown to be safe and effective in patients with solid tumors in early phase clinical trials (38). To determine if PD-1 blockade could enhance response to solid tumor–directed CAR T cells, we evaluated B7-H3.CAR-T activity in PDOTS alone or in combination with PD-1 blockade. B7-H3.CAR-T activity in PDOTS from a subset of our initial cohort of patients (Supplemental Table S2) was similar to that observed with the PDOTS from our initial patient cohort (Supplemental Table S1), with 5 of 7 PDOTS demonstrating a response (Fig. 2A-B). Anti-PD-1 alone demonstrated no responses (0/7), but inclusion of anti-PD-1 improved the response rate to B7-H3.CAR-T to 100% (7/7). Response to combination anti-PD-1 plus B7-H3.CAR-T was particularly potent in PDOTS 10214, derived from a patient with melanoma that was refractory to immune checkpoint blockade (ICB), with evidence of significant improvement of CAR-T activity with the addition of PD-1 blockade (Fig. 2C-D, Supplementary Fig. S4A). Addition of anti-PD-1 to B7-H3.CAR-T was also associated with greater induction of inflammatory cytokines, including effector cytokines (e.g., TNFɑ, IFNγ) (Fig. 2E).

Figure 2 |. PD-1 blockade enhances CAR T-cell activity in monotypic and organotypic tumor spheroids.

A, Waterfall plot for PDOTS (n=7, indicated tumor types) treated with anti-PD-1 (pembrolizumab, 250 μg/mL), B7-H3.CAR-T (E:T 1:1), or combined anti-PD-1 plus B7-H3.CAR-T. Mean values are shown. B, Violin plot of PDOTS viability assessment with the indicated treatments. Mean values (bars) and individual values (open circles) are shown. n = 21 biological replicates, 7 independent specimens. Statistical analysis was performed using one-way ANOVA with Dunn’s multiple-comparison test. C, Viability assessment of anti-PD-1-refractory cutaneous melanoma PDOTS (10214) after the indicated treatments. Mean values (bars) and individual values (open circles) are shown. n = 3. Statistical analysis was performed using one-way ANOVA with Dunn’s multiple-comparison test. D, Representative images of PDOTS viability evaluation for PDOTS 10214 after the indicated treatments (blue = Hoechst; green = acridine orange; red = propidium iodide). E, Heatmap depicting changes in secreted cytokines in PDOTS 10214 after the indicated treatments. F-H, Viability assessment of (F) M160-GFP melanoma spheroids, (G) 10101 melanoma spheroids, and (H) ECC4-GFP tumor spheroids, and (H) in 3D microfluidic culture after the indicated treatments. Mean values (bars) and individual values (open circles) are shown. n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ***P<0.001, ****P<0.0001, ns = not significant).

To determine if the restoration of CAR T-cell function observed in PDOTS could be observed in co-cultures lacking tumor-infiltrating immune cells, we examined the effect of PD-1 blockade using MTS cultures. Addition of an anti-PD-1 enhanced the efficacy of B7-H3.CAR-T in co-cultures with MTS derived from two melanoma cell lines (M160, 10101) and one cholangiocarcinoma cell line (ECC4-GFP) (Fig. 2F-H). These observations suggest the rapid development of CAR T-cell dysfunction in 3D microfluidic co-cultures with monotypic and organotypic tumor spheroids and confirm improved efficacy of CAR T cells in combination with PD-1 blockade.

Targeting TBK1 enhances CAR T-cell efficacy in monotypic and organotypic tumor spheroids

Activation of tumor intrinsic immune evasion pathways contributes to escape from cancer immunotherapy, whereas disruption of key immune evasion genes can enhance antitumor immunity (39–42). TBK1 is a multi-functional cytosolic serine/threonine-kinase with established roles in innate immunity, inflammation, and cell death signaling (43). We recently demonstrated that genetic deletion or pharmacological inhibition of TBK1 enhanced response to cancer immunotherapy both by sensitizing cancer cells to immune attack and by remodeling the tumor immune microenvironment (21). To determine if TBK1 inhibition could also enhance CAR T-cell efficacy, we examined sensitivity of PDOTS to a TBK1 inhibitor (TBK1i; Compound 1 (19)) +/− B7-H3.CAR-T. Single-agent activity of TBK1i was observed in 18% (2/11) of PDOTS, generated with a third subset of our initial cohort of patients (Supplemental Table S3), whilst response to B7-H3.CAR-T was observed in 63% (7/11) (Fig. 3A-B), similar to response rates observed previously for TBK1i (21) and for B7-H3.CAR-T alone (see Fig. 1–2). The addition of TBK1i enhanced the response to B7-H3.CAR-T with a significant decrease in PDOTS viability observed in 100% of cases (11/11) (Fig. 3A-B). Among the strongest responses to TBK1i plus B7-H3.CAR-T were PDOTS 10239 from a patient with uveal melanoma whose disease had progressed despite combination anti-PD-1/anti-CTLA-4 blockade and tumor-infiltrating lymphocyte (TIL) therapy (Fig. 3C). The combination of TBK1i plus B7-H3.CAR-T was also particularly active in PDOTS 10261 and 10250 from treatment-naive patients with pancreatic ductal adenocarcinoma (PDAC) and hepatocellular carcinoma (HCC), respectively (Fig. 3D-F, Supplementary Fig. S4B). Degradation of TBK1 via a proteolysis targeting chimera (PROTAC) (TBK1 PROTAC 3i) similarly enhanced the activity of B7-H3.CAR-T cells in PDOTS (Supplementary Fig. S4C–D). Of note, the addition of PD-1 blockade to TBK1 PROTAC 3i offered little improvement over anti-PD-1 or TBK1 PROTAC 3i alone (Supplementary Fig. S4E).

Figure 3 |. Targeting TBK1 enhances CAR T-cell activity in monotypic and organotypic tumor spheroids.

A, Waterfall plots for PDOTS (n=11, indicated tumor types) treated with TBK1i (1 μM), B7-H3.CAR-T (E:T 1:1), or combined B7-H3.CAR-T plus TBK1i. Mean values are shown. B, Violin plot of PDOTS viability assessment with the indicated treatments. Mean values (bars) and individual values (open circles) are shown. n = 33 biological replicates, 11 independent specimens. Statistical analysis was performed using one-way ANOVA with Dunn’s multiple-comparison test. C-E, Viability assessment of PDOTS from patients with (C) uveal melanoma (10239), (D) pancreatic adenocarcinoma (PDAC, 10261), and (E) hepatocellular carcinoma (HCC, 10250) after the indicated treatments. Mean values (bars) and individual values (open circles) are shown. n = 3. Statistical analysis was performed using one-way ANOVA with Dunn’s multiple-comparison test. F, Representative images of PDOTS viability evaluation for PDOTS 10250 after the indicated treatments (blue = Hoechst; red = propidium iodide). G-I, Viability assessment of (G) M160-GFP melanoma spheroids, (H) 10101 melanoma spheroids, and (I) ECC4-GFP tumor spheroids in 3D microfluidic culture after the indicated treatments. Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ***P<0.001, ****P<0.0001, ns = not significant). J, Incucyte viability assessment ECC4-GFP cholangiocarcinoma cells after the indicated treatments. Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments) +/− s.e.m. (shaded region). One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ***P<0.001, ****P<0.0001, ns = not significant). K, Representative images of ECC4-GFP cholangiocarcinoma cells after treatment with B7-H3.CAR-T alone or in combination with TBK1i in collagen hydrogels (green = GFP; blue = dead).

To confirm and extend these observations, we next examined the effect of TBK1i treatment on B7-H3.CAR-T efficacy using MTS co-cultures. TBK1 inhibition enhanced the efficacy of B7-H3.CAR-T cells in MTS co-cultures using M160 and 10101 patient-derived melanoma cell lines (Fig. 3G-H) and ECC4-GFP cholangiocarcinoma cells (Fig. 3I). We observed similar sensitization to B7.H3-CAR-T in 10101 melanoma spheroids and ECC4-GFP cholangiocarcinoma spheroids using TBK1 PROTAC 3i (Supplementary Fig. S4F–G). Incucyte analysis of ECC4-GFP MTS revealed enhanced tumor cell killing accompanied by accelerated loss of GFP-positive cancer cells following treatment with TBK1i plus B7-H3.CAR-T compared to B7-H3.CAR-T alone (Fig. 3J-K). Thus, targeting TBK1 is an effective strategy to enhance the activity of B7-H3.CAR-T in organotypic tumor spheroid culture, which can be recapitulated using MTS cultures.

TBK1 inhibition prevents CAR T-cell dysfunction

TBK1 has been shown to directly and indirectly regulate T-cell function (19,21,44,45), although the role of TBK1 in regulating CAR T-cell function has not yet been established. To determine if enhanced efficacy of B7-H3.CAR-T in organotypic and monotypic tumor spheroid cultures could be attributed (at least in part) to a CAR T-cell intrinsic effect of TBK1 inhibition, we performed multispectral flow cytometry on B7-H3.CAR-T before and after co-culture with cell line–derived MTS, examining the expression of co-inhibitory receptors and other cell surface markers associated with T-cell exhaustion (e.g., PD-1, CD39, TIM-3, and LAG-3) and FoxP3⁺CD25⁺ regulatory T cells denoting “CAR-Tregs” (46) (Fig. 4A). As we expected, baseline expression of PD-1 was low (1%) in B7-H3.CAR-T but increased following co-culture with MTS, with a significant increase observed in melanoma spheroids (10101, M160-GFP), but not in cholangiocarcinoma spheroids (ECC4-GFP) (Fig. 4B-D, Supplementary Fig. S5). However, upregulation of PD-1 expression was blunted in B7-H3.CAR-T when TBK1i was added to the 3D microfluidic co-culture with MTS, and the extent of inhibition was similar in co-cultures treated with pembrolizumab (anti-PD-1) (Fig. 4B-D, Supplementary Fig. S5). Pharmacologic TBK1 inhibition and anti-PD-1 treatment also blunted the upregulation of CD39 and TIM-3 (Fig. 4E-G, Supplementary Fig. S6), which have previously been shown to mark terminally dysfunctional exhausted T cells (47). We did observe that the extent of PD-1 versus CD39/TIM3 upregulation varied across cell lines examined, although consistent upregulation was observed in 10101 melanoma spheroids (Fig. 4B, 4E). Upregulation of LAG-3 was significantly and consistently upregulated in B7-H3.CAR-T cells following co-culture with all examined MTS, and TBK1i and anti-PD-1 treatment blunted LAG-3 upregulation (Fig. 4H-J, Supplementary Fig. S7). CAR-T expressing hallmarks of Tregs (FOXP3⁺CD25⁺) were also observed following co-culture with MTS, however both TBK1i and anti-PD-1 effectively reduced the generation of CAR-Tregs (Fig. 4K-M, Supplementary Fig. S8). These results demonstrate acute and dynamic upregulation of markers of CAR T-cell dysfunction and generation of immune suppressive cell states within days of encountering antigen-expressing tumor spheroids in 3D microfluidic culture. Together, the observations that PD-1 blockade (Fig. 2) and TBK1i treatment (Fig. 3) enhance B7-H3.CAR-T efficacy suggest that strategies that prevent the formation of immune suppressive and/or dysfunctional CAR T cells may limit the development of resistance to CAR T-cell treatment in solid tumors.

Figure 4 |. Targeting TBK1 prevents dysfunction of CAR T cells.

A, Schematic of co-culture of CAR T cells with cell line-derived (monotypic) tumor spheroids in 3D microfluidic culture with B7-H3.CAR-T added to the side channel with tumor spheroids grown in collagen hydrogels in the center/gel region. B-D, Flow cytometry of PD-1+ B7-H3.CAR T-cells before (control) and after co-culture with (B) 10101 melanoma spheroids, (C) M160-GFP melanoma spheroids, and (D) ECC4-GFP tumor spheroids in 3D microfluidic culture after the indicated treatments for 72 hours. E-G, Flow cytometry of CD39+TIM-3+ double positive B7-H3.CAR T cells before (control) and after co-culture with (E) 10101 melanoma spheroids, (F) M160-GFP melanoma spheroids, and (G) ECC4-GFP tumor spheroids in 3D microfluidic culture after the indicated treatments for 72 hours. H-J, Flow cytometry of LAG3+ B7-H3.CAR-T cells before (control) and after co-culture with (H) 10101 melanoma spheroids, (I) M160-GFP melanoma spheroids, and (J) ECC4-GFP tumor spheroids in 3D microfluidic culture after the indicated treatments for 72 hours. K-M, Flow cytometry of CD25+FOXP3+ B7-H3.CAR-T cells before (control) and after co-culture with (K) 10101 melanoma spheroids, (L) M160-GFP melanoma spheroids, and (M) ECC4-GFP tumor spheroids in 3D microfluidic culture after the indicated treatments for 72 hours. Mean values (bars) and individual values (open circles) are shown (n=3 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ***P<0.001, ****P<0.0001, ns = not significant).

TBK1 inhibition promotes CAR T-cell proliferation and effector function

As CAR T-cell dysfunction is associated with reduced proliferative potential, we next examined the proliferation of B7-H3.CAR-T cells before/after co-culture with cancer cells alone or in combination with TBK1i or anti-PD-1. Treatment with TBK1i and pembrolizumab (anti-PD-1) enhanced the proliferation of B7-H3.CAR-T in co-culture with 10101 melanoma cells assessed at 96 hours (Fig. 5A). To examine the kinetics of T-cell proliferation, we performed Incucyte analysis of fluorescently labeled B7-H3.CAR-T and observed a burst of T-cell proliferation between 24–48 hours in the presence of TBK1i whereas PD-1 blockade resulted in a more gradual, but steady increase in CAR T-cell proliferation up to 96 hours (Fig. 5B). Antitumor immune function of CD8⁺ T cells associated with response to immune checkpoint blockade and autologous TIL therapy is restricted to a specific T-cell state defined by the transcription factor, TCF7, and by absence of surface markers CD39 and TIM3 (47,48). Studies in CAR T cells have confirmed that TCF7-expressing CAR T cells are associated with improved clinical response and persistence of CAR T cells (49). Both TBK1i and anti-PD-1 treatment increased expression of TCF7 in B7-H3.CAR-T (Fig. 5C-D), consistent with reports that TCF7 expression defines a stem/memory-like cell state with continued proliferative potential (47,50). Genetic deletion of TBK1 in CAR T cells (Fig. 5E) increased TCF7 expression, mirroring the effect of pharmacologic TBK1 inhibition on TCF7 expression and CAR T-cell proliferation (Fig. 5F, Supplementary Fig. S8E–F). CAR T cell–specific TBK1 deletion also prevented accumulation of PD-1⁺, CD39⁺/TIM-3⁺, LAG-3⁺, and FOXP3⁺CD25⁺ CAR T-cell subpopulations (Fig. 5G-J, Supplementary Fig. S8G–H).

Figure 5 |. Targeting TBK1 promotes proliferation and effector function of CAR T cells.

A, CAR T-cell viability (Cell Titer Glo) after 96 hours co-cultured with 10101 melanoma cells in combination with anti-PD-1 or TBK1i compared to CAR T-cells alone (control). Mean values (bars) and individual values (open circles) are shown (n=3 independent experiments). B, Incucyte proliferation assessment of B7-H3.CAR-T cells labeled with CellTracker Green during co-culture with 10101 melanoma cells with the indicated treatments. Mean values (closed circles) are shown (n=6, 2 independent experiments) +/− s.e.m. (shaded region). One-way ANOVA with Tukey’s multiple comparisons test (*P<.05, ***P<0.001). C-D, Intracellular staining for TCF7 in B7-H3.CAR-T cells treated with TBK1i (C) and pembrolizumab (D). Mean values (bars) and individual values (open circles) are shown (n=2, 2 independent experiments). E, Western blotting for TBK1 levels in control sgRNA and TBK1 sgRNA B7-H3.CAR-T cells. F, Intracellular staining for TCF7 in control sgRNA and TBK1 sgRNA B7-H3.CAR-T cells. G-J, Flow cytometry of control and TBK1-null B7-H3.CAR-T cells after co-culture with 10101 melanoma spheroids examining surface expression of (G) PD-1⁺, (H) CD39⁺TIM3⁺, (I) LAG3⁺, and (J) CD25⁺FOXP3⁺. K-L, Concentrations of effector cytokines IFNγ (K) and TNFɑ (L) 3 days after co-culture with B7-H3.CAR-T cells with indicated treatments. Mean values (bars) and individual values (open circles) are shown (n=2 across two independent experiments; 1-sided unpaired t-test. M-N, Concentrations of secreted cytokines IFNγ (M) and TNFɑ (N) in 10250 PDOTS 5 days after co-culture with B7-H3.CAR-T cells with indicated treatments. Mean values (bars) and individual values (open circles) are shown (n=2 across two independent experiments); 1-sided unpaired t-test. O, Viability assessment of 10101 melanoma monotypic tumor spheroids in 3D microfluidic culture treated with control B7-H3.CAR-T cells (E:T 1:1), TBK1i-treated B7-H3.CAR-T cells, or TBK1-null B7-H3.CAR-T cells compared to untreated control. Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (*P<.05, **P<.01, ****P<0.0001).

In addition to limited proliferative capacity, dysfunctional T cells are characterized by reduced production of effector cytokines, such as IFNγ and TNFɑ (51). In co-cultures of B7-H3.CAR-T and 10101 melanoma cells, we observed increased levels of secreted IFNγ and TNFɑ in co-cultures treated with TBK1i (Fig. 5K-L). Similarly, in PDOTS (10250 HCC - see Fig. 3E-F), addition of TBK1i enhanced the production of IFNγ and TNFɑ (Fig. 5M-N). Consistent with these observations, TBK1-null CAR T cells demonstrated improved cytotoxicity against 10101 melanoma MTS cultures (Fig. 5O). Taken together, these results suggest that TBK1 restrains CAR T-cell proliferation and effector function and that CAR T cells lacking functional TBK1 are more proliferative, express higher TCF7 levels, are less prone to exhaustion, and possess enhanced effector function.

Single cell analysis of CAR-T/PDOTS 3D co-cultures

To more deeply characterize the dynamic changes of CAR T cells and PDOTS in 3D culture, we performed scRNAseq of PDOTS and B7-H3.CAR-T before (0 hours) and after (24 hours) 3D co-culture with TBK1i and/or anti-PD-1 treatment (Fig. 6A). As enhanced CAR T-cell proliferation and tumor cell killing were observed as early as 48–72 hours (see Fig. 5), we selected the 24-hour time point to capture early transcriptional changes before maximum tumor cell killing was observed. Aggregated data from each treatment group was used to perform clustering and to examine the relative abundance of lymphoid, myeloid, and tumor cell populations in baseline PDOTS compared with PDOTS exposed to CAR T cells +/− indicated treatments for 24 hours (Fig. 6B-C, Supplementary Fig. S9A–B). Melanoma PDOTS (10214) were found to consist of tumor cells, T cells, myeloid cells, and B cells with little change in cellular composition observed between 0 and 24 hours of 3D culture (Fig. 6D). In contrast, treatment with B7-H3.CAR-T resulted in changes in the proportion of tumor and immune cells, most notably with a reduction of tumor cells (Supplementary Fig. S9C).

Figure 6 |. Single cell RNA sequencing of CAR T cells in 3D co-cultures with PDOTS.

A, Schematic of conditions for baseline (0 hr) and 24 hr of PDOTS +/− B7-H3.CAR-T cells in 3D culture with indicated treatment conditions. B-C, Uniform Manifold Approximation and Projection (UMAP) of all tumor and immune cells (77,284 cells) from the conditions indicated (B) with 9 unique populations identified (C). D, Bar plots of proportional changes in unique tumor and immune cell clusters identified in C. E-F, UMAP of 28,227 cells from the conditions indicated (E) and 10 unique populations identified (F) among sub-clustered CAR-T cells. G-H, UMAP of 15,179 cells from the conditions indicated (G) and 5 unique populations identified (H) among sub-clustered cancer cells. I, Bar plots of proportional changes in unique tumor cell clusters identified in H. J, Dot plot for select genes differentially expressed in the 5 sub-clusters of cancer cells. K, tracks plots depicting the changes in B7-H3 (CD276) gene expression across the different cancer cell sub-clusters at baseline (0 hr) and after 24 hr +/− B7-H3.CAR-T cell challenge with/without TBK1i (1 μM), αPD-1 (pembrolizumab, 250 μg/mL), or PD-1 plus TBK1i.

Examination of B7-H3.CAR-T (identified by the presence of the full CAR plasmid and CD4 or CD8) demonstrated marked changes in transcriptional cell state following 3D co-culture with PDOTS, with upregulation of gene programs associated with cytotoxicity and effector function compared to baseline (Fig. 6E-F, Supplementary Fig. S9D–F). Differential gene expression analysis revealed upregulation of IL2RA, GZMB, TNFRSF18 (encoding GITR), TNFRSF4 (encoding OX-40), and LAG3, with concomitant downregulation of LTB, CD52, CD48, and KLF2 in B7-H3.CAR-T after 24 hours in 3D culture (Supplementary Fig. S9E–F). While upregulation of CTLA4 and ENTPD1 (CD39) were also observed in B7.H3-CAR-T, we did not observe a significant effect of TBK1i, anti-PD-1, or both on CAR T-cell composition or expression of TCF7, PDCD1 (encoding PD-1), HAVCR2 (encoding TIM3), ENTPD1 (encoding CD39), and FOXP3 at this early time point. As the maximal effect of PD-1 blockade and TBK1 inhibition on CAR T cells was observed after >36–48 hours (see Fig. 5B), the lack of significant transcriptional differences between the different treatment groups at 24 hours likely reflects the early time point of this analysis. Gene set enrichment analysis revealed enrichment of several hallmark pathways including TNFα/NFкB signaling, hypoxia, and IL-2/STAT5 signaling in B7-H3.CAR-T 24 hours after co-culture with PDOTS (Supplementary Fig. S9G). We also observed changes in TILs following B7-H3.CAR-T challenge, with an increased expression of early activated T cells and a relative decrease in the frequency of naive T cells (Supplementary Fig. S10A–C). Following B7-H3.CAR-T challenge, native TILs upregulated genes associated with early T-cell activation including IL2RA (encoding CD25), CD69, TNFRSF4, CD44, and GZMB, accompanied by decreased expression of markers associated with naive T cells, including SELL and IL7R (Supplementary Fig. S10D), confirming changes in transcriptional cell states of native TILs following CAR T-cell treatment.

Examination of tumor cells in PDOTS revealed a decreased proportion of cells in the M1 cluster (SOX10, MITF, TYR) and relative increases in the proportion of cells in the M2 (DUSP10, KDM5B, CXCL8) and M3 (PSMB9, STAT1, CD47, ISG15) clusters following 24 hours in 3D culture (Fig. 6G-J). In addition, CD276 (encoding B7-H3) expression was preserved in PDOTS after 24 hr 3D culture compared to baseline (0 hr). Moreover, while the overall number of remaining viable melanoma cells was reduced following B7-H3.CAR-T treatment in all conditions, residual expression of CD276 (B7-H3) in tumor cells was observed after 24 hours (Fig. 6K). This may reflect the early time point (24 hours) at which specimens were collected for scRNAseq analysis. However, given the relative enrichment of specific transcriptional cancer cell states (e.g., increased proportion of cluster M2 cancer cells) following 3D culture of PDOTS +/− B7-H3.CAR-T treatment (Fig. 6J), it is also possible that tumor-intrinsic transcriptional cell state changes contribute to CAR T-cell resistance independently of CD276 (B7-H3) expression. Gene set enrichment analysis revealed enrichment of several hallmark pathways including TNFα/NFкB signaling, IFNγ response, inflammatory response, hypoxia, and EMT in ‘M2’ cancer cells compared to ‘M1’ cancer cells (Supplementary Fig. S10E). Similar results were observed between the total cancer cell population at baseline (PDOTS, 0 hr) compared to the 24h non-treated cancer cells (PDOTS, 24 hr) (Supplementary Fig. S10F). These findings indicate that transcriptional cancer cell states associated with CAR T-cell resistance are present at baseline (0 hr), but are more evident in 3D culture and are further enriched following short-term CAR T-cell treatment.

IFNγ sensing is required for CAR T-cell efficacy

CAR T cells eliminate target tumor cells through release of cytolytic molecules and effector cytokines, although differences between solid tumors and hematologic malignancies have been observed (52). For example, IFNγ signaling is critical for the response of solid tumors to CAR T cells (53), whereas IFNγ is dispensable for the efficacy of CD19-targeted CAR T cells in hematologic malignancies (54). To define the role of specific effector cytokines in the elimination of cancer cells by CAR T cells, we performed bead-based cytokine profiling of conditioned media collected from ECC4-GFP MTS alone or treated with B7-H3.CAR-T. Increased secretion of IFNγ and TNFɑ was observed as early as 24 hours following the addition of B7-H3.CAR-T in co-culture with ECC4-GFP MTS with persistent levels up to 96 hours (Fig. 7A-B). To define the contribution of specific effector cytokines (TNFɑ, IFNγ) to CAR T-cell efficacy, we co-cultured ECC4-GFP MTS with B7-H3.CAR-T (E:T 1:1) alone or in combination with neutralizing antibodies against IFNγ (anti-IFNγ), TNFɑ (anti-TNFɑ), or both (anti-IFNγ plus anti-TNFɑ). Neutralization of IFNγ +/− TNFɑ treatment rescued cancer cells from CAR T cell–mediated tumor cell killing (Fig. 7C), demonstrating that IFNγ (and to a lesser extent TNFɑ) are required for MTS elimination by CAR T cells, which is consistent with the observations of Larson et al. (53). Neutralization of chemokines upregulated following B7-H3.CAR-T treatment (e.g., CXCL10, IL8, CCL2, CCL3, CCL4) had no effect on CAR T-cell efficacy alone or in combination (Supplementary Fig. S11A), suggesting that these chemokines do not participate directly in the cytotoxic effect of CAR T cells.

Figure 7 |. Targeting TBK1 sensitizes cancer cells to CAR T cell–derived TNFα/IFNγ.

A, Heatmap depicting changes in secreted cytokines in ECC4-GFP tumor spheroids treated with B7-H3.CAR-T (E:T of 1:1) compared to time-matched controls (L2FC of averages from 5 replicates across 2 independent experiments). B, Concentrations of effector cytokines IFNγ and TNFɑ at 48 hours. Mean values (bars) and individual values (open circles) are shown (n=5 across two independent experiments). Unpaired t-test (**P<0.01, ***P<0.001, ns = not significant). C, Cell viability assessment of ECC4-GFP tumor spheroids treated with B7-H3.CAR-T (E:T 1:1, 72 hours) alone or treated with anti-IFNγ (10 μg/mL), anti-TNFɑ (10 μg/mL), or anti-IFNγ plus anti-TNFɑ compared to untreated control. Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (****P<0.0001). D, Cell viability assessment of control and TBK1-null 10101 melanoma tumor spheroids treated with B7-H3.CAR-T (E:T 1:1, 48 hours) compared to untreated control. Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (**P<0.01, ***P<0.001, ns = not significant). E-F, Mean GR values (n=3 biological replicates) for (E) 10101 melanoma cells and (F) M160-GFP melanoma cells treated with TBK1i (n = 3) across TNFα/IFNγ concentrations. G-H, Viability assessment of (G) 10101 melanoma spheroids and (I) ECC4-GFP spheroids in 3D microfluidic culture treated with CD19.CAR-T or B7-H3.CAR-T (E:T 1:1, 48 hours) alone or with ruxolitinib (JAK1/2i, 0.5 μM). Mean values (bars) and individual values (open circles) are shown (n=6, 2 independent experiments). One-way ANOVA with Tukey’s multiple comparisons test (**P<.01, ****P<0.0001, ns = not significant). I, Viability assessment of PDOTS (10247, 10249, 10252, and 10261), with the indicated treatments. Mean values (bars) and individual values (open circles) are shown (n=12, 4 independent experiments).

Dynamic cytokine profiling confirmed upregulation of multiple effector cytokines and chemokines in three PDOTS responsive to B7-H3.CAR-T (Supplementary Fig. S11B–G). Importantly, a dose (E:T ratio) dependent upregulation of IFNγ and TNFɑ was observed in PDOTS (Supplementary Fig. S11B–C). However, cytokine/chemokine upregulation also was observed in PDOTS that did not respond to B7-H3.CAR-T challenge, and no specific pattern of secreted cytokines/chemokines was clearly associated with response or resistance to B7-H3.CAR-T (Supplementary Fig. S11D–G). Taken together, these data demonstrate the feasibility of CAR T-cell profiling using PDOTS and demonstrate substantial heterogeneity that is not explained solely by alterations in the levels of antigen (B7-H3) expression.

CAR T-cell efficacy is enhanced against cancer cells lacking TBK1

We have previously shown that a lack of TBK1 in cancer cells heightens their sensitivity to cancer immunotherapy because it renders the cancer cells more sensitive to effector cytokines, specifically IFNγ and TNFɑ (21). To determine the effect of cancer cell–specific TBK1 loss on B7-H3.CAR-T efficacy, we generated TBK1-null (TBK1 sgRNA) 10101 melanoma cells (Supplementary Fig. S11H). Cell line–derived MTS derived from TBK1-null 10101 melanoma cells exhibited increased sensitivity to B7-H3.CAR-T compared to spheroids derived from control sgRNA 10101 melanoma cells (Fig. 7D). To determine if TBK1 inhibition altered the cell death threshold in patient-derived melanoma cell lines, we performed normalized growth rate inhibition (GR) analysis (26) across a range TNF𝛼 and IFN𝛾 concentrations. Similar to observations in B16 murine melanoma cells (21), dose-dependent growth inhibition was observed at lower TNF𝛼/IFN𝛾 concentrations with a cytostatic effect at higher concentrations, whereas dose-dependent cytotoxicity was observed in 10101 and M160-GFP melanoma cells treated with TBK1i above threshold concentrations of TNF𝛼 and IFN𝛾 (Fig. 7E-F). These results confirm that genetic deletion of TBK1 and pharmacologic TBK1 inhibition enhance sensitivity to TNF𝛼/IFN𝛾 by converting sub-lethal, cytostatic responses to inflammatory cytokine challenge to a lethal, cytotoxic response.

We previously demonstrated a key role for the JAK/STAT pathway in mediating the cytotoxic effect of combined TNF𝛼/IFN𝛾 treatment in cells lacking TBK1 (21). To determine the requirement for JAK/STAT signaling on CAR T-cell efficacy, we examined the effect of ruxolitinib (JAK1/2 inhibitor) on the efficacy of B7-H3.CAR-T cells +/− TBK1i co-treatment. Using cell line–derived MTS, we observed rescue of cell killing mediated by B7-H3.CAR-T in the presence of JAK1/2i (Fig. 7G-H). Enhanced TNF𝛼/IFN𝛾-driven killing observed in MTS (10101, ECC4-GFP) treated with TBK1 PROTAC 3i was also reversed with JAK1/2i treatment, consistent with earlier observations (Supplementary Fig. S11I–J). Similarly, in cocultures of B7-H3.CAR-T with PDOTS, addition of JAK1/2i resulted in complete loss of tumor cell elimination mediated by B7-H3.CAR-T (Fig. 7I). These data confirm that disrupting tumor intrinsic TBK1 signaling enhances B7-H3.CAR-T efficacy by lowering the cytotoxic threshold to TNF𝛼/IFN𝛾 and confirm a key role for the IFN sensing (JAK-STAT) pathway in determining sensitivity to CAR T cells in solid tumors.

DISCUSSION

There is an unmet need to develop preclinical models to examine patient-specific responses to solid tumor–directed CAR T-cell therapy and to more clearly define the immunosuppressive microenvironmental factors that contribute to their failure, in order to generate novel therapeutic strategies to overcome resistance to CAR T-cell therapy (55,56). Evaluation of patient-derived samples that preserve features of the TME is expected to facilitate identification of novel predictive and/or prognostic biomarkers, accelerate identification of therapies to overcome CAR T-cell resistance, and advance translational research efforts to ultimately guide precision medicine efforts to tailor therapy decisions for individual patients. Here, we demonstrate the feasibility of examining patient-specific responses to solid tumor–directed CAR T cells using PDOTS and reveal that early dysfunction of CAR-T cells in the TME can contribute to therapeutic resistance. Additionally, we establish the utility of PDOTS in testing combination strategies to overcome resistance to CAR T cells using both monotypic and organotypic tumor spheroids. Lastly, we demonstrate that pharmacologic TBK1 inhibition has dual roles in enhancing CAR T-cell efficacy by (i) rendering cancer cells more sensitive to immune attack and (ii) by preventing CAR T-cell dysfunction and enhancing CAR T-cell proliferation and effector function.

While choice of target antigen is a major decision point in the development of CAR T cells, other factors contribute to diminished CAR T-cell efficacy in solid tumors including inter-patient heterogeneity, loss of target antigen expression, impaired trafficking, and the immunosuppressive TME, which limits effector function through multiple poorly understood mechanisms (7–9)(55). Evaluation of CAR T cells to date has largely involved in vitro 2D cell culture and in vivo murine tumor models in which human cancer cells are implanted in immune-deficient mice. In recent years, patient-derived tumor models have been developed to examine patient-specific responses to CAR T cells using organoid technology. Use of patient-derived tumor models such as patient-derived organoids (PDOs) (57) facilitates examination of patient-specific factors that may influence CAR T-cell efficacy. However, PDOs and related model systems rely on prolonged, weeks-long culture to expand the requisite amount of material for in vitro testing and lack key features of the TME (i.e., tumor-infiltrating stromal and immune cells), and generating organoids is time- and labor intensive. In contrast, PDOTS profiling is performed on the day of tumor resection using organotypic tumor spheroids, which retain autologous stromal and immune cells, therefore providing a path for companion functional diagnostic testing to examine the response of a patient’s tumor to a given CAR T-cell product before the patient is treated. Further, evaluation of CAR-T cells using PDOTS in 3D microfluidic culture enables examination of both CAR T-cell trafficking and activity, as CAR T cells are seeded in the side channels of the microfluidic device and must migrate into and through the collagen hydrogels to encounter target antigen and perform their effector function. Studies using MTS enable evaluation of CAR T-cell intrinsic and tumor intrinsic drivers of resistance to CAR T cells, whereas PDOTS enable evaluation of both tumor-intrinsic and tumor-extrinsic drivers of immune evasion. While we did not set out to establish the superiority or inferiority of any particular model system in this study, there are clear conceptual advantages of incorporating PDOTS in the preclinical evaluation of CAR T cells compared to reliance exclusively on patient-derived xenografts (PDXs) or PDOs (58,59).

In the present study, we examined CAR T cells directed against B7-H3 (CD276) broadly expressed across solid tumors. Whilst a higher proportion of B7-H3-expressing cancer cells was associated with improved sensitivity of PDOTS to B7-H3.CAR-T, we also identified other factors influencing sensitivity of PDOTS to CAR T-cell challenge, including CAR T-cell dysfunction, which is associated with reduced clinical efficacy in patients (35,36). Here, we demonstrated that B7-H3.CAR-T rapidly acquired markers of dysfunctionality following interaction with antigen-expressing tumor spheroids. One cause of this dysfunction is through stimulation of immune checkpoint pathways leading to inhibition of T-cell effector function. We confirmed that the PD-1/PD-L1 pathway plays a key role in the dysfunction of B7-H3.CAR-T against both tumor cell line–derived spheroids and PDOTS, with anti-PD-1 immune checkpoint blockade enhancing CAR T-cell functionality. Anti-PD-1 in combination with CAR T-cell therapy may produce synergistic effects in patients, as is being explored in clinical trials (38). Single cell RNA sequencing confirmed early transcriptional changes in CAR T cells in co-culture with PDOTS in 3D culture, as well as transcriptional changes in tumor cells during 3D culture and following CAR T-cell challenge. Furthermore, we identified an increased expression of early activated T cell tumor-infiltrating lymphocytes following B7-H3.CAR-T challenge. While we cannot exclude a contribution of non-transduced donor peripheral T cells, the accumulation of early activated T cells in PDOTS at 24 hrs (lacking CAR T cells) suggests that non-transduced donor T cells alone cannot account for this observation. Given the modest association that we observed between baseline abundance of CD4⁺ and CD8⁺ T-cell infiltration and sensitivity to B7-H3.CAR-T, this observation raises several interesting questions regarding the role of native TILs in influencing the response to CAR T cells.

In hematological malignancies it is established that patients with low tumor burdens at treatment yield better outcomes following CAR T-cell therapy (60). Similarly in solid tumors the size of the tumor burden is thought to directly influence the level of systemic inflammation and immune dysregulation, resulting in CAR T cells that are only able to eradicate smaller metastatic tumors rather than large tumor bulks (61). Although a small sample size, here we found that PDOTS generated from primary tumors and treated with B7-H3.CAR-T (E:T 1:1) exhibited reduced response (5/11, 45%) compared to those taken from metastatic sites (9/11, 82%). This perhaps reflects differences in TME immunosuppressive factors related to the size, morphology, and development of primary and metastatic tumors. It will be vital to develop novel strategies that can reverse the immunosuppressive effects of the TME on CAR T-cell function if effective implementation of this therapy to solid tumor patients is to be achieved. We have shown that CAR T-cell function can be enhanced via targeting of TBK1, which presents a strategy to modulate both cancer and CAR T cells. Examination of peripheral T-cell populations in patients with hypomorphic mutations in TBK1 revealed skewing towards a more activated/effector population (62). Previously, we observed an accumulation of early effector CD8⁺ T cells with concomitant reduction in terminally differentiated effector CD8⁺ T cells in B16-ova tumors from mice treated with TBK1i +/− anti-PD-1 (21), suggesting that TBK1 inhibition may limit T-cell exhaustion and promote effector T-cell function. Consistent with this, here, we demonstrated that pharmacological TBK1 inhibition prevented upregulation of co-inhibitory receptors, enhanced cytotoxic functionality, and increased CAR T-cell proliferative capacity. The effect of TBK1i treatment or TBK1 deletion on CAR T-cell proliferation and dysfunction was observed between 48–96 hours after co-culture, likely explaining why only minor differences in the transcriptional programs of CAR T cells +/− TBK1i treatment were observed 24 hours after co-culture with PDOTS. Furthermore, TBK1 loss increased cancer cell sensitivity to CAR T cells through enhanced sensitivity to TNF𝛼/IFN𝛾. TBK1 presents an intriguing target molecule that may provide synergistic therapeutic effects when targeted both in cancer cells and immune cells. Additional potential therapeutic strategies that can modulate both the TME and CAR T cells, such as PTPN2 (39,63,64) or disulfiram/copper plus ionizing radiation (22), should be further explored through the use of PDOTS.

SYNOPSIS.

The authors demonstrate the feasibility of using 3D microfluidic cultures of patient-derived organotypic tumor spheroids (PDOTS) to evaluate factors influencing the efficacy of solid tumor–directed CAR T-cell therapy and identify combination strategies to overcome CAR T-cell resistance.

Abbreviations:

B7-H3

B7 homolog 3 protein

CAR

chimeric antigen receptor

CRC

colorectal cancer

E:T

effector:target ratio

GFP

green fluorescent protein

HCC

hepatocellular carcinoma

IFNγ

interferon γ

ECC

extrahepatic cholangiocarcinoma

MTS

monotypic tumor spheroids

NTT

non-targeting T cells

PDAC

pancreatic ductal adenocarcinoma

PNET

pancreatic neuroendocrine tumor

PDOTS

patient-derived organotypic tumor spheroids

PDXs

patient-derived xenografts

3D

three dimensional

TNBC

triple negative breast cancer

TME

tumor microenvironment

TMB

tumor mutational burden

TNFɑ

tumor necrosis factor ɑ