Solid tumour CAR-T cells engineered with fusion proteins targeting PD-L1 for localized IL-12 delivery

Cell lines

The mouse prostate cancer cell line PTEN^−/− Kras (PTEN–Kras) derived from 10-week K-ras G12D;PTEN deletion mutant prostate cells was a kind gift from David Mulholland (The Icahn School of Medicine at Mount Sinai)⁴². The Ras/Myc transformed prostate cancer line, RM9, was a kind gift from Timothy C. Thompson (MD Anderson Cancer Centre) and has been previously described⁴⁵. ID8, a cell line originated from C57BL/6 mouse ovarian surface epithelial cells, was a kind gift from Karen Aboody (City of Hope National Medical Center). Cell lines were cultured in complete Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Hyclone), 2 mM l-glutamine (Fisher Scientific) and 25 mM HEPES (Irvine Scientific) (cDMEM). PC3 human prostate cancer cell line (ATCC, CRL-1435) and PEO4 human ovarian cancer cell line (Sigma Aldrich, 10032309) were cultured in RPMI-1640 medium (Lonza) containing 10% FBS (Hyclone), 2 mM l-glutamine (Fisher Scientific) and 25 mM HEPES (Irvine Scientific) (complete (c)RPMI). Virus titres were checked with either HT1080 (ATCC; CCL-121, cDMEM) and/or Jurkat cells (ATCC TIB-152; clone E6-1, cRPMI). Human embryonic kidney cell line 293T (ATCC; CRL-3216) was cultured in cDMEM. To generate a more aggressive ID8 tumour line in vivo, the parental ID8 cells were engrafted in vivo in the peritoneal cavity of mice for ~30 days. Unsorted cells collected from murine ovarian model peritoneal ascites were processed for red blood cell (RBC) lysis and analysed immediately as described. Passaged ID8 cells were then engrafted into tumour-naïve mice confirming increased aggressiveness relative to parental cells.

DNA constructs, lentiviral production and transduction, and retrovirus production

Tumour cells were engineered to express firefly luciferase (ffluc) by transduction with epHIV7 backbone lentivirus carrying the ffluc gene under the control of the EF1α promoter. PTEN–Kras cells were transduced to express human PSCA (hPSCA) (human PSCA, accession no. NM_005672.5), and ID8 and PEO4 cells were engineered to express tumour-associated glycoprotein-72 (TAG72) via transduction with epHIV7 lentivirus carrying the St6galnac-I gene (STn; human, accession no. NM_018414.5, or murine mSTn, accession no. NM_011371), under the control of the EF1α promoter. STn is the unique sialyltransferase responsible for generating surface expression of aberrant glycosylation sialyl-Tn (TAG72)⁷³. The single-chain variable fragment (scFv) sequence used in the TAG72–CAR construct was obtained from a murine monoclonal antibody clone CC49 that targets TAG72 (ref. ⁷⁴). The extracellular spacer domain included the 129-amino acid middle-length CH2-deleted version (ΔCH2) of the murine IgG1 Fc spacer and a murine CD28 transmembrane domain⁷⁴. The scFv sequence from the murine anti-hPSCA antibody (clone 1G8) was used to develop the murine PSCA–CAR construct. The extracellular spacer domain included the murine CD8 hinge region, followed by a murine CD8 transmembrane domain⁴⁵. Both TAG72–CAR and PSCA–CAR had a 4-1BB intracellular co-stimulatory signalling domain. The murine CD3ζ cytolytic domain was previously described^29,45. The CAR sequence was separated from a truncated murine CD19 gene (mCD19t) by a T2A ribosomal skip sequence and cloned in a pMYs retrovirus backbone under the control of a hybrid MMLV/MSCV promoter (Cell Biolabs) as previously described^29,45,74. Bifunctional fusion proteins or controls contain αPD-L1 scFv derived from Avelumab, or mutein non-PD-L1 binding variant (αPD-L1^mut) as described previously⁵⁵. Sequences from full-length murine IL-12(p40p35) (murine IL-12a and IL-12b, NCBI accession nos. NM_001159424.3, NM_001303244.1)⁶⁸, human IL-15/Ra (NCBI human IL15 and IL15RA, accession nos. NM_000585.5, NM_001243539)^30,55 and TGFβRII (US Patent US9676863B2)³⁰ were used to construct their respective fusions. Murine fusions contained linkers using a CH2-deleted version (ΔCH2) of human IgG1 Fc spacer (NCBI genomic DNA accession no. NG_001019.6) (Supplementary Fig. 1a). Human bifunctional fusion proteins and controls contain αPD-L1 scFv derived from Avelumab (which is cross reactive with human and mouse). Human fusions contained linkers using a CH2-deleted version (ΔCH2) of human IgG4 Fc spacer (NCBI accession no. NG_001019.6). Sequences from full-length human IL-12(p35p40) (NCBI accession no. IL-12A, NM_000882.4; IL-12B, NM_002187.3) were used, with the p40 subunit distal in all instances. All constructs were ordered as complete gene fragments (Genscript) and cloned into appropriate epHIV7 lentiviral or pMYs retroviral backbones as previously described^{29,45,74,75,76}. Bifunctional murine and human fusion and CAR construct illustrations are detailed in Supplementary Figs. 1 and 9. Amino acid sequences for murine CARs and bifunctional fusions (human and murine) are detailed in Supplementary Table 1, and human bifunctional fusions are also available in US Provisional Patent Application No. 63/772,399. Amino acid sequence for the human TAG72–CAR is detailed in US Patent US20210308184A1.

Retrovirus was produced by transfecting the ecotropic retroviral packaging cell line, PLAT-E (Cell Biolabs), with addition of murine PSCA–CAR and murine TAG72–CAR retrovirus backbone plasmid DNA using FuGENE HD transfection reagent (Promega). Viral supernatants were collected after 24, 36 and 48 h, pooled and stored at −80 °C in aliquots for future T cell transductions⁴⁵. Lentivirus was generated by plating 293T cells in T-225 tissue culture flasks 1 day before transfection with packaging plasmids and desired CAR lentiviral backbone plasmid. Supernatants were collected after 3–4 days, filtered and centrifuged to remove cell debris, and incubated with 2 mM magnesium and 25 U ml⁻¹ Benzonase endonuclease (EMD Millipore) to remove contaminating nucleic acids. Supernatants were combined and concentrated via high-speed centrifugation (6,080 × g) overnight at 4 °C. Lentiviral pellets were then resuspended in phosphate-buffered saline (PBS)-lactose solution (4 g lactose per 100 ml PBS), aliquoted and stored at −80 °C for later use. Lentiviral titres were determined by transducing Jurkat or HT1080 cells. Flow cytometry was used to quantify expression of either intracellular αFc (fusion constructs) or surface CD19t (CAR+), respectively.

Murine and human T cell isolation, transductions and ex vivo expansion

Murine T cell activation and transduction was performed as described previously^29,75. Briefly, for murine studies, splenocytes were obtained by manual digestion of spleens from male heterozygous hPSCA-KI or female C57BL/6j mice. Enrichment of T cells was performed using the EasySep mouse negative T cell isolation kit following manufacturer protocol (StemCell Technologies). Murine PSCA–CAR or murine TAG72–CAR were achieved with single retrovirus transductions. Dual transduction required mixing of desired murine CAR retrovirus and secondary helper/fusion protein retrovirus to generate CAR-expressing fusion-positive T cells. Subsequent murine T cell expansion was performed as previously described²⁹.

Human leukapheresis products were obtained from consenting healthy donors under City of Hope IRB-approved protocol no. 09025. Peripheral blood mononuclear cells (PBMCs) were isolated via Ficoll-Paque (GE Healthcare) density gradient centrifugation, washed in PBS/EDTA and cryopreserved in CryoStor (StemCell Technologies). T cell activation and transduction was performed in part as described previously^29,76. Briefly, thawed PBMCs (2.0 × 10⁶ CD3⁺) were cultured in X-VIVO medium (Lonza) containing 100 U ml⁻¹ recombinant human IL-2 (rhIL-2, BioTechne) and 0.5 ng ml⁻¹ recombinant human IL-15 (rhIL-15, BioTechne). T cells were activated via CD3/CD28 TransAct (Miltenyi Biotec, 130-111-160) at 1:500 dilution, following manufacturer-suggested protocol, in 24-well tissue culture plates (Corning). Primary human TAG72–CAR transduction was performed in the presence of protamine sulfate (APP Pharmaceuticals), cytokines and human TAG72–CAR-T cell lentivirus as described previously^29,76. Secondary transduction with human helper viruses (αPD-L1–Fc, αhIL-12–Fc or αPD-L1–Fc–hIL-12) involved restimulation with TransAct and cytokine replenishment 24 h following primary transduction. All viruses were used at an MOI of 1. Cells were continuously cultured with cytokine-enriched X-VIVO, refreshed every 2–3 days. At 10 days post transduction, CAR-T cell purity and phenotype were assessed by flow cytometry, detecting surface CAR (hCD3⁺hCD19⁺) and intracellular helper proteins via Fc-linker and hIL-12 expression. At day 14, cells were used for in vitro co-culture assays or cryopreserved in CryoStor CS5 for future use.

Flow cytometry

For flow cytometric analysis, cells were resuspended in FACS buffer (Hank’s balanced salt solution without Ca²⁺, Mg²⁺ or phenol red (HBSS^−/−), Life Technologies) containing 2% FBS and 1× antibiotic-antimycotic (GIBCO, FACS buffer). Single-cell suspensions from mouse tissues or tumours were incubated for 20 min on ice with mouse Fc Block at a dilution of 1:50 (BD, 553140). Cells were then incubated with primary antibodies for 30 min at 4 °C in the dark with either Brilliant Violet 510 (BV510), Brilliant Violet 570 (BV570), Brilliant Violet 605 (BV605), Brilliant Violet 650 (BV650), fluorescein isothiocyanate (FITC), phycoerythrin (PE), peridinin chlorophyll protein complex (PerCP), PerCP-Cy5.5, PE-Cy7, allophycocyanin (APC), or APC-Cy7 (or APC-eFluor780), eFluor506, PE/Dazzle 594, PerCP-eFluor 710, BD Horizon Red 718 (R718), Alexa Fluor 488 (AF488), or PE-Cy5-conjugated antibodies. Antibodies against mouse CD3 (BD Biosciences, 563109, Clone: 17A2), mouse CD4 (ThermoFisher, 340443, Clone: RM4-5), mouse CD8a (BioLegend, 347313, Clone: 53-6.7), mouse CD19 (BD Biosciences, 557835, Clone: 1D3), mouse CD45 (BioLegend, 103145, Clone: 30-F11), mouse CD137 (ThermoFisher, 25-1371-82, Clone: 17B5), mouse NK1.1 (BioLegend, 108733, Clone: PK163), mouse PD-1 (BioLegend, 69-9985-80, Clone: J43), mouse LAG3 (BioLegend, 125227, Clone: C9B7W), mouse TIM-3 (BioLegend, 119704, Clone: RMT3-23), mouse CD11b (BioLegend, 101237, Clone: M1/70), CD44 (BD Biosciences, 103010, Clone: IM7), CD62L (BioLegend, 104412, Clone: MEL-14), CD80 (BD Biosciences, 740130, Clone: 16-10A1), mouse I-A/I-E (MHC Class II) (Biolegend, 64-5321-80, Clone: M5/114.15.2), mouse CD274 (PD-L1) (BioLegend, 124312, Clone: 10F.9G2), Ly6-C (BioLegend, 128029, Clone: HK1.4), mouse CD11c (BioLegend, 117316, Clone: N418), mouse Ly-6G (Biolegend, 127623, Clone: 1A8), mouse CD103 (BioLegend, 121426, Clone: 2E7), mouse F4/80 (BioLegend, 123127, Clone: BM8), mouse IL-12/IL-23 p40 (ThermoFisher, 12-7123-41, Clone: 17.8), human IL-15/Ra (R&D Systems, FAB10900R, Clone: 2639B, 1:20 dilution), human TGFβ Receptor II (BioLegend, 399703, Clone: W17055E), purified non-conjugated anti-human TGFβ Receptor II antibody (BioLegend, 399702, Clone: W17055E), anti-human CD45 (BD Biosciences, 563204; Clone: HI30), anti-human CD25 (Invitrogen, 46-0259-42, Clone: BC96), anti-human CD19 (BD Pharmigen, 557835, Clone: SJ25C1), purified APC anti-human PD-L1 antibody (BD Biosciences, 563741, Clone: MIH1) and PE mouse anti-human IL-12 (p40/p70) (BD Bioscienes, 554575, Clone: C11.5). All antibodies (anti-mouse and anti-human) used for flow cytometery were used at a dilution of 1:100 unless stated otherwise. Cell viability was determined using 4′, 6-diamidino-2-phenylindole (DAPI, Sigma). When necessary, secondary staining of cells was performed by washing twice before a 30-min incubation at 4 °C in the dark. Flow cytometry was performed on a MACSQuant Analyzer 16 (Miltenyi Biotec), and the data were analysed with FlowJo software (v.10.8). Representative flow cytometry gating strategy for lymphoid and myeloid populations can be found in Supplementary Fig. 10.

For intracellular flow cytometry, BD GolgiStop (51-2092KZ) was added to CAR-T cells for blocking intracellular protein transport and incubated for 3–4 h at 37 °C. Cells were transferred to a 96-well plate. Reagents and buffers for flow cytometry processing were pre-chilled on ice unless otherwise stated. Cells were washed with FACS buffer and then fixed in 1× BD Cytofix/Cytoperm (51-2090KZ) at 4 °C for 20 min. Following washing with 1× BD Perm/Wash buffer (51-2091KZ) twice, cells were stained with intracellular antibody: FITC polyclonal goat anti-human Fc (Jackson ImmunoResearch, 109-096-008) for 30 min at 4 °C. Data were acquired on a MACSQuant Analyzer 16 cytometer (Miltenyi) and analysed with FlowJo software (v.10.8).

In vitro PD-L1 binding, tumour killing and functional assays

For murine PD-L1 blocking/binding experiments, PD-L1 expression was first induced on RM9 cells for 4 h via conditioned media collected from CAR-T cell:tumour cell co-cultures. PD-L1-induced tumours were then co-cultured for 1 h at room temperature with supernatants collected from supernatants of dual-transduced CAR-T cells engineered to secrete bifunctional fusion proteins (Fig. 1d). Levels of PD-L1 blockade, from indicated CAR-T cell supernatants or relevant positive and negative controls, were measured in a flow cytometry-based competitive binding assay using a fluorescently conjugated competitively binding αPD-L1 antibody (BioLegend, Clone: 10F.9G2). Simultaneously, PD-L1-induced and fusion protein-blocked tumours were measured for detection of surface-bound fusion cytokines IL-15/Ra, IL-12 and TGFβRII. For TGFβRII detection, tumours were first blocked with cold unconjugated anti-TGFβRII antibody to remove tumour receptor background before adding CAR-T cell supernatants.

For human PD-L1 blocking/binding experiments, PD-L1 expression was first induced on PC3 prostate tumour cells overnight in 20 ng ml⁻¹ recombinant human IFNγ (BioLegend, 570202). PD-L1-induced PC3 tumour cells were then co-cultured for 1 h at room temperature with supernatants collected from supernatants of transduced human Jurkat T cell tumour cells engineered to secrete human bifunctional fusion proteins or controls (Supplementary Fig. 9a). Levels of PD-L1 blockade, from indicated CAR-T cell supernatants or relevant positive and negative controls, were measured in a flow cytometry-based competitive binding assay using an APC fluorescently conjugated competitively binding anti-human PD-L1 antibody (BD Biosciences, 563741, Clone: MIH1; 1:100 dilution). Simultaneously, PD-L1-induced, αPLD1–Fc–IL-12 protein, single-sided controls or untransduced control Jurkat T cell supernatant blocked tumours, were measured for detection of surface-bound fusion cytokine human IL-12 (BD Bioscienes, Clone: C11.5; PE mouse anti-human IL-12 (p40/p70), 554575; 1:100 dilution) and/or Fc-linker protein using goat anti-human Fc-FITC (Jackson ImmunoResearch, 109-096-008; 1:100 dilution).

For mouse tumour cell killing assays, UTD or PSCA–CAR mouse T cells engineered with or without αPD-L1, αPD-L1–TGFβ^trap, αPD-L1^mut–IL-15, αPD-L1–IL-15, αPD-L1^mut–IL-12 or αPD-L1–IL-12 were co-cultured at a primary 1:2 E:T ratio against PTEN–Kras hPSCA mouse prostate tumour cells in complete RPMI medium without cytokines in 96-well plates (Fig. 1g). At indicated timepoints (every 48 h) co-cultures were analysed by flow cytometry as described. At these timepoints, replicate plates were rechallenged with an additional 20,000 PTEN–Kras hPSCA tumour cells every 2 days for a total of 5 tumour challenges. For human tumour cell killing assays, UTD or TAG72–CAR-T cells engineered with or without human αPD-L1–Fc, hIL-12–Fc or αPD-L1–Fc–hIL-12 were co-cultured in triplicate at a primary 1:2 E:T ratio against PEO4-STn, human ovarian cancer cells transduced to express TAG72, in complete X-VIVO media without cytokines in 96-well plates (Fig. 1g). After 48 h, co-cultures were analysed by flow cytometry as described. For both murine and human co-culture assays, tumour cell killing by CAR-T cells was calculated by comparing CD45-negative DAPI-negative (viable) cell counts relative to targets co-cultured with UTD. In addition to flow cytometry measures of co-cultures, cell supernatants were collected from each timepoint to quantify IFNγ (human or mouse) by ELISA.

ELISA cytokine assays

Supernatants from tumour cell killing assays were collected at indicated times and frozen at −80 °C for future analysis. Supernatants from all timepoints were thawed and analysed for murine IFNγ using the murine IFNγ ELISA Ready-SET-GO! kit (mIFNγ; Invitrogen, 88-7314-88), or for human IFNγ using the human IFNγ ELISA kit (huIFNγ; Invitrogen, 88-7316-88) following manufacturer protocol. Mouse serum, plasma, peritoneal ascites or tumour CD45⁺ isolated cell supernatant cytokine levels of IFNγ and IL-12 were measured using murine IFNγ and IL-12 ELISA Ready-SET-GO! ELISA kits (IL-12; Invitrogen, BMS616), following manufacturer protocols. Plates were read at 450 nm using a Cytation3 imaging reader with Gen5 microplate software v.3.05 (BioTek). Mouse Luminex Discovery Assay kit (15-Plex; R&D Systems, LXSAMSM-15) was used to evaluate multiple analytes on collected peritoneal ascites from in vivo ovarian cancer models. Multiplex cytokine expression data were log₂ transformed and displayed in balloon plots created using the gglot2 v.3.5.1 R package.

RT–qPCR

Complementary DNA was prepared from 1 μg of total RNA (matched from RNA-seq analyses) using SuperScript IV reverse transcriptase (ThermoFisher). Quantitative PCR was performed in triplicate using SsoAdvanced Universal SYBR Green Supermix (BioRad). Data were analysed using the comparative threshold method, and gene expression was normalized to murine GAPDH expression on a CFX RT-PCR instrument and CFX Maestro software (BioRad, v.2.3). Murine forward (F) and reverse (R) gene primers: murine gapdh (F: GTCAAGCTCATTTCCTGGTATGACA, R: GTTGGGATAGGGCCTCTCTTG) and murine tgfβ (F: AGCTGCGCTTGCAGAGATTA, R: AGCCCTGTATTCCGTCTCCT) were used.

In vivo studies

All animal experiments were performed under protocols approved by the City of Hope Institutional Animal Care and Use Committee (protocol no. 21025). For all animal studies, mice were housed with 12 h light and 12 h dark cycle at a temperature range of 68–75 °F and humidity between 30–70%. For subcutaneous prostate tumour studies, PTEN–Kras hPSCA cells (1.0 × 10⁶) were prepared in a final volume of 100 μl HBSS^−/− and injected under the skin of the abdomen of 6–8-week-old male heterozygous hPSCA-KI C57BL/6J mice as previously described⁴⁵. Tumour growth was measured using calipers (length × width × height = mm³). For prostate tumour studies, mice were treated by i.p. administration with soluble murine recombinant IL-12 (sIL-12; 1 μg, PeproTech, 210-12) once daily for 5 days starting on the day of T cell treatment. For in vivo i.p. ovarian tumour studies, ID8-mStn (ffluc expressing) cells (5.0 × 10⁶) were prepared in a final volume of 400 μl HBSS^−/− and engrafted in >6-week-old female C57BL/6J (Jackson Laboratories) mice by i.p. injection. Tumour burden was measured via non-invasive bioluminescence imaging (LagoX, Spectral Imaging), and flux signals were analysed with Aura software (v.4.0, Spectral Imaging). Where indicated, mice were treated i.p. with 100 mg kg⁻¹ cyclophosphamide (Cy). Mice received i.v. treatment with indicated T cells (1.0 × 10⁶ CAR+ T cells) in 100 μl final volume 24 h after Cy pre-conditioning. For ovarian tumour studies, mice were i.p. treated at 14 (or 35) days post tumour engraftment as indicated, with CAR+ T cells (3.0–5.0 × 10⁶) in 400 μl final volume without Cy pre-conditioning. For indicated ovarian tumour studies, mice were co-treated i.p. with CAR-T cells and either soluble murine recombinant sIL-12 (0.5 μg i.p.) or with Avelumab (200 μg i.p.; anti-PD-L1, Bavencio, EMD Serono) every other day for a total of 3 doses starting on the day of T cell treatment. For all studies, mice were euthanized upon reaching s.c. tumour volumes exceeding 1,000 mm³ or i.p. tumours showing signs of distress such as a distended belly due to peritoneal ascites, laboured or difficulty breathing, apparent weight loss, impaired mobility or evidence of being moribund.

Peripheral blood was collected from isoflurane-anaesthetized mice by retro-orbital (r.o.) bleed through heparinized capillary tubes (Chase Scientific) into polystyrene tubes containing a heparin/PBS solution (1,000 units ml⁻¹, Sagent Pharmaceuticals). Total volume of each blood draw (~120 μl per mouse) was recorded. RBCs were lysed with 1× Red Cell Lysis buffer (Sigma) according to manufacturer protocol, and then washed, stained and analysed by flow cytometry as described above. When applicable, tumour, liver and spleen were collected from euthanized mice. Spleen weights were measured for analysis of splenomegaly. Serum from syngeneic prostate and ovarian cancer mouse studies was collected via r.o. bleed in non-heparinized capillary tubes as described above. Blood was kept at room temperature for 30 min, centrifuged at 6,000 × g for 10 min at 4 °C, aliquoted and frozen at −80 °C until used for serum cytokine ELISA or chemistry analyses. Serum chemistry analysis was performed by running samples on a VETSCAN VS2 Chemistry Analyzer (Zoetis), using the phenobarbital chemistry panel rotor (Zoetis) for BUN, ALT and AST quantification as described in the manufacturer protocol. At pre-determined timepoints or at moribund status, mice were euthanized, and tissues and/or peritoneal ascites were collected and processed for flow cytometry and immunohistochemistry. Peritoneal ascites was centrifuged at 336 × g for 10 min at 4 °C, aliquoted and frozen at −80 °C until used for cytokine ELISA or Mouse Luminex Discovery Assay multiplex analysis per manufacturer protocol (R&D Systems). For the study of immune cells collected from ovarian i.p. tumour masses, mice from each group were randomly selected and killed at day 6 or 7 post T cell injection. Peritoneal solid tumour masses were physically minced and enzymatically digested using a Miltenyi mouse tumour digestion kit, and CD45 positive selection was performed using magnetic mouse CD45 MicroBeads following manufacturer protocols (Miltenyi Biotec). Isolated CD45⁺ cells were analysed by flow cytometry, and individual replicate supernatants of overnight-cultured isolated CD45⁺ cells from each treatment group were measured for murine IFNγ secretion via ELISA as described above.

Immunohistochemistry and Nanostring GeoMx digital spatial profiling (DSP) analysis

Tumour tissues were fixed for up to 3 days in 4% paraformaldehyde (Boston BioProducts) and stored in 70% ethanol until further processing. Immunohistochemistry was performed by the solid tumour pathology core at City of Hope. Briefly, paraffin-embedded sections (10 μm) were stained with haematoxylin and eosin (H&E, Sigma Aldrich), CD3 (Ventana, Clone: SGV6), PD-L1 (Invitrogen, Clone: JJ08-95), CD4 (Abcam, Clone: EPR19514), CD8 (Cell Signaling, Clone: D4W22) and FOXP3 (Abcam, Clone: EPR22102-37). Images were obtained using the Nanozoomer 2.0HT digital slide scanner and the associated NDP.view2 software (Hamamatzu). For Nanostring GeoMx DSP, similarly prepared tissues and slides were sent for multiplex protein profiling with spatial context. The Nanostring GeoMx system-provided data centre software was used for generating the raw SNR data. Briefly, after quality control of the initial data, the segment and target data were filtered, followed by generation of SNR data for downstream analysis. SNR count protein expression data from 12 tumour ROIs per treatment group were processed and analysed using R v.4.4.0. The normality of data was assessed using the Shapiro–Wilk test. Protein expression values were z-score transformed and displayed in heat maps using the ComplexHeatmap R package (v.2.20.0)⁷⁷. ROIs from different treatment groups were clustered using complete-linkage hierarchical clustering on expression data of 39 proteins subgrouped by cell type or phenotype. The strength and direction of the relationship between the expression of defined pairs of proteins were evaluated using Spearman’s rank correlation method. Unless otherwise specified, a P-value threshold of P < 0.05 was used to determine statistical significance.

Statistical analysis

In this study, we evaluated CAR-T cells for the treatment of solid tumours using in vitro T-cell functional assays, as well as syngeneic tumour models in mice. We engineered CAR-T cells to secrete bifunctional fusion proteins and evaluated therapeutic efficacy in these model systems. All in vitro assays were performed with at least duplicate samples and were repeated in at least two independent experiments. In vivo studies were performed using 6–8-week-old C57BL/6 or C57BL/6 background huPSCA-KI transgenic mice, using at least three mice per group for all in vivo studies to ensure statistical power. Mice were randomized on the basis of tumour volume or bioluminescence imaging to ensure evenly distributed average tumour sizes across each group. In vivo experiments were repeated at least twice. For subcutaneous tumour models, survival was based on the maximum tumour size allowed (~1,000 mm³ maximum volume). Excel (Microsoft, v.16.97) and GraphPad Prism 10 (GraphPad Software) were used to generate bar plots and graphs. Data are presented as mean ± s.e.m., unless otherwise stated. Unless otherwise indicated, P values for pairwise comparisons were generated using an unpaired two-tailed Student’s t-test with assumption of unequal variance, where *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.