Genetically engineered ErbB2 overexpression sensitizes organoid-derived tumors to checkpoint inhibition in a syngeneic model of gastric cancer.

Introduction
HER2/ERBB2 is a receptor tyrosine kinase encoded by the ERBB2 gene in humans (located on chromosome 17q12) and the Erbb2 gene on chromosome 11 in mice.1 HER2 does not have direct ligand-binding activity, but functions as the preferred heterodimerization partner for the three ERBB family members EGFR, ERBB3/HER3, and ERBB4/HER4.24
ERBB2 amplifications and/or mutations, and HER2 overexpression occur at varying frequencies across numerous tumors, most commonly in breast cancer, gastric and gastroesophageal junction (GEJ) carcinoma, salivary gland and ovarian cancer, and less frequently in non-small cell lung cancer (NSCLC) and colorectal cancer.5 Generally speaking, HER2 overexpression or its oncogenic activation by somatic mutation promotes HER2 homodimerization and/or heterodimerization, resulting in the hyperactivation of the downstream RAS/MEK/ERK and PI3K/AKT signaling pathways and driving cell survival and proliferation.4
ERBB2 genetic alterations differ across tumor types, which has implications for the biology of HER2 in these cancer entities and provides opportunities for new treatment options beyond breast and gastric cancer.6 In breast cancer, the ERBB2 genomic locus is subject to amplification in 20% of cases, whereas somatic mutations are rare; ERBB2 amplification correlates very well with overexpression of the protein and is associated with an unfavorable prognosis.3 The HER2-specific humanized antibody trastuzumab targets the extracellular domain IV of HER2, which results in inhibition of downstream signaling and induction of antibody-dependent cell-mediated cytotoxicity and has been approved since 1998 for HER2-positive breast cancer.7 8 In gastric cancer, 11%–16% of patients harbor ERBB2 amplifications and a further 3% harbor ERBB2 mutations, mostly affecting the protein tyrosine kinase domain; trastuzumab combined with chemotherapy (cisplatin plus either capecitabine or 5-fluorouracil) is the standard first-line treatment for patients with advanced-stage HER2-positive gastric cancer ever since FDA approval was granted in 2010.9 Somatic mutations in the tyrosine kinase domain, resulting in constitutive downstream signaling driving tumorigenesis, are more common than ERBB2 amplification in several other cancers, including lung cancer.10 11 The anti-HER2 antibody–drug conjugate (ADC) trastuzumab deruxtecan (T-DXd) shows efficacy in ERBB2-mutant NSCLC,12 13 which has resulted in the accelerated FDA approval of T-DXd for this indication in 2022. Numerous novel HER2-targeted agents have either entered the clinic or are in clinical evaluation; these include small-molecule TKIs, HER2-directed ADCs other than T-DXd, bispecific T cell engagers, adoptive cell therapies, and vaccines.5 14 The advent of novel potent HER2-targeted agents such as T-DXd has on the one hand expanded the indications for HER2-targeted therapy to include HER2-low tumors15 and has led to the histology-agnostic FDA approval in 2024 of T-DXd for HER2-positive solid tumors other than breast and gastric cancer.16
Whereas the cell-intrinsic implications of HER2 overexpression and ERBB2 mutations have been the subject of intense research that in turn has led to the development of ever more potent therapeutics, the consequences of ERBB2 alterations on the tumor microenvironment are overall less well understood, although a strong association between ERBB2 amplification and dense lymphocytic infiltration of the tumor was reported for breast cancer as early as 1990.17 Large numbers of tumor-infiltrating lymphocytes (TILs) at diagnosis have been associated with pathological complete response and event-free survival in HER2-positive early-stage breast cancer treated with a combination of the TKI lapatinib and trastuzumab18 and were positively prognostic in HER2-positive breast cancer patients in a large cohort irrespective of treatment.19 Given the need for gaining a better understanding of possible effects of ErbB2 overexpression on the tumor immune microenvironment in a physiologically relevant and well-controlled setting, we set out to develop a syngeneic immunocompetent, organoid-based model of ErbB2-overexpressing gastric tumorigenesis. We find that ErbB2 overexpression leads to increased T-cell infiltration in subcutaneously and orthotopically growing tumors and provide evidence that such tumor-infiltrating T-cells are capable of controlling ErbB2-expressing tumors on adoptive transfer into immunodeficient hosts. The combination of PD-1 blockade with an anti-ErbB2 antibody is more effective than the single agents at reducing the tumor burden in ErbB2-expressing but not control tumors.

Methods

Mice and organoid-based tumor models
C57BL/6J wild-type mice (strain #: C57BL/6JRj) were obtained from Janvier Laboratories; B6.129S2-Trp53tm1Tyj/J (strain #:002101), C57BL/6J-ApcMin/J (strain #:002020) and B6.129S2-Ifnar1tm1Agt/Mmjax (strain #:010830) were obtained from the Jackson laboratory. Composite Trp53−/−
Apcmin/+ mice were generated in-house, as were Rag2−/−Il2rg−/− mice (B6.Il2rg<tm1Cgn>/J x B6.Rag2<tm1Fwa>, parental lines were obtained from a local mouse repository). All C57BL/6J wild-type and mutant mouse lines were housed in groups of up to five animals per individually ventilated cage at approved animal facilities of the University of Zurich, with ad libitum access to standard chow and water and enrichment such as tunnels and additional nesting material. Male mice were mostly used for experimentation, as AP organoids were of male origin; mice were 6–8 weeks old when included in experiments. Sample sizes were chosen based on feasibility and our previous experience with regard to the mouse-to-mouse variability in our models rather than a formal power analysis. For the subcutaneous tumor model, 300,000 organoid-derived cells were collected per injection and resuspended in 100 µL of Matrigel (#354 230 Corning). Mice were anesthetized via isoflurane inhalation, and the organoid-Matrigel suspension (100 µL) was injected subcutaneously into each flank using a 30G needle. Tumor growth and overall health were monitored and scored three times per week. Tumors were harvested 7 weeks post-injection for further analysis. For the orthotopic tumor model, 40,000 organoid-derived cells were collected per injection and resuspended in 15–20 µL of Matrigel, with several injections performed per stomach. Buprenorphine was administered 30 min prior to surgery for analgesia. Mice were anesthetized via isoflurane inhalation, and Carprofen (Rimadyl) was administered 5 min before surgery. The abdomen was shaved and disinfected, and mice were placed in a supine position on a heated pad. Local anesthesia was provided via subcutaneous injection of lidocaine and bupivacaine. A 1 cm skin incision was made 0.2 cm below the lowest rib using sterile scissors, and underlying connective tissue was carefully loosened with forceps. A 7 mm incision was then made in the peritoneal wall, and the stomach was gently exteriorized. Using a 30G needle, the organoid-Matrigel suspension was injected into the submucosa (1–2 injections per side), followed by flipping the stomach to administer 1–2 additional injections on the opposite side. The injection site was kept moist with prewarmed (37°C) sterile Phosphate-buffered saline (PBS), and the stomach was returned to the peritoneal cavity after a 2 min waiting period. The peritoneal wall was closed using 3–4 separate sutures (Vicryl 5–0), followed by five additional sutures to close the skin incision. If necessary, 2–3 drops of Histoacryl were applied to seal the wound. Postoperatively, Buprenorphine and Carprofen were administered via subcutaneous injection every 12 hours for 72 hours. Mice were monitored and scored three times per week. Tumors were harvested 7 weeks post-injection for further analysis. Anti-ErbB2 (#BE0277), anti-PD1 (#BE0273), and isotype control antibodies (#BE0089, #BE0085) were purchased from BioX Cell (New Hampshire, USA) and administered i.p. twice weekly at 250 µg/dose. All animal experimentation described in this work adheres to institutional, cantonal, and federal guidelines and was approved by the Zurich Cantonal Veterinary Office (ZH150/2022 and ZH023/2025 and their amendments, to AM).

Gastric organoid cultures
Stomachs from 8 to 10 week-old mice were harvested, opened along the greater curvature, and washed in ice-cold DPBS (#14190-094 Gibco). The mucus and muscle/serosa layers were carefully removed under a stereomicroscope using forceps, and the antrum and corpus were separated with a safety margin before being cut into 2–5 mm² fragments. Tissue dissociation was performed using 43.4 mM sucrose (#A2211,5000 Huberlab) and 54.9 mM D-sorbitol (#240 850 Sigma-Aldrich) in DPBS, followed by 5–10 rounds of vigorous pipetting and washing. The tissue was then incubated in 10 mM EDTA (#A1103,1000 BioChemica) at 4°C for 1 hour and 20 min on a rolling shaker. After an additional washing step with the sucrose/sorbitol solution, glands were extracted by applying pressure to the tissue fragments under a glass slide (#AA00000112E01MNZ10 Epredia), releasing individual glands that were collected in Dulbecco's Modified Eagle's Medium (DMEM), filtered through a 40 µm mesh, counted, centrifuged, and mixed with Matrigel (#356231 Corning). The suspension was seeded into 24-well plates (50 µL per well) and incubated at 37°C. After Matrigel solidification, 500 µL of standard 3D organoid culture medium was added on top, consisting of Advanced DMEM/F12 (#12634-010 Gibco) supplemented with 10 mM Penicillin/Streptomycin (#167369 Gibco), 50% (v/v) Wnt-conditioned medium (supernatants from L Wnt3a cells; ATCC CRL-2647), 1 µg/mL R-spondin1 (#315-32 PeproTech), 1 mM N-acetylcysteine (#A9165 Sigma-Aldrich), 50 ng/mL EGF (#AF-315-09 PeproTech), 100 ng/mL FGF (#100-26 PeproTech), 100 ng/mL Noggin (#250-38 PeproTech), 10 nM [Leu15]-Gastrin I (#G9145 Sigma-Aldrich), 1×N2 supplement (#17502-048 Gibco), 1×B27 supplement (#17504-044 Gibco), 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (#15630-056 Gibco), 2 mM Glutamax (#35050-038 Gibco), and 10 µM Y27632 (included only on the seeding day; #S1049 Selleckchem). Organoids were passaged every 3 days at a 1:4 ratio to maintain culture viability.

Cloning, lentivirus production, and transduction of gastric organoids
The full-length wild type coding sequence of rat Erbb2 was amplified from Addgene plasmid #66 945 using the primers fwd: CAGATCGCCTGGAGAATTGGCTAGCATCAACAAGTTTGTACAAAAAAAGCAGGCTACCATGATCATCATGGAGCTG, rev: TCCCCTACCCGGTAGAATTGGATCCAAACCACTTTGTACAAGAAAGCTGGGTTCATACAGGTACATCCAG and cloned into the EcoRI and ClaI restriction sites of the pHIV-EF1a-puro-T2A expression vector. The full-length mouse Erbb2 sequence was amplified from plasmid “MG50714-UT” (Sino Biological Inc) using the primers fwd: TGGAGGAGAATCCTGGCCCAATGGAGCTGGCGGCCT, rev: GCTCCATGTTTTTCTAGGTCTCGATCGAGGTCGACGGTATCGATTACTGGCACATCCAGGCCT and cloned as described above. For IFNAR and STING knockdown, shRNA constructs (IFNAR: fwd: CCGGGAATGAGGTTGATCCGTTTATCTCGAGATAAACGGATCAACCTCATTCTTTTTG rev: AATTCAAAAAGAATGAGGTTGATCCGTTTATCTCGAGATAAACGGATCAACCTCATTC; STING: fwd: CCGGATGATTCTACTATCGTCTTATCTCGAGATAAGACGATAGTAGAATCATTTTTTG rev: AATTCAAAAAATGATTCTACTATCGTCTTATCTCGAGATAAGACGATAGTAGAATCAT) were inserted into the pLKO.1-TRC vector at the Age I and Eco RI restriction sites. For the control shRNA construct, the following oligonucleotides were used: fwd: CCGGCCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGGCGACTTAACCTTAGGTTTTTG; rev: AATTCAAAAACCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGGCGACTTAACCTTAGG. The puromycin selection cassette in pLKO.1-TRC was exchanged for the blasticidin selection cassette. For lentivirus packaging, 293 T cells were seeded at a density of 3.8×10⁵ cells/mL (23 mL per plate) in antibiotic-free growth media (DMEM+10% iFBS) in 15 cm tissue culture plates on day 1 and incubated for 18–22 hours to reach 60%–70% confluence. On day 2, 2 hours before transfection, the media was replaced with 23 mL of prewarmed reduced-serum OptiMEM (#31985062 Gibco). The transfection mixture was prepared by combining 1.15 mL of OptiMEM with 36 µg of total DNA, consisting of the transfer plasmid, psPax2, and pCMV-VSVG in a 4:3:2 ratio, followed by immediate vortexing. A 3:1 PEI:DNA ratio was used, with PEI diluted in 2.3 mL of OptiMEM before being gently added dropwise into the DNA solution. The mixture was incubated at room temperature for 15–20 min before being slowly added to the cells, followed by incubation for 18–22 hours. On day 3, the OptiMEM was removed and replaced with 23 mL of Advanced DMEM/F-12 (#12 634 010 Gibco) supplemented with B-27 (1 x; #17504-044 Gibco), and GlutaMax (1×; #35050-038 Gibco). After 72 hours (until day 6), viral supernatants were collected, centrifuged at 1000 rpm for 5 min at 4°C to pellet cell debris, and filtered through a 0.45 µm filter. The clarified supernatant was mixed at a 3:1 ratio with Lenti-X Concentrator (Takara, 631232) and incubated at 4°C for 16–18 hours. The mixture was then centrifuged at 1,500×g for 45 min at 4°C, and the resulting virus pellets were resuspended in the required volume of standard 3D organoid culture medium supplemented with 8 µg/mL Polybrene.
For transduction, 3D cultured organoids were harvested from culture wells, mechanically dissociated into small pieces by vigorous pipetting (30x), followed by a 5 min incubation in TripLE (#12 605 028 Gibco) at 37°C. The resulting small clusters were collected by centrifugation, combined with 250 µL of the viral suspension described above before being transferred into 48-well plates. The plates were then centrifuged at 32°C for 1 hour to enhance viral transduction, followed by incubation at 37°C for 6 hours. After incubation, transduced organoids were collected, centrifuged at 1000 rcf for 5 min, and resuspended in 50 µl of Matrigel per sample before being seeded into 24-well plates. Once the Matrigel solidified, 500 µL of prewarmed 3D seeding culture medium was added to each well to support further organoid growth. Puromycin or blasticidin selection was initiated 2 days after transduction and lasted for 10 days (2 µg/mL for 3 days, followed by seven more days with 0.6 µg/mL puromycin; 50 µg/mL for 3 days, followed by seven more days with 10 µg/mL blasticidin).

Multiplexed immunofluorescence imaging
Organoid-derived cells were grown in μ-Slide 8 Well chambered coverslips with ibiTreat surface modification (Ibidi, 80806). To allow for multiple rounds of antibody staining and imaging, the 4i multiplexing protocol was applied.20 In brief, cells were fixed with 4% Paraformaldehyde (PFA) in 1x PBS for 30 min. After a brief wash with 1x PBS, cells were permeabilized with 0.1% Triton X-100 for 30 min and blocked with blocking solution (1% Bovine serum albumin (BSA), 200 mM NH4Cl in PBS) supplemented with 150 mM Maleimide for 1 hour. Cells were then incubated with primary antibodies diluted in blocking solution for 2 hours while gently rocking. The following antibodies and dilutions were applied in multiple 4i staining/imaging cycles: anti-Erbb2 mouse monoclonal (Invitrogen, MA5-13675; 1:300), anti-E-cadherin mouse monoclonal (BD Bioscience, 610182, 1:300), anti-phospho Histone H2AX (Ser139) mouse monoclonal (Millipore, 05-636-AF647, 1:300), anti-β-Tubulin Rat Monoclonal (Abcam, ab6160; 1:200), anti-Lamin B1 Rabbit Polyclonal (Abcam, ab61048; 1:500), anti-cGas (Cell Signaling Technology, D3080, 1:200). After the incubation, cells were briefly washed with 1x PBS, followed by an incubation with appropriate secondary antibodies diluted in blocking solution for 1 hour while gently rocking. The following antibodies were applied: Alexa Fluor Plus 488 Goat anti Mouse IgG (Invitrogen, A32723; 1:500), Alexa Fluor 568 Goat anti Rabbit IgG (Invitrogen, A11036; 1:500), Alexa Fluor 568 Goat anti Mouse IgG (Invitrogen, A11036; 1:500), Alexa Fluor Plus 647 Goat anti Rat IgG (Invitrogen, A48265; 1:500). Cells were then washed with 1x PBS three times. For each round of imaging, cells were stained with 10 µg/mL DAPI diluted in 1x PBS for 30 min. Cells were imaged in imaging buffer (700 mM N-Acetylcysteine/dH2O, pH adjusted to 7.4). After each round of imaging, cells were washed with dH2O and antibodies were eluted with elution buffer (0.5 M Glycine, 5 M Urea, 5 M Guanidinium chloride, 70 mM TCEP-HCl; all in dH2O with pH adjusted to 2.5) for 15 min while rocking (150 RPM). The elution step was repeated three times. Chambers were then washed with 1x PBS and re-blocked for 1 hour while gently rocking, followed by the next round of incubation with primary and secondary antibodies. The high-throughput imaging was performed using the high-content imaging system ImageXpress Confocal HT.ai from Molecular Devices equipped with motorized stage, a CMOS sensor, and dual spinning disk technology. Images were acquired using a 50 µm pinhole disk and 40x (NA=1.15) Nikon objective with water immersion. 25–36 images with z-planes spanning a range up to 20 µm were acquired for each well. The images were then rendered in a Fiji ImageJ software using a Maximum Intensity Z-Projection. Image registration was performed using a Fiji-based two-dimensional (2D) descriptor-based registration plugin.

Immunohistochemistry
Formally fixed paraffin-embedded longitudinal sections of stomach or subcutaneous tumor tissue were stained with the following primary antibody clones or polyclonal sera (supplier, dilution, and incubation time are given): CD3 Clone SP7 (Abcam, 1:250, o/n), pan-cytokeratin clone PCK26 (Novus Biologicals, 1:1500, 1 hour) followed by detection with EnVision secondary reagents and DAB. Slides were scanned and imaged using NDP.view V.2 software. Quantification of signals was done using QuPath.

Bulk RNA sequencing
Bulk RNA sequencing was conducted at the Functional Genomics Center Zürich (FGCZ). Total RNA was extracted using the Qiagen RNeasy Mini Kit according to the manufacturer’s protocol. Library preparation was performed with the Illumina TruSeq mRNA Library assay, following standard procedures. Sequencing was carried out on the Illumina NovaSeq 6000 using the S1 Reagent Kit v1.5 (100 cycles), and demultiplexing was performed with the Illumina bcl2fastq Conversion Software. The number of reads per library ranged from 17 million to 42.7 million. RNA sequencing data analysis was conducted using the SUSHI framework, incorporating several steps: read quality assessment using FastQC, adapter trimming with fastp, read alignment via the STAR aligner against the Ensembl mouse genome build GRCm39 (Release 106), and gene-level expression quantification using the ‘featureCounts’ function from the R package Rsubread. Differential expression analysis was performed using the generalized linear model within the DESeq2 Bioconductor R package, while Gene Ontology (GO) pathway enrichment was analyzed using the hypergeometric over-representation test via the ‘enrichGO’ function in the clusterProfiler Bioconductor R package. All analyses were executed in R V.4.2.2 with Bioconductor version 3.16. All raw bulk RNA sequencing data are available at GEO under the accession number GSE300738.

Single-cell RNA sequencing
Single-cell suspensions were prepared and stained with a viability dye (eFluor780, eBioscience), anti-CD8 BV711 (53–6.7, BioLegend), anti-CD4 BUV496 (GK1.5, BD Horizon) and anti-TCRβ PE-Cy7 (H57-597, BioLegend) for live/dead discrimination and T cell identification, respectively. Viable TCRβ+ cells were sorted from four to five mice per sample and labeled with sample tags from the BD Mouse Immune Single-Cell Multiplexing Kit (BD Biosciences) according to group. Subsequently, cells were subjected to single-cell library preparation using the BD Rhapsody Single-Cell Analysis System (BD Biosciences), following the manufacturer’s protocol (Mouse TCR/BCR Full-Length, mRNA Whole Transcriptome Analysis (WTA), and Sample Tag; 23-24365(01)). Sequencing of the libraries was performed at the FGCZ using the NovaSeq X Plus platform with a paired-end 150 bp read configuration. Each library yielded the following number of reads: 25 million for the sample tag, 243 million for V(D)J (TCR), and 312 million for WTA. Following demultiplexing of BCL files using bcl2fastq v2.20.0.422 (Illumina), data were processed using the BD Rhapsody Sequence Analysis Pipeline v2.1, including WTA, sample tag cell demultiplexing, and V(D)J analysis. After quality control and filtering (using scDblFinder and DropUtils), 11 561 cells remained for analysis. Subsequent analyses were carried out in R V.4.4 using the Seurat package V.5. Clonality analysis was performed using scRepertoire.21

Quantitative RT-PCR
RNA was isolated from organoid-derived tumors using the RNeasy Mini kit (QIAGEN) according to the manufacturer’s instructions. Complementary DNA synthesis was performed using Superscript III reverse transcriptase (Thermo Fisher). Real-time quantitative PCR (qRT-PCRs) was done using the Sybr green method. The primers used in this study were the following: IFN-α forward: GGATGTGACCTTCCTCAGACTC, reverse: ACCTTCTCCTGCGGGAATCCAA; IFN-β forward: GCCTTTGCCATCCAAGAATGC, reverse: ACACTGTCTGCTGGTGGAGTTC; ISG15 forward: CATCCTGGTGAGGAACGAAAGG, reverse: CTCAGCCAGAACTGGTCTTCGT; NLRP3 forward: TGCTCTTCACTGCTATCAAGCCCT, reverse: ACAAGCCTTTGCTCCAGACCCTAT; ZBP1 forward: GATCTACCACTCACGTCAGGAAG, reverse: GGCAATGGAGATGTGGCTGTTG; RIPK3 forward: CAGTGGGACTTCGTGTCCG, reverse: CAAGCTGTGTAGGTAGCACATC; MLKL forward: AATTGTACTCTGGGAAATTGCCA, reverse: TCTCCAAGATTCCGTCCACAG; Caspase-1 forward: GGCACATTTCCAGGACTGACTG, reverse: GCAAGACGTGTACGAGTGGTTG. GAPDH forward: TGTGTCCGTCGTGGATCTGA, reverse: CCTGCTTCACCACCTTCTTGAT. The TaqMan gene expression assays used for IFNAR and STING kd verification were purchased from Thermo Fisher (IFNAR: Mm00439536_m1; STING: Mm01158117_m1). Complementary DNA samples were analyzed in duplicate using a Light Cycler 480 detection system (Roche) and gene expression levels for each sample were normalized to HPRT or GAPDH expression. Mean relative gene expression was determined, and the differences were calculated using the 2ΔC(t) method.

Single-cell preparation and spectral flow cytometry
Cells were isolated from subcutaneously or orthotopically grown organoids with subsequent staining for spectral flow cytometry analysis according to our recently developed MARMOT protocol22 with minor modifications. Tissues were enzymatically dissociated at 37°C for 40 min using a shaking incubator. The digestion was performed in RPMI 1640 medium supplemented with 10% fetal calf serum (FCS), 15 mM HEPES, 0.05 mg/mL DNase I, and 500 U/mL type IV collagenase (Sigma-Aldrich). Following digestion, the suspension was filtered through a 70 µm cell strainer. Tissue fragments were mechanically dissociated by pressing them through the mesh using a 1 mL syringe plunger. Both the digestion tube and the filter were subsequently rinsed with PBS/BSA/EDTA, and the cell suspension was centrifuged at 380×g for 10 min. The supernatant was discarded, and the pellet was resuspended in 150 µL of PBS/BSA/EDTA. Single-cell suspensions were transferred to a 96-well plate, centrifuged at 350×g for 5 min, and the supernatant was removed. Cells were then blocked by incubation in 25 µL of DPBS (without calcium and magnesium; Thermo Fisher) containing 1:100 Fc receptor blocking reagent (anti-CD16/CD32, BioLegend) for 20 min at 4°C. For surface marker staining, 25 µL of antibody cocktail targeting the following extracellular markers was added (CD45 BUV395, clone 30-F11; CD4 BUV496, GK1.5; CD19 BUV563, 1D3; Ly6G BUV661, 1A8; CD44 BUV737, IM7; CD103 BUV805, M290, TCRgd BV510, GL3; CD11c BV750, N418; FcεR1α PerCP-eFluor 710 or BUV615, MAR-1; CD206 PE, Y17-505; TCRb PE-Cy7, H57-597; CD69 RB780, H1.2F3; Ki-67 BV 480, B56; all BD Biosciences; SiglecF BV421, E50-2440; CD62L BV570, MEL-14; F4/80 BV711, BM8; NK1.1 BV785, PK136; Ly6C PerCP, HK1.4; IL7Ra PerCP-Cy5.5, A7R34; CD80 APC, 16–10 A1; LAG3 BV421; C9B7W; PD-L1 PE-Cy5, 10F.9G2; PD1 BV605, 29F.1A12; ICOS PE-Cy5, 15F9; ICAM-1 PE, YN1/1.7.4; KLRG1 APC, 2F1/KLRG1; CD25 BV650, PC61; all BioLegend; CD11b eFluor450, M1/70; CD8a AF532, 53–6.7; MHC-II AF700, M5/114.15.2; all Thermo Fisher) including a fixable viability dye in Brilliant Stain Buffer (Thermo Fisher). Cells were incubated for an additional 20 min at 4°C. After surface staining, cells were washed with PBS/BSA, centrifuged at 350×g for 5 min, and fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience, Thermo Fisher), following the manufacturer’s protocol. Cells were washed with the kit’s permeabilization buffer and centrifuged at 500×g for 5 min. Intracellular staining was performed in the same buffer using antibodies against transcription factors and intracellular markers (Foxp3 FITC, FJK-16s; RORgt PE-eFluor610, B2D; c-Maf eFluor660, sym0F1; all Thermo Fisher, T-bet BV605, 4B10; Biosciences) for 50 min at 4°C. Cells were then washed with PBS/BSA, centrifuged at 500×g for 5 min, and finally resuspended in 150 µL of PBS/BSA. Single-stain controls were prepared using either the same cell type or Ultracomp eBeads Plus (Thermo Fisher), processed in parallel. FMO staining was used throughout to accurately determine where to set gates. Data acquisition was carried out on a 5-laser Cytek Aurora flow cytometer (Cytek Biosciences) equipped with 64 detectors.

Transgenic T-cell receptor expression and co-culture assay
To construct the T-cell receptors (TCRs), the two identified TCR α chain (AAG amino acid CDR3 sequence: AAGMNYNQGKLI; AVM amino acid CDR3 sequence: AVMNYNQGKLI) were cysteine modified,23 codon-optimized, and then combined with the shared TCR β chain (amino acid CDR3: ASSLDTGAEQF) to form two complete TCRs (AAG and AVM αβTCR). The complete AAG and AVM TCR sequences were then cloned in sense orientation into plasmid vectors (Genescript) under the murine phosphoglycerate kinase promoter and flanked by pT Internal Tandem Repeats recognized by the SB100x Transposase; α and β chains were divided by a T2A cleaving region. Flat-bottom 12-well plates (non-treated) were coated with anti-CD3 (5 µg/mL) and anti-CD28 (2 µg/mL) antibodies diluted in 2 mL PBS per well and incubated overnight at 4°C. Single-cell splenocyte suspensions were prepared by mechanically dissociating spleens through a 100 µm cell strainer placed over a 50 mL Falcon tube. Cells were washed with PBS and centrifuged at 700 × g for 8 min at 4°C–10°C. Red blood cells were lysed by resuspension in 500 µL ACK lysis buffer per spleen for 3 min, followed by washing in RPMI supplemented with 10% FCS and centrifugation at 700 × g for 5 min. Viable cells were counted manually, and T cells were isolated using the Pan T Cell Isolation Kit II (Miltenyi Biotec) according to the manufacturer’s instructions. For electroporation, 5×10⁶ purified T cells per condition were mixed with 7.5 µg pT-based transposon vector (either TCR α/β AVM or AAG) and 2.5 µg Sleeping Beauty (SB100X) transposase mRNA in P3 buffer (Lonza), and electroporated using the Lonza 4D-Nucleofector system and cuvettes with program DN100, as recommended by the manufacturer. Cells were immediately transferred into pre-warmed Advanced RPMI supplemented with beta-Mercaptoethanol, 10% FCS, and GlutaMAX (Gibco), and human IL-2 (PeproTech, Gibco). After a 2-hour incubation, penicillin/streptomycin (1%) was added. On day 3, the medium was refreshed and supplemented with human IL-7 and IL-15 (both PeproTech, Gibco) to support T cell expansion. On day 8 post-electroporation, engineered T cells were co-cultured with organoid-derived cells in a 96-well plate format for 48 hours at various effector-to-target ratios. Afterward, the supernatants were collected after gentle resuspension of the wells and transferred to a new plate. The plate was centrifuged at 350×g for 8 min, and the cleared supernatants were stored for subsequent IFN-γ analysis by ELISA. The remaining T cells were stained for flow cytometric analysis in 50 µL staining buffer containing the following antibodies: LD eFluor780, CD45 BUV395, CD3 BV570, CD8 AF532, CD4 BUV496, CD69 RB780, and CD44 BUV737 (clones as described above). After staining, cells were washed and fixed using the BD Cytofix Fixation Kit, and data were acquired on a Cytek Aurora spectral flow cytometer.

Statistical analyses
All statistical analyses were performed using Graph Pad Prism V.10.0 software. Non-parametric Mann-Whitney U test was used for comparisons between two groups, and Kruskal-Wallis testing was used for comparisons of more than two groups, followed by Dunn’s multiple comparisons correction. P values are indicated as follows: *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.

Results

ErbB2 overexpression induces a transcriptional signature enriched for immune response and cell adhesion genes
To generate murine gastric organoids that stably overexpress ErbB2, we transduced murine wild-type gastric organoid cells with a constitutive overexpression vector encoding full-length rat ErbB2, under the control of the EF-1α promoter. Cells were selected by puromycin for 3 days; various passages of selected cells (passages 1–4) were subjected to ErbB2-specific qRT-PCR, which confirmed high level expression of the transcript (figure 1A). ErbB2-specific immunofluorescence microscopy alongside nuclear, cytoplasmic, and plasma membrane markers confirmed the specific and exclusive localization of ErbB2 at the plasma membrane in organoid cells grown as 2D monolayers (figure 1B). No overt differences were seen with respect to growth properties such as doubling time, cell morphology of cells grown in 2D, 3D organoid shape, or other parameters (figure 1B and data not shown). RNA sequencing of three batches of cells revealed numerous genes to be upregulated due to ErbB2 overexpression; ErbB2 itself and Sprr1a were the most differentially expressed genes (figure 1C,D). Over-representation analyses revealed the biological processes of cell adhesion and of immune/inflammatory responses to be selectively transcriptionally enriched in ErbB2-overexpressing organoids (figure 1E). KEGG pathway analysis revealed the “cytoplasmic DNA sensing” pathway to be particularly enriched among ErbB2-induced genes, specifically Aim2, Ripk3, Sting, Ikk, and Il33 (online supplemental figure 1A). The combined results suggest that ErbB2 overexpression does not measurably affect the in vitro growth of gastric organoids but triggers a transcriptional program that appears to be defined by the activation of cell adhesion and immune response genes.

ErbB2 overexpression results in enhanced T-cell recruitment in orthotopic and subcutaneous models of gastric organoid-derived tumor formation
Having generated murine gastric organoids overexpressing ErbB2, we asked what the consequences would be of ErbB2 overexpression for organoid engraftment and growth, and for the composition of the organoid-derived tumor microenvironment in a syngeneic, immunocompetent model. To this end, wild-type C57BL/6 organoids expressing ErbB2 or not (the latter transduced with empty vector only) were subcutaneously transplanted onto the flanks of syngeneic C57BL/6 mice. ErbB2 overexpression did not affect the engraftment rate or the tumor size and weight at the study endpoint 7 weeks post-transplantation (figure 2A,B). H&E staining suggested that epithelial (organoid-derived) cells contributed more to the overall tumor size in control tumors than in ErbB2-overexpressing tumors, whereas the latter were more inflamed (figure 2C). We speculated that ErbB2 expression might thus affect the composition of the tumor immune microenvironment and designed a spectral flow cytometry panel that allowed us to capture the main myeloid and lymphocytic populations (figure 2D, online supplemental figure 2A,B). Whereas the proportion of neutrophils, monocytes, macrophages, dendritic cells, B-cells, and regulatory T-cells did not differ as a result of ErbB2 overexpression, we found increased frequencies and absolute counts of CD4+ Foxp3− T-cells and to a lesser extent of CD8+ T-cells in ErbB2-overexpressing tumors (figure 2D,E, online supplemental figure 2C). CD4+ T-cells in ErbB2-overexpressing tumors expressed more CD44, LAG3, PD1, and/or ICAM1 than their counterparts in control tumors (populations 16, 22, 23; online supplemental figure 2D).
Wild-type organoids overexpressing ErbB2 grow slowly and do not exhibit features of gastric cancer or premalignant lesions; rather, they retain the ability to grow as a single-layered epithelium and to form a liquid-filled lumen. To develop a model that more closely reflects the uncontrolled growth of gastric cancer, we transduced organoid cells with our ErbB2 overexpression plasmid that had been generated from a composite mouse strain lacking both alleles of Trp53 (encoding P53) and expressing a truncated, inactive form of Apc (Apcmin/+). Such “AP” organoids grow faster and form larger tumors when transplanted subcutaneously onto the flanks of C57BL/6 mice, with the growth advantage being driven mainly by the Apc mutation.24 ErbB2 overexpression in AP organoids (“APE”) led to smaller tumors at the study endpoint relative to the empty vector control (“APctrl”), although engraftment rates were similar (figure 2F,G). We speculated that this might be due to a more pronounced T-cell infiltration and a possibly improved T-cell-driven immune control of APE tumors. Indeed, CD4+ and CD8+ T-cells were strongly overrepresented in the microenvironment of APE tumors relative to AP control tumors, in terms of both frequencies and absolute counts (figure 2H–J); note that lentiviral vector components, or puromycin resistance conferred by our lentiviral vector are unlikely to contribute to T-cell infiltration as AP and AP control tumors were infiltrated by similar numbers of CD4+ and CD8+ T-cells (online supplemental figure 2E). The populations enriched in APE tumors14 16 18 mainly differed from all other (unchanged) T-cell populations in their PD1 expression (online supplemental figure 2F,G). Other lymphocytic and myeloid populations were unaffected by ErbB2 overexpression (online supplemental figure 2H). The differences in T-cell infiltration between AP control and APE tumors were confirmed by immunohistochemistry of FFPE sections using CD3 as a marker of T-cells; pan-cytokeratin (PCK) staining revealed fewer tumor cells (confirming the reduced tumor weights) in APE tumors (figure 2K–M). The expression of mouse ErbB2 instead of rat ErbB2 in AP organoids confirmed the over-representation of CD4+, but not CD8+ T-cells at two different examined time points (online supplemental figure 2I,J), and did not affect tumor growth (data not shown).
To test the idea that ErbB2 drives T-cell infiltration in a second, more physiologically relevant model, we developed a protocol that allows for the surgical implantation of gastric organoids directly into the gastric submucosa. To this end, the peritoneal cavity is opened by a small incision, and 40,000 organoid cells are directly implanted via injection into the corpus region of the stomach; this number results in reasonably efficient engraftment while not compromising tissue architecture or overall health of the animals. ErbB2 overexpression in wild-type organoid cells did not affect the engraftment rate (figure 2N), but led to a strong overrepresentation of both CD4+ and CD8+ T-cells in the microenvironment of the engrafted organoids (figure 2O,P). The orthotopic implantation of AP control and APE organoids revealed a lower engraftment rate for the latter (figure 2Q); flow cytometric evaluation was not possible due to the tiny size of the lesions. The combined results suggest that ErbB2 overexpression drives T-cell infiltration in two organoid backgrounds (wildtype and AP) and in both a subcutaneous and an orthotopic model of in vivo organoid growth.

ErbB2 overexpression induces the formation of cGAS-positive micronuclei and IFNAR-dependent T-cell recruitment
Oncogene activation, for example, of KRAS, is known to result in micronuclei formation, which in turn can activate cGAS and trigger a type I interferon response. We speculated that ErbB2 overexpression might similarly drive micronuclei formation; we used quantitative immunofluorescence microscopy to examine this possibility. The analysis of 3000 wild-type organoid cells per condition and experiment, transduced or not to overexpress ErbB2, allowed us to determine the numbers and frequencies of micronuclei and of micronuclei formation based on DAPI staining. ErbB2 overexpression led to an increase in micronuclei formation that was consistent across 5 experiments (figure 3A,B). Approximately 50% of micronuclei were not surrounded by lamin B1, indicating that they had ruptured; the fraction of ruptured micronuclei without a lamin B1 ring increased as a result of ErbB2 overexpression (trend only; figure 3C,D), whereas the fraction of intact nuclei did not (online supplemental figure 3A). Micronuclei showed evidence of containing damaged DNA, as judged by a positive γH2AX signal (figure 3C). There was no clear connection between damaged DNA content and an intact, or ruptured lamin B1 coat, with both γH2AX+ and γH2AX- micronuclei randomly surrounded, or not, by lamin B1 (figure 3C). The fraction of γH2AX+ micronuclei did not change on ErbB2 overexpression (figure 3E), and we also could not observe an increase in overall DNA damage (in the main nucleus) due to ErbB2 overexpression (online supplemental figure 3B). DNA in ruptured micronuclei is potentially accessible to the cytosolic enzyme and DNA sensor cyclic GMP-AMP synthase (cGAS); indeed, roughly half of all micronuclei were positive for cGAS, and their proportion increased due to ErbB2 overexpression (figure 3F,G). cGAS+ micronuclei were typically positive for γH2AX, and all lacked an intact lamin B1 ring (figure 3F). The combined results indicate that ErbB2 overexpression results in the increased formation of micronuclei, which show evidence of rupturing and ensuing cGAS recruitment.
To address whether, and how the observed increase in micronuclei might affect the immune response to transplanted organoid cells in vivo, we performed qRT-PCR of AP control and APE tumors harvested on subcutaneous growth in wild-type mice for 7 weeks. Genes encoding the type I interferons, which are known to be directly activated by cGAS/cGAMP/STING signaling via IRF3 activity, numerous other interferon-stimulated genes (ISGs), and other components of the cytosolic DNA sensing machinery were activated by ErbB2 overexpression in vivo (figure 3H, online supplemental figure 3C). Mining the RNA sequencing data generated from in vitro grown organoids confirmed many of the transcripts to be upregulated due to ErbB2 expression also in vitro (online supplemental 3D). This result prompted us to transplant ErbB2-overexpressing organoids on wild-type and AP backgrounds into Ifnar−/− mice, which lack the receptor for type I interferons. Subcutaneous transplantation of APE organoids generated similarly sized tumors in wild-type and Ifnar−/− mice (figure 3I), which, however, differed in their T-cell infiltration. Both CD4+ and CD8+ T-cell infiltrates were reduced in Ifnar−/− mice (figure 3J). Interestingly, the engraftment rate of orthotopically transplanted APE organoids was higher in Ifnar−/− than in wild-type mice (figure 3K). To further dissect whether tumor cell-intrinsic STING and IFNAR signaling are required for T-cell infiltration into APE organoids, we generated organoids with a knockdown of STING or of IFNAR in both AP and APE cells. The knockdown efficiency ranged from 30% to 90% at the transcript level (online supplemental figure 3E). Subcutaneous transplantation resulted in similarly sized tumors if either STING or IFNAR was knocked down (figure 3L, online supplemental figure 3F); CD4+ and CD8+ infiltrates were somewhat reduced in the knock-down organoids in the APE, but not the AP background (figure 3M,N). The combined data suggest that T-cell recruitment to ErbB2-overexpressing tumors is at least partly driven by a STING/type I interferon response, which in turn appears to be triggered by micronuclei formation and rupturing, and the sensing of cytosolic DNA.

CD8+ effector memory T-cells undergo ErbB2-specific clonal expansion as determined by scRNA sequencing
Having identified T-cells as being differentially abundant in ErbB2-overexpressing and control tumors, we performed scRNA sequencing combined with TCR repertoire sequencing on FACS-sorted T-cells from six APE and six AP control tumors that had grown subcutaneously for 7 weeks in WT C57BL/6 mice. The T-cell populations were >90% pure after sorting (data not shown). Seurat clustering identified 24 clusters (figure 4A); among these, clusters 1, 8, 11, 14, and 17 were overrepresented as a consequence of ErbB2 expression, and clusters 0, 9, 13, 16, 18, 20, and 22 were under-represented (figure 4A,B). Clusters that were over-represented in ErbB2-overexpressing tumors were classified as CD8+ effector memory (EM) T-cells on projection onto reference single-cell atlases using ProjecTILs25 or Immgen; signature transcripts of the over-represented populations included transcripts encoding chemokines or their receptors (Ccl4, Ccr5, Cxcr6) and transcripts indicative of chronic activation, cytotoxic properties, and of stem-like exhausted T-cells (Tox, Tcf7, Gzmk) (figure 4C,D, online supplemental figure 4A–C). TCR α-chain and β-chain (TRA, TRB) sequencing revealed that clonal expansion had occurred predominantly in the CD8+ EM T-cell compartment; clones were on average more expansive in APE than in AP control tumors (figure 4E); 18 distinct clones (defined as >50 cells harboring the same paired TCR α-chain and β-chain CDR3 sequences) could be identified overall, of which half (nine clones) were exclusive to the CD8+ EM T-cell compartment of APE tumors (figure 4E); the two dominant clones were represented by >180 cells each. Comparatively minimal clonal expansion was detected in the CD4+ T-cell compartment of either tumor type (figure 4E). The projection of TCR α-chain and β-chain (TRA, TRB) sequences onto our 24 Seurat clusters failed to reveal a particular bias toward any of the clusters beyond the broader category of CD8+ EM T-cells (figure 4F). Manual mapping of select T-cell signature transcripts (of which some, but not all, were among the top 10 transcripts of the populations enriched in APE tumors) onto the Seurat clusters confirmed that Gzmk, Fasl, Tox, and other effector molecules of cytotoxic CD8+ T-cells were indeed expressed, although not exclusively, in the populations enriched in APE tumors (figure 4G,H, online supplemental figure 4D,E). The combined results suggest that in vivo growing AP control, and especially APE organoids, are infiltrated by clonally expanding CD8+ EM cells with cytotoxic properties.

The adoptive transfer of T-cells sorted from ErbB2-expressing organoid-derived tumors controls ErbB2-expressing tumors in immunodeficient recipients
To address whether T-cells that infiltrate ErbB2-overexpressing organoid-derived tumors are capable of controlling tumor growth in an ErbB2-specific manner when adoptively transferred to tumor-bearing immunodeficient recipients lacking T-cells, B-cells, and NK-cells (Rag2–/–Il2r–/–), we sorted TCRα/β+ T-cells from APE or control AP organoid-derived tumors that had grown subcutaneously in WT C57/BL6 mice. 160 000 pure T-cells from APE or AP tumors (of which approximately half were CD4+ and the other half CD8+) were intravenously transplanted into Rag2–/–Il2r–/– recipients that had been subcutaneously injected 4 weeks earlier with either APE or AP organoids and had formed palpable tumors at the time of injection. APE-derived T-cells effectively reduced the tumor burden in APE tumor-bearing, but not AP tumor-bearing Rag2–/–Il2r–/– recipients relative to mock-injected controls receiving PBS only; AP control tumor-derived T-cells had a smaller and insignificant effect on the APE tumor burden (figure 5A,B). The effects on tumor size and volume (figure 5A,B) were confirmed by the immunohistochemical quantification of the epithelial compartment using pan-cytokeratin staining of the AP and APE tumors transplanted or not with T-cells (figure 5C,D). CD3+ T-cells could be detected by immunohistochemistry only in APE tumors of mice transplanted with APE-derived T-cells, but not in any of the control groups (figure 5E). Within APE tumors, CD3+ T-cells were mostly confined to the periphery of the organoid growth, with some cells diffusely infiltrating the single-layered APE epithelium (figure 5E,F). In order to determine the exhaustion state of adoptively transferred T-cells, we repeated the experiment, this time with mouse ErbB2 overexpression and with a spectral flow cytometry readout at the study endpoint. Immune profiling was possible, although less than 4000 (mostly CD8+) T-cells were detected in APE tumors transplanted with APE-derived T-cells, and even fewer T-cells were detected in the controls; the majority (~80%) of adoptively transferred CD8+ T-cells from APE, and also from AP control tumors, were positive for the exhaustion markers LAG3 and PD1, especially if they had been transferred to Rag2–/–Il2r–/– recipients bearing APE tumors; opposite trends were observed for TIM3, KLRG1, and Ki67 (online supplemental figure 5A–D). The combined results indicate that adoptively transferred T cells are sufficient to control the tumor burden, and this effect appears to be driven by, and specific to, ErbB2 expression. Given that we had identified two TCR α-chain and β-chain pairs that had undergone clonal expansion (>180 cells per clone) in vivo, we used sorted splenic T-cells to produce TCR-redirected T-cells expressing one or the other TCR, in addition to their endogenous TCRs. We applied our Sleeping Beauty engineering platform, originally developed for human CAR T-cell production,26 to mouse T cells using sleeping beauty transposition. The resulting transgenic T-cells were expanded and co-cultured with AP or APE organoids, followed by the flow cytometric quantification of expression of the T-cell activation marker CD69. Both CD4+ and CD8+ redirected T-cells became activated on exposure to organoid cells at all examined effector: organoid cell ratios, but this was observed irrespective of organoid ErbB2 expression (online supplemental figure 5E,F). We conclude that the two clonally expanded TCR α/β-chain pairs are AP organoid- but not ErbB2 specific, although they were identified in ErbB2-expressing tumors.

T-cells in ErbB2-overexpressing organoid-derived tumors express PD1 and respond to PD1-targeted checkpoint blockade, especially in combination with an ErbB2-targeting antibody
Having established by adoptive transfer into tumor-bearing Rag2–/–Il2r–/– mice that T-cells are capable of controlling organoid-derived tumors in an ErbB2-dependent manner, we asked whether T-cell-directed immunotherapy using PD1 blockade would reduce the growth of ErbB2-expressing tumors. CD4+ and CD8+ T-cells harvested from ErbB2-expressing WT or AP organoid-derived tumors were more likely to express PD1 than their empty vector counterparts (figure 6A–D); a trend for CD4+ T-cells was also seen if mouse instead of rat ErbB2 was overexpressed in AP organoids (online supplemental figure 6A,B). PD-1 expression by tumor-infiltrating T-cells was at least partly dependent on tumor cell-intrinsic STING and IFNAR signaling in APE tumors (online supplemental figure 6C,D). 3.5 weeks of anti-PD1 treatment were not sufficient to significantly reduce the tumor burden of mice injected subcutaneously with APE or AP cells (figure 6E–G); however, we observed an increase in intratumoral CD4+ and CD8+ T-cells that was particularly pronounced in ErbB2-expressing tumors under PD1 blockade (figure 6H). CD4+ and CD8+ T-cells responded quite differently to PD-1 blockade as determined by flow cytometric profiling of activation and exhaustion markers (online supplemental figure 6E). We speculated that PD-1 blockade might act synergistically with ErbB2 targeting in reducing the tumor burden. This was indeed the case: only mice harboring APE tumors, but not mice harboring AP tumors, benefited from combination treatment with anti-PD1 and an ErbB2-targeting antibody (figure 6I,J). APE, but not AP tumors, were smallest at the study endpoint if both antibodies were administered (figure 6I,J), which coincided with increased T-cell infiltration under PD1 blockade (online supplemental figure 6F,G). The combined results suggest that ErbB2 overexpression provides an opportunity for combinatorial treatment using anti-PD1 and anti-ErbB2 antibodies.

Discussion
In this study, we investigated the immunological consequences of ErbB2 overexpression, mimicking the amplification and mutation of the ERBB2 locus in human gastric cancer, in two newly developed syngeneic organoid-based models of the malignancy. One model was comparatively simple and involved the subcutaneous injection of genetically engineered organoid cells harboring up to three recurrently occurring gastric cancer mutations, affecting the Apc, Trp53, and Erbb2 genes; the advantage of this model was that it generated sufficient material for spectral flow cytometry, immunohistochemistry, and scRNA sequencing, including TCR repertoire sequencing. The other, orthotopic model entailed the surgical implantation of organoid cells into the gastric submucosa and was thus more physiologically relevant, but material for downstream readouts was limiting. In both scenarios, and in organoids of both WT and AP backgrounds, the overexpression of ErbB2 resulted in higher T-cell infiltration, which at least in part could be linked to clonal expansion. The elevated frequencies of T-cells in ErbB2-overexpressing tumors resulted in improved tumor control, both at steady state and under PD1 blockade, and especially in combination with ErbB2-targeted therapy. T-cells in ErbB2-expressing tumors tended to express high levels of granzyme K and of FasL, which likely explains their anti-tumor activity in WT mice, and on adoptive transfer, also in tumor-bearing Rag2–/–Il2r–/– mice.
Both rat and mouse ErbB2 overexpression (achieved overexpression levels were >10 fold vs 2–3 fold, respectively) was used interchangeably in our studies, with generally stronger effects of the rat homolog; rat ErbB2 was initially chosen for practical reasons (such as the availability of in vivo-grade antibodies) and because it represents a high-antigen-density, non-self context that enables robust CD8 priming, analogous to a mutated oncogenic ERBB2. Rat ErbB2-transgenic mice have been used extensively for testing of drugs and vaccines,27 and rat ErbB2-expressing tumor cell lines have been used to explore the mechanism of action underlying ErbB2-targeting antibodies in mice.2830 In contrast, mouse ErbB2 expression represents the more stringent and physiologically relevant self-antigen model; its much lower expression in our organoids appears to be sufficient to support CD4+ T-cell activation (eg, via antigen uptake and presentation by APCs) but falls below the threshold required for robust CD8+ T-cell priming. Although rat and mouse ErbB2 are highly homologous (94% amino acid sequence identity in the extracellular domain), minor sequence differences may generate xenogeneic epitopes that are not subject to central tolerance in the murine immune system; indeed, some of the effects of the immunogenicity of rat ErbB2 overexpression, and of the efficacy of anti-ErbB2 antibody in tumor cell transplantation models2830 were found to be reduced in a syngeneic setting.31 The reduced CD8+ T-cell response observed with mouse ErbB2 highlights the translational challenge of targeting HER2 as a self-antigen.
Our experimental results are in line with the recently published results of a phase 3 clinical trial in which patients with unresectable, metastatic, ERBB2/HER2-positive gastric or GEJ adenocarcinoma were treated in first line with a combination of pembrolizumab with trastuzumab and chemotherapy. This combination provided a significant improvement in overall survival as compared with trastuzumab and chemotherapy alone.32 33 Similar results have been obtained in patients with PD-L1-positive, trastuzumab-resistant, advanced HER2-positive breast cancer.3436 The positive results of the pembrolizumab/trastuzumab/chemotherapy combination32 33 add to an increasing body of evidence supporting the incorporation of PD-1 antibodies in first-line therapy for (HER2-negative) advanced gastric or GEJ adenocarcinoma,37 at least in patients with PD-L1-positive tumors.3840
Mechanistically, our results suggest that T-cell infiltration is caused by interferon responses in ErbB2-expressing cells that are initiated by micronuclei formation. ErbB2-expressing cells produce more microscopically detectable micronuclei than their control counterparts; micronuclei are small, extranuclear bodies containing DNA that arise from errors during the segregation of sister chromatids during mitosis.41 Micronucleation is a well-known consequence of oncogene-induced replication stress, that is, the deregulation of physiologic DNA replication, and has been attributed to aberrant origin firing, replication–transcription collisions, reactive oxygen species, and defective nucleotide metabolism.42 Cyclin E, MYC, and RAS are examples of oncoproteins that are known to induce replication stress.4345 Our data suggest that ErbB2 should be added to a list of >25 oncoproteins that are all linked, in one way or another, to the induction of replication stress and the ensuing genomic instability.42 Roughly half of ErbB2-induced micronuclei show signs of rupturing in our hands, that is, they have lost their lamin B envelope making micronuclear DNA accessible to cGAS. This in turn leads to the activation of IRF3 and the production of type I interferons (IFN) and other proinflammatory cytokines.46 We indeed detect the increased expression of a variety of not only type I IFNs, but also interferon-stimulated genes such as Isg15, by RNA sequencing of ErbB2-expressing tumors. Type I IFNs act directly on T-cells to promote the acquisition of effector and memory functions47 48 and indirectly through cross-presenting dendritic cells for optimal T-cell priming.49 50 Indeed, ErbB2-overexpressing organoids are more likely to form orthotopic tumors in Ifnar–/– mice, which are unresponsive to type I IFN, than in controls and are infiltrated by fewer T-cells when growing in Ifnar–/– mice. Given how important continuous type I IFN production likely is for T-cell responses and tumor control, we believe that PD-1 antibody should be given prior to anti-ErbB2 to achieve optimal synergy between the two therapeutics. Other studies have investigated possible mechanisms underlying the synergy of trastuzumab and PD-1 antibodies. In preclinical models and patient-derived samples, trastuzumab increased ErbB2 internalization, Fc receptor (FcR)-mediated uptake of ErbB2 by dendritic cells and ErbB2-specific T-cell responses.51 The effect of anti-ErbB2 depended on the FcR and adaptive immunity in a subcutaneously transplanted syngeneic cell line model of ErbB2 overexpression, also arguing against the simplified notion that the antibody merely interrupts oncogenic signals and induces FcR-mediated cytotoxicity.29
In summary, our results demonstrate the utility and versatility of organoid-derived tumors as genetically tractable models to study tumor–immune interactions in a controlled but physiologically relevant setting. Our findings link oncogene-induced replication stress and micronucleation to type I IFN responses and T-cell activation, providing a mechanistic explanation for the observed synergy between PD1 antibodies and ErbB2/HER2-targeted therapy in mouse models and in patients with HER2 overexpression.

Supplementary material
10.1136/jitc-2025-012976online supplemental file 110.1136/jitc-2025-012976online supplemental file 2