CAR-T cell therapy targeting MUC17 in gastric tumors.

Introduction
Gastric cancer is recognized as the fifth most common cancer worldwide, with more than one million new cases annually and over 800,000 deaths worldwide.1 2 As gastric cancer progresses rapidly and often lacks early symptoms, most patients are diagnosed at advanced stages. Despite advances in therapeutic strategies for gastric cancer including chemotherapy, targeted therapies, and immune checkpoint blockade, the overall prognosis for patients with advanced gastric cancer remains poor, with median survival rarely exceeding 1 year.2 3 Specifically, conventional treatment typically involves surgical resection, chemotherapy, and radiation, but many patients are diagnosed past the stage where curative surgery is possible.4 Targeted therapies, such as HER2 inhibitors and anti-angiogenic agents, have shown benefits in select patient populations, but resistance often develops.57 Immune checkpoint inhibitors have shown clinical success in various cancers, including gastric cancer, but their efficacy is restricted to a limited subset of patients.8 Collectively, these limitations highlight an unmet need for more effective, durable, and broadly applicable therapeutic strategies.
Chimeric antigen receptor (CAR) T cell therapy has demonstrated remarkable success in the treatment of hematologic malignancies and is now being expanded to target various solid tumors, including gastric cancer.916 However, several challenges continue to hinder the efficacy and persistence of CAR-T cells in solid tumors. For example, a major challenge in solid tumor CAR-T cell therapy is the lack of truly tumor-specific antigens that are abundantly expressed on cancer cells but absent from essential normal tissues.17 18 In addition, the tumor microenvironment presents additional barriers, including poor CAR-T cell trafficking, limited infiltration into tumor sites, and immunosuppressive signals that dampen T cell function.19 Using such tumor-associated antigens (vs tumor-specific) as the CAR target limits cell specificity and increases the risk of off-tumor toxicity.19 Currently, six tumor-associated antigens are under clinical investigation in gastric cancer, including mesothelin and the emerging target Claudin 18.2, which have already been targeted in clinical trials using CAR-T cells.1012 2023 While Claudin 18.2-targeted CAR-T therapy has shown preliminary antitumor activity, including partial responses in phase 1/2 clinical trials, cases of serious on-target off-tumor toxicity have also been reported,10 24 25 underscoring the need for safer and more selective targets in gastric cancer. In addition to these six, several other candidate antigens are currently under investigation in gastric cancer and represent key targets for next-generation CAR-T therapies.18
Gastric adenocarcinoma displays substantial molecular heterogeneity, yet many patients’ tumors share a common feature: the overexpression of membrane-bound mucins, which contribute to the formation of a dense glycocalyx on the tumor cell surface.2628 Gastric mucins, which play an important role in the protection and proper functioning of the gastric mucosa, are altered during neoplasia.28 29 Generally, this mucin-rich barrier not only facilitates tumor progression and metastasis but also impedes the access of immune cells by forming a physical and biochemical shield.30 31 Among cancer-associated mucins, MUC17, a membrane-bound glycoprotein containing 61 tandem repeats of a 59-amino acid motif decorated in serine, threonine, and proline residues, has gained attention as a promising target in gastric cancer.32 33 In gastric tumors, MUC17 expression is predominantly upregulated in epithelial cells, and its levels increase progressively from low-grade intraepithelial neoplasia to advanced gastric cancer.33 MUC17 is overexpressed in approximately 23%–52% of patients with gastric cancer, highlighting its potential as a clinically relevant target.3436 Recently, two T-cell engagers targeting MUC17 in gastric cancer have been investigated in preclinical and clinical evaluation. AMG 199, a MUC17-targeting T cell engager designed to engage MUC17 and CD3, entered clinical trials for advanced gastric and gastroesophageal junction cancers (NCT04117958), underscoring the growing interest in MUC17 as a therapeutic target.37 This program is no longer active, with the phase 1 study listed as terminated/discontinued in recent sources. In parallel, preclinical studies of another T cell engager, SCR-9171, demonstrated robust T-cell activation and cytotoxicity against MUC17-positive cancer cells, with reduced cytokine release compared with other CD3-engaging antibodies.38 To date, no studies have reported the development of MUC17-specific CAR-T cells. compared with T cell engagers, CAR-T cells may offer advantages such as prolonged persistence and enhanced cytolytic activity.39 Alongside other gastric cancer-associated antigens, MUC17 may serve as a novel target for immunotherapeutic strategies.
In this study, we explore the potential of targeting MUC17 with CAR-T cells as a strategy to overcome these therapeutic barriers. We provide preclinical evidence demonstrating that MUC17-targeted CAR-T cells selectively and effectively eradicate gastric cancer cells in vitro and in vivo. Additionally, we demonstrate that MUC17-targeted CAR-T cells are effective against the MUC17-expressing pancreatic cancer cell line ASPC-1, exhibiting robust cytotoxic activity both in vitro and in vivo. Collectively, our findings establish MUC17 as a viable and innovative target for gastric cancer immunotherapy and further demonstrate its therapeutic potential in MUC17-expressing pancreatic cancer models.

Results

MUC17 is a novel tumor-associated antigen for CAR-T cell therapy in gastric cancer
MUC17 has an extensive extracellular domain, comparable to or larger than the commonly studied mucins such as MUC1 or MUC16324043 (figure 1a). To confirm its potential as a CAR-T cell target, we first sought to identify MUC17 expressions in normal tissues and cancer cell types using publicly available datasets from DepMap and GTExPortal. MUC17 expression in normal tissues is mainly restricted to the small intestine and not observed in other normal tissues (figure 1b). Furthermore, DepMap data revealed that colon cancer, gastric cancer, and pancreatic cancer could be ideal targets for therapies targeting MUC17 (figure 1c). We first evaluated expression by immunohistochemistry (IHC) staining on gastric cancer patient specimens using two tissue microarrays. Certain tissues such as ulcer type mucinous adenocarcinoma exhibited very high MUC17 intensity and a high percentage of positive cells (figure 1d). To identify suitable cell line models for experimentation, we used DepMap datasets to profile MUC17 expression across gastric cancer cell lines, alongside known CAR-T targets such as MSLN and CLDN18.2104446 (figure 1e). KATO-3, signet-ring cell gastric carcinoma (diffuse type) derived from pleural effusion, and GSU, diffuse-type gastric carcinoma established from ascites, exhibited high expressions of all three markers, whereas NUGC4, signet-ring cell gastric carcinoma (diffuse type) originally from metastatic lymph node, showed relatively lower mesothelin expression levels. Beyond gastric cancer, the pancreatic cancer cell line ASPC-1 also demonstrated notably high expression of MUC17 (figure 1e,f). We compared the MUC17 expression levels across each cell line using flow cytometry analysis, confirming that GSU had the highest MUC17 expression, whereas NUGC4 and KATO-3 showed moderate levels (figure 1f).
To target MUC17-expressing cells, we generated MUC17-targeted CAR constructs containing an scFv derived from an anti-human MUC17/CD3 T cell engager that has already been evaluated in clinical trials37 (figure 1g). CARs were constructed in two scFv configurations (VH–VL or VL–VH), linked via a CD8α hinge and a 4-1BB/CD3ζ signaling domain. Following standard CAR-T cell production methods, cells were activated using CD3/CD28 beads, de-beaded on day 6, and cultured until day 14. Phenotypic analysis indicated that the engineered CAR-T cells closely resembled mesothelin-targeted CAR-T cells, and no differences were observed between the two scFv configurations across three different human donors (figure 1h and online supplemental figures 1–3). The MUC17 and mesothelin-targeted CARs were constructed on the same second-generation backbone, with identical hinge, transmembrane, costimulatory, and CD3ζ signaling domains; the only distinction between the two constructs is the scFv binding domain. To further compare the two scFv configurations in terms of cytotoxicity, we first used a luciferase-based killing assay, showing that both configurations of MUC17-targeted CAR-T cells exhibited robust killing of MUC17-expressing target cell lines: GSU, NUGC4, and KATO-3 (figure 1i). Likewise, a real-time cytotoxicity assay further confirmed that MUC17-targeted CAR-T cells exhibited enhanced cytotoxicity against these cell lines compared to untransduced T cells, demonstrating that MUC17 is a targetable tumor-associated antigen for CAR-T cells (figure 1j and online supplemental figure 3b,c).

MUC17 CAR-T cells exhibit robust and specific in vitro cytotoxicity
To evaluate the specificity of MUC17-targeted CAR T cells, we next generated MUC17-knockout (KO) in cell lines for each parental cell line via CRISPR-Cas9 and confirmed the MUC17 KO cells by flow cytometry analysis and immunofluorescence microscopy (figure 2a,b and online supplemental figure 4a). We also validated the KO cell lines using MUC3 antibody, as MUC3 shares high sequence similarity with MUC17 in the extracellular domain43 and may help assess potential off-target effects or cross-reactivity (online supplemental figure 4b). We observed no changes in MUC3 expression with MUC17 KO, demonstrating the specificity of the KO. To assess CAR-T cell killing activity against MUC17 KO cell lines, we used real-time cytotoxicity assay with the two scFv configurations CAR-T cells against MUC17 KO cell lines. CAR-T cell-mediated cytotoxicity was significantly reduced against MUC17 KO cell lines and approached levels observed with untransduced T cells, validating the antigen-specific recognition of the MUC17-targeted CAR (figure 2c). Additionally, we compared the killing efficacy of MUC17-targeted CAR-T cells with established mesothelin-targeted CAR-T cells against GSU cells. Both scFv configurations of MUC17 CAR-T cells showed comparable or superior cytotoxicity compared with mesothelin-targeted CAR-T in three different donors (figure 2d).
We next analyze the phenotypes of MUC17 CAR-T cells as well as mesothelin-targeted CAR-T cells after a 24-hour co-culture with target cells. Notably, MUC17-targeted CAR-T cells showed significantly increased proportions of central memory (TCM) subsets, which are associated with increased CAR T cell efficacy and decreased terminally differentiated effector memory (TEMRA) subsets compared with mesothelin-targeted CAR-T cells, suggesting an enhanced potential for persistence and long-term activity47 48 (figure 2e and online supplemental figure 5). Next, we examined the cytokine secretion profiles of MUC17-targeted CAR-T cells after 24-hour co-culture with GSU and NUGC4 cells. The secretion levels of effector cytokines such as TNF-α, IL-2, IFN-γ, and Granzyme B were generally elevated in MUC17-targeted CAR-T cells, with Granzyme B levels significantly exceeding those of mesothelin-targeted CAR-T cells on co-culture with GSU cells (figure 2f). Similar cytokine secretion patterns were observed following co-culture with NUGC4 cells (figure 2g). We further compared cytokine secretion levels across four different cell lines using MUC17-targeted CAR-T cells. To further evaluate MUC17 CAR-T cell function under more physiologically relevant conditions, we generated 3D spheroids composed of NUGC4 cells and pancreatic fibroblasts in agar-coated flat-bottom plates. After 3 days of spheroids culture, CAR-T cells were added to assess cytotoxicity. Consistent with previous results, MUC17-targeted CAR-T cells showed superior efficacy compared with mesothelin CAR-T cells (figure 2h,i). Collectively, these findings highlight MUC17 as a promising CAR-T target for gastric cancers.

MUC17 CAR-T cells mediate potent in vivo tumor control
We next investigated whether these CAR-T cells showed efficacy or potential toxicity in vivo. We injected 2×106 GSU cells subcutaneously into NSG mice and monitored the tumors for 7 days. When the average tumor size reached approximately 150–200 mm3, we injected 3×106 CAR-T cells intravenously into each mouse (figure 3a). Tumor volumes measured by calipers demonstrated effective tumor control without significant differences observed between the two scFv configuration VH–VL and VL–VH constructs (figure 3b). Body weight and overall survival were significantly improved in mice treated with MUC17-targeted CAR-T cells compared with mice treated with untransduced T cells (figure 3c,d). These observations were validated with CAR-T cells derived from a second healthy donor (figure 3e–g). To further evaluate therapeutic potency under more stringent conditions, we next tested MUC17 CAR-T cells in a GSU tumor challenge model using a reduced CAR-T cell dose. NSG mice were injected with 2×106 GSU cells subcutaneously and monitored for 7 days before intravenous injection of 1×106 CAR-T cells per mouse (figure 3h). In this aggressive tumor model, MUC17 CAR-T cells achieved sustained tumor control for up to 90 days. Consistent with their enhanced antitumor activity, MUC17 CAR-T-treated mice exhibited superior survival and maintained body weight compared with those receiving untransduced T cells (figure 3i–k). We further compared the efficacy of MUC17 CAR-T cells with mesothelin-targeted CAR-T cells in a GSU challenge model using a distinct human blood cell donor. In this tumor model, several mice in the CAR-T cell groups reached endpoints early, limiting statistical comparisons. MUC17 CAR-T cells induced tumor regression in a subset of mice, maintained stable body weight, and extended overall survival to mesothelin CAR-T cells, which exhibited minimal therapeutics benefit (figure 3l–n). Although not all mice responded, these results suggest that targeting MUC17 can elicit potent antitumor activity in tumors with high antigen expression. We next collected tumor, spleen, liver, and intestine tissues 43 days after CAR-T cell injection. Histological analysis with these tissue samples revealed that MUC17 expression was predominantly observed in tumor tissues and on the apical surface of villous epithelial cells facing the intestinal lumen. To confirm that the observed staining was not due to vascular leakage or nonspecific diffusion, we performed co-staining with CD31, a marker for blood vessels and vascular structures, which clearly delineated the vascular boundary and supported that MUC17 localization was restricted to epithelial cells. As expected, CAR-T cells circulate within the vasculature and systemically in the bloodstream, suggesting that the moderate expression of MUC17 in normal small intestine tissue is unlikely to result in significant on-target off-tumor toxicity. IHC staining confirmed that MUC17 is expressed in the mouse small intestine. Similar to CD31, MUC17 was localized to the outer epithelial surface, whereas CAR-T cells were detected within the vasculature, suggesting spatial separation between circulating CAR-T cells and MUC17 expression in normal tissue (figure 3o,p). Similar to other CAR-T cells, MUC17-targeted CAR-T cells predominantly localized to tumor and spleen tissues, consistent with their targeted cytotoxic activity and natural homing to lymphoid organs (online supplemental figure 6).

Extended antitumor activity of MUC17 CAR-T cells in MUC17+ pancreatic cancer models
Building on our findings in gastric cancer, we next assessed the efficacy of MUC17-targeted CAR-T cells in a pancreatic cancer model using the ASPC-1 cell line, which expresses both MUC17 and mesothelin. ASPC-1 expresses high levels of both mesothelin and MUC17, so we used ASPC-1 to compare MUC17 and mesothelin CAR-T cells. We first confirmed and compared the cell-surface expression of mesothelin by ASPC-1 cells, along with three gastric cancer cell lines. As expected, ASPC-1 showed higher mesothelin expression compared with the gastric cancer cell lines (online supplemental figure 7a). Next, we compared two different configurations of MUC17 CAR-T cells against ASPC-1. Similar to the results observed in gastric cancer cell lines, both MUC17 CAR configurations exhibited potent cytotoxic activity against ASPC-1 cells in vitro (figure 4a,b and online supplemental figure 7b). Cytokine profiling revealed no statistically significant differences in Granzyme B, IL-2, IFN-γ, or TNF-α secretion between the four cell lines, and both GSU and ASPC-1 cells induced similarly elevated cytokine levels (figure 4c). Phenotypic analysis showed that, compared with mesothelin-targeted CAR-T cells, MUC17-targeted CAR-T cells exhibited an increased proportion of TCM cells and a decreased frequency of TEMRA cells, consistent with results from gastric cancer cell line models (figure 4d). We further evaluated therapeutic efficacy in vivo using a challenge model: 2×106 ASPC-1 cells were subcutaneously injected into NSG mice, and 1×106 CAR-T cells were administered intravenously 14 days later (figure 4e). In this model, MUC17 CAR-T cells achieved superior tumor control compared with mesothelin CAR-T cells, with no appreciable loss in body weight. MUC17 CAR-T treatment also resulted in significantly improved survival compared with both untransduced and mesothelin CAR-T groups (figure 4f–h and online supplemental figure 7c). These findings demonstrate that MUC17 CAR-T cells effectively target not only gastric but also pancreatic cancers, suggesting broader potential clinical applicability of MUC17 CAR-T cells.

Discussion
In this study, we identified MUC17 as a novel and targetable tumor-associated antigen for CAR-T cell therapy in gastric cancer. Using a combination of gene editing, in vitro assays, and in vivo experiments, we demonstrated that MUC17-targeted CAR-T cells exhibit antigen-specific cytotoxicity, enhanced TCM phenotypes, and favorable cytokine release profiles compared with conventional mesothelin-targeted CAR-T cells, suggesting promising potency of MUC17-targeted CAR T cells for solid tumors, particularly in gastric and pancreatic cancer. Future studies directly comparing MUC17-targeted CAR-T cells with Claudin 18.2 CAR-T cells may further delineate the relative benefits of targeting MUC17.
Although there is a potential risk of on-target off-tumor toxicity based on MUC17 expression in the small intestine, MUC17 has apical localization within the small intestine, which likely limits immune cell access under normal physiological conditions.49 Histological analyses confirm that MUC17 expression in non-malignant tissues is sequestered from systemic circulation, mitigating concerns around on-target and off-tumor activity. Nonetheless, pathological disruptions of the intestinal barrier, such as inflammation or mucosal injury, may transiently increase T-cell accessibility and alter this safety profile. While the NSG mouse small intestine does express MUC17, as confirmed by IHC using a cross-reactive antibody, sequence differences between human and mouse MUC17 limit the model’s ability to fully predict on-target off-tumor toxicity in humans. Future studies will therefore require more appropriate systems, such as human MUC17 knock-in mice, as well as advanced CAR engineering strategies such as affinity-tuned CARs, logic-gated circuits (eg, AND gates), and integrated safety switches.
compared with MUC17-directed Bispecific T-Cell Engager (BiTEs; eg, vepsitamab/AMG-199 and SCR-9171), MUC17 CAR-T cells offer durable expansion and memory formation after a single infusion, which may help sustain activity in heterogeneous or immunosuppressed tumors (ref). However, CAR-T exposure is less titratable once infused, so any physiological MUC17 accessible to T cells could pose sustained on-target/off-tumor risk. In contrast, BiTEs provide dose control and rapid discontinuation if adverse events occur, and their small size can aid tissue penetration, but they rely on endogenous T-cell number and T-cell fitness and usually require continuous or repeated dosing for durable control. These platform-level trade-offs inform clinical positioning of MUC17-targeted cell therapy vs BiTEs.
The higher proportion of central-memory T cells observed with MUC17-targeted CAR-T cells relative to MSLN-targeted CAR-T cells may reflect differences in signal strength or signal quality arising from interactions between the binder and antigen, given that the two CARs differ only in the scFv. Although the precise mechanisms remain unclear, several non-mutually exclusive factors could contribute. First, binding kinetics and functional avidity might tune cumulative CAR signaling per contact, where more moderate or intermittent signaling tends to preserve TCM, whereas strong sustained signaling favors effector differentiation and exhaustion.50 Second, epitope geometry and membrane context may differ between MUC17 and mesothelin in ways that alter immunological synapse organization and downstream programming.51 Third, antigen display properties, including density, distribution, and shedding, could change the frequency and duration of productive engagements.52 Finally, differences in intramolecular scFv-scFv interactions between the MUC17- and MSLN-targeting CARs may stabilize CAR dimerization to different extents, thereby influencing the persistence of signaling. These possibilities are hypotheses but provide a potential framework for the higher central-memory fraction observed with the MUC17 binder. Future studies, such as normalizing antigen density and changing binder affinity, will be needed to validate these possibilities.
While our data support the efficacy and specificity of MUC17-targeted CAR-T cells, the heterogeneity of MUC17 expression across tumor types and within individual tumors presents a potential challenge for single-antigen targeting. Such heterogeneity may allow for immune escape by tumor cells with low or absent MUC17 expression. Given that gastric cancer or pancreatic cancers coexpress additional surface markers such as Claudin 18.2 or mesothelin, combinatorial approaches to additional tumor-associated antigens could enhance either specificity or therapeutic coverage.53 In addition, future studies using heterogeneous tumor models containing mixed populations of MUC17+ and MUC17- cells will be valuable to evaluate the efficacy of MUC17 CAR-T cells under more clinically relevant conditions and to guide the design of multi-antigen targeting strategies. are needed to evaluate the safety and feasibility of these multi-antigen targeting strategies.
In summary, our study establishes MUC17 as a promising target for CAR-T cell therapy in gastric cancer, with potential applicability in pancreatic cancer. MUC17-targeted CAR-T cells exhibited robust antigen-specific cytotoxicity, favorable cytokine profiles, and improved T cell phenotypes compared with conventional approaches. These findings lay the groundwork for further development of MUC17-directed immunotherapies and support clinical translation as a new avenue for the treatment of solid tumors.

Materials and methods

Study design
This study was designed to evaluate the efficacy of MUC17-targeted CAR T cells in preclinical models of gastric tumors. To assess the antitumor activity of the engineered CAR T cells, both in vitro cytotoxicity assays and in vivo functional studies using xenograft mouse models were performed.
As a source of T cells, anonymized human peripheral blood mononuclear cells were obtained under approval from the Institutional Review Board (IRB) at Massachusetts General Hospital (MGH) and classified as ‘non-human subjects research’. Mice used in vivo experiments were randomized prior to CAR T cell administration.

Primary cells and cell lines
GSU (Creative Bioarray), NUGC4 (Creative Bioarray), and ASPC-1 (ATCC; CRL-1682) cell lines were cultured in RPMI 1640 media (Thermo Fisher Scientific; Cat# 72400047) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific) and 1x penicillin/streptomycin (Thermo Fisher Scientific) at 37°C in 5% CO2. KATO-3 (ATCC; HTB-103) cell line was cultured in IMDM (Iscove’s Modification of Dulbecco’s Modified Eagle’s Medium; Corning) with 10% fetal bovine serum (Thermo Fisher Scientific) and 1x penicillin/streptomycin (Thermo Fisher Scientific) at 37°C in 5% CO2. Human T cells were isolated (Stem Cell Technologies, 15061) from healthy donor leukopaks obtained through the MGH Blood Bank, following a protocol approved by the institutional review board, and were cultured in RPMI 1640 media supplemented with 10% fetal bovine serum, 20 U/mL recombinant human IL-2 (PeproTech), and 1x penicillin/streptomycin at 37°C in 5% CO2.

Mice
Male and female 6–11-week-old in-house bred NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (The Jackson Laboratory) were housed in groups of up to five under pathogen-free conditions. The animals were kept at temperatures of 21.1°C–24.5°C, 30%–70% humidity, and a 12:12 light-dark cycle. The mouse experiment included at least 4 mice per group, with the exact numbers used for each experiment specified in the figure legend. When the average tumor size reached approximately 150–200 mm³, mice were randomized on day −1, therefore post-tumor injection and 1 day prior to treatment. Sample size was determined based on prior in vivo CAR-T studies showing significant tumor growth suppression with n=5. Mice were included in the study if they demonstrated successful tumor engraftment (tumor size ≥100 mm³ by day 7 postimplantation). All mouse injections were performed by one animal technician, and monitoring was blinded to expected outcomes. Animals that did not meet this threshold or exhibited signs of distress unrelated to tumor burden prior to CAR T cell infusion were excluded from the study. Caliper-based measurements were used to assess tumor burden, and data analysis was performed using GraphPad Prism software. All mice were housed at the MGH Center for Cancer Research, and all care and experiments were conducted in accordance with protocols approved by the MGH IACUC.

Generation of engineered cell lines
Cell lines were transduced to express click beetle green (CBG) luciferase and enhanced GFP (eGFP), then sorted using a BD FACSAria to isolate a clonal or different population of transduced cells. MUC17 KO cell lines were generated by electroporation of 10 µg of Cas9 mRNA and 0.3 nmol of sgRNA (Synthego CRISPRevolution UUCAGACCUCAGUGUGAACA) using BTX ECM 830 electroporator. KO efficiency was measured by flow cytometry, and a pure population was isolated through single-cell clonal expansion.

Generation of CAR constructs
Two anti-MUC17 CAR constructs and one anti-mesothelin CAR construct were synthesized and cloned into a third-generation lentiviral plasmids backbone regulated by a human EF-1α promoter (Genscript)(online supplemental file 2). All CAR constructs including MUC17 and mesothelin CAR contained a CD8 hinge and transmembrane domains, a 4-1BB co-stimulatory domain, and an intracellular CD3ζ signaling domain. All plasmids contained transgene coding for the fluorescent reporter, mCherry, to evaluate transduction efficiency. The scFvs against mesothelin and MUC17 were derived from sequences of SS1 and AMG199, respectively (available to the public). (G4S)3 linker was inserted between VH−VL or VL−VH.

CAR T cell production
An Institutional Review Board protocol approved the purchase of anonymous healthy human donor leukapheresis products from the MGH blood bank. Stem Cell Technologies T cell Rosette Sep Isolation kit was used to isolate T cells. Bulk human T cells were activated on Day 0 using CD3/CD28 Dynabeads (Life Technologies) at a 1:3 T cell:bead ratio to generate CAR T cells and untransduced T cells from the same donors to serve as controls. T cells were grown in RPMI 1640 media with GlutaMAX and HEPES supplemented with 10% FBS, penicillin, streptomycin, and recombinant human IL-2 (20 IU per ml; Peprotech). On day 1 (24 hours after activation), cells were transduced with CAR lentivirus at an MOI of 5–10. CAR-T cells were expanded with IL-2 containing cell culture media addition every 2–3 days to maintain the concentration between 0.5 and 1×106 cells/mL. Dynabeads were removed via magnetic separation on day 5–6, and cells were assessed by flow cytometry with mCherry expression, ALFA-tag, or anti-(G4S)3 linker antibody binding on days 12–14 to determine transduction efficiency prior to cryopreservation.

Flow cytometric analysis
The following antibody clones were used for CAR-T cell analysis: Alexa Fluor 647 conjugated G4S (E7O2V) antibody (69 782S; Cell Signaling Technology), Alexa Fluor 700 conjugated anti-human CD3 antibody (300424; BioLegend), Per/Cy7 conjugated anti-human CD4 antibody (300518; BioLegend), PerCP conjugated anti-human CD8α antibody (301032; BioLegend), FITC conjugated anti-human CCR7 antibody (561271; BD BioSciences), Brilliant Violet 42 conjugated anti-human CD45RA antibody (304130; BioLegend), Alexa Fluor 700 conjugated anti-human CD45 antibody (304024; BioLegend). Adherent cancer cells were detached by incubating with Tryple Express Enzyme (1x; 12-604-013; Fisher Scientific) at 37°C for 5–10 min. APC-conjugated anti-human CD69 (310910; BioLegend) was diluted 1:200 in 2% FBS in phosphate-buffered saline (PBS) and incubated with cells at 4°C for 1 hour for each stain. For analysis of MUC17 cell surface expression level or glycosylation patterns, APC-conjugated anti-human CD17 antibody (395305; BioLegend) and mucin probe (SAE0212; Sigma-Aldrich) were diluted 1:200 in 2% FBS PBS and incubated with cells at 4°C for 1 hour. Cells were washed and stained with DAPI containing 2% FBS PBS to assess cell viability before analyzing on a BD Fortessa X-20.

Cytotoxicity assays
For single-time point cytotoxicity assays, target cells expressing CBG luciferase were incubated with CAR-T cells or untransduced T cells at varying effector-to-target (E:T) ratios in 200 µL of growth media specific to the target cell line, without IL-2, and cocultured in a 96-well plate for 30 hours at 37°C in 5% CO2. Cells were then lysed using the Bright-Glo Luciferase Assay System (E2610; Promega), and luciferase activity was measured with a Biotek Neo2 luminescence plate reader. Specific lysis was calculated using the following formula: Percentage of specific lysis=[(luminescence target cell only)−(luminescence target cell+CAR T cell)]/(luminescence target cell only)×100%. For real-time killing assays, target cells expressing CBG-eGFP were plated in flat-bottom 48-well or 96-well plates (Corning) and incubated for at least 4 hours at 37°C in 5% CO2. CAR-T cells or untransduced T cells were added in triplicate at various E:T ratios as indicated. CAR expression (%) was normalized using untransduced T cells for each donor. Plates were then incubated at 37°C for up to 6 days, with whole wells recorded every 2 hours using the IncuCyte Live Cell Analysis system. Cytotoxicity was measured as the total green fluorescent area and analyzed using the IncuCyte image analysis software.

3D spheroid assay
96-well flat bottom tissue culture plates were coated with 1% agarose to create a low-attachment binding surface for spheroids formation as described previously.54 55 NUGC4 tumor cells (2×10³) were seeded in 50 µL of culture medium in an agarose-coated 96-well plate. Pancreatic stellate cells (PSCs) were subsequently added at varying ratios to tumor cells in an additional 50 µL volume. Spheroid formation was monitored by brightfield microscopy, and compact spheroids were confirmed after 48 hours. For the CAR-T cell killing assay, 2×10³ MUC17 CAR-T cells, mesothelin CAR-T cells, or untransduced T cells were added to preformed spheroids. CAR-T cell-mediated cytotoxicity was monitored using the Incucyte Live-Cell Imaging System.

Immunofluorescence
Target cancer cell lines (GSU, NUGC4, and KATO-3) were plated in 35 mm glass bottom dishes (P35G-1.5–14 C; Mattek), grown for 24 hours. APC-conjugated anti-human CD17 antibody (395305; BioLegend) was diluted 1:200 in 2% FBS PBS and incubated on samples at 4°C for 1 hour. Cells were further treated with Hoechst 33 342 (4082S; Cell Signaling Technology) containing 2% FBS PBS. All samples above were imaged on a LSM 780 confocal microscope using 63x (NA: 1.4 Oil) objectives (Zeiss).

Immunohistochemistry
After tumor-engrafted mice were euthanized, tumors and organs were extracted and fixed in 4% PFA overnight, washed with PBS, and serial washed with 30%, 50%, 70% ethanol and stored in 70% EtOH until staining. Tissue slides were then made and embedded in paraffin. Slides were stained for CD3, MUC17 (ab122184; abcam) by the specialized histopathology services core facility at MGH. All IHC slides were imaged on an Axio Scan.Z1 microscope and quantified the respective stains by QuPath (V.0.5.1).

Statistical methods, sample sizes, data collection, and assumptions
The sample sizes were selected on the basis of standards in the field and not pre-determined using statistical methods. For statistical comparisons, data distributions were assumed to be normal. Normality was tested for conditions with approximately ten or more data points. All the statistical analyses were performed using GraphPad Prism 8 software. All the experimental data are presented as mean±SD or as box-and-whisker plots with the first and third quartiles (boxes), median and range of data, unless stated otherwise within figure legends. Appropriate statistical tests were used to analyze the data, as described in the figure legends.

Supplementary material
10.1136/jitc-2025-013120online supplemental file 112310.1136/jitc-2025-013120online supplemental file 2