Intraperitoneal CAR T-cell therapy for peritoneal carcinomatosis from gastroesophageal cancer: preclinical investigations to a phase I clinical trial (NCT06623396)

Abstract

Peritoneal carcinomatosis is a frequent metastatic condition in gastroesophageal cancer and is associated with poor prognosis and limited therapeutic options. Here, we establish clinically relevant mouse models of peritoneal carcinomatosis to evaluate the efficacy of a second-generation mesothelin-targeted chimeric antigen receptor (CAR) T-cell therapy. Our model recapitulates key clinical features, including ascites, bowel obstruction, and an immunosuppressive tumor microenvironment characterized by tumor-cell programmed death-ligand 1 expression and elevated TGF-β levels in ascites. To overcome T-cell exhaustion, we engineered a CAR T-cell construct (M28z1XXPD1DNR) that incorporates a programmed cell death protein-1 decoy receptor lacking the intracellular signaling domain, which enhances functional persistence. We demonstrate that intraperitoneal (regional) administration of CAR T cells at low doses achieves superior antitumor efficacy, longer survival, and sustained functional persistence, compared with intravenous (systemic) administration. Of note, intraperitoneal treatment also exhibits potency against distant disease sites. These findings provide a strong rationale for clinical translation; we are now conducting a clinical trial (NCT06623396) to evaluate intraperitoneal administration of M28z1XXPD1DNR CAR T cells in patients with gastroesophageal cancer peritoneal carcinomatosis.

Background

Peritoneal carcinomatosis, which is observed in one-third of patients who present with metastatic gastroesophageal adenocarcinoma, is resistant to chemotherapy and immunotherapy and is associated with a median survival of 3–5 months.1,4

The immune microenvironment in gastroesophageal adenocarcinoma peritoneal metastases—which is characterized by M2 macrophages, inhibitory cytokines, and low levels of tumor-infiltrating lymphocytes—hinders the response to immunotherapy.5,7 To tilt the immune environment in favor of effector responses, chimeric antigen receptor (CAR) T-cell therapy has been administered to boost the number of tumor-infiltrating lymphocytes in these resistant tumors.⁸ CARs are synthetic receptors that have been engineered to direct T cells to target a cancer-associated antigen, such as mesothelin (MSLN), which is overexpressed on gastroesophageal adenocarcinoma cells and is associated with tumor aggressiveness.⁹ To overcome the challenges of poor infiltration of adoptively transferred immune effector cells into solid tumors, we investigated intrapleural (regional) delivery of CAR T cells, which, in preclinical studies, achieved superior efficacy, compared with intravenous (systemic) administration, even at 30-fold lower doses, and augmented the CD4 CAR T-cell helper function to sustain a higher number of CD8 CAR T cells.¹⁰ Intrapleural delivery of MSLN-targeted CAR T cells was translated to a phase I clinical trial (n=27 patients) and was shown to be safe and effective; 41 patients have been treated to date in phase I/II studies, with systemic circulating CAR T cells detected >100 days after a single intrapleural infusion.¹¹

Patients with peritoneal carcinomatosis from gastroesophageal adenocarcinoma often present with extraperitoneal sites of metastases;^{2 4} we therefore investigated the efficacy of intraperitoneal CAR T-cell therapy in mouse models of combined peritoneal and distant sites of disease.

Methods

Tumor cells

KYAE-1 and SK-GT-4 gastroesophageal adenocarcinoma cells were obtained from Sigma-Aldrich. Because of the variable nature of the MSLN expression in cancer cells in culture, cells were transduced with human MSLN variant 1 (isolated from human ovarian cancer cells OVCAR-3) subcloned into an SFG retroviral vector to generate MSLN-overexpressing tumor cells (KYAE1-M and SKGT4-M). Cells were retrovirally transduced to express green fluorescent protein (GFP)–firefly luciferase fusion protein (KYAE1-GM, SKGT4-GM) to facilitate the performance of in vivo bioluminescence imaging (BLI). All cell lines used were periodically authenticated at the Antibody and BioResource Core Facility at Memorial Sloan Kettering Cancer Center using short tandem repeat profiling. In addition, all lines were routinely tested for mycoplasma contamination to ensure cell line identity and culture integrity throughout the study.

γ-retroviral vector construction and viral production

An MSLN-specific CAR was generated by linking a single-chain variable fragment specific for human MSLN to the CD28/CD3ζ signaling domains. A programmed cell death protein-1 (PD-1) dominant negative receptor (DNR) was generated by linking the extracellular portion of the PD-1 receptor to the M28z vector sequence.⁸ The CD3ζ domain was modified to contain a single functional immunoreceptor tyrosine–based activation motif, termed 1XX, resulting in a M28z1XXPD1DNR CAR construct. Each CAR sequence was inserted into the SFG γ-retroviral vector (provided by I Riviere, Memorial Sloan Kettering Cancer Center) and linked to a Myc-tag sequence. The CAR-encoding plasmids were then transfected into 293T H29 and 293VecRD114 packaging cells to produce the retrovirus. The sequences for the 28z MSLN CAR, PD-1 DNR, and 1XX CAR constructs used in this study are available in the following patent applications: WO2015188141A9,¹² WO2017040945A1,¹³ and EP3732191A2.¹⁴

T-cell isolation and gene transfer

Peripheral blood leukocytes were isolated from the blood of healthy volunteer donors under a Memorial Sloan Kettering Cancer Center Institutional Review Board–approved protocol. Peripheral blood mononuclear cells (PBMCs) were isolated by low-density centrifugation on Lymphoprep (STEMCELL Technologies) and activated for 2 days with phytohemagglutinin (2 ug/mL; Remel). Two days after isolation, PBMCs were transduced with 293VecRD114-produced retroviral particles encoding for CARs and PD-1 DNR and spinoculated for 1 hour at 1800 g on plates coated with retronectin (15 µg/mL; r-Fibronectin, Takara Bio). Transduced PBMCs were maintained in interleukin-2 (IL-2) (20 UI/mL; Novartis).⁸ Both untransduced and prostate-specific membrane antigen (PSMA)–targeted CAR T cells were used as controls. We have previously reported that a 28z CAR targeting PSMA, an irrelevant antigen in this context, showed no efficacy in MSLN tumor models, comparable with untransduced T cells,¹⁵ with minimal background activity.

CAR T-cell cytotoxicity assays

The cytotoxicity of CAR T cells was determined using a standard ⁵¹Cr-release assay. M28z1XXPD1DNR CAR T cells and control untransduced T cells were cultured with MSLN-overexpressing target cells (KYAE1-M, SKGT4-M) at effector to target (E:T) ratios ranging from 1:4 to 64:1. Cytotoxicity was determined after 18 hours of incubation.¹⁵ Data are reported as the mean of triplicate measurements±SE and were analyzed using Microsoft Excel (Microsoft) or GraphPad Prism (GraphPad Software).

In vitro CAR T-cell proliferation and accumulation assays

CAR T cells were stained with CellTrace Violet cell proliferation stain using a 1 µM working solution, in accordance with the manufacturer’s instructions (Thermo Fisher Scientific). M28z1XXPD1DNR CAR T cells (180e3 CAR T cells/24-well plate) and MSLN-overexpressing tumor cells (60e3 tumor cells/24-well plate; KYAE1-M, SKGT4-M) were plated at an E:T ratio of 3:1 in triplicate. After incubation for 4 and 7 days, T cells were collected, T-cell numbers were counted using a hemacytometer, and CAR T cells were assessed for CAR+ percentage and CellTrace Violet fluorescence intensity using flow cytometry. Plotted CAR T-cell accumulation numbers were adjusted for CAR+percentage.

In vitro CAR T-cell repeat-antigen-stimulation assay

To assess accumulation after repeat antigen stimulation, MSLN-overexpressing tumor cells (KYAE1-M, SKGT4-M) were irradiated and plated at a density of 30e3 tumor cells per cm² in 6-well plates in triplicate. At 24 hours after tumor cell plating, M28z1XXPD1DNR CAR T cells were plated at an E:T ratio of 2:1. After 2 days of incubation, T cells were collected and counted using a hemacytometer and were assessed for CAR+percentage. Plotted CAR T-cell accumulation numbers were adjusted for CAR+ percentage. T cells were then restimulated under the same conditions, for a total of three antigen stimulations.

In vitro cytokine-release assay

M28z1XXPD1DNR CAR T cells (180e3 CAR T cells/24-well plate) and MSLN-overexpressing tumor cells (60e3 tumor cells/24-well plate; KYAE1-M, SKGT4-M) were plated at an E:T ratio of 3:1 in triplicate. After 20 hours of incubation, cell-culture supernatant was collected. Human interferon-gamma, IL-2, and tumor necrosis factor-alpha were quantified using the Luminex Assay (Thermo Fisher Scientific).

Flow cytometry and T-cell proliferation labeling

Fluorochrome-conjugated antibodies to MSLN (anti-MSLN PE-conjugated Rat IgG2A FAB32652P; R&D Systems), CD45, CD3 (PE/Cy7 anti-human CD3 Clone HI3A, 300316; BioLegend), CD4 (APC Mouse Anti-Human CD4, 55349; BD Pharmingen), CD8 (FITC Mouse Anti-Human CD8, 561948; BD Pharmingen), programmed death-ligand 1 (PD-L1), PD-1, and Myc-tag (PE Mouse Anti-Human Myc, 9B11; Cell Signaling Technology) were used. All flow cytometric analyses were performed on either a BD FACSCalibur or an LSR II (BD Biosciences) flow cytometer; data were analyzed using FCS Express software (V.7). Flow cytometry of in vivo xenograft tumor samples was performed after filtration in a 40 µm filter and blocking with an antimouse Fc block.

Gastroesophageal cancer peritoneal carcinomatosis mouse models

The experimental procedures used for the animal studies were approved by the Institutional Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center. NSG mice (The Jackson Laboratory) were injected intraperitoneally with 5e6 KYAE1-GM or SKGT4-GM gastroesophageal adenocarcinoma cells expressing MSLN and GFP-luciferase in 200 µL of serum-free media. M28z1XXPD1DNR CAR T cells or control (untransduced) T cells were injected in 200 µL of serum-free media by intraperitoneal or intravenous injection in doses of 1e5, 2e5, or 5e5 cells per mouse. Flank tumor was established via injection of 5e6 KYAE1-M or KYAE tumor cells in serum-free media. Flank-tumor volumes were calculated as tumor volume (mm³) = [width (mm)]² × [length (mm)]/2. Ascites fluid was obtained from the mouse peritoneum after euthanasia, and the TGF-β1 concentration was determined using the Luminex assay (Thermo Fisher Scientific).

Imaging

Tumor growth was monitored by BLI through intraperitoneal injection of D-luciferin. The peritoneal BLI signal was quantified after obtaining a ventral image (0.5 s exposure). After euthanasia, the abdomen of mice with KYAE1-GM peritoneal tumor burden was imaged using a Zeiss Lumar version 12 stereoscope at 6.4× magnification. Fluorescein isothiocyanate-channel images (representing GFP-tagged tumor cells) were overlaid on bright-field images. Ultrasound images of mouse peritoneal tumor were obtained using the Vevo 2100 Imaging Platform (VisualSonics) with a 40-MHz probe.

Statistics

Data were analyzed using GraphPad Prism software (V.6.0) and are presented as mean±SE of the mean, as stated in the figure legends. Results were analyzed using unpaired Student’s t-tests (two-tailed). Survival curves were analyzed using the log-rank test. Statistical significance was defined as p<0.05.

Results

We first established clinically relevant mouse models of peritoneal carcinomatosis in NSG mice using gastroesophageal adenocarcinoma cells (KYAE-1 and SK-GT-4) transduced with MSLN and GFP-luciferase (to facilitate monitoring of tumor burden by BLI); these models demonstrated ascites with high levels of TGF-β (median ascites TGF-β1 level, 1,800–6,000 pg/mL) and bowel obstruction (figure 1A–E). We next established a model with peritoneal carcinomatosis and distal antigen-expressing left-flank and non-antigen-expressing right-flank tumors to investigate the response to therapy at the primary peritoneal site and the distant metastatic site after intraperitoneal versus intravenous treatment (figure 1F). An MSLN-targeted CAR (M28z1XXPD1DNR; figure 2A) was genetically engineered to express a CD3ζ domain containing a single functional immunoreceptor tyrosine-based activation motif, termed 1XX (which increases persistence), CD28 costimulation, and a PD-1 DNR; the DNR serves as a decoy receptor that lacks an inhibitory intracellular signaling domain, thereby reducing T-cell exhaustion.^{16 17} When cocultured with M281XXPD1DNR CAR T cells, gastroesophageal adenocarcinoma cells with variable expression of MSLN and PD-L1 (figure 2B) demonstrated antigen intensity-dependent cytotoxicity (figure 2C, right panel), accumulation of CAR T cells (figure 2C, middle panel), and effector cytokine secretion (online supplemental figure 1). Intraperitoneal administration of M28z1XXPD1DNR CAR T cells achieved better antitumor efficacy than intravenous administration, even at a fivefold lower dose (figure 2C, right panel). Similar results were observed in a different model, with SKGT-4 cells (figure 2D, E).

Figure 1. Gastroesophageal adenocarcinoma peritoneal carcinomatosis models. (A) Representative images from the peritoneal carcinomatosis mouse model showing tumor burden by BLI at days 9 and 23 after tumor-cell injection in NSG mice. (B) Fluorescent stereoscope imaging at 6.4× magnification demonstrates peritoneal carcinomatosis (left panel), and ultrasound imaging using a high-frequency probe shows ascites, peritoneal tumor, and bowel obstruction (right panel). Data shown are representative of eight independent biological replicates for each cell line. (C) Representative images of peritoneal tumors showing H&E staining (left panel) and MSLN immunohistochemical staining (right panel), confirming tumor establishment and high MSLN expression in vivo. (D) Bar chart showing TGF-β1 levels detected in malignant ascites derived from two different tumor models: KYAE1 (n=10 of 12 mice) and SKGT4 (n=7 of 12 mice). (E) Median survival after intraperitoneal tumor injection was 34 days for the KYAE1-GM model and 65 days for the SKGT4-GM model. Each experiment included eight mice per cell line. Data are representative of five independent experiments. (F) Schematic representation of mouse models with peritoneal and distant tumor sites. NSG mice were engrafted with an antigen-expressing tumor in the left flank, an antigen-negative tumor in the right flank, and an antigen-expressing GFP-luciferase+peritoneal tumor. After the establishment of both flank and peritoneal tumors, mice were treated with either intraperitoneal or intravenous CAR T-cell administration. P values were determined by two-tailed, unpaired Student’s t-test. **p<0.01, ***p<0.001. BLI, bioluminescence imaging; CAR, chimeric antigen receptor; GFP, green fluorescent protein; IHC, immunohistochemical; IP, intraperitoneal; IV, intravenous; MSLN, mesothelin.

Figure 2. Intraperitoneal CAR T-cell administration demonstrates antitumor efficacy and longer survival, compared with intravenous administration. (A) MSLN-targeted CAR M28z1XXPD1DNR expressing a CD28 costimulatory domain, a PD-1 DNR, and a CD3ζ domain containing a single functional immunoreceptor tyrosine–based activation motif, termed 1XX. (B) Flow cytometry histograms demonstrate in vitro MSLN surface expression in KYAE-1 cells (KYAE1-G with native MSLN and KYAE1-GM with MSLN overexpression). PD-L1 surface expression was measured with and without IFN-γ (3.13 ng/mL and 200 ng/mL) for 24 hours. Each condition was assessed in triplicate (n=3 technical replicates); data shown are representative of three independent experiments. (C) In vitro chromium cytotoxicity assay demonstrating M28z1XXPD1DNR CAR T-cell and control untransduced T-cell cytotoxicity against KYAE-G (first panel) and KYAE-GM (second panel). Cytotoxicity at 18 hours was assessed over a range of effector to target ratios. CAR T-cell accumulation on repeated stimulation with KYAE1-GM cells in vitro (third panel) and Kaplan-Meier survival curve showing survival differences after intraperitoneal (green line, 1e5 dose) versus intravenous (blue line, 5e5 dose) administration of M28z1XXPD1DNR CAR T cells in KYAE1-GM tumor–bearing mice, compared with untreated mice (black line, control). Data shown in the first three panels, from left to right, are representative of three independently repeated experiments using at least three independent donors; the fourth panel shows data from eight mice per cohort. (D) MSLN and PD-L1 expression measured for SK-GT-4 cells. Each condition was assessed in triplicate (n=3 technical replicates); data shown are representative of three independent experiments. (E) In vitro sensitivity to lysis by CAR T cells in a chromium-release assay and CAR T-cell accumulation on repeated stimulation with SKGT4-GM cells in vitro. The Kaplan-Meier survival curve shows a comparison of an intraperitoneal low dose (green line, 5e4) versus an intraperitoneal high dose (red line, 1e5 dose) of M28z1XXPD1DNR CAR T cells versus no treatment (dotted black line, control). Data shown in the first three panels, from left to right, are representative of three independently repeated experiments using at least three independent donors; the fourth panel shows data from eight mice per cohort. P values were determined by two-tailed, unpaired Student’s t-test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. CAR, chimeric antigen receptor; DNR, dominant negative receptor; E:T, effector to target; FMO, fluorescence minus one; IFN-γ, interferon-gamma; MSLN, mesothelin; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1; SA, splice acceptor; SD, splice donor; UT, untransduced.

To investigate the functional persistence of intraperitoneal M28z1XXPD1DNR CAR T cells, mice in which tumor burden was eradicated, with long-term survival of >84 days after injection of 1e5 CAR T cells, were rechallenged with repeat intraperitoneal injection of tumor cells. We observed functional persistence of CAR T cells, with a reduction in BLI to background levels by day 5 after rechallenge (figure 3A). The PD-1 DNR construct likely contributes to the sustained CAR T-cell activity observed in long-term studies (figure 3A) by preventing PD-1 signaling and supporting persistence in the tumor microenvironment. A similar benefit of PD-1 DNR was previously reported in a different tumor model, where CAR T cells expressing PD-1 DNR outperformed those lacking the PD-1 DNR construct.¹⁵

Figure 3. Antitumor efficacy at regional and distant sites of disease after intraperitoneal administration of CAR T cells. (A) Schematic of the tumor rechallenge study showing a reduction in initial BLI signal after CAR T-cell treatment and a temporal increase in BLI signal after tumor rechallenge on day 84, which began to subside to baseline by day 89, demonstrating functional persistence of CAR T cells (n=10 biological replicates). (B) Left panel: schematic of the dual-flank model with intraperitoneal tumor burden from KYAE1-GM cells; MSLN antigen is expressed in the left flank (KYAE1-M), and antigen is not expressed in the right flank (KYAE1-E). Middle panel: differences in the reduction of left-flank tumor volume after intraperitoneal (blue line) or intravenous (green line) administration of CAR T cells, compared with treatment with PSMA-targeted CAR T cells. Right panel: corresponding reduction in BLI signal of peritoneal tumor in the same mice. Data shown are representative of two technical replicates using cells from two independent donors. PSMA group, n=5; IP and IV groups, n=7. (C) Left panel: schematic showing experimental cohorts where ex vivo–activated CAR T cells were administered intraperitoneally (1a) or non-activated CAR T cells were administered either intraperitoneally (1b) or intravenously (1c) to dual-flank model mice without peritoneal tumors, and non-activated CAR T cells were administered either intraperitoneally (2a) or intravenously (2b) to KYAE1-GM peritoneal tumor-bearing dual-flank model mice. Middle panel: the corresponding dynamics of reduction in left-flank tumor burden in mice from cohort 1a and 2a and quantification of CAR T cells per gram of tumor detected in flank tumors collected from each cohort are shown (right panel). Data shown are representative of two technical replicates using cells from two independent donors with n=7 mice per cohort. P values were determined by two-tailed, unpaired Student’s t-test. **p<0.01, ***p<0.001, ****p<0.0001. BLI, bioluminescence imaging; CAR, chimeric antigen receptor; IP, intraperitoneal; IV, intravenous; L, left; MSLN, mesothelin; PSMA, prostate-specific membrane antigen; R, right; UT, untransduced.

Patients with metastatic gastroesophageal adenocarcinoma present with multiple sites of metastases in 22–78% of cases. To address this issue, NSG mice were established with a combined tumor model consisting of antigen-expressing left-flank tumor, antigen-negative right-flank tumor, and antigen-expressing GFP-luciferase+peritoneal tumor. After the establishment of both flank and peritoneal tumors (figure 3B), mice were treated with 2e5 PSMA M28z1XXPD1DNR CAR T cells via either intraperitoneal or intravenous delivery. All mice (n=7) that received intraperitoneal CAR T cells exhibited reduced distal antigen-positive flank-tumor volume by day 11 and eradication of peritoneal tumor by day 7 (figure 3B). Reduction in flank-tumor burden was slower in mice that received intravenous CAR T cells (day 17). Median survival was not reached in either group, but all mice that received intraperitoneal CAR T cells were alive at 30 days; in contrast, three of seven mice that received intravenous CAR T cells were euthanized on days 11 to 12 because of tumor burden.

We next investigated whether the superior antitumor efficacy at distant sites after intraperitoneal CAR T-cell administration was attributable to mere antigen-driven proliferation (in vivo dose expansion) or additional improved distal tumor infiltration. To test this, we designed experiments with mouse cohorts (n=7) with or without (figure 3C) peritoneal antigen-expressing tumors, in addition to MSLN-expressing left-flank tumors and antigen-negative right-flank tumors. Intravenous administration of ex vivo antigen-activated CAR T cells resulted in pulmonary toxicity in most mice (antigen-activated CAR T cells are larger and granular, with higher pulmonary capillary sequestration, and they secrete cytokines). Intraperitoneally administered ex vivo antigen-activated CAR T cells infiltrated flank tumors at similar levels as CAR T cells activated by intraperitoneal MSLN-expressing tumors, whereas intraperitoneally administered CAR T cells without ex vivo antigen activation and CAR T cells administered against antigen-negative intraperitoneal tumors had weak infiltration (figure 3C). To assess CAR T-cell specificity, we analyzed the contralateral MSLN⁻ tumors across all treatment groups. As shown in online supplemental figure 1D, CAR T-cell accumulation and CD8⁺ T-cell frequency in MSLN⁻ tumors remained low across all conditions, regardless of delivery route. These findings confirm minimal infiltration into antigen-negative sites and support the antigen-dependent activity of CAR T cells in this model. Antigen expression in the left flank tumor and no antigen expression in the right flank tumor were confirmed by immunohistochemistry of harvested flank tumors after euthanasia.

Discussion

In clinically relevant mouse models of gastroesophageal adenocarcinoma with peritoneal carcinomatosis that included ascites, bowel obstruction, and an immune inhibitory microenvironment with high levels of TGF-β and with tumor-cell PD-L1 expression, intraperitoneally administered MSLN-targeted CAR T cells that had been genetically tuned to the tumor immune microenvironment were effective both intraperitoneally and at distant sites. Intraperitoneally delivered CAR T cells displayed functional persistence and were associated with superior survival; similar results have been demonstrated after intracerebroventricular and intratumoral delivery for brain tumors in early-phase clinical trials.¹⁸ Intraperitoneal CAR T cells were enriched with CD8+ cells and had higher levels of HLA-DR activation, compared with intravenously delivered or in vitro-activated cells. Additionally, in situ antigen activation in the peritoneum was associated with better CAR T-cell expansion and infiltration into flank tumors, compared with ex vivo activation.

On the basis of these observations, a phase I trial of intraperitoneal MSLN-targeted CAR T-cell therapy for esophagogastric adenocarcinoma with MSLN-positive peritoneal carcinomatosis has been initiated, has received Investigational New Drug, Institutional Review Board, and US Food and Drug Administration approvals, and is actively recruiting (ClinicalTrials.gov identifier: NCT06623396).¹⁹ Participants will receive intravenous lymphodepletion with cyclophosphamide followed by escalating doses of a single infusion of intraperitoneal M28z1XXPD1DNR CAR T cells (figure 4).

Figure 4. Phase I trial of intraperitoneal MSLN-targeted CAR T-cell therapy for esophagogastric adenocarcinoma with MSLN-positive peritoneal carcinomatosis. (A) Study design and treatment timeline. Eligible patients undergo leukapheresis for CAR T-cell manufacturing (∼3–4 weeks). Bridging therapy may be administered until the CAR T-cell product is available. Preconditioning with intravenous fludarabine or cyclophosphamide is administered. Patients with esophagogastric adenocarcinoma receive one of five CAR T-cell doses via intraperitoneal infusion on Day 0. After infusion, patients are monitored in the outpatient clinic for 4 weeks. Imaging is performed every 8 weeks for 6 months and then every 8–12 weeks for up to 2 years. (B) Cohort design and dose-escalation strategy. A minimum of 4 and a maximum of 18 patients will receive escalating doses of CAR T cells. Patients will be enrolled in cohorts in accordance with the study design, with an initial dose level (dose level 2) requiring a 14-day staggered interval between the first two patients. Dose-limiting toxicities will be monitored from Day 0 to Day 30 after administration. All patients in a cohort must complete the 30-day dose-limiting toxicity observation period before the next cohort is advanced to. The modified continuous reassessment method will guide dose escalation, with patients accrued in cohorts of two. Only eligible and evaluable patients will be considered for dose determination. CAR, chimeric antigen receptor; MSLN, mesothelin.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.