Advances, challenges, and innovative strategies of CAR-T cell therapy in pancreatic cancer

Abstract

Pancreatic cancer remains one of the most lethal malignancies worldwide, with pancreatic ductal adenocarcinoma (PDAC) accounting for the majority of cases and exhibiting a persistently poor prognosis. Although chimeric antigen receptor T (CAR-T) cell therapy has achieved remarkable success in hematological malignancies, its therapeutic efficacy in pancreatic cancer remains limited. This review summarizes recent advances in CAR-T cell therapy for pancreatic cancer, with a focus on representative tumor-associated targets, including mesothelin (MSLN), claudin 18.2 (CLDN18.2), prostate stem cell antigen (PSCA), CD155, CD276, growth arrest-specific protein 6 (GAS6), and glypican-1 (GPC1), while also highlighting emerging next-generation CAR-T engineering strategies, including nanobody-based antigen recognition, cytokine-armored CAR-T cells, allogeneic CAR-NKT platforms, dual-target and logic-gated CAR systems, and innovative delivery approaches. Current preclinical and early clinical evidence suggests that several targets, particularly MSLN and CLDN18.2, show promising antitumor activity; however, durable clinical responses remain difficult to achieve. The major barriers include the dense desmoplastic stroma, highly immunosuppressive tumor microenvironment (TME), antigen heterogeneity, antigen loss, limited CAR-T persistence and expansion, T-cell exhaustion, and on-target, off-tumor toxicity. To overcome these obstacles, emerging strategies have focused on remodeling the TME, engineering armored or dual-target CAR-T cells, developing logic-gated CAR systems, improving CAR-T persistence, and optimizing delivery approaches through nanomaterials, oncolytic viruses, in situ CAR-T generation, alternative immune-cell carriers, and locoregional administration. Overall, CAR-T therapy for pancreatic cancer is progressing from preclinical exploration toward clinical translation. Future success will likely depend on rational target selection, multi-dimensional TME modulation, advanced CAR engineering, precision delivery, and biomarker-guided patient stratification.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40364-026-00978-0.

Introduction

Pancreatic cancer (PC) remains one of the most lethal malignancies worldwide, with pancreatic ductal adenocarcinoma (PDAC) representing the predominant histological subtype, accounting for nearly 90% of cases [1]. Despite notable geographic variations in incidence, epidemiological data indicate a rising global trend in PDAC diagnoses [2]. Projections suggest that by 2040, PDAC will surpass colorectal cancer, becoming the second leading cause of cancer-related mortality after lung cancer [3]. Advances in molecular and clinical research have elucidated the complex interplay between genetic predisposition and environmental risk factors; however, the overall prognosis remains dismal, with a 5-year survival rate of approximately 10% [4]. Surgical resection remains the primary curative modality, yet its applicability is limited, as most patients present with locally advanced or metastatic disease, precluding operative intervention [5]. Consequently, the development of novel therapeutic strategies with favorable safety profiles is critical to improve clinical outcomes.

Immunotherapeutic approaches leveraging the host immune system have emerged as a central focus in oncology, among which chimeric antigen receptor T cell (CAR-T) therapy has garnered considerable attention. In 1989, Gross and colleagues first demonstrated the feasibility of genetically redirecting cytotoxic T lymphocytes to recognize tumor cells, proposing that T cells engineered with tumor-specific chimeric antigen receptors could mediate antitumor activity [6]. This seminal work laid the foundation for the subsequent development of CAR-T therapies. CAR-T therapy has achieved remarkable clinical efficacy in hematological malignancies, including diffuse large B-cell lymphoma, B-cell non-Hodgkin lymphoma, multiple myeloma, and acute B-cell lymphoblastic leukemia [7, 8]. However, its application in solid tumors faces significant challenges [9]. Among solid tumors, pancreatic cancer represents one of the most challenging indications for CAR-T therapy. Although CAR-T approaches targeting antigens such as CLDN18.2 in gastric cancer, GPC3 in hepatocellular carcinoma, EGFRvIII/B7-H3 in glioblastoma, and CEA in colorectal cancer have shown encouraging preclinical or early clinical results, durable clinical responses remain difficult to achieve across solid tumors. Pancreatic cancer presents additional obstacles due to its dense desmoplastic stroma, highly immunosuppressive tumor microenvironment, low immunogenicity, and pronounced antigen heterogeneity, which collectively restrict CAR-T infiltration, persistence, and therapeutic efficacy. (Fig. 1).

Fig. 1.

Overview of CAR-T cell therapy in pancreatic cancer. This schematic illustrates CAR-T cell therapy targeting pancreatic cancer. Key tumor-associated antigens (Mesothelin, CLDN18.2, PSCA, CD133, CD155, CD276, GPC1, GAS6, IL-10) are shown within the tumor microenvironment. The figure highlights challenges including short-term persistence, limited expansion, immunosuppressive TME, and off-target toxicity. Experimental models for evaluation, including cell, animal, and clinical studies, are also depicted

Despite these challenges, CAR-T therapy represents a promising avenue for PDAC treatment. This review systematically evaluates the current state of CAR-T therapy in clinical studies for PDAC, delineates tumor-specific obstacles, and highlights emerging strategies designed to overcome these barriers. Beyond summarizing conventional antigen-targeted CAR-T approaches, this review further emphasizes innovative CAR engineering concepts, including alternative antigen-recognition platforms, immune-cell redesign, TME-remodeling strategies, and advanced therapeutic delivery systems, which may collectively reshape the future landscape of CAR-T therapy for pancreatic cancer. By synthesizing recent advances and identifying potential avenues for therapeutic optimization, this work aims to provide a comprehensive framework for the rational development of CAR-T therapies in pancreatic cancer.

CAR-T cell therapy

Chimeric antigen receptor T (CAR-T) cell therapy is an immunotherapeutic approach in which a patient’s own T cells are genetically engineered in vitro to specifically recognize and attack tumor cells [10]. The process begins with isolating T cells from the patient’s peripheral blood, followed by the introduction of a gene encoding a chimeric antigen receptor (CAR) capable of binding tumor-associated surface antigens. The modified T cells are then expanded in vitro and reinfused into the patient via intravenous administration [11]. Since 2017, the U.S. Food and Drug Administration (FDA) has approved six CAR-T cell therapies for the treatment of hematological malignancies such as leukemia and lymphoma [12].

CAR-T cells have been developed into five successive generations. First-generation CAR-T cells: The simplest structure, composed of an extracellular antigen recognition domain linked to an intracellular CD3ζ signaling domain, which mimics natural T cell activation. However, these cells exhibit limited persistence and minimal antitumor activity, resulting in low therapeutic efficacy [13]. Second-generation CAR-T cells: FDA-approved CAR-T therapies are based on this generation, which incorporates one co-stimulatory domain into the intracellular region, enhancing persistence and antitumor efficacy in vivo [14]. Third-generation CAR-T cells: These cells further integrate an additional co-stimulatory domain, augmenting T cell activation, proliferation, and cytotoxic potential [15]. Fourth-generation CAR-T cells: Also referred to as TRUCKs or “armored” CAR-T cells, these are engineered to secrete cytokines, express anti-PD-1 antibodies, or include suicide genes, thereby enhancing therapeutic potential by overcoming inhibitory factors [16]. Fifth-generation CAR-T cells: Similar to the second generation but containing a truncated intracellular IL-2 receptor β-chain domain, which enables drug-dependent regulation of CAR-T activity, potentially optimizing therapeutic outcomes (Fig. 2) [17].

Fig. 2.

Five generations of CAR-T cell engineering and functional improvements. Structural evolution of CAR-T cells across five generations. First-generation CAR-T contains only the CD3ζ signaling domain. Second- and third-generation CAR-Ts include one or two co-stimulatory domains to enhance activation and proliferation. Fourth-generation “Armored CAR-T” modulates the tumor microenvironment through cytokine secretion (e.g., IL-12/IL-15/IL-18). Fifth-generation CAR-T incorporates a JAK/STAT signaling module for improved persistence, activation, and therapeutic efficacy

While CAR-T therapy has demonstrated remarkable efficacy in hematological malignancies, its application in solid tumors such as pancreatic cancer remains limited. Several factors contribute to this challenge: (1) the dense physical barrier of solid tumors impedes CAR-T cell infiltration; (2) the immunosuppressive tumor microenvironment secretes inhibitory factors, such as transforming growth factor-β (TGF-β), that impair CAR-T function; and (3) CARs typically target surface antigens, whereas solid tumors often lack unique tumor-restricted antigens that can be safely targeted without harming normal tissues [18, 19].

Current status of CAR-T cell therapy in pancreatic cancer

Clinical studies of chimeric antigen receptor T (CAR-T) cell therapy in pancreatic ductal adenocarcinoma (PDAC) have primarily focused on identifying and validating suitable tumor-associated antigens. Due to the high heterogeneity and immunosuppressive tumor microenvironment (TME) of PDAC, single-target CAR-T strategies often exhibit limited efficacy. Consequently, multiple antigen targets and combinatorial strategies, including engineered CAR-T modifications, have been explored to enhance therapeutic outcomes (Fig. 4). This section summarizes the progress of CAR-T therapy targeting key antigens, including mesothelin (MSLN), claudin 18.2 (CLDN18.2), prostate stem cell antigen (PSCA), CD133, and HER2, which represent the main directions of current research.

Fig. 4.

Major barriers to CAR-T cell therapy in pancreatic cancer. The figure summarizes key obstacles to CAR-T efficacy in pancreatic cancer. Barriers include: (1) physical barriers from dense ECM and CAFs; (2) immunosuppressive cytokines (TGF-β, IL-10, VEGF, PD-L1, IDO1) inhibiting CAR-T expansion/function; (3) checkpoint signaling (PD-1/PD-L1) inducing CAR-T exhaustion; (4) Fas/FasL-mediated apoptosis; (5) limited CAR-T infiltration due to stromal and TME constraints

Mesothelin (MSLN)

Mesothelin (MSLN) is a glycoprotein broadly expressed in multiple malignancies, including pancreatic, ovarian, lung, and gastric cancers, and has therefore become an important target for pancreatic cancer immunotherapy. The study titled Proton radiation boosts the efficacy of mesothelin-targeting chimeric antigen receptor T cell therapy in pancreatic cancer demonstrated in subcutaneous and orthotopic mouse models of PDAC that proton radiation markedly upregulated MSLN expression on the surface of tumor cells and promoted CAR-T cell infiltration into tumor sites. Combination therapy not only significantly inhibited local tumor growth and prolonged survival but also induced distal antitumor effects in bilateral tumor models [20].

In contrast, Systematic Interrogation of Tumor Cell Resistance to Chimeric Antigen Receptor T-cell Therapy in Pancreatic Cancer investigated resistance mechanisms to MSLN-targeted CAR-T therapy using genome-wide clustered regularly interspaced short palindromic repeats-associated protein 9 (CRISPR-Cas9) screening. Validation in multiple PDAC cell lines and xenograft mouse models revealed both antigen-dependent and antigen-independent mechanisms of resistance. In particular, loss of genes involved in glycosylphosphatidylinositol (GPI)-anchor biosynthesis led to loss of surface MSLN expression, representing a major mechanism of antigen-dependent resistance [21].

Lee HH et al. [22]. screened MSLN-specific single-chain variable fragments (scFvs) with optimal binding activity and constructed MSLN CAR-T cells. In vitro assays using PDAC cell lines such as MIA PaCa-2 and AsPC-1, as well as orthotopic pancreatic cancer mouse models, showed that MSLN CAR-T cells specifically recognized and killed MSLN-high PDAC cells, resulting in significant tumor regression and even complete remission in animal models.

Liu X et al. [23]. analyzed proteolytic cleavage sites of MSLN and developed a novel monoclonal antibody, 15B6, targeting the juxtamembrane region of MSLN. This antibody not only inhibited MSLN shedding but was also used to generate CAR-T cells. In cellular and animal models, 15B6 CAR-T cells exhibited stronger antitumor activity than conventional SS1 CAR-T cells targeting distal epitopes, and their activity was not blocked by shed MSLN.

Zhu Y et al. [24]. explored the potential application of natural killer T (NKT) cells and γδ T cells in MSLN-targeted immunotherapy, beyond conventional αβ T cells. Using an MSLN/CD3 bispecific antibody, the study compared the antitumor efficacy and safety of peripheral blood mononuclear cells (PBMCs), NKT cells, and γδ T cells in vitro and in vivo using pancreatic cancer xenograft models. In the presence of the bispecific antibody, γδ T cells demonstrated superior tumor-killing capacity compared with PBMCs and NKT cells, while inducing lower cytokine release and milder systemic toxicity in animal models.

Stock S et al. [25]. developed a modular CAR-T platform targeting the P329G-mutated Fc region of effector-silenced human immunoglobulin G1 (IgG1) antibodies. This P329G-CAR platform indirectly targets tumor antigens by recognizing antibodies carrying P329G LALA Fc mutations. In this study, a P329G-mutated anti-MSLN antibody was used as an adaptor, and its efficacy was evaluated in MIA PaCa-2-MSLN pancreatic cancer cells and xenograft mouse models. P329G-CAR-T cells combined with anti-MSLN antibodies effectively killed pancreatic cancer cells in vitro and significantly inhibited tumor growth in vivo, with antitumor efficacy comparable to conventional directly MSLN-targeted CAR-T cells.

Li D et al. [26]. investigated the enhancing effect of long-acting interleukin-7 (rhIL-7-hyFc, NT-I7) in combination with multiple CAR-T platforms, including MSLN-targeted CAR-T cells, for solid tumor therapy. In a pancreatic cancer model using KLM-1 cells expressing MSLN to establish an intraperitoneal xenograft mouse model, low-dose MSLN CAR-T cells at 1 × 10⁶ cells were combined with NT-I7. MSLN CAR-T monotherapy induced transient tumor regression but failed to prevent tumor regrowth, whereas combination therapy with NT-I7 sustained tumor suppression, significantly prolonged survival, and caused no obvious toxicity.

Chu GJ et al. [27]. systematically evaluated the influence of MSLN antigen density on the cytotoxicity of two clinical-stage anti-MSLN CAR-T products, m912 and SS1. The study found that SS1 CAR-T cells showed stronger cytotoxicity against pancreatic cancer and mesothelioma cell lines expressing relatively low levels of MSLN, below approximately 170,000 molecules per cell, whereas m912 CAR-T cells effectively lysed only high-MSLN-expressing cells. The study further demonstrated for the first time that MSLN shedding inhibitors, including lanabecestat and TMI-1, increased surface MSLN density and significantly enhanced m912 CAR-T cytotoxicity.

Liu XF et al. [28]. focused on overcoming the inhibitory effect of shed MSLN on CAR-T cells and validated their findings in animal models and clinical samples. The team developed a humanized antibody, h15B6, targeting a membrane-proximal MSLN epitope without binding shed MSLN, and used it to construct CAR-T cells. In refractory pancreatic cancer patient-derived xenograft (PDX) models, h15B6 CAR-T cells induced complete tumor regression, whereas SS1 CAR-T cells targeting shed epitopes were ineffective.

Aboalela MA et al. [29]. constructed an oncolytic herpes simplex virus expressing human MSLN (HSV-MSLN) and combined it with MSLN CAR-T cells in mouse PDAC models. HSV-MSLN effectively infected tumor cells in vitro and in vivo and induced MSLN expression, thereby providing a target for CAR-T cells. Mechanistically, HSV-MSLN not only increased tumor antigen density but also remodeled the immunosuppressive TME, as reflected by increased CD8-positive T cell infiltration, reduced regulatory T cells (Tregs) and exhausted T cells, enhanced dendritic cell maturation and migration, and increased intratumoral CAR-T cell activation and proliferation. This approach achieved synergistic effects through viral oncolysis, antigen provision, and immune microenvironment remodeling.

Wehrli M et al. [30]. engineered MSLN CAR-T cells to secrete T cell-engaging antibody molecules (TEAMs) targeting fibroblast activation protein (FAP) on cancer-associated fibroblasts (CAFs) and CD3 on T cells, thereby simultaneously targeting pancreatic cancer cells and stromal fibroblasts within the TME. In preclinical models, mesoFAP CAR-TEAM cells effectively eliminated both tumor cells and fibroblasts, enhancing antitumor efficacy.

Wang Y et al. [31]. developed MSLN-targeted CAR-T cells secreting programmed death ligand 1 (PD-L1)-blocking single-chain variable fragments, referred to as Sec-MesoCAR-T cells, to overcome immune suppression in pancreatic cancer and improve therapeutic efficacy. Sec-MesoCAR-T cells exhibited stronger antitumor activity than conventional MesoCAR-T cells both in vitro and in vivo. They also enhanced the secretion of cytokines such as interleukin-2 (IL-2), interleukin-6 (IL-6), and interferon-γ (IFN-γ), improved CAR-T expansion and persistence, and thereby increased therapeutic efficacy against pancreatic cancer.

Claudin 18.2 (CLDN18.2)

Claudin 18.2 (CLDN18.2) is frequently expressed in malignant tumors, including PDAC. The high efficacy of claudin18.2-targeted CAR-T cell therapy in advanced pancreatic cancer with an antibody-dependent safety strategy reported early results from three patients with advanced pancreatic cancer treated with CLDN18.2-targeted CAR-T cells carrying a truncated epidermal growth factor receptor (tEGFR) safety switch. All patients achieved disease regression after a single infusion, including one complete response (CR) and two partial responses (PRs), indicating substantial antitumor activity in advanced pancreatic cancer. However, all patients developed varying degrees of gastric mucosal injury, defined as on-target off-tumor (OTOT) toxicity, which was associated with physiological CLDN18.2 expression in normal gastric tissue. Severe OTOT was controlled by cetuximab-mediated elimination of tEGFR-expressing CAR-T cells, supporting the clinical feasibility of antibody-dependent safety-switch strategies [32].

CXCR4-modified CAR-T cells suppresses MDSCs recruitment via STAT3/NF-κB/SDF-1α axis to enhance efficacy against pancreatic cancer investigated the therapeutic potential of CLDN18.2-targeted CAR-T cells coexpressing C-X-C chemokine receptor type 4 (CXCR4) in PDAC. Using immunocompetent mouse models, including PANC02-A2 and KPC allograft models, the study found that CXCR4-coexpressing CAR-T cells infiltrated tumor sites more efficiently and displayed stronger antitumor activity in vivo [33].

CT041 CAR T cell therapy for Claudin18.2-positive metastatic pancreatic cancer reported two patients with metastatic pancreatic cancer treated with CT041, a CLDN18.2-targeted CAR-T product. Both patients received CT041 infusion after failure of standard therapy and achieved PR. One patient showed marked shrinkage of lung metastases, whereas the other achieved complete remission of pulmonary metastases, with responses maintained until the last follow-up. During treatment, both patients developed mild to moderate cytokine release syndrome (CRS), which was controlled with tocilizumab [34].

Complete remission of advanced pancreatic cancer induced by claudin18.2-targeted CAR-T cell therapy: a case report described a 72-year-old male patient with advanced PDAC who achieved CR after CLDN18.2-targeted CAR-T therapy. The patient had previously undergone pancreaticoduodenectomy and multiple lines of chemotherapy but developed liver, peritoneal, and cervical lymph node metastases. After enrollment in a CLDN18.2 CAR-T clinical trial (NCT05620732), the patient received CAR-T cells at a dose of 1.2 × 10⁶ cells/kg. Grade 2 CRS and gastric mucosal injury occurred after infusion but were controlled with supportive management. CR was achieved one month after treatment and maintained for eight months, followed by CLDN18.2-negative relapse [35].

FAP-targeted CAR-T suppresses MDSCs recruitment to improve the antitumor efficacy of claudin18.2-targeted CAR-T against pancreatic cancer investigated the synergistic effects and mechanisms of sequential FAP-targeted CAR-T and CLDN18.2-targeted CAR-T therapy in pancreatic cancer. In immunocompetent mouse models, including PANC02 and KPC1199 pancreatic cancer allografts, initial infusion of FAP CAR-T cells effectively depleted CAFs in the TME and suppressed recruitment of myeloid-derived suppressor cells (MDSCs). This remodeling of the immunosuppressive TME enhanced subsequent CLDN18.2 CAR-T infiltration and antitumor activity. Sequential therapy produced stronger tumor suppression than either CAR-T therapy alone and caused no obvious toxicity [36].

Preclinical development and a case report of a nanobody-based CLDN18.2 CAR-T IMC002 with reduced on-target off-tumor toxicity described IMC002, a nanobody-based CLDN18.2 CAR-T product. Preclinical studies in multiple cell-derived xenograft and PDX models of gastric and pancreatic cancers demonstrated its antitumor activity and safety. Compared with conventional scFv-based CAR-T constructs, IMC002 showed lower OTOT toxicity and improved tolerability, with a maximum non-severe toxic dose of 5 × 10⁸cells/kg. Pharmacokinetic data showed enrichment of IMC002 in tumor tissue, with delayed and limited accumulation in gastric tissue. The clinical case section reported one patient with advanced gastric cancer who achieved pathological CR after IMC002 treatment and remained recurrence-free during follow-up [37].

Local radiotherapy polarized tumor-associated macrophages enhance the efficacy of Claudin18.2-targeted CAR-T therapy in pancreatic cancer generated second-generation anti-CLDN18.2 CAR-T cells using CLDN18.2-overexpressing Panc02 cells and mPAKPC pancreatic cancer cell lines. In vitro assays confirmed cytotoxicity against antigen-positive cells, and radiotherapy at 8 Gy significantly enhanced CAR-T-mediated killing of CLDN18.2-positive pancreatic cancer cells. In Panc02-CLDN18.2 and mPAKPC subcutaneous tumor models, local radiotherapy combined with CAR-T therapy significantly inhibited tumor growth, prolonged survival, and induced complete remission in some mice, particularly in the context of cyclophosphamide preconditioning [38].

Although CLDN18.2-targeted CAR-T therapy has shown encouraging preliminary activity in pancreatic cancer, its clinical evidence remains immature. Current data are mainly derived from small-scale early-phase clinical studies and individual case reports, with limited patient numbers, relatively short follow-up periods, and no randomized controlled trials available to date. Therefore, whether CLDN18.2 CAR-T therapy can achieve durable clinical benefit, improve overall survival, and provide advantages over existing treatment strategies remains unclear. Larger prospective clinical trials with longer follow-up and appropriate comparative designs are required to further establish its clinical efficacy and safety profile.

Prostate Stem Cell Antigen (PSCA)

Prostate stem cell antigen (PSCA) is a potential pancreatic cancer biomarker characterized by increased transcriptional expression in PDAC and association with reduced survival [39]. Off-the-shelf invariant NKT cells expressing anti-PSCA CAR and IL-15 promote pancreatic cancer regression in mice investigated cryopreserved allogeneic invariant NKT cells expressing anti-PSCA CAR and soluble interleukin-15 (sIL-15), referred to as CAR-iNKT cells, for pancreatic cancer therapy. In cellular models, CAR-iNKT cells specifically killed PSCA-high pancreatic cancer cell lines, including Capan-1 and MIA PaCa-2, and overcame gemcitabine resistance. In metastatic and orthotopic pancreatic cancer mouse models, CAR-iNKT cells showed potent antitumor activity after intraperitoneal or combined intravenous administration, significantly suppressing tumor progression and prolonging survival without obvious systemic toxicity or graft-versus-host disease (GvHD). Compared with CAR-T cells carrying the same CAR, CAR-iNKT cells achieved comparable antitumor efficacy with improved safety [40].

IL-7 armed binary CAR T cell strategy to augment potency against solid tumors constructed a dual-antigen CAR-T system targeting PSCA and mucin 1 (MUC1). The authors further designed a binary system in which IL-7 secretion was introduced into PSCA CAR-T cells (C.P7), whereas IL-7 receptor overexpression was introduced into MUC1 CAR-T cells (C.M7R), thereby providing local cytokine support and enhancing T cell persistence. In two-dimensional coculture systems and three-dimensional tumor spheroid models based on the CAPAN-1 pancreatic cancer cell line, this binary system showed stronger tumor suppression and T cell expansion than single-target or unmodified dual-target CAR-T cells. In NSG mouse subcutaneous pancreatic cancer models, combined C.P7 and C.M7R therapy also demonstrated enhanced antitumor activity and sustained in vivo T cell expansion [41].

Comparative efficacy and safety of PSCA CAR-engineered Vδ1 γδ T cells for immunotherapy of pancreatic cancer evaluated PSCA CAR-engineered Vδ1 γδ T cells in pancreatic cancer. In vitro killing assays using PSCA-high pancreatic cancer cell lines such as Capan-1 and MIA PaCa-2 showed that PSCA CAR-Vδ1 γδ T cells effectively lysed tumor cells through a CAR-dependent mechanism, accompanied by granzyme B and IFN-γ release. These cells retained activity after cryopreservation, supporting their potential as an off-the-shelf therapy. In metastatic and orthotopic pancreatic cancer xenograft mouse models, PSCA CAR-Vδ1 T cells significantly inhibited tumor growth, promoted tumor infiltration, and prolonged survival. Compared with PSCA CAR-αβ T cells, they did not induce GvHD or CRS. Single-cell transcriptomic analysis further showed lower exhaustion than PSCA CAR-Vδ2 T cells. Comparative analysis of PSCA CAR-Vδ1, Vδ2, and αβ T cells indicated similar antitumor efficacy, with Vδ1 cells showing advantages in safety and persistence [42].

Pre-conditioning modifies the TME to enhance solid tumor CAR T cell efficacy and endogenous protective immunity used an innovative human PSCA knock-in (hPSCA-KI) immunocompetent mouse model that mimics physiological PSCA expression in normal tissues such as prostate, bladder, and stomach, enabling simultaneous assessment of safety and antitumor efficacy. In immunocompetent mice, PSCA CAR-T monotherapy showed limited efficacy against subcutaneous or bone-metastatic PSCA-positive prostate and pancreatic cancer models. However, cyclophosphamide preconditioning before CAR-T infusion markedly enhanced antitumor responses, induced complete remission, and prolonged survival. Mechanistically, cyclophosphamide not only mediated lymphodepletion but also remodeled the TME by promoting inflammatory myeloid conversion, enhancing antigen presentation, and improving the infiltration and expansion of infused CAR-T cells and endogenous T cells. Importantly, treatment did not cause obvious toxicity in PSCA-expressing normal tissues and induced antigen-independent protective immune memory, enabling cured mice to resist rechallenge with PSCA-negative tumors [43].

CD family targets

CD133 is the first identified cancer stem cell (CSC) marker in pancreatic cancer. CD133-positive CSCs are highly tumorigenic, resistant to standard chemotherapy, and critical for metastasis. Yao Y et al. [44]. synthesized CD133-targeted multifunctional nanomicelles, CD133-FITC/PFH@OSMs, for dual-modality imaging and synergistic high-intensity focused ultrasound (HIFU) ablation in pancreatic cancer. In mouse pancreatic cancer models, multiphoton fluorescence microscopy after injection of CD133-FITC/PFH@OSMs showed specific accumulation of the nanomicelles in pancreatic tumor tissue, and the nanomicelles enhanced the efficacy of HIFU therapy.

Wang R et al. [45]. developed multifunctional CD133-targeted hybrid nanovesicles, CD133-grafted Cy5.5/PFOB@P-HVs, which not only targeted CD133-positive pancreatic CSCs but also enabled fluorescence and ultrasound imaging. In vitro experiments demonstrated their ability to kill CD133-positive pancreatic CSCs, and in vivo biodistribution analysis showed preferential accumulation in tumor regions. These two studies support the biological and therapeutic relevance of CD133 in pancreatic cancer, although they represent targeted nanomedicine approaches rather than direct CD133 CAR-T studies.

CD155 is a single-pass transmembrane cell-surface protein highly expressed in pancreatic cancer and represents a promising antitumor target. Xiong et al. [46]. found that the pancreatic cancer cell line Capan-1 highly expressed CD155. In vitro experiments demonstrated that CD155 CAR-T cells effectively killed Capan-1 cells in a dose-dependent manner and significantly increased secretion of tumor necrosis factor (TNF) and IFN-γ. In animal models, NSG mice were intraperitoneally injected with luciferase-labeled Capan-1 cells to establish a pancreatic cancer model. CD155 CAR-T therapy significantly reduced tumor burden and prolonged survival. Mechanistically, CD155 CAR-T cells specifically recognized and bound CD155 on tumor cells, activating T cell-mediated cytotoxicity and enabling targeted elimination of CD155-high tumor cells. In addition, a humanized CD155 CAR-T construct, Hu-B03, retained antitumor activity while reducing immunogenicity and showed favorable systemic safety in immunocompetent mice, without significant body weight loss, organ injury, or neurotoxicity.

CD276, also known as B7-H3, is highly expressed in multiple solid tumors, including pancreatic cancer. Deng T et al. [47]. investigated CD276-targeted CAR-T therapy and introduced a rapid manufacturing method termed Dash CAR-T, which shortened production time. CD276-targeted CAR-T cells showed significant antitumor activity both in vitro and in mouse pancreatic cancer models. Dash CAR-T cells were successfully manufactured within 48–72 h and exhibited stronger proliferative capacity and greater persistence than conventionally manufactured CAR-T cells, providing a potentially more efficient therapeutic approach for pancreatic cancer.

Other targets and engineering strategies

Although MSLN and CLDN18.2 remain the most extensively investigated targets in pancreatic cancer CAR-T therapy, recent advances have expanded the field beyond conventional scFv-based CAR designs. Emerging approaches now include tumor-specific glycoform recognition, cytokine-armored CAR-T cells, alternative immune-cell platforms, nanobody-based antigen-recognition domains, and CAR architectures integrating enhanced specificity and functional regulation. These strategies aim not only to identify improved tumor targets but also to overcome the fundamental limitations imposed by antigen heterogeneity, immunosuppressive TME, poor persistence, and safety concerns.

Tumor-specific glycoform antigens as emerging CAR-T targets

Beyond conventional protein-based tumor-associated antigens, tumor-specific glycoforms have recently emerged as a novel class of targets for CAR-T cell therapy, offering a potential solution to the limited tumor selectivity of conventional antigen recognition strategies in solid tumors. Malignant transformation is frequently accompanied by abnormal glycosylation patterns, including truncated O-glycosylation, altered mucin processing, and increased expression of tumor-associated carbohydrate antigens. These cancer-associated glycan structures can generate tumor-restricted epitopes that are absent or minimally expressed in normal tissues, thereby providing opportunities to improve CAR-T specificity and reduce on-target, off-tumor toxicity [48, 49].

Among glycoform-directed CAR-T strategies, tumor-associated MUC1 glycoforms represent one of the most extensively investigated examples. Normal epithelial MUC1 is highly glycosylated, whereas malignant transformation leads to incomplete O-glycan maturation and exposure of cryptic peptide epitopes, such as the Tn-MUC1 glycoform. Posey et al. developed CAR-T cells specifically recognizing the cancer-associated Tn-MUC1 glycoform and demonstrated selective killing of MUC1-positive adenocarcinoma cells while minimizing recognition of normal tissues, providing a proof-of-concept that post-translational modifications can serve as tumor-selective CAR-T targets beyond conventional protein expression levels [50].

This strategy may be particularly relevant for pancreatic ductal adenocarcinoma (PDAC), which exhibits extensive glycosylation abnormalities, including increased expression of mucin-associated glycoantigens and altered cell-surface carbohydrate structures. Recent studies have further demonstrated that tumor-associated glycan recognition can be integrated into CAR-T design to overcome the physical and biological barriers associated with solid tumors. Park et al. developed a glycan-targeting CAR-T approach using a high-affinity Tn-MUC1-binding domain to enhance recognition of pancreatic cancer cells with dense surface glycocalyx structures. This strategy improved CAR-T engagement with pancreatic tumor cells and enhanced antitumor activity in preclinical models, suggesting that glycan-directed CAR-T engineering may represent a promising approach for PDAC immunotherapy [51].

Despite these advances, glycoform-directed CAR-T therapy remains largely at the preclinical stage. Major challenges include the structural complexity and heterogeneity of tumor glycan profiles, the difficulty of generating highly specific glycoform-recognizing antibodies, and potential variability of glycan expression among different tumor subtypes. Moreover, antigen glycosylation itself may influence CAR recognition and T-cell activation, indicating that glycan composition should be considered an important factor in CAR-T antigen selection and optimization [52]. Future studies integrating glycoform-specific recognition with dual-target CAR systems, logic-gated CAR circuits, and TME-remodeling strategies may further enhance the precision and durability of CAR-T therapy in pancreatic cancer.

Cytokine-armored CAR-T strategies: IL-10 engineering

Interleukin-10 (IL-10) is an anti-inflammatory cytokine mainly secreted by T cells, B cells, macrophages, and other immune cells. One mechanism of immune escape in pancreatic cancer is associated with increased IL-10 levels in the TME, where IL-10 suppresses immune responses and facilitates tumor immune evasion. Zhao Y et al. engineered CAR-T cells to express IL-10, using IL-10 secretion to counteract T cell dysfunction within the TME and enhance CAR-T proliferation and effector function. IL-10 improved mitochondrial function and oxidative phosphorylation, thereby enhancing the metabolic fitness of CAR-T cells. IL-10-expressing CAR-T cells showed significant efficacy across multiple solid tumor models, including pancreatic cancer, mediated durable clearance of tumors and metastases, and induced stem-like memory responses that provided long-term immune protection. In this context, IL-10 should be regarded as an engineering strategy rather than a tumor antigen [53].

GAS6-based CAR-T cells targeting tumor cells and immunosuppressive myeloid populations

Growth arrest-specific protein 6 (GAS6) is a natural ligand of the TYRO3, AXL, and MERTK (TAM) receptor family. These receptors are highly expressed in multiple cancer cells and tumor-associated macrophages (TAMs) and contribute to tumor growth, dissemination, drug resistance, and immune escape. GAS6-based CAR-T cells exhibit potent antitumor activity against pancreatic cancer showed that GAS6-CAR-T cells significantly inhibited tumor growth in pancreatic cancer mouse models, with sustained tumor clearance lasting 42 days. Tumor volume was markedly reduced and T cell infiltration increased in the CAR-T-treated group, indicating effective targeting and tumor elimination. In pancreatic cancer PDX models, GAS6-CAR-T cells also showed favorable efficacy, eliminating tumor cells and reducing AXL-positive cells among TAMs, thereby further supporting the multi-target therapeutic potential of GAS6-CAR-T cells [54].

GPC1-targeted CAR-T cells for pancreatic cancer

Glypican-1 (GPC1) is a GPI-anchored cell-surface heparan sulfate proteoglycan that is highly expressed in pancreatic cancer. The IgG4 hinge with CD28 transmembrane domain improves V(H)H-based CAR T cells targeting a membrane-distal epitope of GPC1 in pancreatic cancer used hybridoma technology and phage display to isolate the monoclonal antibody HM2, which recognizes a membrane-proximal GPC1 epitope, and the camelid nanobody D4, which recognizes a membrane-distal epitope. D4-based CAR-T cells specifically killed pancreatic cancer cell lines with both high and low GPC1 expression, including T3M4, BxPC-3, and Panc-1. Replacing the CD8 hinge with a shorter IgG4 hinge and combining it with a CD28 transmembrane domain generated the optimized D4-IgG4H-CD28TM construct, which significantly enhanced recognition and clearance of low-antigen-density tumor cells while promoting cytokine secretion. In multiple immunodeficient mouse models, including peritoneal dissemination and orthotopic pancreatic cancer models, optimized D4-IgG4H-CD28TM CAR-T cells showed potent antitumor activity, induced tumor regression, and prolonged survival. Mechanistically, disulfide bond-mediated CAR dimerization through the IgG4 hinge enhanced T cell signaling, including PLCγ/NF-κB pathway activation, and promoted multifunctional and memory-like CAR-T phenotypes. Upregulation of genes such as HMGB1 and ID1 further supported the durable antitumor potential of this approach. The study also suggested that GPC1 participates in Wnt signaling in pancreatic cancer and that CAR-T therapy can downregulate this pathway [55].

Alternative immune-cell platforms: allogeneic CAR-NKT cells

Beyond conventional CAR-T cells, allogeneic CAR-NKT cells have emerged as an alternative immune-cell platform with potential advantages for solid tumor therapy. Unlike autologous CAR-T products generated from individual patients, allogeneic CAR-NKT cells can be manufactured from healthy donor-derived hematopoietic stem and progenitor (HSP) cells. Recent studies established an off-the-shelf CAR-NKT manufacturing platform by introducing invariant NKT-cell receptor genes together with tumor-targeting CAR constructs into CD34⁺ HSP cells using lentiviral vectors, followed by sequential ex vivo expansion, NKT-cell differentiation, maturation, and large-scale expansion. This strategy enables the generation of highly pure CAR-NKT products suitable for cryopreservation and scalable production. Furthermore, incorporation of additional immune-modulatory genes, such as IL-15, may enhance persistence and antitumor activity, providing a potential solution to the manufacturing complexity and limited accessibility of conventional autologous CAR-T therapies [56, 57].

Overall, current evidence indicates that MSLN and CLDN18.2 are the most extensively investigated targets in pancreatic cancer CAR-T research. CLDN18.2 has demonstrated encouraging preliminary clinical activity; however, its therapeutic potential remains to be validated through larger-scale clinical studies with longer follow-up and controlled trial designs. PSCA, CD155, CD276, GAS6, and GPC1 remain primarily supported by preclinical studies, whereas IL-10 represents an engineering enhancement strategy rather than a tumor antigen. The evidence landscape across these targets and strategies is summarized in Fig. 3; Table 1. Detailed information on CAR-T targets and preclinical evidence is provided in Supplementary Table 1.

Fig. 3.

Evidence matrix of CAR-T targets and strategies in pancreatic cancer. Matrix representation of preclinical (blue) and clinical (orange) evidence for CAR-T targets across study models (in vitro, animal, PDX/orthotopic, case report, early clinical study). Columns represent tumor-associated surface antigens (MSLN, CLDN18.2, PSCA, CD155, CD276, GPC1, GAS6) and an engineered strategy (IL-10). Blank cells indicate no clear evidence. Insets highlight CAR-T antigen engagement, TME microenvironmental adjustments, and IL-10 engineering strategies

Table 1.

Clinical evidence of CAR-T therapy in pancreatic cancer

Target	CAR-T product/strategy	Study type	Patients	Response	Major toxicity	Reference
CLDN18.2	CLDN18.2 CAR-T with tEGFR safety switch	Early clinical study	3 advanced pancreatic cancer patients	1 CR,2 PR	Gastric mucosal injury / OTOT	Zhong et al.,2025 [32]
CLDN18.2	CT041	Case series / early clinical report	2 metastatic pancreatic cancer patients	PR; pulmonary metastasis regression	Mild–moderate CRS	Qi et al.,2023 [34]
CLDN18.2	CLDN18.2 CAR-T	Case report	1 advanced PDAC patient	CR for 8 months, then CLDN18.2-negative relapse	Grade 2 CRS, gastric mucosal injury	Zhong et al.,2024 [35]
EGFR	EGFR CAR-T	Phase I clinical trial	Metastatic pancreatic cancer	Limited efficacy / safety data	Mucosal and skin toxicity, pleural effusion, pulmonary exudation	Liu et al.,2020 [82]
MSLN	Anti-MSLN CAR-T	Phase I / early clinical study	Advanced PDAC	Favorable safety but limited efficacy	T-cell exhaustion	Aznar et al.,2025 [79]

Nanobody-based CAR-T cells: an emerging antigen-recognition platform

Nanobody-based CAR-T cells represent an emerging strategy to optimize antigen recognition by replacing conventional antibody-derived single-chain variable fragments (scFvs) with camelid-derived single-domain antibodies (VHHs). Compared with conventional scFvs, nanobodies possess smaller molecular size, high structural stability, improved tissue penetration, and flexible epitope recognition characteristics, making them attractive components for next-generation CAR designs. These advantages may facilitate the recognition of membrane-proximal epitopes and improve CAR-T functionality in tumors with complex antigen accessibility [58].

In hematological malignancies, nanobody-based CAR-T platforms have demonstrated promising therapeutic potential, particularly against CD19- and BCMA-expressing tumors. CD19-targeted nanobody CAR-T cells have shown efficient activation, proliferation, and cytotoxicity against B-cell lymphoma models, supporting the feasibility of replacing conventional scFv-based antigen-recognition domains with VHH structures [59]. Furthermore, humanized CD19-specific nanobody CAR-T cells have been developed to reduce potential immunogenicity while maintaining antigen binding capacity, cytokine secretion, and antitumor activity, providing additional support for clinical translation of VHH-based CAR-T approaches [60].

BCMA-directed nanobody CAR-T cells have also demonstrated encouraging activity in multiple myeloma. Han et al. developed single VHH-directed BCMA CAR-T cells and reported potent antitumor activity with induction of remission in patients with relapsed/refractory multiple myeloma, highlighting the potential of nanobody-based recognition domains for plasma cell malignancies [61]. These studies indicate that nanobody-based CAR-T platforms may overcome some limitations associated with conventional scFv-based CAR structures, including molecular complexity and potential immunogenicity.

Beyond hematological malignancies, nanobody-based CAR-T strategies have increasingly been explored in solid tumors. In pancreatic cancer, Du et al. developed a nanobody-based CLDN18.2 CAR-T product (IMC002) with reduced on-target/off-tumor toxicity. Compared with conventional scFv-based CLDN18.2 CAR-T cells, IMC002 demonstrated favorable tumor accumulation, delayed distribution in normal gastric tissues, and potent antitumor activity in pancreatic and gastric cancer models, suggesting that nanobody-based CAR structures may provide advantages in balancing efficacy and safety for solid tumor immunotherapy [37].

Despite these advances, several challenges remain, including optimization of VHH affinity, potential immunogenicity of non-human nanobody sequences, limited clinical validation in solid tumors, and the need to integrate nanobody recognition domains with advanced CAR architectures. Future studies combining nanobody-based CARs with dual-target strategies, armored CAR designs, and logic-gated circuits may further enhance the specificity and durability of CAR-T therapy for pancreatic cancer.

Limitations of CAR-T Cell therapy in pancreatic cancer

The immunosuppressive tumor microenvironment of pancreatic cancer

The interaction between tumor cells and the TME is a key driver of cancer initiation and progression. Pancreatic cancer is characterized by a highly suppressive TME, marked by dense physical barriers and extensive infiltration of immunosuppressive cells [62]. These barriers collectively restrict CAR-T cell trafficking, infiltration, activation, expansion, persistence, and cytotoxic function (Fig. 4).

First, CAFs and stromal components constitute a major obstacle. CAFs are one of the predominant stromal cell populations in pancreatic cancer, accounting for 15–85% of stromal cells. By producing extracellular matrix (ECM) proteins, they contribute to the formation of a dense physical barrier [63]. CAFs can be divided into three major subsets. The first subset is located near cancer cells and displays myofibroblastic and potentially antitumor characteristics. The second subset is activated by IL-1 and contributes to immunosuppression. The third subset, termed antigen-presenting CAFs, remains insufficiently characterized and may also exert immunosuppressive functions [64]. In pancreatic cancer, pancreatic stellate cells (PSCs) are activated by transforming growth factor-β (TGF-β) and platelet-derived growth factor (PDGF), express α-smooth muscle actin (α-SMA), and become an important source of CAFs [65]. Tissue-resident fibroblasts can also be reprogrammed through interactions with growth factors, chemokines, and cancer cell-derived metabolites. Through ECM remodeling, paracrine signaling, and immune suppression, CAFs support tumor growth and metastasis and suppress antitumor immunity by directly inhibiting CD8-positive T cells or recruiting MDSCs [66].

Second, pancreatic cancer contains a complex network of immunosuppressive cells. The PDAC TME is among the most immunosuppressive of all solid tumors and is characterized by extensive infiltration and cooperative interactions among Tregs, MDSCs, and TAMs [67]. Tregs accumulate in the pancreatic cancer TME and directly suppress CD8-positive cytotoxic T lymphocytes through secretion of IL-10 and TGF-β, while also impairing effector T cell activation and proliferation through contact-dependent mechanisms. MDSCs suppress immunity through multiple pathways. They highly express arginase-1 and inducible nitric oxide synthase, leading to arginine depletion and production of reactive oxygen species and nitric oxide, thereby impairing T cell function and survival. MDSCs can also recruit and expand Tregs, forming a positive feedback loop of immune suppression [68]. TAMs in pancreatic cancer are predominantly M2-polarized. They maintain an immunosuppressive environment by secreting IL-10 and TGF-β and upregulate lipid transport molecules such as CD36 and fatty acid-binding proteins, enhancing fatty acid uptake and utilization. This metabolic adaptation allows TAMs to maintain suppressive activity in nutrient-deprived TMEs [69]. These suppressive cell populations do not operate independently but instead form a coordinated immunosuppressive network through cytokine and chemokine signaling. Even when a small number of CAR-T cells successfully cross the dense stromal barrier and infiltrate tumors, local immunosuppressive signals can rapidly induce functional exhaustion, characterized by expression of immune checkpoint molecules such as programmed cell death protein 1 (PD-1), T cell immunoglobulin and mucin-domain containing-3 (TIM-3), lymphocyte activation gene 3 (LAG-3), cytotoxic T lymphocyte-associated protein 4 (CTLA-4), and T cell immunoreceptor with Ig and ITIM domains (TIGIT). Pancreatic cancer cells further reinforce this suppressive network by secreting IL-10, TGF-β, IL-23, CXCL1-3, CXCL5, CXCL12, CCL2, and vascular endothelial growth factor (VEGF), thereby activating stromal cells and recruiting additional immunosuppressive cells [70]. Therefore, the immunosuppressive network dominated by Tregs, MDSCs, and TAMs represents a central barrier to durable CAR-T efficacy in pancreatic cancer.

Third, pancreatic cancer is characterized by low immunogenicity and defects in antigen presentation. PDAC has a tumor mutational burden (TMB) substantially lower than many other solid tumors, resulting in a limited number of tumor-specific neoantigens and weakening immune recognition [71]. However, because CAR-T cells recognize surface antigens independently of major histocompatibility complex (MHC) presentation, MHC-I defects should be interpreted primarily as impairing endogenous CD8-positive T cell immunity and limiting potential synergy between CAR-T cells and endogenous antitumor responses, rather than directly preventing CAR-T recognition. MHC-I downregulation is common in PDAC and is closely associated with resistance to immunotherapy. Although mutations in genes encoding MHC-I molecules or antigen-presentation machinery are relatively rare in PDAC compared with other cancers, MHC-I loss is prominent, suggesting non-mutational regulatory mechanisms [72]. Canel et al. [73]. showed that focal adhesion kinase (FAK) suppresses immunoproteasome subunits such as Psmb8 and MHC-I-mediated antigen presentation through a kinase-independent nuclear translocation mechanism, thereby weakening CD8-positive T cell recognition and killing. Inhibition of FAK signaling increased immunoproteasome expression, enhanced MHC-I-mediated antigen presentation, improved peptide-MHC-I binding affinity, and promoted infiltration of tumor-reactive CD8-positive T cells. In addition, pancreatic cancer cells can selectively degrade MHC-I molecules through autophagy, further reducing surface antigen presentation and enabling immune evasion. Post-translational modifications (PTMs), including aberrant glycosylation, O-GlcNAcylation, phosphorylation, and citrullination, can also reshape the pancreatic cancer immunopeptidome and promote immune escape through reversible enzyme-driven processes [74].

Fourth, tumor-intrinsic immune escape mechanisms further limit CAR-T efficacy. Antigen escape is one of the most direct mechanisms by which pancreatic cancer resists CAR-T attack. CAR-T cytotoxicity depends on recognition of specific surface target antigens; however, pancreatic cancer exhibits marked antigen heterogeneity. Under selective pressure against a single antigen, tumor cells may downregulate or completely lose the target antigen, resulting in clonal escape [75]. Genome-wide CRISPR-Cas9 screening has shown that disruption of genes involved in GPI-anchor biosynthesis can cause loss of GPI anchoring of MSLN, preventing CAR-T recognition and killing. The same screening also identified antigen-independent resistance pathways, suggesting that tumor cells can alter death receptor-related pathways and reduce CAR-T-mediated cytotoxicity [21]. Pancreatic cancer cells also broadly express Fas ligand (FasL), whereas activated T cells express its receptor Fas. When CAR-T cells infiltrate tumor sites, tumor-derived FasL can bind Fas on CAR-T cells and induce apoptosis, thereby promoting immune escape. This mechanism affects not only CAR-T cells but also endogenous bystander T cells, further amplifying immunosuppression [76].

Finally, persistent antigen stimulation and metabolic stress in the pancreatic cancer TME readily drive CAR-T cells into functional exhaustion. Long-term exposure to high antigen burden and immunosuppressive signals induces progressive upregulation and coexpression of checkpoint molecules such as PD-1, LAG-3, TIM-3, and CTLA-4, leading to impaired proliferation, reduced effector cytokine secretion, and loss of cytotoxic activity. Sun et al. [77]. showed that dual-target CAR-T strategies can improve pancreatic cancer immunotherapy by enhancing effector function and suppressing T cell exhaustion, supporting exhaustion as a key factor limiting CAR-T efficacy. Metabolic factors further aggravate exhaustion. Indoleamine 2,3-dioxygenase 1 (IDO1), a tryptophan-metabolizing enzyme, is markedly upregulated in the pancreatic cancer TME. IDO1 catalyzes tryptophan depletion and kynurenine accumulation, causing effector T cell proliferative arrest and promoting Treg expansion. IDO1 deletion in CAR-T cells has been reported to prolong survival and improve tumor control in murine pancreatic cancer models, suggesting that IDO1-mediated metabolic immune escape is an important mechanism weakening CAR-T therapy [78].

Short-term persistence and limited expansion of CAR-T cells

Short-term persistence and limited expansion of CAR-T cells in patients with pancreatic cancer represent major bottlenecks limiting clinical efficacy. In contrast to hematological malignancies, where CAR-T cells can expand by several orders of magnitude and maintain durable responses for years, CAR-T cells in pancreatic cancer often enter functional exhaustion shortly after infusion and lose sustained antitumor activity. A phase I clinical trial reported by Aznar et al. in 2025 in Cell Reports Medicine showed that anti-MSLN CAR-T cells administered intravenously and locally to patients with advanced PDAC had a favorable safety profile but limited efficacy. Single-cell genomic analysis revealed prominent exhaustion signatures after infusion, including upregulation of the transcription factors ID3 and SOX4 and enrichment of a GZMK-positive phenotype. Dual knockout of ID3 and SOX4 in CAR-T cells prolonged relapse-free survival in xenograft models, demonstrating the feasibility of enhancing CAR-T persistence by targeting exhaustion-associated transcription factors [79].

At the molecular level, the NR4A transcription factor family has been identified as a key regulator of CAR-T exhaustion and mitochondrial dysfunction. Nakagawara et al. showed that simultaneous CRISPR-Cas9-mediated knockout of NR4A1, NR4A2, and NR4A3, referred to as NR4A triple knockout (TKO), in human HER2-targeted CAR-T cells enabled resistance to exhaustion under repeated antigen stimulation and preserved stronger tumor-killing activity in vitro and in vivo. Mechanistically, NR4A TKO CAR-T cells exhibited enhanced mitochondrial oxidative phosphorylation and regulated mitochondrial gene expression to maintain persistence and stemness, thereby supporting prolonged intratumoral activity [80].

Off-target toxicity and safety concerns

Off-target toxicity and safety concerns represent another major obstacle for applying CAR-T therapy to pancreatic cancer and other solid tumors. CAR-T cytotoxicity depends on recognition of specific surface antigens on tumor cells. However, most antigens currently targeted in pancreatic cancer CAR-T therapy, including MSLN, carcinoembryonic antigen (CEA), epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2 (HER2), and PSCA, are TAAs rather than tumor-specific antigens (TSAs). Because these antigens can also be physiologically expressed in normal tissues at varying levels, OTOT toxicity constitutes a major clinical risk [81].

Clinical studies have provided evidence of such toxicity. In a phase I clinical trial of EGFR-targeted CAR-T cells for metastatic pancreatic cancer, Liu et al. reported reversible OTOT toxicities after CAR T-EGFR infusion, including grade 3 or higher mucosal and skin toxicities, pleural effusion, and pulmonary interstitial exudation [82]. CLDN18.2-targeted CAR-T therapy has similarly revealed OTOT toxicity. Zhong et al. reported that CLDN18.2 CAR-T cells induced substantial antitumor activity in patients with advanced pancreatic cancer, including one CR and two PRs among three patients, but gastric mucosal injury occurred due to CLDN18.2 expression in normal gastric mucosa. Cyclophosphamide and cetuximab were used to eliminate CAR-T cells and successfully control this toxicity, demonstrating the feasibility of antibody-dependent safety-switch strategies for CAR-T toxicity management [32].

In addition to OTOT toxicity, CAR-T therapy can induce systemic immune-related adverse events. CRS is one of the most common and potentially life-threatening systemic toxicities associated with CAR-T therapy. It arises from massive CAR-T activation after tumor antigen recognition and subsequent explosive release of proinflammatory cytokines such as IL-6, IFN-γ, and TNF-α. A pooled analysis by Qi et al. of CLDN18.2-targeted CAR-T therapy (CT041) in patients with refractory metastatic pancreatic cancer showed that CRS and gastrointestinal dysfunction were the most commonly reported grade 1–2 adverse events, indicating that systemic toxicities such as CRS should not be overlooked in pancreatic cancer CAR-T therapy [83].

Beyond CRS and on-target/off-tumor toxicity, several additional challenges may influence the clinical translation of CAR-T therapy for pancreatic cancer. Immune effector cell-associated neurotoxicity syndrome (ICANS), although more frequently reported in hematological malignancies, remains an important safety consideration as CAR-T platforms continue to expand into solid tumors. Long-term persistence of CAR-T cells is another critical determinant of therapeutic durability, as insufficient persistence may lead to tumor relapse, whereas uncontrolled persistence may increase the risk of prolonged toxicity. In addition, viral vector-based CAR-T manufacturing raises concerns regarding genomic integration and potential insertional mutagenesis, highlighting the need for safer gene-delivery technologies.

From a translational and commercial perspective, manufacturing complexity, batch-to-batch variability, and limited product standardization remain major obstacles. Autologous CAR-T production requires individualized manufacturing processes, resulting in high costs, prolonged production times, and restricted accessibility. Emerging strategies, including allogeneic CAR-T platforms, non-viral gene delivery systems, and standardized manufacturing pipelines, may help improve scalability and reduce production barriers. Furthermore, regulatory frameworks for advanced cellular therapies continue to evolve, requiring rigorous evaluation of product quality, long-term safety, and clinical efficacy before widespread implementation (Table 2).

Table 2.

Major barriers and corresponding strategies for CAR-T in pancreatic cancer

Barrier	Mechanism	Potential Strategy
Dense stroma / CAFs	Impaired CAR-T infiltration	FAP-targeted CAR-T, collagenase nanogel backpack, local delivery
Immunosuppressive TME	Tregs, MDSCs, TAMs, IL-10,TGF-β	RIG-I agonists, armored CAR-T, checkpoint blockade
Antigen heterogeneity	Antigen loss / clonal escape	Dual-target CAR-T, logic-gated CAR
Poor persistence	Exhaustion, metabolic stress	IL-7,IL-10,NR4A/ID3/SOX4 modulation
OTOT toxicity	TAA expression in normal tissues	Safety switch, synNotch CAR, local administration

Strategies to overcome current limitations

Given the unsatisfactory efficacy of CAR-T therapy in pancreatic cancer, a single therapeutic strategy is unlikely to be sufficient. Based on the limitations discussed above, effective treatment may require an integrated approach combining TME remodeling, CAR-T cell engineering, and optimized delivery strategies. These major therapeutic directions are summarized in Fig. 5.

Fig. 5.

Optimization strategies for CAR-T cell therapy in pancreatic cancer. Strategies to enhance CAR-T efficacy include: (1) target optimization via multi-target CAR-T, juxtamembrane epitope targeting, and antigen-density enhancement; (2) enhanced tumor infiltration using CXCR4 modification, local delivery, and radiotherapy-assisted trafficking; (3) remodeling the tumor microenvironment through modulation of CAFs, TAMs, MDSCs, and oncolytic virus therapy; (4) improving CAR-T persistence and metabolic fitness (Armored CAR-T, IL-7, IL-10); (5) controlling toxicity using safety switches, drug-inducible CAR-T, local administration, and dual-antigen recognition

Remodeling the tumor microenvironment

The PDAC TME is characterized by dense desmoplastic stroma, profound immune suppression, and metabolic abnormalities, all of which limit CAR-T cell infiltration and effector function. Therefore, remodeling the PDAC TME from an immunologically “cold” tumor into a “hot” tumor permissive to CAR-T activity represents a promising but clinically challenging strategy for improving CAR-T therapy.

At the level of CAR engineering, dual-function CAR-T designs that combine tumor antigen targeting with stromal remodeling have achieved important progress. Wehrli et al. reported mesoFAP CAR-TEAM cells, which express an anti-MSLN CAR and secrete T cell-engaging molecules targeting FAP on CAFs and CD3 on T cells. This design allows simultaneous killing of tumor cells and depletion of CAFs that contribute to immune barriers. In patient-derived organoid and matched CAF coculture models, patient-derived organotypic tumor spheroids (PDOTS), and primary and liver metastatic pancreatic cancer mouse models, mesoFAP CAR-TEAM cells were significantly more effective than single-target CAR-T cells in eliminating both PDAC cells and CAFs [30].

Xiao et al. [84]. further demonstrated mechanistically that depletion of FAP-positive CAFs using FAP-targeted CAR-T cells disrupted the integrity of dense desmoplastic stroma and rendered previously treatment-resistant tumors responsive to MSLN-targeted CAR-T cells and anti-PD-1 antibody therapy. The mechanisms included overcoming stroma-dependent restriction of T cell extravasation and perivascular invasion, reversing immune exclusion, reducing myeloid cell accumulation, and increasing infiltration of endogenous CD8-positive T cells and natural killer (NK) cells.

Innate immune signaling pathways can also be reactivated within the TME to overcome pancreatic cancer resistance to CAR-T therapy. Senz et al. demonstrated the potential of retinoic acid-inducible gene I (RIG-I) agonists, specifically synthetic 5′-triphosphate RNA (3p-RNA), in combination with CAR-T cells. 3p-RNA rapidly induced type I interferon responses and secretion of chemokines such as CCL5 and CXCL9/10/11, generating chemokine gradients that recruited CCR5-positive and CXCR3-positive CAR-T cells into tumors. Combination therapy achieved complete tumor clearance in 60–70% of tumor-bearing mice, whereas either monotherapy was largely ineffective. Moreover, cured animals rejected rechallenge with CAR antigen-negative tumor cells, indicating that RIG-I activation induced endogenous T cell-mediated antigen spreading. These findings suggest that intratumoral RIG-I agonists can reprogram immunotherapy-resistant pancreatic cancer TMEs into CAR-T-permissive environments [85].

Targeting the glycocalyx barrier has emerged as another TME-remodeling strategy. The highly expanded glycocalyx on pancreatic cancer cells, especially overexpressed Tn-MUC1 glycoproteins, forms a dense physical barrier on the tumor cell surface and prevents effective contact between CAR-T cells and target antigens. Park et al. integrated a non-signaling, high-affinity Tn-MUC1-binding domain, termed HPA-bridge, into CAR-T cells. This design enables CAR-T cells to anchor to the glycocalyx before target antigen recognition, thereby overcoming the physical obstruction imposed by glycoproteins and enhancing immune synapse formation between CAR-T cells and tumor cells. HPA-bridge-equipped CAR-T cells showed improved activity in pancreatic cancer models [51].

Alternative immune-cell carrier platforms provide another dimension of TME remodeling beyond conventional αβ T cells. Li et al. developed an allogeneic off-the-shelf CAR-invariant NKT (CAR-iNKT) platform, Allo15MCAR-NKT cells, generated by engineering human hematopoietic stem and progenitor cells to express both an MSLN-targeted CAR and IL-15. Compared with conventional CAR-T cells, Allo15MCAR-NKT cells can target CD1d-positive tumor cells and simultaneously eliminate or reprogram suppressive myeloid populations, thereby reshaping the TME in a manner that supports durable antitumor immunity [86].

Engineering CAR-T cells

To overcome the severe immunosuppressive barriers and antigen heterogeneity of pancreatic cancer, CAR-T engineering has become a central research direction. Current advances mainly involve armored CAR-T cells, dual-targeting CAR-T cells, and logic-gated systems.

Armored CAR technology enables CAR-T cells to secrete immunomodulatory cytokines or express engineered receptors, thereby actively remodeling the local TME and enhancing intrinsic CAR-T function. Recent reviews have summarized strategies to overcome solid tumor microenvironmental barriers, including engineering CAR-T cells to express dominant-negative receptors or cytokine-release constructs such as interleukin-12 (IL-12) and interleukin-18 (IL-18), as well as introducing chemokine receptor knock-ins to improve tumor infiltration. These strategies can enhance CAR-T function in solid tumors at both molecular and cellular levels [87].

Dual-target CAR strategies represent a key engineering approach to address antigen heterogeneity and antigen escape in pancreatic cancer. CAR-T cells designed to recognize two distinct TAAs, such as MSLN and CLDN18.2, may more effectively cover heterogeneous tumor cell populations and reduce immune escape caused by loss of a single antigen. Combination approaches involving CAR-T therapy, immune checkpoint inhibitors, and cytokine modulation are also being explored to improve CAR-T activity in pancreatic cancer [75].

At the level of safety and precise control, logic-gated CAR systems represent an important frontier in CAR engineering. Shirzadian et al. reviewed the integration of synthetic Notch (synNotch) receptors with CAR-T therapy. SynNotch receptors act as molecular logic gates and enable precise multi-antigen-controlled T cell activation by regulating gene expression in a spatiotemporal manner. This allows engineered T cells to distinguish cancer cells from normal cells with improved specificity and controllability. AND-gated, OR-gated, and NOT-gated logic systems may substantially reduce OTOT toxicity caused by single-antigen expression in normal tissues and provide an important safety framework for pancreatic cancer CAR-T therapy [88].

Optimization of therapeutic delivery strategies

The dense fibrotic stroma of pancreatic cancer not only establishes an immunosuppressive barrier but also physically restricts the intratumoral infiltration of chimeric antigen receptor T (CAR-T) cells. Even when CAR-T cells are functionally competent in vitro, abnormal vascular architecture and stromal density often prevent them from efficiently reaching the tumor core in vivo. Therefore, optimization of therapeutic delivery has emerged as a critical strategy. Current approaches mainly include nanomaterial-assisted delivery, in situ CAR-T generation through direct in vivo immune-cell engineering in situ CAR-T generation, oncolytic virus-mediated tumor retargeting, engineered immune-cell carrier platforms, and locoregional administration (Fig. 5).

In terms of nanomaterial-assisted delivery, Zhao et al. [89]. developed a collagenase nanogel backpack system. The nanogel was prepared by crosslinking collagenase with oxidized sodium alginate and was further modified with the CXCR4 antagonist peptide DV1. Through DV1–CXCR4 receptor–ligand interactions, the nanogel specifically attached to the CAR-T cell surface, forming a cell-backpack delivery unit. Once CAR-T cells reached the tumor stroma, the nanogel backpack gradually released collagenase to degrade collagen barriers within the extracellular matrix (ECM). Meanwhile, DV1 blocked the CXCL12/CXCR4 axis, relieving chemokine-mediated spatial restriction and enabling CAR-T cells to overcome both physical and chemotactic barriers to penetrate deeper into the tumor core.

In situ CAR-T generation represents a shift from ex vivo manufacturing to in vivo engineering, bypassing the complex procedures of T-cell collection, ex vivo genetic modification, expansion, and reinfusion required for conventional CAR-T therapy. At the 2025 American Association for Cancer Research (AACR) Annual Meeting, Bajbouj et al. reported a targeted lipid nanoparticle (tLNP)-based strategy for in vivo generation of fibroblast activation protein (FAP)-directed CAR-T cells. The team developed an anti-CD5-conjugated tLNP platform to deliver modified messenger RNA (mRNA) encoding a FAP-targeted CAR. FAP is highly expressed on cancer-associated fibroblasts (CAFs) in pancreatic cancer, which are key cellular components of the dense immunosuppressive stromal barrier. After a single intravenous injection of 30 µg FAP-CAR mRNA tLNPs, transient FAP-CAR expression was detected in more than 45% of splenic CD3-positive T cells, more than 69% of peripheral blood CD3-positive T cells, and more than 35% of tumor-infiltrating CD3-positive T cells. Compared with conventional ex vivo retroviral transduction followed by adoptive transfer, this in situ mRNA CAR-T engineering strategy achieved superior tumor growth inhibition (74% vs. 48%), without the need for lymphodepleting preconditioning, while also avoiding the potential risks associated with viral vector integration [90].

Beyond pancreatic cancer-specific FAP-CAR engineering, in vivo CAR engineering has rapidly evolved into a broader therapeutic platform aimed at overcoming the manufacturing limitations of conventional CAR-T therapies. Current approaches include targeted lipid nanoparticle (LNP)-based delivery of CAR-encoding mRNA or DNA, polymeric nanoparticles, lentiviral and adeno-associated virus (AAV)-based delivery systems, as well as emerging bioinstructive scaffolds and biomimetic delivery vehicles. These platforms directly introduce CAR genetic materials into endogenous immune cells, enabling the generation of CAR-engineered cells within the patient and potentially eliminating the need for individualized leukapheresis, ex vivo modification, expansion, and reinfusion processes [91].

Compared with conventional ex vivo CAR-T manufacturing, in vivo CAR engineering provides several potential advantages, including simplified production workflows, reduced manufacturing costs, shortened treatment timelines, and improved accessibility through off-the-shelf therapeutic formats. In addition, non-integrating approaches such as LNP-mRNA delivery may provide transient CAR expression, allowing improved control over therapeutic exposure and potentially reducing long-term safety concerns associated with permanent genomic modification [91].

However, several challenges remain before widespread clinical translation. These include achieving precise immune-cell targeting, minimizing off-target transduction, optimizing CAR expression kinetics, controlling immunogenicity, and establishing long-term safety profiles. For example, although LNP-based platforms enable efficient and transient CAR expression, repeated administration may be required to maintain therapeutic effects, whereas viral vector-based approaches provide more durable expression but raise concerns regarding insertional mutagenesis and immune recognition. Recent clinical advances using targeted lentiviral vectors and LNP-based systems have demonstrated the feasibility of generating CAR-engineered immune cells directly in vivo, supporting the future development of scalable and accessible CAR-based immunotherapies [92].

Oncolytic virus-mediated tumor retargeting provides a strategy to overcome the lack of ideal tumor-specific antigens in pancreatic cancer. Chen et al. used the engineered oncolytic virus CF33, which has high affinity and specificity for solid tumors, to deliver a non-signaling truncated CD19 antigen (CD19t). This approach labels pancreatic tumor cells that do not naturally express CD19 with a CD19 target, thereby enabling recognition and killing by clinically approved CD19-directed CAR-T cells. In vitro experiments showed that more than 90% of tumor cells expressed CD19t after CF33-CD19t infection. In pancreatic cancer xenograft models, combination therapy induced significant tumor regression, with tumor volumes of 128 ± 14 mm³ in the combination group compared with 485 ± 20 mm³ in the control group (p < 0.0001) [93].

Optimization of engineered immune-cell carriers may also help overcome metastatic dissemination and immune escape in pancreatic cancer. Li et al. developed an allogeneic off-the-shelf CAR-natural killer T (CAR-NKT) cell platform, termed Allo15MCAR-NKT cells. This therapy was generated by engineering human hematopoietic stem and progenitor cells to produce invariant natural killer T cells expressing both a mesothelin-targeted CAR and interleukin-15 (IL-15). Compared with conventional CAR-T cells, Allo15MCAR-NKT cells demonstrated multiple delivery-related advantages. They mediated antitumor activity through dual CAR- and natural killer receptor-dependent cytotoxic mechanisms, achieved superior tumor control in orthotopic and metastatic human pancreatic cancer xenograft models, showed enhanced migration and infiltration into tumor sites, maintained effector and cytotoxic phenotypes, reduced the expression of exhaustion markers, and exhibited favorable safety, with no graft-versus-host disease (GvHD) and only mild cytokine release syndrome (CRS) observed [86].

Locoregional delivery is another promising strategy. Lee et al. investigated the feasibility of regional B7-H3 CAR-T cell delivery combined with liver-directed radiotherapy for pancreatic cancer liver metastases. The study found that 4 Gy radiotherapy significantly upregulated B7-H3 expression on the pancreatic cancer cell line PANC-1. In vivo, liver-directed radiotherapy increased the proportion of B7-H3-positive cells in liver metastases from 72% to 90%. More importantly, when CD8-positive T cells were administered through the portal vein rather than the conventional tail vein route, CD69 expression, a marker of tissue residency, was significantly increased in liver-isolated T cells 24 h later (36% vs. 16%, p < 0.001). These findings suggest that portal vein delivery enhances hepatic T-cell residency and engraftment, while radiotherapy pretreatment further improves the cytotoxic efficacy of B7-H3 CAR-T cells [94].

Discussion and future perspectives

This review systematically summarizes the progress, unique challenges, and innovative therapeutic strategies of CAR-T therapy in pancreatic cancer. Although CAR-T therapy has achieved revolutionary success in hematological malignancies, its application in pancreatic cancer remains at an early exploratory stage, and its therapeutic efficacy remains far below expectations. At present, the dense physical barrier, highly immunosuppressive tumor microenvironment (TME), antigen heterogeneity, and low immunogenicity of pancreatic cancer collectively constitute major obstacles to effective CAR-T cell therapy.

In recent years, preclinical and early clinical studies targeting mesothelin (MSLN), claudin 18.2 (CLDN18.2), prostate stem cell antigen (PSCA), and other targets have demonstrated that CAR-T therapy possesses measurable biological activity in pancreatic cancer, despite limited clinical efficacy. Meanwhile, the field is rapidly moving beyond conventional antigen-targeted CAR-T approaches toward multifunctional platforms integrating improved antigen recognition, immune-cell engineering, microenvironmental remodeling, and precision delivery technologies. These studies have also continued to reveal the mechanisms underlying therapeutic failure. In the future, integrating intrinsic functional enhancement of CAR-T cells, systematic TME remodeling, and innovative delivery strategies may provide more feasible and comprehensive therapeutic approaches for pancreatic cancer.

Despite encouraging preclinical results, the clinical translation of TME-remodeling strategies remains challenging. Most TME-targeting approaches have been evaluated in simplified mouse models, which may not fully recapitulate the spatial heterogeneity, cellular complexity, and dynamic evolution of human pancreatic cancer. The heterogeneity of cancer-associated fibroblasts (CAFs) and the tumor immune microenvironment represents a major challenge for stromal-targeting therapies, as depletion or modulation of specific stromal populations may have context-dependent effects and requires careful consideration of potential impacts on normal fibroblast functions [63, 65]. Although FAP-targeted strategies can reduce stromal barriers and enhance immune infiltration in experimental models, their clinical application requires consideration of FAP expression heterogeneity, tumor-independent fibroblast populations, and the complexity of stromal remodeling in patients [64]. Similarly, cytokine-armored CAR-T cells may enhance local immune activation; however, uncontrolled cytokine activity, insufficient spatial restriction, and limited persistence within solid tumors remain major barriers to clinical application [87]. Furthermore, transient modulation of the TME may not be sufficient to overcome persistent immunosuppressive networks maintained by CAFs, TAMs, MDSCs, and tumor cells, indicating that durable clinical responses will likely require integrated strategies combining TME remodeling with advanced CAR-T engineering and rational combination therapies [67].

Several directions may be critical for advancing CAR-T therapy in pancreatic cancer. First, multi-target strategies should be developed to address antigen heterogeneity and reduce off-target toxicity. Ideally, CAR-T activation should strictly depend on tumor-specific antigen combinations or be initiated only under specific signals within the TME, thereby enabling more precise tumor killing. Shen et al. [95]. emphasized that the development of multi-target CAR constructs for pancreatic ductal adenocarcinoma (PDAC), such as carcinoembryonic antigen (CEA)/MSLN dual-target CAR-T cells, has become an important engineering strategy to overcome antigen heterogeneity and immune escape. In addition, Shirzadian et al. systematically described the integration of synthetic Notch (synNotch) receptors with CAR-T therapy. SynNotch receptors function as molecular logic gates that enable precise multi-antigen regulation of T-cell activation through spatiotemporal control of gene expression, thereby improving the ability to distinguish cancer cells from normal cells [88]. AND-gated, OR-gated, and NOT-gated logic systems may substantially reduce on-target, off-tumor toxicity caused by single-antigen expression in normal tissues and provide an important safety framework for CAR-T therapy in pancreatic cancer.

With the development of next-generation engineered CAR-T cells, fifth-generation and more advanced CAR structures will continue to evolve. Future designs may further enhance the intrinsic ability of CAR-T cells to resist suppressive TME signals while actively remodeling the TME, for example by secreting enzymes that degrade ECM components or expressing chemokines that recruit beneficial immune cells. In parallel, systematic combination regimens should be developed, including CAR-T therapy combined with immune checkpoint inhibitors, targeted therapies such as KRAS G12C inhibitors where applicable, chemotherapy, radiotherapy, tumor vaccines, and agents targeting CAFs or tumor-associated macrophages (TAMs). Moreover, to overcome the low efficiency and high toxicity associated with systemic administration, local delivery approaches such as intra-arterial infusion, intraperitoneal infusion, and endoscopic ultrasound-guided intratumoral injection should be further explored.

To facilitate clinical translation, future CAR-T strategies for pancreatic cancer should be evaluated according to their evidence maturity rather than solely based on preclinical efficacy. Some approaches have already entered clinical practice or early clinical validation, such as lymphodepleting preconditioning, which has become an essential component of adoptive cell therapy by enhancing CAR-T expansion and persistence through immune niche remodeling [43].

In contrast, other strategies remain at an early translational stage. Locoregional CAR-T delivery, including intraperitoneal, intra-arterial, or intratumoral administration, has demonstrated improved tumor exposure and reduced systemic toxicity in preclinical or early clinical studies, but requires further validation in larger patient cohorts. Similarly, oncolytic virus-mediated tumor sensitization represents a promising approach by increasing antigen availability and remodeling the immunosuppressive microenvironment; however, issues including viral delivery efficiency, host immunity, and manufacturing complexity remain unresolved [29, 93].

More speculative approaches, including fully synthetic logic-gated CAR circuits, in situ CAR-T generation, and highly engineered immune-cell platforms, provide attractive solutions to antigen heterogeneity, manufacturing limitations, and safety concerns, but currently lack sufficient clinical evidence to determine their long-term feasibility [88].

Importantly, combination strategies involving immune checkpoint inhibitors should be interpreted cautiously. Although PD-1/PD-L1 blockade has demonstrated substantial activity in several malignancies, clinical trials of immune checkpoint inhibitors in pancreatic cancer have generally produced disappointing results, likely due to the profoundly immunosuppressive and immunologically “cold” nature of pancreatic tumors. Therefore, future CAR-T and checkpoint inhibitor combinations should be guided by mechanistic biomarkers and rational patient selection rather than empirical combination approaches [71].

Overall, the application of CAR-T therapy in pancreatic cancer is gradually moving from theoretical feasibility toward clinical translation. In recent years, dual-target CAR-T cells and logic-gated systems have provided potential solutions to antigen heterogeneity and off-target toxicity, while TME-reprogramming strategies such as retinoic acid-inducible gene I (RIG-I) agonists have begun to show advantages in reshaping the immunosuppressive microenvironment. Meanwhile, the preliminary validation of in situ CAR-T generation marks a conceptual transition from ex vivo manufacturing to in vivo engineering. However, modification of any single strategy is unlikely to fundamentally overcome the multidimensional therapeutic resistance of pancreatic cancer. Future progress may depend on integrating intelligent CAR-T engineering, multidimensional TME remodeling, combination therapies, and precision delivery strategies. As molecular subtyping and immune profiling of pancreatic cancer become increasingly refined, biomarker-based patient stratification and personalized CAR-T treatment regimens may improve the clinical outcomes of CAR-T therapy in pancreatic cancer.

Conclusion

CAR-T therapy has brought new hope to the treatment of pancreatic cancer, but substantial limitations remain. Although many scientific and clinical challenges persist, continued advances in basic research, biotechnology, and clinical trial experience are expected to drive the optimization and innovation of CAR-T therapy. Ultimately, refined CAR-T strategies may play an important role in the multidisciplinary treatment of pancreatic cancer and offer new opportunities to improve patient prognosis.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 ^{(63.5KB, docx)}

Author contributions

Ru-yan Xiao conducted the literature review, created the figures, and drafted the initial manuscript. Ye-fan Jiang contributed to the literature collection, data collation, and manuscript writing. Wei Zhao and Qing-qing Wang participated in reference screening, data interpretation, and figure optimization. You-qing Huang provided critical comments and assisted in manuscript revision. Xin-yu Li and Yu-jie Xiao critically reviewed and edited the manuscript and supervised the writing process. Rui-wu Dai, and Hao Yao conceived the review topic, guided the overall structure and direction of the manuscript, and were responsible for funding acquisition. Hao Yao supervised the study and finalized the manuscript for submission.

Funding

This work was supported by the Sichuan Science and Technology Program [grant numbers 2026NSFSC0667].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.