Introduction: The Promise and the Problem
Chimeric antigen receptor T-cell (CAR-T) therapy has fundamentally transformed outcomes in relapsed or refractory B-cell malignancies and acute lymphoblastic leukemia, achieving durable remissions with multiple FDA-approved products. However, this success has proven difficult to replicate in solid tumors, which account for the overwhelming majority of cancer mortality worldwide. Between 2023 and 2026, a wave of next-generation engineering strategies and early-phase clinical trials has generated genuine proof-of-concept signals in glioblastoma, mesothelioma, gastric cancer, and hepatocellular carcinoma—yet efficacy remains substantially lower than in hematologic disease. Understanding why solid tumors resist CAR-T therapy, and how emerging strategies are systematically addressing each barrier, is essential for oncologists and translational researchers navigating this rapidly evolving field 12.
Why Solid Tumors Are Biologically Distinct from Blood Cancers
The fundamental advantage of hematologic CAR-T therapy lies in its target landscape: CD19 and CD20 are expressed with high density and near-uniform consistency on malignant B cells, making antigen escape rare and immune access relatively unimpeded. Solid tumors, by contrast, exhibit profound intratumoral and intertumoral antigen heterogeneity. Tumor cells within a single lesion display variable target antigen density, and metastatic deposits may differ markedly from the primary. This spatiotemporal heterogeneity creates selective pressure favoring antigen-negative escape variants under CAR-T attack 126.
Beyond antigen biology, the solid tumor microenvironment (TME) is actively hostile to adoptively transferred T cells. Dense extracellular matrix (ECM), cancer-associated fibroblasts (CAFs), and hypoxic niches create physical and metabolic obstacles that hematologic tumors simply do not impose. Collectively, these features demand a fundamentally different engineering approach compared to blood cancers 23.
Table 1. Major Barriers to CAR-T Efficacy in Solid Tumors
| Barrier | Key Mechanism | Clinical Impact | Emerging Solutions |
|---|---|---|---|
| Antigen heterogeneity | Variable antigen density; genetic and epigenetic downregulation; trogocytosis | Antigen-loss relapse; incomplete response | Dual/multi-target CARs; logic-gated designs 1 |
| Lack of tumor-specific antigens | Most TAAs also expressed on normal tissues | On-target/off-tumor toxicity; dose-limiting events | HLA-restricted NOT-gates; AND-gate designs 826 |
| Poor T-cell trafficking | Stromal barriers; insufficient homing signals; FAP-expressing CAFs | Minimal tumor infiltration | Regional delivery; FAP-targeted stromal remodeling 57 |
| Immunosuppressive TME | TGF-β, IL-10, Tregs, MDSCs, TAMs | CAR-T exhaustion; impaired expansion | Armored CAR-T; dnTGFβRII; SMAD7 co-expression 24 |
| Hypoxia/metabolic stress | Glucose depletion; HIF-1α; IDO/kynurenine/AhR axis | Metabolic dysfunction; reduced persistence | Hypoxia-responsive promoters; FOXP3 metabolic reprogramming 320 |
| T-cell exhaustion | Chronic antigen stimulation; PD-1/TIM-3/LAG-3 upregulation | Loss of effector function | Checkpoint-resistant CAR-T; PD-1 nanobody secretion 41 |
| Physical stromal barrier | Dense ECM; desmoplastic stroma; CAF enrichment | Impaired infiltration and migration | FAP-targeting dual-CARs; oncolytic virus TME remodeling 714 |
| Manufacturing/logistics | 3–5 week ex vivo production; cost $300,000–500,000 per patient | Treatment delays; limited access | Allogeneic off-the-shelf; in vivo engineering 2627 |
Current Clinical Experience: Signals Across Solid Tumor Types
Despite formidable barriers, early-phase trials are producing reproducible—if still modest—clinical activity across a range of targets and histologies. The most compelling efficacy signals through mid-2026 emerge from mesothelin, GPC3, HER2, CLDN18.2, and selected PSMA programs 6.
Mesothelin remains among the most clinically mature solid-tumor targets. ATA2271, incorporating cell-intrinsic PD-1 dominant-negative signaling and delivered intrapleurally, achieved an overall response rate (ORR) of 63% (10/16 patients) including three complete responses in malignant pleural mesothelioma when combined with pembrolizumab, with 12-month overall survival of 80% and no CAR-T-related toxicities exceeding grade 2 6. In a separate program, the aPD1-MSLN JL-Lightning CAR-T (anti-PD-1 nanobody-secreting, mesothelin-targeting) achieved 100% ORR at dose level 2 in three mesothelioma patients including one complete response lasting over nine months, though grade 3–4 CRS and immune-mediated pneumonia were observed 626.
GPC3-directed CAR-T in hepatocellular carcinoma (HCC) has produced some of the strongest signals in the dataset. The C-CAR031/AZD-7003 program reported an ORR of 56.5% overall (75% at DL4) with tumor reduction in 91.3% of patients 6. A separate armored GPC3 CAR-T construct (Ori-CAR-001) achieved ORR of 44.4% and disease control rate (DCR) of 77.8% in nine evaluable HCC patients 626.
HER2-targeted constructs show repeated activity signals across gastric, breast, and colorectal cancers. CCT303-406 achieved a deep partial response with 100% reduction in sum of longest diameters in one gastroesophageal junction patient at 36 weeks, with 12-month survival of 83.3% in the highest-dose cohort and no dose-limiting toxicities or ICANS 626.
CLDN18.2 programs are emerging as a credible target in gastrointestinal and pancreatic cancers. The ELIMYN18.2 Phase 1b study in 19 patients (seven gastric/GEJ, 12 pancreatic) demonstrated ORR of 26.3% overall (42.9% in gastric/GEJ), clinical benefit rate of 42.1%, and median OS of 12.8 months from leukapheresis, with no dose-limiting toxicities 29. Crucially, China's NMPA in June 2026 approved satricabtagene autoleucel (satri-cel), an autologous anti-CLDN18.2 CAR-T, for pretreated CLDN18.2-positive, HER2-negative advanced gastric/GEJ adenocarcinoma—representing the world's first approved CAR T-cell therapy for any solid tumor 28.
Glioblastoma has seen intensive investigation with locoregional delivery. A systematic review of 13 Phase I trials (128 patients total) across six molecular targets—predominantly EGFR family and IL13Rα2—found overall survival from infusion of 2.9 to 14.5 months, with CAR-T cells typically becoming undetectable within one month and neurological adverse events (headache, seizures) being the most common serious safety signals 10. Direct CNS delivery (intratumoral, intraventricular) markedly outperformed intravenous administration in CAR-T localization; one study found little to no CAR-T in brain in 6/7 intravenous-only patients 10. The CART-EGFR-IL13Rα2 dual-target approach achieved tumor shrinkage in 85% of evaluable rGBM patients with a median 35% target lesion reduction by modified RANO criteria, though grade 3 ICANS occurred in 56% 26.
B7-H3 represents a particularly broad platform target. The BrainChild-03 trial (NCT04185038) demonstrated feasibility of repeatedly dosed intraventricular B7-H3 CAR-T in pediatric DIPG, with one patient showing sustained improvement through 12 months and CAR-T cells detectable in CSF through course 9—with no dose-limiting toxicities across 40 infusions 9. Systemically, the STRIvE-02 program (NCT04483778) at Seattle Children's represents the first-in-human systemic B7-H3 CAR-T study in pediatric non-CNS solid tumors, incorporating bispecific (B7H3×CD19) and pembrolizumab combination arms 6. B7-H3 CAR-armored EBVSTs also show in vitro and in vivo activity across multiple tumor types while potentially mitigating MDSC-driven suppression, since B7-H3 is upregulated on MDSCs 19.
Safety remains heterogeneous across programs. Fatal ICANS was observed in the TmPSMA-02 prostate cancer CAR-T study, which was terminated after two patient deaths 6. The PSMA-directed P-PSMA-101 program, by contrast, showed PSA decline in 71% and ≥50% PSA reduction in 36% of evaluable mCRPC patients, with ICANS described as manageable 6. This divergence underscores that "solid tumor CAR-T safety" cannot be treated as a uniform category.
Table 2. Representative Targets, Clinical Programs, and Key Evidence
| Target | Tumor Type | Key Program | ORR/DCR Signal | Notable Safety Event |
|---|---|---|---|---|
| Mesothelin | Mesothelioma | ATA2271 + pembrolizumab | ORR 63%, 12-mo OS 80% | Grade 3 pneumonitis (attributed to pembrolizumab) 6 |
| Mesothelin | Mesothelioma | aPD1-MSLN JL-Lightning | ORR 100% at DL2 (n=3) | Grade 3-4 CRS; immune-mediated pneumonia 626 |
| GPC3 | HCC | C-CAR031/AZD-7003 | ORR 56.5%, DCR 91.3% | CRS grade 1-2; no ICANS 6 |
| HER2 | GEJ/Breast | CCT303-406 | 100% SLD reduction (1 pt); 12-mo OS 83.3% (high dose) | No DLTs, no ICANS 626 |
| CLDN18.2 | Gastric/PDAC | ELIMYN18.2 | ORR 26.3% (42.9% gastric) | CRS in 17/19; 2 grade 3 CRS; 2 grade 1 ICANS 29 |
| CLDN18.2 | Gastric/GEJ | Satri-cel (NMPA approved June 2026) | Regulatory approval milestone | Ongoing post-marketing safety surveillance 28 |
| B7-H3 | DIPG (pediatric) | BrainChild-03 (ICV) | Sustained improvement in 1/3 patients | No DLTs across 40 infusions 9 |
| B7-H3 | rGBM | Phase I (IT/ICV) | Median OS 14.6 months | TIAN in 81% of infusions; managed with anakinra 26 |
| EGFR+IL13Rα2 | rGBM | CART-EGFR-IL13Rα2 | 85% tumor shrinkage, -35% median | Grade 3 ICANS 56% 26 |
| GD2 | Neuroblastoma | KUR-501 (CAR-NKT) | ORR 25%, DCR 58% | No DLTs, grade 2 CRS in 1 patient 6 |
| PSMA | mCRPC | P-PSMA-101 | 71% PSA decline, 36% ≥50% PSA reduction | Grade ≥3 CRS in 2; ICANS in 2 6 |
| PSMA | Prostate | TmPSMA-02 | Trial terminated | 2 fatal ICANS events 6 |
| CEA | CRC liver mets | HAI CAR-T | 57% recurrence-free (high dose) | No grade ≥3 adverse events 26 |
Next-Generation Engineering Strategies
Armored and Cytokine-Secreting CAR-T
Fourth-generation "armored" CAR-T cells are engineered to secrete immunostimulatory cytokines—IL-12, IL-15, IL-18—that reshape the TME from immunosuppressive to pro-inflammatory, enhance CAR-T persistence, and activate bystander immune cells. IL-15-armored GPC3-CAR-T showed a DCR of 66% and ORR of 33% in GPC3-positive solid tumors with manageable toxicity 1. DLL3-targeted LB2102, armored with dominant-negative TGFβRII (dnTGFβRII) to counteract TGF-β-mediated suppression, demonstrated dose-dependent activity in small-cell lung cancer without dose-limiting toxicities 26. Uncontrolled cytokine release, however, necessitates inducible safety switches such as inducible caspase-9 (iC9) systems activated by the chemical inducer AP1903 1.
TGF-β Resistance Engineering
SMAD7, a suppressor of TGF-β signaling, has been co-expressed with HER2-targeted CAR-T cells to confer resistance to TGF-β-induced exhaustion. SMAD7-coexpressing CAR-T cells demonstrate high cytolytic efficacy, sustained tumoricidal capacity after continuous antigen exposure, and substantially reduced inflammatory cytokine production—with enhanced infiltration confirmed in patient-derived tumor organoids 4.
Logic-Gated and Dual-Target CARs
To address both antigen heterogeneity and on-target/off-tumor toxicity, logic-gated designs have reached clinical evaluation. AND-gate CARs require co-expression of two tumor-associated antigens for activation, improving selectivity. The HLA loss-of-heterozygosity-based Tmod platform (e.g., A2B694: mesothelin activating CAR plus HLA-A*02 blocking CAR) distinguishes tumor from normal cells and is being evaluated in ovarian, pancreatic, and lung cancers without dose-limiting CRS or neurotoxicity in early cohorts 826. A TALEN-edited allogeneic FAP/mesothelin IF/THEN-gated CAR demonstrated in vivo anti-tumor activity exclusively in FAP-expressing tumor compartments, with no off-tumor effects in FAP-negative contralateral tumors in preclinical models 7. The EGFRvIII-triggered synNotch circuit (NCT06186401), which induces downstream targeting of EphA2/IL13Rα2 only when EGFRvIII engagement occurs, represents another clinically advancing logic-gated approach 8.
Regional Delivery
Locoregional administration is rapidly becoming a standard feature of solid-tumor CAR-T trial design. Intrapleural delivery for pleural mesothelioma, intraperitoneal delivery for peritoneal carcinomatosis, hepatic artery infusion (HAI) for liver-confined disease, and intratumoral/intraventricular administration for CNS tumors each improve local exposure while reducing systemic toxicity. In colorectal peritoneal disease, IP-delivered CEA-targeting CAR-T achieved 75% tumor-free survival at day 34 in preclinical models versus partial relapse with IV delivery 5. For glioblastoma, dual intratumoral/intraventricular delivery was associated with statistically better survival than single-route delivery in a 41-patient cohort 10.
Combination with Checkpoint Inhibitors and Oncolytic Viruses
Radiotherapy normalizes tumor vasculature, upregulates homing chemokines (CXCL9-11, CCL5), induces immunogenic cell death, and can upregulate CAR-targeted antigens such as CD20, CEA, and MUC-1 5. Oncolytic viruses synergize with CAR-T through viral-mediated TME remodeling, DAMP release, and direct tumor lysis 14. The allogeneic NKG2D-targeting CYAD-101 combined with pembrolizumab in the KEYNOTE-B79 trial for colorectal cancer achieved a DCR of 73.3% in the preceding alloSHRINK trial 1.
Allogeneic and Off-the-Shelf Platforms
Allogeneic CAR-T and CAR-NK platforms address manufacturing bottlenecks by enabling pre-manufactured, off-the-shelf availability. The allogeneic CD70-targeting ALLO-316 in renal cell carcinoma achieved 20% ORR (33% in CD70-high tumors) without graft-versus-host disease, though three grade-5 adverse events were attributed to the investigational product 626. CAR-NK cells offer complementary advantages—broader cytotoxic pathways including ADCC, lower graft-versus-host risk, reduced manufacturing costs, and potentially less exhaustion susceptibility—and are increasingly evaluated in solid-tumor settings 121621. Virus-free CRISPR knockout of KLRC1 combined with GD2 CAR knock-in in primary NK cells demonstrated 98% knockout efficiency and up to 23% transgene integration, overcoming HLA-E-mediated inhibition in GD2+ melanoma models 1722.
Future Directions and Clinical Implications
Patient Selection and Biomarkers
Molecular subtyping will be critical to identifying patients most likely to benefit. Among colorectal cancer subtypes, CMS1 (MSI-immune) with high tumor-infiltrating lymphocytes represents the most favorable CAR-T candidate, whereas CMS4 (mesenchymal) with dense stroma poses the greatest challenge 1. Central antigen expression profiling, TME characterization, and CAR-T expansion kinetics as pharmacodynamic biomarkers should be incorporated into trial design from inception.
Response Assessment and Toxicity Management
RECIST v1.1 criteria are inadequate for capturing CAR-T-specific phenomena such as pseudoprogression and delayed response; modified RANO criteria and iRANO for neuro-oncology provide improved frameworks 26. CRS and ICANS grading using CTCAE v5.0 with early tocilizumab or anakinra intervention is standard, but inflammation-associated neurotoxicity (TIAN) in CNS-delivered programs requires additional anakinra and dexamethasone protocols 26. The fatal outcomes observed in TmPSMA-02 mandate rigorous, prospective neurotoxicity surveillance across all solid-tumor CAR-T programs 6.
Trial Design Innovations
Adaptive phase I/II seamless designs with real-time dose-escalation informed by efficacy and safety, regional delivery comparison arms, combination strategies with checkpoint inhibitors or ablation modalities, and biomarker-stratified enrollment represent best practices for the current investigational landscape. Long-term safety monitoring for subsequent malignant neoplasms, as documented in anti-CD19 CAR-T cohorts, should extend to solid-tumor programs as follow-up matures 23.
Conclusion
CAR-T cell therapy has crossed the threshold from theoretical feasibility to genuine early-phase clinical activity in solid tumors. The NMPA approval of satri-cel for CLDN18.2-positive gastric cancer in June 2026 marks the world's first regulatory approval for a solid-tumor CAR-T product 28. Yet responses remain inconsistent, persistence is often short-lived, and the therapeutic window is narrow. Success will require integrated solutions: optimized CAR design tailored to low-antigen environments 18, TME remodeling through armoring and combination strategies, regionally targeted delivery, robust patient selection via molecular biomarkers, and rigorous standardized toxicity management. As this convergence of engineering, immunology, and clinical science matures, CAR-T therapy holds genuine potential to extend its transformative impact from hematologic malignancies to the far more complex landscape of solid tumors 2425.