CAR-T Cell Therapy Faces Its Next Test: Solid Tumours, Safer Designs and Lower Costs

Chimeric antigen receptor T-cell therapy has delivered one of the most dramatic shifts in modern oncology, converting certain relapsed and refractory blood cancers from death sentences into manageable, sometimes curable, diseases. Engineered T cells carrying synthetic receptors that recognise CD19 or BCMA have produced durable remissions in B-cell acute lymphoblastic leukaemia, non-Hodgkin lymphoma and multiple myeloma, and multiple products including tisagenlecleucel, axicabtagene ciloleucel, lisocabtagene maraleucel, idecabtagene vicleucel and ciltacabtagene autoleucel have now received regulatory approval. Yet a comprehensive review published in Frontiers in Molecular Medicine argues that this success story masks substantial unresolved problems: relapse through antigen escape, life-threatening inflammatory toxicities, manufacturing bottlenecks, prohibitive costs, and a conspicuous failure to replicate results in solid tumours. The review’s central message is that the field’s future depends as much on engineering, biomarker science and health economics as on receptor design itself.

The most intensively explored strategy for overcoming relapse is multi-antigen targeting. When tumours are pushed by a single-target CAR, they frequently survive by losing, downregulating or masking the targeted antigen, a phenomenon well documented after CD19-directed therapy. Three architectural solutions dominate current development. Pooled approaches administer two independently manufactured monospecific products, such as separate CD19- and CD22-directed cells, preserving modular design flexibility but risking dominance of one population after infusion. Bicistronic dual-CAR T cells carry two independent receptors within the same cell, creating an OR-gate in which either antigen triggers cytotoxicity, providing more uniform cell-by-cell coverage at the cost of a larger genetic payload and more complex signalling. Tandem CARs fuse two binding domains into a single receptor chain, ensuring both specificities on every engineered cell, but their performance is acutely sensitive to structural configuration, including domain order, linker flexibility, hinge length and binder affinity, and steric hindrance can compromise recognition of single-antigen-positive cells. Clinical studies of CD19/CD22 and CD19/CD20 constructs have shown feasibility and antitumour activity in acute lymphoblastic leukaemia and large B-cell lymphoma, but no strategy has yet established universal superiority over optimised single-target products.

Understanding why requires a look at the receptor generations themselves. First-generation CARs coupled an antibody-derived single-chain variable fragment directly to the CD3ζ signalling domain and, lacking costimulation, persisted poorly in patients. Second-generation constructs added a single costimulatory domain such as CD28 or 4-1BB, dramatically improving expansion, persistence and resistance to exhaustion, and they remain the clinical standard underlying every approved therapy. Third-generation receptors combining two costimulatory domains have shown stronger preclinical activity but no consistent clinical advantage, and may amplify inflammatory toxicity. Fourth-generation armoured CARs add inducible transgenes encoding cytokines such as IL-12 or IL-18 that remodel the immunosuppressive tumour microenvironment and recruit endogenous immune cells, an approach considered particularly attractive for solid tumours, though controlling transgene expression and avoiding cytokine-mediated toxicity remain significant barriers to translation.

Solid tumours remain the field’s hardest problem, and clinical trials have delivered sobering lessons. A landmark Phase I trial of EGFRvIII-directed CAR T cells in glioblastoma showed that engineered cells successfully trafficked to tumour sites yet produced no durable responses; post-treatment analyses revealed rapid loss of the target antigen alongside increased regulatory T cells, tumour-associated macrophages and PD-L1 expression, a textbook demonstration of adaptive immune resistance. An earlier HER2-directed trial suffered a fatal on-target, off-tumour pulmonary toxicity caused by low-level HER2 expression in normal lung, an event that reshaped the entire discipline’s approach to antigen selection and dose escalation. Mesothelin-directed cells have been safe in mesothelioma, pancreatic and ovarian cancers but have mostly achieved disease stabilisation rather than regression. The recurring barriers are heterogeneous antigen expression, dense desmoplastic stroma that physically blocks infiltration, abnormal vasculature, and a microenvironment saturated with TGF-β, IL-10, adenosine, regulatory T cells and myeloid-derived suppressor cells, all of which drive CAR-T-cell exhaustion through inhibitory receptors such as PD-1, TIM-3 and LAG-3.

Against this backdrop, a genuine milestone has now been reached. Satri-cel, a CAR targeting claudin 18 isoform 2, a tight-junction protein exposed on gastric cancer cells but largely concealed in healthy mucosa, significantly prolonged median progression-free survival from 1.77 to 3.25 months compared with physician’s-choice treatment in the randomised Phase II CT041-ST-01 trial in heavily pretreated gastric and gastro-oesophageal junction cancer. In June 2026, China’s National Medical Products Administration approved the product for CLDN18.2-positive, HER2-negative advanced gastric cancer after at least two prior systemic lines, making it the first regulator-approved CAR-T-cell therapy anywhere in the world indicated specifically for a solid tumour. The review’s authors are careful to frame this as proof of principle rather than a universal solution: responses still depend on antigen density and heterogeneity, grade 3 or worse adverse events were frequent, and the treatment demands lymphodepletion and specialist monitoring.

Toxicity management is simultaneously becoming more sophisticated and more predictive. Cytokine release syndrome, driven largely by IL-1 and IL-6 from activated macrophages, remains the most common acute complication and is treated first-line with tocilizumab, with corticosteroids reserved for severe cases because excessive immunosuppression may erode CAR-T-cell persistence. Immune effector cell-associated neurotoxicity syndrome, linked to endothelial activation and blood-brain barrier disruption, can progress from encephalopathy to cerebral oedema, and severe CRS can also trigger coagulopathy and prolonged cytopenias. The review highlights an emerging biomarker toolkit that goes well beyond IL-6: persistently elevated ferritin and C-reactive protein track with severe CRS, soluble IL-2 receptor alpha adds independent predictive value as a T-cell activation marker, and endothelial injury markers including angiopoietin-2, von Willebrand factor and D-dimer illuminate the vascular dimension of toxicity. Machine-learning models, particularly gradient-boosting algorithms applied to multiparametric datasets, have shown encouraging accuracy in flagging patients destined for severe ICANS or high-grade CRS before clinical deterioration begins.

Cost and accessibility constitute what may ultimately be the field’s defining constraint. Approved autologous CAR-T products routinely cost between 400,000 and 500,000 US dollars per patient once manufacturing, hospitalisation, lymphodepletion and toxicity management are counted, and the individualised production workflow, from leukapheresis through genetic modification, expansion, quality control and cryopreserved transport, typically takes two to four weeks during which rapidly progressive disease may render a patient ineligible. Allogeneic off-the-shelf products manufactured from healthy donors, enabled by CRISPR/Cas9 disruption of the endogenous T-cell receptor and HLA molecules to prevent graft-versus-host disease and rejection, promise immediate availability and economies of scale. Early candidates such as ALLO-501 and its successor ALLO-501A, which adds CD52 disruption permitting selective lymphodepletion, produced clinically meaningful responses in lymphoma with no reported graft-versus-host disease, though host rejection, limited persistence and small early-phase datasets keep their advantage over autologous products unproven. Automated closed-system manufacturing, non-viral transposon-based gene transfer, point-of-care production and rational combinations with checkpoint inhibitors or targeted agents such as BTK and BCL-2 inhibitors are all being pursued to compress costs further.

Alternative cellular platforms are broadening the engineering arsenal in parallel. γδ T cells, which recognise stressed and transformed cells independently of MHC presentation and carry a naturally low risk of graft-versus-host disease, offer a compelling route around antigen escape, but they constitute only 1 to 5 percent of circulating lymphocytes, their ex vivo expansion is highly donor-dependent, subset heterogeneity complicates manufacturing, and their in vivo persistence falls short of conventional αβ CAR-T cells. CAR-NK cells share the off-the-shelf appeal and a favourable safety profile, with early CD19-directed trials showing encouraging responses, but genetic modification remains inefficient and persistence is limited. More exotic platforms, including CAR macrophages that phagocytose tumour cells and remodel stroma, CAR neutrophils capable of penetrating glioblastoma, invariant NKT cells, double-negative T cells, MAIT cells and fibroblast activation protein-targeted cells that attack cancer-associated fibroblasts, all show preclinical promise but face unresolved questions of scalability, persistence and long-term safety. Switchable CAR systems that decouple recognition from activation through soluble adaptor molecules add a further layer of control, allowing clinicians to withdraw activity in the event of severe toxicity or retarget cells to new antigens without manufacturing new products.

The review’s synthesis is deliberately unsentimental. CAR-T-cell therapy is a major advance, not a universally effective cancer treatment, and its next decade will be judged on dimensions that extend far beyond antitumour potency: safer multi-target and logic-gated constructs, validated biomarker panels integrated with artificial intelligence for toxicity prediction and patient selection, automated and decentralised manufacturing, rigorous randomised evidence for allogeneic and alternative platforms, and health-economic and regulatory frameworks capable of delivering these therapies beyond specialised centres in high-income countries. The approval of the first solid-tumour CAR-T product demonstrates that rational antigen selection can crack previously resistant cancers; sustaining that momentum across tumour types, and making the treatments affordable and equitable, is now the central scientific and societal challenge.

Subject of Research: Future directions of chimeric antigen receptor T-cell therapy, including multi-target designs, solid tumour challenges, toxicity biomarkers and cost reduction strategies

Article Title: The future directions of CAR-T Cell therapy: unlocking the potential of immunotherapy in cancer treatment

Article References: The future directions of CAR-T Cell therapy: unlocking the potential of immunotherapy in cancer treatment. (n.d.). https://doi.org/10.3389/fmmed.2026.1931159

Image Credits: AI Generated

DOI: 10.3389/fmmed.2026.1931159

Keywords: CAR-T cell therapy, immunotherapy, solid tumours, antigen escape, cytokine release syndrome, allogeneic CAR-T, γδ T cells, biomarkers, manufacturing costs, claudin 18.2, CAR-NK cells, genome editing