The core idea is deceptively simple. Natural T cells do not rely on their antigen receptor alone; they depend on co-stimulatory receptors such as CD28 and ICOS, the inducible T cell co-stimulator, to fine-tune activation, survival and differentiation. The researchers reasoned that if a therapeutic T cell receptor could be fused directly to a co-stimulatory signaling domain, every encounter with a tumor cell would deliver not just a recognition signal but a survival and persistence signal at the same time. The construct they engineered, which they call TCR:ICOS, combines the extracellular variable and constant domains of a conventional T cell receptor with a CD28 transmembrane segment and intracellular domains drawn from both ICOS and CD3 epsilon. The result is a single hybrid receptor that couples antigen specificity to co-stimulation in one molecular unit.

The functional consequences in mice were dramatic. T cells carrying the TCR:ICOS receptor showed enhanced, antigen-specific production of inflammatory cytokines, the chemical weapons that help recruit and coordinate an immune attack. More importantly, the cells persisted far longer within tumors than their conventional counterparts, and this persistence translated into clinical outcomes: delayed recurrence of melanoma and, in a subset of animals, complete and durable cures. For a field in which the transience of engineered T cells is a central bottleneck, the demonstration that a receptor-level modification can extend the working lifespan of the cells inside a hostile tumor microenvironment is a significant proof of concept.

Beneath the phenotypic changes lies a defined signaling logic. The team found that TCR:ICOS activation engaged two major pathways, PI3K and NF-kappa B, while paradoxically restraining the activation of AKT, a kinase downstream of PI3K that is often associated with terminal differentiation and metabolic burnout in T cells. This combination appears to be the key to the durability. Sustained PI3K and NF-kappa B signaling supports inflammatory function and survival, while blunted AKT activity helps the cells avoid the hyperactive, exhausted state that typically shortens the life of tumor-infiltrating lymphocytes. When the researchers genetically ablated the ICOS-PI3K pathway, the long-term anti-tumor effects disappeared, confirming that this signaling axis is not a side effect but the mechanistic engine of the durable response.

Single-cell level analysis reinforced the picture of a fundamentally altered differentiation program. TCR:ICOS T cells were enriched for a stem-like state, a less differentiated, self-renewing condition that immunologists regard as the hallmark of long-lived, re-challenge-capable T cell populations. They also showed resistance to exhaustion, the progressive loss of function marked by inhibitory receptors such as PD1, LAG3 and TIM3 that plagues conventional tumor-infiltrating cells. In practical terms, the engineered cells behaved less like short-lived commandos and more like a renewable garrison, capable of maintaining pressure on the tumor over time rather than expending themselves in a single burst of activity.

Translating the concept from mouse to human T cells required an additional round of molecular engineering. Early versions of the hybrid receptor did not express efficiently on the surface of human cells, a common obstacle in receptor design where folding, assembly and trafficking can fail silently. The team solved this by identifying a single amino acid change in the cytosolic tail of the receptor that enabled functional surface expression. Crucially, the optimized receptor avoided two well-known hazards of introducing engineered T cell receptors into patient cells: mispairing with endogenous TCR chains, which can create unpredictable and potentially dangerous specificities, and competition for limited CD3 molecules, which can impair the function of the T cell’s native receptor complex.

The human-cell experiments went beyond a single antigen. The researchers showed that the optimized TCR:ICOS format sustained the functional performance of human T cells across repeated stimulation cycles and could be extended to multiple tumor antigens, including targets relevant to solid cancers such as NY-ESO-1 and ROPN1, without eroding T cell fitness. This uniform applicability matters for the field. Many receptor-engineering advances are idiosyncratic, working for one TCR or one antigen but failing when generalized. A format that preserves T cell quality across different specificities offers a modular platform: any tumor-reactive TCR could, in principle, be equipped with the same built-in co-stimulation and inherit the same durability advantages.

The findings arrive at a moment when the cell therapy field is intensely focused on next-generation designs. Chimeric antigen receptors, or CARs, already incorporate co-stimulatory domains such as CD28 or 4-1BB, and that choice profoundly shapes how CAR T cells persist and differentiate. TCR-based therapies, which can recognize intracellular tumor antigens presented by HLA molecules and therefore access a much larger pool of cancer targets, have lacked an equivalent, systematic way to embed co-stimulation. The TCR:ICOS design fills that gap, and its mechanism, favoring stem-like persistence through PI3K and NF-kappa B while restraining AKT, offers a template that other groups can rationally modify.

There are, of course, steps between a mouse melanoma model and approved therapy. The study was conducted in preclinical systems, and human trials will need to establish safety, particularly given that engineered co-stimulation could, in principle, amplify off-tumor reactivity if a TCR recognizes healthy tissue. Dosing, manufacturing consistency and the behavior of TCR:ICOS cells in the complex, immunosuppressive environment of human solid tumors all remain to be tested. The authors note that the work was supported by the Dutch Cancer Society and a Health Holland public-private partnership, and several team members report pending patents related to the receptor design, signaling commercial interest in bringing the platform toward clinical evaluation.

Even so, the study offers a compelling answer to one of adoptive cell therapy’s most persistent questions: how to make engineered T cells last. By fusing recognition and co-stimulation into a single receptor, the Rotterdam team has shown that durability can be designed into the therapeutic product itself rather than bolted on with cytokines, checkpoint blockade or lymphodepleting chemotherapy. If the format performs in human trials as it has in mice, TCR:ICOS could become a foundational component of fitter, longer-lived T cell products for solid tumors, turning a transient spark of immune attack into a sustained campaign against cancer.

Subject of Research: Engineering T cell receptors with built-in ICOS co-stimulation to extend the durability of adoptive T cell therapy against solid tumors

Article Title: T cell receptors equipped with ICOS provide T cells with durable anti-tumor response

Article References: Marraffa, A., Berrevoets, C., Mosiello, M., Coelho, R. M., Roelofs, D., van Brakel, M., Wijers, R., Kroese, K., Peeters, M. J., Dik, W. A., Kunert, A., Abbott, R. J., Hammerl, D., Schliehe, C., & Debets, R. (2026). T cell receptors equipped with ICOS provide T cells with durable anti-tumor response. Molecular Cancer, 25(1), Article 216. https://doi.org/10.1186/s12943-026-02765-9

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02765-9

Keywords: adoptive T cell therapy, T cell receptor, ICOS, co-stimulation, T cell exhaustion, solid tumors, melanoma, PI3K signaling, NF-kappa B, stem-like T cells, gene engineering, Molecular Cancer

Immunotherapy, particularly Chimeric Antigen Receptor (CAR) T cell therapy, has already redefined the landscape of hematologic cancers such as leukemia and lymphoma by harnessing the patient’s own immune cells to eradicate malignant cells. The process involves extraction of T cells, their genetic reprogramming with a synthetic receptor, and reinfusion into the patient’s bloodstream. Despite its seminal success against blood cancers, this modality faces formidable hurdles when applied to solid tumors. The tumor microenvironment’s complexity and the difficulty of CAR T cells to adequately recognize and penetrate solid masses call for novel receptor configurations and sophisticated cell engineering approaches.

The combinatorial explosion of possible CAR T cell designs, coupled with the growing wealth of accumulated experimental data and literature, presents a daunting analytical challenge. To tackle this, Miskov-Zivanov aims to build an AI-powered system capable of sifting through vast bodies of scientific literature and heterogeneous data repositories to integrate expert knowledge and raw experimental insights. This system will intelligently recommend superior therapeutic lymphocyte designs, including both CAR T cells and tumor-infiltrating lymphocytes (TILs), by synthesizing disparate sources of information into actionable engineering guidance.

Drawing on her unique background as a computer engineer with extensive postdoctoral experience in computational and systems biology, Miskov-Zivanov emphasizes automation in a field traditionally dominated by labor-intensive manual processes. She envisions her computational framework as a catalyst that automates the complex tasks typically performed by biologists, thereby accelerating and refining the design cycle for immunotherapeutic cells. This aspiration springs from her conviction that the convergence of computation and biology can unveil novel pathways that manual curation might never reveal.

Building on her earlier NSF-funded EAGER award, which developed a prototype tool utilizing Natural Language Processing (NLP) to extract pertinent data from scientific texts, she now evolves the approach to incorporate state-of-the-art large language models (LLMs) and neural networks. This hybrid system will not only parse and analyze scientific papers but also interpret experimental datasets to conduct comprehensive in silico experiments. By simulating thousands of potential cell designs computationally, this framework will perform hypothesis-driven screening prior to laboratory validation.

A critical innovation in Miskov-Zivanov’s project lies in developing improved prompting techniques for AI models, enabling more precise and relevant extraction of meaningful data from the overwhelming corpus of biomedical literature. Instead of forcing researchers to navigate tens of thousands of papers, many irrelevant to their queries, the system will pinpoint high-impact insights and knowledge, distilling the essence of complex biological narratives. This capability could dramatically reduce time and resources consumed in immunotherapy research and design.

To represent and utilize the extracted knowledge efficiently, Miskov-Zivanov converts science-derived data into knowledge graphs (KGs)—structured semantic networks encoding relationships among biological entities like proteins, signaling pathways, and cellular behaviors. These KGs serve as a scaffolding layer upon which graph neural networks (GNNs) operate. GNNs, leveraging their prowess in modeling graph-structured data, analyze interconnections within the KGs to predict the efficacy of various immunotherapeutic cell configurations. This synergistic blend amplifies predictive accuracy beyond what isolated datasets or traditional statistical models can achieve.

Understanding the imperative for educating emerging engineers in these frontier methodologies, Miskov-Zivanov has introduced a novel graduate-level course focused on knowledge graphs and their construction, interpretation, and application. She believes that equipping the next generation of researchers with computational tools capable of integrating structured knowledge and data-driven learning models is vital for addressing increasingly complex biomedical challenges. By nurturing interdisciplinary expertise, this educational initiative seeds future innovation in synthetic biology and therapeutic design.

Underlying this ambitious technological endeavor is the goal to establish a reliable methodology for engineering and systematically testing thousands of immunotherapeutic cell designs with diverse receptor systems. Success could catalyze breakthroughs in developing cellular therapies that effectively infiltrate and neutralize solid tumors—an enduring challenge in oncology. Moreover, the project aspires to contribute novel algorithmic innovations to identify, present, and validate trustworthy predictive data in biomedical research.

Reflecting on her motivation, Miskov-Zivanov shares a poignant narrative of how a childhood news story about a young leukemia patient cured by immunotherapy ignited her passion. Her dual lens as a computer engineer and a scientifically curious individual fuels her drive to forge impactful applications of computing technologies in life-saving medical research. Her work epitomizes the compelling convergence of artificial intelligence and biotechnology, promising to reshape cancer treatment paradigms.

Her department chair, Alan George, lauds her as a rising star and innovator whose research lab, the MeLoDy (Mechanisms and Logic of Dynamics) Laboratory, bridges digital circuits, synthetic biology, AI, and dynamic systems. The award spotlights Miskov-Zivanov’s pioneering approach to designing immunotherapies and teaching complex computational methods, setting the stage for profound future contributions in science and engineering.

Dr. Miskov-Zivanov’s project embodies the forefront of biomedical innovation, where AI-powered automation intersects with molecular engineering to tackle the enduring challenge of cancer therapy. By weaving together computational linguistics, graph theory, machine learning, and synthetic biology, she charts a new course toward more efficient, interpretable, and impactful immunotherapy design. The convergence of these fields promises to accelerate discovery and ultimately transform patient outcomes in oncology.

Subject of Research: Artificial Intelligence-driven design of immunotherapy cells, focusing on CAR T cells and tumor-infiltrating lymphocytes.

Article Title: Artificial Intelligence-Driven Framework Poised to Revolutionize Immunotherapy Design

News Publication Date: Not specified in the provided content.

Web References:

• Natasa Miskov-Zivanov Faculty Page
• NSF Award Detail
• Pitt News on NSF EAGER Award
• MeLoDy Laboratory

Keywords: Cancer immunotherapy, Generative AI, Computer science, Artificial intelligence, Deep learning, Systems neuroscience, T lymphocytes, Immune system