Conventional CAR T cell therapy requires a multi-step ex vivo process involving the isolation of a patient’s T cells, their activation, genetic modification, expansion, and rigorous quality control assays. This workflow commonly extends over two to three weeks, during which time delicate cellular manipulations can compromise T cell functionality, and rapid disease progression may outpace treatment availability. Furthermore, the personalized nature of such therapies restricts scalability, confining benefits to select patients within specialized centers.

The cutting edge concept of in vivo CAR T cell production foregoes extracorporeal cell processing by delivering CAR genetic constructs directly into T cells within the patient’s body. This approach utilizes finely engineered viral vectors, such as lentiviruses and adeno-associated viruses (AAVs), alongside emerging nonviral delivery systems, including lipid nanoparticles. Upon administration, these vectors specifically transduce T cells in situ, effectuating genetic reprogramming that endows them with tumor-targeting capabilities. This innovation has the potential to drastically reduce production complexities, cut timelines, and improve therapeutic potency by preserving T cell phenotypes in their native milieu.

One of the foremost advantages of in vivo CAR T therapy lies in its inherent scalability and potential to yield “off-the-shelf” CAR T cell products. Contrasting with the “one patient, one batch” model of ex vivo manufacturing, in vivo strategies could harness a more universal delivery modality, enabling broader patient reach and cost efficiencies unimaginable with current standards. Moreover, retaining T cells within their physiological environment mitigates the risk of functional exhaustion seen in cultured cells, thus enhancing efficacy and durability of tumor eradication.

Nanoparticle-based delivery systems exemplify a promising nonviral vector class facilitating efficient CAR gene transfection with minimal immunogenicity. Their ability to encapsulate nucleic acids and traverse biological barriers allows for targeted T cell modification without integrating viral components, thereby alleviating concerns regarding insertional mutagenesis. Advances in materials science have led to the design of nanoparticles optimized for stability, biodistribution, and cell-specific uptake, which are critical parameters for clinical translation.

Viral vectors such as lentiviruses and AAVs remain pivotal due to their high transduction efficiency and ability to confer stable CAR expression. Lentiviral vectors integrate into the T cell genome, ensuring persistent CAR expression, whereas AAVs tend to remain episomal, offering a safer but transient modification profile. The refinement of vector tropism and promoter elements continues to improve transgene expression specificity and intensity, enhancing the precision of in vivo CAR T cell engineering.

Despite the promise, the transition to in vivo CAR T cell production is not without formidable challenges. Precise targeting is essential to avoid off-target modification of non-T cell populations, which could provoke adverse effects or diminish therapeutic efficacy. Immunogenic responses to vector components or newly expressed CAR proteins pose risks of rapid clearance, reduced transgene expression, or systemic inflammation. Additionally, insertional mutagenesis induced by integrating vectors remains a safety concern necessitating rigorous preclinical assessment.

The rapidly progressing biology of certain malignancies makes the expedited timeline of in vivo CAR T cell generation especially compelling. Bypassing ex vivo expansion could dramatically shorten the interval between diagnosis and treatment administration, potentially altering disease trajectories. Furthermore, overcoming manufacturing bottlenecks could democratize access to CAR T therapy beyond specialized centers, fostering more equitable cancer care.

An important consideration in advancing in vivo CAR T therapies is balancing transfection efficiency with cost-effectiveness and safety profiles. While viral vectors offer superior gene transfer efficiencies, their production costs and biosafety infrastructure requirements can be prohibitive. Conversely, nonviral systems promise more affordable manufacturing and flexibility but often suffer from lower transduction rates. Intensive research aims to optimize these platforms, perhaps combining the strengths of both approaches to achieve the ideal therapeutic index.

Ongoing studies are exploring the integration of synthetic biology and genome editing tools to refine the specificity and functionality of in vivo-generated CAR T cells. Innovations such as inducible CAR expression systems and multispecific CAR constructs may be harnessed to enhance tumor targeting while minimizing off-tumor toxicity. Additionally, multiplexed delivery systems could facilitate simultaneous modification of multiple immune cell types, broadening the scope of adoptive immunotherapy.

In summary, in vivo CAR T cell therapy stands at the frontier of personalized medicine, poised to overcome the scalability and logistical obstacles of traditional CAR T manufacturing. Its capacity for rapid, efficient, and cost-effective generation of functional CAR T cells could revolutionize clinical oncology, especially for aggressive cancers needing urgent intervention. While challenges surrounding safety, targeting specificity, and delivery vector optimization remain, the trajectory of current research augurs well for the translation of this approach into routine clinical practice.

The evolution of CAR T cell engineering from complex ex vivo bioprocesses to streamlined in vivo genetic modification mirrors the broader trend in gene therapy toward minimally invasive, patient-centric interventions. As the field progresses, collaborative efforts among immunologists, bioengineers, and clinicians will be paramount to harnessing the full potential of in vivo CAR T cell production, ultimately transforming the landscape of cancer treatment and patient outcomes.

Subject of Research: In vivo production of CAR T cells and its therapeutic potential in cancer treatment.

Article Title: In vivo production of CAR T cell: Opportunities and challenges.

News Publication Date: 1-Nov-2025.

References: Zhiqiang Song, Yi Zhou, Binbin Wang, Yuke Geng, Gusheng Tang, Yang Wang, Jianmin Yang, In vivo production of CAR T cell: Opportunities and challenges, Genes & Diseases, Volume 12, Issue 6, 2025, 101612, DOI: 10.1016/j.gendis.2025.101612.

Image Credits: Genes & Diseases.

Keywords: Cancer genetics, CAR T cell therapy, in vivo CAR T production, gene therapy, viral vectors, nanoparticle delivery, hematological malignancies, immunotherapy.

This research effort, led by a multidisciplinary team including clinicians and computational scientists, sought to assess whether widely available large language models (LLMs) such as ChatGPT and Gemini could be harnessed to generate individualized meal plans and comprehensive grocery lists adapted to the complex needs of cancer patients. By inputting variables such as cancer stage, co-morbid health conditions, socioeconomic constraints, geographic location, and cultural food preferences, the team endeavored to push these AI models beyond generic recommendations toward nuanced, actionable dietary strategies. Their findings, published recently in a peer-reviewed journal, reveal the remarkable potential, as well as current limitations, of LLM-guided nutritional support.

At the heart of this study is the recognition that cancer nutrition is inherently multifactorial. Patients not only contend with fluctuating appetites, treatment side effects, and altered metabolism but also with diverse cultural and economic realities. Senior author Dr. Nicole Simone, a radiation oncologist affiliated with the Sidney Kimmel Comprehensive Cancer Center, underscores the transformative capacity of this technology. By leveraging AI’s ability to process vast datasets and contextualize patient-specific factors, the team aimed to bridge longstanding gaps in personalized care that often leave vulnerable patients underserved.

The first phase of the investigation involved detailed prompt engineering, conducted by Julia Logan, a medical student whose hands-on summer research tested the response fidelity of the LLMs across a wide array of patient scenarios. These prompts simulated real-world conditions that reflect the complexity clinicians face, including budgetary limits and local ingredient availability. The artificial intelligence systems were able to generate meal plans that reflected these parameters with a surprising level of granularity. For example, the AI considered the cost-effectiveness of ingredient choices while incorporating ethnic foods familiar to patients, thereby supporting not only nutritional goals but also cultural adherence and patient satisfaction.

Wookjin Choi, PhD, a computational physicist and co-author, emphasized the pragmatic strategy behind the research. Instead of developing a specialized AI model from the ground up, which is resource-intensive and time-consuming, the team adapted existing large language models. These robust generalist LLMs served as adaptable foundations, upon which layers of domain-specific evaluation and modification were applied. This iterative process involved scoring the outputs, calibrating prompt structures, and gradually creating a refined model attuned to oncology nutrition nuances.

One of the most striking outcomes reported by the researchers was the LLMs’ capacity to integrate geographical and socioeconomic data effectively. The AI successfully accounted for local grocery store inventories, ensuring recommended ingredients were realistically obtainable by patients—a critical component often overlooked in digital health solutions. This adaptability signifies a paradigm shift, where AI tools may not only tailor medical advice to the individual’s biology but to their lived environment, enhancing practical utility.

The research team acknowledged, however, that AI-generated nutritional plans are not infallible. While the outputs demonstrated impressive alignment with dietitian-level recommendations in many cases, inherent risks exist regarding errors or omissions that could affect patient safety. Thus, the study posits an AI-human collaborative approach: using LLMs as scalable aides for preliminary dietetic guidance with subsequent professional oversight to verify and customize care plans further. This model could alleviate the burden on clinical nutritionists by streamlining routine advisory tasks while preserving the critical human element.

Furthermore, the study’s implications extend beyond oncology. The methodological framework of customizing AI-driven dietary advice based on multidimensional patient characteristics could revolutionize nutritional care across chronic diseases. The integration of socioeconomic factors and cultural sensitivity ensures recommendations resonate deeply with diverse patient populations, potentially reducing disparities in health outcomes linked to nutrition access and education.

Ethical considerations accompanying AI’s increasing role in healthcare naturally arise. Transparency in how dietary advice is generated, the explainability of AI decisions, and the safeguarding of patient data are vital components that the researchers plan to address in subsequent work. Additionally, ongoing evaluation of AI reliability and clinical validation remain paramount before widespread adoption.

In sum, this pioneering research underscores a promising frontier where artificial intelligence catalyzes patient-centered nutrition support in oncology. By systematically merging clinical insights, computational power, and real-world constraints, Thomas Jefferson University’s endeavor illuminates a path toward more equitable, personalized, and effective dietary assistance for cancer patients. While challenges remain, notably in defining the thresholds for AI autonomy versus professional supervision, this work lays critical groundwork for the future convergence of technology and compassionate cancer care.

As AI models continue to evolve in accuracy and contextual understanding, their deployment in healthcare settings must be executed thoughtfully. The success of this project not only highlights the potential of LLMs but elevates the conversation about integrating digital tools into clinical workflows. Embracing AI as an adjunct facilitator could ultimately enhance patient empowerment and adherence, especially for those grappling with the multidimensional burdens of cancer.

The journey ahead includes rigorous validation studies, broader demographic testing, and exploration of longitudinal impacts on patient health outcomes. Moreover, fostering interdisciplinary collaborations will be essential to translate these technological advances into clinical standards and patient education. By navigating these complexities, the medical community can harness AI’s transformative capacity while maintaining the trust and safety central to patient care.

The narrative emerging from this research is one of hope and innovation, where technology acts as a bridge over existing healthcare gaps. The ability to democratize expert-level dietary guidance means that nutritional intervention—a cornerstone of cancer therapy support—may become more accessible, scalable, and individualized than ever before. The implications resonate well beyond clinical nutrition, offering a template for AI-driven, culturally competent, and socioeconomically sensitive healthcare in the 21st century.

Subject of Research: Use of Large Language Models for Personalized Nutritional Guidance in Cancer Patients

Article Title: (Not specified in the content)

News Publication Date: (Not specified in the content)

Web References:

• https://www.mdpi.com/2072-6643/17/7/1176
• https://research.jefferson.edu/labs/researcher/simone-laboratory/team.html
• https://www.jefferson.edu/academics/colleges-schools-institutes/skmc/departments/radiation-oncology/faculty/directory/choi.html

References: (No additional references provided)

Image Credits: (No images provided)

Keywords: Generative AI, Cancer patients, Artificial intelligence, Oncology, Nutritional support