The development of in vivo CAR-T therapy depends on advances in gene delivery and editing. Unlike adoptive therapy, the therapeutic outcome in this approach is determined during the in vivo generation process, which relies on the precision, efficiency, and accuracy of CAR gene delivery. Consequently, the selection of an appropriate delivery vector and strategy becomes a critical consideration in the manufacturing process. To date, three main technical platforms have been employed for in vivo CAR-T generation, each offering distinct characteristics and advantages.

CAR: chimeric antigen receptor, LNPs: lipid nanoparticles.

Lentiviral vectors and adeno-associated viruses (AAV) are the most commonly used viral vectors. Lentiviral vectors, capable of stable gene integration and widely used in ex vivo CAR-T, are promising for in vivo delivery. However, their in vivo application still faces several challenges, such as host immune responses that may clear the virus, difficulties in specifically targeting T cells, and the potential tumor risk associated with random genomic integration.

To enhance the efficiency and safety of in vivo CAR-T generation, researchers have developed various optimization strategies. For example, some lentiviral platforms display multi-domain fusion proteins (MDFs) on the viral surface, simultaneously providing T cell activation signals and co-stimulatory signals, thereby improving T cell transduction efficiency. Other studies focus on targeting specific T cell subsets (such as CD4+ or CD8+ T cells), with the resulting CAR-T cells demonstrating enhanced persistence and antitumor activity in vivo. AAVs have also been explored, successfully reprogramming T cells and demonstrating antitumor effects in models.

Despite promising preclinical results, genomic integration risks must be carefully monitored to avoid potential malignant transformation. Thus, viral vectors hold great potential, but safety and controllability remain key considerations.

Nanoparticle (NP) platforms, mainly lipid nanoparticles (LNPs), offer non-viral gene delivery for in vivo CAR-T cell generation. Compared to viral vectors, nanoparticles are more readily engineered to target T cells, offer improved safety profiles, and are relatively cost-effective. Moreover, mRNA delivery avoids risks associated with genomic integration. They are also suitable for large-scale production, transport, and storage. LNPs typically contain ionizable lipids, phospholipids, cholesterol, and PEGylated lipids, with surface modifications enhancing T-cell targeting and delivery efficiency.

Early studies explored biodegradable polymeric nanocarriers and transposon systems for in vivo CAR-T cell engineering. Although initial efficiency was limited, these studies demonstrated the feasibility of in situ CAR-T generation. The mRNA-LNP technology advanced by COVID-19 vaccines has opened new opportunities for immune cell therapy. For example, CD5-targeted LNPs delivering CAR mRNA successfully edited T cells and cleared fibroblasts in mouse models, improving cardiac function. Strategies that co-express CARs and cytokines have further enhanced therapeutic efficacy. The latest targeted LNP (tLNP) technologies further reduce off-target delivery to the liver and have shown significant antitumor or B-cell depletion effects in humanized leukemia mouse models and non-human primates, along with favorable safety profiles.

To address limitations such as mRNA persistence and delivery efficiency, researchers have also investigated circular RNA (circRNA) and DNA platforms. CircRNA, due to its circular structure, is more stable than linear mRNA and can prolong protein expression. DNA carriers, through transposon systems, enable stable integration and sustained expression. Despite optimization, challenges remain in persistence, immunogenicity, and editing efficiency. Overall, nanoparticle platforms offer a flexible, safe, and scalable option for in vivo CAR-T therapy, representing a promising direction for future immune cell therapies.

Beyond direct injection of viruses or nanoparticles, scaffold implants have emerged as a promising strategy for localized in vivo CAR-T delivery. Scaffold-mediated in vivo CAR-T generation is primarily achieved in two approaches.

The first approach pre-isolates T cells from patient blood, co-encapsulates them with a CAR vector in a scaffold, and implants it at the tumor site for in vivo CAR-T generation. Although it requires brief ex vivo isolation, T cell activation and expansion occur entirely in vivo.

The second approach directly implants a scaffold with the CAR vector at the disease site to recruit and edit T cells in vivo. For example, Inamdar et al. used a porous collagen implant to recruit immune cells. The vector assembles CAR molecules, followed by rapid T cell expansion and release into tumors, suppressing growth and prolonging survival in mice^[4]. Compared to direct vector injection, implants enable rapid local CAR-T generation and reduce off-target expression.

In summary, implants provide an in situ platform for CAR integration, T cell expansion, and activation. This approach addresses limitations in CAR-T numbers and enables sustained local release. Compared to systemic approaches, it offers better efficacy and safety and could become a versatile platform for CAR-T engineering.