A pioneering form of immunotherapy has emerged for cancer treatment: Chimeric Antigen Receptor T-cell (CAR-T) therapy. This highly innovative approach involves genetically modifying a patient’s T cells to express chimeric antigen receptors (CAR) on their surface. These receptors are engineered to specifically recognize antigens present on the surface of tumor cells. After modification, CAR-T cells are reinfused into the patient, where they target and eliminate tumor cells. The concept of leveraging T cells to combat cancer was first pioneered by Steven Rosenberg and colleagues in 1988, when they treated metastatic melanoma patients with tumor-infiltrating T lymphocytes (TILs) isolated from the tumor, expanded ex vivo and reintroduced into patients (1). Moreover, lifileucel, a TIL based therapy was approved in 2024 for metastatic melanoma treatment (2). By creating the first generation of CAR-T cells termed « T-bodies », Zelig Eshhar and his colleagues made significant progress in the field in 1993 (3). CAR-T cells demonstrated impressive outcomes in treating hematological malignancies within the past ten years. In 2017, the U.S. Food and Drug Administration (FDA) approved Kymriah (tisagenlecleucel), the first CAR-T treatment for refractory B-cell acute lymphoblastic leukemia (ALL), (4). It was followed by other CAR-T therapies for lymphomas such as TECARTUS (brexucabtagene autoleulcel), (5) for refractory mantle cell lymphoma and ALL, YESCARTA (axicabtagene ciloleucel), (6) for refractory large B-cell (LBCL) and follicular lymphomas, BREYANZI (lisocabtagene maraleucel, 7) for refractory LBCL and chronic lymphocytic leukemia (CLL), and for refractory multiple myeloma (MM) such as ABCMA (idecabtagene vicleucel), (8) and CARVYKTI (ciltacabtagene autoleulcel, 9). Although ongoing efforts to develop CAR-T therapies for solid tumors, and overcome numerous obstacles related to the hostile tumor microenvironment (TME), further progress is needed before these therapies can reach clinical approval.

Over the years, CAR designs have rapidly improved, resulting in the development of five distinct generations (Figure 1). Each generation has added unique characteristics to enhance the efficacy and persistence of CAR-T cells (10, 11). These developments have significantly improved the intrinsic CAR-T cell activity and enhanced the ability to overcome immune evasion of cancer cells and challenges occurring within the hostile TME. While preclinical and early data are promising (12, 13), none of these constructs have yet received FDA approval, excepted CAR-Ts from the second generation. The first-generation CAR construct consists of an extracellular antigen-recognition domain linked to the intracellular CD3ζ signaling domain. In the second-generation construct, a co-stimulatory domain (CD28 or 4-1BB) is added alongside CD3ζ to enhance T cell activation and proliferation. The third-generation CAR construct is similar to the second one but incorporates more than one co-stimulatory domain to enhance CAR T cell activity and persistence. The fourth-generation CAR construct, also called T-cells redirected for universal cytokine killing (TRUCK), is designed to secrete cytokines upon activation to modulate the TME, attract innate immune cells, and improve antitumor efficacy (14). The fifth-generation CAR construct combines a STAT3-binding motif between the co-stimulatory domain and the CD3ζ signaling domain with a truncated cytoplasmic region of the IL-2 receptor β-chain. Without requiring external cytokine support, this design allows for antigen-dependent activation of the JAK-STAT pathway to enhance T cell proliferation and persistence (15).

In addition to design improvement, CAR-T manufacturing has been enriched by several innovations such as CRISPR-Cas9 gene editing to remove immune checkpoint inhibitors such as LAG-3, PD-1 or TIM-3 and enhance resistance to exhaustion and antitumor efficacy (16–18), adenoviral transduction providing efficient gene expression without genome integration and low risk of mutagenesis but offering a short-lived expression (19), retroviral transduction with stable gene integration for a long-term expression but limited to dividing cells with a risk of insertional mutagenesis (19, 20), and lentiviral transduction with stable gene integration for a long-term expression on a broader cell targeting range with lower risk of insertional mutagenesis but with high cost and manufacturing complexity (19). There are also mRNA electroporation, lipidic nanoparticles and transposon-mediated approaches (Figure 2) (21, 22). Moreover, CAR-T cells can be manufactured either ex vivo by extracting cells from the patient, engineering them, and reinfusing back or in vivo by directly injecting the CAR construct into the patient’s body (23). While in vivo CAR-T therapies represent a promising alternative to conventional ex vivo approaches, the first clinical trials have recently begun, and their efficacy and safety remain to be fully established. This is in contrast with second-generation CAR-T therapies produced ex vivo, some of which have already been approved by the FDA. Despite these advances, challenges remain such as immunosuppressive TME, tumor infiltration, antigen downregulation, and CAR-T cell expansion and persistence issues still arise in preclinical and clinical trials. Although results from in vitro characterization could differ from those obtained in in vivo mouse models and do not always predict outcomes in human clinical trials. Therefore, three-dimensional (3D) co-culture systems appear as attractive intermediate systems between in vitro and in vivo models to provide more physiologically relevant models for rapid assessment of CAR-T efficacy.

There are some advantages of using 2D co-culture systems (Table 1), such as the ease of use and reproducibility, because these systems are straightforward to establish and follow standardized protocols. Another advantage is the real-time observation because the single-layer arrangement of cells enables direct assessment of CAR-T activity, including cell adhesion, tumor cell lysis and proliferation, through techniques like microscopy and flow cytometry. The absence of a TME enables the targeted study of key parameters such as cytokine production, cytotoxicity, and CAR-T cell proliferation without interference from external factors. Moreover, the 2D cultures can easily integrate a variety of analytical techniques, including ELISA, flow cytometry, lactate dehydrogenase (LDH), or calcium flux assays, facilitating a comprehensive evaluation of CAR-T function. Finally, 2D models remain accessible for widespread use because of their lower cost and minimal specialized equipment required compared to in vivo or preclinical research.

Nevertheless, 2D co-culture systems have several limitations (Table 1), such as their inability to recreate a realistic TME. This is due to their failure to replicate oxygen and nutrient gradients as well as the complex interactions between tumor cells and extracellular matrix (ECM). Importantly, the 2D co-culture systems lack the ability to simulate immunosuppressive elements of the TME, such as cancer-associated fibroblasts (CAFs), M2 macrophages, and regulatory T cells (Tregs), which are critical obstacles to solid tumor therapy. The oversimplification can lead to an overestimation of CAR-T efficacy (27). Tumor cells in 2D models use cell lines, naturally expressing the target antigen, which may cause a loss of their native tumor phenotype and lead to false-positive results regarding CAR-T efficacy (28). Moreover, 2D culture seems not suitable for some primary cells, such as primary CLL and MM cells for hematological malignancies or primary pancreatic ductal adenocarcinoma (PDAC) which undergo rapid apoptosis or senescence in vitro, and the use of cell lines does not fully recapitulate the biology of the disease. However, cultivating primary cells in 3D models might provide the right microenvironment to support cancer cell survival, making it possible to study CAR-T cell activity for a longer time than in 2D cocultures. Furthermore, in a physiological context, CAR-T cells must overcome physical and biochemical barriers before interacting with tumors cells. This complexity is not adequately represented in 2D models, where CAR-T cells directly interact with tumor cells (29). Thus, 2D systems do not capture dynamics like CAR-T cell migration, penetration into dense tumor tissues, or exhaustion in hypoxic, poorly vascularized environments, and fail to include essential stromal and vascular components, which significantly influence tumor growth, immune evasion, and CAR-T cell functionality (28). Because of their simplicity to use, low cost and analytical compatibility, 2D culture systems remain indispensable for initial CAR-T cell research. However, their inability at accurately replicating the complexities of the TME highlights the necessity to switch to more physiologically relevant models. To fill these gaps, researchers increasingly depend on 3D co-culture systems to provide a transitional step between basic in vitro studies and in vivo validation. This allows for a more accurate evaluation of CAR-T behavior and efficacy in realistic tumor environments (Figure 3).

The first method for generating spheroids is direct formation by letting cells aggregate spontaneously together in a flat-bottom plate such as the liquid overlay technique, or in a U-bottom plate such as the hanging drop spheroid (22, 37–41). In these techniques, cell aggregation can be facilitated using low or ultra-low attachment plates coated to prevent cell adherence to the plastic surface. This coating promotes cells to interact with each other, forming small spherical structures (42–51). Additionally, 1% of agarose can be added to the plate bottom to maintain the stemness of cells when necessary, as seen in organoid formation using stem-cell-derived beta cells using plate shaking techniques (37, 38, 52),. Alternatively, direct spheroid formation can be fostered by adding specific molecules and factors to mimic the TME and promote cell growth and aggregation. For instance, for brain cancer and glioblastoma models, epidermal growth factor (EGF), basic fibroblast growth factor (bFGF) or insulin-like growth factor (IGF) are used along with neurobasal medium supplemented with B27 or N2 to replicate brain tumor signaling and support neural cell growth (Figure 5; 51, 53),. For CAR-T cell studies, CAR-T cells are subsequently introduced into the established 3D models using different strategies, depending on the research goals. For instance, to study CAR-T cell-mediated cytotoxicity, CAR-T cells can be added directly with tumor cells for co-culture into the spheroid during the initial aggregation phase. Alternatively, to investigate infiltration dynamics, CAR-T cells can be introduced after spheroid formation.

The second method for producing spheroids or organoids is to use a matrix as a scaffold (Figure 6). There are several matrices available such as matrigel (54–63), hydrogel (64–73), and collagen (74–76) from rat or mouse tails. In this model, cells are cultured in the matrix as in 3D space instead of monolayer. Another kind of matrix is the ECM derived from tissues and organs. ECM can provide a more natural microenvironment compared to synthetic hydrogels, such as polyethylene glycol (PEG) or poly-lactic-co-glycolic acid (PLGA), which facilitate the creation of more physiologically and biochemically relevant models for spheroid and organoid formation (54, 60, 66, 69), (73), (75),,– (77). Another way is to produce the spheroid using a low-attachment plate before transferring it into a gel bubble to let it grow in 3D (55, 56),. CAR-T cells can be directly added into pre-formed spheroids or organoids in the matrix or into the culture medium surrounding the spheroid/organoid (65, 67),. To investigate the early interactions with tumor cell aggregates, or CAR-T activation and tumor killing (72), CAR-T cells can be co-cultured before or during the formation of the spheroids/organoids. On the other hand, to evaluate CAR-T infiltration, migration as well as killing tumor cells, CAR-T cells can also be added after the spheroid/organoid formation as a well-established tumor structure. This method mimics the conditions of treating an existing tumor rather than targeting cells in the early stages of aggregation.

The third method for creating spheroids and organoids involves microfluidic, organs-on-chip, or bioprinting technologies (Figure 6). Several devices and chips have been engineered to deliver oxygen or nutrients, regulate and study chemotaxis (78, 79), mimic blood vessels and vascularization (80), remove dead cells and deliver CAR-T cells using microdroplets (81) or microhydrogels (45, 82, 83),. In these approaches, spheroids, organoids, or biopsy-derived cells are cultured on a matrix and connected to microfluidic systems (78, 84). Similar to microfluidic systems, organ-on-chip provides precise regulation of biochemical and physical microenvironments, enabling accurate modeling of tumor microenvironment complexity and CAR-T cell behavior. Additionally, organ-on-chip systems simulate blood flow to allow the investigation of CAR-T cell infiltration and interactions with endothelial barriers, which is critical for studying solid tumors (82, 84–86). Using multiple channels, these devices allow simultaneous testing on various conditions, such as cytokine gradients or different CAR construct injections (86). Most recently, bioprinting has emerged as a cutting-edge technology offering precise tissue construction (87). Thus, these technologies facilitate the 3D dynamic studies of CAR-T cell migration, proliferation, and cytotoxicity under well-controlled parameters, giving kinetic results that static cultures cannot achieve.

The fourth method is patient-derived organotypic tumor or tumor slice culture assays (Figure 7). This technique is related to ex vivo studies, where human tumors are extracted from the patient and cultured on a matrix (21, 88–94). Depending on the size of the tumor and the microenvironment components extracted, the tumor can be cultured whole or sliced into thin sections (93). These cultures can also be connected to a microfluidic system to supply medium, and nutrients or to facilitate CAR-T cell injection (95). Alternatively, CAR-T cells can be introduced using similar methods to those described in the second method of organoid manufacturing.

As with 2D analysis, several methods can be used to study CAR-T cell behavior in a 3D co-culture system. Indeed, confocal microscopy and live-cell imaging are used to track CAR-T activity such as the infiltration into the spheroid/organoid, the spatial distribution of tumor and CAR-T cells, CAR-T migration, and cytotoxicity. To track CAR-T and tumor cells in the spheroid/organoid, cells are labelled with a fluorescent molecule or a reporter system. Flow cytometry can also be used after the dissociation of the co-culture system into single-cell suspensions to further analyze CAR-T cell activity, cytotoxicity, persistence, expansion and proliferation, differentiation phenotype, and exhaustion. Coupled with flow cytometry, ELISAs can be employed to analyze release cytokine production for CAR-T cells such as TFN-α, IFNγ, and IL-2. Viability assays using LDH release assays or live-dead staining can be used to quantify tumor cell damage or death before and after CAR-T treatment. Finally, compared to 2D co-culture systems, histological analysis can be applied to fixed spheroids/organoids using immunohistochemistry (IHC) or immunofluorescence to detect tumor killing or apoptosis, CAR-T cell infiltration and phenotype (97).