Translational safety assessment of T cell engagers: the critical role of humanized mouse models

T cell engagers (TCEs), including bispecific antibodies, have emerged as a powerful class of immunotherapies capable of redirecting cytotoxic T cells toward tumor cells through simultaneous engagement of CD3 and a tumor-associated antigen (TAA). This mechanism enables potent, MHC-independent tumor killing and has translated into multiple clinical approvals, particularly in hematological malignancies (Amoozgar et al., 2025; Radtke et al., 2024). However, the same mechanism that underpins their efficacy, robust T cell activation, also drives significant safety liabilities. As TCEs continue to expand into solid tumors and next-generation multispecific formats, ensuring an accurate assessment of safety early in development has become essential.

Safety liabilities of TCEs

A defining feature of TCE-associated toxicity is the induction of cytokine release syndrome (CRS), an acute systemic inflammatory response triggered by widespread immune activation. CRS results from excessive cytokine secretion following CD3-mediated T cell engagement and can manifest as fever, hypotension, hypoxia, and, in severe cases, multi-organ dysfunction (Radtke et al., 2024). Importantly, CRS is not solely driven by T cells but is amplified by myeloid and dendritic cells, highlighting the complexity of immune interactions underlying toxicity (Godbersen-Palmer et al., 2020). In addition to CRS, immune effector cell-associated neurotoxicity syndrome (ICANS) represents another major safety concern, associated with blood-brain barrier disruption and cytokine-driven neuroinflammation. Other adverse events, including infections, cytopenias, and on-target off-tumor toxicity, further complicate the clinical development of TCEs (Géraud et al., 2024). Together, these liabilities underscore the need for predictive preclinical systems capable of capturing both the magnitude and the mechanisms of immune-mediated toxicity.

Limitations of conventional preclinical models for TCE safety assessment

Despite this need, traditional preclinical models remain poorly adapted to TCE evaluation. Wild-type mice are inherently unsuitable for testing human CD3-targeting therapies due to species-specific differences in CD3 structure and function, which limit antibody cross-reactivity and prevent accurate modelling of human T cell activation. As a result, these models fail to reproduce key pharmacological processes such as cytokine release, leading to misleading safety and efficacy readouts. Furthermore, commonly used PBMC-reconstituted models introduce additional limitations, including xenogeneic graft-versus-host disease (GvHD) leading to a short therapeutic window, non-physiological immune activation driven by xenogeneic T cell responses against mouse tissues, leading to systemic, antigen-independent cytokine release, a lack of human myeloid cells, and variability depending on donor characteristics, which can obscure drug-specific effects. These limitations highlight a critical gap in the translational pipeline: the absence of models that faithfully recapitulate human CD3 biology within a functional immune environment.

To address these challenges, mouse models with a human immune system (HIS) and knock-in models with humanized genes, have become indispensable tools for the preclinical evaluation of TCEs. By enabling the engagement of human CD3 and, in some cases, additional human immune targets, these models allow mechanistic investigation of T cell activation, cytokine release, and dose-dependent toxicity. Notably, these models have demonstrated the ability to simultaneously assess efficacy and cytokine-driven toxicity, thereby providing a more integrated view of the therapeutic index (Yang et al., 2023). However, the predictive value of these systems depends critically on the degree of humanization and the preservation of physiological immune interactions, particularly those involving myeloid compartments that contribute to CRS.

genOway models for mechanistic and predictive TCE safety assessment

The development of TCEs requires a careful balance between potent immune activation and acceptable safety profiles. Given the central role of CD3 engagement in both efficacy and toxicity, preclinical models must accurately reproduce human T cell biology and the downstream immune cascades that drive adverse events. In this context, genOway has developed a portfolio of mouse models specifically designed to overcome the translational limitations of conventional systems. The genO-panhCD3 model constitutes a cornerstone for TCE safety assessment by expressing a humanized CD3 complex (ε, γ, δ) under physiological regulation. This enables endogenous T cell development, preserves native CD3/TCR signaling, and supports dose-dependent T cell activation and cytokine release, thereby providing a robust platform for evaluating CRS risk in vivo and enabling the ranking of new TCEs (Mhyre et al., AACR 2026). Complementing this model, the genO-panhCD3/hCD28 and genO-panhCD3/hCD137 (4-1BB) models allow the evaluation of next-generation TCEs incorporating co-stimulatory signals, which are increasingly used to enhance efficacy while modulating safety profiles. These dual-humanized models recreate the coordinated delivery of signal 1 (CD3) and signal 2 (CD28), enabling a more accurate assessment of both T cell activation dynamics and cytokine responses (Sônego et al, AACR 2024, Martin-Jeantet et al., SITC 2022).

Beyond receptor humanization, models integrating a human immune system (HIS) provide additional insights into complex safety mechanisms. The genO-BRGSF-HIS model, reconstituted with human CD34⁺ hematopoietic stem cells, generates functional human lymphoid and myeloid compartments. This feature is particularly critical for modelling CRS, as it enables the study of immune crosstalk and cytokine amplification driven by T and non-T cell populations (Martin et al, 2024).

By combining precise genetic humanization with functional immune reconstruction, platforms such as those developed by genOway enable a more predictive evaluation of TCE safety, ultimately accelerating the development of safer and more effective immunotherapies.

References

• Amoozgar, B. et al. (2025). From Molecular Precision to Clinical Practice: A Comprehensive Review of Bispecific and Trispecific Antibodies in Hematologic Malignancies. International Journal of Molecular Sciences, 26(11), 5319. DOI: https://doi.org/10.3390/ijms26115319.
• Géraud, A. et al. (2024). Reactions and adverse events induced by T-cell engagers as anti-cancer immunotherapies, a comprehensive review. European Journal of Cancer, 205, 114075. DOI: 10.1016/j.ejca.2024.114075
• Godbersen-Palmer, C. et al. (2020). Toxicity induced by a bispecific T cell redirecting protein is mediated by both T cells and myeloid cells in immunocompetent mice. Journal of Immunology, 204(11), 2973–2983. DOI: https://doi.org/10.4049/jimmunol.1901401.
• Martin et al. (2024). Myeloid and dendritic cells enhance therapeutics-induced cytokine release syndrome features in humanized BRGSF-HIS preclinical model. Frontiers in Immunology, 15:1357716. DOI: 10.3389/fimmu.2024.1357716.
• Radtke, K.K. et al. (2024). Clinical Pharmacology of Cytokine Release Syndrome with T-Cell–Engaging Bispecific Antibodies: Current Insights and Drug Development Strategies. Clinical Cancer Research, 31(2), 245–257. DOI: https://doi.org/10.1158/1078-0432.CCR-24-2247.
• Yang, J. et al. (2023). Simultaneous evaluation of treatment efficacy and toxicity for bispecific T-cell engager therapeutics in a humanized mouse model. FASEB Journal, 37(6), e22995. DOI: https://doi.org/10.1096/fj.202300040R.