Preclinical evaluation of co‑stimulatory T‑Cell Engagers in IO: from mechanism to translational models

Bispecific T‑cell engagers (BiTEs or TCEs) have emerged as a powerful immunotherapy strategy, redirecting cytotoxic T cells toward tumor cells through simultaneous engagement of CD3 and tumor‑associated antigens. Their clinical impact is well established for hematologic cancers, with several TCEs now approved and many more advancing through clinical development. As experience with these first‑generation molecules has grown, opportunities to further enhance their efficacy, durability, and safety, particularly in solid tumors, have become apparent. Classical TCEs rely on CD3‑mediated activation, which can limit sustained T‑cell function and contribute to systemic immune activation. These insights have driven the development of next‑generation TCEs that incorporate physiological co‑stimulation, building on the strengths of existing therapies to enable more complete and durable T‑cell responses.

Co-stimulatory TCE design and mode of action

First‑generation BiTEs provide only signal 1 through CD3 engagement. However, delivery of the co‑stimulatory signal 2 is required for robust activation, proliferation, persistence, and durable anti-tumor responses. As a result, these TCEs can drive suboptimal and transient T‑cell responses that lead to exhaustion, while the CD3 activation can trigger systemic toxicities such as cytokine release syndrome (CRS). Consequently, next‑generation TCEs aim to incorporate physiological co‑stimulation of T cell receptors to achieve a more complete and durable T‑cell activation^1,2.

Co‑stimulation delivers several distinct signals, collectively referred to as signal 2, that ensure full T‑cell activation. CD28 provides the canonical early co‑stimulatory signal, CD137 (4‑1BB) promotes survival and memory formation, and CD2 enhances adhesion and sustained activation. The absence of signal 2 renders traditional BiTEs susceptible to immunosuppressive cues that dominate the tumor microenvironment, particularly in solid tumors characterized by poor T‑cell infiltration and chronic antigen stimulation. Incorporating co‑stimulatory pathways directly into the TCE architecture offers a rational strategy to overcome these barriers^3,4.

The resulting multispecific TCEs integrate tumor antigen recognition, CD3 activation, and co‑stimulatory engagement within a single molecular scaffold. This design provides spatially and temporally aligned signaling that enhances T‑cell expansion, improves persistence, and sustains effector function⁵. CD28‑ or CD137‑containing formats have shown superior cytotoxicity and improved tumor control in preclinical models, outperforming classical BiTEs even in tumors with hostile microenvironments. Providing activation and co‑stimulatory signals together in a single TCE improves T‑cell infiltration, as illustrated by DLL3×CD137 TCEs in small‑cell lung cancer models⁶.

Beyond improving effectiveness, integrated co‑stimulation offers a safety advantage. Requiring simultaneous engagement of the TAA, CD3, and the co‑stimulatory receptor restricts activation to the tumor site, thereby reducing systemic cytokine release and widening the therapeutic window. This tumor‑restricted delivery of signal 2 stands in contrast to traditional BiTEs, where CD3 activation can occur independently of the tumor when the TAA is expressed in other tissues (on-target off-tumor toxicity), generating dose‑limiting toxicity.

Preclinical mouse models enabling the evaluation of multispecific TCEs

Evaluating co‑stimulatory TCEs in preclinical settings presents different challenges. These molecules require the simultaneous engagement of human CD3 and human co‑stimulatory receptors such as CD28 or CD137, which may not be cross‑reactive with their murine counterparts. As a result, standard CD3‑humanized models cannot reproduce the full signaling cascade triggered by next‑generation TCEs, capturing only the CD3 activation arm of the response. This limits the ability to assess the effectiveness of coordinated signal‑1 and signal‑2 delivery, consequently affecting the durability of the T‑cell response, memory formation, the risk of off‑tumor activation, and cytokine release patterns that inform safety and dosing strategies.

To test these next-gen TCEs, two types of models can be employed: syngeneic mouse models or humanized mouse models containing a human immune system. Syngeneic mouse models, in which murine tumors are implanted into immunocompetent mice, are valuable for studying general immune dynamics, tumor growth inhibition, and interactions within an intact mouse immune system and a fully functional tumor-stroma crosstalk. Syngeneic mouse models with knock-in humanized targets such as CD3 and a co-stimulator such as CD28 or CD137 (4-1BB) are therefore ideal models for testing co-stimulatory surrogate TCEs. Additionally, these models can also be implanted with murine tumor cells expressing a human TAA, enabling the assessment of clinical TCEs targeting a human TAA, human CD3 and a human co-stimulatory receptor. As a result, genOway has developed double‑humanized mouse models such as genO-panhCD3/hCD28 and genO-panhCD3/hCD137, which recreate the receptor context needed for accurate cross-linking, signaling, and cytokine responses. These models display physiological expressions of their humanized targets, and maintain a functional CD3-TCR interaction as well as functional T and B cell cooperation.

As an alternative, models with a human-like immune system can also be used to evaluate next-gen TCEs. The genO-BRGSF-HIS is a mouse model reconstituted with CD34⁺ human hematopoietic stem cells, enabling the development of a functional human immune system comprising both lymphoid and myeloid compartments. This means that its T cells are of human origin, and consequently express human T cell receptors, such as CD3, CD28, 4-1BB, OX40, among others. Additionally, this model can be engrafted with tissue from cancer patients (patient derived xenograft, PDX models) or cell line-derived xenografts (CDX), enabling the study of clinically relevant constructs in the context of a human immune response. Beyond enabling the efficacy assessment of new TCEs, this model also enables the safety assessment of these constructs, thanks to the presence of human myeloid cells^7,8. This is a particular advantage over PBMC-reconstituted models, which quickly develop graft-versus-host disease (GvHD), resulting in short therapeutic windows, and which exhibit biased cytokine release syndrome (CRS) responses due to the absence of human myeloid cells. Despite its advantages, it is important to note that the stroma in this model remains of murine origin and consequently it might not fully capture the tumor-immune cell-stroma interactions.

Together, these models provide a translationally relevant and mechanistically accurate foundation for de‑risking multispecific TCE programs.  

‍

References:

1. Radtke, K. et al. Cytokine Release Syndrome: Trends with T Cell Engaging Bispecifics from a Systematic Review of Licensing Applications. Blood 144, 5798–5798 (2024).

2. Muth, A. et al. T cell exhaustion in bi- and trispecific T cell engager therapy in hematologic malignancies: Mechanisms and implications. Med 101031 (2026) doi:10.1016/j.medj.2026.101031.

3. Velasquez, M. P. et al. CD28 and 41BB Costimulation Enhances the Effector Function of CD19-Specific Engager T Cells. CancerImmunol. Res. 5, 860–870 (2017).

4. Passariello, M. et al. Tri-specific tribodies targeting 5T4, CD3, and immune checkpoint drive stronger functional T-cell responses than combinations of antibody therapeutics. Cell Death Discov.11, 58 (2025).

5. Sun, Y. et al. Leveraging T cell co-stimulation for enhanced therapeutic efficacy of trispecific antibodies targeting prostate cancer. J. Immunother. Cancer 13, e010140 (2025).

6. Mikami, H. et al. Engineering CD3/CD137 Dual Specificity into a DLL3-Targeted T-Cell Engager Enhances T-Cell Infiltration and Efficacy against Small-Cell Lung Cancer. Cancer Immunol. Res. 12,719–730 (2024).

7. Martin, G. et al. Myeloid and dendritic cells enhance therapeutics-induced cytokine release syndrome features in humanized BRGSF-HIS preclinical model. Front. Immunol. 15, 1357716 (2024).

8. Martin, G. H. et al. Tumor-dependent myeloid and lymphoid cell recruitment in genO-BRGSF-HIS mice: a novel tool for evaluating immunotherapies. Front. Immunol. 16, 1624724 (2025).

9. Garfall, A. L. & June, C. H. Trispecific antibodies offer a third way forward for anticancer immunotherapy. Nature 575, 450–451 (2019).

‍