P-02.04 Developing T cell bispecific antibodies with reduced-affinity anti-CD3 binders for a potent, efficacious, and low-cytokine response

<h3>Background</h3> T cell bispecifics (TCBs) were conventionally considered to require high affinities for both CD3 and tumor target binders to activate T cells and achieve efficient tumor cell killing. Based on this approach, TCBs have successfully achieved efficacy in treating patients with hematological and more recently solid tumors, but their mode-of-action has been associated with cytokine release which has been dose-limiting in some cases. Recent studies suggested that lowering CD3 binder affinity may result in effective tumor killing with reduced cytokine levels. However, the possible interdependence of CD3 affinity with additional parameters to achieve this outcome still remains untackled. This study aims to understand the impact of the CD3 and tumor-target binder affinity on differential T cell-dependent cytotoxicity and cytokine secretion. Therefore, a series of TCBs with different affinities for CD3, along with two target binders of differing affinities (high and low) targeting FOLR1 and CEACAM5, were developed in a 2:1 TCB format. This configuration allows for bivalent, preferential binding to the tumor antigen and monovalent binding to CD3, providing a distinct opportunity to enhance the activity of conventional TCBs. <h3>Materials and Methods</h3> In vitro, 2:1 TCBs were tested using human peripheral blood mononuclear cells (PBMCs) and human tumor cell lines to evaluate T cell-mediated cytotoxicity and cytokine release. In vivo, we investigated the impact our TCBs have on efficacy, cytokine release, and pharmacokinetic properties in stem cell-engrafted humanized mouse models. <h3>Results</h3> CD3 binders were categorized into high, intermediate, low, and very low affinity groups based on SPR analysis and functional assays. CD3intermed produced lower cytokine levels compared to CD3high while maintaining T cell mediated cytotoxicity regardless of the target. Additionally, CD3low exhibited a favorable profile in comparison to CD3high only when coupled with a high-affinity target binder, resulting in even lower cytokine levels than those induced by CD3intermed. By utilizing advanced humanized mouse models, we demonstrated the in vitro to in vivo translatability of our findings across targets using CD3 affinity-attenuated binders, as opposed to the typical use of PBMC-engraftment models in published studies. Lastly, we observed that CD3 affinity does not impact PK properties in vivo. <h3>Conclusions</h3> We confirmed that reducing CD3 affinity in our 2:1 TCB platform can maintain anti-tumor efficacy while lowering cytokine release. Most importantly, the selection of the target binder affinity and the appropriate preclinical models are crucial for determining the optimal CD3 affinity threshold. <b>O. Abdelmotaleb:</b> A. Employment (full or part-time); Significant; Roche. E. Ownership Interest (stock, stock options, patent or other intellectual property); Significant; Roche. <b>C. Klein:</b> None. <b>I. Waldhauer:</b> None. <b>J. Sam:</b> None. <b>T. Hofer:</b> None. <b>M. Lechmann:</b> None. <b>A. Freimoser-Grundschober:</b> None. <b>M. Eigenmann:</b> None. <b>A. Bransi:</b> None. <b>A. Odermatt:</b> None. <b>P. Brünker:</b> None. <b>S. Colombetti:</b> None. <b>A. Schneider:</b> None.