Abstract 1809: In vitro and in vivo characterization of human CD3EDG triple knock-in mice (CD3EDG HuGEMM™) for CD3 bispecific antibody evaluation

Introduction: At present, there are >110 bispecific antibodies in clinical development, and >180 in preclinical development, among which CD3-based bispecific antibodies are the main focus. There are four CD3 subtypes that make up the TCR, including CD3ϵ, CD3δ, CD3γ and CD3ζ, with CD3δ/CD3ϵ or CD3γ/CD3ϵ heterodimers that bind to α/β chain of TCRs. CD3 bispecifics usually recognize the CD3ϵ subunit of the heterodimers and activate T cells to promote tumor killing. However, during assembly of the TCR heterodimers, CD3ϵ/CD3δ and CD3ϵ/CD3γ are not necessarily formed at 1:1 ratios, which may lead to big differences among different batches of anti-CD3 antibodies due to conformational variation of the heterodimer proteins. This could therefore be a challenge when developing CD3 bispecifics. To better understand and recapitulate T cell development and signaling via targeting human CD3 in vivo, we developed a fully humanized CD3EDG model (CD3EDG HuGEMM) expressing human CD3ϵ, δ, and γ proteins, to evaluate CD3 bispecifics. Methods: CD3EDG HuGEMM was developed based on three-step BAC transgenes via ES cell targeting, in which the mouse Cd3e/Cd3d/Cd3g tandem on the same allele of chromosome 9 was completely replaced by human CD3E/CD3D/CD3G tandem, making it feasible to generate both heterozygous or homozygous CD3EDG HuGEMM mice via cross-breeding. The integration of human CD3E/CD3D/CD3G was confirmed by PCR-based genotyping and sequencing as well as southern blotting. The expression of human CD3E/CD3D/CD3G proteins was confirmed by FACS. In addition, we have engineered syngeneic tumor cell lines to express human tumor-associated antigens (hTAA) or other human targets (HuCELL™). Results: We first characterized T/B/NK cell lineage development in the heterozygous and homozygous CD3EDG mice, which was proportionately normal compared to wild type mice. We then performed binding assays for 4 different anti-hCD3 antibodies via FACS, leveraging the CD3EDG HuGEMM-derived splenocytes and human PBMC as control. We found differential binding between anti-hCD3 antibodies and human CD4+ or CD8+ T cells. In light of this finding, we further investigated in vitro T cell activation of cells treated with these 4 antibodies via FACS and ELISA, and found varying levels of cytokine release (IFN-γ, TNF-α, IL-6, IL-10 and IL-2). Finally, MC38-hEpCAM syngeneic tumors in the CD3EDG HuGEMM mice were treated with different anti-hEpCAM CD3 bispecifics and data will be reported. Conclusions: Our