405 Background: Globally, gastrointestinal (GI) cancers, which account for around 28% of all cancers, has increased constantly at a CAGR of 2% (2014-2019), while the treatment options remain limited and often ineffective. This is largely due to late diagnosis and lack of new tumor associated antigens for most of the GI malignancies. Cadherin-17, CDH17 (aka CA17), is being evaluated as a potential therapeutic target and molecular marker for GI cancers, of which aberrant expression of CDH17 was reported in 50-90% in tumor samples (including stomach, pancreas, liver/bile duct and colon) by multiple groups. The present study aims to develop a humanized CDH17-CD3 bispecific T-cell engager antibody, ARB202, to treat advanced pancreatic and metastatic colon cancers. Methods: The ARB202 bispecific was generated from anti-CDH17 monoclonal ARB102 as previously described (Lum Y et al, Mol. Pharm. 2020) by linking a CD3-binding scFv in the format of IgG4-scFv. Biophysical properties, efficacy and safety were evaluated. Efficacy was addressed in cytotoxicity assays, IL2-release assays and subcutaneous xenograft models using multiple GI tumor target lines. Preclinical safety profiles include Tissue cross-reactivity assays and GLP repeat-dose toxicity studies in NHP. Results: The ARB202 bispecific antibody was efficiently produced (yield ~1 g/L) and remained stable for at least 3 months under accelerated stress tests. Binding affinity (Kd [M]) of ARB202 to CDH17 and CD3 was at 10-10 and 10-9 respectively, with cellular affinities of 10-9 and 10-8. In vitro cytotoxicity assays demonstrate that ARB202-mediated tumor cell killing is CDH17-dependent and requires immune synapse formation with effector T cells. By contrast, normal cells and CDH17-negative cancer cells are resistant. Pancreatic AsPC1 xenograft model revealed an ARB202 (co-inoculated with human PBMC) dose-dependent inhibition of tumor growth. Human IL-2 release in mice blood samples was consistent with functional CDH17 and CD3 engagement. ARB202 PK profile in rodents and NHP was found to be excellent relative to other bispecifics and GLP toxicity studies showed no significant findings nor adverse effects. Conclusions: ARB202 bispecific antibody promises to be a first-in-class biologic, which may tackle approximately 50% of GI cancers that which have been reported CDH17 positive. First-in-human clinical study with advanced GI cancers is scheduled to start in Q1 2021.