Abstract
The nonclinical development of T-cell receptor (TCR)-based therapeutics is uniquely challenged by the human-exclusive specificity of the peptide-HLA target, rendering conventional animal models unsuitable for safety assessment. To support first-in-human (FIH) studies in patients with solid tumors with the novel TCR-like T Cell Bispecific (TCB) antibody MAGE-A4-TCB, targeting the MAGE-A4 peptide presented by HLA-A*02:01 and engaging T-cells via the CD3ϵ chain, we implemented a comprehensive, prediction-independent safety strategy relying on New Approach Methodologies (NAMs). This approach utilized 21 human cellular systems, including 2D and 3D in vitro models of vital organs (such as liver, heart, kidney, and lung), co-cultured with allogeneic peripheral blood mononuclear cells to identify potential off-target toxicity and a whole blood assay for cytokine release risk assessment. The concentration-effect relationship and Minimal Anticipated Biological Effect Level (MABEL) were established using the A375 cancer cell line, chosen for its comparable sensitivity to ovarian carcinoma organoids, which were the most sensitive out of 22 tested tumor organoids. Nonclinical testing demonstrated an absence of off-target reactivity in vital organs, though the whole blood assay predicted cytokine release syndrome as the primary risk. The calculated FIH starting dose was safe in the subsequent Phase I study (NCT05129280), and the most commonly reported adverse events were treatment-related skin and subcutaneous tissue disorders (73% of patients). These skin-related adverse events, which appeared in contradiction to the initial in vitro results in skin co-culture models, were hypothesized to be driven by the target-independent activation capability of the UCHT1-based CD3 binder, potentially enhanced in patients with a systemic pro-inflammatory profile. This study provides a translational framework for TCR-based immunotherapies, demonstrating the successful application of NAMs to define a safe starting dose while highlighting the critical need for refining these methodologies to accurately predict complex, systemic immune-related adverse events, particularly those involving barrier organs.