Normalized internalization predicts potency of carcinoembryonic antigen-related cell adhesion molecule 6-targeting antibody-drug conjugates.

Antibody-drug conjugate (ADC) efficacy in solid tumors frequently fails to track with target abundance or binding affinity, reflecting delivery bottlenecks that remain difficult to quantify during lead selection. Here, we develop a discovery-to-function workflow for carcinoembryonic antigen-related cell adhesion molecule 6 (CEACAM6), a glycosylphosphatidylinositol (GPI)-anchored antigen with context-dependent trafficking and show that intracellular delivery-related parameters can better organize ADC potency than antigen abundance or binding affinity alone in the CEACAM6 model system. Using multiplexed single-cell screening, we generated a mechanistically diverse anti-CEACAM6 antibody panel spanning CEACAM-family specificity, domain-level epitope usage, kinetic interaction modes, and uptake behavior. Three leads were humanized while preserving specificity and epitope identity and were conjugated to MMAE using a cleavable MC-VC-PAB linker with comparable drug loading across constructs. Despite target-dependent cytotoxicity, ADC potency and selectivity varied markedly by clone and cellular background and could not be consistently predicted by antigen expression or binding affinity alone. We defined a unitless normalized internalization degree (NID), a composite ranking metric derived from conventional surface binding and internalization assays, to integrate surface engagement with internalization efficiency. NID was associated with potency across CEACAM-expressing contexts and provided a delivery-integrated descriptor distinct from conventional internalized-fluorescence amount metrics. This delivery-related interpretation was supported by pHrodo-based tracking of accumulation in acidic compartments relevant to linker processing. The CEACAM6-selective lead h9738, prioritized based on NID profiling, exhibited robust, dose-dependent antitumor activity in a Capan-1 pancreatic cancer xenograft model, achieving complete responses (CRs) at higher doses. Collectively, these findings support a delivery-aware framework for CEACAM6 ADC optimization and identify NID as a practical, context-sensitive ranking metric for candidate prioritization for delivery-limited, GPI-anchored targets such as CEACAM6.