OR04-08 Ligand Class Analysis Enables Predictive Models for Effective Antagonism of Endocrine Resistant Breast Cancer

Abstract Disclosure: J.C. Nwachukwu: None. C.K. Min: None. R.R. Kobylski: None. T.J. Kim: None. Y. Hou: None. S. Kim: None. G.R. Hancock: None. T. Izard: None. S.W. Fanning: None. B.S. Katzenellenbogen: None. J.A. Katzenellenbogen: None. K.W. Nettles: None. Drugs targeting the estrogen receptor-α (ER) fall into two chemical classes, selective estrogen receptor modulators and degraders (SERMs and SERDs), which are effective treatments for hormone-responsive breast cancer. Compounds in these chemical classes typically show similar optimized efficacy in pre-clinical models of breast cancer, including in the context of ER-activating mutations (e.g. ER Y537S) and other resistance pathways, limiting our ability to identify small differences that may be more impactful clinically. To understand structural and molecular drivers of ER antagonism, we expanded the repertoire of ER targeting strategies. We characterized 108 compounds derived from a high-affinity adamantyl scaffold, incorporating 8 distinct classes of pharmacophores, each featuring a diverse array of chemical groups. The chemically diverse ligands revealed that non-canonical targeting approaches produced complex structure activity relationships across a panel of hormone sensitive and resistant breast cancer cell lines and reporter systems, activity that was often more efficacious than existing ER-targeted therapies. We obtained X-ray crystal structures of 25 adamantyl ligands bound to ER resistance mutants. This revealed how the different ligand chemical groups generated diverse structural effects on the surface binding site for transcriptional regulatory proteins that regulate gene expression. Molecular dynamics simulations showed that ER Y537S and D538G stabilized different antagonist conformers, or substates, from the wild type receptor, suggesting a structural basis for resistance to ER antagonists. We selected a subset of 11 adamantyls for a computational approach called “ligand class analysis” (LCA). LCA leverages the wide range of growth inhibitory effects of these related compounds (i.e. a compound class) to provide the statistical robustness for machine learning the underlying mechanisms of action. RNA-seq from MCF-7 cells expressing the ER Y537S resistance allele revealed ∼1600 genes differentially regulated by the 11 adamantyl ligands. We used the growth inhibitory effect of each ligand as the dependent variable in machine learning and identified a gene set response pattern with >95% predictive power for ligand efficacy. Defining causal links from ligand to receptor structure and the cellular mediators of growth inhibition revealed basic principles of allosteric signaling and transcriptional regulation. LCA differentiated these outcome-focused ER signaling pathways from the full set of ligand-regulated cellular effects, while diversification of the ligand-receptor structures identified new ways of modulating ER activity for targeting endocrine resistant breast cancer. Supported by NIH R01CA275142, R01CA220284, R37CA27934, and the Breast Cancer Research Foundation BCRF-083 and BCRF-084 Presentation: Saturday, July 12, 2025