High-throughput AR dimerization assay identifies androgen disrupting chemicals and metabolites.

Analysis of streamlined computational models used to predict androgen disrupting chemicals revealed that assays measuring androgen receptor (AR) cofactor recruitment/dimerization were particularly indispensable to high predictivity, especially for AR antagonists. As the original dimerization assays used to develop the minimal assay models are no longer available, new assays must be established and evaluated as suitable alternatives to assess chemicals beyond the original 1,800+ supported by the current data. Here we present the AR2 assay, which is a stable, cell-based method that uses an enzyme complementation approach. Bipartite domains of the NanoLuc luciferase enzyme were fused to the human AR to quantitatively measure ligand-dependent AR homodimerization. 128 chemicals with known endocrine activity profiles including 43 AR reference chemicals were screened in agonist and antagonist modes and compared to the legacy assays. Test chemicals were rescreened in both modes using a retrofit method to incorporate robust cytochrome P450 (CYP) metabolism to assess CYP-mediated shifts in bioactivity. The AR2 assay is amenable to high-throughput screening with excellent robust Z'-factors (rZ') for both agonist (0.94) and antagonist (0.85) modes. The AR2 assay successfully classified known agonists (balanced accuracy = 0.92) and antagonists (balanced accuracy = 0.79-0.88) as well as or better than the legacy assays with equal or higher estimated potencies. The subsequent reevaluation of the 128 chemicals tested in the presence of individual human CYP enzymes changed the activity calls for five compounds and shifted the estimated potencies for several others. This study shows the AR2 assay is well suited to replace the previous AR dimerization assays in a revised computational model to predict AR bioactivity for parent chemicals and their metabolites.