Resolving specificity over a large footprint distinguishes androgen and glucocorticoid receptor DNA binding

Abstract

The glucocorticoid (GR) and androgen receptors (AR) share an identical DNA binding interface, bind a similar core motif, yet occupy largely distinct genomic loci to regulate different gene programs. Their interplay is particularly important in the context of some castration resistant prostate cancers (CRPC), where GR can functionally substitute AR to drive the growth of tumor in the absence of androgen signaling. Some differences in the genomic localization of AR and GR can be explained by selective association with binding partners, however it is not clear how they discriminate the remaining sites. Examination of the crystal structures suggested that specificity might extend over a larger footprint (≥20bp) which had not been accessible to experimental and computational techniques previously. We therefore developed a refined version of SELEX‐seq coupled to an iterative algorithm, SelexGLM, to model the free energy of binding over a ~30bp binding site. The position specific affinity matrices (PSAMs) for these two factors reveled difference both within and outside the 15bp core motif, with AR being less tolerant to changes and thus more specific. Part of this specificity was derived from readout not just of specific base pairs, but also the shape of DNA. AR also exhibited preference for A‐tracts flanking the core sequence that narrow minor groove width. In contrast, GR specificity was largely restricted to the core sequence, with preference for a wider minor groove within that motif. The enhanced specificity of AR appears to be intrinsic to the protein, and thermodynamically driven. Isothermal titration calorimetry revealed that AR DNA‐binding is more enthalpically driven than GR binding, which is more entropically driven. This demonstrates, on a comprehensive scale, that, much like direct readout, DNA shape driven binding is enthalpically driven, and that entropically driven binding is associated with decreased specificity. Thus, this work provides a biophysical rationale for how GR can functionally substitute for AR in some castration resistant prostate cancers: in the absence of AR, overexpression of the less selective GR is able to bind AR sites. Further, this work demonstrates that combining high‐throughput and classical biophysical approaches provides the deep understanding of specificity that is necessary to dissect the regulation of fine‐tuned systems.Support or Funding InformationAcknowledgments: This research was supported by the National Institutes of Health (NIH) K99/R00 CA149088, the Roy J. Carver Charitable Trust 01–224, and the National Science Foundation CAREER grant 1552862 (MAP), and grant R01HG003008 (HJB).