Steroid receptor-DNA interactions: toward a quantitative connection between energetics and transcriptional regulation

Abstract

Steroid receptors comprise an evolutionarily conserved family of transcription factors. Although the qualitative aspects by which individual receptors regulate transcription are well understood, a quantitative perspective is less clear. This is primarily because receptor function is considerably more complex than that of classical regulatory factors such as phage or bacterial repressors. Here we discuss recent advances in placing receptor-specific transcriptional regulation on a more quantitative footing, specifically focusing on the role of macromolecular interaction energetics. We first highlight limitations and challenges associated with traditional approaches for assessing the role of energetics (more specifically, binding affinity) with functional outcomes such as transcriptional activation. We next demonstrate how rigorous in vitro measurements and straightforward interaction models quantitatively relate energetics to transcriptional activity within the cell, and follow by discussing why such an approach is unexpectedly effective in explaining complex functional behavior. Finally, we examine the implications of these findings for considering the unique gene regulatory properties of the individual receptors.

Figure 1.

Schematic of steroid receptor domain structure and phylogenetic tree. (A) Schematic of steroid receptor domains and number of amino acids. Functional domains as labeled: DBD, DNA binding domain; LBD, ligand-binding domain; AF, activation functions are present in both the N-terminal region and LBD. (B) Phylogenetic tree representing divergence of the steroid receptor family. Filled circle represents the ER-like common ancestor for subfamily 3A (ER-α and ER-β) and subfamily 3C (PR, AR, GR and MR). The two PR isoforms are not shown, as they are generated from the same gene via alternate transcriptional or translational start sites.

Figure 2.

Quantitative analysis of GR-HRE binding energetics. (A) Representative quantitative footprint titration image of GR binding to the Pal sequence. Schematic to the right indicates position of the HRE and approximate location of the transcriptional start site. Fractional saturation () was determined by integrating band intensities across the entire HRE. (B) GR-HRE assembly model depicting the total binding reaction and macroscopic product constant (K_tot), the total affinity for assembling two GR monomers at a palindromic response element. (C) Fractional saturation of the Pal sequence from two independent footprint titrations. Solid line represents global fit to both data sets using the K_tot binding model in Panel (B) and Equation 2 (SD = 0.087); dashed line represents fit to Equation 1 (SD = 0.126).

Figure 3.

K_app versus fold-activation for seven HREs. (A) Plot of fold-activity values ± SEM for the seven HRE sequences shown in Table 1; dashed line represents linear regression. An identical R² result is obtained if the data are plotted as a function of total binding affinity (K_tot). (B) Plot of simulated fold-activity for the seven HREs as a function of nanogram GR expression vector. Data points and dashed lines represent cross-sectional analysis used to generate plot in following panel. (C) Plot of simulated fold-activities as a function of K_app for five GR expression vector doses (3, 32, 100, 316, 1000 and 1500 ng). (D) Plot of simulated error-perturbed fold-activities ± SEM (n = 3) for the seven HRE sequences at 100 ng GR expression vector dose; dashed line represents linear regression. Error added was identical to that in panel (A); see (10).

Figure 4.

Linear regression of log-transformed fold-activities and K_app values at 100 ng GR expression vector. (A) Log–log transformation of the experimental data presented in Figure 2A. (B) Same as Panel (A) using the simulated error-perturbed data in Figure 2D. For both panels, the same R² value is obtained if the data are analyzed in units of K_tot rather than K_app.

Figure 5.

Relative fold-activities for seven HREs as a function of simulated mutagenesis and coactivator knockdown. Simulated activity differences relative to wild-type (gray) for 100 (top) and 1000 (bottom) ng GR-expression vector doses, when K_tot is reduced 10-fold (red) or fold-activity (FA) is reduced 2-fold (yellow) (*P < 0.05; **P < 0.005).

Figure 6.

Global fitting of dose–response curves indicates that DNA binding energetics largely account for sequence-specific transcriptional activation. (A) Plots of dose–response curves for seven HREs as a function of GR expression vector dose. Filled circles indicate fold-activation ± SEM (n = 3). Lines represent global fit of all dose–response curves using a statistical thermodynamic dimer-binding model, K_tot for each respective HRE and global scaling factors. (B) Same as Panel (A) using simulated error-perturbed fold-activation values. Error was identical to that in panel (A). The data shown here were also used to determine the extent of correlation presented in Table 2.

Figure 7.

Predicted and experimentally determined dose–response curves for four additional HREs. Dose–response curves ± SEM (n = 3) for GR-induced activity for the four HREs (filled circles). Dashed lines represent predicted dose–response curves using respective K_tot determined previously (10) and d, e and f scaling factors resolved in Figure 6A. DNA sequence and experimentally determined GR binding affinity for the four HREs were previously presented (10).

Figure 8.

Sequence-specific activation is maintained in multiple promoter-types, cell lines and in chromatin environment. (A) pA3-Pal and pA3-TAT₄ dose–response curves in COS7 cells (black) overlaid with dose–response curves ± SEM (n = 3) from the respective sequences in pGL3 vector (red). (B) pA3-Pal and pA3-TAT₄ dose–response curves in COS7 cells (black) overlaid with dose–response curves ± SEM (n = 3) for respective sequences in U2OS cells (red). (C) TA-induced activity ± SEM (n ≥ 2) of pGL3-TAT₄ (green), pGL3-Pal (blue) and pGL3-TAT_4-Y [red; (10)] determined in transient and stably transfected COS7 cells (1 μg GR expression vector). Dashed line represents linear regression.

Figure 9.

Microstate energetics of steroid receptor assembly at a simple two-site promoter. Circles represent receptor dimerization affinity (k_dim) and squares represent inter-site cooperativity (k_c). As dimerization was not observed for wild-type AR, T877A and GR, downward arrows have been added to indicate that plotted values represent lower limits. Error bars represent 67% confidence intervals. Because the dimerization and cooperativity terms each represent a microscopic rather than a macroscopic interaction (e.g. K_tot), they are represented by a lower case k.

Figure 10.

Predicted probabilities of the fully ligated promoter state under competitive binding conditions. (A) Simulation of competitive binding to an isolated half-site by three receptors differing in dimerization and cooperative energetics. Red, k_dim = 10 μM and k_c = 200; Blue, k_dim = 1 μM and k_c = 50; Green, k_dim = 16 nM and k_c = 1. Affinity of monomer binding to half-site was assumed to be an identical 1 μM for all receptors. Strength of dimerization and cooperativity terms is indicated schematically by font size of each parameter. (B) Same as (A) but now binding to an isolated palindrome. Affinity of pre-formed dimer binding was assumed to be an identical 10 nM for all receptors. (C) Same as above, but binding to a promoter containing two palindromic sites. (D) Same as above, but binding to a promoter containing one half-site and one palindrome. (E) Same as above, but binding to a promoter containing two half-sites.