Dose-specific GR binding across the human genome. (A) Intensities of ChIP-seq signals across all GR binding sites. Briefly, A549 cells were treated in biological duplicate for 1 h with doses of DEX as indicated in the columns. Each row represents 1 of the 5,898 GR binding sites observed at the highest (50 nM) dose of DEX. Binding sites were classified and grouped by the lowest dose of DEX at which a binding site was called, and that class is indicated on the right as hypersensitive (white), medium sensitive (orange), and low sensitive (blue). The color within the ChIP-seq data indicates the ChIP-seq signal intensity, expressed in units of aligned reads per kilobase of the binding site and per million aligned ChIP-seq reads (RPKM). (B) The total number of binding sites in each class, using the same color scheme as in panel A. (C) The consensus GR binding motifs identified in each sensitivity class of GR binding sites.

Expression of PER1 is uniquely sensitive to low doses of glucocorticoids. (A) Dexamethasone dose-response curves for all 28 genes responsive to 5 nM DEX, as determined by RNA sequencing. Cells were treated for 1 h with increasing doses of DEX (x axis), and gene expression relative to control-treated cells was plotted (y axis). Of the responsive genes, PER1 (indicated in blue) was particularly sensitive and showed the strongest response to ≤5 nM DEX. (B) Immunoblot of PER1 and actin (as a loading control) after 4-h treatment with the indicated dose of DEX. (C) GR occupancy, as measured by ChIP-seq, in the region surrounding the PER1 transcription start site. Absent DEX, no occupancy was detected in the region (top line), whereas binding upstream of PER1 occurred after addition of 0.5 nM DEX (second line), and binding in the first intron of PER1 only occurred after addition of 5 nM DEX.

The GR binding sites flanking the PER1 transcription start site are sufficient to drive gene expression outside the genome. (A) Diagram of the genomic regions that were cloned into a promoterless luciferase reporter vector. Three regions were cloned from the PER1 locus. Two regions (“Full” and “Upstream,” shown in red and blue, respectively) contain the GR binding site upstream of the PER1 transcription start site. A third region (“Intron,” in purple) only contains the intronic GR binding site. For both the full and intron reporter, luciferase was maintained in phase with the PER1 start codon; for the upstream reporter, cloning included the endogenous promoter and stopped at the first nucleotide of the annotated transcription start site. As an additional control, a pair of GR binding sites upstream of the STOM gene were also cloned into the same reporter vector. (B) Luciferase expression, as a percentage of maximal response (y axis), across increasing doses of DEX (x axis) for each reporter construct. Each point indicates mean ± the standard error of the mean (SEM) for 8 biological replicate measurements, and the line indicates a fitted dose-response curve with fixed minimum and maximum but variable Hill slope. (C) Response to DEX of the PER1 reporter fragments with the upstream GR binding sequence converted to match the intronic GR binding sequence (inset). As for panel B, points indicate means ± SEM for 8 biological replicates, and lines indicates the fitted dose-response curve with variable Hill slopes.

Minimal enhancer region sufficient to drive a hypersensitive glucocorticoid response. (A) Regions surrounding the hypersensitive GR binding site upstream of the PER1 transcription start site were cloned in front of luciferase with a minimal promoter. Also, regions surrounding the nonhypersensitive GR binding site in the first intron of PER1 were cloned into the same reporter (not shown) and are referred to as the “intronic enhancer.” (B) Reporter plasmids were transiently transfected into A549 cells and treated for 1 h with doses of DEX (x axis). For each construct, the response (log fold change) was calculated compared to control-treated cells and then normalized to the maximal response observed. Error bars indicate standard errors of the means, and curves indicate the fitted dose-response curve with a fixed maximum and minimum and a variable slope. Each line indicates a different enhancer, as indicated, with regions containing the hypersensitive GR binding site upstream of PER1 shown in the same colors as in panel A. MMTV results show the dose-response curve of the MMTV promiter, driving the same luciferase. (C) Dose-response curves of the PER1 enhancer regions in response to cortisol treatment. (D) EC₅₀s, estimated by fitting dose-response curves to each experiment's results. Error bars indicate 95% confidence intervals of the estimates. (E) Hill coefficients of the fitted dose-response curves for each enhancer, with error bars indicating 95% confidence intervals. Values close to 1 indicate first-order response kinetics, whereas values close to 2 indicate second-order kinetics.

Select nucleotides within the minimal hypersensitive enhancer are necessary for the low-dose response. (A) Diagram of mutations introduced into the hypersensitive GR response element via site-directed mutagenesis. The blue box indicates the GR binding motif in the enhancer. WT indicates the wild-type sequence and, for both mutated sequences, the changed bases are colored and underlined. (B) A549 cells were transiently transfected with mutated plasmids and then treated for 4 h with increasing doses of DEX (x axis), and luciferase activity relative to a control treatment was calculated (y axis) and plotted, with error bars representing standard errors of the means. Lines indicate dose-response curves fit to each reporter response. The wild-type sequence (green) had the most sensitive response, whereas the mutated sequences (blue and orange) had less hypersensitive and normally sensitive responses, respectively. Black indicates a 69-bp enhancer element derived from the same binding site but which lacks hypersensitivity. (C) EC₅₀ values were calculated from dose-response curves for wild-type and mutated enhancer elements, with error bars indicating 95% confidence intervals. (D) Hill coefficients for all four enhancer elements presented in panel B, with error bars indicating 95% confidence intervals. While the wild-type enhancer had second-order kinetics, one mutant (blue) had intermediate kinetics, and the second mutant (orange) had first-order kinetics that were indistinguishable from the 69-bp enhancer region. (E) Mutation scanning results for the enhancer region. Each point represents the effect on sensitivity to DEX of replacing 10 bp of wild-type DNA with (AC)₅. Error bars represent 95% confidence intervals of the fit to the log₁₀(EC₅₀). WT and 69 bp indicate 95% CIs of the wild-type sequence of the 235-bp enhancer and of the truncated 69-bp enhancer, respectively.

Hypersensitive PER1 expression is general to the cell line and conserved in the mouse. (A) Dose-response curves for enhancer elements, as described for Fig. 5A, but transfected into the human endometrial cell line ECC-1, using the same protocol as used for A549 cells. As in Fig. 5, points indicate mean responses, with error bars indicating standard errors of the means. The two enhancer regions that contained the GR binding site upstream of PER1 (green and blue) showed a more sensitive dose response than an enhancer derived from the GR binding site in the first intron of PER1. (B) EC₅₀s, calculated for the enhancer elements presented in panel A, with error bars indicating 95% confidence intervals. The hypersensitive GR binding sites (green and blue) both resulted in EC₅₀s close to 2 nM DEX, whereas the nonhypersensitive GR binding site was about half as sensitive (EC₅₀, ∼4 nM). (C) Comparison of Hill coefficients for activities of hypersensitive reporters between A549 cells and ECC-1 cells. (D) EC₅₀ for each of 20 genes responsive to 5 nM DEX in the mouse pituitary cell line AtT-20 and as measured using RNA sequencing. The lines indicates the means, and error bars indicate standard deviations of the distribution. (E) Time course of gene expression response of circadian rhythm genes after transient induction of PER1 expression. A549 cells were treated for 1 h with 0.5 nM DEX, after which cells were rinsed and returned to DEX-free medium (zero on the x axis). Cells were collected over the following 12 h, and expression was measured in a TaqMan reverse transcriptase quantitative PCR. For each of the five biological replicates, threshold cycles were calculated as the average over four technical replicates, and expression was normalized to that of GAPDH and plotted relative to A549 cells prior to any DEX treatment. Points indicate means, and error bars indicate standard deviations over the five biological replicates. (F) Expression of circadian rhythm genes after 4 h of continuous exposure to 500 pM DEX. Columns indicate mean relative expression levels, and error bars indicate ranges observed across 3 biological replicates.

Overlap of hypersensitive GR binding with open chromatin and transcriptional cofactors (A) Fraction of GR binding sites in each sensitivity class that overlap regions of DNase I hypersensitivity. (B) For each GR binding site, we estimated the degree of open chromatin by calculating the DNase I signal as the RPKM. Plotted is the distribution of DNase I signal across each class of GR binding site. The inset shows the mean ± standard deviation of each class of GR binding site. (C) Fraction of GR binding sites in each sensitivity class that had overlapped calls of binding sites for cofactors, as indicated on the x axis. As a control, we also performed ChIP-seq for USF1, a factor not known to interact with the GR, and dashed lines indicate the fraction of overlap with USF1 for each class. (D) Receiver operating characteristic (ROC) curve, showing the sensitivity and specificity of predicting hypersensitivity of a GR binding site based on the occupancy of each cofactor. The ROC was calculated for the observed enrichment for the ability to predict a hypersensitive GR binding site versus predicting a medium sensitive GR binding site.