Whole-genome cartography of estrogen receptor alpha binding sites

Abstract

Using a chromatin immunoprecipitation-paired end diTag cloning and sequencing strategy, we mapped estrogen receptor alpha (ERalpha) binding sites in MCF-7 breast cancer cells. We identified 1,234 high confidence binding clusters of which 94% are projected to be bona fide ERalpha binding regions. Only 5% of the mapped estrogen receptor binding sites are located within 5 kb upstream of the transcriptional start sites of adjacent genes, regions containing the proximal promoters, whereas vast majority of the sites are mapped to intronic or distal locations (>5 kb from 5' and 3' ends of adjacent transcript), suggesting transcriptional regulatory mechanisms over significant physical distances. Of all the identified sites, 71% harbored putative full estrogen response elements (EREs), 25% bore ERE half sites, and only 4% had no recognizable ERE sequences. Genes in the vicinity of ERalpha binding sites were enriched for regulation by estradiol in MCF-7 cells, and their expression profiles in patient samples segregate ERalpha-positive from ERalpha-negative breast tumors. The expression dynamics of the genes adjacent to ERalpha binding sites suggest a direct induction of gene expression through binding to ERE-like sequences, whereas transcriptional repression by ERalpha appears to be through indirect mechanisms. Our analysis also indicates a number of candidate transcription factor binding sites adjacent to occupied EREs at frequencies much greater than by chance, including the previously reported FOXA1 sites, and demonstrate the potential involvement of one such putative adjacent factor, Sp1, in the global regulation of ERalpha target genes. Unexpectedly, we found that only 22%-24% of the bona fide human ERalpha binding sites were overlapping conserved regions in whole genome vertebrate alignments, which suggest limited conservation of functional binding sites. Taken together, this genome-scale analysis suggests complex but definable rules governing ERalpha binding and gene regulation.

Figure 1. ChIP-PET Analysis Identified 1,234 High Confidence ER Binding Sites in MCF-7 Cells

(A) A total of 635,371 PETs were sequenced and resolved into 1,234 binding site clusters after filtering for ambiguous and redundant mapping and local noise in amplified regions of the genome. (B) This diagram illustrates binding site mapping by a cluster of ChIP-PETs. ER binding site is situated in the region of overlap between the ChIP fragments and their PETs. (C) All of the 1,234 high confidence clusters have at least three PETs in their region of overlap (moPET3+) with the largest number of clusters being moPET3s.

Figure 2. The ERE Sequence Motif Is Enriched in ER Binding Sites

Overall, 71.6% of the 1,234 high confidence ChIP-PET clusters encode at least one ERE motif, and 3.73% contained no ERE or half site motif.

Figure 3. Putatively Higher Affinity Binding by ER Is Associated with the ERE

(A) ChIP-PET ER binding sites are validated by conventional ChIP followed by quantitative PCR. Data shown represent average of duplicate experiments. Binding sites are considered validated if the binding ratio (ER ChIP/irrelevant antibody control) ≥ 2. Validated sites are grouped by presence of EREs (allowing for up to two base deviation from consensus), half ERE sites, and no ERE motifs. (B) The frequency of ERE motifs increase as the size of binding site clusters increase. ERE motifs are only present in 7.05% similarly sized genomic fragments shown as the background.

Figure 4. ERE Sequences in ER ChIP-PET Binding Sites Are Functional Transcriptional Enhancers

(A) ER ChIP-PET binding sites were cloned into the pGL4-TATA luciferase reporter construct and transfected into MCF-7 cells that have been grown in hormone depleted medium for at least three days. The cells were treated with either ethanol or 10 nM estradiol for 18–24 h before harvesting for luciferase activity. pGL4-TATA and pGL4-2ERE-TATA (two copies of the vitellogenin ERE cloned upstream of TATA box of pGL4-TATA) were used as negative and positive controls, respectively. The cells were also cotransfected with the HSV-TK renilla vector as an internal control for transfection efficiency. The data represent the average of three individual experiments ± standard error of mean. The binding site coordinates and their adjacent genes are: ERE1 and ESR1, Chromosome 6: 152029288–152029705; ERE2 and ESR1, Chromosome 6: 152071268–152071889; ERE3 and FOXA1, Chromosome 14: 37189409–37189699; ERE4 and GREB1, Chromosome 2: 11589053–11589737; ERE5 and GREB1, Chromosome 2: 11621762–11622024; ERE6 and GREB1, Chromosome 2: 11622967–11623504; ERE7 and GREB1, Chromosome 2: 11630097–11630780; ERE8 and PGR, Chromosome 11: 100554271–100554807; ERE9 and PGR, Chromosome 11: 100712072–100712428; ERE10 and CA12, Chromosome 15: 61467060–61467460; ERE11 and TFF1, Chromosome 21: 42669273–42670075. (B) Putative ERE motifs in ER ChIP-PET binding sites were mutated and examined in transient transfection studies as described in (A).

Figure 5. Comparison of ER Binding Sites Discovered by ChIP-PET and Published ChIP-on-Chip Experiments Indicate Sites Common to Both Technologies and Platform-Form Specific Sites

(A) In human Chromosome 21 and 22 studies, 57 ER binding sites were discovered by Carroll et al. [15], and 36 sites were identified in this study. There is an overlap of 20 sites between the two studies. (B) Validation of select binding sites from both studies by ChIP and qPCR suggest that the common sites (both) are high affinity sites, whereas sites unique to each technology tend to have more moderate affinity for ER. (C) Venn diagram of the overlap between the 3,665 sites discovered by Carroll et al. and the 1,234 sites identified in this study. The 624 binding sites identified in this study actually correspond to 633 binding regions reported by Carroll et al. [16].

Figure 6. Comparative Analysis of Binding Site Affinity and Location Adjacent to E2 Up-Regulated Genes Versus Down-Regulated Genes

(A) Binding site affinity measured by ChIP and qPCR for 22 ChIP-PET sites within 100 kb of E2 responsive genes detected in the microarray studies. Up-regulated genes are denoted in red bars and down-regulated genes in green bars. Each binding site is furthered characterized for the presence of EREs, half EREs, or no EREs. (B) Locations of binding sites adjacent to up- (blue line) and down- (red line) regulated genes are mapped relative to the transcripts. Relative location to a random set of genes from the UCSC KGs database is included as a reference.

Figure 7. ER Binding Sites Are Adjacent to Genes Associated with ER-Status and Disease Outcome in Breast Cancer Patients

(A) Expression profiles of genes adjacent to the 1,234 ER binding sites (<100 kb) cluster 260 breast cancer patients into ER+ and ER− groups. ER status is indicated by blue (ER+) and orange (ER−) bars beneath each patient sample. (B) Kaplan-Meier analysis of disease outcome indicates significantly longer survival for patients with the ER+ profile (red) and compared to those in the ER− cluster (black).

Figure 8. Most of the Sequences Flanking the 1,234 ER Binding Sites Are Not Conserved through Evolution

Measure of species conservation at all 1,234 ER binding sites from the center of the ChIP-PET cluster is depicted in the blue line. The green line measures species conservation of randomly selected fragments. The red line depicts the degree of conservation in 22% (273/1,234) of the binding sites bearing conserved elements, and the yellow line shows the degree of species conservation of the remaining 78% (961/1,234) of the binding sites that harbor no conserved elements. These results show that most of the conservation signal is driven by a minority of the binding sites. Conservation was measured by base-by-base comparisons.

Figure 9. Impact of Sp1 Knock-Down in MCF-7 Cells on E2 Stimulation of Target Genes Adjacent to ChIP-PET Clusters with Predicted Sp1 Binding Sites

(A) Cells were transfected with GL3 luciferase siRNA control or Sp1 siRNA constructs 72 h prior to treatment with 0.1% ethanol vehicle or 1nM E2 for 4 h. Expression of target genes was analyzed by quantitative real-time PCR. Values and error bars are based on the mean of three determinations. (B) Knock-down of Sp1 impacts ERα recruitment to E2-regulated genes. ChIP assays using ERα antibody were performed after transfection of MCF-7 cells with GL3 luciferase siRNA control or Sp1 siRNA for 72 h followed by 45 min treatment with 0.1% ethanol vehicle or 1nM E2. (C) Sp1 is present at ChIP-PET regions of E2-target genes, but its presence is not affected by E2 treatment. ChIP assays were performed using Sp1 antibodies after 45 min of 0.1% ethanol vehicle or 1nM E2 treatment. Enrichment of ERα or Sp1 at ChIP-PET regions was evaluated by quantitative real-time PCR and normalized to IgG control antibody. Results average two to four independent determinations.

Figure 10. Nonrandom Positional Distribution of AP-1, SF1, MAF, PAX3, PAX2, CDX, and AML1 (as Negative Control) Binding Sites in the 500-bp Window Centered on the Main ERE

The y-axis represents the cumulative frequency of the specific transcription factor motif, and the x-axis represents the position of that motif relative to the ERE centered at position 250. Motif hits are marked in red “+” and green “X” indicating forward and reverse strand hits respectively. Multiple hits on the same sequence are depicted as multiple marks on the same y-value sequence.

Figure 11. Alignment of TFBSs Enriched in ChIP-PET Clusters with Overlapping Sequence Motifs with the ERE

The consensus string is a representation of the matrix based on the following rules [48]: A single nucleotide is shown if its frequency is greater than 50% and at least twice as high as the second most frequent nucleotide; a double-degenerate code indicates that the corresponding two nucleotides occur in more than 75% of the underlying sequences, but each of them is present in less than 50%; all other frequency distributions are represented by the letter “n;” the letters in red indicate bases identical to the ERE consensus.