This section provides detailed clinical and experimental information for the primary breast tumors and distant metastases studied in the research. Patients with good outcome tumors (G1 to G8) were all ER+, PR+, and HER2-, while those with poor outcome tumors (P1 to P7) were ER+, but either PR- or HER2+. ChIP-seq was performed on intact solid tumors, except for Met 1, where epithelial cells were purified from pericardial fluid before analysis. Supplementary Figure 2 shows Illumina sequencing data for all samples, including the number of aligned and filtered reads, and the number of peaks identified using MACS or SWEMBL for tumor samples and cell line experiments. Supplementary Figure 3 details the heterogeneity within specific tumors by comparing ER ChIP-seq data from different sections of three tumors (G5, P5, and P6). The correlation in ER binding profiles between sections of the same tumor was higher than between unmatched samples. Supplementary Figure 4 presents the number of peaks overlapping in different ER+ tumors and the global ER binding signal in differential ER binding events from patients with different outcome tumors and metastases. Supplementary Figure 5 shows the expression of genes adjacent to 484 core ER binding events in ER+ breast cancers, which are elevated compared to ER- breast cancers in nine independent datasets. Supplementary Figure 6 illustrates survival analyses based on core ER binding events, using a gene signature to stratify patients and compare those with the top and bottom one-third risk indices. Supplementary Figure 7 represents a heatmap of patient tumor samples clustered based on 1,791 differential ER binding events, with P5 being the only tumor that clusters incorrectly due to its long-term survival. Supplementary Figure 8 identifies enriched processes and pathways based on differential ER binding events using the Genome Region Enrichment Analysis Tool (GREAT). Supplementary Figures 11 and 12 provide survival analyses in independent datasets, comparing genes adjacent to ER binding events enriched in patients with poor and good outcome tumors. Supplementary Figure 13 details multivariate analyses of genes near poor and good outcome ER binding events, showing their independence from factors like grade, size, and stage, and comparisons with established prognostic signatures. Supplementary Figure 14 shows survival analysis of a gene predictor identified in ER+ tumors but tested in ER- tumor cohorts, demonstrating that the predictor has predictive value only in ER+ tumor cohorts. Supplementary Figure 16 demonstrates the reprogramming of ER binding events by a cocktail of mitogens (EGF, IGF-1, IL-6, and TNF-α) after 90 minutes of treatment, with enriched motifs identified in both induced and reduced binding regions.This section provides detailed clinical and experimental information for the primary breast tumors and distant metastases studied in the research. Patients with good outcome tumors (G1 to G8) were all ER+, PR+, and HER2-, while those with poor outcome tumors (P1 to P7) were ER+, but either PR- or HER2+. ChIP-seq was performed on intact solid tumors, except for Met 1, where epithelial cells were purified from pericardial fluid before analysis. Supplementary Figure 2 shows Illumina sequencing data for all samples, including the number of aligned and filtered reads, and the number of peaks identified using MACS or SWEMBL for tumor samples and cell line experiments. Supplementary Figure 3 details the heterogeneity within specific tumors by comparing ER ChIP-seq data from different sections of three tumors (G5, P5, and P6). The correlation in ER binding profiles between sections of the same tumor was higher than between unmatched samples. Supplementary Figure 4 presents the number of peaks overlapping in different ER+ tumors and the global ER binding signal in differential ER binding events from patients with different outcome tumors and metastases. Supplementary Figure 5 shows the expression of genes adjacent to 484 core ER binding events in ER+ breast cancers, which are elevated compared to ER- breast cancers in nine independent datasets. Supplementary Figure 6 illustrates survival analyses based on core ER binding events, using a gene signature to stratify patients and compare those with the top and bottom one-third risk indices. Supplementary Figure 7 represents a heatmap of patient tumor samples clustered based on 1,791 differential ER binding events, with P5 being the only tumor that clusters incorrectly due to its long-term survival. Supplementary Figure 8 identifies enriched processes and pathways based on differential ER binding events using the Genome Region Enrichment Analysis Tool (GREAT). Supplementary Figures 11 and 12 provide survival analyses in independent datasets, comparing genes adjacent to ER binding events enriched in patients with poor and good outcome tumors. Supplementary Figure 13 details multivariate analyses of genes near poor and good outcome ER binding events, showing their independence from factors like grade, size, and stage, and comparisons with established prognostic signatures. Supplementary Figure 14 shows survival analysis of a gene predictor identified in ER+ tumors but tested in ER- tumor cohorts, demonstrating that the predictor has predictive value only in ER+ tumor cohorts. Supplementary Figure 16 demonstrates the reprogramming of ER binding events by a cocktail of mitogens (EGF, IGF-1, IL-6, and TNF-α) after 90 minutes of treatment, with enriched motifs identified in both induced and reduced binding regions.