ChIP-seq

ChIP-seq for GR and STAT3 was performed on 4 basal-like cell lines (SUM159, MDA-MB-231, HCC1937, and HCC70) and 4 luminal cell lines (MDA-MB-361, BT-474, MDA-MB-453, and MCF-7).

For ChIP experiments, protein-DNA complexes were covalently cross-linked by incubating cells in 1% formaldehyde for 10 min at room temperature. Cells were incubated with 0.125 M glycine for 5 min, to quench cross-linking reaction. Cells were washed and scraped with PBS (pH 7.4) (Lonza). Cells were lysed with Farnham Lysis Buffer (5mM PIPES at pH 8.0, 85 mM KCl, 0.5% NP-40) containing protease inhibitor (Roche). Cell lysate was centrifuged at 2,000 rpm for 5 min at 4 °C. The crude nuclear extract contained in the supernatant was stored at −80 °C. ChIP-seq was performed using antibodies for GR (sc-1003, Santa Cruz Biotechnology) and STAT3 (sc-482, Santa Cruz Biotechnology). A thorough version of the ChIP-seq protocol used in this study is available on the ENCODE Project website:

https://www.encodeproject.org/documents/df9dd0ec-c1cf-4391-a745-a933ab1af7a7/@@download/attachment/Myers_Lab_ChIP-seq_Protocol_v042211.pdf

STAT3 ChIP-seq datasets have been deposited in the NCBI Gene Expression Omnibus (GEO) accession numbers {"type":"entrez-geo","attrs":{"text":"GSE85579","term_id":"85579"}}GSE85579 and {"type":"entrez-geo","attrs":{"text":"GSE152203","term_id":"152203"}}GSE152203.

Fastq files from ChIP-seq were aligned to the hg19 build of the human genome using Bowtie with the following parameters: -m 1 -t -best -q -S -l 32 -e 80 -n 2. ChIP-seq peaks were identified by comparing GR ChIP-seq in cells induced with dexamethasone to GR ChIP-seq in cells treated with ethanol, and STAT3 ChIP-seq to input control libraries. Peaks were called using Model-Based Analysis of ChIP-seq-2 (MACS2) (26) with a p value cutoff of 1e −10 and the mfold parameter constrained between 15 and 100. Bedtools merge (17) was used to merge bed files of MACS2 narrow peak calls from each of the ChIP-seq experiments. Bedtools coverageBed (17) was used to extract read counts under each merged peak in each ChIP-seq experiment in each cell line. DESeq2 version 1.20.0 (20) was used to identify peaks with significantly different read depth between basal-like and luminal cell lines (adjusted p-value < 0.05), and significantly different read depth between GR and STAT3 ChIP-seq experiments (adjusted p-value < 0.05). A multivariate model was used to identify shared GR and STAT3 peaks whose read depth was significantly different between subtypes in both GR and STAT3 ChIP-seq experiments (adjusted p-value < 0.05), but not significantly different between GR and STAT3 in the same subtype (adjusted p-value > 0.05). Pheatmap package version 1.0.10 and Deeptools version 3.1.0(25) computeMatrix and plotHeatmap functions were used to create the heatmaps of ChIP-seq data.