Genomic analysis of estrogen cascade reveals histone variant H2A.Z associated with breast cancer progression

Sujun Hua^1,2,3,a, Caleb B Kallen^4,a, Ruby Dhar^1,2, Maria T Baquero⁵, Christopher E Mason⁶, Beth A Russell^1,2,6, Parantu K Shah^1,2, Jiang Liu^1,2, Andrey Khramtsov⁷, Maria S Tretiakova⁸, Thomas N Krausz⁸, Olufunmilayo I Olopade⁷, David L Rimm⁵ & Kevin P White^1,2

^aThese authors contributed equally to this work

Article highlights

Synopsis

One promise of the Human Genome Project was to stimulate the translation of large-scale functional analyses of the human genome to produce novel targets useful in the diagnosis or treatment of complex diseases. Here, we demonstrate an integrated approach to the study of a transcriptional regulatory cascade involved in the progression of breast cancer and we identify a protein associated with disease progression. The estrogen (E2) response is mediated in part by the estrogen receptor- (ER) (Smith and O'Malley, 2004), a ligand-activated and DNA sequence-specific transcription factor that plays an integral role in the initiation, development, and metastasis of breast and uterine cancers (Yager and Davidson, 2006). The oncogene c-myc, which is upregulated by ER in response to E2 (Dubik and Shiu, 1992), is one of the most broadly overexpressed oncogenes in human cancer.

Because nearly 70% of breast cancers express ER, antiestrogens remain a mainstay of therapy and prevention for these cancers. Unfortunately, this therapy is hindered by the subsequent development of tumors resistant to antiestrogen treatment through a process that often involves an E2-autonomous c-MYC (Jeng et al, 1998; Rodrik et al, 2005). A complete understanding of antiestrogen resistance in these tumors may depend on developing an integrated, comprehensive understanding of E2-dependent oncogenic pathways.

Microarrays now permit large-scale gene expression changes to be measured under selected tissue conditions. In addition, tiling arrays that represent the entire genomic sequence with high resolution have made it possible to locate the sites bound by specific transcription factors in a process that is known as 'location analysis' or 'binding site detection.' This is often done using antibodies to immunoprecipitate transcription factors that have been chemically crosslinked to their target sites in vivo. The precipitated DNA is then purified, amplified, labeled, and hybridized to genomic arrays. Combining gene expression data, in the presence and absence of an activated transcription factor, with data from transcription factor localization identifies the activated or repressed genes that are direct targets of the transcription factor (Gao et al, 2004).

We analyzed targets of E2 signaling in MCF7 breast cancer cells by performing genome-wide binding site detection of ER and c-MYC. We identified a total of 1615 ER-bound regions (P<1e- 5) throughout the human genome. The distribution of ER-binding regions ranged from proximal (<1 kb) to the nearest transcription start site (TSS) of a gene to over 500 kb from the closest TSS (Figure 1A). Similarly, we detected 311 MYC-bound regions across the human genome in MCF7 cells. The distribution of MYC-bound loci in relation to annotated TSSs is depicted in Figure 1B.

Figure 1

Genomic distribution of ER- and MYC-binding sites. The distributions of 1615 ER- (A) and 311 MYC- (B) binding sites in E2-stimulated MCF7 cells relative to known genes. Where multiple genomic probes indicated a transcription factor-bound region, the center of each ER- or MYC-binding region was designated as the bound position. Within annotated genes, binding sites were classified as follows: within 5' untranslated regions (5' UTR), within coding sequences (CDS), within 5'-most intron or the first intron, within other introns, and within 3' untranslated regions (3' UTR). ER or MYC binding in intergenic regions was further classified based on the distance to the nearest annotated gene (0–10, 10–50, and >50 kb).

Full figure and legend (109K)Figures & Tables index

We examined ER- and MYC-binding regions for the presence of transcription factor-binding motifs and detected statistically significant enrichment of canonical estrogen response elements (EREs, AGGTCAnnnTGACCT) in ER-bound regions and E-boxes (CACGTG) in MYC-bound regions. It is known that ER may exert effects at non-ERE-containing chromatin targets through protein–protein interactions with DNA-bound transcription factors, including AP-1 (Webb et al, 1999; Cheung et al, 2005). Consistent with this observation, AP-1 sites were found to be highly enriched in ER-bound regions. We also found a high frequency of ER-binding sites near binding site motifs for the forkhead transcription factor FOXA1. These data are consistent with a recently proposed model in which a subset of ER regulatory regions are first targeted for transcriptional regulation by FOXA1 (Carroll et al, 2005; Laganiere et al, 2005). Additionally, we discovered significant enrichment of binding sites for the GATA factor, CREB1, and MSX1. Similar computational analyses for binding site enrichment were performed for MYC-bound regions and significant enrichment was detected for CREB1, CTCF, AP-2, and Sp1. Overall, our results identify several transcription factors that may converge with ER and/or MYC at particular cis-regulatory modules to exert their effects on target gene transcription.

We examined the evolutionary rates of regulatory regions bound by ER and MYC and found that these differ. ER-binding regions sustained relatively high sequence conservation between human and mouse or human and rat (about 75 million year divergence times) whereas MYC-binding regions did not show sequence conservation above background levels.

Data from these location analyses were integrated with E2-dependent gene expression profiling to identify the E2-dependent direct gene targets for ER and MYC. We reasoned that such direct gene targets would represent excellent candidates for regulating the progression of the cancer phenotype. We similarly expected that some direct targets would serve as potential markers for metastatic potential and might serve as prognostic markers and candidate genes for cancer therapy. Depletion of such factors from cancer cells might be expected to impair cell proliferation. Consistent with this hypothesis, RNAi-mediated knockdown of gene targets of ER and c-MYC revealed novel mediators of E2-stimulated cell proliferation in MCF7 cells. One such gene coded for a histone variant known as H2A.Z. H2A.Z was directly regulated by c-MYC (Figure 8A and B) in response to E2 (Figure 8C).

Although gene knockdown and proliferation assays are suggestive of important gene functions, the significance of such findings for real tumors in living humans can be tested only by correlating the expression levels of these genes in primary tumor specimens with specific clinical features. Tissue microarray screening of over 500 breast tumors revealed that high expression of this epigenetic factor, H2A.Z, was significantly associated with lymph node metastasis and decreased breast cancer survival (Figure 8D and E). Detection of high H2A.Z levels independently increased the prognostic power of biomarkers currently in clinical use. This integrated approach has accelerated the identification of a molecule linked to breast cancer progression, has implications for diagnostic and therapeutic interventions, and can be applied to a wide range of cancers.