Oestrogen regulates cellular behaviour in diverse target tissues, although this is mostly limited to the reproductive organs. During development, oestrogen plays a crucial role in the functioning of female reproductive organs, including the mammary gland, where it generally functions as a growth-promoting factor. Oestrogen can also have growth-promoting properties in developed mammary glands, which can in some cases lead to hyperplasia and breast cancer. The link between breast cancer growth and oestrogen was described more than a century ago, and currently it is known that oestrogen functions by binding to two intracellular oestrogen (estrogen) receptors: ER-α and ER-β. ER-α (from herein called ER) activity can be modulated clinically with the goal of blocking oestrogen action in breast cancer cells. As an example, the selective ER modulator (SERM) tamoxifen is used as an antiproliferative drug in treating breast cancer – although, interestingly, tamoxifen can act as a proliferative drug in other reproductive tissues, such as the endometrium (Ref. Reference Osborne1). Tamoxifen is known to bind to the ER and induce changes in gene transcription that culminate in cellular growth responses (Refs Reference MacGregor and Jordan2, Reference Jordan3), but the mechanistic basis for the tissue-specific tamoxifen–ER response is currently unclear. A better understanding of ER function may ultimately lead to more-effective ways of modulating ER function in the various hormonally responsive tissues, particularly with regard to breast cancer treatment.

Analysis of the molecular mechanisms of ER activity involves several distinct areas of research: assessment of gene expression profiles regulated by oestrogen, identification of the proteins and cofactors that associate with the ER complex to promote or inhibit transcription, characterisation of histone and DNA modifications that can regulate the transcriptional potential of specific genes, and identification of the ER-binding sites within the genome that define cis-regulatory elements used by the ER. The most significant advances have involved characterisation of the gene targets of the ER and the factors that participate in the ER–multiprotein complex that regulates transcription. Chromatin immunoprecipitation (ChIP) assays have allowed the investigation of candidate protein association with one or two specific loci in promoter-proximal regions of target genes and have revealed the diverse network of proteins that participate in ER-mediated transcription. A complete understanding of the proteins involved in ER-mediated transcription is paramount, since changes in stoichiometry of specific cofactors can have profound physiological effects (Refs Reference Anzick4, Reference Lavinsky5, Reference Shang and Brown6, Reference Torres-Arzayus7). However, it is of equal importance to define the genomic features that dictate where and when ER complexes can associate with chromatin, since it is these specific features that likely determine what gene transcription changes occur in response to oestrogen stimulation. A major hurdle in delineating this part of the problem is the complexity of the mammalian genome and the lack of technical approaches capable of interrogating the genome in a comprehensive and meaningful manner. This review focuses primarily on the genomic investigations that have helped define a functional ER-binding site and the significance of this with regard to ER-mediated gene regulation and breast cancer.

Identification of ER-binding motifs

Characterisation of motifs and ER binding at known ER targets

Analysis of promoter-proximal regions of oestrogen-regulated gene targets resulted in the definition of two distinct modes of ER binding. The first, termed ‘classical’ binding, is ER association through direct physical interaction with a 13 bp inverted palindromic oestrogen-responsive element (ERE) – GGTCANNNTGACC – although various derivations of this motif also appear to be functional in both in vitro and in vivo systems (Ref. Reference Klinge8). The observation that this motif occurred in the regulatory region of an oestrogen-target gene was first made in the promoter of Xenopus vitellogenin gene (Ref. Reference Klein-Hitpass9), but the motif was later shown to occur in promoter regions for several other known oestrogen-target genes, including TFF1 (trefoil factor 1) (Refs Reference Berry, Nunez and Chambon10, Reference Brown11), EBAG9 (‘ER-binding site associated 9’) (Ref. Reference Ikeda12) and cathepsin D (Refs Reference Cavailles, Augereau and Rochefort13, Reference Krishnan, Wang and Safe14, Reference Wang15) (Fig. 1). The second mode of ER binding is termed ‘nonclassical’ and involves ER associating with DNA indirectly via other motifs, including those for the transcription factors AP-1 and Sp1, likely through interaction with these factors. Much of this work focused on specific gene targets that had proliferative roles in cell cycle progression, including MYC, and the genes for insulin-like growth factor 1 and cyclin D1 [via Sp1, AP-1 and NF-κB (nuclear factor κB) elements, respectively]. Interestingly, both cyclin D1 and MYC were shown to be sufficient to promote cell cycle progression in MCF-7 breast cancer cells (Ref. Reference Prall16), indicating the importance of understanding how the ER can regulate their transcription.

Figure 1.

A basic schematic model of oestrogen action. Oestrogen (E2) is a steroid and as such can diffuse into cells readily, where it associates with an oestrogen receptor (ER)-α homodimer. This complex can subsequently find specific motifs (oestrogen-responsive elements; EREs) within the DNA and physically bind to those regions of the DNA to induce transcription. However, the ER has also been shown to bind to DNA via motifs that are specific to other transcription factors.

Almost all of these conclusions about DNA motifs required for ER-mediated transcription were drawn from in vitro studies assessing promoter activity in a histone-free reporter assay system, which does not completely recapitulate the pressures and restrictions imposed by chromatin (Ref. Reference Cheung, Schwabish and Kraus17). Despite this, the mapping of a putative functional ERE within the promoter-proximal regions of TFF1 and cathepsin D genes provided the opportunity for in vivo characterisation of temporal protein associations on a chromatin target. Using ChIP assays (explained in more detail below), several research groups have mapped in extensive detail the slew of proteins that associate with the ER on a known regulated promoter at different time points after synchronisation of transcription. That numerous proteins, including several cofactors and the transcription machinery, participate with the ER in the association with chromatin adjacent to target genes is not overly surprising. However, what was unexpected from these studies was the observation that ER complexes associated with chromatin in an oscillating, cyclic fashion (Refs Reference Shang18, Reference Metivier19) (Fig. 2). Subsequent to the initial observation that the ER and a majority of assessed cofactors cycle on and off the chromatin of the promoter regions of one or two target genes, two schools of thought have evolved about the periodicity in nuclear-receptor cycling kinetics. Using fluorescence recovery after photobleaching (FRAP), glucocorticoid receptor and ER association with chromatin was shown to cycle with rapid kinetics, measured in seconds (Refs Reference Stenoien20, Reference Becker21), whereas ChIP suggested slower cycling kinetics for the ER, measured in minutes or hours (Refs Reference Shang18, Reference Metivier19). Although these differences likely result from the different approaches used to measure kinetic changes, it is probable that both conclusions are correct – that is, that nuclear receptors oscillate on and off the chromatin rapidly, but that overall global changes are detected in minutes or hours, correlating with the detectable accumulation of transcript levels.

Figure 2.

Cyclic association of the oestrogen receptor transcription complex with a promoter. (a) Chromatin immunoprecipitation (ChIP) of the oestrogen receptor (ER) and suspected cofactors on the cathepsin D promoter after increasing periods of oestrogen stimulation reveals cyclic association of the transcription complex, supporting (b) a model whereby cofactors associate with ER to potentially regulate chromatin structure. In this ChIP approach, MCF-7 breast epithelial cells are deprived of steroids to induce cell cycle arrest, and then transcriptional events are synchronised by the addition of oestrogen for increasing time periods. The cells are crosslinked and the chromatin is collected and sonicated. Precipitation by an antibody against the protein of interest enriches for DNA associated with the protein. DNA is purified and PCR is performed of a known binding site within a promoter of a regulated gene. Abbreviations: AcH, acetylated histone; AIB1, amplified in breast cancer 1; CBP, CREB-binding protein; p/CAF, p300/CREB-binding protein associated factor; Pol II, RNA polymerase II. The ChIP data in part a are reprinted from Ref. Reference Shang18 (© 2000 Elsevier), with permission from Elsevier.

Experimental approaches for finding ER-binding sites

Most experimental approaches to define ER-binding sites in an unbiased manner have used ChIP to isolate bound regions of the chromatin, followed by various methods for identifying the specific fragments of DNA associated with the ER. In recent years there has been a rapid increase in studies mapping transcription-factor-binding sites using these approaches, and analyses focusing specifically on mapping ER-binding sites have constituted a significant fraction of this work. ChIP exploits the strong interaction between proteins and chromatin in vivo, allowing identification of regions of the genome bound directly to the ER as well as regions bound indirectly via other transcription factors or coregulators. Also, since these approaches generally do not have any preconceived ideas of specific target genes, ER-binding sites can be defined in an unbiased manner.

ChIP–sequencing strategies

One potential approach for defining the DNA fragments that associate with the ER, as identified by ChIP, is the direct cloning of bound DNA and the subsequent sequencing of clones. This avenue has not been extensively pursued, although Giguere's group found 11 new ER-binding sites as a result of cloning, sequencing and verification as genuine ER-binding sites (Ref. Reference Laganiere, Deblois and Giguere40). A significant problem with the cloning and sequencing approach is that most DNA that associates with the ER in ChIP is background genomic DNA. The fact that billions of base pairs of background DNA are precipitated along with several thousand specific ER-binding sites attests to the problems associated with indiscriminant sequencing of precipitated chromatin. As such, any fragments that are represented only once require subsequent validation to ensure that they are genuine in vivo ER-binding sites.

High-throughput sequencing approaches can be employed to identify binding sites based on the fact that these particular loci are over-represented when compared with the majority of background genomic DNA and are therefore sequenced multiple times. The paired-end ditag (PET) approach, producing concatemers of small tags of genomic sequence, significantly increases the information generated by a single sequencing event, allowing the identification of numerous genomic loci (tags) at any one time. This method, when conducted on a large enough scale, allows binding sites to be deciphered based on the recurrence of specific tags, which can then be associated with genomic locations. Although significant sequencing is required, this approach has been successfully used to map binding sites for several important transcription factors, including CREB (cAMP-response-element-binding protein) and p53 (Refs Reference Impey41, Reference Wei42) – and more recently the ER (Ref. Reference Lin43). This ER-binding site map, using PET technologies, revealed 1234 high-quality ER-binding sites, which were used to characterise underlying properties of ER function (Ref. Reference Lin43). Recent advances in DNA sequencing technologies will permit generation of much more sequence information with less effort and time. The most developed of these technologies are: 454 Life Sciences (CT, USA), which uses an emulsion-based PCR followed by high-throughput sequencing (Ref. Reference Ng44); Solexa (Illumina, CA, USA), which utilises novel reversible terminator sequencing; and ABI (CA, USA) SOLiD, which utilises an oligonucleotide ligation and detection methodology. Other companies are developing independent methods for ultra-high-throughput sequencing, including Helicos BioSciences (MA, USA). All these approaches produce orders of magnitude more sequence information than current standard sequencing. Clearly these technologies will change the pace at which we can map transcription-factor-binding events in the near future.

ChIP-on-chip

Another approach to defining ER-binding sites in an unbiased manner is the use of microarrays that contain probes that cover extensive regions of the genome (Fig. 3). Initally, these approaches were limited to specific regions of the genome, primarily because of the technical inability to cover the entire genome on a reasonable number of microarrays. A pilot experiment used ChIP–microarray (or ChIP-on-chip) methodology to map ER-binding sites across the nonrepetitive sequence of chromosomes 21 and 22, which constitutes ~3% of the genome and contains about 750 known coding genes (Ref. Reference Carroll45) (Fig. 4). As a result of the complex nature of repetitive DNA, binding sites in these regions cannot be assigned a unique locus and, as such, probes representing these regions were omitted from the microarrays. This approach revealed several novel findings with regard to ER transcription, including the fact that the ER rarely binds to promoter-proximal regions but mainly to intergenic regions (Ref. Reference Carroll45). Mining of the ER-binding sites identified on the 35 Mb of nonrepetitive sequence on chromosomes 21 and 22 revealed an enrichment of ERE and forkhead motifs. Subsequent biochemical work proved that the forkhead protein FOXA1 binds to chromatin and is required for the ER to associate with DNA: the specific silencing of FOXA1 inhibited the ability of ER to bind to DNA and induce gene transcription of several target genes (Ref. Reference Carroll45). This study highlighted the fact that focusing on promoter regions is insufficient to describe the complete network of elements required for ER transcription. It is also likely that the ER can bind to and utilise regulatory elements located in repetitive DNA, an area of research that it might be possible to address using high-throughput sequencing in combination with ChIP (Ref. Reference Lin43).

Figure 3.

Schematic of oestrogen receptor ChIP-on-chip procedure. MCF-7 cells are deprived of hormones to induce cell cycle arrest, and transcription is subsequently reinitiated by adding oestrogen. Chromatin is then crosslinked by adding formaldehyde and the chromatin sheared into smaller fragments by sonication. The chromatin is incubated with an antibody against the oestrogen receptor (ER) and the chromatin precipitated from background: this enriches for fragments of chromatin that are bound by the ER. The chromatin is purified and the DNA is validated, amplified and labelled. Labelled DNA is hybridised to microarrays. Total genomic DNA (input) is also included as a control and ER-binding sites are determined as regions of the genome enriched above the input background levels. Increased binding of DNA to specific probes is reflected as increased signal on the microarray.

Figure 4.

Graphical representation of binding sites of the oestrogen receptor on chromosome 21, as determined by ChIP-on-chip. The oestrogen receptor (ER) binds to a limited number of sites, most of which are distant from promoter regions. ER ChIP-on-chip was performed by stimulating cells with oestrogen, crosslinking the chromatin and immunoprecipitating the ER. The chromatin was hydridised to tiling microarrays covering all the nonrepetitive sequence from chromosomes 21 and 22 at 35 bp resolution (Ref. Reference Carroll45). This means that a probe existed every 35 bp along the nonrepetitive region of the chromosome. The blue bars indicate gene locations; the black bars represent predicted ER-binding sites in the form of an oestrogen-responsive element (ERE); the red bars represent the experimentally defined in vivo ER-binding sites. The numbers under the chromosome represent base pairs. These data confirm that ER binds only to a small fraction of predicted binding sites that contain an ERE and therefore that a majority of EREs are not functional. Below is an expanded view of the TFF1 gene (in its genuine 3′ to 5′ orientation), showing ER binding at the promoter and an upstream enhancer. This gene represents one of the few ER-binding sites at a promoter region. A second binding site exists upstream of the TFF1 gene and functions as an enhancer of TFF1 transcription, despite the fact that it is located within an adjacent gene, TMPRSS2.

ChIP-on-chip has also been successfully employed to map ER-binding sites in an unbiased manner on most promoter-proximal regions. One study mapped a number of ER-binding sites on the ~18 000 promoter regions represented (Ref. Reference Laganiere46), and also showed that the FOXA1 promoter is a direct target of the ER and subsequently confirmed that FOXA1 plays an essential role in ER-mediated transcription. A similar approach focused on the sequence contained within ER-binding sites determined by ER ChIP-on-chip using promoter arrays (Ref. Reference Cheng47) and identified a bias of MYC-binding sites (E-boxes) in conjunction with EREs. This particular study subsequently showed that MYC and ER have coregulatory roles in oestrogen-mediated transcription (Ref. Reference Cheng47). The unbiased identification of underlying biological pathways, such as Forkhead and MYC proteins, involved in ER-mediated transcription emphasises the power of this reverse-biological approach where the sequence within in vivo binding sites is used to identify genuine interacting transcription programmes. A more recent method combined ER ChIP with Illumina-based beadchip technology representing all human promoter regions (Ref. Reference Kwon48). The beadchip technology has unparalleled sensitivity, and when combined with ER ChIP revealed approximately 500 promoter regions bound by the ER. Extension of this work to the remainder of the genome will likely prove to be very informative in defining the number of ER-binding sites within the genome.

ChIP-on-chip approaches had focused primarily on promoter-only regions or contiguous sequence of one or two chromosomes, mainly because of the large number of probes required to cover such regions. Technological advances in microarray production have produced the entire 1.5 billion base pairs of nonrepetitive human genomic sequence on a seven microarray set. ER ChIP-on-chip analysis was performed using these whole human genome microarrays to define the ER-binding profile (Ref. Reference Carroll49). Using a relatively stringent statistical cut-off, representative of a false discovery rate (FDR) of 1%, 3665 ER-binding sites were identified within the nonrepetitive sequence of the genome. Similarly, RNA Pol II ChIP-on-chip performed on the whole genome identified 3629 sites. A combination of the binding sites for ER and RNA Pol II with gene expression profiles after oestrogen stimulation of cells for different time points, reveals novel aspects of ER transcription, particularly with regard to downregulated genes, which constitute a significant fraction of all oestrogen-mediated gene changes (Refs Reference Carroll49, Reference Frasor50) yet are poorly understood. Although ER binding is significantly enriched within 50 kb of oestrogen-upregulated genes, no discernable ER binding was seen within 50 kb of early (3 h) oestrogen-downregulated genes, suggesting that downregulation of genes immediately following oestrogen exposure results from indirect mechanisms – possibly physiological squelching (the sequestration of essential and rate-limiting components away from these genes, for use in transcription of other genes). Surprisingly, however, a bias of ER binding occurred within 50 kb of genes that are downregulated at a later time point of 12 h, which are likely to be indirect transcriptional targets. Analysis of the sequence contained within these ER-binding sites adjacent to the late downregulated genes revealed a bias of AP-1 motifs with very few EREs (Ref. Reference Carroll49). Exploration of potential factors that may function as direct transcriptional repressors via AP-1 motifs within the ER-binding sites led to the discovery that NRIP-1/RIP140 (Ref. Reference Cavailles25), which is induced by oestrogen at the early 3 h time point, directly binds to and represses several assessed late oestrogen-downregulated gene targets (Refs Reference Carroll49, Reference Teyssier51).

Although the first genome-wide maps of ER binding are only now becoming available, it is clear that the different approaches have very good overlap. The intersection between the ER ChIP–PET dataset and the ER ChIP-on-chip dataset (Refs Reference Lin43, Reference Carroll49) is greater than 50%, supporting the reproducibility of approaches using ChIP in combination with different methods for identifying bound fragments of DNA (Table 1).

Table 1.

The benefits and disadvantages of ChIP-on-chip and ChIP–sequencing to define transcription factor binding events

Abbreviation: ChIP, chromatin immunoprecipitation.

Future directions

The genomic revolution has permitted a unique opportunity to evaluate gene expression, expression of noncoding RNAs and, importantly, profiles of transcription factor binding on a genome-wide scale and in an unbiased manner. However, despite the fact that the term ‘genome-wide’ has been used liberally in recent years to describe location analyses, to date only a handful of genuine genome-wide location analyses have been performed for transcription factors (Refs Reference Impey41, Reference Wei42, Reference Carroll49, Reference Kim52, Reference Yang53, Reference Kim54, Reference Mikkelsen55, Reference Johnson56). The next few years will likely produce a large number of maps of binding sites of popular transcription factors, but several issues remain before it will become clear how cell-type specific a transcription factor binding profile is. Very little information exists about the overlap in binding sites for the same transcription factor in different cells types or in cells from different organs. As an example, it is unclear whether ER-binding sites identified in MCF-7 breast epithelial cells will be the same as the binding sites in other oestrogen-responsive tissue, such as bone or endometrium. Similarly, little is known about the overlap of binding sites of different transcription factors within the same cell line or cell type. Also, it is not completely clear what overlap exists in transcription factor binding profiles between species, although initial experiments on promoter regions suggest that several liver-specific factors have only 30% concordance in binding profiles between human and mouse (Ref. Reference Odom57). The recent first ER ChIP-on-chip in mouse liver supports the theory that transcription-factor-binding sites are significantly different between species (Ref. Reference Gao58), although it remains to be defined whether these differences are species specific or whether the differences are due to comparisons between liver and breast epithelial cells.

Chromatin structure will inevitably dictate activity of an enhancer region. Significant work has already been performed mapping specific histone marks to regions of the genome (Refs Reference Ballestar59, Reference Miao and Natarajan60, Reference Bernstein61, Reference Huebert62). Certain histone marks correlate with 5′ regions of actively transcribed genes, as was clearly shown at oestrogen-regulated promoter regions (Ref. Reference Cheng47). Similarly, we are only now starting to identify histone marks that likely influence chromatin at distal, nonpromoter regions where ER generally binds (Ref. Reference Lupien63). In the β-globin gene region, distal enhancers have been shown to be required for regulation of downstream promoter-proximal regions (Refs Reference Sawado30, Reference Bender64, Reference Schubeler, Groudine and Bender65), suggesting that interplay between distal enhancers and the promoters of target genes is not a simple matter of bringing the transcription machinery to the promoter. Clearly, our understanding of the communication between ER binding at enhancers and the promoters of oestrogen-regulated genes requires further work.

A further issue is the disproportionate number of binding sites compared with gene expression events in the cell line (MCF-7) experiments. The number of binding sites found in the genome-wide location analyses is usually in the thousands, and the ChIP-on-chip methods tend to use stringent statistical cut-offs, suggesting that the real number of binding sites may be in the order of about 10 000 or more. Surprisingly, however, the number of direct differentially regulated genes is less than 300, supporting the hypothesis that a large majority of ER-binding sites are not functional in MCF-7 cells in the defined experimental conditions. Since oestrogen is known to regulate different gene expression profiles in other target tissues, it is possible that these ‘nonfunctional’ ER-binding sites determined in MCF-7 cells are active and responsible for different gene expression patterns in other hormone-dependent tissue. Another possibility is that some of the ER-binding sites may be regulating noncoding RNAs. It was recently shown that oestrogen can influence microRNA levels (Ref. Reference Cohen66), and microRNAs have an established role in regulating breast cancer growth (Refs Reference Iorio67, Reference Ma, Teruya-Feldstein and Weinberg68).

A major difficulty faced by researchers in the field is the technical problem of having to assign distal ER-binding sites with specific gene promoters. Although methods such as chromosome conformation capture (CCC) have been used to identify suspected nuclear receptor enhancer–promoter interactions (Refs Reference Carroll45, Reference Wang, Carroll and Brown69), an unbiased method for identifying which cis-regulatory element regulates which gene(s) is a requirement. Modification of CCC (Ref. Reference Dekker70), potentially in combination with high-throughput sequencing approaches, may yield useful information about ER enhancer–promoter interactions. Use of CCC for the discovery of novel chromatin interactions has been recently shown in a number of simultaneous approaches identifying unknown chromatin regions that interacted with a particular sequence of interest (Refs Reference Dostie71, Reference Zhao72, Reference Simonis73). A combination of these genomic tools will help provide a complete map of the binding activity of each transcription factor.

A major hurdle in the field is extending on the data generated using cell lines into human tumours in order to validate the findings observed in tissue culture experiments. A critical step in achieving this goal is the establishment of chromatin tools, such as ChIP, in primary human tumour samples. A requirement of all established ChIP methods to date has been a cell number of >1 × 10⁵ cells, although methods for ChIP using smaller numbers have been described recently (Ref. Reference O'Neill, VerMilyea and Turner74). In order to perform genome-wide binding assays in tumour material, ChIP-on-chip or ChIP–sequencing techniques must be reproducibly applied to the limited cell numbers that can be obtained from stored primary tumour material.

Clinical implications

Until recently, the field had attempted to define oestrogen/ER-induced growth of cancer cells by focusing on the promoter regions of key target oncogenes. Understanding how endocrine therapies, such as tamoxifen, work and what happens when they ultimately fail has also focused on discrete promoter regions of putative gene targets. Despite the fact that reporter assays of the promoter regions revealed insight into some of the cooperating transcription factors that can associate with ER at these promoters, the past few years have shown the importance of chromatin structure and histone modifications, all of which are devoid from in vitro reporter assays. The ER ChIP-on-chip and ChIP–PET datasets suggest that the regulatory regions are generally far from the target gene, but these cis-regulatory elements may be the key molecular switches that control transcription of target genes. Mutations or single-nucleotide polymorphisms (SNPs) in the regulatory elements may contribute significantly to alterations in transcription factor binding and transcription of important oncogenes or tumour suppressor genes. This is of particular importance given the fact that a majority of the most significant breast-cancer-associated SNPs are not in coding sequence and exist in nongenic regions (Ref. Reference Easton75), where enhancers and cis-regulatory elements may be influenced. Combining the ER binding information with the breast-cancer-associated SNPs may reveal insight into expression changes influenced by the presence of specific SNPs in enhancer regions.

Genome-wide analyses such as the ChIP–PET and ChIP-on-chip studies also reveal information about cooperating proteins that assist in ER activity. An example of this is FOXA1, which was shown to be essential for ER binding to chromatin and transcriptional activity (Refs Reference Carroll45, Reference Laganiere46). Subsequent studies have shown that FOXA1 is a better prognostic marker for breast cancer outcome than the traditional target progesterone receptor (Ref. Reference Badve76). Since a majority of endocrine-therapy-resistant breast cancers express ER (Ref. Reference Osborne1) and have functional ER pathways (Refs Reference Howell77, Reference Johnston78, Reference Johnston79), a better approach for inhibiting growth of ER-positive cancers may be to target the pioneer factors, such as FOXA1. Inhibition of proteins such as FOXA1 might have more potent antigrowth properties than the ER antagonists currently used in the clinic.