To bind or not to bind - FoxA1 determines estrogen receptor action in breast cancer progression

Chromatin immunoprecipitation followed by massively parallel sequencing (ChIP-seq) is rapidly enabling the comprehensive characterization of genome-wide transcription factor-binding sites, thus defining the cistrome (cis-acting DNA targets of a trans-acting factor). Estrogen receptor (ER) ChIP-seq studies have been performed mainly in cell lines, but Ross-Innes and colleagues have now completed the first such study in clinical breast cancer samples. The study aimed at determining the dynamics of ER binding and differences between more and less aggressive primary breast tumors and metastases. The authors found that ER bound to DNA in both aggressive and drug-resistant tumors but to different sites and with different affinities. Given previous findings from cell lines, FoxA1 appears to play a critical role in this reprogramming of ER binding.


The article
Ross-Innes and colleagues [8] analyzed ER ChIP-seq data from 15 ER + tumors (eight with a good prognosis and seven with a poor prognosis) and three distant metastases. Th e authors found a core set of 484 ER-binding events present in at least 75% of all ER + tumors (but not in the ER − controls). Intriguingly, ER-binding signal intensity was highest in metastatic samples and lowest in patients with good outcomes, suggesting that binding intensity may correspond to disease progression. Diff eren tial binding analysis found 1,192 ER-binding events that were stronger in the poor prognosis/metastasis group in comparison with the good outcome samples and found 599 binding events more prevalent in the good outcome tumors. Motif analysis revealed the presence of estrogen response element (ERE) sites in all three groups but a unique enrichment of FoxA1 sites in the poor prognosis/metastatic tumors. An important fi nding was that ER still bound DNA in tumors resistant to hormonal therapies but was recruited to novel sites in the genome. Th ese sites are functionally and biologically relevant since a gene expression predictor based on genes within a 20-kb window of the binding sites was associated with survival in ER + data sets.
Treatment of breast cancer cells (whose ER cistrome closely overlaps with that of poor prognosis/metastatic tumors, presumably because the cell lines were isolated from metastases) with a mitogenic cocktail also resulted in rapid enrichment of novel ER-binding sites.

Abstract
Chromatin immunoprecipitation followed by massively parallel sequencing (ChIP-seq) is rapidly enabling the comprehensive characterization of genome-wide transcription factor-binding sites, thus defi ning the cistrome (cis-acting DNA targets of a trans-acting factor). Estrogen receptor (ER) ChIP-seq studies have been performed mainly in cell lines, but Ross-Innes and colleagues have now completed the fi rst such study in clinical breast cancer samples. The study aimed at determining the dynamics of ER binding and diff erences between more and less aggressive primary breast tumors and metastases. The authors found that ER bound to DNA in both aggressive and drugresistant tumors but to diff erent sites and with diff erent affi nities. Given previous fi ndings from cell lines, FoxA1 appears to play a critical role in this reprogramming of ER binding.
In triguingly, half of these binding sites occurred in regions to which FoxA1 was already bound or to which FoxA1 was recruited in response to mitogenic stimulus. High and correlated expression of ER and FoxA1 in metastatic samples further supports the idea that FoxA1 might direct the reprogramming of ER binding in advanced disease.

Viewpoint
Th is study presents an important and exciting milestone in our eff orts to understand the plasticity of ER function in breast tumors. Although the numbers of samples are small, the data strongly suggest that diff erential ER binding is associated with the outcome of patients with breast cancer. Increasing the number of samples will allow the analysis of poor outcome tumors separately from metastases, a critical question given the lack of knowledge about the role of ER in metastasis.
An interesting fi nding of the study is the observation that the average ER-binding signal was highest in the metastatic samples. Additional studies are necessary to determine whether this is a general phenomenon and, if so, the underlying mechanisms. Th is could include altered levels or post-translational modifi cation of ER (or both) or altered interaction of ER with co-regulator proteins. Alternatively, this may be simply a result of increased tumor cellularity or decreased tumor cell heterogeneity in metastatic samples or both. In any case, this is a curious fi nding and deserves further study.
For us, however, the most interesting observation is that reprogramming of ER binding seems to be associated with co-recruitment of FoxA1 and frequently with recruitment to sites pre-bound by FoxA1. Although this study does not include FoxA1 ChIP-seq data from clinical samples, the in vitro data point toward a unique mechanism of FoxA1 binding to 'pre-marked' DNA in poor outcome tumors, and this may have critical clinical relevance. Th e further use of frozen tumor specimens for ChIP-seq studies and the use of paraffi n-embedded samples as recently described by Fanelli and colleagues [9,10] will undoubtedly shed more light on these questions.