The ER subtypes, ERα and ERβ, are ligand-dependent transcription factors that interact with DNA and control transcription of ER target genes in response to estradiol (E2). In breast cancer cells, thousands of ER target genes and binding sites have been identified through genome-wide approaches (6, 15, 19, 31, 35). In the classical mechanism of ER action, binding to DNA occurs at palindromic estrogen response elements (EREs). However, many ER binding sites do not contain recognizable EREs (5, 35, 38); hence, in addition to direct binding to DNA sequences, ER tethering to other transcription factors appears to play a significant role in mediating estrogen action (21, 27, 33, 39, 47). The NF-κB pathway is activated by a variety of extracellular stimuli, such as proinflammatory cytokines and growth factors. In the canonical pathway, activation of the upstream IκB kinase (IKK) kinase complex leads to downstream phosphorylation and subsequent proteasomal degradation of the IκB inhibitory proteins. This liberates NF-κB family members p65 and p50, which then translocate to the nucleus, where they bind to their response elements (NF-κB-REs) and transactivate gene expression. Like ER, NF-κB can interact with many other transcription factors to regulate target gene expression.

ER and NF-κB family members have been shown to influence each other's transcriptional activity. Much work has been done to delineate the multiple mechanisms by which ER can repress NF-κB action to exert an anti-inflammatory effect (7, 13, 28, 41). Similarly, there are several pieces of evidence indicating that NF-κB can repress ER expression and transcriptional activity (2, 10, 14, 23, 36). However, there are relatively few examples of ER and NF-κB working together to increase transcription (1, 30, 45, 53). Recently, we found that these factors can work cooperatively on a number of genes, including PTGES, which codes for prostaglandin E synthase, and ABCG2, which codes for a drug efflux pump (17, 43). For these genes, we demonstrated that NF-κB may potentiate ER action by stabilizing ER occupancy on DNA at functional EREs. For the PTGES gene, cross talk occurs at the ERE itself, whereas for the ABCG2 gene, a weak NF-κB-RE located downstream of the ERE is required. This suggests that many gene-specific mechanisms of positive cross talk can occur between ER and NF-κB and that the specificity may be based on the nature and arrangement of the regulatory elements in the gene.

Previous studies have shown that cytokines regulate BIRC3 expression in an NF-κB-dependent manner through the first (NF1) and third (NF3) of three NF-κB-REs located in the promoter of the gene (24). We find that inhibition of the NF-κB pathway with IκBα-DN (Fig. 2A) or mutation of both NF-κB sites (Fig. 2B) not only prevents TNF-α from regulating expression of BIRC3, but also completely blocks E2 from potentiating TNF-α action, suggesting that both E2 and TNF-α require NF-κB activity to mediate their effects on the BIRC3 gene. We confirmed that IκBα-DN potently inhibits nuclear translocation of NF-κB family members, p65 and p50, in MCF-7 cells (see Fig. S1 in the supplemental material). The NF-κB family member p65 has been shown to bind to the promoter region containing NF1 and NF3 in response to cytokines (24). p65 recruitment to this region was assessed in breast cancer cells by ChIP assays, which demonstrated that the combination of E2 and TNF-α had little effect on p65 occupancy compared to that seen with TNF-α alone (Fig. 2C). It is therefore unlikely that E2 potentiates TNF-α action by influencing p65 recruitment to the promoter, which is in contrast to what we have previously shown for the ABCG2 gene (43).

Further analysis of the BIRC3 promoter sequence revealed a putative ERE located 156 bp upstream of NF1 (Fig. 3A). Deletion (−247) or mutation of this ERE (mERE) completely prevented E2 from potentiating TNF-α action but had little effect on TNF-α alone (Fig. 3B). This finding suggests that although the ERE does not appear to be functional in response to E2 alone, it possesses a latent functionality that becomes apparent in response to both E2 and TNF-α—or, in other words, TNF-α allows this ERE to function. Although the sequence of the ERE is near consensus (GGGCA TAT TGACC), we mutated it to a full consensus site (GGTCA CCG TGACC) according to MatInspector (44) because the single-nucleotide difference is at an essential position in the sequence. A BIRC3 promoter construct containing the consensus ERE also failed to respond to E2, indicating that the lack of functionality of the ERE is not because of a weak binding sequence (Fig. 3C). In addition, TPBM, a small molecule that prevents ER binding to DNA (37), prevented potentiation of BIRC3 expression by E2 plus TNF-α over TNF-α alone, suggesting that ER binding to the BIRC3 ERE is essential for enhanced regulation of the gene (Fig. 3D).

To examine recruitment of ER to the BIRC3 promoter, ChIP assays were carried out. In agreement with the lack of an E2 effect on BIRC3 expression, E2 had no effect on ER recruitment to the BIRC3 promoter region containing the ERE (Fig. 4A; see Fig. S2 in the supplemental material). However, the combination of E2 and TNF-α led to a rapid and robust recruitment of ER, which was sustained over 60 min and completely prevented by blockade of the NF-κB pathway with IκBα-DN (Fig. 4A and B). These results indicate that E2 alone cannot bring about ER recruitment to the BIRC3 gene and that activation of NF-κB is necessary to allow ER to be recruited by the combination of E2 and TNF-α. This mechanism appears to be gene specific since TNF-α can actually enhance (17, 43), repress, or have no effect on (see Fig. S3 in the supplemental material) ER recruitment to other ER target genes.

We next conducted a series of experiments to examine whether TNF-α influenced ER recruitment through chromatin remodeling around the ERE of the BIRC3 gene. Formaldehyde-assisted isolation of regulatory elements (FAIRE), micrococcal nuclease (MNase) sensitivity, and restriction enzyme accessibility assays all failed to detect any changes in chromatin architecture at the ERE, but chromatin remodeling was observed at the NF-κB-REs (see Fig. S4 in the supplemental material) (data not shown). However, TNF-α induced a robust enrichment of acetylated histones 3 and 4 (AcH3 and AcH4) around the ERE (Fig. 5A; see Fig. S5 and S6 in the supplemental material). A similar increase in acetylation throughout this region, as well as increased AcH4 nearer the transcription start site, was observed in response to E2 plus TNF-α (see Fig. S6). The effect of TNF-α on acetylation around the ERE was blocked by siRNA targeting p65 (Fig. 5B and C), indicating that TNF-α acting through NF-κB leads to an increase in histone acetylation at this site. This finding suggests that TNF-α-stimulated histone acetylation may be a potential mechanism by which NF-κB influences ER recruitment to the BIRC3 promoter.

To determine which histone acetyltransferases (HATs) may be involved in histone acetylation around the ERE and subsequent ER recruitment, siRNAs for a number of coactivators and HATs known to interact with ER and p65 were used (see Fig. S7 in the supplemental material). As expected, knockdown of ER and p65 reduced regulation of BIRC3. Of the HATs, we found that CBP significantly reduced BIRC3 expression in response to E2 plus TNF-α treatment (Fig. 6A). CBP knockdown, as well as ER knockdown, significantly reduced ER recruitment to the BIRC3 promoter in the presence of E2 plus TNF-α (Fig. 6B). Furthermore, knockdown of CBP significantly reduced AcH3 and AcH4 enrichment around the ERE (Fig. 6C). The presence of CBP at the BIRC3 promoter was observed by ChIP assay but was not regulated by E2 and/or TNF-α treatment (Fig. 6D). Taken together, these data suggest that CBP is required for regulation of BIRC3, and this may be due to its ability to acetylate histones around the ERE and promote ER access to the gene, thereby leading to potentiation of BIRC3 gene expression.

At first we expected that the lack of functionality of the ERE could be the result of its sequence. It has been demonstrated that the sequence of the ERE can affect the binding affinity of ER for its response element and in turn influences transcription of the target gene (18, 32). Furthermore, we have shown that cooperativity between ER and NF-κB occurs between a nonconsensus NF-κB-RE and an ERE for another gene whose expression is increased by E2 and TNF-α (43). However, the nature of the BIRC3 ERE is nearly identical to a consensus sequence, and conversion to a perfect consensus sequence does not significantly increase ER activity, suggesting another factor may be preventing ER from interacting with this binding site. This is in accordance with the finding that far more high-confidence ER binding sites have been identified computationally than are actually detected by genome-wide ChIP-on-chip studies for ER recruitment with E2 alone (6, 52).

We also considered that the endogenous chromatin structure may be preventing ER from interacting with its binding site. The fact that many transcription factor binding sites are occluded during DNA packaging has been proposed as a common mechanism to confer a higher degree of regulation to gene expression in response to various stimuli (11). Furthermore, previous studies have demonstrated that other transcription factors, such as FOXA1, are necessary to facilitate ER recruitment to specific regions of chromatin (26). NF-κB has previously been shown to enhance DNA accessibility around the NF-κB-RE to which it binds. For example, at the granulocyte-macrophage colony-stimulating factor gene (GM-CSF) promoter, NF-κB causes increase in chromatin accessibility via remodeling through recruitment of the ATPase component of the SWI/SNF chromatin remodeling complex, BRG1 (4, 22). More recently, constitutive activation of NF-κB along with AP1 has been associated with increased accessibility of the IL-6 promoter in highly invasive breast cancer cells (40). However, in the case of the BIRC3 gene, NF-κB had little to no effect on the chromatin structure around the ERE, suggesting that NF-κB may be influencing ER recruitment through another mechanism. We did observe a significant enrichment of acetylated histones around the ERE in the presence of TNF-α or E2 plus TNF-α, which suggested that NF-κB interaction with HATs may increase accessibility of the ERE to ER. Previous studies have shown that histone acetylation can contribute to DNA accessibility to transcription factors by loosening histone-DNA contacts (3).

In our study, a number of HATs were identified that are required for E2 plus TNF-α-induced expression of BIRC3, several of which are known to interact independently with ER or NF-κB. CBP was of particular interest to us since it was required not only for enhanced BIRC3 expression, but also for ER recruitment as well as histone acetylation around the ERE. Previous studies have shown that ER recruitment to the CRH promoter is accompanied by an increase in H3 acetylation, as well as CBP recruitment (34). CBP has also been shown to be important for NF-κB action on the BIRC3 gene (25). In MCF-7 breast cancer cells, CBP appears to be present constitutively at the BIRC3 promoter. This has also been shown for other genes where CBP is thought to play a role in the rapid induction of transcription (42, 46). For the c-Fos gene, CBP is constitutively present at a serum response element where it interacts with the transcription factor, Elk-1 (42). Similarly, CBP is constitutively present at the mitogen-activated protein kinase phosphatase 1 gene (MKP-1) promoter, potentially through interactions with Sp3 and CREB transcription factors (46). The BIRC3 promoter contains a large number of putative response elements in close proximity to the ERE, including binding sites for AP-1, SP1, PPAR, C/EBP, and STAT transcription factors. All of these are known to interact with CBP, but determining which factors are responsible for maintaining CBP at the BIRC3 promoter requires further investigation.

Notably, CBP has been described to play a role in both synergy and repression between transcription factors. CBP appears to be required for synergy between NF-κB and STAT1 on the CXCL10 gene (12), while several studies describe a role for CBP, or the related factor p300, in mutual transrepression between ER and NF-κB (20, 41, 50). These studies suggest that transrepression may occur through (i) a competition between ER and NF-κB for the limited amounts of cellular CBP or p300 or (ii) ER causing displacement of CBP from NF-κB target gene promoters (41, 50). One study suggested that a repressive complex involving all three players may occur. However, this was not supported by a mammalian two-hybrid study that demonstrated a synergistic action between ER, NF-κB, and CBP (20). In the case of BIRC3 regulation, we propose that CBP plays a synergistic role between ER and NF-κB, potentially because of the nature and arrangement of the ER and NF-κB binding sites. Loss of CBP is correlated with reduced histone acetylation and ER binding, suggesting that the function of CBP is to assist ER in accessing the ERE, potentially through increased histone acetylation. Alternatively, CBP could function to acetylate nonhistones, such as ER and p65. These posttranslational modifications are associated with increased DNA binding and activity of both transcription factors (9, 29) and therefore require further investigation. A third possibility is that CBP may play a scaffolding role, allowing formation of a more stable complex, consisting of ER, p65, and other HATs, which cause acetylation of histones around the ERE.