Signaling by estrogens



By regulating activities and expression levels of key signaling molecules, estrogens control mechanisms that are responsible for crucial cellular functions. Ligand binding to estrogen receptor (ER) leads to conformational changes that regulate the receptor activity, its interaction with other proteins and DNA. In the cytoplasm, receptor interactions with kinases and scaffolding molecules regulate cell signaling cascades (extranuclear/nongenomic action). In the nucleus, estrogens control a repertoire of coregulators and other auxiliary proteins that are associated with ER, which in turn determines the nature of regulated genes and level of their expression (genomic action). The combination of genomic and nongenomic actions of estrogens ultimately confers the cell-type and tissue-type selectivity. Recent studies have revealed some important new insights into the molecular mechanisms underlying ER action, which may help to explain the functional basis of existing selective ER modulators (SERMs) and provide evidence into how ER might be selectively targeted to achieve specific therapeutic goals. In this review, we will summarize some new molecular details that relate to estrogen signaling. We will also discuss some new strategies that may potentially lead to the development of functionally selective ER modulators that can separate between the beneficial, prodifferentiative effects in bone, the cardiovascular system and the CNS as well as the “detrimental,” proliferative effects in reproductive tissues and organs. J. Cell. Physiol. 213:610–617. © 2007 Wiley-Liss, Inc.

Estrogens comprise a group of structurally related, hormonally active molecules that regulate critical cellular signaling pathways and, by doing so, control cell proliferation, differentiation and homeostasis. The existence of estrogens, established between 1923 and 1938, opened important new opportunities for hormonal drug research and development. Estrogens occur in nature in several forms. In women with active menstrual cycles, the ovaries produce between 70 and 500 µg of 17β-estradiol (E2) daily, which is converted to estrone and to a lesser extent estriol. The two metabolites of E2, estrone and estriol, although bind ERα with high affinity (Kuiper et al., 1997), are much weaker agonists on estrogen receptors (ERs). These metabolites were previously thought to be inactive, but recent evidence suggests that they may have tissue-specific roles (Gruber and Huber, 1999).

Beginning in the late 1950s into the 1970s, the laboratories of Drs. Jensen and Gorski proposed that interactions between a receptor and E2 are involved in transducing the cellular effects of this hormone (Toft and Gorski, 1966; Jensen, 1973). Their laboratories also performed the isolation and initial characterization of ER (Toft and Gorski, 1966; Jensen, 1973). In 1985, soon after the glucocorticoid receptor (Hollenberg et al., 1982), the human ER cDNA (Walter et al., 1985) was cloned. Since then, multiple cDNAs encoding phylogenetically related members of the nuclear receptor family have been described (Committee NRN, 1999). Sequence analysis of the various receptor cDNAs demonstrated a high degree of similarity and led to their inclusion in a superfamily of nuclear receptors possessing a defined structure–functional organization (Evans, 1988). This growing family now includes receptors for the sex and adrenal steroids, thyroid hormones, retinoids, vitamin D3, and eicosinoids. Together with receptor-like proteins with no known ligand, termed orphan receptors, the family now includes 48 distinct members (reviewed in Evans (1988) and Mangelsdorf et al. (1995)).

In vitro expression of the wild type and mutated ER in mammalian (Kumar and Chambon, 1988) and yeast (Metzger et al., 1988) cell systems in combination with cell-free in vitro transcription (Lees et al., 1989a,b) systems and their functional analysis allowed detailed structure–function characterizations/mapping of the distinct functional domains of ER (Kumar et al., 1987; Lees et al., 1989a,b; Danielian et al., 1992). Similar to other nuclear receptors, the N-terminal region (A/B domain) of the ER contains a constitutively active transactivation region (AF-1). The most conserved region is the DNA-binding domain (DBD, C domain), which contains the P-box, a short motif responsible for DNA-binding specificity to a sequence typically containing a palindromic hexanucleotide 5′-AGGTCA-3′ motif, and is involved in ER dimerization. The DBD is followed by a less conserved region (D domain) that behaves as a flexible hinge between the C and E domains, and contains the nuclear localization signal (NLS), which extends into the C domain. The largest domain is the moderately conserved among nuclear receptors ligand-binding domain (LBD, E domain), whose secondary structure of 12-helices is better conserved than the primary sequence. The E domain is responsible for many functions, which are primarily regulated by ligand binding. The AF-2 transactivation function, a dimerization interface, an additional NLS, and a repression function are associated with this domain. The very C-terminal part of ER molecule also contains the F domain, which exerts a complex modulatory role on activity of the receptor (Montano et al., 1995; Koide et al., 2007).

In the 1990s, significant progress in nuclear receptor biology was made by the discovery of a large number of coregulator proteins that modulate the transcriptional activity of nuclear receptors. The involvement of coregulator proteins in receptor signaling was first postulated when nuclear receptors (NRs) were found to functionally cross-react (squelch) with each other and other classes of transcription factors (Schule et al., 1990). Since then, biochemical and genetic approaches have been used to identify and clone numerous nuclear receptor associated proteins, called coregulators.

In late 1990s, multiple laboratories reported the discovery of a second type of ER in the rat (Kuiper et al., 1996), mouse (Tremblay et al., 1997), and human (Mosselman et al., 1996). This newly discovered receptor was named ERβ, with the classical ER being referred to as ERα. This discovery raised the possibility of cell-specific gene targets and models of action that involved cooperation as well as potential competition between the two ER subtypes. The two receptors are not isoforms of each other, but rather different proteins encoded by separate genes located on different chromosomes. As the role of ERβ remains less well defined (subject of a number of recent publications (reviewed in Heldring et al. (2007)), the focus of this review will be ERα.

Sixty-five years ago, it was first reported that steroid hormones might have very rapid action on cells, too rapid to involve transcriptional regulation (Selye, 1942). Now it is well documented that in addition to direct regulation of gene expression (genomic action), steroid hormones regulate cell-signaling phosphorylation cascades. This process is insensitive to inhibitors of RNA and protein biosynthesis and, in some cases, can take place in the absence of a nucleus, with isolated cell membranes or enucleated cytoplasts (Welshons et al., 1988). All members of the steroid hormones, from the corticosteroids (glucocorticoids and mineralocorticoids) to the gonadal hormones (estrogens, progestins, and androgens), vitamin D3 and thyroid hormone can exhibit nongenomic effects. These effects range from activation of adenylyl cyclase (AC), mitogen-activated protein kinases (MAPKs), phosphatidylinositol 3-kinase (PI3K), to rise in intracellular calcium concentrations (for some recent reviews see Watson and Gametchu, 2003; Boonyaratanakornkit and Edwards, 2004; Cheskis, 2004; Norman et al., 2004; Shupnik, 2004; Levin, 2005; Lange et al., 2007). In recent years, significant progress has been made in understanding the molecular mechanisms of the nongenomic action of the steroid/nuclear receptors. Major questions that remain to be addressed are the nature of the receptors that are responsible for certain aspects of nongenomic action; the molecular mechanisms that integrate hormonal action in the regulation of cell signaling pathways, and most importantly, the physiological role of rapid nongenomic actions of steroid hormones.

Nuclear Action

ERE-dependent/independent transcriptional regulation

In common with other transcription factors, ERα and β stimulate transcription by recruiting a pre-initiation complex. According to the “classical model,” “ligand-activated” ERα binds to specific DNA sequences, as in the case of the prolactin (Maurer and Notides, 1987), progesterone receptor (Savouret et al., 1991), or c-fos (Weisz and Rosales, 1990) genes, and interact with basal transcription factors and coregulator proteins, which ultimately regulate the transcription of target genes (Mangelsdorf et al., 1995; Glass et al., 1997; McKenna et al., 1999; McKenna and O'Malley, 2002; O'Malley, 2007). However, in recent years, mechanisms of gene regulation by ERs that deviate from this “classical model” have also been described. These include gene regulation by ERs that does not involve direct receptor binding to DNA, but rather via ERs participation in the formation of the pre-initiation complex via protein–protein interactions, such as the AP1 complex (Johnston et al., 1999; Webb et al., 1999), as in the case of the collagenase (Paech et al., 1997) and IGF-I (Umayahara et al., 1994) genes. In addition, ERs may compete for the interaction with other transcription factors, which may affect target gene expression (Meyer et al., 1989; Schule et al., 1990; Sheppard et al., 1999).

Although the classical ER signaling pathway has been well characterized at the molecular level, nonclassical ER signaling is less well understood. A number of studies have recently been carried out in vitro to understand the molecular basis of ER action at AP-1 sites (Kushner et al., 2000; Jakacka et al., 2001). Recently, using an in vitro assay Cheung et al. (2005) have described altered pharmacology and distinct coactivator use by ERα at AP1 compared with ERE sites, further adding to the increasing complexity of ER signaling. However, the exact mechanisms of ERα actions at AP1 sites (or for that matter at any other site involving ER interactions with other transcription factors) remains a topic of debate, with different studies describing somewhat varying in vitro results (Paech et al., 1997; Cheung et al., 2005).

To assess the relative importance of classical ERα signaling, a nonclassical ERα knock-in (NERKI) mice model was recently generated. To date, this model has been used to characterize the in vivo roles of the ERE-independent pathway in the uterus (O'Brien et al., 2006), bone (Jakacka et al., 2002) and in the female hypothalamic–pituitary–gonadal axis (Glidewell-Kenney et al., 2007). These mice express an ERα, which carries two amino acid substitutions (E207A/G208A) in the first zinc finger of the DBD. In vitro, this mutant receptor failed to activate reporter constructs containing EREs, but was active in regulating transcription from an AP1 site (Jakacka et al., 2001). Results from these studies suggest that E2 action is mediated by the complex crosstalk between ERE-dependent and independent ERα signaling pathways, and their alterations can result in a markedly aberrant response to E2 (Syed et al., 2005, 2007).


NRs exist as large multiprotein complexes with coregulator molecules. Based on their ability to stimulate or inhibit transcription, coregulators can be subdivided into two classes. Coactivators enhance and corepressors repress transcription. Current evidence indicates that this definition can be modified by gene, cell, and signaling context for any one coregulator (subject of numerous comprehensive reviews (Spiegelman and Heinrich, 2004; Margueron et al., 2005; Rosenfeld et al., 2006; O'Malley, 2007) to name a few).

The structural determinants of coregulator–NR interaction have been well defined. The ER LBD, similar to other NRs, consists of a three-layered, antiparallel, α-helical sandwich in which a central core layer of three helices packed between two additional layers of helices forms the ligand-binding cavity. An additional helix required for ligand-dependent transcriptional activation (AF2) resides at the C terminus of the LBD and adopts different positions depending on the presence or absence of ligands (Brzozowski et al., 1997). In the presence of agonists, the activation helix 12 (H12) is configured to form a “charge clamp” in which a conserved glutamate in the AF2 helix and a conserved lysine in helix 3 of the LBD grip the ends of helical motifs that contain an LXXLL consensus sequence present in most coregulators (Heery et al., 1997). The leucine residues of the LXXLL helix pack into a specific hydrophobic pocket at the base of the charge clamp that stabilizes the interactions (Darimont et al., 1998; Shiau et al., 1998). Many coregulator molecules contain several LXXLL motifs, which may be used in a nuclear receptor-specific fashion (McInerney et al., 1998; Cheskis et al., 2003). Similarly, corepressors that include the nuclear receptor corepressor (NCoR) and silencing mediator of retinoic acid and thyroid hormone receptor (SMRT) interact with unliganded nuclear receptors, including ER, through an elongated helix of sequence LXX I/H IXXX I/L, alternatively referred to as the CoRNR-box (Nagy et al., 1997; Hu and Lazar, 1999; Webb et al., 2000). This extended helix can occupy the same hydrophobic pocket contacted by LXXLL motifs in the absence of agonist binding because it can displace the H12 (helix 12, of AF-2 domain). In contrast, the extended helices of NCoR/SMRT are too long to be accommodated by this pocket when the H12 assumes the charge clamp configuration in response to agonist binding. Thus, agonist binding reduces the affinity of nuclear receptors for CoRNR-box-containing corepressors and increases affinity for LXXLL-containing coactivators.

The interactions of coregulators with NRs can be controlled at several levels, including coregulator expression, post-translational modifications, and ligand binding. For many coregulators there is evidence for genome-wide competition for their recruitment to specific DNA-binding factors, which is an important factor that determines gene expression. One of the well-known examples of the “limiting” levels of a cofactor is in the Rubenstein-Tabi syndrome, in which haploinsufficiency for CBP results in severe developmental/regulatory abnormalities (Robert and Miller, 1995). In the case of SRC-3, six phosphorylation sites are required for coactivation of ERs, but not all of these sites are required for coactivation of NFkB (Wu et al., 2004). Phosphorylation of several residues of SRC3 is required for its interaction with CBP (Torchia et al., 1997; Chen et al., 1999; Wu et al., 2004). Furthermore, different combinations of site-specific phosphorylations of SRC-3 are necessary for regulation of endogenous genes involved in inflammation or transformation. Biochemical studies support the concept that modulation of SRC3 phosphorylation alters its interactions with potential activator/coactivator partners, allowing it to function as an integrator for diverse signaling pathways.

Evaluation of the kinetics of promoter occupancy by NRs has revealed that, there is a cyclical pattern of coregulator recruitment and release in the presence of E2. In the case of ERα binding to the pS2 promoter, ERα turnover was observed with a cycle time of approximately 40 min (Metivier et al., 2003). Furthermore, recycling of liganded ERα on the pS2 promoter was dependent on proteosome activity (Reid et al., 2003). Similarly to ERα, there was a specific order of recruitment/release of the p160 factors, HATs, TAFs, Mediator, ASC2, PARP1, Mediator, Pol II, chromatin remodeling complexes; and methyltransferases, Mi2/HDACs/NcoR (Metivier et al., 2003). More recently, the ChIP-microarray (ChIP-chip) approach that covers the entire nonrepetitive sequence of chromosomes 21 and 22, has been utilized to identify ER binding sites on a chromosome-wide scale (Carroll et al., 2005). Interestingly, these binding sites represent only a small fraction of predicted EREs, confirming that the presence of an ERE within DNA is insufficient to determine an ER binding site.

These results suggest that the enzymatic modification of histone and transcription factors, strictly controlled in signal, time and space dependent fashion, leads to changes in chromatin structure, which determine transcriptional regulation.

Extranuclear/Nongenomic Action of Estrogens

Receptors that mediate the nongenomic action of steroid hormones

There is substantial evidence that a subpopulation of conventional steroid hormone and vitamin D receptors, mediate rapid effects of hormones on cell signal transduction pathways. Several lines of evidence support this conclusion. In experiments with cultured cells, rapid activation of various signaling pathways by all classes of steroid hormones and vitamin D has been shown to be dependent on conventional steroid/NRs either by reconstitution experiments with receptor negative cell lines, by knock down of receptors with siRNA or antisense RNAs, by the use of highly specific steroid receptor antagonists, or by studies with receptor knock out mice (reviewed in Edwards (2004)). For example, the onset of rapid electrical responses to vitamin D3 were recently shown to be lost in primary osteoblast cells derived from VDR knock out mice (Zanello and Norman, 2004). At the same time novel membrane receptors unrelated to conventional steroid receptors have also been implicated. An orphan member of the G protein-coupled receptors (GPCR) superfamily, termed GPR30, has been reported to act independently of classical ERs to trigger rapid signaling by estrogens (Filardo et al., 2000; Revankar et al., 2005). E2 treatment of GPR30 transfected cells that lack ERs expression caused activation of a stimulatory G protein (Gs) that is directly coupled to this receptor and increased adenylyl cyclase activity (Thomas et al., 2004). GPR30 is localized to endoplasmic reticulum and binds E2 with nanomolar affinity (Revankar et al., 2005). However, while siRNA to ERα prevented rapid signaling in MCF7 and endothelial cells, GPR30 knockout or overexpression had no effect on E2 action (Pedram et al., 2006). The biological relationship between GPR30 and conventional ERs is currently unknown; further studies are required to determine the relative contribution of either or both of these pathways to estrogen signaling. It is possible however that the rapid E2 signaling is mediated by a complex network of proteins that consists of conventional steroid receptors and other steroid binding proteins such as GPR30 (Cheskis, 2004).

Membrane localization of estrogen receptors

Although the majority of ERα is localized in the nucleus, there is evidence that a small fraction of the receptor is localized at or near the cell membrane in either the presence or absence of E2. Immunocytochemical staining has demonstrated positive staining of ERα at the plasma membrane of different cells (Watson et al., 1999; Zivadinovic and Watson, 2005). Confocal microscopy showed that E2 treatment of MCF7 cells rapidly induced membrane ruffles, pseudopodia and translocation of ERα to the cell membrane (Levin, 2002; Figtree et al., 2003; Li et al., 2003). Endogenous ERα was biochemically isolated from plasma membranes and caveolae fractions of endothelial cells and E2 stimulated signaling in these isolated membrane fractions (Levin, 2002).

While the precise mechanism(s) are currently unknown, it has been proposed that ERα translocation to cell membrane is mediated by its interactions with membrane proteins. Candidate interacting proteins include caveolin-1/-2 and the 110-kDA caveolin-binding protein–striatin. Caveolae are specialized regions of the plasma membrane that assemble and organize signaling protein complexes (Anderson, 1998). Endogenous ERα has been reported to interact with caveolin-1 and -2 in an E2-dependent manner in MCF-7 and in vascular smooth muscle cells. Overexpression of caveolin-1 in MCF-7 cells has been shown to increase the E2-dependent ERα translocation to the plasma membrane (Razandi et al., 2002).

Striatin is a calmodulin-binding member of the WD-repeat family of proteins. It has been reported to anchor ERα to the cell membrane, and to serve as a scaffold for the formation of an ERα–Gαi complex, which is critical for E2 activation of eNOS (Lu et al., 2004). In contrast with these results, using pull-down experiments with purified recombinant proteins, direct interactions of ERα with Gαi and Gßγ were also documented (Kumar et al., 2007).

It has been also proposed that ERα can be targeted to the cell membrane by the adaptor protein Shc (Pelicci et al., 1996). The SH2 domain of Shc can directly interact with the N-terminal part of ERα (Song et al., 2002). ERα, Shc and insulin-like growth factor type 1 receptor (IGF1R) have been shown to interact on the cell membrane of MCF7 cells through Shc binding to phosphorylation sites of the intracellular domain of the IGF1 receptor. Further supporting the importance of these interactions for ERα translocation to the cell membrane, treatment of MCF7 cells with siRNA for Shc, or IGFR1, attenuated the E2-induced ERα translocation to the cell membrane and E2 stimulation of MAPK phosphorylation (Song et al., 2004).

Recently another membrane adaptor protein the p130Cas (Crk-associated substrate) has also been reported to interact with ERα-cSrc complex in T47D breast cancer cells and to potentiate the E2 activation of Src (Cabodi et al., 2004).

Finally, palmitoyl acyl transferase (PAT)-dependent S-palmitoylation of ERα was recently reported to promote ERα association with the plasma membrane and interaction with caveolin-1 (Acconcia et al., 2004, 2005). Mutation of the palmitoylation site in ERα (Cys 447) amino acid or inhibition of palmitoylation with 2-bromo-palmitate resulted in a significant decrease in receptor localization at the plasma membrane. Furthermore, cystine 447-mutated ERα did not stimulate activation of MAP and PI3 kinases (Acconcia et al., 2005). A terminally truncated 46 kDa variant of ERα has been found to be preferentially palmitoylated and enriched in plasma membrane of several cell types (endothelial, osteoblasts, and MCF-7 cells) (Denger et al., 2001; Marquez and Pietras, 2001; Longo et al., 2004). Inhibitors of palmitoylation blocked membrane localization of 46 kDa ERα. It has been suggested that the truncated ERα through altered protein folding may expose sites for fatty acid acylation that are not accessible in full-length 66 kDa ERα (Li et al., 2003).

In conclusion, several membrane proteins have been identified that interact with classical receptors and influence their nongenomic action. However, the precise role of these proteins in receptor regulation of the cell signaling remains to be further investigated. It is possible that the composition of ER complexes at the plasma membrane is cell context dependent, which may potentially explain the cell type selectivity of nongenomic action.

ER activation of cell signaling molecules

One of the most interesting questions that remain to be answered is how the conformational changes in receptor molecules induced by the binding of steroid hormones are converted into activation of cell signaling molecules. The nature of the upstream receptor targets remains to be better established. Multiple lines of evidence suggest that activation of the tyrosine kinase cSrc represents one of the initial steps in ERα-mediated cell signaling in MCF7 cells (Migliaccio et al., 2002). The essential role of Src kinase in the nongenomic action of steroid receptors was demonstrated in experiments with embryonic fibroblasts derived from cSrc−/− mice. These cells did not show rapid activation of the MAP kinase pathway in response to AR and ERα activation, whereas wild-type cSrc+/+ cells did (Kousteni et al., 2001).

cSrc can be activated either by dephosphorylation of the C-terminal inhibitory phosphotyrosine site (or in oncogenic variants by loss of the C-terminal tail), or by binding of high-affinity ligands to the SH2 or SH3 domains. These domains are modular polypeptide units that mediate protein–protein interactions and are found together on many proteins, suggesting that their activities can be coordinated and that they can cooperate in Src regulation (Cohen et al., 1995; Hubbard et al., 1998).

It has been initially proposed that ERα can activate Src by a direct interaction with the SH2 domain (Migliaccio et al., 1998; Ballare et al., 2003). However, later it was found that although ERα was capable of interacting directly with the SH2 domain of Src in E2 dependent manner (Migliaccio et al., 2000; Castoria et al., 2001), this interaction did not appear to be sufficient for E2 induced activation of Src (Boonyaratanakornkit et al., 2001; Wong et al., 2002). An adaptor protein Modulator of Nongenomic Action of Estrogen Receptor (MNAR) has been identified that is required for E2 induced ERα activation of cSrc and downstream MAP kinase pathway (Wong et al., 2002). MNAR is homologous to a protein that had been previously isolated by pull-down with the Src homology domain 2 (SH2) of p56lck (Lck) (Joung et al., 1996). The protein, referred to as p160 (Joung et al., 1996), was later designated as PELP1 (proline-, glutamic acid-, leucine-rich protein) (Vadlamudi et al., 2001). MNAR/PELP1 is a ∼120 kDa scaffold protein that contains multiple protein–protein interaction domains. The N-terminal portion of the MNAR molecule contains 10 LXXLL motifs similar to those in p160 family of coactivators that can potentially mediate hormone agonist-dependent interaction with LBDs of nuclear receptors (Heery et al., 1997), and 3 PXXP motifs that are similar to SH3 domain interaction sequences. Purified MNAR alone simulates cSrc enzymatic activity. However, purified ERα and MNAR together synergize to produce a strong E2 dependent activation of cSrc (Wong et al., 2002). Interaction between endogenous ERα, MNAR and cSrc was demonstrated using co-immunoprecipitation from the cell extracts of MCF7 cells. As evidence that MNAR and ERα cooperate to activate Src in intact cells, overexpression of MNAR enhanced E2 stimulation of Src enzymatic activity and phosphorylation of MAPK in MCF-7 cells, while expression of antisense oligonucleotides to MNAR attenuated E2 activation of the Src/MAPK pathway (Wong et al., 2002). Mutational analysis and functional evaluation of MNAR and the use of ERα and cSrc mutants revealed that MNAR interacted with the SH3 domain of Src via its N-terminal PXXP motif (designated PXXPP motif #1). Mutation of this motif abolished both the MNAR-induced activation of Src/MAP kinase pathway. ERα interacted with Src SH2 domain using phosphotyrosine 537 and this complex was further stabilized by MNAR-ERα interaction. The region responsible for the MNAR interaction with ERα was maped to the two N-terminal LXXLL motifs of MNAR (designated as LXXLL motifs 4 and 5). Mutation of these motifs prevented ERα-MNAR complex formation and eliminated activation of the Src/MAP kinase pathway (Barletta et al., 2004).

It has been recently demonstrated that in MCF7 cells treated with E2, endogenous MNAR, ERα and cSrc also interacted with p85, the regulatory subunit of the PI3 kinase. Further, ERα-MNAR activation of cSrc, led to MNAR phosphorylation on Tyr 920 which was required for its interaction with SH2 domain of p85 (Greger et al., 2007) and activation of the PI3K/Akt pathway. Mutation of tyrosine 920 to alanine (Y920A) abrogated the interaction between MNAR and p85 and the E2-induced activation of the PI3K/Akt pathway, which was required for the E2 protection of MCF7 cells from apoptosis.

Presence of multiple LXXLL motifs suggests that MNAR can potentially interact with multiple NRs. Indeed, MNAR also interacts in a hormone-agonist dependent manner with several other NRs including AR, GR, PR, and VDR (Wong et al., 2002; Barletta et al., 2004). However, it is not clear whether all NRs would require MNAR for the activation of cell signaling pathways. Progesterone receptor beta (PRβ), for example, can directly interact with the SH3 domain of Src and this interaction is sufficient for its activation of cSrc (Boonyaratanakornkit et al., 2001). Existing data indicate that MNAR is a scaffold, which is promoting receptor binding to Src and stabilizing ERα-Src complex. Therefore, it is reasonable to postulate that the affinity of ERα binding to cSrc-MNAR complex is higher than that of ERα binding to Src alone. Thus, formation of this complex can take place at lower concentrations of ERα, Src, and E2. Some receptors however may not require an adaptor molecule, because they may interact with cSrc with higher affinity, or their expression level is sufficient (for recent review see Edwards, 2004).

Crosstalk with growth factors signaling

Crosstalk between growth factors and ERs takes place in both nuclear and cytoplasmic compartments. ERα transcriptional activity can be regulated by ligand-independent mechanisms, via ERα and coregulator phosphorylation (Kato et al., 1995) (reviewed in Cenni and Picard (1999)). In addition, both EGF and IGF can also activate ERα transcriptional activity in the presence of E2 (Vignon et al., 1987; Ignar-Trowbridge et al., 1992; Hewitt et al., 2005). Treatment with EGF or insulin-like growth factor-I (IGF-I) resulted in increase in uterine weight and proliferation of the uterine epithelial cells of ovariectomized mice. Importantly, these effects were not observed in ERα knockout mice, suggesting that ERα is required for growth factor action (Hewitt et al., 2005). In addition as discussed above, E2 may rapidly activate the two main signaling cascades coupled to the IGF-I and the EGF receptors: the PI3K/Akt and the Src/MAPK signaling pathways. Initially an indirect mechanism of ERα activation of epidermal growth factor receptor (EGFR) has been proposed to explain these effects. According to this hypothesis, ERα bound to caveolin-1 in the cell membrane could interact with a G-protein-coupled receptor, which in response to estrogen- or tamoxifen-binding may directly or indirectly interact with and activate specific G proteins. The subsequent activation of cSrc leads to activation of matrix metalloproteinases, which in turn cleave heparin-binding epidermal growth factor (EGF) from the membrane. This form of EGF then binds to surface EGFR in an autocrine or paracrine manner to activate the receptor and its downstream kinases including ERK 1/2 MAPK and Akt (Levin, 2002, 2003).

Recent evidence however indicates that ERα and Src may play a direct role in EGF activation. In MCF7 cells, EGF stimulation of cell proliferation was blocked by pure antiestrogen, ICI 182,780. Similar effect was also observed in the presence of the Src kinase family inhibitor, PP2, as well as in cells transiently transfected with siRNA for ERα. ICI 182,780 also blocked EGF activation of the Src/MAPK signaling pathway in MCF7 cells (Migliaccio et al., 2005). Importantly, treatment of MCF7 cells with EGF induced association of ERα, AR, MNAR with Src and EGFR, which was blocked by selective inhibitors of the EGFR tyrosine kinase, Iressa (ZD 1839), and the anti-erb-B2 antibody herceptin (Trastuzumab) (Migliaccio et al., 2005). Direct interaction has also been documented for ERα and IGF receptor both in vitro (Kahlert et al., 2000) and in vivo (Mendez et al., 2003). The interaction was coincident with the increase in tyrosine phosphorylation of IGF-I receptor, suggesting a possible causal relationship (Kahlert et al., 2000; Mendez et al., 2003).

Existing results therefore suggest that the intracellular signaling by estrogens and growth factors are closely interdependent and may be mediated by the direct interactions of their cognate receptors and/or the receptor' auxiliary proteins. It is also possible that the composition and potentially the dynamics of these complexes are cell-type and signal specific.

Functional consequences of steroid activation of cell signaling pathways

One of the best-characterized extranuclear actions of estrogens is the rapid activation of the Ras/Raf/MAPK pathway. In neuronal cells, E2 rapidly triggers Erk 1/2 activation, leading to cFos gene expression (Watters et al., 1997). Rapid activation of this pathway was also found in osteoblasts (Endoh et al., 1997) and white adipose tissue (Garcia Dos Santos et al., 2002). E2 activated growth of human colon carcinoma-derived Caco-2 cell was found to be mediated through rapid and reversible stimulation of the cSrc and cYes, and subsequent activation of ERK1 and ERK2 kinases (DiDomenico et al., 1996). In the MCF-7 human breast cancer cell line, E2 triggered rapid increase in the active form of p21ras, rapid tyrosine phosphorylation of Shc and p190, and association of p190 with the guanosine triphosphatase (GTPase) activating protein (GAP). Both Shc and p190 are substrates of activated cSrc, and once phosphorylated, can interact with other proteins and stimulate p21ras. E2-mediated stimulation of Ras/Raf/ERK pathway promotes MCF7 cell proliferation (Migliaccio et al., 1996). The MAPK pathway is involved in the control of many fundamental cellular functions that include cell proliferation, survival, differentiation, apoptosis, motility, and metabolism. Some of these functions are mutually exclusive, such as E2 triggered proliferation in MCF7 cells (Migliaccio et al., 1996) versus cell-cycle arrest and differentiation in osteoblasts (Chen et al., 2004). Activation of the MAPK pathway by E2 exerts anti-apoptotic effects on osteoblasts/osteocytes, but pro-apoptotic effects on osteoclasts. It has been postulated that the kinetics of ERK phosphorylation and the length of time that phospho-ERKs are retained in the nucleus determine the pro-versus anti-apoptotic effects of E2 (Chen et al., 2004). It has long been recognized that transient and sustained signaling from the Ras/ERK pathway can lead to the different biological outcomes of proliferation and differentiation, respectively (reviewed in Marshall (1995)).

A well-characterized and biologically important action of E2 is the acute effect on blood vessels to stimulate vasodilation and protect against vascular injury. This action is mediated by a subpopulation of ERα in plasma membrane of endothelial cells through the activation of eNOS and the stimulation of NO production via the PI3 kinase/Akt signaling pathway. cSrc, which is upstream of PI3K, also appears to be important. As evidence of the biological importance of this action of E2, mice treated with E2 show increased eNOS activity and decreased vascular leukocyte accumulation after ischemia and reperfusion injury in a manner dependent on PI3K and eNOS. ERα knockout mice lost the acute protective effect of E2 on the vascular injury response, which indicates that conventional receptor mediated this rapid effect of E2 (Karas et al., 2001). One of the important down stream targets of PI3K is the threonine-serine kinase Akt/protein kinase B. Activation of PI3K/Akt by E2 has also been shown to be important in breast cancer cells in mediating E2-stimulation of cell cycle progression (Castoria et al., 2001) and inhibition of apoptosis (Campbell et al., 2001).

Many cell-signaling pathways converge upon and regulate the phosphorylation status and hence activity of multiple transcription factors, which affects gene expression. Several examples of this mode of regulation have been reported, including ERα-dependent E2 regulation of the c-fos gene mediated by Src/MAP and Src/PI3 kinase pathways converging on Elk-1 and SRF, respectively; E2 regulation of cyclin D1 mediated by PI3K/Akt pathway and E2 regulation of the Egr-1 gene mediated by MAP kinase activation of SRF (Duan et al., 2001, 2002).

Protein phosphorylation cascades rapidly stimulated by steroids also play an important role in gene regulation by affecting receptor stability and transcriptional activity. PR and RARγ2 undergo ligand-dependent degradation mediated by the ERK and p38 pathways, respectively (Lange et al., 2000; Gianni et al., 2002). ERα is phosphorylated on multiple serine/threonine residues in the N-terminus by MAPK and other kinases, and these phosphorylations are important for intrinsic transcriptional activity of the receptor (Feng et al., 2001; Rochette-Egly, 2003). Members of p160 family of steroid receptor coactivators, SRC1 and GRIP1, are also direct targets of MAP kinases. For both SRC1 and GRIP1, ERK pathway activation led to enhanced coactivation function (Rowan et al., 2000; Lopez et al., 2001).

By affecting many intracellular pathways (Fig. 1), estrogens mediate a broad spectrum of physiologic functions ranging from regulation of the menstrual cycle and reproduction to the modulation of bone development and cholesterol transport. At present, our basic understanding of how different aspects of their action in vitro manifest into physiological activities is very fragmentary. However, the discovery of these connections may open the door for rational drug discovery to create ligands of ERs with site-specific action; ligands that would antagonize ER action in female reproductive organs, but would protect from bone loss, and not interfere with the beneficial effects of E2 in the brain.

Figure 1.

Signaling by estrogens is mediated by a complex interface of direct control of gene expression (genomic action) and by regulation of cell signaling/phosphorylation cascades (nongenomic action). Both the genomic and the nongenomic aspects of E2 action are interdependent. It is feasible however that some genes, and because of that some manifestations of E2 action, can be regulated purely via genomic, or nongenomic mechanisms.


Functionally selective estrogens

For years, estrogens have been successfully used to relieve the symptoms of menopause in women. Furthermore, hormone replacement therapy (HRT) has also been prescribed for desired long-term benefits extending beyond menopausal symptom relief. In addition to the prevention of bone loss, favorable changes in serum lipid levels and other markers of cardioprotection (Mendelsohn and Karas, 2007) as well as changes in cognitive functions (reviewed in Zhao et al. (2005)) have been attributed to HRT. However, long-term treatment with estrogens has been known to increase the risk of breast and endometrial cancers. This issue has been extensively discussed based on data from two randomized clinical trials, HERS (Hulley et al., 1998) and the Women's Health Initiative (Writing Group for the Women's Health Initiative I, 2002).

These trials have further incited an intense search for therapies related to estrogens that might provide significant benefits, but avoid some or all of the negative aspects associated with HRT treatment. A great deal of interest has been dedicated to a group of compounds called selective estrogen receptor modulators (SERMs), because they presented important evidence that the actions of ER ligands can be cell-type and tissue-selective. Initially SERMs were developed as ER antagonists that block its proliferative action in reproductive tissues and organs. One of these compounds, tamoxifen, has well-established antiproliferative action in the breast. In addition, both the skeletal and the cardiovascular benefits of tamoxifen have been demonstrated in postmenopausal women who receive the drug either as primary or adjuvant therapy for breast cancer (Love et al., 1992; Rutqvist and Anders Mattsson for the Stockholm Breast Cancer Study, 1993). These potential benefits have led to studies of tamoxifen in a chemoprevention setting in normal women. However, this approach has not been approved, largely due to the adverse effects of tamoxifen in the uterus (Force USPST, 2002). Another antiestrogen, raloxifene, which was also developed because of its antiproliferative action, also showed skeletal antiresorptive properties (Sarkar et al., 2002). Importantly, there is no evidence of an increased risk of ovarian or endometrial cancers associated with raloxifene treatment (Goldstein et al., 2000). One of the raloxifene side effects, however, is increased incidence of hot flashes (Davies et al., 1999).

Currently, it is not clearly understood how SERMs, such as tamoxifen and raloxifene, can be antiestrogenic in some cells and estrogenic in others. Most of the unique pharmacology of SERMs can perhaps be explained by several interdependent mechanisms. (1) Upon binding to receptor, they stimulate different conformational changes. (2) Both in the cytoplasm and in the nucleus, they impose different affinities on ER interactions with other proteins, which lead to the activation of different nongenomic/genomic cellular programs and the expression of different sets of genes (Zajchowski et al., 2000; Frasor et al., 2006). (3) They may have different effects on ER proteasomal stability, or (4) they may be ERα, or ERβ selective.

Recent developments shed light on cellular mechanisms that are responsible for the proliferative action of estrogens in breast cancer (Castoria et al., 2004; Migliaccio et al., 2005, 2006; Kininis et al., 2007; Stender et al., 2007) and in uterus (O'Brien et al., 2006; Groothuis et al., 2007; Naciff et al., 2007). Some new insights are also emerging on the prodifferentiative and anti-apoptotic mechanisms of estrogens action in bone (Aguirre et al., 2007), cardiovascular system and CNS (Karas et al., 2001; Lu et al., 2004; Turgeon et al., 2006). With this new data, it is becoming increasingly clear that most of the therapeutic benefits can be obtained with SERMs that may separate between the beneficial, prodifferentiative effects of the hormone in the cardiovascular system, bone and in CNS, and their “detrimental,” proliferative effects, in mammary gland and uterus.

To develop new drug discovery approaches, it is crucial to understand which pathways are involved in actions of estrogen, and how their effects are integrated into the balance of risks and benefits. Development of these “ideal” estrogens may provide unique therapeutic opportunities for the treatment and prevention of human diseases.