I. Introduction

Top Next References

In 1958, Lerner and coworkers published their landmark paper on the pharmacological properties of the first nonsteroidal antiestrogen ethamoxytriphetol or MER-25 (fig. 1). Lerner later wrote, "the compound was appealing not only because it completely inhibited the uterine response to estradiol (E₂)^b but also because it was devoid of uterine stimulatory properties. This was an added bonus. Here was a possible tool for the study of oestrogen requirements and involvement in bodily functions. Was the inhibition of estrogenic activity competitive or noncompetitive? Various doses of MER-25 were studied against a single dose of oestradiol benzoate, and various doses of the oestrogen were studied against a single dose of the antagonist. The results of these studies demonstrated dose response relationships with competitive antagonism" (1981).

View larger version (8K): [in this window] [in a new window]   Fig. 1.   The first nonsteroidal antiestrogen, MER 25, was never developed clinically because of high toxicity and low potency. The triphenylethylene compound, MRL-41, or clomiphene developed in 1961 is now the standard therapy for infertility. The triphenylethylene compound, tamoxifen, is the endocrine treatment of choice for the treatment of breast cancer.

However, in the 1950s, the main roadblock to further progress was that no one knew how estrogen produced its effects. The target site-specific actions of estrogen on the reproductive system had been well known since the work of Allen and Doisy (1923), who identified and assayed ovarian "estrus-stimulating" hormones. But why did one tissue like the uterus and vagina respond to estrogen whereas another like muscle did not? Early studies with ¹⁴C-labeled hormones could not detect any target site localization (Twombly and Shoenewaldt, 1951; Hanahan et al., 1953). It subsequently would be discovered that the specific activity was too low, and tritium-labeled compounds, with high specific activity, would be necessary for success.

In 1959, in Vergennes, Vermont, Gregory Pincus and Erwin Vollmer organized a conference sponsored by the Cancer Chemotherapy National Service Center of the National Cancer Institute, entitled "Biological Activities of Steroids in Relation to Cancer." Jensen and Jacobson, from the Ben May Laboratory for Cancer Research at the University of Chicago, had synthesized [6,7-³H]E₂ and for the first time illustrated the target tissue specificity of a natural hormone. They injected [³H]E₂ into immature female rats and noted that the radioactivity, which they proved was E₂, was bound in and retained by estrogen target tissues (uterus, vagina, and anterior pituitary) but was not retained in nontarget tissues (muscle, kidney, and liver). These pivotal studies (Jensen and Jacobson, 1960) opened the door for the subsequent identification and study of steroid receptors. At the meeting, Dr. Gerald Mueller, then Lasker Professor of Cancer Research at the McArdle Memorial Laboratory at the University of Wisconsin, commented, "Dr. Jensen, you certainly have filled a tremendous gap in the information that we have wanted for a long time; that is, the state of hormones in the tissue during response to hormone. This beautiful work is an example of experimentation executed with good command of organic chemistry and good knowledge of the biological picture" (Mueller, 1960).

The study of estrogen and antiestrogen action converged when Pincus, the father of oral contraceptives and then Director of the Worcester Foundation for Experimental Biology (now the Worcester Foundation for Biomedical Research) in Shrewsbury, Massachusetts, invited Jensen and Jacobsen to present their findings at a Laurentian Hormone Conference in Mont Tremblant in 1961 (Jensen and Jacobson, 1962). The talk was entitled "Basic Guide to the Mechanism of Estrogen Action." Again, the authors elegantly described the target tissue specificity of estrogen, but additionally, Jensen (1962) described the first studies that demonstrated that the antiuterotropic activity of MER-25 depends, at least in part, on its ability to prevent the incorporation and retention of administered E₂ in the rat uterus. Thus, a foundation for the molecular mechanism of action of antiestrogens was established.

In 1963, Lerner reviewed progress in the development of antiestrogens at the Laurentian Hormone Conference (Lerner, 1964). MER-25 was not to become a clinically useful agent because of toxicity and low potency (Lerner, 1981); however, a triphenylethylene MRL-41 or clomiphene (fig. 1), as it became known (Holtkamp et al., 1960), was showing promise for the induction of ovulation in subfertile women (Greenblatt et al., 1962). The drug is now standard therapy for the treatment of infertility in anovulatory women.

After 1964, progress toward an understanding of antiestrogen action and the clinical utilization of antiestrogens was slow and largely ignored. However, by the late 1970s, with the successful clinical development of tamoxifen (fig. 1) for the treatment of breast cancer (Lerner and Jordan, 1990; Jordan, 1994), the prospects for new drug discovery changed dramatically.

Twenty years after Lerner completed the first review of nonsteroidal antiestrogens (1964), we reviewed the important developments that had occurred in our understanding of the receptor-mediated mechanism of action and the then state-of-the-art structure-activity relationships (Jordan, 1984). However, during the past dozen years, there have been enormous and far reaching changes in our basic knowledge and a new appreciation of the potential of antiestrogens as targeted agents to treat diseases associated with the menopause. This is because tamoxifen is an antiestrogen in the breast but has estrogen-like properties in other target tissues such as bone. Be that as it may, tamoxifen is used exclusively for the treatment of all stages of breast cancer (Jordan, 1997b), and clinical trials are testing the worth of tamoxifen as a preventive for breast cancer (Jordan, 1993, 1995b). By contrast, new and novel antiestrogens are being evaluated currently not only for breast cancer therapy but also for the prevention of osteoporosis (Gradishar and Jordan, 1997).

At a time when there is enormous interest in this topic, it is most appropriate to dedicate our review to Drs. Leonard Lerner and Elwood Jensen, whose seminal discoveries laid the foundations for all the subsequent research in this area. Our title is an adaptation of the original "Basic Guides to the Mechanism of Estrogen Action" used by Jensen and Jacobson at the Laurentian Hormone Conference in 1961 (Jensen and Jacobson, 1962).

We have organized our current review into two major parts. First, we will discuss the problems and inadequate understanding of antiestrogen action that occurred in 1984 and describe the enormous progress that has been achieved in understanding the fundamentals of estrogen action. Second, we will consider the current problems and potential of antiestrogens as valuable therapeutic agents and highlight the new knowledge that is emerging about the target site-specific mechanisms of estrogen and antiestrogen action. We recommend that readers refer to earlier articles for the history of the development of antiestrogens (Jordan, 1997a,b) and for a broad review of structure-activity relationships (Jordan, 1984; Lerner and Jordan, 1990).

	II. Unresolved Issues in 1984

Top Previous Next References

In 1984, we concluded our article in Pharmacological Reviews with the statement that several key issues concerning the pharmacology and mode of action of antiestrogens remained unresolved (Jordan, 1984). We wrote:

A unifying theory of antiestrogen action is, however, impractical because there are several unexplained observations with antiestrogens that require further study. (a) The species differences in the pharmacology of antiestrogens is perplexing. Although it is possible that the triphenylethylene type I antiestrogens (tamoxifen) are metabolized to estrogens in rodents, no convincing evidence has been presented to show metabolic differences between chickens and rodents. (b) Most antiestrogens exhibit agonist or partial agonist actions in vivo but in vitro, the compounds usually have zero intrinsic efficacy. The reason for this is unknown. (c) Tamoxifen binds to the so called "antiestrogen binding site" with precise structural specificity and high affinity. The binding site requires definition biochemically and its physiological role needs to be established.

As an introduction to our current review, we will briefly consider progress in addressing our previously "unresolved" issues. This retrospective illustrates the impressive progress that has been made.

There is still no satisfactory explanation for the different pharmacology of tamoxifen in the mouse, for example, where the drug is an estrogen in short-term tests, but in the chicken, it is an antiestrogen. Nevertheless, new facts have emerged to demonstrate that perhaps the answer lays more in tissue specificity than in species specificity. Long-term tamoxifen treatment of ovariectomized inbred (Jordan et al., 1990) or athymic mice (Gottardis and Jordan, 1988) results in an early estrogen-like effect in the uterus, but eventually the tissue response changes so the uterus is refractory to estrogen action. Tamoxifen is a preventive for the development of mammary tumors in mice (Jordan et al., 1991b); similarly, human breast tumor cell lines, which grow in response to estrogen in athymic mice, will not grow initially in response to tamoxifen (Gottardis et al., 1988a). After many months, however, tamoxifen-stimulated breast tumors will grow (Gottardis and Jordan, 1988), but interestingly enough, the uterus becomes refractory to estrogen in the same animal. As a result of these findings, one could ask whether tamoxifen-stimulated tumor growth is species-specific. The answer is "no," because the tamoxifen-stimulated human tumors derived from the athymic mouse model also will grow in response to tamoxifen in the athymic rat (Gottardis et al., 1989a). This excludes the possibility of species-specific metabolism.

Thus, an elucidation of the complexities of the target site-specific actions of antiestrogens may hold the most promise for resolving the unusual species differences. A combined effort to exploit the emerging molecular biology of receptor function and an understanding of the pharmacology of novel agents will prove instructive for future progress.

This issue has been resolved, for the most part, with the discovery that culture media contains estrogens (Berthois et al., 1986). We describe this fundamental discovery in detail in Section VI.A. There are now reasonable parallels with the partial agonist actions of compounds in vivo and in vitro.

Tamoxifen and the other triphenylethylene antiestrogens bind with high affinity to microsomal sites in tissues throughout the body. We previously reviewed progress in this area (Jordan and Murphy, 1990), but no one has succeeded yet in identifying a function for the binding protein itself. In parallel studies, Lubahn and colleagues (1993) have addressed the issue indirectly by showing that an estrogen receptor (ER) knock-out transgenic mouse does not elicit a uterotropic response to 4-hydroxytamoxifen (4-OHT). Thus, if the antiestrogen binding site plays a role in the pharmacology of antiestrogens, it does not seem to be as pivotal as the ER. Conceptually, this becomes a key issue. In the earlier review, we wrote, "Finally, it is perhaps naive to believe that a clear view of the mechanism of action of antiestrogens can be described when the molecular mechanism of estrogen-controlled protein synthesis and cell division is as yet unknown (Jordan, 1984)."

At that time, before the precise structure of the ER was known, crude models of the interaction of estrogens and antiestrogens were proposed to describe the agonist, partial agonist, and antagonist actions of various ligands. These models were summarized in our earlier article (Jordan, 1984), but the proposal was based on experimental studies with ER antibodies and radiolabeled E₂ and 4-OHT conducted in collaboration with Elwood Jensen (Tate et al., 1984) and in an extensive series of structure-activity relationship studies that started with a collaboration with Jack Gorski (Liebermann et al., 1983a,b; Jordan et al., 1984). Essentially, each study supported a model of ligand binding sites that would anchor estrogen but then be locked by a conformational folding of ER like the closing of the jaws of a crocodile. By contrast, an antiestrogen-like tamoxifen could be wedged into the ligand binding site, but the protein could not close around it correctly. The antiestrogenic molecule would be like a stick jammed into the jaws of a crocodile.

Progress to understand estrogen and antiestrogen action has been dramatic with the cloning and sequencing of the ER. The realization that the ER is a nuclear transcription factor, and just one of a superfamily of transcription factors, with many as yet unknown functions, has had a profound effect on scientific thinking during the past decade. Indeed, the conventional ER is now referred to as ER because a second receptor ER has been discovered recently (Kuiper et al., 1996).

Currently, evidence that our simple models of estrogen and antiestrogen action (Jordan, 1984) were close to the true state of affairs is developing. The ER has recently been crystallized with estrogens and antiestrogens revealing a similar locking of the estrogenic ligand by the mobile protein tail of the ER (Brzozowski et al., 1997). Nevertheless, the overall consequences of ligand binding are now known to be far more complex. Various levels of intrinsic efficacy are related to a range of conformations (McDonnell et al., 1995), and there is now knowledge of the essential role of associated proteins, or coactivators, to construct a transcriptional unit (Katzenellenbogen et al., 1996).

In our review, we will first describe the progress that has been made in the understanding of the molecular biology of estrogen action and use this as a basic foundation to consider the multifaceted actions of antiestrogens and their potential clinical applications. Finally, we will summarize the proposed molecular mechanism of action of the antiestrogen raloxifene (see Section XVII.) and suggest future studies that are necessary for a complete understanding of the multifaceted actions of a spectrum of drugs.

	III. The Estrogen Receptor

Top Previous Next References

The first evidence for a connection between estrogen and breast cancer growth was presented in 1896 when Beatson, a British physician, discovered that by removing the ovaries of premenopausal women, he could cause a regression of advanced breast tumors. Shortly thereafter, Stanley Boyd (1900) reported a study that established that one-third of all patients with breast cancer who had an oophorectomy would see a regression of their disease. However, the mechanism by which this occurred in these patients, and who would respond, were not to be discovered until 60 years later. The ER first was described in the uterus of rats (Jensen and Jacobson, 1962; Toft and Gorski, 1966; Jensen et al., 1968), and extensive early literature on the basic biochemistry of the ER quickly developed (Jensen and DeSombre, 1973; Gorski et al., 1968; Williams, 1974). Jensen and colleagues (1971) translated the basic science into clinical utility by proposing a predictive test, the ER assay, to determine which patients would respond to endocrine ablation, i.e., oophorectomy in premenopausal patients and adrenalectomy in postmenopausal patients. It was then established that patients with ER-rich tumors respond to endocrine therapy, whereas patients with ER-negative tumors are unlikely to respond (McGuire et al., 1975). These pivotal observations provide an excellent example of basic research that translated to the treatment of human disease.

Nuclear hormone receptors are a family of hormone-activated transcription factors that can initiate or enhance the transcription of genes containing specific hormone response elements. The human ER, which belongs to this family, was cloned and sequenced from MCF-7 human breast cancer cells (Green et al., 1986, Greene et al., 1986). The ER protein consists of 595 amino acids with a molecular weight of 66 kDa (Green et al., 1986) that has been separated into six different functional domains (fig. 2) (Kumar et al., 1986, 1987). Two of these functional domains are highly conserved in the primary sequence of members of the nuclear hormone receptor superfamily. One of the domains, the DNA binding domain (DBD), contains two zinc fingers that mediate receptor binding to hormone response elements in the promoters of hormone-responsive genes. In the C-terminal region, the hormone binding domain (HBD) contains two regions of sequence homology with other hormone receptors and bestows hormone specificity and selectivity (Carson-Jurica et al., 1990; Krust et al., 1986; Kumar et al., 1987; Kumar and Chambon, 1988; Orti et al., 1992). The human ER is located on chromosome 6q sub band 25.1 (Menasce et al., 1993), and the mouse ER is located on chromosome 10 (Sluyser et al., 1988).

View larger version (18K): [in this window] [in a new window]   Fig. 2.   The ER consists of six functional domains (AF) transcribed by eight exons. The functional domains are labeled accordingly and the relevant mutations, illustrated in the text, are highlighted with arrows.

	IV. A Second Receptor

Top Previous Next References

Recently, a novel member of the nuclear hormone receptor superfamily was cloned from a rat prostate complementary DNA (cDNA) library (Kuiper et al., 1996; Katzenellenbogen and Korach, 1997). This novel sequence encodes a protein of 485 amino acid residues, and the molecular weight has been calculated to be 54.2 kDa (fig. 3). ER bears substantial homology to ER especially in the DBD (95%) and the HBD (55%), and these proteins are functionally homologous in that ER binds estrogen with high affinity as shown by saturation ligand-binding analysis. The functional homology of ER and ER has been determined by measuring transcriptional activity of ER in a system designed to test the functionality of ER. It has been determined by the activation of transcription of a vitellogenin A2 estrogen response element (ERE)-containing reporter plasmid in the presence and absence of estrogen that ER is functionally homologous (Kuiper et al., 1996).

View larger version (16K): [in this window] [in a new window]   Fig. 3.   Comparison of the rat (r) ER and rER proteins and percent amino acid homology in the functional regions.

Recently, the mouse homolog of the rat ER was cloned and mapped to chromosome 12 (Tremblay et al., 1997). The ER gene has been designated Estrb and is expressed in several transcripts. The corresponding cDNA has been shown to encode a 485-amino-acid protein and has 97% identity to the DBD of mouse (m)ER and 60% identity to the LBD of mER. The most interesting question after the identification of this novel ER is whether it has the same pharmacological properties as ER. Tremblay and colleagues (1997) have shown that mER binds to the vitellogenin A2 ERE although with a lower affinity than that of mER. More importantly, mER can transactivate reporter genes containing EREs in transient transfection experiments with the same efficiency as mER in HeLa and Cos-1 cell lines.

As would be expected, similarities and differences exist between the mER and mER such as different aspects of regulation. For example, it is possible that mER can be activated via phosphorylation through the mitogen-activated protein kinase pathway as shown for ER (Kato et al., 1995; Bunone et al., 1996). This would be predicted because of the conservation of serine 60 which could be phosphorylated in the mouse, rat and human ER sequences. A few differences surfaced in the pharmacology of ER when 4-OHT was tested in transient transfection reporter assays. The partial agonism that 4-OHT expresses in cells with ER is not present when cells are transfected with ER (Tremblay et al., 1997). One possible explanation is the lack of homology in the amino-terminal domains of these proteins where the activation function-1 (AF-1) resides (see Section V.A.). The AF-1 is thought to be responsible for the partial agonist activity of tamoxifen in cells that express ER (McInerney and Katzenellenbogen, 1996).

Clearly, the most important question is the distribution of ER in tissues and the relative importance of ER and ER for the pharmacological action of antiestrogens. In addition to the presence of ER in the rat prostate and the mouse ovary, in situ hybridization studies have determined that the granulosa cells of the rat ovary also express ER (Kuiper et al., 1996). Previous studies tested an ER knock-out mouse that does not express functional ER for its ability to respond to estrogen (Lubahn et al., 1993). The female knock-out mice that were ER negative were infertile and did not develop normal uteri and ovaries. Thus, if ER was expressed in the ovaries of these ER knock-out mice, it was not functioning to compensate for the loss of ER. Alternatively, ER could regulate the expression of ER so that in the absence of ER, ER is down-regulated. However, recent studies by Korach and colleagues (personal communication) suggest this is not highly probable. Further studies with ER knock-out mice show residual estrogen binding of approximately 5 to 10% of the ER level (Lubahn et al., 1993; Couse et al., 1995; Korach et al., 1996). What is particularly interesting is the fact that there are very high circulating levels of E₂ in the ER knock-out mice that could be interacting with ER to produce the pathological states observed in the mice.

The presence of two different ERs could explain the mechanism of the target site specificity seen with antiestrogens or differential transcriptional AFs on estrogen-responsive genes (Kuiper et al., 1997). Even though the evolutionarily conserved regions of these two ERs are homologous, various nonconserved regions exist which probably account for the differences seen between ER and ER. We will discuss the issue again in Complexity of Antiestrogen Action (Section XII.).

	V. Estrogen Receptor

Top Previous Next References

The model for estrogen action via the ER (henceforth referred to as ER) has evolved considerably during the past 40 years. The first realistic conceptual model was proposed by Mueller and colleagues (1957) to explain the initiation of metabolic events in the rat uterus by estrogen. Since then, several models have evolved that address the mechanism of how the ER functions in the nucleus and how it activates the transcription of estrogen-responsive genes in the presence of estrogens (Gorski et al., 1984, 1993), an effect differentially blocked by antiestrogens. We will describe the emerging data about the functional domains of the ER to lay the foundation for our discussion of receptor regulation and antiestrogen action.

The six structural domains of the ER are regions that have been defined based on the putative functions that are contained in each area. The A/B domain contains one of the two transcriptional AFs present in the ER (fig. 2). AF-1 and AF-2 activate transcription in a cell and promoter context specific manner (Gronemeyer, 1991) and AF-1 and AF-2 are autonomous in that they are located at the N- and C-termini, respectively. In early studies, the existence of AF-1 initially was not discovered because ER deletion mutants in the A/B region retained the ability to activate the transcription of vit-tk-CAT reporter genes (Kumar et al., 1987). Unlike AF-2, which is induced upon hormone binding to the receptor (Kumar et al., 1987; Webster et al., 1988, 1989; Lees et al., 1989; Tora et al., 1989), we now know that AF-1 is constitutively active.

AF-1 acts in a cell type-specific fashion as shown in experiments using chimeric receptors. When the A/B region of the ER was expressed with the DBD of the yeast transcriptional activator Gal-4, this chimera was able to activate transcription of Gal-4-responsive promoters in chicken embryo fibroblasts but not in HeLa cells, thus demonstrating a cell type-specific function (Berry et al., 1990; Tora et al., 1989). The AF-2, which is located in the E region containing the HBD, when associated with Gal-4 showed activation of Gal-4-responsive promoters in both HeLa and chicken embryo fibroblasts (Webster et al., 1988). Thus, it is thought that AF-1 is responsible for the promoter-specific transcriptional activation independent of the presence of ligand and that AF-2 provides ligand-specific activation (Berry et al., 1990; Webster et al., 1988).

The C region contains the DBD and a dimerization domain. The DBD is the most highly conserved region in the nuclear hormone receptor superfamily. The DBD consists of two zinc fingers that fold into two helical domains upon the coordination of one zinc to four cysteines and a third helix that extends from the zinc fingers (Schwabe et al., 1993). These zinc fingers are essential components of the ER because when the ER lacks the DBD, it cannot bind DNA in vitro or in vivo (Kumar and Chambon, 1988; Kumar et al., 1987). However, the C region alone is not sufficient to bind an ERE. As stated above, the A/B region can be deleted without compromising the DNA binding ability but deletion of the basic amino acids (amino acids 256 to 270) located downstream of the zinc fingers does impair the ability of the receptor to bind EREs (Kumar and Chambon, 1988; Chambraud et al., 1990).

There are many similarities in the zinc finger regions among different steroid hormones receptors, but there are precise differences that account for the specificity of each receptor. It is believed that the specificity of a certain receptor is afforded by the first of the two zinc fingers. These conclusions are based on mutagenesis in the region of the first zinc finger. The results prove that the receptor binds to specific nucleic acid residues in the major groove of the DNA helix. The second zinc finger is responsible for stabilizing this interaction through ionic bonds with the phosphate groups in the DNA backbone (Umesono and Evans, 1989; O'Malley, 1990; Parker and Bakker, 1991). In addition to these mutational studies, domain-swapping experiments in which the ER DBD was exchanged with the DBD of the glucocorticoid receptor showed that the chimeric protein activates glucocorticoid responsive genes in the presence of estrogen (Green and Chambon, 1991).

In addition to the basic requirement for DBD activity, the C region may bind to heat shock protein 90 (Chambraud et al., 1990) and also be responsible for nuclear localization of the receptor. The C region contains three lysine- and arginine-rich proto-nuclear localization signals (NLSs) that are ligand-independent. Several NLSs have been identified in the ER, one in the DBD and three others in the HBD (within amino acids 256 to 303) (Ylikomi et al., 1992). One NLS in the HBD has been shown to be ligand inducible, and the other NLSs are ligand independent. The inducible and constitutive NLSs cooperate in the presence of estrogen (Ylikomi et al., 1992).

The E region, the HBD, contains the AF-2 (ligand-dependent and promoter-specific), heat shock protein 90 binding function, a NLS (ligand-dependent), and a dimerization domain. The HBD is found in the C-terminus and is responsible for specific ligand recognition because it allows the ER to be transcriptionally active in a specific and selective manner. The HBD is thought to coordinate with the DBD and upon ligand binding, the coordination is lost and the receptor protein changes conformation, releases the DBD, and becomes transcriptionally active (reviewed in Gronemeyer, 1991; Parker et al., 1993).

Estrogen diffuses through the plasma membrane of cells where it binds to the ER (Rao, 1981). For many years, it generally was thought that estrogen bound to the ER in the cytoplasm and translocated into the nucleus, but it is known now that the ER is a nuclear transcription factor that initially interacts with estrogen in the nucleus (King and Greene, 1984; Welshons et al., 1984). Once estrogen binds to the ER, heat shock proteins dissociate and a change in conformation and homodimerization occurs (fig. 4).

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Estradiol (E2) action is transduced through ER located in the nucleus. On estrogen binding, the ER homodimerizes and interacts with EREs located in the promotor region of estrogen-activated genes. These events trigger an estrogenic response in the cell.

Although phosphorylation of steroid hormone receptors enables them to become transcriptionally active, until recently, the role of phosphorylation of the ER was still in question (Orti et al., 1992). Phosphorylation of the ER from MCF-7 and calf uterus is estrogen-dependent and, in addition, increases the receptor's affinity for specific DNA sequences (Denton et al., 1992). The basal level of ER phosphorylation increases three- to four-fold upon treatment with estrogen and antiestrogens (Le Goff et al., 1994). However, the key to elucidating the mechanism of estrogen action is the identification of the selective sites for phosphorylation. Several serines in the amino-terminal portion of the human ER may play a role in hormone-regulated phosphorylation. However, when phosphopeptide maps of wild-type and mutant ERs treated with estrogen or antiestrogens are compared, the results are similar indicating that differential phosphorylation between these receptors cannot account for any differences in function (Lahooti et al., 1994). An alternate approach might be the identification of enzymes responsible for phosphorylation. There are several protein kinases thought to be involved in phosphorylation of the ER (ER kinase, DNA-dependent kinase, Ser-Pro kinases, protein kinase C, protein kinase A, and casein kinase II) (reviewed in Kuiper and Brinkman, 1994). Recently, a mitogen-activated protein kinase also was implicated in phosphorylation of the ER on Ser 118 resulting in the activation of ER AF-1 (Kato et al., 1995). Interestingly, another consequence of phosphorylation of the ER is the regulation of homodimerization through phosphorylation of tyrosine 537 (Arnold et al., 1995).

Although phosphorylation may play a part in receptor activation, exciting progress has been made in understanding how the receptor cooperates with other proteins to assemble a transcription unit for gene activation. The receptor can be viewed as a skeleton to assemble the unit as a prelude to DNA unwinding and the transcription of selected mRNAs. To achieve this, the receptor eventually must interact with other proteins as well as bind to one or several EREs. We will dissect the process by describing the areas needed for receptor activation, ligand binding, DNA binding, and protein-protein interactions.

The ER contains two areas called AFs: AF-1 is located in the amino-terminal region of the ER, and AF-2 is located in the carboxyl-terminal region in the ligand binding domain (LBD) of the ER; these are synergistic when the ER is activated by estrogen. Katzenellenbogen and colleagues (1995) used mammalian cells to show that the AF-1 and AF-2 regions, when expressed as separate polypeptides, functionally interact in response to estrogen and antiestrogens. The authors found that this interaction could activate transcription in response to estrogen. In addition, when mutations were made in AF-1 or AF-2 that abrogated the functional activity of these domains, no transcriptional activity was seen. Additionally, when mutations were made in the LBD that eliminated estrogen binding, no transcriptional activity could be detected. These experiments suggest that estrogen binding to the ER facilitates a conformational change that brings AF-1 and AF-2 in direct association with one another leading to synergy that results in transcriptional activation. These elegant experiments provide a mechanistic explanation for the role of the two AFs in mediating hormone-regulated transcription.

In addition to understanding the mechanism through which the ER becomes transcriptionally active, many of the amino acids important in the binding of ligand to the ER have been identified. Harlow and coworkers (1989) showed a covalent attachment between Cys530 and both an estrogen agonist and an antagonist. This work also suggested that other cysteine residues present in the LBD may be important for ligand-mediated transcriptional activation. Further mutant ERs have been constructed with mutations at the other cysteine residues present in the LBD. Each of these mutants showed an affinity similar to that of the wild-type ER (Reese and Katzenellenbogen, 1992). When these mutants were tested in reporter assays, the mutants C530A and C530S showed unaltered binding to estrogens and antiestrogens, but the transactivation response to both estrogens and antiestrogens had changed. After showing that the C530 is involved in discriminating between ligands, Pakdel and Katzenellenbogen (1992) examined the role of amino acids adjacent to the other cysteines in the LBD of the ER. The results showed that the amino-terminal domain of the LBD was important in differential transcriptional activation but not in binding affinity. When the carboxyl-terminal region of the LBD is mutated, this renders the protein transcriptionally inactive although it can still bind ligand, making this a very powerful dominant negative ER. Thus, there is a distinction between the hormone binding and the transactivation functions.

Once the ER has bound estrogen and dimerized, it binds to EREs present in the promoter region of genes. These EREs are 13 base pair palindromic sequences located upstream from the transcriptional start site. The EREs function by enhancing the transcriptional potential of a gene. EREs have been identified and defined using reporter systems to test the enhancer ability when exposed to different compounds (Gronemeyer, 1991). Also, deletional analysis has allowed the definition of the sequence of EREs. Optimally, it consists of two inverted repeats separated by any three base pairs. The exact sequence of EREs varies between species and genes (Klein-Hitpass et al., 1986).

Some models of estrogen action predict that when the dimerized hormone-receptor complex binds to the palindromic ERE that it forms a looped structure allowing the ER to interact with the transcriptional apparatus at the RNA initiation site. It is thought that the hormone-receptor complex can recruit components of the transcriptional complex and serves as a nucleation site. Previous studies focus on the interaction of the ER with EREs, but more recently, there has been a shift toward the study of ER receptor interactions with ancillary proteins in the nucleus.

For example, Gorski and colleagues (1993) have suggested that the ER binds DNA in a heterodimer structure involving a variety of other proteins such as transcription factors or other DNA binding proteins. It also has been shown that estrogen is not essential for ER binding to DNA (Murdoch et al., 1990) but that this does increase the ER's affinity for nuclear components. Another aspect of this study suggests that in the traditional reporter assays generally used to study these mechanisms lack the complexity that exists in the nucleus and in the nucleosome-chromatin structure.

Currently, there is intense interest in the identification of possible coactivators that can enhance ER-dependent transcription. The first candidate for a transcriptional coactivator, SPT6, was isolated from Saccharomyces cerevisiae and was shown to be capable of modulating ER-mediated transcription in yeast and mammalian cells and to interact specifically with the carboxyl-terminal portion of the ER (Baniahmad et al., 1995). Another steroid receptor coactivator, SRC-1, was sequenced and characterized using the yeast two-hybrid system (Oñate et al., 1995). SRC-1 has been shown to interact specifically with the progesterone receptor (PR) and enhance its transcriptional activity. When SRC-1 was tested with the thyroid hormone receptor (TR), retinoic acid receptor (RAR), ER, and glucocorticoid receptor, it enhanced the transcriptional activity of each of these steroid hormone receptors. In fact, SRC-1 may have a complex role to play in steroid receptor regulation. For example, the ER can interfere with transcriptional activation by PR but SRC-1 will inhibit the effects of the ER.

Another recent discovery is that the transcription factor cAMP response element-binding protein (CREB) has an associated protein termed the CREB-binding protein (CBP) (Smith et al., 1996). CBP has been shown to interact specifically with RNA polymerase II (Kee et al., 1996), TFIIB (Kwok et al., 1994), and with CREB in its phosphorylated form (Chrivia et al., 1993). It has been postulated that the ability of CBP to stimulate transcription is through the targeted recruitment of RNA polymerase II to the promoters of genes. In addition to the above-described proteins, CBP can interact specifically with members of the steroid hormone nuclear receptor family and is able to enhance transcriptional activity in some instances (Kamei et al., 1996). Thus, CBP can function as a coactivator for a rapidly growing number of transcription factors.

Ectopic expression of CBP can enhance estrogen-dependent ER transcriptional activity approximately ten-fold compared with the ectopic expression of SRC-1 (Smith et al., 1996). Again, CBP is partially able to reverse the transcriptional interference that activated ER has on PR-mediated transcriptional activity. Most importantly, these data suggest that CBP may be present in limited quantities in particular cells and may be able to modulate the activity of the steroid receptors. When SRC-1 and CBP are coexpressed ectopically, ER- and PR-mediated transcriptional activity is enhanced in a synergistic manner, which suggests that these two proteins are not functionally homologous.

In addition to coactivators, another category of molecules that are able to repress basal transcription induced by hormone receptors has been identified. Two corepressors termed the silencing mediator for retinoic and thyroid hormone receptors (SMRT) (Chen and Evans, 1995) and nuclear receptor corepressor (N-CoR) (Hörlein et al., 1995; Kurokawa et al., 1993) have been cloned using a yeast two-hybrid system. Both SMRT and N-CoR can interact with TR and RAR through specific homologous domains that have been shown to bear some homology to each other (Perlmann and Vennstrom, 1995). This finding suggests that a family of evolutionally conserved corepressors may exist that interact with other steroid hormone receptors. Corepressors that act on the ER have not yet been identified, but there is every reason to believe that they could exist.

Both SMRT and N-CoR associate with specific unliganded receptors but are released once the ligand has bound (Chen and Evans, 1995; Hörlein et al., 1995). This is consistent with present dogma because when hormone receptors are unliganded, their ability to activate transcription presumably is compromised, but when ligand binds, thereby activating the receptors, the repression is alleviated leading to either an active receptor or possibly one that is open to activation by coactivators. Further evidence that these corepressors can silence receptor activity has been shown in mutational studies. The hinge region of TR and RAR which connects the DBD and the HBD has been shown to be important for a receptor's susceptibility to a repressor. When mutations are introduced into the hinge regions of the TR and the RAR, interaction with the corepressor is ablated and basal transcription levels are repressed (Chen and Evans, 1995; Hörlein et al., 1995; Kurokawa et al., 1993). The characterization of these corepressors could offer new insights into the molecular basis of nuclear hormone receptor modulation of transcription.

Overall, there has been enormous progress in understanding the growing levels of complexity involved in estrogen action. The key to understanding antiestrogen action is the ER, so we will now review progress in the regulation of the protein as it pertains to issues in breast cancer and antiestrogen responsiveness.

	VI. Estrogen Receptor Regulation

Top Previous Next References

The discovery of the ER and the fundamental role it plays in estrogen and antiestrogen action naturally has focused interest on the regulation of this nuclear transcription factor. However, progress in elucidating regulatory pathways between 1970 and 1986 had been slow partly because of the misinterpretation of data derived from the available laboratory models. In this section, we will review the change that has occurred in our basic understanding of estrogen action in cell culture.

Estrogen withdrawal is one of the principal treatment strategies for breast cancer (reviewed in Santen et al., 1990; Jordan and Murphy, 1990). Nevertheless, throughout the 1970s and early 1980s, the direct effects of estrogen on breast cancer cell growth in culture were extremely difficult to demonstrate and results were hard to interpret. The discovery that the standard laboratory cell culture model was flawed is an important lesson that has multiple ramifications in science. Lippman and Bolan (1975) first showed that the ER-positive MCF-7 breast cancer cell line was growth inhibited by the antiestrogen tamoxifen, but this effect could be reversed by the addition of E₂. The action of E₂ alone, compared with controls, was not particularly dramatic. The inability of the research community to provoke breast cancer cell growth reproducibly in cell culture was the subject of an intense debate for approximately a decade (1975 to 1986) and there were even suggestions that because estrogen could only cause MCF-7 cells to grow into tumors in estrogen-treated athymic animals (Shafie, 1980) but estrogen could not cause growth in vitro, then a second hormonal messenger was necessary in vivo to support growth. At the time, this was not unreasonable because both estrogen and prolactin were required for the growth of dimethylbenzanthracene-induced rat mammary tumors (Welsch, 1985).

Despite the inability to demonstrate a direct effect of estrogen-stimulated growth in all laboratories, Lippman's group did show that ZR-75 cells would respond to estrogen in a defined medium (Allegra and Lippman, 1978), and a reproducible model of estrogen-stimulated prolactin synthesis in primary cultures of cells from primary tumors also was established (Lieberman et al., 1978). This latter model was used to define the structure-activity relationships of numerous antiestrogens (Lieberman et al., 1983a,b; Jordan et al., 1984, 1986, 1988a; Jordan and Lieberman, 1984). However, there was no adequate explanation for the finding that antiestrogens always depressed control values despite vigorous removal of all known estrogen from the culture system through serum stripping with charcoal.

The breakthrough came with the discovery that the pH indicator, phenol red, was present in micromolar concentrations in cell culture media (Berthois et al., 1986). The structure of phenol red is reminiscent of the estrogens originally synthesized (fig. 5) by Sir Charles Dodds in the 1930s (Dodds and Lawson, 1936). Removal of phenol red indicator from culture media dramatically altered the cellular response to exogenous estrogen. Now, control values were not depressed by antiestrogens but E₂ did cause a huge increase in the growth response of ER-positive breast cancer cell lines in culture. As predicted, antiestrogens competitively inhibited estrogen-stimulated growth and exhibited partial agonist actions (Berthois et al., 1986).

View larger version (21K): [in this window] [in a new window]   Fig. 5.   The structure of the estrogenic contaminant of phenol red indicator present in culture media (Katzenellenbogen et al., 1996; Bindal and Katzenellenbogen, 1988). The contaminant is a bisphenolic compound reminiscent of the nonsteroidal estrogens first reported in the 1930s (Dodds and Lawson, 1936). At that time anol was believed to be the minimal structure with estrogenic activity (Dodds and Lawson, 1937a,b), but a dimerization product was found to contaminate some samples (Campbell et al., 1938a,b). The structure is very similar to the potent synthetic estrogen, diethylstilbestrol (Dodds et al., 1938a).

Clearly, breast cancer cells were grown unintentionally in a fully estrogenized medium, so studies of exogenous estrogen action and estrogen withdrawal were impossible. Estrogen was always present. To place this in perspective, we now know that the growth response to estrogen is so exquisitely sensitive that less than 10¹⁰ M will produce maximal effects. The concentration-response curve that extends between 10¹² and 10¹⁰ M is within the lower range of circulating levels of estrogen in postmenopausal women but often beyond the range of routine radioimmunoassays. In contrast to the profound sensitivity of replication to estrogen stimulation, the action of estrogen to induce differentiation functions of progesterone receptor or prolactin synthesis requires ten times more estrogen.

Interestingly enough, phenol red was not the actual estrogenic stimulus. Different lots of phenol red from different manufacturers had different levels of estrogenicity (Welshons et al., 1988), but John and Benita Katzenellenbogen demonstrated that the phenol red alone could not account for the estrogenicity seen (Bindal et al., 1988). They isolated a contaminant, produced during manufacture, that was a potent estrogen (Bindal and Katzenellenbogen, 1988). The compound is a dimerization product of components used in the synthesis of phenol red (fig. 5).

The discovery of an estrogenic contaminant in phenol red indicator is analogous to a research problem encountered in the 1930s during the first synthetic attempts to define the minimal structure of an estrogen. Anol, a simple phenol derived from anethole (fig. 5), was reported to possess extremely potent estrogenic activity with lng inducing estrus in all rats (Dodds and Lawson, 1937a). These results were not confirmed using different preparations of anol (Dodds and Lawson, 1937b; Zondek and Bergman, 1938), but it was discovered that dimerization of anol to dianol can occurr during drug synthesis and this impurity, which was know to have estrogenic properties (Campbell et al., 1938b), was responsible for the anomalous results (Campbell et al., 1938a). At approximately the same time, Dodd's group discovered that the diethyl substitution at the ethylenic bond of stilbesterol produces the potent estrogen diethylstilbesterol (Dodds et al., 1938a,b). This discovery was to revolutionize therapeutics with estrogen, and high-dose diethylstilbesterol therapy became the standard endocrine treatment for breast and prostate cancer before the discovery of antiestrogens (Haddow et al., 1944).

The regulation of ER expression in human breast cancer cells is a complex and multifaceted process that varies between different cell types and is also differentially regulated by estrogen and antiestrogens. The understanding of how different estrogens and antiestrogens affect the expression of ER in different cell types may be important in optimizing the development of new antiestrogen therapies that do not promote progression to hormone nonresponsive phenotypes. Currently, two models of ER regulation have been proposed (Pink and Jordan, 1996) that begin to elucidate how estrogens and antiestrogens direct the expression of the ER in T47D and MCF-7 cells.

In the MCF-7 cell culture system, Model I regulation dictates the response of the cell to treatment with estrogen or antiestrogens. This model is defined by down-regulation of ER expression at both the mRNA and the protein level with estrogen treatment. However, the partial antiestrogen 4-OHT (see Section VII.A.) has no effect on the mRNA levels but causes a net accumulation of ER protein by stabilization. The pure antiestrogen, ICI 182,780 (see Section VII.B.), causes a marked reduction in ER protein levels but has no effect on the mRNA levels. Thus, each of these compounds has a dramatically different effect on the expression of the ER both at the mRNA and at the protein level.

The T47D human breast cancer cell line exhibits Model II regulation. This is defined by an increase mRNA expression and a maintenance of ER protein levels with estrogen treatment. Upon treatment with 4-OHT, there is little effect on the steady-state ER mRNA levels. On the other hand, ICI 182,780 causes a marked reduction in ER protein levels and lowers levels of ER mRNA. These examples illustrate two very different mechanisms of estrogen and antiestrogen effects on ER expression in two ER-positive human breast cancer cell lines. These short-term studies could explain the response of these breast cancer cells to long-term estrogen deprivation (see Section VI.C.).

The transcriptional regulation of the ER in breast cancers seems very complicated; however, there have been recent advances in elucidating a mechanism. The control of ER expression allows the cell to increase or decrease the levels of ER in the cell according to the requirements for survival. The regulation of ER expression also plays an important role in the ER status of a cell during tumor progression. Clearly, discovery of the mechanisms for receptor regulation or re-activation hold the promise of being a valuable therapeutic target to maintain antiestrogen sensitivity.

Transcription of the ER can be initiated at two separate promoters, P₀ or P₁ (Keaveney et al., 1991), although the principal transcriptional start site is P₁ (Green et al., 1986). deConinick and colleagues (1995) found that there is an important transcriptional regulatory element in the 5'-untranslated leader sequence in the ER gene. They showed that this sequence contains two binding sites for a trans-acting DNA-binding protein called ER factor 1 (ERF-1). ERF-1 is expressed in higher levels in ER-positive and endometrial carcinomas and in lower amounts in normal human microvascular endothelial cells. This suggests that a correlation exists between the expression of ERF-1 and the amount of ER expressed in a given cell. The challenge is to discover whether the expression of ERF-1 is tightly regulated or whether it is susceptible to subtle changes in the cellular environment.

Recently, McPherson and colleagues (1997) cloned the gene for the ERF-1 transcription factor and also showed that ERF-1 is a member of the developmentally regulated AP-2 transcription factor family. Using a 30 base pair imperfect palindromic sequence that has been defined as a high-affinity binding site for ERF-1, they showed that ERF-1 bound specifically so they used this concept to affinity purify the ERF-1 protein. The ERF-1 is approximately 50 kDa and the predicted peptide sequence shares 65% identity and 83% similarity with AP2 and is the same as AP2. In vitro translated ERF-1 showed activity similar to native ERF-1 and an AP2 polyclonal antibody that specifically reacts with ERF-1. The mechanism for ERF-1 to activate transcription of the ER has yet to be elucidated.

Other positive regulatory elements exist in the ER gene further upstream from the transcriptional start site (3778 to 3744) (Tang et al., 1997). The cis-acting element is of a 35 base pair element termed ER-EH0 that is active in ER-positive but not ER-negative cells. ER-EH0 contains not only an AP-1 but also flanking sequences that bind an as yet unknown factor. Both of the flanking sequences are required for enhancer activity. Tang et al. (1997) suggest that the ER-EH0 enhancer element is the predominant cis-acting factor in differential ER expression.

We believe it is important to stress that the regulation of the ER is a primary therapeutic target. Further progress can be facilitated by the description of models for the loss of receptor regulation. However, this is an area of some controversy. Although dogma dictates that breast tumors progress from ER positive to ER negative, the principle is not demonstrated easily in cell culture.

Studies of long-term estrogen deprivation of MCF-7 breast cancer cells in culture illustrate that selection pressure occurs with an initial increase in ER content so that the cells now grow maximally in the apparent absence of estrogen (Katzenellenbogen et al., 1987; Welshons and Jordan, 1987). The cells, however, still respond to antiestrogens with an inhibition of growth. Either they have become hypersensitive (Masamura et al., 1995) to other environmental estrogens leached from laboratory plasticware or the cells have devised alternative growth pathways. To the first point, several chemicals have been identified that might be responsible for supporting the growth of breast cancer cells in an "estrogen-free" environment (fig. 6) (Krishnan et al., 1993; Soto et al., 1991; White et al., 1994) and it would illustrate the need for a breast cancer treatment strategy in patients that blocks the ER continuously. Highly estrogen-sensitive clones will be selected to develop and grow toward any source of weak estrogens. To the second point, numerous estrogen-unresponsive clones of MCF-7 cells have been developed from the original stocks kept in a phenol red-free environment for many months. One cell line, MCF-7/5C, is ER positive but does not respond to either estrogens or antiestrogens. The ER is not mutated but the receptor is incapable of initiating progesterone receptor synthesis in the presence of estrogen (Jiang et al., 1992b). The cell type is reminiscent of the clinical situation of breast cancers that are ER positive but progesterone receptor negative and are less responsive to endocrine therapy (Jordan et al., 1988b).

View larger version (16K): [in this window] [in a new window]   Fig. 6.   The structure of environmental laboratory estrogens that can support the growth of breast cancer cells in an "estrogen-free" environment (Soto et al., 1991; White et al., 1994; Steinmetz et al., 1997).

The cell line, MCF-7/2A, is another clonal estrogen-independent cell line derived from MCF-7 stocks maintained in an estrogen-free state for several years (Pink et al., 1995). The cells are unique because they express wild-type ER as well as an ER that has a duplication of exons 6 and 7 in the LBD (Pink et al., 1996b). The high molecular weight ER does not bind estrogens or antiestrogens (Pink et al., 1997). However, there is no evidence that this mutant receptor is responsible for estrogen-independent growth.

All of the studies of estrogen deprivation so far described have used one single cell line, MCF-7, and the results from different groups demonstrate that numerous clones develop to survive the loss of the primary growth stimulus, estrogen (Cho et al., 1991; Clark et al., 1989b). However, it is now apparent that different cell types respond differently to estrogen withdrawal than MCF-7 cells. ER levels seem to be regulated in different ways (Pink and Jordan, 1996).

Murphy et al. (1989, 1990b) first illustrated the progression of an ER-positive T47D breast cancer cell line to an ER-negative state after prolonged estrogen deprivation. Pink et al. (1996a) subsequently demonstrated that the loss of the ER at the mRNA and protein level in this T47D cell line was irreversible The resulting cell line (T47D: C4:2) is resistant to antiestrogens and grows maximally in estrogen-free media. This raised the questions that if the receptor is lost, how is it lost and can it be reactivated?

One area of intense investigation is the hypermethylation of CpG islands in the 5'-promoter region of the ER gene that could silence ER synthesis. ER-negative human breast cancer cells grown in culture have an enhanced ability to methylate DNA which may explain the silencing of ER expression. Additionally, using the ER-negative MDA-MB-231 breast cancer cell line, treatment with DNA methylation inhibitors actually caused the re-expression of the ER at the protein level (Ferguson et al., 1995). This re-expressed ER is functional because it can activate the transcription of estrogen-responsive genes. However, this is not a universal cellular phenomenon, so further studies need to be undertaken. We have noted in our T47D cell lines that the CpG islands are not hypermethylated when the ER is lost (Chen et al., 1997).

The finding that ER can be retained in some cell lines in response to estrogen deprivation but not in others has clinical relevance. The levels of expression of the ER in clinical tumors as they progress to a hormone-independent state has become controversial. A recent review proposes that the actual loss of ER expression in ER-positive tumors does not occur (Robertson, 1996). However, the primary endocrine therapy today is tamoxifen and this has estrogen-like properties and may, as a result, preserve ER status. This is consistent with the observations in both cell and tumor models of antiestrogen resistance (Mullick and Chambon, 1990; Gottardis and Jordan, 1988; Katzenellenbogen et al., 1995). The receptor is not lost. However, the loss of the ER may occur in tumors that become resistant to the pure antiestrogens (see Section VII.B.). In part, the difference in the biological response may be a result of the different mechanisms of action for tamoxifen and pure antiestrogens on the ER signal transduction pathway (see Section VIII.). This is the focus of current clinical investigations.

	VII. Antiestrogen Classification

Top Previous Next References

Antiestrogens can be classified into two major groups: analogs of tamoxifen or its metabolites (type I) which have mixed estrogenic/antiestrogenic actions in laboratory assays and pure antiestrogens (type II) that have no estrogen-like properties in laboratory assays. There is emerging information to suggest that the classification may also be based on different mechanisms of action (see Section VIII.).

The triphenylethylene structure of tamoxifen has provided the basis for several new analogs that are being investigated in the clinic. The finding that tamoxifen is metabolized to 4-OHT, a potent antiestrogen (Jordan et al., 1977), also has provided a central theme for drug development (fig. 7).

View larger version (22K): [in this window] [in a new window]   Fig. 7.   The principal tamoxifen analogs currently under investigation are toremifene, idoxifene, droloxifene, and TAT-59. Each triphenylethylene is a derivative of tamoxifen, the parent drug, or its metabolite 4-OHT.

The principal tamoxifen analogs currently under investigation are illustrated in figure 7. Toremifene, or chlorotamoxifen, has been investigated thoroughly as an antiestrogen and antitumor agent in the laboratory (Kangas et al., 1986; Kangas, 1990) and currently is being used for the treatment of advanced breast cancer and tested as an adjuvant therapy. The compound is of interest because it does not produce DNA adducts in rat liver and, as a result, is not a potent carcinogen in rat liver (Hard et al., 1993; Hirsimaki et al., 1993) (see Section XIII.).

Idoxifene is a metabolically stable analog of tamoxifen synthesized to avoid toxicity reported with tamoxifen in the rat liver (fig. 7) (McCague et al., 1989, 1990). Substitution of halogens in the 4-position of tamoxifen is known to reduce antiestrogen potency by preventing conversion to 4-OHT (Allen et al., 1980) and it was argued that reduced demethylation of the side chain also would avoid the formation of formaldehyde in the liver (McCague et al., 1989, 1990). Idoxifene is a 4-iodopyrrolidino derivative of tamoxifen that has antiestrogenic and antitumor properties in laboratory rats (Chander et al., 1991).

Droloxifene, or 3-hydroxytamoxifen, has been studied extensively as an antiestrogen and an antitumor agent in the laboratory (fig. 7) (Hasman et al., 1994). This drug does not form DNA adducts under laboratory conditions (White et al., 1992) or produce liver tumors in rats (Hasman et al., 1994). Extensive clinical testing has shown activity in the treatment of advanced breast cancer in postmenopausal patients (Rausching and Pritchard, 1994).

TAT-59 is a prodrug that is being developed for the treatment of advanced breast cancer (fig. 7). TAT-59 has been shown to inhibit the growth of ER-positive, DMBA-induced rat mammary carcinomas (Toko et al., 1990). The drug inhibits the growth of estrogen-stimulated, ER-positive breast cancer cells transplanted into athymic mice (Koh et al., 1992; Iino et al., 1994). The drug is activated metabolically to a dephosphorylated form (Toko et al., 1990) that binds with high affinity to the ER (Toko et al., 1992). Clinical studies using TAT-59