Selective estrogen receptor modulators (SERMs) such as tamoxifen are effective in the treatment of many estrogen receptor-positive breast cancers and have also proven to be effective in the prevention of breast cancer in women at high risk for the disease. The comparative abilities of tamoxifen versus raloxifene in breast cancer prevention are currently being compared in the Study of Tamoxifen and Raloxifene trial. To better understand the actions of these compounds in breast cancer, we have examined their effects on the expression of ∼12,000 genes, using Affymetrix GeneChip microarrays, with quantitative PCR verification in many cases, categorizing their actions as agonist, antagonist, or partial agonist/antagonist. Analysis of gene stimulation and inhibition by the SERMs trans-hydroxytamoxifen (TOT) and raloxifene (Ral) or ICI 182,780 (ICI) and by estradiol (E2) in estrogen receptor-containing MCF-7 human breast cancer cells revealed that (a) TOT was the most E2-like of the three compounds, (b) all three compounds either partially or fully antagonized the action of E2 on most genes, with the order of antagonist activity being ICI > Ral > TOT, (c) TOT and Ral, but not ICI, displayed partial agonist/partial antagonist activity on a number of E2-regulated genes, (d) several stimulatory cell cycle-related genes were down-regulated exclusively by ICI, (e) the estrogen-like activity of Ral nearly always overlapped with that of TOT, indicating that Ral has little unique agonist activity different from that of TOT, and (f) some genes were specifically up-regulated by TOT but not Ral, ICI, or E2. Hence, gene expression profiling can discern fundamental differences among SERMs and provides insight into the distinct biologies of TOT, Ral, and ICI in breast cancer.
Selective estrogen receptor modulators (SERMs) are characterized by their diverse range of agonist/antagonist actions on estrogen receptor (ER)-mediated processes. They have the ability to act as either ER antagonists by blocking estrogen action through its receptor, as ER agonists by displaying estrogen-like actions, or as ER partial agonists/antagonists with mixed activity. Frequently, these differences in SERM activity depend upon the target gene promoter, as well as the cell or tissue background (1) . Two of the best characterized SERMs are tamoxifen and raloxifene (Ral), which are both considered to act predominantly as estrogen antagonists in breast cancer cells, blocking the effects of estrogens. However, even in breast cancer, SERMs have the potential to act as estrogen-like ER agonists when resistance to these compounds develop or in the case of certain ER mutations in which antiestrogens behave as estrogens (2, 3, 4) . Despite this potential, tamoxifen has been successfully used for many years as adjuvant therapy for hormone-responsive breast cancer (5) . Furthermore, in a large clinical trial, tamoxifen was shown to effectively prevent breast cancer in many women at high risk for the disease (6) . In addition, tamoxifen reduced the occurrence of bone fractures in these women; however, some detrimental side effects such as an increased risk of endometrial cancer, stroke, and pulmonary embolism were also associated with tamoxifen treatment (7) . Ral was examined in the Multiple Outcomes of Raloxifene Evaluation trial and found to be effective in reducing the incidence of osteoporosis in postmenopausal women, as well as the incidence of breast cancer but, unlike tamoxifen, without the increased risk of endometrial cancer (8 , 9) . On the basis of the positive outcome of these trials, the Study of Tamoxifen and Raloxifene trial was begun in 1999 to directly compare the effects of these two SERMs, tamoxifen and Ral, in prevention of breast cancer (10 , 11) .
The beneficial effects of SERMs on breast cancer were originally attributed to their ability to antagonize the actions of endogenous estrogens by competition for ER binding. More recently, X-ray structural work has demonstrated that when different ER ligands such as estradiol (E2), tamoxifen, Ral, and the antiestrogen ICI 182,780 (ICI) interact with the ligand binding domain of the receptor, distinctly different conformations of the receptor are induced (12, 13, 14, 15) . In addition, introduction of different mutations into the ER ligand binding domain demonstrated that the chemical nature of different ligands can determine which residues of the ligand binding domain can make contact with the ligand, additionally supporting the idea that ligands induce different conformations of the ER (3 , 16 , 17) . It has been suggested that as a result of these different ligand-induced conformations, the ER can thereby recruit different coregulator proteins to target gene promoters and differentially mediate gene transcription (18) . Although ER agonists like E2 recruit transcriptional coactivators such as those of the p160 family that can enhance gene transcription, SERMs and ICI generally do not appear to recruit coactivators; rather, they promote the recruitment of corepressors such as NCoR and SMRT that can actively suppress transcription of E2 target genes, thereby additionally antagonizing the action of estrogens (19, 20, 21, 22) .
SERMs, however, do not always act as ER antagonists, and they can often display estrogen-like agonist activity or mixed agonist/antagonist activity. The mechanisms for these mixed effects of SERMs depend on several factors, including the differing chemical nature of the ligand, the cell background, and the context of the gene promoter (23 , 24) . For example, tamoxifen has been shown to act either as a partial or full agonist on different DNA regulatory elements such as particular estrogen response elements, Sp1 sites, or activator protein 1 sites (25, 26, 27, 28, 29) . Several lines of evidence also indicate that SERM agonist activity depends on different regions of the receptor than does E2 for its agonist activity. In particular, the NH2-terminal transactivation domain activation function-1 is important for tamoxifen agonist activity, whereas the synergistic action of both activation function-1 and activation function-2 is important for estrogen-induced activity (30, 31, 32, 33) . It is also clear that tamoxifen and E2 use different regions of activation function-1, as was demonstrated with receptor mutations and different peptides that could antagonize either E2 or tamoxifen agonism (32 , 34 , 35) . The agonist activity of SERMs at particular genes may be accompanied by the recruitment of the same (24) or distinct (36) coactivators to the ER as are recruited by E2. It has also been suggested that the availability of coregulators can determine whether tamoxifen will act as an agonist or antagonist. For example, in mouse embryonic fibroblast cells from an NCoR knockout model, tamoxifen displayed agonist activity through the ER, whereas it showed antagonist activity in wild-type mouse embryonic fibroblast cells (37) . Also, in MCF-7 cells overexpressing SRC-1, tamoxifen but not Ral acted as an agonist on particular genes (24) .
Although tamoxifen and Ral prevent E2-induced proliferation of breast cancer cells and are assumed to antagonize many of the actions of estrogens, it is not known globally which genes and gene networks in the cell are altered by SERMs, on which genes SERMs antagonize E2 regulation fully or only partially, and on which genes SERMs exert significant E2-like agonist effects in breast cancer cells. It is also not known to what extent the SERMs or ICI can regulate the expression of genes independently from E2 action, as we observed previously for quinone reductase (38) . Furthermore, in contrast to tamoxifen and Ral, the antiestrogen ICI is thought to act as a pure antagonist through the ER, although its effects on gene expression have not been fully explored (39) .
With the advent of microarray technology, the effects of SERMs on gene expression can be compared on a very large number of genes without any prior selection bias. In fact, gene expression profiling in breast cancer using microarrays has been carried out in a number of studies investigating genes overexpressed in breast cancer (40) or genes associated with clinical outcome or prognosis (41, 42, 43) , response to chemotherapy (44) , tumor aggressiveness (45) , or classification of primary tumors (46, 47, 48, 49) . Several studies have also investigated gene expression patterns associated with the ER status of breast cancers (50, 51, 52, 53, 54) . Although there have been several microarray studies examining the actions of SERMs (55, 56, 57, 58) , these have not directly compared the SERMs tamoxifen and Ral and ICI and their agonistic and/or antagonistic actions on a large set of estrogen-regulated genes. Therefore, in an effort to examine the effects of these three compounds on E2-regulated gene expression in ER-positive breast cancer cells, we carried out gene expression profiling using oligonucleotide microarrays. Our findings indicate that although the regulation of the majority of E2-regulated genes is either partially or fully reversed by the SERMs and by ICI, distinct differences can be observed among these ligands in their balance of agonist, partial antagonist, or full antagonist activities on the spectrum of E2-regulated genes. In addition, a unique subset of genes, encoding proteins that may have beneficial effects, was found to be regulated by the SERMs and/or ICI but not by E2.
MATERIALS AND METHODS
Cell Culture and RNA Extraction.
MCF-7 human breast cancer cells were routinely cultured in MEM (Sigma Chemical Co., St. Louis, MO) supplemented with 5% calf serum (Hyclone, Logan, UT) and antibiotics. Four days before E2 treatment, cells were switched to phenol red-free MEM containing 5% charcoal dextran-treated calf serum. Media were changed on day 2 and day 4 of culture. Cells were treated with 10 nm E2 alone or in the presence of 1 μm ICI, Ral, or trans-hydroxy-tamoxifen (TOT) or with 1 μm ICI, Ral, or TOT alone in the absence of E2 for 8 or 48 h. Because multiple combinations of compounds were examined, these time points were chosen based on our earlier time course study of E2 regulation of gene expression in these cells (59) . We observed generally similar patterns of gene stimulation at 4 and 8 h, but with more genes down-regulated at 8 versus 4 h; hence we selected 8 h as our first time point. Because we observed very similar gene regulations at 24 and 48 h, we selected 48 h as our later time point. Real-time PCR determinations involved three independent experiments and microarray determinations involved two independent experiments. Total RNA was prepared using Trizol Reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. RNA was additionally purified using RNeasy columns (Qiagen, Valencia, CA) and treatment with RNase free-DNase I (Qiagen).
Total RNA was used to generate cRNA, which was labeled with biotin according to techniques recommended by Affymetrix (Santa Clara, CA). cRNA was then hybridized to Affymetrix Hu95A GeneChips, which contain oligonucleotide probe sets for ∼12,000 human genes. After washing, the chips were scanned and analyzed using MicroArray Suite 5.0 software (Affymetrix). Average intensities for each GeneChip were globally scaled to a target intensity of 150. Additional analysis was performed using GeneSpring software V5.0.1 (Silicon Genetics, Redwood City, CA) to obtain fold-change and Ps for each gene for each treatment relative to the vehicle control. The entire microarray data set will be available through the Gene Expression Omnibus accession no. GSE848. 3
To identify genes significantly regulated by E2, gene lists were created in GeneSpring using a fold-change cutoff of 2.5 for up-regulated genes and 0.4 for down-regulated genes (i.e., 2.5-fold down-regulation). We then compared this list to one we generated recently using a confidence scoring method in detailed E2 time course gene expression microarray profiling experiments also in MCF-7 cells (59) . Only genes that were regulated in both studies were used because this set represents genes that are reproducibly and robustly regulated by E2. The SERMs (TOT and Ral) and ICI were then analyzed for their agonist or antagonist activity on this set of E2-regulated genes.
To identify unique genes, i.e., genes that are regulated only by SERMs or ICI but not by E2, we used the following criteria: (a) a fold change for up-regulated genes of ≥2.0 for the SERMs or ICI but <1.3 for E2; (b) a fold change for down-regulated genes of ≤0.5 for the SERMs or ICI but >0.78 for E2; (c) P of <0.1; (d) present calls with SERM or ICI treatment; and (e) a raw expression level of >20. These cutoffs enable the identification of robust changes in gene expression, as documented in previously published microarray work of our lab and others (59 , 60) .
Real-time PCR was carried out to verify regulation of gene expression by E2, SERMs, or ICI. One μg of total RNA was reverse transcribed in a total volume of 20 μl using 200 units of reverse transcriptase, 50 pmol random hexamer, and 1 mm deoxynucleotide triphosphates (New England Biolabs, Beverly, MA). The resulting cDNA was then diluted to a volume of 100 μl with sterile water. Each real-time PCR reaction consisted of 1 μl of diluted reverse transcription product, 1× SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA), and 50 nm forward and reverse primer. Reactions were carried out in an ABI Prism 7700 Sequence Detection System (Applied Biosystems) for 40 cycles (95°C for 15 s and 60°C for 1 min) after an initial 10 min incubation at 95°C. For the genes, the regulation of which is described in detail in this article, the primers used for real-time PCR are listed in Table 1 ⇓ . The fold change in expression of each gene was calculated using the ΔΔCt (threshold cycle) method, with the ribosomal protein 36B4 mRNA as an internal control (61) .
Effects of SERMs and ICI on Genes Significantly Regulated by E2.
Gene expression profiling was carried out on RNA from MCF-7 cells treated with E2, the SERMs Ral or TOT, or ICI, either alone or in combination with E2 for 8 or 48 h. As shown in Fig. 1 ⇓ and as we recently reported (59) , E2 inhibited the expression of about twice as many genes as E2 stimulated in these breast cancer cells. Two sets of genes, one highly and reproducibly up-regulated by E2 (n = 40) and one down-regulated by E2 (n = 89), were then identified as described in “Materials and Methods” and used for examining the effects of the SERMs or ICI either alone or in combination with E2. Gene cluster analysis, as shown in Fig. 1 ⇓ , reveals that the up-regulation (in red) and the down-regulation (in blue) of gene expression elicited by E2 were in large part reversed by each of the three compounds. This is clearly visible in the clusters labeled A (Fig. 1) ⇓ for E2 up-regulated genes and B for E2 down-regulated genes. However, several exceptions to this as well as some additional interesting patterns of regulation can be observed. Cluster C shows several genes stimulated more robustly by E2 and TOT than by Ral or ICI, and cluster D contains genes down-regulated by both E2 and TOT. Cluster E highlights two genes encoding the drug-metabolizing enzymes Cyp1A1 and Cyp1B1 that were stimulated by all four ligands, especially so at the 8 h time point.
SERM and ICI Agonistic and/or Antagonistic Activity on E2-Regulated Genes.
To examine what effect the SERMs or ICI might have on E2-regulated gene expression, the activity of each compound was first assessed on those genes found to be significantly up- or down-regulated by E2 (Fig. 2) ⇓ . We defined a compound as either having no agonistic activity (≤35% of E2 activity), partial agonistic activity (if the compound alone evoked activity > 35% but ≤70% of E2 activity), or full agonistic activity (if the compound evoked >70% of E2 activity). As expected, on those genes where SERMs displayed full agonistic activity, no antagonism of E2 action by the SERM was observed. Also, on those genes where SERMs displayed partial agonistic activity, the SERM generally acted as a partial antagonist, reducing E2 activity to the level seen with the SERM alone (see below). On those genes where the compound alone displayed no E2-like agonistic activity, we defined the compound as a full antagonist if the effect of E2 could be reversed by the compound by ≥50% or as a partial antagonist if the compound reversed the E2 effect by 30–50%. If reversal of E2 activity by the compound was <30%, we considered this to be no antagonism. In every case where the compound had no agonist activity, we found that it acted as either a full or partial antagonist of E2 action.
As shown in Fig. 2 ⇓ , ICI had the greatest antagonist activity among the 3 compounds evaluated. ICI antagonized E2 action on 95% of E2 up-regulated genes and 91% of E2 down-regulated genes. Ral acted as an antagonist on 67% of E2 up-regulated genes and 33% of E2 down-regulated genes, whereas TOT antagonized E2 action on only 47% of E2 up-regulated genes and 26% of E2 down-regulated genes.
Both Ral and TOT displayed partial agonist/antagonist activity on a larger proportion of the genes than did ICI, with this being 25% for Ral and 30% for TOT on E2 up-regulated genes. On E2 down-regulated genes, Ral and TOT displayed more partial agonist/antagonist activity, with Ral showing partial agonist/antagonist activity on 63% of the genes and TOT displaying partial agonist/antagonist activity on 43% of these genes. Only TOT acted as a full agonist on a substantial percentage of genes, 23% of the E2 up-regulated genes, and 31% of the E2 down-regulated genes. In contrast, both Ral and ICI had full agonist activity on ≤8% of E2 up- or down-regulated genes.
We next examined whether TOT, Ral, or ICI were acting as agonists, partial agonist/antagonists, or antagonists on the same genes, or whether each of the compounds was regulating expression of different subsets of E2-regulated genes. To address this question, we compared the activities of ICI, Ral, and TOT on each gene and identified four major combinations of activities. As shown in Fig. 3 ⇓ , all three compounds (white area) acted as antagonists of E2 action (i.e., reversed partially or fully the stimulation or the suppression of gene expression by E2) on 47% of the genes up-regulated by E2 and on 20% of the genes down-regulated by E2. These genes, on which all of the compounds acted as antagonists of E2 action, are given in Table 2 ⇓ . On 25% of E2 up-regulated genes and 54% of E2 down-regulated genes, only ICI antagonized E2 action (stippled area, Fig. 3 ⇓ ), whereas TOT and Ral acted as either full or partial agonists on these genes (Table 3) ⇓ . On 23% of E2 up-regulated genes and 16% of E2 down-regulated genes (black area, Fig. 3 ⇓ ), both ICI and Ral acted as antagonists, whereas only TOT acted as a partial or full agonist (Table 4) ⇓ . On only a small percentage of genes (5% of up-regulated and 10% of down-regulated, striped area, Fig. 3 ⇓ ), all three compounds acted as either full or partial agonists (Table 5) ⇓ .
Distinct patterns of gene regulation by the SERMs were verified by real-time PCR, as demonstrated in Figs. 4 ⇓ and 5 ⇓ . On several E2 up-regulated genes, both Ral and TOT but not ICI displayed agonist activity; this is seen for the calcitonin receptor (Fig. 4A) ⇓ and insulin-like growth factor binding protein-4 (Fig. 4B) ⇓ . Of the three compounds, only TOT displayed substantial E2-like activity on Rab31 (Fig. 4C) ⇓ and phospholipase C-like 2 (PLC-L2) gene expression (Fig. 4D) ⇓ . For E2 down-regulated genes, including MAX dimerization protein 4 (MAD 4), B-cell linker (BLNK), and RAP1 GTPase activating protein 1 (RAP1GAP) (Fig. 5) ⇓ , both TOT and Ral displayed partial agonist/antagonist activity, whereas ICI alone had no E2-like down-regulating activity and instead displayed antagonist activity, largely reversing the effect of E2.
Overall, these findings indicate that TOT is capable of acting as a full agonist on more E2-regulated genes than either Ral or ICI. In contrast, ICI’s activity is primarily as an antagonist of the E2 effect on gene expression, with little agonistic activity. Of note, Ral had no agonistic effects on any genes different from those on which TOT was an agonist, such that all of Ral’s agonistic activity overlapped with that of TOT.
Unique Gene Expression Regulation by SERMs and ICI but not E2.
Using the technique described in “Materials and Methods,” we identified 24 genes that were up-regulated and 51 genes that were down-regulated by the SERMs or ICI that were not significantly regulated in the same manner by E2. Gene cluster analysis was performed for these genes to identify any major patterns of regulation, and the results are shown in Fig. 6 ⇓ . Cluster A demonstrates up-regulated genes that appear to be stimulated to a greater extent by the three compounds than by E2. In contrast, cluster B demonstrates that most of the down-regulated genes were down-regulated specifically by ICI, with only a few being down-regulated by Ral or TOT but not E2. Also, a set of genes that is up-regulated specifically by TOT but not by E2, Ral, or ICI is shown in cluster C. The identity of these SERM- and ICI-regulated genes and their fold change in gene expression in response to these ligands are given in Table 6 ⇓ . Immediate early response 3, also called IEX-1, represents an additional gene found to be highly and specifically induced by TOT but not Ral or ICI, but it is also significantly down-regulated by E2 (Table 2) ⇓ . Real-time PCR was performed to verify regulation by the SERMs or ICI. Rab30 is an example of a gene stimulated by TOT but not by E2, Ral, or ICI (Fig. 7A) ⇓ . Additional studies demonstrated that regulation of Rab30 by TOT is mediated by the ER because 100-fold excess of E2 could reverse the effect of TOT (Fig. 7B) ⇓ . The genes cyclin A2 and cdc2 have their expression markedly down-regulated by ICI only and this is reversed by E2 (Fig. 7, C and D) ⇓ . Interestingly, these two genes and the majority of genes down-regulated by ICI (Table 6) ⇓ are associated with the control of cell cycle progression.
Tamoxifen and Ral are of significant value in breast cancer treatment and prevention and are currently under investigation in the Study of Tamoxifen and Raloxifene trial (4, 5, 6, 7, 8, 9, 10, 11) , but the genes that are regulated by these SERMs have not been broadly examined. Gene expression profiling using oligonucleotide microarrays can provide this information. Indeed, our findings indicate that in the MCF-7 breast cancer cell line, where TOT and Ral and the antiestrogen ICI all act as physiological antagonists of E2 action in terms of cell proliferation, each of these ER ligands evokes a very different pattern of gene regulation. Overall, ICI, Ral, and TOT acted as either partial or full antagonists on the majority of E2 up- and down-regulated genes. However, in contrast to ICI, which displayed primarily antagonistic activity only, both Ral and TOT displayed partial agonist/antagonist behavior on a substantial proportion of E2 up- and down-regulated genes. Furthermore, TOT had more estrogen-like activity that either Ral or ICI, with full agonist behavior on 23% of E2 up-regulated and 31% of E2 down-regulated genes. Interestingly, any agonist activity that was seen with Ral overlapped with that of TOT. In addition to these effects on E2-regulated genes, TOT, Ral, and ICI were observed to regulate a small subset of genes that were not regulated by E2, but which appear to be mediated by ER because a 100-fold excess of E2 reversed the SERM effect.
SERMs and ICI as Antagonists of Estrogen Action through the ER.
As might be expected, ICI, the pure antiestrogen, antagonized E2 action on >95% of E2-regulated genes. Similarly, Ral acted as an antagonist on >90% of E2 regulated genes, whereas TOT antagonized E2 action on fewer genes (∼70%). Despite the anticipated nature of these results, two interesting points can be made. First, on all genes where the SERMs and ICI did not have any partial or full agonist activity, they always antagonized E2 action to some extent, and second, very little of the antagonism by ICI was accompanied by partial agonism, but for Ral and TOT, partial agonist/antagonist activity was more frequent.
Also of note is the functional nature of the genes on which all of the SERMs acted as antagonists (Table 2) ⇓ . As we have previously demonstrated, E2 up-regulates a number of genes that would have stimulatory effects on cell proliferation such as cell cycle-associated genes, growth factors, and transcription factors, as well as down-regulating numerous genes that would inhibit cell proliferation (59) . On the basis of the results reported here in this study, it is apparent that many of the genes on which the SERMs act as antagonists could affect cell proliferation. For example, all three of the compounds antagonized the E2 up-regulation of the transcription factor c-fos, the DNA synthesis regulator CDC6, and the growth stimulatory growth factors amphiregulin and chemokine ligand 12, the last of which is also known as SDF-1 and has previously been shown to be antagonized by ICI (62) . Similarly, all three compounds antagonized the E2 down-regulation of growth inhibitory factors such as transforming growth factor β2, inhibin βB, and IEX-1, the last of which is also known as IER3 and has been shown to inhibit breast cancer cell growth (63) . These findings suggest that although the SERMs show agonist activity on some E2-regulated genes in MCF-7 cells (as discussed below), their ability to block the E2 stimulation of cell proliferation suggests that the genes they antagonize are those that are essential for the stimulatory effect of E2 on cell proliferation.
Agonist Activities of SERMs and ICI.
Although ICI activity was almost always antagonistic to E2 action, TOT and Ral displayed a fairly high degree of partial agonist/antagonist activity, whereas only TOT displayed any substantial full agonist activity, indicating that these ligands have very different natures. On the other hand, any agonistic activity seen with Ral nearly always overlapped with that of TOT, which suggests that the agonist activity of Ral may be through a similar mechanism as TOT.
Perhaps one of the more interesting findings revealed by this study is that TOT had full agonist activity on a number of genes on which Ral displayed only partial or no agonist activity, as was the case with phospholipase C-like 2 and Rab31 (Fig. 4) ⇓ . These findings suggest that these ligands have different activities at different target gene sites because of their abilities to induce different conformations in the receptor (12, 13, 14, 15 , 23) . The functional significance of these differences is not currently known but is under investigation.
Unique Patterns of Gene Regulation by the SERMs.
In addition to the agonist/antagonist activities of the SERMs and ICI on E2-regulated genes, several genes were identified that were specifically regulated by the SERMs or ICI but were not significantly regulated by E2. Very few genes showing this unique “reverse pharmacology” have previously been identified. However, quinone reductase, an important suppressor of DNA damage in breast cancer cells, was previously identified by this laboratory as an antiestrogen induced gene (38 , 64) . In the current study, one potentially important gene displaying a similar reverse pharmacology is retinoblastoma 1 coiled coil protein, which was significantly up-regulated by all of the SERMs but not by E2 (Table 6) ⇓ . This gene has been proposed to act as a tumor suppressor by its ability to up-regulate retinoblastoma 1 levels, and it could provide an additional beneficial effect of SERMs on breast cancer cells (65 , 66) . Although the mechanism of SERM-regulated expression of retinoblastoma 1 coiled coil protein is not known, recent evidence suggests that one mechanism by which SERMs can regulate gene expression independently from E2 action could be through a squelching mechanism, whereby SERM interaction with ER leads to the recruitment of corepressors to the ER and away from genes that are under some basal level repression (67) .
In addition to genes up-regulated by all of the SERMs, several genes were identified in this study as being specifically up-regulated by TOT but not E2, Ral, or ICI. This included Rab30, a small GTPase, the enzyme 5α-reductase type I, which converts testosterone to dihydrotestosterone, tropomyosin 1, which is a cytoskeletal protein (Table 6) ⇓ , and IER3/IEX-1 (Table 2) ⇓ . All of these genes appear to be regulated through the ER because excess E2 or ICI can block up-regulation of these genes (Fig. 7 ⇓ and data not shown). This finding additionally supports the idea that TOT is capable of inducing a different and unique receptor activity, most likely through a different receptor conformation, from that of the other ligands (23) .
It is of interest that several of these genes have potential tumor suppressor or antiproliferative activities in breast cancer cells and could contribute to the beneficial effects of TOT in breast cancer. For example, the up-regulation of 5α-reductase could potentially reduce local E2 levels in vivo through the conversion of androgens to more potent, nonaromatizable androgens rather than to estrogens. Furthermore, 5α-reductase expression has been detected in breast cancer cells and is inversely correlated with proliferation markers such as Ki67 (68) . There is also evidence that dihydrotestosterone decreases breast cancer cell proliferation (69 , 70) . Both tropomyosin and IEX-1, which are up-regulated by TOT, have been shown to inhibit proliferation of breast cancer cells, although the mechanisms for these antiproliferative effects are not known (63 , 71) . These findings suggest that TOT, in addition to antagonizing estrogen action through the ER at certain genes, may have additional beneficial effects through its ability to up-regulate other specific target genes.
Unique and specific gene regulation was also seen with ICI, but in contrast to TOT, these genes were specifically down-regulated by ICI but not Ral or TOT, and this down-regulation was reversed by E2. The majority of these genes appear to be regulators of the cell cycle, cell proliferation, and DNA synthesis. This is supportive of observations that ICI very effectively arrests the proliferation of breast cancer cells in the G0 phase of the cell cycle (72) . Therefore, by down-regulating the expression of these genes, ICI may have an additional beneficial effect over the other SERMs. Several of these genes are known to actually be up-regulated by E2; however, as we have shown previously, E2 stimulation of these genes occurs at late time points only and may be secondary responses to E2 (59) . Two potential upstream transcriptional regulators of these genes are c-Myc and E2F1, which are also down-regulated by ICI (Table 6) ⇓ . One mechanism that might explain the down-regulation of these genes by ICI is that ICI can increase turnover and decrease ER protein levels in breast cancer cells, thereby suppressing any potential ligand-independent activity of the receptor in these cells (73) . Thus, ICI may be suppressing growth factor activity through the ER, which has been previously demonstrated for epidermal growth factor and insulin-like growth factor I actions in both breast cancer and uterine cells (74, 75, 76, 77) .
Here, we demonstrate that the major actions of the SERMs tamoxifen and Ral and of the antiestrogen ICI are largely antagonistic of E2 action. Because E2 regulates a large number of genes in several different pathways that promote cell proliferation, decrease apoptosis, and regulate other activities in these breast cancer cells (59) , these antagonistic, tumor-suppressive actions of SERMs should be very desirable. Despite these common antagonistic actions, clearly distinct patterns of gene regulation were observed by microarray profiling for each of these three ligands, indicating that conformational differences in these ER-ligand complexes translate into different pharmacological phenotypes. Some of the genes that are regulated uniquely by the SERMs or ICI might also be contributing to the beneficial and somewhat different effects of these compounds when they are used as endocrine therapies for breast cancer.
We thank Mark Band, Dennis Akan, and Lei Liu of the University of Illinois Biotechnology Center for their contributions to these studies.
Grant support: NIH Grant CA18119 and Breast Cancer Research Foundation (to B. S. Katzenellenbogen) and NIH Grant T32 HD07028.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Dr. Benita S. Katzenellenbogen, University of Illinois, Department of Molecular and Integrative Physiology, 524 Burrill Hall, 407 South Goodwin Avenue, Urbana, IL 61801-3704. Phone: (217) 333-9769; Fax: (217) 244-9906; E-mail:
↵3 Internet address: http://www.ncbi.nlm.nih.gov/geo.
- Received October 22, 2003.
- Revision received December 3, 2003.
- Accepted December 9, 2003.
- ©2004 American Association for Cancer Research.