Estren Behaves as a Weak Estrogen Rather than a Nongenomic ...

0013-7227/06/$15.00/0 Printed in U.S.A.

Endocrinology 147(5):2203–2214 Copyright © 2006 by The Endocrine Society doi: 10.1210/en.2005-1292

Estren Behaves as a Weak Estrogen Rather than a Nongenomic Selective Activator in the Mouse Uterus Sylvia C. Hewitt, Jennifer Collins, Sherry Grissom, Katherine Hamilton, and Kenneth S. Korach Receptor Biology Section, Laboratory of Reproductive and Developmental Toxicology (S.C.H., K.H., K.S.K.), and Microarray Group (J.C., S.G.), National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina 27709 A proposed membrane-mediated mechanism of rapid nongenomic response to estrogen has been the intense focus of recent research. Estren, a synthetic steroid, is reported to act selectively through a rapid membrane-mediated pathway, rather than through the classical nuclear estrogen receptor (ER)-mediated pathway, to maintain bone density in ovariectomized mice without uterotropic effects. To evaluate the mechanism and physiological effects of estren, we studied responses in adult ovariectomized mice. In a 3-d uterine bioassay, we found that 300 ␮g estren significantly increased uterine weight; in comparison, a more maximal response was seen with 1 ␮g estradiol (E2). The estren response was partly ER␣ independent, because ER␣ knockout (␣ERKO) uteri also exhibited a more moderate weight increase. Estren induced epithelial cell proliferation in wild-type, but not ␣ERKO, mice, indicating ER␣ dependence of the epithelial growth response. Examination of estren-regulated uterine genes by microarray indicated that early (2 h) changes in gene expression are similar to the early responses to E2. These gene responses are ER␣

T

HE RODENT UTERUS is a useful in vivo model to study estrogen receptor ␣ (ER␣)-mediated responses to estrogen. Uterine sensitivity to estrogenic substances is reflected in well-characterized biological responses that culminate in increased tissue weight and epithelial cell proliferation. Loss of these responses in mice lacking the ER␣ (␣ERKO) and their preservation in mice lacking the ER␤ (␤ERKO) illustrate the essential role of ER␣ in mediating uterine responses to E (1). We have previously described a microarray study that examined the acute endogenous uterine gene responses to estradiol (E2). A temporal biphasic pattern to the gene changes was apparent, with distinct gene changes occurring early (within 2 h) and late (within 24 h) that reflected biological changes in the uterus that also occur in a biphasic manner (2). The roles of ER␣ and ER␤ were also evaluated using ERKO tissues in this same study, which indicated that ER␣ is essential to estrogen-induced gene changes in the uterus (2). Additionally, the uterine response to ligands other than E2, such as environmental toxicants or First Published Online February 9, 2006 Abbreviations: AR, Androgen receptor; BrdU, bromodeoxyuridine; DHT, dihydrotestosterone; E2, estradiol; ER, estrogen receptor; ERE, estrogen response element; ICI, ICI 182,780; KO, knockout; 19NT, 19nortestoterone; TBST, 20 mm Tris (pH 7.4), 200 mm NaCl, and 0.1% Tween 20; WT, wild type. Endocrinology is published monthly by The Endocrine Society (http:// www.endo-society.org), the foremost professional society serving the endocrine community.

dependent, because they are not seen in ␣ERKO mice. Later estren-induced changes in gene expression (24 h) are blunted compared with those seen 24 h after E2. In contrast to early genes, these later estren responses are independent of ER␣, because the ␣ERKO shows a similar response to estren at 24 h. We found that E2 or estren treatments lead to depletion of ER␣ in the uterine cytosol fraction and accumulation in the nuclear fraction within 30 – 60 min, consistent with the ability of estren to regulate genes through a nuclear ER␣ rather than a nongenomic mechanism. Interestingly, estren, but not E2, induces accumulation of androgen receptor (AR) in the nuclear fraction of both wild-type and ␣ERKO samples, suggesting that AR might be involved in the later ER␣-independent genomic responses to estren. In conclusion, our studies suggest that estren is weakly estrogenic in the mouse uterus and might induce nuclear ER␣- and AR-mediated responses. Given its activity in our uterine model, the use of estren as a bone-selective clinical compound needs to be reconsidered. (Endocrinology 147: 2203–2214, 2006)

activators of various cell-signaling pathways, such as those initiated by growth factors, can be examined and compared with E2 responses, leading to better understanding of their mechanisms (3). The mechanism of gene regulation by the ER␣ involves activation of nuclear-localized ER␣ protein by estrogenic ligands, allowing assembly of appropriate transcriptional regulators, leading to changes in transcript levels (1). Recent studies have described an alternate mechanism of estrogen response in which estrogen interacts with plasma membrane-associated ER␣ protein, leading to rapid activation of cytosolic signaling through pathways such as phosphatidylinositol kinase-3 or MAPK pathways and ultimately to altered gene expression (4). The natural ER agonist E2 interacts with and activates both the classical nuclear ER␣ pathway and the more novel membrane ER␣ pathway. A synthetic steroid, estren, has recently been reported to selectively activate this membrane ER␣ signaling pathway, although it lacks the ability to interact with nuclear ER␣ (5). Initial studies suggested that estren’s biological effects are targeted to bone maintenance, but lack uterotropic activity (5, 6). Subsequent analyses, however, indicate that estren does cause a significant uterine weight increase (7), although this is blunted compared with the effect of E2. Additionally, estren can induce an estrogen-responsive reporter construct via a conventional nuclear ER␣ mechanism (7). Because estren should lead to regulation of uterine genes that are targets of the membrane ER␣ pathway, whereas E2 will regulate uter-

2203

2204

Endocrinology, May 2006, 147(5):2203–2214

ine genes that are targets of the membrane pathway in addition to genes that are regulated by the nuclear ER␣ pathway, we used the estren compound in our microarray study design with the expectation that its membrane signaling selective mode should result in gene responses that are a subset of those of E2 and represent targets of the membrane ER␣ pathway in the uterus. Additionally, the role of ER␣ in the estren gene response can be evaluated by comparing the estren response of the ␣ERKO. The uterine genomic responses after estren treatments were not consistent with a membrane-selective signaling mechanism, prompting additional investigation of the biological and biochemical effects of estren in the mouse uterus.

Hewitt et al. • Estren Responses in Mouse Uterus

genes represented (Agilent Technologies, Palo Alto, CA). Five hundred nanograms of total RNA was amplified and labeled using the Agilent Low RNA Input Fluorescent Linear Amplification Kit according to the manufacturer’s protocol. For each two-color comparison, 750 ng of each Cy3- and Cy5-labeled cRNA were mixed and fragmented using the Agilent In Situ Hybridization Kit protocol. Hybridizations were performed for 17 h in a rotating hybridization oven according to the Agilent 60-mer oligo microarray processing protocol before washing and scanning with an Agilent scanner (Agilent Technologies, Wilmington, DE). Each sample pair was hybridized on two replicate chips. Data were obtained using Agilent Feature Extraction software (version 7.5), with defaults for all parameters. The Feature Extraction software performs error modeling before data are loaded into the Resolver system. Images and GEML files, including error and P values, were exported from the Agilent Feature Extraction software and deposited into Rosetta Resolver (version 5.0, build 5.0.0.2.48; Rosetta Biosoftware, Kirkland, WA).

Materials and Methods Animals

Data analysis

All animal studies were carried out using a protocol approved by the National Institute of Environmental Health Sciences (NIEHS) animal care and use committee and in accordance with National Institutes of Health guidelines for the humane use of laboratory animals. Adult female C57BL6J mice were either purchased from Charles River Laboratories (Raleigh, NC) or generated in the NIEHS ␣ERKO colony at Taconic Farms (Germantown, NY). Mice were ovariectomized, either at the vendor (Charles River Laboratories) or at NIEHS and were fed standard NIH-31 chow for 10 –14 d to clear endogenous ovarian steroids before use.

The Rosetta Resolver system performs a squeeze operation that creates ratio profiles by combining replicates while applying error weighting. The error weighting consists of adjusting for additive and multiplicative noise. The resultant ratio profiles were combined into ratio experiments as described by Stoughton and Dai (8). A P value is generated and propagated throughout the system. The P value represents the probability that a gene is differentially expressed. The Resolver system allows users to set thresholds, below which genes of a P value are considered to be significantly expressed. In this study, the threshold was set at P ⬍ 0.001 for each comparison.

Uterine bioassay

Real-time RT-PCR

Ovariectomized mice were treated with daily sc injections of sesame oil (vehicle), E2 (Steraloids, Newport, RI; 1 ␮g/mouse), or estren (Steraloids; 300 ␮g/mouse) for 3 d. On the fourth day, uteri were collected and weighed. The weights of each group were analyzed by ANOVA, and sample groups were compared with the vehicle group using Dunnett’s test.

Mice were treated with vehicle or estren as described for microarray analysis; some were also injected with 45 ␮g of the ER antagonist ICI 182, 780 (ICI; Tocris, Ellisville, MO) or 2 mg of the androgen receptor (AR) antagonist flutamide (Sigma-Aldrich Corp.) 30 min before the injection of estren. RNA from three individual uteri per group was prepared using TRIzol. cDNA was synthesized and analyzed by real-time PCR using SYBR Green dye as previously described (3). Relative transcript levels were quantified in comparison with wild-type (WT) vehicle control and normalized to 18S or Rpl7 as a reference using the model described by Pfaffl (9). Primer sequences were selected using Primer Express (Applied Biosystems, Foster City CA) and were purchased from Sigma-Genosys (St. Louis, MO). Primer sequences are shown in Table 1.

Immunohistochemical analysis Mice were treated with a single injection of estren (Steraloids); 22 h later, bromodeoxyuridine (BrdU; 1 mg/mouse; Sigma-Aldrich Corp., St Louis, MO) was injected. Uteri were collected 2 h later (24 h after the estren injection), fixed in 10% formalin, and embedded in paraffin. Uterine cross-sections were stained for BrdU using the Oncogene BrdU kit (HCS30; Oncogene, Cambridge, MA) following the manufacturer’s protocol. Ki67 was detected after decloaking with heat and pressure for 3 min in citrate buffer in a decloaking chamber (Biocare, Walnut Creek, CA), blocking with 3% H2O2 (Fisher Scientific, Fairlawn, NJ) for 10 min, followed by 5% normal rabbit serum (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 10 min. Anti-Ki67 (TEC 3; DakoCytomation, Carpinteria, CA) diluted 1:80 in automation buffer (Biocare) was applied for 1 h at room temperature, followed by biotinylated rabbit antirat IgG (Vector Laboratories, Inc., Burlingame CA) diluted 1:300 for 30 min at room temperature. Avidin-biotin peroxidase complex (Vector Laboratories, Inc.) was applied for 30 min, and 3,3⬘-diaminobenzidine substrate (DakoCytomation) was used to develop the signal. Slides were counterstained in hematoxylin (Sigma-Aldrich Corp.), dehydrated, and coverslipped.

Microarray analysis Mice were treated with sesame oil vehicle (Sigma-Aldrich Corp.), E2 (Steraloids; 1 ␮g/mouse), estren (300 ␮g/mouse), dihydrotestosterone (DHT; Steraloids; 50 ␮g/mouse) or 19-nortestoterone (19NT; Steraloids; 100 ␮g/mouse), and uteri were collected 2 or 24 h after treatment and snap-frozen in liquid nitrogen. Three or four uteri from each treatment group were pooled, and RNA was prepared using TRIzol reagent (Invitrogen Life Technologies, Inc., Carlsbad CA), and the RNeasy clean-up protocol (QIAGEN, Valencia, CA). Gene expression profiling was conducted using Agilent Mouse Oligo arrays with approximately 20,000

Nuclear/cytoplasmic protein extracts and Western blot analyses Mice were treated with saline vehicle, E2, estren (50, 150, or 300 ␮g), DHT (Steraloids; 50 ␮g/mouse), or 19 nor-testosterone (Steraloids; 100 ␮g/mouse) by ip injection as described above, and uteri were collected after 1 h or another indicated time and snap-frozen in liquid nitrogen. Tissue was pulverized, then homogenized in TEGM buffer [10 mm Tris (pH 7.6), 1 mm EDTA, 10% glycerol, and 3 mm MgCl2] supplemented with protease inhibitors (7.5 mm EGTA and 50 ␮g/ml each of antipain, leupeptin, chymostatin, and soybean trypsin inhibitor) using a Polytron homogenizer (PT1200C, Brinkmann Instruments, Westbury NY). Homogenates were filtered through Nitex (Tetko, Briarcliff Manor, NY) and centrifuged at 1000 ⫻ g for 10 min at 4 C to pellet the nuclei. The supernatant was cleared at 42,000 rpm for 45 min using a 42.2 Ti rotor (Beckman Coulter, Fullerton, CA) at 4 C and collected as the cytosol fraction. The nuclear pellet was washed twice by resuspension in TEGM buffer and centrifugation at 1000 ⫻ g for 10 min at 4 C. The nuclei were resuspended in 50 ␮l TEGM buffer. Fifty microliters of 10% SDS and 9 ␮l 5 m NaCl were then added and mixed using a pestle, and the sample was heated at 95 C for 5 min to extract nuclear proteins. Nuclear extracts were cleared using a 42.2 Ti rotor at 42,000 rpm for 1 h and 30 min at 25 C. Protein concentrations in the extracts were determined using a bicinchoninic acid assay (Pierce Chemical Co., Rockford, IL). Two micrograms of nuclear extract or 4 ␮g cytosol was loaded onto 10% NuPage gels (Invitrogen Life Technologies, Inc., Carlsbad, CA), separated, and


TABLE 1. Primers used for real-time PCR analysis Primers and sequences

18s F GAAACTGCGAATGGCTCATTAA 966 –987 R GAATCACCACAGTTATCCAAGTAGGA 1046 –1021 Apoe F GAGCCGGAGGTGACAGATCA 106 –125 R GGTAATCCCAGAAGCGGTTCA 184 –164 Ccnb1 F TTGTGTGCCCAAGAAGATGCT 786 – 806 R GTACATCTCCTCATATTTGCTTGCA 861– 837 Ccnb2 F ATGTCAACAAGCAGCCGAAAC 301–321 R GAGGACGATCCTTGGGAGCTA 378 –358 Cdc2a F GGACGAGAACGGCTTGGAT 872– 890 R GGCCATTTTGCCAGAGATTC 947–928 Cdkn1a (p21) F CAGCGACCATGTCCAATCC 193–211 R CGAAGAGACAACGGCACACTT 264 –244 Cyr 61 F CAGGATGCTCCAGTGTCAAGAA 1085–1106 R GCAGCACCGGCCATCTAC 1152–1135 Fos F GGAATGGTGAAGACCGTGTCA 328 –348 R CCTCTTCAGGAGATAGCTGCTCTAC 409 –385 Inhbb F GCTCATCGGCTGGAACGA 84 –101 R GCCCTCACAGTAGTTCCCGTAGT 147–125 Mad211 F TGGTAGTGTTCTCCGTTCGATCT 22– 44 R GCAGGGTGATGCCTTGCT 88 –71 Rpl7 F AGCTGGCCTTTGTCATCAGAA 353–373 R GACGAAGGAGCTGCAGAACCT 430 – 410 Sdf2l1 F GCGGCCAACAGTCGGTAA 246 –263 R GCGAATCCGCCAGTAACTATTG 310 –289 Sin3b F CGGTCTGGAGATGGGATAAGC 1123–1143 R TGGGAAGTGCCCGATAGC 1201–1184 Sox4 F GGCCCATGAACGCCTTT 838 – 854 R TCGGGCGACTGCTCCAT 904 – 888 Timp1 F GCAGATATCCGGTACGCCTACA 241–262 R TGCGGTTCTGGGACTTGTG 313–295 Tnnt2 F TTCTGCGAAACCGGATCAAT 844 – 863 R CAACGCCCGGTGACTTTG 916 – 899 Txnip F ACCACTTTCTCGGATGTTGGA 1714 –1734 R GGAAAGACAACGCCAGAAGGT 1797–1777 Ube2c F CCTCATGACATCTGGTGACAAAG 175–197 R ATGGTCCCCACCCACTTGA 249 –231 Primers were designed using Primer Express software (Applied Biosystems). Forward (F) and reverse (R) primer sequences are given for each gene as well as the bases to which they anneal in the mouse sequence. transferred to nitrocellulose membranes according to the manufacturer’s protocol. Membranes were stained with Ponceau S solution (SigmaAldrich Corp.) to ensure even loading and transfer of proteins. Membranes were blocked in 20 mm Tris (pH 7.4), 200 mm NaCl, and 0.1% Tween 20 (TBST) with 5% milk (ICN Biomedicals, Irvine, CA) for 1 h at room temperature. Anti-ER␣ (sc 7207, Santa Cruz Biotechnology, Inc., Santa Cruz, CA) or anti-AR (sc 816, Santa Cruz Biotechnology, Inc.) were diluted 1:1000 in TBST-5% milk and incubated with the membrane for 1 h. Membranes were washed in TBST, then incubated with horseradish peroxidase-antirabbit IgG (Cell Signaling Technology, Beverly MA) at a


2205

dilution of 1:5000 in 5% milk-TBST for 1 h. Blots were washed and developed using ECL Plus (GE Healthcare, Piscataway, NJ). Blots were initially analyzed for ER␣, then stripped using Restore Western blot stripping buffer (Pierce Chemical Co.) and reanalyzed for AR.

Results Evaluation of physiological and genomic uterine responses to estren

In a 3-d uterine bioassay, estren caused a significant increase in uterine wet weight (2.1-fold; Table 2), although the increase was less pronounced than that resulting from E2 treatment (4.1-fold). Interestingly, a significant, but less prominent, uterine wet weight increase was seen in estrentreated ␣ERKO mice (1.6-fold), which lack the ER␣ and thus both nuclear and membrane signaling pathways, indicating an ER␣-independent component to the increase. Because the uterotropic response is the result of both increased fluid content and epithelial proliferation, uterine epithelial cells were evaluated for indications of proliferative response. Tissue was collected 24 h after treatment with estren, because this represents the time of peak DNA synthesis, and sections were stained for BrdU incorporation or the presence of Ki67 antigen, both markers of proliferating cells. BrdU and Ki67 are clearly detected in estren-treated uterine sections (Fig. 1). ER␣ dependence of this response is illustrated by the lack of proliferative markers in epithelial cells from estren-treated ␣ERKO tissues (Fig. 1). The acute genomic responses of the uterus to estren were evaluated using a microarray to assess differentially represented genes in uterine RNA from animals treated with vehicle, E2, or estren for 2 or 24 h (representing early and late gene changes, respectively). In addition, ␣ERKO samples were examined to determine the ER␣ dependence of gene responses (Fig. 2 and Table 3). As previously reported, few gene responses were seen in the E2-treated ␣ERKO uterus, indicating ER␣ dependence of the E2 response (2). The diminished response of ␣ERKO mice to 2 h of estren compared with that of WT mice indicates that the early estren response was also ER␣ dependent (Fig. 2A and Table 3). In the WT uterus, estren gene responses were very similar to those seen with E2, because the cluster of significantly regulated genes showed a similar pattern and signal intensity. A correlation plot of the genes generated in Rosetta Resolver indicated almost perfect correlation of common signature genes (genes regulated in both conditions) of WT E2 or estren responses after 2 h (0.954) and a strong correlation of all signature genes (genes regulated in either condition; 0.881; Fig. 2C). The mostly overlapping responses to E2 and estren are inconsistent with estren’s proposed membrane ER␣-selective mechanism. The estren response at 24 h was similar to that of E2 in TABLE 2. Uterine weights (milligrams per uterus ⫾ 3-d bioassay

Vehicle E2 Estren a b

SE),

WT

␣ERKO

16.6 (0.4) 68.0 (4.9)a, 4.1-fold 34.8 (2.2)a, 2.1-fold

15.3 (0.6) 14.1 (1.5), 0.9-fold 24.9 (1.1)b, 1.6-fold

P ⬍ 0.05 vs. WT vehicle. P ⬍ 0.05 vs. ␣ERKO vehicle.

2206



FIG. 1. ER␣-dependent epithelial cell proliferation is induced by estren. Ovariectomized WT or ␣ERKO mice were injected with estren. After 22 h, BrdU was injected. Uterine tissue was collected 2 h later and processed as described in Materials and Methods to detect Ki67 antigen or BrdU. Each section is representative of three individual uteri. Scale bar, 40 ␮M. Arrows indicate antigen-positive epithelial cells.

terms of the gene responses, but was less robust, as illustrated in the cluster showing a similar pattern, but decreased signal intensity (Fig. 2B and Table 3). A correlation plot of genes regulated by E2 or estren at 24 h shows nearly perfect correlation of common signature genes (0.938) and a strong correlation of all signature genes (0.822; Fig. 2D). Additionally, the estren responses of the WT and ␣ERKO at 24 h appear to be similar in pattern and intensity (Fig. 2B), suggesting ER␣-independent estren responses for many genes. As previously reported (2, 3), the E2 responses at 24 h are mostly ER␣ dependent, because the ␣ERKO response to E2 is greatly diminished compared with the WT E2 responses (Fig. 2B). Several gene responses noted at 2 h were evaluated by real-time RT-PCR to confirm the regulation indicated by the microarray. Selected genes included those previously indicated to be increased or decreased by E2 in the uterus and to require ER␣ (2, 3). Microarray data indicated that estren also regulated these genes after 2 h in an ER␣-dependent manner. RT-PCR analysis confirmed that, as reported previously for E2 (2, 3), estren decreases Txnip and Sox4 transcripts (Fig. 3). This decrease is ER␣ dependent, as illustrated by the reversal of the Txnip and Sox4 repression responses seen with the ER antagonist ICI in the WT mice and the lack of regulation of either gene by estren in the ␣ERKO mice (Fig. 3). Similarly, as was seen previously with E2 (2, 3), estren increases the transcripts for Mad2l1, Inhbb, Cyr61, Fos, and Cdkn1a (p21; Fig. 3). ICI inhibited the increase in Inhbb and Mad2l1 by estren, and these transcripts were not increased in the ␣ERKO, indicating the ER␣ dependence of these responses. Although ICI inhibited the estren-induced increases in Cdkn1a (p21), Cyr61, and Fos in the WT mice, indicating ER␣ dependence, there was a residual increase in these three transcripts in the ␣ERKO mice, suggesting that there may be an ER␣-independent component to the regulation of these genes by estren. The minimal estren initiated increases in Cdkn1a (p21), Cyr61, and Fos in ␣ERKO mice are not altered by ICI, indicating that these residual responses are not ER mediated.

Therefore, the effect of the AR antagonist, flutamide, was tested to indicate whether AR signaling might play a role in the estren regulation of these transcripts. Flutamide did not alter these estren responses in the WT mice, but did inhibit the minimal increases in Fos and Cyr61 transcripts in the ␣ERKO mice, suggesting a role for AR in this estren response, whereas no effect was seen on the Cdkn1a (p21) transcript. Evaluation of ER subcellular localization in response to estren

Although the 2 h gene responses were ER␣ dependent, they also included most of the genes regulated by E2. Because of the mechanism described previously (5, 6), we expected that E2 would regulate genes through both the membrane ER␣ and the nuclear ER␣ pathways, whereas estren would only regulate those E2-responsive genes that used the membrane ER␣ pathway. The similar early gene response profiles of E2 and estren are not consistent with the previously described membrane ER␣-selective mode of action of estren. Therefore, the effect of estren on the subcellular localization of ER␣ was evaluated. Previous studies have shown that the ER␣ is depleted in the cytosolic fraction and is increased in the nuclear fraction of uterine homogenates isolated from E2-treated ovariectomized mice. This nuclear translocation can be seen within minutes and is maintained for several hours (10). The ER␣ in cytosolic and nuclear fractions of uterine homogenates from ovariectomized mice treated for 1 h with vehicle, E2 or estren was analyzed by Western blot (Fig. 4A). Both E2 and estren decreased the levels of cytosolic ER␣ and increased nuclear ER␣ levels. Estren has also been reported to have a higher affinity for AR than for ER␣ (11); therefore, the effect of estren on uterine AR translocation was evaluated (Fig. 4A). Estren, but not E2, caused a decrease in cytosolic AR and increased nuclear AR. As controls, the AR agonists DHT and 19-nor-testosterone were evaluated, and these compounds did not alter ER␣ localization but did cause AR nuclear translocation.



2207

FIG. 2. Uterine gene responses to E2 and estren. Ovariectomized mice were injected with vehicle, estren, 19NT, or E2. Uterine tissue was collected after 2 or 24 h, and RNA was isolated for microarray analysis as described in Materials and Methods. Two hour (A) and 24 h (B) significant (P ⬍ 0.001) changes in gene expression between vehicle and each treatment are displayed as a heat map. Each horizontal row indicates significant differences in gene expression from analysis of paired sample hybridizations (E, estradiol; Es, estren). Each vertical line represents a gene that is increased (red) or decreased (green) compared with vehicle, with the color intensity indicating the degree of change. A gene list was generated and exported to Excel (see supplemental Tables S1 and S2, published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org). C, A correlation plot of Veh vs. E2 2 h and WT Veh Vs. estren 2 h was generated in Rosetta Resolver. Nonweighted and error-weighted (using Resolver’s error model) values are displayed. Genes with a value of P ⱕ 0.001 in both comparisons are plotted (Visible). Genes that were anticorrelated, but were significant at P ⫽ 0.001 in both comparisons are not displayed, but were considered in the correlation coefficient calculation for Common Signature. All Signature refers to genes that were significant at P ⫽ 0.001 in either comparison, but not necessarily both comparisons. All refers to the correlation of all genes on the array that were not flagged (no fails) or were not controls. D, A correlation plot of Veh vs. E2 24 h and WT Veh Vs. estren 24 h was generated in Rosetta Resolver as described above for C. (Figure continues on next page.)

Because estren is described as possessing a rapid mode of activity, the time courses of ER␣ and AR receptor translocation were analyzed in tissue extracts from animals treated with E2 or estren for 30 min, 1 h, 8 h, or 16 h. Both E2 and estren induced ER␣ nuclear translocation within 30 min (Fig. 4B), which was maintained at 1 h and then began to decrease. Our studies used a dose of 300 ␮g/mouse because the reported Kd of estren for ER␣ was 200- to 300-fold higher (⬃250 nm) than that of E2 (⬃1 nm), whereas the Kd of estren for AR was lower than that for ER␣ (⬃50 nm) (11, 12). Therefore, the

effect of estren dose on ER␣ and AR translocation was studied to determine whether the higher affinity of estren for AR would be reflected in a differential ability to translocate ER␣ and AR at the lower doses (Fig. 4C). Estren was able to effect ER␣ nuclear translocation only at the highest dose, but caused AR translocation at the lower doses as well. Because ␣ERKO uterine tissue contains a residual truncated ER␣ protein, (E1-R␣), present at low abundance (13), the effect of estren on E1-ER␣ nuclear translocation was also examined (Fig. 4D). E1-ER␣ can be visualized as a more

2208



FIG. 2. Continued

rapidly migrating band on the Western blot (Fig. 4D, arrow). Similar to WT-ER␣, estren causes nuclear translocation of E1-ER␣ as well as AR in the ␣ERKO uterus, whereas E2 causes translocation of E1-ER␣, but not AR. Mechanism of 24-h ER␣-independent estren responses

The preservation of estren gene responses in the absence of ER␣ at 24 h was unexpected, because the membrane ER mechanism for estren previously described involved ER␣ (6). Our observation of AR translocation by estren in the uterine tissue (Fig. 4) and by others in cell culture (12) in addition to reports of estren metabolism to the AR agonist 19NT (11) suggest that these 24 h gene responses may be AR mediated. To investigate this possibility, AR-mediated gene regulation

in the uterus was evaluated by microarray analysis after DHT treatment for 24 h. In agreement with a previous report by Nantermet et al. (14), as illustrated in the cluster analysis (Fig. 5A), uterine gene responses to DHT were similar to those of E2 but were less robust, as indicated by the lower color intensity, and did not include all the responses seen to E2 or estren treatment. A correlation plot of responses to estren or DHT showed a strong correlation of common signature genes (0.844; Fig. 5B), although not as substantial as that observed with the E2 and estren comparisons (Fig. 2, C and D), and a lower, but still stronger, correlation (0.63) for all signature genes, also lower that the previous correlation (Fig. 2, C and D). This suggests that AR might mediate some, but not all, of the estren responses at 24 h. Alternatively, the

TABLE 3. Numbers of genes significantly regulated in each condition Experiment

WT E2, 2 h Estren, 2 h ␣ERKO E2, 2 h Estren, 2 h WT E2, 24 h Estren, 24 h 19NT, 24 h ␣ERKO E2, 24 h Estren, 24 h 19NT, 24 h

Significant at P ⬍ 0.001 (up)

Significant at P ⬍ 0.001 (down)

Significant at 2-fold and P ⬍ 0.001 (up)

Significant at 2-fold and P ⬍ 0.001 (down)

1577 1370

1121 1031

517 356

227 167

556 502

245 322

103 121

3 41

2208 1181 1087

1726 834 1054

676 173 76

514 115 59

959 857 183

601 676 151

83 114 14

14 80 5


FIG. 3. Verification of ER␣-dependent, 2-h estren responses by RTPCR. Genes previously shown to be E2 regulated (2, 3) were selected for additional analysis by RT-PCR. WT or ␣ERKO samples were prepared after mice were treated with estren (Es) for 2 h. Some were also treated with the ER antagonist ICI or the AR antagonist, flutamide (Flut). Each data point is the average of three uterine samples; error bars indicate the SD. Txnip, Thioredoxin-interacting protein; Sox4, SRY box containing gene 4; Mad2l1, mitotic arrest-deficientlike 1; Inhbb, inhibin ␤-B; Cdkn1a cyclin-dependent kinase inhibitor 1A (P21); Fos FBJ, osteosarcoma oncogene; Cyr61, cysteine-rich protein 61.

changes in the tissue seen after 24 h of estren (increased weight and proliferation), but not with DHT treatment might account for the greater genomic response observed with estren compared with DHT. Additionally, 19NT has been shown to be a rapidly formed estren metabolite and to mediate AR transcriptional activity through an estrogen response element (ERE) sequence in a reporter gene assay (11). To assess whether later estren gene responses might be mediated by this ERE-directed AR-19NT metabolite complex, gene responses to 24-h treatment with 19NT were examined. The response was similar to that of estren; however, unlike the estren result, the responses were completely ER␣ dependent, because gene regulation in the


2209

␣ERKO was minimal (Fig. 2B and Table 3). A correlation plot of WT responses to estren or 19NT showed almost perfect correlation of common signature genes (0.92; Fig. 5C) and lower, but still strong, correlation (0.69) for all signature genes, indicating mostly overlapping responses in the WT mice. 19NT is a substrate for aromatase (15) and is probably converted to E2 in mice, thus leading to ER␣-dependent gene responses. The different responses seen in ␣ERKO mice with estren or 19NT treatment suggests that the estren conversion to 19NT is not complete, and the ER␣-independent estren responses seen after 24 h are due to remaining estren or other metabolites. Several up-regulated genes observed in the microarray analysis were also examined by RT-PCR. Ccnb1, Ccnb2, Cdc2a, Cnn1, Sdf2l1, Timp1, Tnnt2, and Ube2c were all increased by E2 or estren in WT mice on the microarray. All except for Sdf2l1 were also increased by DHT in the microarray, suggesting a role for AR. When assayed by RT-PCR, estren increased the levels of Ccnb1, Ccnb2, Cdc2a, Cnn1, Sdf2l1, Timp1, Tnnt2, and Ube2c in WT mice. Several genes (Ccnb1, Ccnb2, Cdc2a, and Ube2c) were not increased in the microarray dataset by estren in ␣ERKO mice. This observation was confirmed by RT-PCR. RT-PCR analysis shows that estren increased the levels of Cnn1, Sdf2l1, and Tnnt2 in ␣ERKO mice, indicating that estren-mediated increases of these are ER␣ independent. To evaluate the relative roles of ER and AR in this mechanism, the effects of the ER antagonist ICI or the AR antagonist flutamide were examined. In the presence of ER␣ (WT), ICI inhibited the estren-mediated increase in Ccnb2 and Ube2c and partially inhibited the increase in Ccnb1, confirming the apparent ER mediation of these responses by estren. ICI did not alter the level of any of the transcripts induced by estren in ␣ERKO mice. In WT mice, flutamide inhibited only the increase in Tnnt2. Interestingly, in ␣ERKO mice, flutamide partly or fully inhibited the estren-mediated increases in Cnn1, Sdf2l1, and Tnnt2, indicating a role for AR. This effect was not seen with Sdf2l1 or Cnn1 in the presence of ER␣, suggesting a difference in the response of these models, potentially due to metabolic or cellular differences. Flutamide also prevented a small residual estren induction in Ccnb1, Ccnb2, and Cdc2a in ␣ERKO mice. Clearly, there are mechanistic differences in these models that are more complex than the presence or absence of ER␣. Three down-regulated genes observed in the microarray analysis were examined by RT-PCR. On the microarray, ApoE, Sin3b, and Txnip were repressed by E2 or estren in WT mice and by estren in ␣ERKO mice. By RT-PCR Sin3b, Txnip, and ApoE are repressed by estren (Fig. 6) and less effectively than with E2 (not shown) in WT uterus. The estren responses of these down-regulated genes in ␣ERKO mice are similar to those in WT mice, indicating the ER␣ independence of these estren responses (Fig. 6). In WT mice, neither ICI nor flutamide reversed down-regulation of these transcripts, whereas in ␣ERKO mice, flutamide partly inhibited estrenmediated down-regulation of Sin3b and effectively inhibited the decrease in Txnip. ICI was also partially effective in ␣ERKO mice in both responses.

2210



FIG. 4. Western blot: subcellular localization of ER␣ and AR. A, Cytosolic and nuclear proteins were extracted from uterine tissues 1 h after the injection of various compounds (Cpd): vehicle (V), E2 (E), estren (Es), DHT (D), or 19NT (NT). E2 causes selective translocation of ER␣, whereas estren leads to translocation of both ER␣ and AR, as seen by a decrease in the cytosolic level and an increase in the nuclear extract. AR agonists, DHT or 19NT, selectively induce translocation of AR. This blot is representative of three blots, each run with independent samples. B, Time course of receptor translocation. Uterine tissue was collected 0.5, 1, 8, or 16 h after E2 or estren injection. Cytosolic and nuclear proteins were extracted. ER␣ translocates into the nucleus as early as 30 min after the injection of E2 or estren. AR increases in the nuclear extracts within 30 min of estren injection. This blot is representative of three blots, each run with independent samples. C, The effect of estren dose on receptor translocation. Cytosolic and nuclear proteins were extracted from uterine tissues 1 h after the injection of vehicle (V), E2, or three different doses of estren (50, 150, or 300 ␮g/mouse). Estren caused ER␣ to increase in the nuclear fraction only at the highest dose, but caused AR translocation at all three doses. This blot is representative of three blots, each run with independent samples. D, Estren induces translocation of E1-ER␣ residual splice variant in ␣ERKO uterus. Cytosolic and nuclear proteins were extracted from uterine tissues 1 h after injection of vehicle (V), E2, or estren treatment in WT or ␣ERKO mice. ER␣ protein is easily detected in WT extracts, but is not detected in the ␣ERKO extracts when the Western blot is exposed to film for 5 sec. After a 30-sec exposure, the E1-ER␣ splice variant was detected in the ␣ERKO (arrow), and either E2 or estren led to nuclear translocation of the variant. Estren caused AR translocation in both WT and ␣ERKO extracts. This gel is representative of two experiments, each run with independent samples.

Discussion

In this study microarray analysis of endogenous uterine genes was initially undertaken in an attempt to discern responses specific to the membrane ER␣-associated rapid signaling pathway from those resulting from nuclear ER␣-mediated mechanisms. However, after 2 h of estren treatment, the estren gene response was similar to that of E2 and was ER␣ dependent, as shown by the lack of responses in ␣ERKO

mice. These observations, which suggest that estren is not acting solely through membrane-mediated mechanisms, prompted additional evaluation of the biological and biochemical changes in the uterus in response to estren. In a 3-d uterine bioassay, estren resulted in a significant increase in uterine weight compared with animals treated with vehicle, although the increase was attenuated compared with that induced by E2. Our observation is consistent with



2211

FIG. 5. Microarray analysis of 24-h DHT uterine transcripts. As in Fig. 2, uterine tissue was collected 24 h after the injection of vehicle or DHT, and RNA was isolated for microarray analysis as described in Materials and Methods. A, Significant (P ⬍ 0.001) gene changes are displayed as a heat map compared with the gene change seen with E2 or estren (Es). A list was exported from Rosetta Resolver representing the genes displayed in the heat map in A (see supplemental Table S3). B, A correlation plot of Veh vs. estren 24 h and WT Veh Vs. DHT 24 h was generated in Rosetta Resolver as described in Fig. 2C. C, A correlation plot of Veh vs. estren 24 h and WT Veh vs. 19 NT 24 h was generated in Rosetta Resolver as described in Fig. 2C.

a previous report of estren-induced uterine weight increase (7). Our study also indicated that this increase was, at least in part, ER␣ independent, because a significant weight increase from estren was also apparent in ␣ERKO mice. In agreement with a non-ER␣ component, neither ER (ICI) nor AR (flutamide) antagonists inhibited the estren response of uterine weight increase, whereas the combination of ICI and flutamide together was effective (data not presented), suggesting that the full effect or estren is mediated by a combination of AR and ER␣ pathways. In a previous study, no uterine weight increase was seen by Moverare et al. (7) in response to estren in ␣␤ERKO double-knockout animals. Several differences in study design may reconcile the disparate observations. Moverare et al. (7) used a lower dose of

estren (75 ␮g) than our dose (300 ␮g). We chose our dose based on the reported Kd of estren for ER␣ (11, 12). We show in this study that the lower (75 ␮g) dose would selectively translocate the AR, but not the ER␣, to the nucleus, because neither 50 nor 150 ␮g was effective. This suggests that the uterine weight increase we report in ␣ERKO mice requires the higher dose (300 ␮g), which would translocate the E1ER␣, whereas Movare et al. (7), in contrast, may have selectively activated AR in ␣␤ERKO mice, resulting in no uterine weight response. Moverare et al. (7) administered E2 or estren for 4 wk, resulting in a more pronounced increase in uterine weight with E2 (⬃10-fold), whereas we used a 3-d assay resulting in an more modest increase (⬃4-fold). The uterine weights of WT mice after their 4-wk estren exposure or our

2212



FIG. 6. Analysis of ER-independent estren responses in WT and ␣ERKO by real-time RT-PCR. Top, Minicluster displaying microarray data for genes selected for verification. RT-PCR, Uterine RNA was from mice treated for 24 h with vehicle (V) or estren (Es). Some were also treated with the ER antagonist ICI or the AR antagonist, flutamide (Flut). Each point is an average of three samples. The error bars are the SD. Ccnb1 and Ccnb2, Cyclins B1 and B2; Cdc2a, cell division cycle 2 homolog A; Sdf2l1, stromal cell-derived factor 2-like 1; Timp1, tissue inhibitor of metalloproteinase 1; Tnnt2, troponin T2, Ube2c ubiquitin-conjugating enzyme E2C; Apoe, apolipoprotein E; Sin3b transcriptional regulator; SIN3B, Txnip thioredoxin-interacting protein.


3-d treatment were comparable (⬃2- to 3-fold). The more pronounced uterine weight increase in WT mice after the 4-wk E2 exposure compared with a minimal estren response in ␣␤ERKO mice even after prolonged exposure may mask detection of a statistically significant weight increase in ␣␤ERKO mice. Finally, the model used by Moverare et al. (7) lacked both ER␣ and ER␤, which is reported to result in more pronounced uterine hypoplasia than in the ␣ERKO mice we used (16). Thus, the lack of both receptors might result in the development of a less responsive tissue in general. Our study clearly demonstrates that ER␣-dependent uterine epithelial cell proliferation results from estren administration to WT mice, whereas no proliferation occurs in ␣ERKO tissue. The early ER␣-dependent gene responses seen in our microarray study may allow identification of the genes that underlie this ER␣-dependent proliferative response. Moverare et al. (7) reports that estren activates an ERdependent reporter gene in transfected Hek293 cells and concludes that estren is a selective ER mediator, with contextdependent ER-mediated activity, rather than a nongenomic selective estrogen. Our observation of estren-induced nuclear translocation of ER␣, together with the genomic response patterns indicate that estren is a weak estrogen that uses a conventional nuclear receptor pathway. Other weak or short-acting estrogens, such as estriol or 16␣-estradiol, bind ER␣, but unlike more potent ER agonists such as E2, they lack the ability to maintain ER nuclear localization and therefore exhibit early estrogenic activity, but provide diminished biological response without repeated administration (17, 18). Our observations of estren-induced early gene changes similar to those induced by E2, but attenuated late gene response, and the blunted 3-d uterotropic response induced by estren compared with E2 indicate that estren lacks full estrogenicity and therefore acts as a weak estrogen. The estren-induced gene responses observed at the later time point (24 h) were similar to those of E2, but were blunted and were ER␣ independent, because they were also observed in ␣ERKO uteri. ER␤ mediation seems unlikely to account for these responses, because our previous study clearly showed a dependence on ER␣ for E2-induced gene changes (2). Several elements suggest that there may be an AR-mediated component to the later ER␣-independent estren gene responses, including 1) estren’s affinity for the AR (Kd ⫽ 53 nm for AR, 247 nm for ER␣) (11, 12), 2) estren’s metabolism to the AR agonist 19NT by the liver enzyme 3␣-hydroxysteroid dehydrogenase (11), and 3) our observation of AR nuclear translocation in response to estren in the uterus and by others in cell culture (12). The premise of AR-mediated late gene responses proved difficult to corroborate. We examined gene responses to the AR agonist DHT, with the expectation that similar responses after 24-h treatments with estren or DHT would point to the involvement of AR in the mechanism. The number of DHTinduced gene changes was modest compared with those of E2 or estren, but some overlap is observed, suggesting that some genes may be regulated by both AR and ER␣. In fact, 19NT has been reported to allow AR-mediated regulation of a normally ER-responsive ERE-controlled reporter gene in vitro (11). Conversion of estren to 19NT in our in vivo model and AR-mediated (and therefore ER␣-independent) regula-


2213

tion of estrogen-responsive uterine genes would account for our observation of an ER␣-independent estren gene response at 24 h. However, genes regulated by 19NT overlap with estren responses, but require the presence of ER␣ for the 19NT response. This is probably due to efficient conversion of 19NT to estradiol by aromatase (15). This finding suggests that when injected into mice, estren is not efficiently metabolized to 19NT, as was observed in previous studies in bone cells (11). If estren were completely converted to 19NT, its uterine gene responses would not greatly differ from those of 19NT; instead, estren responses are ER␣ independent, whereas 19NT responses are ER␣ dependent, potentially due to conversion to E2. This suggests that the estren conversion to 19NT is not complete, and the ER␣-independent estren responses seen after 24 h are due to remaining estren or other metabolites. A second approach to elucidate the ER␣-independent, 24-h estren gene responses was additional analysis of genes selected from the microarray study based on their estren regulation in an ER␣-independent manner. The preserved regulation of several of these genes by estren in ␣ERKO uteri supports the lack of a role for ER␣ in some of the 24-h estren responses. These observations are merely suggestive and do not directly indicate the mechanism mediating the ER␣-independent estren responses. The simultaneous activation of ER and AR responses by estren and its metabolites may result in interaction between pathways, illustrating that there are multiple regulators for some genes in our uterine model. The requirement for ER␣ for the proliferative response to estren is clear. Several of the ER␣-dependent gene responses are known to be involved in control of cell cycle progression, including Mad2l1, Cdkn1a (p21), and Cdc2a. Additional analysis of the networks of regulated genes might illustrate the role of ER␣ in permitting cell cycle progression. In conclusion, although estren has been promoted as a nongenomic estrogen and as a selective treatment for bone maintenance (5, 6), our study and other recent reports indicate that estren possesses biological activity in the uterus (7). Whether estren mediates biological processes in other tissues, such as breast, remains to be determined. In addition, estren’s mechanism of action appears to be more complex than originally described (5–7, 11, 12); therefore, the AR- and nuclear ER-mediated responses (7, 11, 12) need to be considered when assessing its potential medical uses. Acknowledgments We acknowledge Page Myers, James Clark, and Clay Rouse for ovariectomies; Vickie Walker and Linwood Koonce for maintenance of mouse colonies; the National Institute of Environmental Health Sciences Microarray Group for data acquisition and analysis; and Grace Kissling for statistical analysis of uterine bioassay data. We are grateful to Drs. Richard Hochberg and Evan Simpson for advice regarding the potential metabolic products of administered compounds. Received October 12, 2005. Accepted February 2, 2006. Address all correspondence and requests for reprints to: Sylvia C. Hewitt, Receptor Biology Section, Laboratory of Reproductive and Developmental Toxicology, National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina 27709. E-mail: [email protected].

2214


This work was supported by Intramural Research at National Institute of Health. None of the authors has conflicts to disclose.

References 1. Hewitt SC, Harrell JC, Korach KS 2005 Lessons in estrogen biology from knockout and transgenic animals. Annu Rev Physiol 67:285–308 2. Hewitt SC, Deroo BJ, Hansen K, Collins J, Grissom S, Afshari CA, Korach KS 2003 Estrogen receptor-dependent genomic responses in the uterus mirror the biphasic physiological response to estrogen. Mol Endocrinol 17:2070 –2083 3. Hewitt SC, Collins J, Grissom S, Deroo B, Korach KS 2005 Global uterine genomics in vivo: microarray evaluation of the estrogen receptor-growth factor cross-talk mechanism. Mol Endocrinol 19:657– 668 4. Kelly MJ, Levin ER 2001 Rapid actions of plasma membrane estrogen receptors. Trends Endocrinol Metab 12:152–156 5. Kousteni S, Bellido T, Plotkin LI, O’Brien CA, Bodenner DL, Han L, Han K, DiGregorio GB, Katzenellenbogen JA, Katzenellenbogen BS, Roberson PK, Weinstein RS, Jilka RL, Manolagas SC 2001 Nongenotropic, sex-nonspecific signaling through the estrogen or androgen receptors: dissociation from transcriptional activity. Cell 104:719 –730 6. Kousteni S, Chen JR, Bellido T, Han L, Ali AA, O’Brien CA, Plotkin L, Fu Q, Mancino AT, Wen Y, Vertino AM, Powers CC, Stewart SA, Ebert R, Parfitt AM, Weinstein RS, Jilka RL, Manolagas SC 2002 Reversal of bone loss in mice by nongenotropic signaling of sex steroids. Science 298:843– 846 7. Moverare S, Dahllund J, Andersson N, Islander U, Carlsten H, Gustafsson JA, Nilsson S, Ohlsson C 2003 Estren is a selective estrogen receptor modulator with transcriptional activity. Mol Pharmacol 64:1428 –1433 8. Stoughton RS, Dai H (inventors), Rosetta Inpharmatics, I(Asignee) 2002 Statistical combining of cell expression profiles. US Patent 6,351,712 9. Pfaffl MW 2001 A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res 29:2002–2007


10. Golding TS, Korach KS 1988 Nuclear estrogen receptor molecular heterogeneity in the mouse uterus. Proc Natl Acad Sci USA 85:69 –73 11. Centrella M, McCarthy TL, Chang WZ, Labaree DC, Hochberg RB 2004 Estren (4-estren-3␣,17␤-diol) is a prohormone that regulates both androgenic and estrogenic transcriptional effects through the androgen receptor. Mol Endocrinol 18:1120 –1130 12. Krishnan V, Bullock HA, Yaden BC, Liu M, Barr RJ, Montrose-Rafizadeh C, Chen K, Dodge JA, Bryant HU 2005 The nongenotropic synthetic ligand 4-estren-3 17␤-diol is a high-affinity genotropic androgen receptor agonist. Mol Pharmacol 67:744 –748 13. Couse JF, Curtis SW, Washburn TF, Lindzey J, Golding TS, Lubahn DB, Smithies O, Korach KS 1995 Analysis of transcription and estrogen insensitivity in the female mouse after targeted disruption of the estrogen receptor gene. Mol Endocrinol 9:1441–1454 14. Nantermet PV, Masarachia P, Gentile MA, Pennypacker B, Xu J, Holder D, Gerhold D, Towler D, Schmidt A, Kimmel DB, Freedman LP, Harada SI, Ray WJ 2005 Androgenic induction of growth and differentiation in the rodent uterus involves the modulation of estrogen-regulated genetic pathways. Endocrinology 146:564 –578 15. de Gooyer ME, Oppers-Tiemissen HM, Leysen D, Verheul HA, Kloosterboer HJ 2003 Tibolone is not converted by human aromatase to 7-methyl-17-ethynylestradiol (7-MEE): analyses with sensitive bioassays for estrogens and androgens and with LC-MSMS. Steroids 68:235–243 16. Dupont S, Krust A, Gansmuller A, Dierich A, Chambon P, Mark M 2000 Effect of single and compound knockouts of estrogen receptors (ER) and ␤ (ER␤) on mouse reproductive phenotypes. Development 127:4277– 4291 17. Lan NC, Katzenellenbogen BS 1976 Temporal relationships between hormone receptor binding and biological responses in the uterus: studies with short- and long-acting derivatives of estriol. Endocrinology 98:220 –227 18. Katzenellenbogen BS 1984 Biology and receptor interactions of estriol and estriol derivatives in vitro and in vivo. J Steroid Biochem 20:1033–1037

Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the endocrine community.