Estrogen-mediated Regulation of Igf1 Transcription and Uterine Growth Involves Direct Binding of Estrogen Receptor α to Estrogen-responsive Elements*

Estrogen enables uterine proliferation, which depends on synthesis of the IGF1 growth factor. This proliferation and IGF1 synthesis requires the estrogen receptor (ER), which binds directly to target DNA sequences (estrogen-responsive elements or EREs), or interacts with other transcription factors, such as AP1, to impact transcription. We observe neither uterine growth nor an increase in Igf1 transcript in a mouse with a DNA-binding mutated ERα (KIKO), indicating that both Igf1 regulation and uterine proliferation require the DNA binding function of the ER. We identified several potential EREs in the Igf1 gene, and chromatin immunoprecipitation analysis revealed ERα binding to these EREs in wild type but not KIKO chromatin. STAT5 is also reported to regulate Igf1; uterine Stat5a transcript is increased by estradiol (E2), but not in KIKO or αERKO uteri, indicating ERα- and ERE-dependent regulation. ERα binds to a potential Stat5a ERE. We hypothesize that E2 increases Stat5a transcript through ERE binding; that ERα, either alone or together with STAT5, then acts to increase Igf1 transcription; and that the resulting lack of IGF1 impairs KIKO uterine growth. Treatment with exogenous IGF1, alone or in combination with E2, induces proliferation in wild type but not KIKO uteri, indicating that IGF1 replacement does not rescue the KIKO proliferative response. Together, these observations suggest in contrast to previous in vitro studies of IGF-1 regulation involving AP1 motifs that direct ERα-DNA interaction is required to increase Igf1 transcription. Additionally, full ERα function is needed to mediate other cellular signals of the growth factor for uterine growth.

Estrogen is a critical mediator of female reproductive system development and function. Additionally, estrogen is involved in nonreproductive tissues, such as the cardiovascular and skeletal systems, and is implicated in several diseases, including cancer and osteoporosis (1,2). Estrogen interacts with the nuclear estrogen receptor (ER), 2 which binds directly to palindromic target DNA sequences, estrogen-responsive elements (EREs).
In vitro studies in model systems have shown that ERs can also interact with or be "tethered" to other transcription factors, such as AP1, to impact genes regulated by the corresponding motifs (3). A mutation that disrupts the direct DNA binding ability of the ER␣ (4, 5) has been "knocked in" at the ER␣ locus of a mouse (4,6). Female mice that carried a single copy of this nonclassical ER knock-in "NERKI" mutation were infertile because of ovarian and uterine defects. To circumvent this issue, NERKI/ER␣WT males were crossed to female mice heterozygous for the ER␣ null allele (␣ERKO/WT) to produce "KIKO" animals that express the NERKI mutant allele as their only functional ER␣ (7).
The ovariectomized rodent uterus exhibits a robust and rapid response to a single dose of E 2 , culminating in a synchronous wave of epithelial cell mitosis within 18 -24 h (8). The uterine response to E 2 is modulated by stromal factors, such as IGF1, that are induced by E 2 and then impact epithelial responses (9). The Igf1 transcript is increased with a concomitant decrease of Igfbp3 (10) and activation of the Igf1 receptor and downstream effectors following E 2 treatment (11). Igf1 transcript is increased in both stromal and epithelial compartments of the uterus by E 2 , with greater signal apparent in the stroma (12). Igf1 has been demonstrated to play an essential role in the uterine growth response, because Igf1 null mice lack a full uterine proliferative response and, more specifically, lack G 2 /M progression of the epithelial cells following E 2 stimulation (13). Additionally, transgenic mice overexpressing Igfbp1, which sequesters and therefore decreases, the amount of available IGF1, have an attenuated uterine response to E 2 (14). Uterine response is restored by transplanting Igf1KO uterine tissue into a WT host (15). Further, E 2 treatment results in the activation of downstream mediators of Igf1 signaling, including the Igf1 receptor, IRS1 (11), AKT, and GSK3␤ (12). Additionally, inhibitors of AKT and GSK3␤ inhibit E 2 -stimulated uterine growth (12).
In vitro studies that employed a reporter gene to characterize the chicken Igf1 promoter indicated that Igf1 was an example of a transcript whose E 2 regulation was mediated by indirect tethering (16), specifically involving association with AP1. However, we see no increase in vivo in Igf1 transcript in the tethered selective ER␣ containing KIKO uterus following E 2 stimulation, indicating that direct ERE binding was involved in E 2 induction of uterine Igf1 transcripts.
The growth hormone signaling activated transcription factor, STAT5, is also a regulator of Igf1 transcript levels in the rodent liver via interaction with growth hormone-responsive element (GHRE) sites in the Igf1 gene (17,18). The rodent uterus contains STAT5 protein as well, and in this study we observed a WT ER␣-dependent increase in Stat5a transcript. Although estrogen regulation of the Igf1 transcript has been extensively described in the rodent uterus, precise regulatory sequences have not been elucidated. The evaluation of potential mechanisms of estrogen regulation of the mouse Igf1 gene is important; therefore in this study we used KIKO and WT uterine models to identify ERE sequences in the Igf1 and Stat5a genes that could be involved in regulation by estrogen.

MATERIALS AND METHODS
Animals-All of the animal studies were done in accordance with National Institutes of Health Guidelines and a NIEHS Animal Care and Use Committee-approved animal studies proposal. The animals used were either an ER␣ null line (␣ERKO) (19), maintained at Taconic Farms (Germantown, NY) or were obtained by crossing NERKI/ ϩ males (from Jameson lab) with ER␣ Ϫ/ϩ from our Taconic colony. These crosses were done at Charles River (Wilmington, MA). DNA was made from tail biopsies using Direct PCR reagent (Viagen Biotech, Inc., Los Angeles, CA) according to manufacturer's protocols. Offspring were screened for the presence of the NERKI allele as previously described (6) except that 2ϫ RedTaq mix (Sigma) was used. The offspring were then screened for the presence of the ER␣ null allele using the following primers (purchased from Sigma): Esr1 Exon2 forward, CTGTGTTCAACTACCCCGAGG; Esr1 Intron2 reverse II, GGCGCGGGTACCTGTAGAA; Neo forward II, GATATCATAATTTAAACAAGCAAAACCAA in RedTaq reagent in a MBS Satellite PCR machine (Thermo Fisher, Milwaukee WI) with the following conditions: 95°C for 2 min, 95°C for 45 s, 54°C for 45 s, 72°C for 45 s (35 cycles) and then 72°C for 7 min. The results showed either a band at ϳ350 bp (WT or NERKI allele) or bands at ϳ450 and 350 bp (ER␣ null and WT or NERKI alleles, respectively). Females that carried one copy each of the ER␣ null and the NERKI alleles (KIKO) and their ER␣ WT littermates were shipped to NIEHS and used in studies. The mice were ovariectomized after reaching at least 10 weeks of age, rested for 10 -14 days, and then used in studies. The animals were then treated by injecting 100 l of saline (vehicle controls) or E 2 (100 l of 2.5 g/ml in saline) intraperitoneally. Long R3 Igf1, a synthetic IGF1 with low affinity for IGF-binding proteins (Cell Sciences, Canton, MA), was given by osmotic pumps (Durect Corporation, Cupertino, CA) as described previously (20) except the IGF1 was dissolved at 0.5 mg/ml in 0.1 N acetic acid. Uterine tissue was collected at indicated times (1, 2, 6, 18, or 24 h after injection or pump insertion) and either snap frozen in liquid nitrogen for RNA, chromatin, or protein isolation or fixed in 10% formalin for immunohistochemical analysis.
Real Time PCR-RNA was isolated from frozen uteri, and cDNA was prepared as previously described (10). 10 ng of cDNA was used in a 25-l reaction containing Power SYBER Master Mix (ABI, Foster City, CA) and 5 pmol each forward and reverse primers. The primers were designed using Primer Express software (ABI) and synthesized by Sigma and are listed in supplemental Table S2. An ABI 7900 instrument and SDS 2.1 software was used to carry out the PCRs. The values were calculated using the method of Pfaffl (21).
Identification of Potential ERE Sequences-The genomic sequence of Igf1 (NM_010512) plus its 2-kilobase pair upstream promoter sequence (81,992 bp in total) was downloaded from the UCSC Genome Browser (build mm9). We then scanned this sequence for putative ERE using the position weight matrix constructed from 48 experimentally identified ERE (15 bp in length) (22). We computed the log likelihood ratio score of each sliding window of the 15-bp segment for both the plus and reverse complementary strands. A segment was declared a putative ERE when the p value of its log likelihood ratio score is less than or equal to 0.0002 (23). Similarly, we scanned the Ϫ5000-bp promoter sequence of Stat5a (NM_011488) for putative ERE using the same position weight matrix. A putative ERE was found at Ϫ287bp upstream from the Stat5a transcription start site.
Chromatin Immunoprecipitation (ChIP)-Pools of frozen uterine tissue (3-4 uteri/pool; 75-100 mg of tissue) were crushed using a metal pulverizer, and powdered frozen tissue was resuspended in 0.5% formaldehyde in phosphate-buffered saline at 1.5 ml/uterus and incubated on a rotating platform at room temperature for 5 min. Cross-linking was stopped by adding 225 l/uterus of 1 M glycine and incubating 5 more min at room temperature. Tissue was collected by centrifugation at 3000 rpm for 5 min. Liquid was decanted, and the tissue was resuspended in 4°C phosphate-buffered saline with added phosphatase inhibitors 1 and 2 (Sigma) and protease inhibitors (Sigma; 20 g/ml each aprotinin and leupeptin, 4 g/ml ␣-phenylmethylsulfonyl fluoride) and allowed to sit on ice for 2 min. The samples were then centrifuged at 3000 rpm for 5 min. The phosphate-buffered saline wash was repeated, and the tissue was resuspended in 0.5 ml/uterus of 50 mM Tris (Lonza, Rockland ME; pH 7.4), 1% Nonidet P-40 (Sigma), 0.25% deoxycholic acid (Sigma), 1 mM EDTA (Ambion/ABI) with added phosphatase and protease inhibitors. Tissue was homogenized with a PT1200C Polytron (Brinkman, Westbury, NY). Chromatin was pelleted by centrifuging for 15 min at 4°C at 14,000 rpm (Eppendorf 5417R, Westbury NY) and resuspended in Lysis buffer (1% SDS, Ambion, 20 mM Tris, pH 8; Lonza) containing phosphatase and protease inhibitors. Resuspended chromatin was incubated on ice for 15 min, split into 250 -300-l aliquots, and sonicated in a 4°C room in an ice water bath with a Bioruptor (Diogenode, Sparta, NJ) on high for two 3-min cycles of 30 s on and 30 s off with 1 min rest on ice between cycles. Sonicated chromatin was centrifuged for 1 min at 4°C to remove insoluble material; supernatant was stored at Ϫ80°C until use.
Chromatin was diluted 1:5 with HIP buffer (50 mM Tris, pH 7.5 (Lonza), 150 mM NaCl (Lonza) 1% Triton X-100 (Sigma)) containing protease and phosphatase inhibitors and was precleared with 60 l/ml 10% protein A-Sepharose CL-4B (Suspended in HIP buffer); GE Healthcare) containing 200 g/ml salmon sperm DNA (Stratagene, La Jolla, CA) for 1 h at 4°C. Precleared supernatant was divided into 1-ml aliquots, and antibody was added to each. For ER␣, sc7207 (Santa Cruz, Santa Cruz, CA) was added at a 1:25 dilution. For a control, rabbit IgG (Santa Cruz) was added at 1:25. For STAT5a, sc1081 (Santa Cruz) was added at 1:25 dilution. For SP3, sc644 (Santa Cruz) was added at 1:25. All were incubated for 4 -18 h at 4°C. 100 l of protein A-Sepharose containing 200 g/ml single-stranded DNA was added and incubated at 4°C for 3 h. The tubes were centrifuged at 3000 rpm for 3 min to collect the pellets, which were then washed three times in 1 ml of HIP containing phosphatase and protease inhibitors for 10 min at 4°C each wash. Cross-linked chromatin was then eluted twice using 100 l of 1% SDS in 0.1 M NaHCO 3 at room temperature for 15 min. An aliquot was taken for Western blot analysis. The cross-link was reversed using a final concentration of 200 mM NaCl overnight at 65°C, and then RNase (200 g/ml final concentration) was added and incubated at 37°C for 30 min, followed by inactivation with EDTA and proteinase K (Bioline Taunton MA) (at final concentrations of 5 mM and 1 g/l, respectively) at 45°C for 1 h. The DNA was purified using the Qiaquick PCR purification kit (Qiagen). Purified DNA was analyzed by real time PCR using Power Syber master mix (ABI, Foster City, CA). Primer sequences for real time PCR of ChIP samples are shown in supplemental Table S3. The values were calculated relative to vehicle-treated samples.
Gel Shift Assays-Gel shift assays were done with the Gelshift ER kit (Active Motif, Carlsbad, CA) using the provided MCF7 extract according to the manufacturer protocol. The consensus ERE was end-labeled using [␥-32 P]ATP (MP Biomedicals, Solon, OH) and T4 polynucleotide kinase (New England Biolabs, Ipswich, MA). Other EREs were tested by synthesizing oligonucleotides (Sigma) substituting the Igf1 EREs (Ϫ6215 bp, Ϫ29 bp, or intron 3-4) or Stat5a ERE sequences (supplemental Table S4), which were annealed by heating to 95°C for 5 min and then slowly cooling to room temperature in buffer containing 10 mM Tris, 1 mM EDTA, and 50 mM NaCl. These EREs or the WT ERE or mutated ERE provided in the kit were added to reaction mixtures to compete binding to 32 P-labeled consensus ERE. The products were separated on DNA retardation gels (Invitrogen) in 0.5ϫ TBE buffer, dried, and visualized with PhosporImager screens (GE Storm 860; GE Biosciences).
Western Blots-Total protein homogenates were made from pulverized frozen uterine tissue by homogenizing with a Polytron in HIP buffer with containing 2.5 mg/ml sodium deoxycholate (Sigma), phosphatase, and protease inhibitors. Homogenates were centrifuged at 14,000 rpm for 10 min, the supernatant was collected, and protein levels were assayed using BCA assay (Pierce). Nuclear protein fractions were isolated from pulverized frozen uterine tissue using the NPER kit (Pierce). 10 g of total uterine protein or 5 g of nuclear protein was separated on a 10% NuPage gel using LDS loading buffer and reducing agent and MOPS SDS running buffer according to the manufacturer's instructions (Invitrogen). The gels were then transferred to nitrocellulose filters using iBlot apparatus and gel transfer stacks (Invitrogen). The filters were stained with Ponceau Red (Sigma) to ensure even loading and transfer of proteins, then blocked with 5% milk (Santa Cruz) in TBST (20 mM Tris, pH 7.4, 180 mM NaCl, 0.1% Tween 20) for 30 min, and then incubated with primary antibody (details in figure legends) for 1 h. The filters were washed in TBST; incubated with anti-rabbit horseradish peroxidase IgG (Cell Signaling Technologies, Danvers, MA) for all but anti-␤-actin, which was incubated with anti-goat IgG horseradish peroxidase (Santa Cruz); diluted 1:5000 in 5% milk for 1 h; and then washed in TBST. Signal was developed using ECL Plus reagent (GE Healthcare) and detected with Hyblot CL x-ray film (Denville Scientific, Metuchen, NJ).
Immunohistochemistry-Formalin-fixed uterine pieces were embedded on end in paraffin, and cross-sections were cut in 4-m slices, mounted on Superfrost charged slides (Fisher), deparaffinized, and hydrated. Ki67 was detected as previously described (24). Phosphoserine 10 histone H3 was detected using a similar method, except blocking buffer contained 1.5% goat serum (Santa Cruz) and 1% bovine serum albumin and primary antibody (catalog number 06-570, Upstate Cell Signaling Solutions, Lake Placid, NY) diluted 1:500 in blocking buffer and incubated on slides for 1 h.

RESULTS
Uterine Growth Is Impaired in the KIKO-A previous report showed that although the KIKO uterus lacked any uterine weight increase in response to E 2 , epithelial proliferation was retained (7). However, we were unable to detect this uterine response in our colony, as demonstrated by the lack of increase in the Ki67 proliferation marker or the perimitotic marker Ser(P) 10 histone H3 24 h after E 2 treatment (Fig. 1). Additionally, transcripts that are associated with cell cycle stages G 1 -S FIGURE 1. E 2 does not increase uterine proliferation markers Ki67 or phosphoserine 10 histone H3 in the KIKO. Uterine cross-sections from ovariectomized mice that were treated for 24 h with E 2 , representative of three to five uteri sampled were incubated with anti-Ki67 or anti-Ser(P) 10 histone H3 (Ph-ser10H3) as markers of cell cycle activity. Only the WT epithelia had easily detectable amounts of either marker. The bar in the Ki67 section is 0.1 M, and the bar in the anti-Ser(P) 10 histone H3 section is 0.2 M. The arrow shows a perimitotic epithelial cell.
(supplemental Fig. S1A) or G 2 -M (supplemental Fig. S1B) were assessed by RT-PCR. All were induced in WT uteri, but responses were lost or dramatically diminished in the KIKO and absent in the ␣ERKO (not shown), suggesting a lack of cell cycle progression.
Igf1 Is Not Increased by E 2 in the KIKO-Igf1 is critical to the uterine growth responses (12,13,17), and the Igf1 transcript is reported to be an example of a gene response mediated by the tethered ER␣ mechanism, where ER␣ indirectly interacts with other transcription factors on promoter sequences rather than directly binding to DNA (16). Surprisingly, no increase in the Igf1 transcript was seen in the KIKO mouse uterus, which has a mutation in the ER␣, rendering it unable to bind directly to ERE sequences, but selectively preserves its ability to interact with other transcription factors (5,6). In WT mice, Igf1 transcript increases maximally at 6 h after E 2 ( Fig. 2A) (25,26). The lack of Igf1 regulation in the KIKO and ER␣ null (␣ERKO) uterus indicates the sole dependence on ER␣ for the response ( Fig. 2A) (26) and suggests the possible presence of ERE sequences in the Igf1 gene.
The mouse Igf1 gene has two promoters that result in production of transcripts with different first exons ( Fig. 3) (27,28).
To determine whether the E 2 -mediated Igf1 increase was selectively regulated by one of these promoters, exon 1-or 2-selective probes were designed. RT-PCR indicated that both Igf1 promoters are E 2 -regulated because both exon 1-and 2-containing transcripts were increased by E 2 (Fig. 2B), although exon 1 is more robustly increased. Neither transcript increased in the KIKO, indicating that both promoters require direct ER␣ DNA binding for estrogen-mediated response.
Potential ERE Sequences That Mediate E 2 Regulation of Uterine Igf1-Sequence analysis of the Igf1 gene identified several putative ERE sequences upstream of and within intronic regions of the mouse Igf1 gene (supplemental Table S1). Using ChIP, we examined ER␣ binding to three of these putative EREs, selected because they varied from the consensus ERE sequence (GGTCAnnnTGACC) by no more than two bases and were located such that they might impact either promoter (Fig. 3).
ChIP Analysis of EREs in Igf1 Promoter-ER␣ binding to a previously described ERE sequence (29) upstream of both promoters 1 and 2 (Ϫ6215 and Ϫ8270 bp, respectively) was demonstrated by ChIP. E 2 increased ER␣ association with this ERE by more than 5-fold after 1 h (Fig. 4A), and the level decreased to about 2-fold of the vehicle-treated level after 2-6 h. ER␣ binding to an ERE sequence adjacent to promoter 1 (Ϫ29 bp) but also upstream of promoter 2 (Ϫ2084 bp) was enhanced 10-fold above vehicle levels after 1 h of E 2 treatment and decreased to 4 -5-fold after 2-6 h (Fig. 4B). ER␣ binding to the ERE sequence in intron 3-4 was enriched 9 -10-fold compared with the vehicle control 1-2 h after E 2 injection and decreased after 6 h (Fig. 4C). Control experiments to demonstrate specificity of the ER␣-chromatin interaction indicated there was no ER␣ interaction in a region 3000 bp distal from the ERE sequences assayed that lacked ERE and AP1 motifs (data not presented).
ER␣ is not detected on these three predicted ERE sequences in chromatin isolated from KIKO uterine tissues (Fig. 4, A-C).
Western blot of the immunoprecipitated chromatin indicates that the ER␣ in the KIKO samples is only 10% of the level found in WT samples (Fig. 4D). No ER␣ binding could be seen even when the amount of KIKO chromatin used in the real time PCR analysis was increased by 4-fold (data not presented). Total uterine RNA was isolated from uteri from ovariectomized WT, KIKO, or ␣ERKO mice that had been treated with vehicle (0 h) or with E 2 for 2, 6, or 24 h. Transcripts were quantified by real time PCR as described under "Materials and Methods" and calculated relative to WT vehicle-treated (WTV). WT, KIKO, and ␣ERKO E 2 treated were tested versus their vehicle control by two-way ANOVA with Bonferroni correction (*, p Ͻ 0.05; ***, p Ͻ 0.001) in GraphPad Prism software (GraphPad Software Inc, La Jolla, CA). B, Igf1 transcripts from both promoter 1 and 2 are increased by E 2 . Total RNA from uteri treated with vehicle or with E 2 for 6 h was quantified by real time RT-PCR using primers that are specific for exons 1 or 2, reflecting use of promoter 1 or 2 (P1 and P2), respectively. Both transcripts are increased by E 2 and are not induced in the KIKO. The data were tested by two-way ANOVA t test with a Bonferroni correction (***, p Ͻ 0.001) compared with vehicle control.
Previous reports have described an AP1 element in the chicken Igf1 promoter that mediates estrogen regulation (16,30). Comparison of sequences in this vicinity of the chicken and mouse promoters indicated that no homologous region is present in mammals (supplemental Fig. S2 and Table S5). Evaluation of the promoter region of the mouse Igf1 gene for AP1-like motifs indicated the presence of an Ap1-like sequence (number 15 in supplemental Table S5) adjacent to promoter 1. ChIP analysis of this sequence indicates E 2 -dependent ER␣ enrichment in this region in both WT (2.5 Ϯ 0.4-fold increase after 1 h) and KIKO (2.3 Ϯ 0.4-fold enrichment after 1 h).
ChIP Analysis of EREs in Stat5a Promoter-Igf1 transcript can be induced in the rat liver via binding of STAT5 to regulatory sequences in intron 2-3 (17,31). Interestingly, Stat5a transcript is increased by E 2 in our mouse samples after 2 h (Fig.  5A), which precedes the peak of Igf1 increase at 6 h ( Fig. 2A). The increase in Stat5a transcript does not occur in the KIKO or ␣ERKO, indicating that ER␣ DNA binding is required. A potential ERE sequence was found 287 nucleotides upstream of the Stat5a gene promoter. ChIP assays showed that ER␣ binding to this potential ERE sequence was enriched about 7-fold within 1 h of E 2 treatment (Fig. 5B) and that the binding to this ERE was not detected in the KIKO.
ERE Sequences Bind ER␣ in Gel Shift Assay-ER␣ binding to Igf1 and Stat5a EREs was tested in a gel shift assay. ER␣ that was complexed with 32 P-labeled WT consensus ERE was coincubated with 100-fold excess unlabeled WT ERE, mutant ERE, or 100 or 200-fold excess unlabeled ERE from Stat5a or Igf1 genes (Fig. 6). The consensus WT ERE sequence effectively competed the complexes, whereas the mutant ERE sequence did not. The Stat5a ERE and the Igf1 ERE from intron 3-4 were equally effective at competing as the WT ERE at 100-fold excess. The Igf1 ERE from Ϫ6215 bp was less effective, requiring 200fold excess to achieve competition similar to the WT ERE at 100-fold excess. The Igf1 ERE from Ϫ29 bp was the weakest, with only partial competition at 200-fold excess. The relative strength of these EREs in the gel shift assay agrees with the differences in sequence between each ERE and the consensus ERE sequences, because the Stat5a and Igf1 intron 3-4 oligonucleotides contain EREs that differ by one base from the consensus GGTCAnnnTGACC and differ from the sequence in the 32 P labeled WT ERE probe by three bases (supplemental Table S3). The Igf1 Ϫ6215-bp ERE also varies from consensus  . ER␣ is recruited to ERE sequences after E 2 treatment. ChIP was carried out on chromatin isolated from WT or KIKO uteri after ovariectomy and injection with vehicle (0 h) or E 2 for 1, 2, or 6 h. A-C, chromatin immunoprecipitated with anti-ER␣ antibody or normal IgG was analyzed by real time PCR using primers that flanked the ERE sequence identified: upstream of both promoters 1 and 2 (Ϫ6215 and 8270 bp, respectively) (A); adjacent to exon 1 (Ϫ29/Ϫ208 4bp) (B); or the ERE sequence in intron 3-4 (C). The levels are expressed relative to the values for WT vehicle anti-ER␣ samples, which are plotted as time 0. The time 0 ER␣ value was compared with each time point ER␣ value by two-way ANOVA t test with Bonferroni correction. For each time point, p Ͻ 0.001 (***). D, ER␣ Western blot of input (In) and chromatin-associated proteins in ChIP sample aliquots after IP with normal IgG control (IgG) or anti-ER␣. Samples were from WT or KIKO treated with saline vehicle (V) or treated for 1 h with E 2 (E1h). ER␣ antibody: sc542, Santa Cruz diluted 1:1000 in TBST with 5% milk. Gel images were quantified using ImageQuant software (GE Healthcare), and the values were calculated as percentages of WT input, shown below the gel. IP, immunoprecipitation; IB, immunoblot. by one base but differs from the WT probe by four bases (supplemental Table S3); this ER-ERE interaction was weaker and required 200-fold excess to compete the complex. The Igf1 Ϫ29 ERE varies from the consensus ERE by two bases and also differs from the WT probe by five bases, and its weaker interaction did not effectively compete with the WT probe at 200-fold excess.
STAT5a Binds to a GHRE1 Site in Igf1 Intron 2-3-In total tissue homogenates, STAT5a protein is present in both WT and KIKO (Fig. 7A), and E 2 treatment for up to 6 h does not change the amount of STAT5a protein. In contrast, when only nuclear proteins were extracted from uteri, the STAT5a protein in the nuclear extracts from WT samples was very low but increased after 6 h of treatment with E 2 , indicating that it is being translocated to the nuclei after E 2 treatment (Fig. 7A). The amount of STAT5a protein found associated with the nuclear fraction in vehicle-treated KIKO was higher than the WT but did not increase after E 2 injection, indicating that E 2 treatment does not increase overall uterine STAT5a protein levels but leads to nuclear accumulation of STAT5a selectively in the WT.
We hypothesized that E 2 increases nuclear STAT5a protein, which may contribute to the observed increase of Igf1 transcription. ChIP was used to determine whether STAT5a is recruited to a previously described STAT5-binding site in intron 2-3 of the Igf1 gene GHRE-1; (31)). STAT5a recruitment to this region was enriched nearly 5-fold within 1 h of E 2 treatment (Fig. 7B) and was retained throughout the 6 h. There is less recruitment of STAT5a to the intron 2-3-binding site in the KIKO (2.5-fold enrichment after 1 h of E 2 ; Fig. 7B). Western blot of proteins in the WT and KIKO immunoprecipitated chromatin indicates STAT5a is present in the KIKO at ϳ25% of the WT level (Fig. 7C), which correlates with the decreased amount of STAT5a binding observed in the ChIP assay. In the KIKO, STAT5a binding to the GHRE1 is not seen in the 2-or 6-h samples, unlike the WT (Fig. 7B).
To ensure in vivo function of the KIKO ER␣ on a uterine gene, recruitment of the ER␣ to a previously described SP1 site in the Cdkn1a (p21) promoter (32) was evaluated. SP3 and ER␣ are both enriched in E 2 -treated chromatin from WT or KIKO mice (supplemental Fig. S3).
Exogenous Igf1 Does Not Rescue Proliferation in KIKO-Because of the lack of IGF1 expression, KIKO mice were treated with exogenous IGF1, alone or in combination with E 2 , to determine whether epithelial proliferation could be stimulated. The combined treatments were either administered simultaneously (E 2 for 24 h and IGF for 24 h), or the IGF1 was given 6 h after E 2 (E 2 for 24 h and IGF1 for 18 h) to mimic the time at which IGF1 increases following E 2 dosing. The treatment was successful in WT mice, as demonstrated by the increase in the Ki67 proliferation marker but was not effective in KIKO uterine samples (Fig. 8), indicating that IGF1 replacement does not suffice to rescue the inactivity of the mutant KIKO ER␣ in uterine proliferative response and  emphasizes the additional requirement for direct ER␣ DNA binding (ERE) activity.
Igf1 receptors were present in the KIKO uterus at levels modestly lower than those of WT samples (supplemental Fig. S4). AKT and GSK3␤ are both downstream mediators of IGF1 signaling. Examination of the phosphorylated forms of these proteins in WT and KIKO uterine extracts indicates that E 2 leads to phosphorylation of both proteins more effectively in the WT samples (supplemental Fig. S5), indicating that this aspect of the signaling is impaired in the KIKO.

DISCUSSION
Although in vitro Igf1 has been reported to be E 2 -regulated via indirect tethering (16), we saw no increase of Igf1 transcript in the KIKO uterus. Similarly, O2OS cells stably expressing the NERKI ER␣ lacked E 2 -dependent Igf1 regulation (33), and a different DNA ER␣ binding mutant also failed to increase uterine Igf1 transcript in response to E 2 (34). Comparison of the previously described AP1 element in the chicken Igf1 promoter (16,30) to mammalian sequences indicated that there was no homologous region in the mammalian Igf1 genes (supplemen-tal Table S5 and Fig. S2). An AP1like motif near promoter 1 of mouse Igf1 had E 2 -dependent ER␣ binding but apparently did not contribute to increased Igf1 transcription, because the KIKO ER␣ was also enriched in this motif, but no increase in transcript occurred. Because the reproductive cycle of avian egg-laying species differ greatly from mammals, mechanisms of Igf1 regulation may have evolved to accommodate the different modes of reproduction, where a more involved temporal regulation, including other factors such as STAT5, has developed.
The mouse Igf1 gene is encoded by six exons, with as many as six to eight reported splice variant transcripts. The five RefSeq transcripts utilize either exon 1 or 2 as their first exon; thus two alternate promoters adjacent to these exons have been described (27). Our RT-PCR analysis indicated that both of these promoters are targets for ER␣-dependent regulation, because exon 1-and 2-containing transcripts were both increased by E 2 (Fig. 2B), as are transcripts containing exons 4 ( Fig. 2A), 5, and 6 (data not presented); therefore, estrogen regulatory elements should be common to both of the promoters. A potential ERE sequence upstream of both promoters was described as a result of genome-wide sequence analysis for high affinity estrogen-responsive elements (29). Our analysis identified 32 more putative ERE sequences; of these nine are half-sites (supplemental Table S1). Among the remaining 23 palindromic EREs, one is upstream of exon 1, two are in intron 2-3, 15 are in the very large 48 kb intron 3-4, and five are in intron 5-6. In this initial study we focused on locations that might facilitate E 2 regulation of both promoters; however, other ERE sequences could be involved in ER␣-dependent mediation of Igf1 transcript. We selected for study potential EREs that varied from the consensus ERE by two or fewer bases. We examined an ERE adjacent to exon 1 and about 2 kb upstream from exon 2 (Ϫ29/Ϫ2084 ERE) because it is located near the promoters, and a second ERE in intron 3-4 (intron 3-4 ERE), because it varied very slightly from the ERE consensus sequence by only one base. Additionally, we studied a previously described putative ERE (29) sequence Ϫ6215 and Ϫ8270 bp upstream of promoters 1 and 2, respectively (Ϫ6215/ Ϫ8270 ERE). ChIP analysis demonstrated E 2 -dependent recruitment of ER␣ to all three of these EREs in uterine tissue. Less ER␣ protein is present in KIKO uteri (Fig. 4D) (35); however, no ER␣ binding to these EREs was detected in the KIKO samples when the amount of KIKO IP chromatin was increased 4-fold (data not presented). The complexes detected in the WT samples are consistent with a mechanism of Igf1 transcript regulation by E 2 that utilizes direct ER␣-ERE interaction. Additionally, all of the tested ERE sequences competed away the interaction between ER␣ and a consensus ERE in a gel shift assay with effectiveness that mirrored their degree of variation from the consensus sequence (Fig. 6).
Binding to the ERE sequences was temporally correlated to the E 2 -stimulated increase in the Igf1 transcript, with binding occurring within 1 h after E 2 injection, which precedes the maximal increase of Igf1 transcript after 6 h. Together, our observations support a mechanism by which ER␣ interacts in an E 2 -dependent manner with the ERE sequences located within the vicinity of the mouse Igf1 gene promoter to increase mouse uterine Igf1 transcription. Our future studies will develop experimental reporter gene systems in vitro to examine the remaining potential ERE sequences and will evaluate whether these interactions are regulating Igf1 transcription.
Although interactions between estrogen and Stat5 signaling have been investigated, the Stat5a gene as a target for ER has only recently been suggested in a study involving mammary tissue (36), where levels of STAT5a protein were low following ovariectomy but increased after E 2 and progesterone treatments. We noted that uterine Stat5a transcript was increased by E 2 after 2 h in the WT and that no increase was seen in either KIKO or ER␣ null uteri. We also identified an ERE sequence in the vicinity of the mouse Stat5a promoter that recruits ER␣ binding after E 2 stimulation; thus we propose Stat5a is a target for ER␣-mediated regulation and involves direct DNA binding.
Two highly homologous isoforms of Stat5, Stat5a and Stat5b, are produced by neighboring genes (37). Stat5b is the predominant isoform found in the liver, whereas Stat5a is more abundant in the mammary gland (37,38). In liver and breast cancer cell models, E 2 -dependent effects on STAT5 phosphorylation, interaction with ER␣ or ER␤, nuclear accumulation, and transcriptional activity have been reported (39 -43). Studies in the rat liver (31) show Igf1 regulation via interaction between STAT5b and two STAT5-binding sites (GHRE1 and GHRE2) in the intron between exons 2 and 3. These STAT5-binding sites are conserved in the mouse (31).
In our studies, we detected abundant STAT5a protein in the WT uterus prior to E 2 injection, and no increase in total STAT5a protein was observed after the E 2 treatment. Although the total uterine level of WT STAT5a protein was unchanged, it is interesting that nuclear accumulation increased after 6 h, which coincides with the peak of increasing Igf1 transcript. More STAT5a protein was in the nuclear fraction of the KIKO uterus than the WT prior to E 2 treatment, and the nuclear accumulation of STAT5a was unaffected by E 2 in the KIKO, which correlates with the lack of Igf1 regulation in the KIKO uterine samples. Experimentally, we were able to see E 2 -dependent STAT5a binding to the GHRE1-binding site in intron 2-3 in the WT, preceding the peak of Igf1 increase.
Estrogen-dependent interaction between ER␣ and STAT5a proteins has been described in human kidney (HEK293) cells transfected with ER␣, STAT5a, and prolactin receptor (44). Similarly, interaction between GST-ER␣ and STAT5b has been reported in HC11 mammary epithelial cells and shown to involve the DNA-binding domain of the ER␣ (40). These and other studies have lead to a proposed mechanism of integration of growth factor and estrogen signals via interaction between ER and STAT proteins on promoters (39). A similar DNAbinding mutant form of the ER␣ (point mutations in the first zinc finger at amino acids C201A/C204A) is unable to interact with STAT5 (40). Thus, this tethered mechanism of STAT5a interaction with ER␣ signaling could be disrupted in the KIKO (which also has point mutations in the first zinc finger but at amino acids E207A/G208A). Because STAT5a nuclear accumulation was inhibited in the KIKO, it is difficult to address whether STAT5/ER␣ interaction is altered in KIKO tissues. Our studies did show that STAT5a protein was present in levels comparable with those of WT (Fig. 7A); however, KIKO ER␣ did not interact with the Igf1 EREs (Fig. 4), and there was little STAT5a binding to GHRE1 detected (Fig. 7B), suggesting a potential effect of ER␣-ERE binding on the STAT5a recruitment to GHREs. Future analysis in model systems will facilitate more thorough evaluation of relationships or interactions between STAT5a and ER␣ recruitment and uterine Igf1 regulation.
Previously, it was reported that E 2 was not able to increase the uterine weight of the KIKO but was effective in increasing Ovariectomized WT or KIKO mice were treated with E 2, longR3 Igf1, or both. Either they were treated simultaneously, or Igf1 was administered 6 h after the E 2 injection to mimic the peak of Igf1 induction that follows E 2 stimulation. Uterine tissue was collected after 24 h, and cross-sections were evaluated for the cell proliferation marker Ki67. The bar represents 0.1 m; the sections are representative of the treatment group, which included three mice each. epithelial cell proliferation (7) as evidenced by increases in epithelial cell height and DNA synthesis as assessed by bromodeoxyuridine incorporation. Since that report, we established a separate colony for further analysis using NERKI/WT males on a mixed background (129/SvJ and C57Bl/6J) (6, 7) from the Jameson lab and ER␣KO/WT females from our ␣ERKO colony, which are on a pure BL6 background. Female mice expressing only the NERKI ER␣ allele no longer show any indication of epithelial proliferation as evidenced by the proliferative marker Ki67 or the perimitotic marker Ser(P) 10 histone H3 (Figs. 1 and 8) or epithelial cell height increase (data not presented). The reason for the difference between our current observations and those reported previously are unclear at the present time. They may be explained by the drift of the mixed genetic background of the current breeders. Whatever the reason, it is clear that in our colony, we are unable to detect any E 2 -dependent uterine growth.
Studies described elsewhere have used microarray to demonstrate that E 2 regulates numerous ER␣-dependent uterine transcripts in the KIKO, indicating that the NERKI ER␣ allele is functional (35). Cdkn1a (p21) is one example of an ER␣-dependent transcript response that is preserved in the KIKO (supplemental Fig. S1A) (7). This gene has been demonstrated to recruit ER␣ to SP1 motifs in its promoter (32). We observe E 2dependent recruitment of both ER␣ and SP3 to one of the SP1 sites in the p21 promoter in the KIKO (supplemental Fig. S3).
Because Igf1 plays such an essential role in uterine response to E 2 , the lack of increase in Igf1 in the KIKO might have explained the lack of growth in this model. However, the supply of systemic exogenous IGF1 either alone or together with E 2 did not restore uterine response (Fig. 8). Our previous studies have shown that the ER␣ is required for the uterine proliferative response to EGF or IGF1, because ␣ERKO mice administered either growth factor lacked uterine epithelial proliferation (45,46). We conclude that Igf1 signaling together with other ER␣mediated ERE-dependent gene regulations are necessary for a full uterine response to E 2 .
Together, our observations suggest that, in contrast to previous reports indicating a tethered interaction between AP1 and ER␣ on the Igf1 promoter (16), E 2 -mediated regulation of Igf1 in the mouse uterus requires the DNA binding ability of ER␣, and we propose this mechanism of mouse uterine Igf1 regulation: E 2 induces ER␣ and Stat5 recruitment to EREs and GHREs in the Igf1 gene as early as 1 h after E 2 treatment, leading to increased transcription. Igf1 transcript does not increase in the KIKO uterus because of a lack of nuclear accumulation of STAT5a and impaired recruitment of ER␣ to the Igf1 gene. In addition, the KIKO uterus lacks other responses to E 2 that would be necessary for uterine proliferation, because no increase in growth can be induced by E 2 even if supplemented with exogenous IGF1, indicating that both Igf1 signaling and ER␣-ERE-mediated responses are necessary for a full uterine response to estrogen and growth.