A Single-Nucleotide Polymorphism in a Half-Binding Site Creates p53 and Estrogen Receptor Control of Vascular Endothelial Growth Factor Receptor 1

Abstract

Interactions between master regulatory pathways provide higher-order controls for cellular regulation. Recently, we reported a C→T single-nucleotide polymorphism (SNP) in the vascular endothelial growth factor receptor 1 (VEGFR-1/Flt1) promoter that merges human VEGF and p53 pathways. This finding suggested a new layer in environmental controls of a pathway relevant to several diseases. The Flt1-T SNP created what appeared to be a half-site p53 target response element (RE). The absence of information about p53 gene responsiveness mediated by half-site REs led us to address how it influences Flt1 expression. We now identify a second regulatory sequence comprising a partial RE for estrogen receptors (ERs) upstream of the p53 binding site. Surprisingly, this provides for synergistic stimulation of transcription specifically at the Flt1-T allele through the combined action of ligand-bound ER and stress-induced p53. In addition to demonstrating direct control of Flt1 expression by ER and p53 proteins acting as sequence-specific transcription factors at half-site REs, we establish a new interaction between three master regulatory pathways, p53, ER, and VEGF. The mechanism of joint regulation through half-sites is likely relevant to transcriptional control of other targets and expands the number of genes that may be directly controlled in master regulatory networks.

Sequence-specific transcription factors (TFs) can modulate the expression patterns of many target genes both temporally and quantitatively through dynamic interaction with unique, cis-regulatory response element (RE) sequences. These REs, which are typically 5 to 20 bases long and located in proximity to the transcription start sites of genes can differ considerably, as exemplified by consensus sequence degeneracy (6, 40).

In response to a variety of stress signals, such as DNA damage, activated tumor suppressor p53 acts as a master transcriptional regulator to directly control several biological outcomes, including growth arrest, apoptosis, DNA repair, senescence, and angiogenesis (52). The tumor suppressor function of p53 is correlated with direct transcriptional activation of promoters containing p53 REs (43). Based on in vitro DNA binding assays, these consensus REs are comprised of two copies of the 10-bp motif 5′-PuPuPuCWWGPyPyPy-3′ separated by 0 to 13 bp (10). Recent studies using genome-wide in vivo p53 binding in human cells (7, 19, 55), as well as results with yeast model systems (49, 50), point to a more defined consensus p53 RE with severe restrictions on spacer length.

Variation in the individual RE sequences among target genes can result in network plasticity with the possibility of fine-tuning overall gene expression programs to environmental challenges (45). Plasticity, robustness, and specificity within a regulatory network could also be achieved through concerted action of multiple TFs. Evolutionary selection of suboptimal REs would represent one means of requiring dependence of gene transcription on multiple TFs, as the input of an individual TF might not be sufficient, although required, to change transcription rates.

Diversity in an RE sequence between individuals can result in human variation in transcriptional control. This in turn could lead to variation in stress response master regulatory networks and ultimately biological outcomes. Single-nucleotide polymorphisms (SNPs) represent the largest class of genetic variation within the human population (60). Functional SNPs in gene regulatory sequences that affect gene expression levels are an important but relatively unexplored source of genetic variation with potential health significance, particularly in the case of master regulators such as p53, which has a direct role in preventing cancer and other diseases. For example, a regulatory SNP in the promoter of the Mdm2 gene, a negative modulator of p53 activity, correlates with increased risk of sarcoma (4). Using computational as well as experimental approaches, we recently identified many functional polymorphisms in p53 REs (50), including an SNP that confers p53 responsiveness to vascular endothelial growth factor receptor 1 (VEGFR-1) (41).

VEGF is a key mediator of blood vessel formation (angiogenesis and vasculogenesis) in a variety of physiological and pathophysiological biological processes, including embryogenesis, wound healing, tumor growth, and chronic inflammatory diseases (5). VEGF can exert its angiogenic effects via the two tyrosine kinase receptors VEGFR-1 (Flt1) and VEGFR-2 (Flk1/KDR) (13). The exact physiological role and mechanism of Flt1 have not been resolved. Flt1 is the highest-affinity VEGFR, but unlike VEGFR-2 it exhibits weak tyrosine kinase activity, resulting in low signal transduction. These biochemical features, the existence of an alternative splice variant resulting in a soluble form (26), and its crucial role for normal embryonic vasculature development (14, 25) suggested that the receptor may act as a negative modulator of VEGFR-2 as well as the angiogenesis process (15, 33).

The Flt1 gene appears to be expressed in several cancers, where it can have a role in survival, proliferation, and migratory potential (12, 17, 53). In a mouse model system, Flt1-mediated signaling in normal bone marrow-derived cells appears to be required for the formation of tumor-induced premetastatic niches, contributing to stromal and endothelial permeability changes and to chemokine signaling that allows homing of epithelial cancer cells to metastatic sites, even if the tumor cells do not express Flt1 (24). Genetic manipulation or pharmacological inhibition of Flt1 could revert the observed phenotypes, indicating the functional relevance of Flt1 expression on epithelial cancer cells as well as on bone marrow-derived cells recruited at premetastatic niches. Those observations underscore the value of understanding mechanisms of Flt1 gene expression regulation, particularly in response to environmental challenges, which can be related to tumorigenesis.

Recently, we found that activated p53 could stimulate transcription at the Flt1 promoter in human cells, but only if the promoter contained a C→T SNP. The SNP creates a complete consensus p53 half-site RE located 710 nucleotides (nt) upstream of the transcription start site within a sequence that is only loosely related to a p53 full site, GGACA(c/T)GCTCccctgGGACcTGagC. The ccctg spacer and the multiple mismatches (lowercase) would prevent any functional activity as a complete target RE. We, therefore, sought to better define the mechanism(s) by which p53 could control Flt1 gene expression through the apparent half-site RE.

MATERIALS AND METHODS

Plasmids.

To ectopically express human wild-type p53, ERα, and ERβ, we used, respectively, pC53-SN3, pSG5-ERα, and pSG5-ERβ. Luciferase reporter constructs were constructed in pGL3-Basic or pGL3-Promoter vectors (Promega, Madison, WI). pGL3-B-pS2 contains a 1.3-kb proximal promoter fragment derived from the estrogen-responsive TFF1/pS2 gene. pGL3-P-P21 contains a 20-bp insert corresponding to the distal p53 RE present in the human p21waf1 gene. pGL3-B-Flt1 (T or C alleles) contains a 1.4-kb fragment from the Flt1 promoter, as previously described (41). Disruption of the estrogen response element (ERE) within pGL3-B-Flt1-T was done by site-directed mutagenesis, according to the manufacturer's protocol (QuikChange site-directed mutagenesis kit; Stratagene, La Jolla, CA). Smaller Flt1-T promoter fragments were PCR generated using primers containing restriction sites and cloned into pGL3-P, double digested by XhoI and KpnI, and then ligated by T4 DNA ligase (New England BioLabs, Ipswich, MA). The small chimeric construct containing only the ERE and the p53 RE-T sequences was obtained with pGL3-P double digested by XhoI and KpnI using a pair of complementary oligonucleotides generating compatible cloning ends. The identity of the inserts was confirmed by DNA sequencing. Details on primer sequences and plasmid construction are available upon request. In all transient transfection experiments, pRLV40, containing the Renilla luciferase cDNA (Promega), was used to normalize for transfection efficiency.

Human cell lines.

Human U2OS (HTB-96, p53 wild type) and SaOS2 (HTB-85, p53 null) osteosarcoma-derived cells were obtained from the American Type Culture Collection (Manassas, VA). The HCT116 (p53 wild type) colon carcinoma-derived cell line and the isogenic derivative that lacks p53 (HCT116 p53^−/−) were provided by Bert Vogelstein (Johns Hopkins University School of Medicine, Baltimore, MD). Cells were grown in McCoy's A5 medium supplemented with 10% fetal bovine serum and 1× penicillin-streptomycin (Invitrogen, Carlsbad, CA) at 37°C in 5% CO₂. Where indicated, cells were trypsinized (0.05% trypsin, 0.02% EDTA) and transferred to phenol red-free RPMI medium containing 10% stripped fetal bovine serum, 2 mM l-glutamine, and 1× penicillin-streptomycin. Cells were incubated for 1 to 2 days before addition of 17β-estradiol (E2; 10⁻⁹ M) with or without the estrogen antagonist ICI 182,780 (Faslodex) (10⁻⁷ M). 17β-Estradiol was purchased from Sigma (St. Louis, MO). ICI 182,780 was a gift from Ken Korach. Charcoal-stripped fetal bovine serum was purchased from HyClone Laboratories, Inc. (Logan, UT), and cell culture medium was from Invitrogen. To induce p53, cells were treated with 0.3 μg/ml doxorubicin (Doxo; Sigma).

Immunoblot analysis of protein expression was performed as previously described (41). Specific antibodies against p53 (DO-1, 1801), actin (C-11), and ERα (H-184) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); the ERβ antibody (68-4) was from Upstate Biotechnology (Charlottesville, VA).

Luciferase assays.

Luciferase reporter assays were performed as described previously (41). Briefly, cells were plated in 24-well plates and grown to ∼70% confluence. Cells were transfected with the indicated plasmids using Fugene6, following the manufacturer's protocol (Roche, Indianapolis, IN). Where indicated, cell cultures were treated with Doxo (0.3 μg/ml) 18 h prior to the reporter assay. At 48 h posttransfection, extracts were prepared using the Promega dual-luciferase assay system following the manufacturer's protocol and luciferase activity was measured with a Victor Wallac multilabel plate reader (Perkin-Elmer, Boston, MA). For each construct, relative luciferase activity is defined as the mean value of the firefly luciferase/Renilla luciferase ratios obtained from at least three independent experiments.

ChIP analysis.

Chromatin immunoprecipitation (ChIP) assays were performed as described previously (41) or using the ChIP kit (Upstate Biotechnology) following the manufacturer's instructions. A panel of specific monoclonal antibodies from commercial sources was used. PCR analysis was performed using a set of primers specific for different regions of the Flt1 promoter, and the products were analyzed by gel electrophoresis (see supplemental material).

mRNA relative quantification by real-time RT-PCR.

Cells were plated in 100-mm dishes and treated with the indicated doses of Doxo, E2 and ICI 182,780. After 24 h, cells were washed in phosphate-buffered saline and total RNA was isolated from cell pellets using the RNeasy kit (QIAGEN, Valencia, CA). cDNAs were made using TaqMan reverse transcription reagents from Applied Biosystems (Foster City, CA). To determine the Flt1 mRNA levels, real-time reverse transcription-PCR (RT-PCR) was carried out using the ABI Prism 7000 sequence detection system. Probes and primer sets for Flt1 and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) were purchased as predeveloped assays from Applied Biosystems. Relative quantification was obtained using the threshold cycle method after verification of primer performance, following the manufacturer's guidelines.

RESULTS

ER and p53 cooperate in transactivation of the Flt1-T promoter allele.

The identification of SNP-dependent p53 responsiveness of the Flt1 gene had been based on gene expression measurements in cell lines heterozygous for the polymorphism and on functional analyses of the individual allelic variants using 1.4-kb fragments of the proximal promoter cloned upstream of a luciferase reporter (41). Several luciferase vectors containing Flt1 promoter fragments (Fig. 1A) were created to address whether the p53 RE created by the C→T polymorphism was sufficient for p53 responsiveness. Initial experiments were performed in the p53 null, osteosarcoma-derived cell line SaOS2, where the p53 contribution to transcription can be assessed by coincident transfection with a p53 expression plasmid.

FIG. 1.

p53-dependent transactivation from the Flt1-T allele promoter is dramatically increased by the presence of a putative half-site estrogen response element. (A) Diagram of the human Flt1 promoter fragments examined and the corresponding derivative reporter constructs encompassing the p53RE-T allele. Included are the p53RE-C allele promoter region as well as constructs with only the 25-bp RE regions. The transcriptional start site is indicated by an arrow. SaOS2 cells were transfected (see Materials and Methods) with the indicated reporter constructs in the absence (open bars) or presence (solid bars) of pCMV-p53wt. After 48 h, the induction of the luciferase reporter was assessed and is presented as relative increase in luciferase activity compared to that in the pGL3 plasmid lacking the promoter (mock). The isolated p53REs as well as the 0.25-kb constructs lack transactivation activity. SV40, simian virus 40. (B) Schematic diagram of the human Flt1 promoter showing the localization of the putative ERE with a complete half-site ∼225 nt upstream of the p53RE. Mismatches to the consensus ERE are shown in lowercase letters. The construct with the mutated ERE is also presented. (C) ERα protein levels in SaOS2 parental cells and cells transfected with the pCMV-ERα expression vector. ERα protein levels in MCF7 cells served as a positive control to indicate endogenous levels of ER in an untransfected cell line. (D) Synergy between p53 and ERα in Flt1 transactivation. SaOS2 cells were cotransfected with the individual ERα or p53 expression vector or with both, along with constructs containing the Flt1 promoter T or C allele. The constructs with the mutagenized ERE and with the 0.25-kb promoter fragment were also analyzed. At 48 h posttransfection, luciferase induction was evaluated. The strong synergistic transactivation in the construct containing the T allele when ERα and p53 are coexpressed was lost when the ERE was mutated. Presented are the average and standard deviation values from at least three independent experiments.

As shown in Fig. 1A, we confirmed that coexpression of p53 with the 1.4-kb Flt1 promoter construct containing the T, but not the C, allele resulted in an approximately sixfold induction of transcription. Smaller fragments of the Flt1 promoter, all encompassing the p53 RE-T sequence, were then used to determine if additional cis-acting sequences were required to confer p53 responsiveness. Comparable luciferase activities were detected using the 1-kb and 0.75-kb regions, while the 0.5-kb fragment resulted in somewhat reduced activity relative to the 1.4-kb fragment. In contrast, the 0.25-kb fragment was unable to support p53-mediated-transcription even though the originally described Flt1-T p53 RE was present, indicating that additional sequence in the promoter contributes to the p53 responsiveness of the Flt1-T allele.

The 1.4-kb promoter region was screened for binding sequence(s) of transcription factors that might be involved along with RE-T in establishing p53 responsiveness. Using a bioinformatics algorithm (MatInspector) and promoter analysis tools (6), we identified a putative ERE 225 nt upstream of the p53 RE (Fig. 1B) that was absent in the 0.25-kb construct. The consensus ERE sequence is a 13-bp palindromic inverted repeat: 5′-GGTCAnnnTGACC-3′ (9, 29). However, in most ER target genes, the EREs do not match the consensus (1, 30), and even half-sites can mediate activity, usually in association with Sp1 (16, 28, 42). The putative ERE found in the Flt1-T promoter is composed of only one functional half-site (Fig. 1B): 5′-GGTCAgagTcACt-3′. The first 5 bases are a perfect match to an ERE half-site. The TcACt half-site is expected to be nonfunctional because of the two mismatches (lowercase).

We addressed the role that expressed ER might play in p53-dependent transactivation of Flt1. Although SaOS2 cells are ERα positive, the ERα levels are low compared to those in breast cancer MCF-7 cells (Fig. 1C) and there is no detectable expression of ERβ (not shown). As noted above, and previously reported, p53 was able to induce transcription of the 1.4-kb Flt1-T promoter region. However, when ER was expressed along with p53, there was an approximate fourfold further increase in Flt1-T transcription (Fig. 1D). Remarkably, this level of induction is comparable to the activity of the strong, full-site p53 RE from the human p21 gene promoter (Fig. 1A). Overexpression of ERα alone had a minimal effect on transcription in the absence of p53.

To directly test if the ERE binding site is responsible for the ER/p53 synergy in transcription, we mutated the ERE sequence to render it nonfunctional (GGTCA to aaTCA; see Fig. 1B). The large reduction in p53-dependent transactivation establishes that the effects of ERα are mediated through the cis-element located upstream of the p53 element (Fig. 1D). As additional controls, we examined the 0.25-kb Flt1 fragment that does not contain the ERE as well as the C-allele variant of the 1.4-kb fragment. In both cases, ERα expression alone or in conjunction with p53 did not result in stimulation of transcription. The role of the ERE in the Flt1-T transactivation by p53 was further supported using an artificial short chimeric construct containing just the putative ERE and the p53 RE-T separated by only 5 bp (see Fig. S1 in the supplemental material). Expression of ERα and p53 resulted in synergistic stimulation of transcription. A low level of transcription was detected when ERα or p53 was expressed alone.

Overall, these results demonstrate a synergistic interaction between p53 and ER that is mediated through their respective REs. The p53 responsiveness of the 1.4-kb Flt1-T allele that depends on the ERE is likely due to the residual level of ERα present in SaOS2 cells.

ERα and ERβ can mediate p53-dependent, allele-specific transactivation of Flt1 in a ligand-dependent manner.

Estrogen receptor is a ligand-activated enhancer/transcription factor that is a member of the large steroid/nuclear receptor superfamily. The interaction with agonist ligands renders ERs capable of acting as sequence-specific TFs. Ligand-bound receptors can also interact with cofactors that can modify chromatin as well as reposition nucleosomes so as to affect expression of genes containing EREs in their promoters. To examine the potential of ERs to cooperate with naturally induced p53 on the Flt1 promoter and the role of ligands, we extended our experiments to U2OS osteosarcoma-derived cells, which exhibit wild-type p53 expression. These cells are also well suited for hormone studies since they do not express ERα or ERβ (Fig. 2A and see Fig. S2 in the supplemental material). Transient transfection assays with luciferase plasmids containing Flt1 promoter fragments also included either ERα or ERβ expression plasmids (Fig. 2 and see Fig. S2 in the supplemental material). To evaluate whether the interaction with p53 was dependent on an ER ligand, cells were grown in stringent, estrogen-depleted medium that was supplemented with the ER ligand 17β-estradiol (E2). The ER antagonist ICI 182,780 was used to modulate the functionality of expressed ER proteins. p53 was induced by treating cells with the chemotherapeutic DNA-damaging agent doxorubicin (see Materials and Methods).

FIG. 2.

ERα cooperates with p53 to induce transcription from the Flt1-T allele promoter. (A) Immunoblot of ERα protein in ER-negative U2OS cells after transient transfection with pSG5-ERα expression vector. ERα protein detection in MCF7 cells was used as a positive control to indicate levels of ER in cells with normal expression. (B) U2OS cells were cotransfected with pSG5 ERα along with different constructs derived from the Flt1-T promoter as indicated, and cells were grown in estrogen-depleted medium. At 24 h posttransfection, cells were treated with 0.3 μg/ml of Doxo in the presence or absence of 17β-estradiol (E2) and/or the ER antagonist ICI 182,720. Transcriptional activity was measured 24 h later using the luciferase reporter assay. As a positive control for E2 ligand-dependent activation, luciferase expression was determined from the promoter region of pS2, a known ER target gene. The synergistic effect of ERα and p53 for Flt1 transactivation is allele specific and requires a functional ERE. mut, mutant.

As shown in Fig. 2B, ERα expression did not result in transactivation of Flt1 constructs, regardless of the presence of E2. (The estrogen-responsive pS2/TFF1 promoter construct was included as a positive control.) Doxo treatment alone led to fivefold induction of the Flt1-T promoter, but not the Flt1-C promoter, even if the ERE site was disrupted. However, combined addition of estradiol and Doxo generated a nearly 20-fold induction of the Flt1-T promoter that was strongly suppressed by addition of the ER-specific inhibitor ICI 182,780. ERβ exhibited comparable effects (see Fig. S2 in the supplemental material). The contribution of ERβ was also observed using HCT116 cells, which are a colon cancer-derived, p53 wild-type cell line that is also ERβ positive and ERα negative (see Fig. S3 in the supplemental material). The importance of ligand and the ERE was confirmed by examining transactivation from promoter constructs in SaOS2 cells grown under estrogen-depleted conditions (see Fig. S4 in the supplemental material).

To establish that just the p53 RE complete half-site in the natural Flt1-T allele (GACATGCTC ccctg GGACcTGagC) is sufficient for transactivation in combination with the ERE, we created additional mismatches (lowercase) in the last 10 bases (GGACATGCTCccctgGcACAccTgT). The induction still required p53, and it was greatly stimulated by the presence of a functional ERE in the promoter plus expressed ER (see Fig. S5 in the supplemental material).

In summary, the heterologous gene reporter assays demonstrate that either of the ERs can cooperate with p53 to transactivate the Flt1 promoter. The ER/p53 synergistic response requires several factors: p53, ER, and an ER ligand, along with partial p53 RE and ERE sites.

Synergistic cooperation between ERβ and p53 in the induction of the endogenous Flt1 gene.

We reported previously (41) that p53 can mediate the up-regulation of endogenous Flt1 mRNA in cancer cell lines heterozygous for the p53 RE SNP in the promoter region of this gene, following p53 activation by Doxo or other DNA-damaging agents. Those experiments were conducted under normal cell culture conditions where estrogens may be present. Having established the potential involvement of ER plus ligand on p53-mediated expression of the Flt1-T allele, we examined ER/p53 cooperativity towards expression of the endogenous Flt1 gene under estrogen-depleted or reconstituted culture conditions. Induction was determined in HCT116 cells that are heterozygous Flt1 C/T, ERβ⁺, and p53⁺ using quantitative RT-PCR.

As shown in Fig. 3, Doxo alone results in an approximately fivefold increase in Flt1 mRNA, while E2 treatment alone did not change the basal expression level, which is very low in HCT116 cells. However, activation of p53 in the presence of E2 led to a synergistic increase in the Flt1 mRNA levels to 10-fold over the level in untreated cells. This effect was ligand dependent, since pretreatment of cells with ER inhibitor ICI 182,780 abolished the synergy, returning Flt1 expression to that observed with Doxo alone. We also examined whether the E2 treatment might increase levels of endogenous p53 protein and found that, in fact, this was not the case, as shown in Fig. S6. Even when cells were grown in the absence of E2, the inhibitor reduced Flt1 induction by Doxo slightly, indicating the presence of residual ER ligands. Similar experiments were conducted in the isogenic p53 null HCT116 cell line. There was no Flt1 induction in the absence of p53, supporting the requirement for active p53 in the Doxo-dependent stimulation of this gene (data not shown).

FIG. 3.

Synergistic Flt1 transcriptional responses due to ER and p53 in HCT116 cells containing the Flt1-T and -C alleles. Cells of the ERβ-positive HCT116 line (p53^+/+ [open bars]) and of its isogenic derivative (p53^−/− [solid bars]), which are heterozygous (C/T) for the p53 RE SNP in the promoter region of this gene, were grown in estrogen-depleted medium and treated with Doxo to activate the endogenous p53, in the presence or absence of E2 and/or the ER inhibitor ICI 182,780. After 24 h, the Flt1 mRNA levels were measured by real-time PCR (see Materials and Methods). The activation of p53 in the presence of E2 leads to an ∼10-fold increase in Flt1 mRNA. This effect was blocked when cells were treated with ICI 182,780. No induction of Flt1 mRNA was detected in the HCT116 p53^−/− cells.

p53 allele-specific binding at Flt1-T sets up ER binding to ERE.

The transcriptional cooperativity of p53 and ER led us to address their occupancy at the Flt1 promoter. Using ChIP binding assays, p53 was shown to be recruited to the T allele of the Flt1 gene, consistent with the allele-specific transcriptional regulation (41).

The ER might increase transcription directly through binding to the ERE half-site, as suggested by the ERE sequence requirements. Alternatively, the ER might bind directly to p53 to enhance its transactivation capability, as found with other TFs such as Sp1 (54). To determine if ER and the presence of the ERE site affect the formation of p53 protein/Flt1 promoter complexes, ChIP assays were performed on plasmids containing the different Flt1 promoter regions which had been transfected into SaOS2 cells along with the p53-expressing plasmid. As shown in Fig. 4A, there were comparable levels of p53 occupancy (four- to fivefold) for the parental Flt1-T construct and the corresponding fragment with the mutated ERE, whereas the Flt1 promoter constructs with the C allele exhibited little, if any, p53-specific binding. Thus, the ERE or mutated ERE sequence does not affect allele-specific binding of p53 to its cognate RE sequence in the Flt1 promoter. Consistent with these results, similar differences in allele-specific p53 occupancy were obtained for cells grown in estrogen-depleted media.

FIG. 4.

p53 allele-specific binding at Flt1-T precedes and enables ER binding to the ERE. (A) The Flt1 promoter constructs containing the T or C allele in the presence of the functional or disrupted ERE half-site were cotransfected along with p53 into SaOS2 cells. Cells were cultured in normal or estrogen-depleted medium. After 24 h, ChIP analysis was conducted (see Materials and Methods) to evaluate the p53 promoter occupancy on the Flt1 promoter. (B) ChIP analysis of the indicated Flt1 promoters in SaOS2 cells that overexpress p53 and ERα. (C) p53 and ERβ occupancy on the Flt1 promoter in HCT116 cells containing heterozygous C/T alleles. Cells were growth in estrogen-depleted medium, and E2 (1 × 10⁻⁹ M) was added to the cultures prior to transfection. Specific primers spanning p53RE and ERE were used. (D) ChIP analysis for theTRAP220 subunit in SaOS2 cells transfected with plasmids overexpressing ERα and p53 as well as Flt1 promoter constructs that contain the T or the C allele and the normal or the disrupted ER half-site.

The transactivation results with the mutagenized ERE construct strongly support a direct mechanism of ER interaction with the Flt1 promoter. ChIP assays were performed using primers that amplify only the ERE-containing region. As shown in Fig. 4B, there was a threefold enrichment for DNA occupancy with the ERα antibody when cells were transfected with both the p53- and ER-expressing plasmids along with the Flt1-T promoter construct, unlike the situation with the Flt1-C promoter, which also contains the ERE. Thus, p53 binding appears to be required for ER recruitment. Moreover in the absence of p53, no ERα promoter occupancy was detected (data not shown). We also examined occupancy by ERβ at the endogenous Flt1 promoter when HCT116 cells were grown under estrogen-depleted or reconstituted culture conditions, with or without ligand (Fig. 4C). The addition of the E2 ligand did not result in any significant increase in ERβ binding to its RE. While treatment with Doxo led to a small net increase, there was a synergistic increase in ERβ interaction with its endogenous ERE when p53 was induced by Doxo in the presence of ligand. As shown in Fig. 4C, and described previously (41), p53 occupancy was detected in untreated cells and increased after Doxo treatment.

Target gene activation by nuclear hormone receptors, including ERs, can be mediated by a variety of interacting cofactors. The large multisubunit TRAP/Mediator complex is a key coactivator for many TFs, allowing the direct communication with the general transcription machinery (38). The TRAP/Mediator complex can interact with ERα and ERβ through the MED1/TRAP220 subunit to enhance ER-dependent transcription in vitro (23). The TRAP/Mediator complex also interacts with p53 through TRAP80 (22, 59). We examined whether this complex might be involved in the p53/ER transactivation of the Flt1 promoter. SaOS2 cells were cotransfected with Flt1 promoter constructs along with p53 and ERα expression vectors, and ChIP assays were performed with an antibody specific to TRAP220. As seen in Fig. 4D, TRAP220 bound the ERE-containing region of the Flt1-T promoter fragment, whereas there was no binding to the same region in the Flt1-C promoter fragment. The binding was dependent on a functional ERE site (compare “Flt1-T” with “mut ERE/Flt1-T” in Fig. 4) and on the presence of p53. These results do not allow us to establish conclusively whether p53 recruits the TRAP complex, which in turn helps in recruiting ER, or whether ER recruitment precedes and enables TRAP220 association with the Flt1 promoter. Regardless, the results suggest that the synergy between p53 and ERα is mediated at least in part through TRAP220.

DISCUSSION

While human p53 has many functions in DNA metabolic processes, its role as a master transcriptional regulator is central to p53-mediated tumor suppression. Hence, the identification of all genes regulated by p53 is critical to a full understanding of the p53 network and how modulations of the network relate to various possible biological outcomes, including the prevention of cancer. Several approaches have been taken to reveal potential p53 sequence-specific gene targets in the human genome, including extensive transcriptome analyses and genome-wide identification of in vivo p53-DNA binding sites (7, 19, 55). These approaches have been based on p53 consensus sequences of the canonical RRRCWWGYYY(n)_0-13RRRCWWGYYY. However, the p53 master regulatory network is highly complex and such studies do not address individual variation in putative p53 target sequences, substantial departures from canonical consensus, strength of interaction with p53, and other factors that might contribute to transactivation by p53. Our findings expand the universe of functional p53 REs by showing that a half-site can function in cooperation with another master regulator, estrogen receptor, to result in regulatory cross talk between the two pathways to influence a third pathway that relates to VEGF-mediated biological responses. As shown in Fig. 5, the interaction at the Flt1-T allele provides for a synergistic responsiveness of this gene to environmental stresses.

FIG. 5.

Transcriptional cooperation between p53 and ER master regulators caused by an SNP in the Flt1 VEGFR: a model for synergy in transcriptional responses to environmental stresses. Presented are activated master regulatory proteins p53 and ER. The common Flt1-C allele fails to bind either p53 or ER plus ligand, and there is no transcriptional induction by these activated proteins. The half-site p53 RE of the Flt1-T allele can bind p53 and may lead to limited transactivation of the Flt1 gene. However, if activated ER plus ligand is present, it can subsequently bind to its ERE half-site and interact with the p53-bound region, possibly through other factors such as the TRAP/Mediator complex, so as to synergistically increase Flt1 expression.

Cooperativity between p53 and ER.

The density of various TF-DNA binding sequences identifiable in regulatory regions of genes including promoters suggests that several TFs can be brought together in response to a given stimulus, or even diverse stimuli, as we show for Flt1, and cooperate to modulate gene expression. Indeed, gene expression programs elicited by physiological or environmental stimuli may involve synergistic or antagonistic interactions of TFs with the promoter sequences of target genes (18). In the case of p53, sequence-specific DNA binding can be sufficient in some contexts to recruit cofactors and initiate a cascade of events leading to preinitiation complex assembly and activation, as shown for p53 control of the p21 gene (11). However, for other target genes, the p53 regulation of gene expression, either positive or negative, may require the concerted activation of specific cofactors (32) or it may be mediated through other transcription proteins recruited to promoters by p53 (51).

The C→T SNP in the promoter region of the VEGFR Flt1 gene creates a consensus p53 target half-site that enables p53 to up-regulate the gene in response to DNA-damaging agents (41). A combination of bioinformatics and reporter assays using portions of the promoter identified a half-site ERE located 225 nt upstream of the p53 RE. Although the canonical consensus ERE comprises two pentameric half-sites arranged in inverted fashion with a short spacer, there is wide variation in ERE sequences and arrangements among ER target genes, from a single half-site to multiple EREs (16). Furthermore, few of the EREs are perfect matches to the consensus (1). Importantly, an imperfect ERE or ERE half-site in combination with the Sp1 site can confer estrogen responsiveness of target genes (28, 54).

Finding a half-site ERE near the putative half-site p53 RE in the promoter region of the Flt1 gene, along with the observation that the p53 RE alone supported limited p53-dependent transactivation in SaOS2 cells, led us to determine the role that the ERE sequence might play in p53 responsiveness. We established that p53/ER synergy in regulation of the human Flt1 gene is accomplished through sequence-specific binding of both proteins to their respective half-sites. Based on occupancy, ER recruitment requires the half-site ERE and is dependent on p53 recruitment to the p53 RE; either ERα or ERβ could cooperate with p53. Disruption of the ERE site abolished the synergy between ER and p53, and in the absence of activated p53, the ER was unable to interact with the Flt1 promoter region. Furthermore, the ER stimulation was ligand dependent and required bound p53 (at the Flt1-T sequence). These findings differ from other reported mechanisms of ER influence on p53 transactivation, which are based on direct physical interaction or on the modulation of Mdm2 functions (4, 34-36, 56, 57). It is interesting that the distance between ER and p53 half-sites is approximately one nucleosome (∼200 bp), raising the possibility that they are in topographical proximity within chromatin. An alternative unlikely explanation for our results is that the binding of p53 to its RE displaces other TFs or cofactors that act in a negative fashion at the ER half-site.

Further evidence for cooperativity is demonstrated by transactivation through the combined action of these TFs on a construct containing only the half-sites (see Fig. S4 and S5 in the supplemental material). We note that while synergy is clearly seen in the various systems used and with the several constructs, there were differences in the extent of synergy as well as p53 transactivation from its target sequence (see Fig. 2 and 3 and Fig. S2, S3, and S4 in the supplemental material). The greater synergy observed with the Flt1::luciferase constructs transfected along with p53 and ER in SaOS2 (Fig. 1 and see Fig. S4 in the supplemental material) or in the HCT116 cells (see Fig. S3 in the supplemental material) in the presence of exogenously added E2 may simply be due to the availability of more components than in the endogenous system.

In addition to sequence-dependent binding of ER and p53, we established the presence of TRAP220, a component of the TRAP/SMCC/Mediator complex that can interact with DNA/TF complexes and RNA polymerase II, enhancing the recruitment of RNA polymerase II and other general TFs to the promoter region. The interaction of ERs with the mediator complex occurs through the TRAP220 subunit (23), while p53 interacts with the TRAP80 subunit (22). Consistent with these observations, MED1/TRAP/Mediator subunits can be selectively recruited to estrogen as well as p53 target genes (59). Although we did not address directly possible interactions between p53, ER, and TRAP subunits, our promoter occupancy analysis showed that at least the TRAP220 subunit of the complex interacts specifically with the Flt1-T promoter along with ERα, but only when p53 was bound to its RE. There was little, if any, TRAP220 bound to the Flt1-T promoter containing a mutated half-site ERE. We suggest that the synergy between ER and p53 in modulating the expression of the Flt1 gene may involve the MED1/TRAP/Mediator complex serving as a bridge between ER and p53 as well as a source of other cofactors such as CBP, p300, pCAF, and SRC1, which also can interact with both p53 and ER. More direct experiments such as the use of small interfering RNA to silence TRAP subunits or other cofactors would help elucidate this mechanism.

Merging of the p53, estrogen, and VEGF pathways through the Flt1-T SNP.

As part of the VEGF pathway, the Flt1 protein plays a central role in normal development as well as disease (8, 31, 37), and variations in Flt1 levels can greatly impact the VEGF pathway (48). The soluble extracellular sFlt1 resulting from alternative splicing can bind VEGF-A with high affinity (26) so as to sequester it (20, 46) and inhibit VEGF-induced endothelial cell proliferation (2, 15, 27, 47). Furthermore, breast cancer patients that have soluble Flt1 levels at least 10-fold higher than tumor VEGF levels have a markedly favorable prognosis compared with patients whose tumors had low sFlt1/VEGF ratios (39, 53). The nonsoluble Flt1 form has been implicated instead in tumor progression (58).

Overall, it is clear that levels of Flt1 and sFlt1 and the relative balance with VEGF can markedly impact human health. Up-regulation of Flt1 could result in autocrine-paracrine VEGF/Flt1 loops, where increased Flt1/sFlt1 ratios may affect proliferation and migration and have potentially distinct biological consequences in neoplastic cells and in stromal and tumor-infiltrating inflammatory cells, as well as in bone marrow-derived cells involved in neoangiogenesis and metastasis. Since p53 is frequently mutated in human tumors, the status of both p53 and ER could be considered as risk factors in individuals with an Flt1-T allele. We note that another level of p53 control in the VEGF pathway may involve repression of VEGF expression so as to suppress angiogenesis (3). Although p53 does not appear to bind the VEGF promoter and directly control its transcription, p53 inhibits hypoxia-inducible HIF1-dependent VEGF transcription (44), and VEGF expression is enhanced in p53^−/− cells (9, 21, 55).

Implications.

The observed functional interaction by master regulators at half-sites has important implications for understanding complexity within networks, identification of transcriptional controls of target genes, and network cross-talk. The near-palindromic RE consensus sequences for p53 and ER are well described, and there is limited variability allowed at several positions in the recognition sequences. Although half-sites have not been identified as functional individual elements for most mammalian REs, there are examples of cooperation of half-sites with full sites, as described above for ER and SP1. Our results also demonstrate that even half-sites can provide function and enable cooperation between TFs, vastly expanding opportunities for gene regulation. Current bioinformatics tools will have to be expanded to accommodate rules for TF binding to half REs in predicting enhancer sites in human promoters.

Along this line, we investigated the evolution of the partial REs described in this study (data not shown). The ERE in the Flt1 promoter is not conserved in rodents (Rattus norvegicus and Mus musculus). There is a putative p53 RE site, but the large number of mismatches would prevent it from being responsive to rodent p53 (human and rodent p53 proteins are highly conserved). The ERE and p53 RE sites are conserved in primates (Pan troglodytes and Macaca mulatta). However, the half-site p53 RE contains the p53 nonresponsive “C” nucleotide and contains an additional nonconsensus mismatch that would prevent any p53 responsiveness. Thus, the regulation of Flt1 via a combination of p53 and ERE half-sites appears to be unique to humans.

In conclusion, our study has established that the Flt1 gene can be the target of cooperative interaction between stress-activated p53 and ligand-bound ER, mediated by two suboptimal REs, one of which contains an SNP. These results depict a novel level of functional interactions among ERs and p53 and establish that both proteins can act as sequence-specific TFs at half-site REs. Since there is a dramatic increase in transactivation only when both environmentally responsive factors are expressed, this is an important example of where environmentally distinct agents (i.e., ligands and radiation) could create a synergistic response (Fig. 5) through a common, biologically important gene whose changes in expression may affect the risks associated with angiogenesis-driven diseases. Epidemiological evaluations are currently under way to address cancer risks associated with the Flt1 alleles.