Background
In the United States, prostate cancer is the most frequently diagnosed cancer in men with more than 200,000 new cases each year and the second most deadly, killing roughly 30,000 men annually [
1]. Prostate cancer growth is dependent upon an adequate blood supply, which is controlled by Vascular Endothelial Growth Factor (VEGF), a regulator of tumor angiogenesis. Several factors are known to modulate VEGF expression including growth factors, cytokines, and hypoxia. Previous studies have also shown that androgen increases VEGF levels [
2‐
5], but the mechanism(s) involved are unknown.
The
VEGF promoter lacks a TATA box, is GC rich, and is regulated by multiple transcription factors, such as AP-2, HIF-1, Egr1, and WT1 [
6‐
10]. Previously we have reported the identification of functional WT1 binding sites within the proximal
VEGF promoter [
7,
11], and others have reported interaction of WT1 and HIF1-α in the regulation of VEGF [
8]. Additionally, Sp1/Sp3 binding sites located in the core promoter are known to play a role in transcriptional regulation of
VEGF in a variety of cell lines including NIH3T3 cells [
12], ZR-75 breast cancer cells [
13], Y79 retinoblastoma cells [
14], NCI-H322 bronchioloalveolar cells [
15], and PANC-1 pancreatic cells [
16]. Members of the Sp family have a conserved C-terminal DNA binding domain, so they can potentially bind the same sequence of DNA and indeed Sp1, 3, and 4 bind preferentially bind at GC-boxes [
17]. However, binding at different sites within a promoter region may also confer different functional responses for Sp1 and Sp3 [
18]. A cluster of Sp1/3 sites in the proximal promoter mediates regulation of VEGF by TNF-α in human glioma cells [
19]. Sp1/3 sites are also required for IL-1β induction of
VEGF transcription in cardiac myocytes [
20] and for TGF-β1 stimulation of
VEGF transcription in cholangiocellular carcinoma cells [
21]. In Panc-1 pancreatic cells, the regulation of VEGF by Sp1 has been extensively documented [
16,
22] and both constitutive Sp1 activity and a 109 bp core promoter region containing Sp1 sites are essential for VEGF expression [
16]. Overall, the transcriptional regulation of
VEGF is cell specific involving different stimuli and factors, but Sp1 plays a prominent role in many cell types.
Since estrogen mediated regulation of VEGF expression in ZR-75 breast cancer cells was shown to require Sp1 sites in the core
VEGF promoter [
13], we asked whether androgen might behave similarly in prostate cancer cells. Previous studies have demonstrated that
VEGF mRNA levels are elevated by androgen treatment of both human fetal prostatic fibroblasts and LNCaP prostate cancer cells [
2,
4,
5]. Also, VEGF protein levels are increased after treatment with hormone [
3] and flutamide, an anti-androgen, has been shown to block this up-regulation [
23]. However, the hormone responsive region of the
VEGF promoter was never identified in these earlier studies, nor was the mechanism of androgen induction of
VEGF promoter activity and
VEGF mRNA expression determined.
This report characterizing the hormone responsive regions and binding sites within the
VEGF promoter is a continuation of earlier studies analyzing conserved putative binding sites in promoters of genes expressed in prostate cancer [
11] that identified potentially important non-classical AR binding sites adjacent to other zinc finger transcription factor binding sites in the promoter of
VEGF and other genes [
24]. Here we identified and characterized the hormone responsive regions of the
VEGF promoter, including a required Sp1 binding site within the core promoter.
Discussion
We and others have previously demonstrated that androgens up-regulate VEGF expression [
3,
4,
7,
23], however, mechanisms involved were not elucidated. Therefore, a molecular understanding of how androgens regulate VEGF in prostate cancer cells was sought. In this study, we firmly established
VEGF as a hormone responsive gene and demonstrated that the Sp1.4 binding site located 50 bp downstream of the transcription start site of
VEGF was necessary for androgen activation of the
VEGF core promoter, a region lacking any potential ARE binding sites, yet responsive to hormone treatment. Consistent with this finding was the observation that Mithramycin A, which inhibits Sp1 binding to GC rich promoter regions, significantly decreased
VEGF mRNA levels. Sp1 mediated hormone activation of the
VEGF core promoter likely involves both DNA binding and AR protein interaction, as demonstrated by ChIP and co-immunoprecipitation assays. These results support a tethering model for hormone activation of the
VEGF core promoter, i.e., that ligand-bound AR is recruited and then held in place by chromatin-bound Sp1 at the core promoter.
In delineating the hormone responsive regions of the
VEGF promoter, we initially focused on three potential ARE sites and demonstrated DNA binding by ChIP and transcriptional activation by R1881. However, mutations of the three functional ARE half-sites merely attenuated, but did not eliminate activation of
VEGF by androgen. Androgens are known to regulate a multitude of genes, with the most well studied androgen regulated gene being
PSA. There are three known dimeric ARE binding sites in the regulatory region of
PSA[
26] and, similar to the
VEGF promoter, all three sites are involved in androgen regulation. Although mutation of the two AREs in the proximal promoter of
PSA significantly decreased activation by R1881, mutation of the ARE in the distal enhancer located 4kb upstream of the transcription start site completely blocked androgen activation of
PSA. In contrast to
PSA, the ARE sites in the
VEGF promoter are monomeric sites, not canonical dimeric ARE sequence, and are located within 2 kb of the start site. One possibility is that binding at these non-classical sites in the
VEGF promoter may not be as strong as that in the
PSA promoter, but interaction with other TFs might enhance or stabilize binding of AR, increasing
VEGF expression.
Since the
VEGF promoter region is highly GC-rich we investigated the role of other zinc finger transcription factors known to bind GC-rich promoter regions, such as Sp1. Sp1/Sp3 binding sites in the core promoter region are known to control
VEGF transcriptional regulation in a number of different cell lines. Sp1 mediates regulation of VEGF in the presence of specific stimuli, such as stress [
27], estrogen [
13], retinoic acid [
14], TGF-β1 [
21], and PDGF [
12] depending on the cell type. Androgens have been known to act in concert with other zinc finger transcription factors such as GATA [
28] and Sp1 [
25,
29] to regulate androgen responsive genes such as
PSA,
p21, and
NRIP. Previously, Sp1 sites have been shown to be involved in androgen induction of both the
p21 gene and the
NRIP (
nuclear receptor interaction protein) gene and co-IP demonstrated that AR interacts with Sp1 to regulate their expression [
25,
29]. Similarly we have demonstrated that Sp1 plays a role in the androgen responsiveness of
VEGF by forming a complex with AR and binding to the
VEGF promoter. While there are four Sp1 binding sites in the core promoter region, mutation of a single binding site, Sp1.4, eliminated androgen induction of this region of
VEGF. The
VEGF promoter is similar to the
NRIP promoter, in that both are TATA-less GC rich promoters that are induced by androgen in prostate cancer cells [
29]. The
NRIP promoter also contains three Sp1 sites and two hormone responsive elements (ARE and GRE). Similar to our findings in the
VEGF promoter, mutation of these ARE/GRE sites did not eliminate hormone response; and Sp1 and AR were shown to cooperatively interact by several methods, including sequential chromatin immunoprecipitation and co-IP. In both these promoters, the association of AR with Sp1 appeared to cooperatively regulate promoter activity.
The mechanism of androgen mediated regulation of
VEGF identified by this study is analogous to estrogen mediated regulation of
VEGF in breast cancer cells [
13]. We show here that androgen up-regulates the
VEGF core promoter, a region lacking ARE binding sites, but containing four binding sites in which Sp1 or Sp3 can bind. In ZR-75 breast cancer cells, estrogen regulation of
VEGF expression is thought to act through ER- α/Sp1 and ER- α/Sp3 interactions with GC-rich motifs [
13]. These authors showed that treatment with estradiol increased
VEGF mRNA levels greater than fourfold. Additionally, the GC-rich region of the
VEGF core promoter (−66 to −47) was required for E2 activation of
VEGF, despite a lack of classical ER binding sites. Both Sp1 and Sp3 were demonstrated to bind the
VEGF promoter
in vitro by EMSA and
in vivo by ChIP, further supporting their functional relevance in E2-mediated regulation of
VEGF. While these non-classical mechanisms of hormone mediated
VEGF regulation operate under normoxic conditions, under hypoxic conditions HIF-1α is known to regulate
VEGF expression. Androgen regulation of
VEGF by HIF-1α is thought to occur indirectly through an autocrine loop involving EGF/phosphatidylinositol 3
′-kinase/protein kinase B, which activates HIF-1α and HIF-1α regulated expression of
VEGF under hypoxic conditions [
23].
Conclusions
Androgen mediated regulation of
VEGF expression required a specific Sp1/3 binding site in the GC-rich
VEGF core promoter. Although ARE sites within the
VEGF promoter bound AR and their mutation dampened
VEGF expression, mutation of a key Sp1 binding site in the core promoter of
VEGF blocked promoter activation by hormone. Our findings with androgen reflect those of others examining regulation of
VEGF by other hormones [
13‐
15]; overall these studies demonstrate the complexity of hormone activation of
VEGF and the importance of protein-protein interactions. Regulation of
VEGF by zinc finger transcription factors, such as Sp1, and the importance of their interactions with AR, suggests that they may play a positive role in promoting angiogenesis and prostate cancer progression. Thus, elevated expression of these zinc finger transcription factors may indicate a worse prognosis. Therapy disrupting AR-Sp1 complexes and thereby suppressing VEGF would be expected to limit angiogenesis and maintain the indolent form of prostate cancer.
Methods
Cell culture and hormone treatment
LNCaP (ATCC CRL-1740) and CWR22Rv1 (ATCC CRL-2505) prostate cancer cells were cultured in RPMI media. All cells were grown in media supplemented with 10% FCS and 100ug/ml penicillin/streptomycin in a 37°C incubator with 5% CO2. For hormone treatment, cells were grown to 60-80% confluency and then serum starved overnight in either serum-free media or media supplemented with 5% charcoal-dextran stripped FBS RPMI. The synthetic androgen methyltrienolone (R1881) (Perkin Elmer, Boston, MA) was then added to the charcoal-dextran stripped FBS RPMI media and cells were treated with 5nM R1881 for 24 hours unless otherwise noted in figure legends. For inhibition of AR, 10μM bicalutamide/casodex (LKT Labs, St. Paul, MN) was added 2 hours prior to treatment with R1881.
Chromatin immunoprecipitation
Two million cells were treated with formaldehyde to crosslink proteins to DNA and lysed as per manufacturer’s recommendations using Millipore EZ ChIP Assay (Upstate Biotechnology Inc., Billerica, MA). Chromatin was sheared by sonication (Biosonik III, Bronwill Scientific, Rochester, NY) to fragments of 200–1,000 bp in length. The supernatant was pre-cleared by incubation with Protein G Agarose and incubated overnight at 4°C with either anti- AR (Santa Cruz), Sp1 (Santa Cruz) and control polymerase II antibodies or non-immune IgG (Upstate Biotechnology Inc.). The complexes were recovered from Protein G magnetic beads, crosslinks were reversed and DNA was purified. Four percent of both immunoprecipitated and input chromatin were amplified by PCR using Taq polymerase (Applied Biosystems by Roche Molecular System, Inc) and the appropriate primers (ARE I (FOR): 5′-TTCGAGAGTGAGGACGTGTG-3′, ARE I (REV): 5′-AGGGAGCA GGAAA GTGAGGT-3′, ARE II (FOR): 5′-TCACTGACTAACCCCGGAAC-3′, ARE II (REV): 5′-TTTGG GACTGGAGTTGCTTC-3′, ARE III (FOR): 5′-GGCTCTTTTAGGGGCTGAAG-3′, ARE III (REV): 5′-AGGCTGATGAACGGGATATG-3′, VEGF V88 (FOR) 5′-CCGCGGGCGCGTGTC TCTGG-3′, VEGF V88 (REV) 5′-TGCCCCAAGCCTCCGCGATCCTC-3′). Following an initial 10 min denaturation at 95°C, DNA was amplified by 32–35 cycles of: 1) 20 sec denaturation at 95°C, 2) 30 sec annealing at either 52°C (for ARE III primers), 53°C (for ARE II primers), 58°C (for ARE I and VEGF V88 primers) and 3) 30 sec extension at 72°C; amplification was completed with a 2 min final extension at 72°C. PCR products were electrophoresed on 1% agarose gel, and ethidium bromide stained DNA was visualized by a gel doc system (Biorad, Hercules, CA).
For quantitation of immunoprecipitated chromatin by quantitative real-time PCR (qRT-PCR), purified DNA samples were amplified in an ABI 7000 thermocycler using primers listed above and following manufacturer’s recommendation for SYBR Green Q-PCR (Applied Biosystems, Foster City, CA). Each PCR reaction was carried out in triplicate and average Ct values were normalized to total input (non- immunoprecipitated) DNA. The amount of DNA immunoprecipitated with the target antibody from hormone treated cells R1881 was compared to that of control samples treated only with vehicle. Shown are input normalized Ct values from chromatin of treated cells relative to untreated control cells.
RNA isolation and quantitative real-time PCR
RNA was isolated from subconfluent cells using the GenElute Mammalian Total RNA Miniprep Kit (Sigma, St. Louis, MO) as per manufacturer’s recommendation. Following quantitation, 1μg of RNA was reverse transcribed using the High Capacity cDNA ReverseTranscription Kit (Applied Biosystems, Carlsbad, CA). qRT-PCR was performed using either Taqman Universal Master Mix with pre-designed Taqman Gene Expression Assay probe sets for
VEGFA (Hs00900057_m1) and
18S (Hs99999901_s1) or SYBR Green Master Mix with primers specific for
VEGF,
Sp1,
GAPDH, and
Beta-actin:
VEGF (FOR): 5
′-CGAAACCATGAACTTTCTGC-3
′,
VEGF (REV): 5
′-CCTCAGTGGGCACACACTCC-3
′,
Sp1 ( FOR) 5
′-TGCATTTCAAGGAATGGAAT-3
′,
Sp1 (REV) 5
′-GCTTCCTTGGTGTGAAGAGA-3
′,
GAPDH (FOR): 5
′-CCATCACCATCTTCCAGGAG-3
′,
GAPDH (REV): 5
′-GGATGATGTTCTGGAGAGCC-3
′,
Beta-actin (FOR): 5
′-GTGGGGCGCCCCA GGCACCA-3
′,
Beta-actin (REV): 5
′-GTCCTTAATGTCACGCACGATTTC-3
′). The comparative Ct method [
30] was used to analyze gene expression differences between control (untreated) cells and cells treated with R1881 alone or with the anti-androgen casodex.
Western blot
Subconfluent monolayers of LNCaP and 22Rv1 cells were washed in PBS and proteins were extracted using RIPA lysis buffer (1% NP40, 0.5% Na Deoxycholate, 0.1% SDS, and 150 mM NaCl) containing protease inhibitors. To quantify the amount of proteins present in each lysate, bicinchoninic acid (BCA) assays (Pierce, Thermo Fisher Scientific, Rockford, IL) were performed and absorbance was measured at 600nm on a Dynex Technologies (Chantilly, VA) MRX Revelation plate reader. Proteins (25-50ug) were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). After transfer to a Polyvinylidene Fluoride (PVDF) membrane and blocking with 5% casein, blots were probed overnight at 4°C with polyclonal VEGF and AR antibodies (Santa Cruz), and monoclonal β-actin antibody (GenScript). Washed blots were then incubated for 1 hr in either HRP-conjugated anti – rabbit (GenScript) or anti – mouse (Santa Cruz) antibodies. Proteins were visualized by incubating the membrane in a luminol ECL solution followed by chemiluminescent detection using a Fuji LAS 3000 (GE, Piscataway, NJ) detection system. Bands were quantified using ImageJ analysis and normalized to actin levels.
Co-Immunoprecipitation
Nuclear extracts from 22Rv1 cells were prepared using Active Motif’s Universal Magnetic Co-IP kit (Carlsbad, CA) as per manufacturer’s recommendations. Cells were swelled in Hypotonic Buffer containing phosphatase-, deacetylase-, and protease- inhibitors, then lysed in 5% detergent. This suspension was then centrifuged at 14,000 x g and the supernatant was discarded leaving the nuclear fraction which was enzymatically sheared in the presence of the same inhibitors. Nuclear extracts (150-200μg) were then combined with 5μg of either AR (Santa Cruz) or Sp1 (Upstate) antibodies, or negative control IgG in the presence of the same inhibitors. Following antibody incubation, complexes were pulled down with Protein G magnetic beads. After washing, these complexes were separated by SDS-PAGE and identified by Western blot analysis (as described above).
Plasmid Transfection and Luciferase Assay
The pGL3-VEGF luciferase reporter constructs (V88, V411, or V2274) were generously provided by Dr. Xie [
16] and DNA purified by the Qiagen plasmid Maxi prep kit (Qiagen, Valencia, CA). LNCaP and 22Rv1 cells were plated in 12-well plates as described above. After reaching ~70-90% confluency, cells were serum-starved for 18–24 hours with serum-free RPMI and then transfected with VEGF reporter constructs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) as described [
7]. After 4–6 hours, transfection media was replaced with appropriate growth media. For hormone induction, 5nM R1881 was added to media with 5-10% charcoal-dextran stripped FBS as described above. After 48 hours cells were lysed and luciferase activity was measured using a Promega luciferase assay kit (Promega, Sunnyvale, CA) and a Turner luminometer (Promega, Sunnyvale, CA) following manufacturer’s recommendations. The luciferase activity was normalized to total cell protein, using a micro BCA protein assay, as described above. All experiments were done in triplicate and repeated at least three times. Standard errors of the mean were determined using GraphPad InStat software (San Diego, CA). Significance was determined by Student’s t-test.
Site directed mutagenesis
Potential binding sites in the
VEGF promoter were identified using MatInspector, as previously described [
11]. Predicted AR and Sp1 binding sites in the
VEGF promoter construct were then mutated using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, Agilent Technologies, Santa Clara, CA). Primers were designed according to the manufacturer’s suggestions using the QuikChange Primer Design Program. Primers containing the desired mutation (shown in bold) are listed below:
(ARE I (FOR):
5′-CTCTATCGATAGGTACCGTGGTCAG CTCTCCCC ACCCGTC CCTGTC-3′,
ARE I (REV): 5′GACAGGGACGGGTGGGGAGAGCTGAC CACGGTACCTATCGA TAGAG-3′,
ARE II (FOR): 5′-GGAACCACACAGCTTCCCACTG TCAGCTCCACA AAC TTGG-3′,
ARE II (REV): 5′-CCAAGTTTGTGGAGCTGACAGT GGGAAGCTGTGTGGTTCC-3′,
ARE III (FOR): 5′-GCCCCAAGATGTCTACAGCTTACGG TCCTGGGGTGC-3′,
ARE III (REV): 5′-GCA CCCCAGGACCGT AAGCTGTAGACATCTTGGGGC-3′,
Sp1.2/Sp1.3 (FOR): 5′-GCCCC CCG GTT CGGGCCGGGTT CGGGGTCCC-3′,
Sp1.2/Sp1.3 (REV): 5′-GGGACCCCG AA CC CGG CCC GAA CCGGGGGGC-3′,
Sp1.4 (FOR): 5′-GGGTCCCGGCGGTT CGGAGCCATGCG-3′,
Sp1.4 (REV): 5′-CGCATGGCTCCGAA CCGCCGGGACCC-3′).
PCR was performed using the V88, the V411, or the V2274 luciferase reporter constructs and the appropriate mutant primers. After PCR amplification, parental strands were digested with DpnI and XL1-Blue super competent cells were transformed with remaining mutant DNA. Individual colonies were grown, plasmids were purified (Qiagen, Valencia, CA) and sequenced to verify that the correct base pairs were changed (Cleveland Clinic Genomics Core, Lerner Research Institute, Cleveland, OH). Luciferase assays were performed using mutant constructs as described above.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
KE and GF conceived and designed study; analyzed and interpreted data; drafted and revised the manuscript. KE performed experiments and the statistical analyses. CJB and MM assisted with luciferase and qRT-PCR assays. AB assisted with sequence analyses and co-immunoprecipitations. All authors read and approved the manuscript.