| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Endocrine Service, Sainte-Justine Hospital Research Center, University of Montreal, Montreal, Quebec, Canada H3T 1C5
Address all correspondence and requests for reprints to: Cheri Deal, Ph.D., M.D., Endocrine Service, Department of Pediatrics, Ste-Justine Hospital, 3175, Côte Ste-Catherine, Montreal, Quebec, Canada H3T 1C5. E-mail: Cheri.L.Deal{at}umontreal.ca.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Serum levels of IGFBP-3 vary according to age, sex, and stage of pubertal development. In addition, twin studies have determined that over 50% of the variability in circulating IGFBP-3 is explained by genetic factors (2). Recent research has revealed the presence of five single-nucleotide polymorphisms (SNPs) in the IGFBP3 promoter region (3). Genotyping at two of these loci, the –202 A/C and –185 C/T promoter polymorphisms, has been correlated with IGFBP-3 serum levels in independent cohorts (3, 4). We have reported that this correlation is influenced by factors such as body mass index (BMI), height, and retinol levels. Clinically, measurements of circulating IGFBP-3 have proved useful for the investigation of growth disorders in children and pertinent to epidemiological studies in cancer. Indeed, a high IGF-1 to IGFBP-3 ratio is associated with increased risk for a number of solid cancers, including breast, colon, lung, bladder, and prostate (5).
The IGFBP3 gene is located on chromosome 7p14–12 (6) and consists of five exons encoded within 8.9 kb of sequence. The genomic sequence of IGFBP3 is conserved among the bovine, rat, and porcine species. A single mRNA of 2.5 kb codes for a polypeptide of 264 amino acids. The minimal promoter lies between nucleotides –431 to +72 relative to the transcription start site and contains a TATA box at –30 and a GC box at –97, typical of eukaryotic gene promoters (7).
IGFBP3 expression in a variety of species is regulated by hormones and other agents that stimulate or inhibit growth. Increased levels of IGFBP-3 mRNA and/or protein are observed in vitro and/or in vivo with antiestrogens, retinoic acid, osteogenic protein-1, vitamin D, PTH, insulin, p53, IL-1, TNF
, TGFß, sodium butyrate, and p53, whereas decreased levels of mRNA and/or protein are observed with estrogens, androgens, and glucocorticoids (8, 9, 10, 11, 12, 13, 14, 15). Hormonal regulation of IGFBP3 can also be biphasic, with low concentrations inhibiting and high concentrations stimulating IGFBP3 transcription, as in the case of androgens (16).
Functional studies exploring response elements for IGFBP3 regulators are now emerging. Response elements for the antiproliferative agents trichostatin A and sodium butyrate have been located within the minimal promoter; these histone deacetylase inhibitors increase IGFBP3 transcription, and in many cancer cell lines, specificity protein 1 (Sp1) binding sites have been implicated (14, 17). Several p53 binding sites (decamer repeats) also occur within the first 300 nucleotides upstream from the transcription start site (18). Recently, a retinoic response element was identified at nucleotides –356 to –337 (19). Further upstream, other investigators have found a 1,25-dihydroxyvitamine D3 response element between nucleotides –3296 and –3282 capable of recruiting heterodimers of vitamin D and retinoid X receptor (20) and an androgen response element located at –2879/–2865 (16), both functioning as activators of IGFBP3.
Ironically, although IGF-I and insulin were among the first hormones known to increase IGFBP-3 levels both in vivo and in vitro, little is known about the molecular mechanisms involved. An insulin response element (IRE) has been localized to nucleotides –1150 to –1124 in the rat IGFBP3 promoter (21); the human IRE(s) is/are unknown. Furthermore, the mechanisms by which IGFBP3 promoter haplotypes affect the interaction of known regulators with this gene have yet to be explored, although changing methylation status is an attractive explanation given the known suppressive effect of promoter CpG dinucleotide methylation on IGFBP3 transcriptional activation (22).
The main purpose of this study was therefore to search for putative transcription factors (TF) capable of interacting with the –202 A/C and –185 C/T promoter polymorphisms, particularly methyl-sensitive TFs, as well as to investigate whether haplotype influences regional methylation status of the IGFBP3 promoter. Because it is also known that there is a good correlation between BMI and insulin levels (23), we focused our initial study on the insulin stimulation of promoter constructs in cellular models. We then identified a response element for the methylation-dependent transcription factor, upstream stimulatory factor (USF), in close proximity to the IGFBP3 promoter polymorphisms and show the participation of USF binding and promoter haplotype on the reporter gene transcriptional response to insulin in SK-HEP-1 cells. Finally, we report haplotype-dependent methylation profiles in human leukocyte DNA that could potentially modulate binding of USF and other transcription factors in vivo.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cloning of human IGFBP3 promoter and upstream sequence
A BAC clone containing a portion of human chromosome 7 was obtained from Genome Systems Inc. (St. Louis, MO). A 6.6-kb EcoRI fragment containing the minimal promoter of human IGFBP3 (7) was subcloned in pBluescript II KS (Stratagene, La Jolla, CA).
IGFBP3 promoter constructs for transient transfection
The pGL3-Basic (Promega Corp., Madison, WI) vector containing the luciferase reporter gene and polycloning site was modified to include new EcoRI and NdeI sites to facilitate insertion of IGFBP3 promoter fragments varying in length from 4.4 kb to 81 bp and extending 3' to nucleotide +19 relative to the mRNA cap site. All constructs except the 1.1-kb construct were prepared by PCR with primers including restriction sites for EcoRI 5' and XbaI 3'. Amplifications were done in the Tgradient Thermocycler (Whatman Biometra, Goettingen, Germany) using the high-fidelity Pfu DNA polymerase with manufacturer’s recommended conditions (Stratagene). PCR annealing temperature was adjusted according to the forward primers and a unique reverse primer. The longest fragment (4.4 kb) was amplified from the genomic EcoRI 6.6-kb clone. After digestion of extremities (EcoRI/XbaI), the product was inserted into the modified pGL3 Basic vector and used for PCR generation of IGFBP3 promoter 5'-deletion fragments. A 1.1-kb fragment was obtained by purification of a double-digestion product of the –4.4-kb construct with EcoRI/Eco81I and religation with T4 DNA ligase (see Table 1
for all primers).
|
The integrity of the IGFBP3 cloned sequences was verified by automated sequencing.
Site-directed mutagenesis
The haplotype AT was created using the protocol QuikChange Site-Directed Mutagenesis (Stratagene) and pBP3/–442A–202C–185 as a template in the mutagenesis reaction. An E-box mutant was also obtained using the same protocol. Primers for these reactions are given in Table 1
.
Transient transfections
Cells were plated in a 12-well cell culture dish the day before transfection at a density of 3 x 105. Cell density was 50–80% when transfected using Fugene 6 transfection reagent (Roche Applied Science, Laval, Quebec, Canada) in ratios of Fugene 6 to DNA of 3 µl to 2 µg (1.5 µg IGFBP3 luciferase construct plus 0.5 µg pSVß-galactosidase transfection control plasmid) using the manufacturer’s protocol. Cells were incubated for 6 h in serum-free medium before a change for fresh serum-free medium supplemented with 0.1% BSA (A-4378; Sigma-Aldrich Canada Ltd., Oakville, Ontario, Canada) and or without freshly diluted insulin (Humulin R; Eli Lilly Canada, Toronto, Ontario, Canada) at 170 nM insulin for 24 h. For the luciferase assay, cells were lysed in potassium phosphate buffer containing 1% Triton X-100, and light emission was detected with a luminometer after addition of luciferin (Wallac 1420; PerkinElmer Canada Inc., Woodbridge, Ontario, Canada). Values are expressed as arbitrary light units normalized to the ß-galactosidase activity of each sample.
Preparation of nuclear extracts
Nuclear extracts were prepared from SK-HEP-1 cells or HepG2 cells grown in T75 flasks to a density of 60–80%. Monolayers were then incubated in serum-free medium for 24 h before treatment with insulin or with vehicle in MEM supplemented with 0.1% BSA (see legends of Figs. 4 and 5 for insulin concentrations and incubation times). After washing in PBS, cells were lysed by freeze-thaw cycles in buffer A. Nuclei were collected by brief centrifugation and then resuspended for lysis in buffer C. Buffers used are as described in Ref. 24 . Nuclear proteins were collected in the supernatant fraction after centrifugation. Protein concentration was determined using the protein assay reagent from Bio-Rad Laboratories, Inc. (Mississauga, Ontario, Canada).
|
|
. Cohesive ends were labeled using -32dATP (Amersham Biosciences Corp., Baie d’Urfé, Québec, Canada) and Large Fragment DNA Polymerase I (Invitrogen Corp., Carlsbad, CA). Labeled probes were purified by centrifugation with ProbeQuant G-50 Micro Column (Amersham).
Two E-boxes potentially binding USF were studied using EMSA with or without cold competitors (10- to 100-fold molar excess) as shown in Table 1. Binding reactions were done in 20 µl of the buffer previously described (25) with 5 µg nuclear extracts from insulin-treated or control SK-HEP-1 cells for 30 min on ice. The probe (1 x 105 cpm) was then added and incubation continued for 20 min. Reactions were stopped by adding loading buffer and subjected to electrophoresis on 4% polyacrylamide gels (29:1, acrylamide to bis) in 0.5x Tris-borate-EDTA at room temperature. Gels were dried and the complexes revealed by autoradiography.
Competitors and supershift assays
Specificity of transcription factor binding to IGFBP3 wild-type and mutated promoter fragments labeled as described above was studied using competitors (see Table 1) and antibodies as discussed in detail in Results. For supershift experiments, 2 µg rabbit polyclonal antibodies against USF-1 and USF-2 (sc-229X and sc-861X, respectively, from Santa Cruz Biotechnology Inc., Santa Cruz, CA) were added to the reaction mixtures and incubated for 30 min at room temperature before addition of the probe.
The effect of competitor methylation on transcription factor binding was examined by performing in vitro methylation with 160 µM S-adenosylmethionine and M.SssI methylase (New England BioLabs Ltd., Pickering, Ontario, Canada) at 37 C for 5 h. The extent of methylation was assessed by digestion with the methylation-sensitive restriction enzyme HpaII or MspI as digestion control.
Chromatin immunoprecipitation (ChIP)
SK-HEP-1 and HepG2 cells were grown in 100-mm dishes as described above. After 24 h serum starvation, cells were exposed to vehicle or to insulin. The length of insulin treatment needed to confirm binding differed between the two cell lines and is indicated in the legend to Fig. 7. Formaldehyde cross-linking, cell lysis, preparation, and lysis of nuclei were performed according to the ChIP-IT manufacturer’s instructions (Active Motif, Carlsbad, CA). Lysed nuclei were sonicated six times for 20 sec in a 1.7-ml conical centrifuge tube on ice, using a Sonics Vibra Cell, model VC130PB set at amplitude 50% (Sonics and Materials Inc., Newtown, CT). For the HepG2 cell line, three pulses of 20 sec were used. The average DNA fragment size obtained ranged from 200–900 bp. A small aliquot of cross-linked, precleared chromatin was kept as input DNA. Aliquots were also incubated with rotation for 16 h at 4 C with 4 µg rabbit polyclonal antibodies against USF-1, USF-2, p53, or Sp1 (Santa Cruz Biotechnology). Positive and negative controls contained anti-TFIIB antibody and IgG, respectively. Five microliters of a 10-fold dilution of input control DNA or 5 µl of immunoprecipitated, proteinase K-treated DNA were subjected to PCR in the presence of 2% dimethylsulfoxide with a 56 C annealing temperature to amplify a 264-bp fragment of the IGFBP3 promoter (primers BP3–247, Table 1). The positive and negative controls also included amplification of a GAPDH promoter fragment of 166 bp containing a TFIIB site and amplification of a region of genomic DNA (174 bp) between the GAPDH gene and the chromosome condensation-related structural maintenance of chromosome-associated protein (CNAP1) gene, respectively. Primers and conditions for these controls are described in the ChIP-IT instruction manual.
|
was performed using materials and conditions from the Thermo Sequenase Radiolabeled Terminator Cycle Sequencing Kit (USB Corp., Cleveland, OH). Sequencing products were resolved in a 6% denaturing polyacrylamide gel (19:1, acrylamide to bis) in the presence of 8 M urea and visualized by autoradiography.
Statistics
Results are given as mean ± SD or mean ± SEM where appropriate. Individual experiments were done in triplicate and repeated at least three times unless otherwise specified. Nonpaired t tests with a confidence interval of 95% were performed for group comparisons using Prism 3.0 software.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To localize the upstream sequences required for insulin regulation of human IGFBP3 gene transcription, we developed a series of 5' deletion constructs containing genomic sequences of the IGFBP3 promoter and upstream region. All plasmids contained the luciferase reporter gene and were designed with a common genotype at nucleotides –202 and –185 and a common 3'-end, preserving the putative TATA box (nucleotide –30) and the GC box (nucleotide –97) as described by others (7). We observed an optimal level of both basal and insulin-stimulated luciferase activity with the reporter gene construct containing approximately 250 nucleotides; larger fragments of the 5' region (
600 nucleotides or more) resulted in significantly lower transcriptional activity under both basal and insulin-stimulated conditions (Fig. 1A, representative experiment performed in triplicate). When data from four different experiments were pooled and expressed as fold increases in luciferase activity with insulin stimulation (induction ratios, Fig. 1B), the values were as follows (mean ± SEM): BP3/–247, 1.67 ± 0.20; BP3/–442, 1.31 ± 0.11; BP3/–1073, 1.33 ± 0.12; and BP3/–1956, 1.27 ± 0.05. Interestingly, the shortest construct (BP3/–80) that did not include the Sp1-binding GC box described by Cubbage et al. (7) (nucleotides –97 to –91) but that retained the TATA box and at least two other Sp1 binding sites was inducible after 24 h insulin treatment (fold induction, 1.62 ± 0.01; n = 2 experiments). Although not statistically significant, there was a trend for lower fold stimulation in the larger constructs (P = 0.158). Because our two polymorphisms of interest lay within the minimal promoter, it was of interest to explore their impact on insulin regulation of gene transcription.
|
). The AT haplotype, which has not been found in several cohorts studied previously (3, 4), was created by mutagenesis of the AC construct. One-way ANOVA suggested a significant genotype effect (P 
0.001); we consistently observed a significantly higher luciferase activity with pBP3/–442 AC compared with the other haplotypes whether under basal conditions (Fig. 2A: AC > CC, P < 0.001; AC > CT, P < 0.001; AC > AT, P < 0.001) or insulin-stimulated conditions (Fig. 2B: AC > CC, P = 0.002; AC > CT, P < 0.001; AC > AT, P = 0.011). Under basal conditions, no genotype differences were noted between CC, CT, and AT constructs, although when transfected cells were insulin treated, the CT genotype consistently gave the lowest relative luciferase activity (CT < CC, P = 0.016; CT < AT, P = 0.028).
|
), both the CC and the AT genotypes showed the greatest relative increment in transcription when insulin was present (CC > AC, P = 0.03; CC > CT, P = 0.03; AT > AC, P = 0.03; AT > CT, P = 0.04). We concluded that IGFBP3 promoter haplotype is capable of influencing transcription under basal and insulin-stimulated conditions.
Transcription Element Search System (TESS) analysis of the nucleotide sequence flanking the –202 and –185 polymorphisms
The result obtained in the TESS analysis is shown in Fig. 3A for the sequence containing the –202 and –185 IGFBP3 promoter polymorphisms and flanking nucleotides; the schematic in Fig. 3B situates this region within a larger region of the promoter. Among the many putative response elements, of particular interest was the presence of two E-boxes (–200CAC GAG and –191CAG GTG), which potentially bind the basic helix-loop-helix leucine zipper (bHLH-zip) transcription factor family, such as the USF proteins (Fig. 3A). Although these E-boxes do not represent the canonical E-box and flanking nucleotides usually recognized by USF (28), binding of USF to the –200CAC GAG E-box has been reported for the Herpes simplex virus promoters (29). Furthermore, an IRE present in the fatty acid synthase (FAS) promoter contains an E-box DNA-binding motif recognized by USF (30). We therefore investigated this region further for USF binding using several different experimental approaches.
|
, A and B, lanes 2 and 3). Competition with a 20-fold molar excess of cold oligonucleotide corresponding to the FAS gene E-box resulted in a complete loss of these complexes, as did incubation with a USF1 antibody that is directed to the C-terminal region of USF, adjacent to the DNA-binding domain (Fig. 4A, lanes 5 and 6). Moreover, a supershift was obtained with a USF2 antibody, which recognizes the N-terminal region of USF2 not involved in USF homo-/heterodimerization or DNA binding (Fig. 4A, lane 7).
Because it has been reported that binding of USF dimers to DNA is abolished by methylation of the central CpG (31), we next used cold in vitro methylated competitors in the gel shift assay (Fig. 4B). Addition of a 100-fold molar excess of the unmethylated A–202C–185 competitor (wild-type E-boxes) resulted in a complete loss of the retarded bands, suggesting that the observed binding is specific to the probe used (lane 5). Second, when the same oligonucleotide was methylated with M.SssI and used as a competitor at 100-fold molar excess, only a reduction of the intensity of the retarded bands was obtained, confirming that binding of the USF factors to this site is methylation dependent (lane 7).
To examine the relative contribution of each E-box to USF binding, we first compared the intensity of DNA-protein complexes obtained with wild-type probe or one in which a single nucleotide was mutated within the upstream E-box (see Table 1, E-box mutant). We observed a reduction but not an abolition of USF binding, as evidenced by the decreased band intensity, suggesting that the downstream E-box also participates in the binding of USF (Fig. 5A, lanes 3 and 4). Supershifted complexes were again seen with only USF2 antibodies, and the bands were more intense when wild-type probe was used (Fig. 5A, lanes 9 and 10).
We next designed a series of E-box mutants (Table 1) to use as cold competitors in gel shift analysis experiments, in which one or both of the E-boxes was mutated, allowing us to determine the relative importance of each site. Competition with 50- and 100-fold molar excess of mut1 and mut2 oligonucleotides (two different mutations in upstream E-box) and mut3 oligonucleotide (downstream E-box mutation) resulted in a reduction in the intensity of retarded bands. However, use of a competitor (mut4) in which both E-boxes were altered proved ineffective in displacing the probe, confirming the implication of both E-boxes in USF binding to this region. We also noted the presence of a smaller DNA-protein complex that could be abolished with the wild-type, mut1, and mut2 oligonucleotides, indicating the binding of another as yet unidentified protein to the –191 E-box. Taken together, our results indicate that USF proteins are able to bind both E-boxes. Furthermore, we have observed an unidentified protein capable of binding the 3' E-box (Fig. 5B).
Participation of USF in insulin responsiveness
Given that the upstream E-box contains a central CpG dinucleotide that is important for methylation-sensitive USF binding, we mutated the upstream E-box located at nucleotides –200 to –195 (CAC GAG) and tested its ability to modulate insulin-stimulated transcription of an IGFBP3 promoter-luciferase construct (see Table 1). This mutation was shown to abolish binding of USF in a competition assay and to lower by 3- to 4-fold basal transcriptional activity of the HSV-2 promoter-CAT plasmid (29); in the gel shift analysis of Fig. 5, a cold oligonucleotide bearing the same mutation (mut1) was an effective competitor.
Levels of transcription obtained with SK-HEP-1 cells transfected with pBP3/–442 AC containing either a wild-type E-box or our mutated E-box are shown in Fig. 6. The first observation was that basal levels of luciferase activity were similar with both wild-type and mutated constructs (P = 0.177), although a trend toward decreased basal activity was observed. Second, transcription activation in the presence of insulin was greatly diminished with a mutated E-box construct compared with the wild-type (P = 0.0006), although a significant level of insulin stimulation was still obtained (P = 0.0047). We therefore concluded that this E-box may function as an IRE.
|
We observed an enrichment of the IGFBP3 promoter with both antibodies against USF1 and USF2 when used with chromatin of cells treated with insulin, whereas chromatin from control cells did not produced similar enrichment (Fig. 7A, lanes 3–6). These results demonstrate the binding of both USF1 and USF2 to the IGFBP3 promoter and validate the data obtained using gel shift analysis. Moreover, the ChIP analysis showed an increased binding of these proteins to the IGFBP3 promoter of insulin-treated SK-HEP-1 cells as well as binding of USF1 to the IGFBP3 promoter from untreated cells. We also used anti-p53 antibodies to test for binding of p53 to our target promoter sequence, because binding to this region has been reported. As shown in the right panel of Fig. 7A, we confirmed binding of p53 but noted no stimulatory effect of insulin treatment at the time point studied.
We also performed ChIP experiments using another cell line (HepG2); binding of USF1 and USF2 to the IGFBP3 promoter was also confirmed using chromatin from both basal and insulin-stimulated cells (Fig. 7B). Cells were insulin treated for shorter periods of time (2 and 6 h); SK-HEP-1 cells exposed to insulin for 2–6 h also demonstrated USF binding (data not shown). We thus demonstrated USF binding to the IGFBP3 promoter within the context of chromatin from two cell types as well as its increased recruitment in response to insulin.
Given our results showing the importance of the E-box to the insulin response in the transfections discussed in Fig. 6 as well as the ChIP data above, we next tested the ability of one to three copies of the USF binding sites placed upstream of two heterologous promoters, thymidine kinase (pTKluc) and SV40 (pGL3PROM), both containing the luciferase reporter gene. We were unable to show significant insulin stimulation compared with the empty vectors, suggesting that promoter context and/or other transcriptional regulators are important for the insulin response (data not shown).
IGFBP3 promoter haplotypes influence the local CpG methylation pattern
Because the USF binding site is sensitive to CpG methylation, we next examined the methylation of genomic DNA by bisulfite sequencing of the region surrounding the –202 and –185 polymorphisms. For this purpose, we chose to use leukocyte DNA from individuals homozygous for the different haplotypes. Genomic DNA from nine individuals (–202 AA/–185 CC, n = 3; –202 CC/–185 CC, n = 3; –202 CC/–185 TT, n = 3) was subjected to bisulfite treatment followed by direct sequencing. A representative sequencing gel is shown in Fig. 8. The global methylation was more abundant in individuals with –202 C and –185 C haplotypes compared with –202 A and –185 C. The individuals with a –202 C have one more methylated CpG and higher methylation level of the neighboring cytosines. Moreover, if we look at the –202 C and –185 T individuals, there was less methylation of the –210, –202, and –183 cytosines compared with what is seen for –202 C and –185 C individuals. Therefore, we concluded that the promoter haplotype influences at least the local methylation pattern, suggesting that this could, in turn, influence USF binding and insulin-stimulated IGFBP3 transcription.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The present paper takes this observation further and provides evidence for the importance of promoter haplotype at two proximal SNPs (–202 A/C and –185 C/T) on basal and insulin-stimulated transcription. We identified two E-boxes adjacent to these SNPs that effectively bind USF1/2 in a methylation-dependent manner. Because we also show that haplotype can influence local methylation of leukocyte genomic DNA, we propose that this may be one of the operative mechanisms explaining how SNPs can influence transcription factor binding and, ultimately, explain biological variability in circulating IGFBP-3. Finally, in keeping with the observed in vivo response of circulating IGFBP-3 to insulin, in vitro studies have shown that liver-derived cell lines and dermal fibroblasts increase IGFBP3 transcription, mRNA, and protein in response to insulin (11, 38, 39); this paper extends this finding to the SK-HEP-1 line.
Using a series of deletion constructs, we explored basal luciferase activity and found that the highest luciferase activity was obtained with the –247 to +19 promoter fragment (Fig. 1A, representative experiment). Optimal insulin-stimulated reporter gene transcription was also obtained with this construct when data from four experiments were analyzed, despite the interassay variability that can be seen with transfection methodology (Fig. 1B). An equivalent fold induction was seen with a construct containing only 80 upstream nucleotides (Fig. 1A); insulin induction with this short construct was also reproducible (fold stimulation, 1.62 ± 0.01; n = 2 experiments). Previously, Cubbage et al. (7) had mapped the IGFBP3 minimal promoter to a fragment of 431 bp, but these investigators did not explore constructs of smaller length. Thus, deleting the Sp1-binding GC box and preserving only the TATA box (–81 to +19) still resulted in significant transcriptional activation under both basal and insulin-stimulated conditions. This is similar to the recent findings of Ongeri et al. (40) for FSH activation of the highly homologous region in the pig Igfbp-3 gene. These investigators mapped an FSH response element capable of binding Sp1 and lying 5' and adjacent to the TATA box (–61 to –48 bp); both this site and the TATA box are necessary and sufficient for FSH stimulation in porcine granulosa cells. Because signaling pathways not only for the FSH receptor but also for the insulin receptor include the phosphoinositol 3-kinase/AKT kinase cascade and lead to an increase in both Sp1 activation and nuclear translocation (41, 42), it is possible that an Sp1-TBP transcriptional complex also participates in the insulin-stimulated IGFBP3 expression in addition to USF and the upstream E-boxes. Indeed, Sp1 is an important factor in the activation of many genes in response to insulin, through increased binding and/or transactivation properties leading to increased interaction with the basal transcriptional machinery (43).
Our results in the SK-HEP-1 cell line also confirm our previous findings using IGFBP3 promoter constructs encompassing nucleotides –441 to +91 obtained under basal conditions both in the SK-HEP-1 cell line and in a breast cancer cell line (MCF-7) for the –202 A/C SNP (3). In the present study, the AC haplotype was the most efficient at stimulating transcription of the luciferase reporter gene, whether under basal or insulin-stimulated conditions (Fig. 2). Interestingly, in the several cohorts studied, it is also the most prevalent haplotype.
For the reporter gene studies, the AT haplotype was created by mutagenesis, because it has not been found either by us or by other laboratories in any of the patient populations studied. It is not yet clear whether this can simply be explained by the fact that –185 C/T is a more recent polymorphism; indeed, in some, but not all cohorts, it has not reached Hardy-Weinberg equilibrium. However, an alternative explanation could be that this haplotype is subject to negative selection. The AT haplotype affected reporter gene transcription in a manner similar to the CC haplotype whether under basal or insulin-stimulated conditions (Fig. 2), suggesting that it does not possess any anomalous behavior under our assay conditions.
Luciferase constructs containing longer IGFBP3 promoter fragments resulted in significantly lower transcriptional activity under both basal and insulin-stimulated conditions. Hanafusa et al. (18) also obtained lower luciferase levels with a –1.9-kb IGFBP3 promoter construct used in transient transfections with or without cotransfection with a p53 expression vector. This prompted us to look for putative repressor protein binding sites by in silico analysis. We found two putative E-box motifs located at nucleotides –744 to –739 and –654 to –649 that could potentially bind c-Myc and/or USF. It has been reported that c-Myc or one of its binding partners, MAD [MAX (MYC-associated factor X) dimerization protein 1], could act as a repressor of transcription (44); additional studies are necessary to explore this possibility.
Little is known about human IGFBP3 IREs. An IRE of the rat Igfbp-3 upstream sequence has been identified in hepatic nonparenchymal cells that binds an insulin-responsive protein (IRE-BP1) (21, 45). Although BLAST analysis using the rat IRE sequence found a mouse sequence with 96% homology corresponding to upstream mouse Igfbp-3 sequence (46), we found no homology with the human IGFBP3 promoter sequence.
We directed our study to USF as a potential transcription factor implicated in the insulin response for several reasons. Our initial studies showed that the maximal insulin response was obtained with promoter fragments ranging in size from nucleotides –442 to –250. In silico transcription factor searches for potential cis-acting elements showed two putative E-boxes within the first 250 nucleotides of the promoter, located adjacent to the –202 and –185 SNPs. E-boxes are known to bind to a series of transcription factors of the basic helix-loop-helix leucine zipper (bHLH-zip) family, including the insulin-responsive factors, USF1 and USF2. These upstream stimulatory factors have previously been shown to regulate insulin-responsive genes involved in lipid metabolism, notably fatty acid synthase gene (FAS) (30), acetyl-CoA carboxylase- gene (ACC-) (47), and apolipoprotein A5 gene (APOA5) (48). Recently, they have also been shown to activate genes involved in cell cycle control and growth, including the cyclin B1 gene and at least two genes in the IGF axis, IGF2R (49) and IGFBP-1 (50). The transcriptional response (activation vs. repression) appears to depend, in part, on whether homo- or heterodimerization has occurred (51), although in the majority of cells, heterodimers predominate (52).
Using both gel shift assays and ChIP, we demonstrated that both USF1 and USF2 could bind to the IGFBP3 E-boxes adjacent to our SNPs of interest, and furthermore, their binding was methylation dependent as shown by competition in gel shift assay (Fig. 4). Both E-boxes appear to be important for maximal USF binding on the basis of competition experiments (Fig. 5) and ChIP analysis (Fig. 7). The E-box (–200/–195) may be more important considering the inability of mut2 to further compete the probe even at a molar excess of 100 times (Fig. 5B, lane 8). Indeed, in vitro reporter gene assays confirmed that mutation of only one E-box significantly decreased insulin-stimulated gene transcription (Fig. 6). Interestingly, equivalent binding of USF was observed in gel shift experiments using nuclear extracts from both control and insulin-stimulated cells, whereas increased USF binding was seen in the ChIP assay with the chromatin from insulin-treated cells. These discordant results may be explained by insulin-induced chromatin rearrangement at the IGFBP3 promoter, promoting USF binding, as suggested by a recent paper studying the insulin-dependent regulation of the ADD1/SREBP1c gene via interaction with the SWI/SNF chromatin remodeling complex (53). Additional evidence that promoter context is important for insulin-stimulated USF binding was also provided by our inability to show activity of USF E-boxes placed immediately upstream of heterologous promoters.
Gel shift analysis with two probes representing IGFBP3 –202 and –185 haplotypes AC and CC confirmed USF binding, although we were not able to detect any significant association between the amount of USF binding and haplotype (data not shown). Quantification of protein binding is hard to assess using this technique. However, we do show that the global methylation of CpGs surrounding the –202 SNP is haplotype dependent. In effect, bisulfite sequencing analysis of individuals homozygous for the various haplotypes at the –202 and –185 polymorphisms showed different methylation patterns involving not only the upstream E-box but also neighboring CpG dinucleotides (Fig. 8). Moreover, cell lines with varying haplotypes for the –202/–185 SNPs also have different patterns of methylation (not shown). Taken together with the methylation sensitivity of USF binding, we suggest that this provides an epigenetic level of gene regulation contributing to interindividual variability.
Transcription repression by CpG methylation has been shown to be an important determinant of human IGFBP3 transcription in tumor-derived cell lines, such as hepatocellular carcinoma cell lines and gastric cancer cell lines (22, 54). IGFBP3 promoter hypermethylation has also been associated with prognosis in both non-small-cell lung cancer and in ovarian cancers (55, 56). Recently, Chang et al. (57) have shown that methylation of the Sp1/Sp3 binding element (nucleotides –97 to –91) in the IGFBP3 promoter decreased binding of Sp1 and increased binding of transcriptional repressors such as methyl-CpG-binding protein-2 (MeCP2) and histone deacetylase (HDAC). A methylated IGFBP3 promoter luciferase construct showed a significantly reduced transcriptional activity.
The –200CACGAG–195 sequence found in the IGFBP3 upstream E-box represents a noncanonical E-box that could also potentially bind c-Myc oncogene, as has been previously described (58). It has been proposed that USF, especially USF2, may act as an antiproliferative factor counterbalancing the tumorigenic effects of c-Myc (59, 60).
Among the other transcriptional regulators of IGFBP3 potentially interacting at the site of the –202/–185 polymorphisms, the tumor suppressor p53 has been shown to bind to this region in a methylation-sensitive manner (18). The –202 A genotype (but not the –202 C) alters the p53 consensus sequence, whereas either polymorphism at –185 preserves the consensus sequence. Our ChIP analysis confirmed p53 binding to this region of the SK-HEP-1 cells’ IGFBP3 promoter (Fig. 7). We genotyped this cell line as –202 C/C and –185 C/C, and bisulfite sequencing revealed hypomethylation of this region of the promoter (data not shown). These results are in complete agreement with the recent finding of Hanafusa et al. (18) regarding the reduced binding of p53 in the presence of methylated binding sites.
In silico analysis of the IGFBP3 promoter and 5' upstream sequences revealed the presence of multiple putative Sp1 binding sites. In a study of the deoxycytidine kinase gene (dCK), pull-down assays have suggested a possible interaction between Sp1 and USF (61). A putative Sp1 binding site immediately upstream of the –202 polymorphism was confirmed by gel shift assay with a probe for nucleotides –216 to –176 and a competitor representing a consensus Sp1 binding site. Supershift assays using anti-USF antibodies did not reveal any interaction between USF and Sp1, because the USF complex, which is also bound to the same oligonucleotide, was supershifted, but not the Sp1 complex (not shown). Interestingly, it is reported that binding of USF to DNA induces a bend toward the minor groove, and the presence of adjacent binding sites can induce a triple-bended structure (62). Therefore, a potential for interaction between USF binding to the E-boxes between the –202 and –185 polymorphisms and Sp1 binding either to the GC box or to the other putative sites (Fig. 3) exists but remains to be explored further.
The USF proteins are under consideration as possible tumor suppressors, because loss of function has been associated both with prostate hyperplasia and carcinoma as well as with breast cancer (59, 63). IGFBP-3, because of its ability to regulate IGF-I levels as well as its direct antiapoptotic effects, is also under extensive investigation as a modulator of cancer risk, and circulating levels have been shown to be negatively associated with cancer risk (64). We have previously studied the association of the IGFBP3 –202 polymorphism with colon cancer and were unable to demonstrate a genotype-dependent effect on cancer risk (3). However, in Asian populations, the IGFBP3 –202 polymorphism was associated with breast, prostate, and lung cancer risks (34, 35, 37). Associations with premenopausal breast cancer risk have not been seen uniformly within Caucasian populations and may reflect heterogeneity in sample size, study design, and statistical analysis (65), although in large case-control series, high IGFBP-3 levels have been shown to be protective and associated with the IGFBP3 –202 A allele (66).
We also have shown significant interactions between IGFBP3 –202 genotype and BMI, a measure of adiposity, on circulating IGFBP-3 levels (3). Furthermore, obesity is also associated with risk for several cancers, and part of this risk has been attributable, among other things, to both increased insulin and IGF-I levels. The contribution of insulin- or IGF-I-induced IGFBP-3 gene regulation to cancer risk is undoubtedly only one small part of the total picture, although it adds impetus to further molecular studies on the IGFBP-3 gene aimed at understanding interindividual variability in IGFBP3 regulation.
In conclusion, IGFBP3 can be added to the genes regulated by insulin that bind the insulin-modulated transcription factor USF in a methylation-sensitive manner. We have shown that the promoter polymorphisms, –202 A/C and –185 C/T, influence the basal and insulin-stimulated transcriptional activity of IGFBP3 and thus correlate with the in vivo effects previously reported. Two putative binding sites for USF were confirmed by gel shift and ChIP assays and were shown to participate in the insulin response. Methylation-sensitive USF binding is relevant to our observation that the IGFBP3 haplotype at the –202 and –185 polymorphisms influences local methylation status of the promoter in human leukocytes. Better understanding of the IGFBP3 promoter regulation and the implication of sequence variations and epigenetic modifications will help explain biological diversity and complex traits such as cancer.
|
Footnotes |
|---|
Disclosure Statement: All authors have nothing to declare.
First Published Online September 6, 2007
Abbreviations: BMI, Body mass index; ChIP, chromatin immunoprecipitation; IGFBP-3, IGF-binding protein 3; IRE, insulin response element; SNP, single-nucleotide polymorphism; Sp1, specificity protein 1; TESS, Transcription Element Search System; TF, transcription factors; USF, upstream stimulatory factor.
Received December 22, 2006.
Accepted for publication August 27, 2007.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
,25-dihydroxyvitamin D3 growth inhibition in the LNCaP prostate cancer cell line through p21/WAF1. J Urol 165:1319–1324[CrossRef][Medline]
), numbered relative to the transcription start site, and the previously described minimal promoter (
) overlapping the polymorphisms are also shown (
