Activating Transcription Factor-1-mediated Hepatocyte Growth Factor-induced Down-regulation of Thrombospondin-1 Expression Leads to Thyroid Cancer Cell Invasion*

Hepatocyte growth factor (HGF) plays a major role in the pathogenesis of a variety of human epithelial tumors including papillary carcinoma of the thyroid. Previous reports demonstrated that HGF, acting through the Met receptor, repressed thrombospondin-1 (TSP-1) expression. To study the mechanisms by which HGF down-regulated TSP-1 expression, we transiently transfected a panel of deleted human TSP-1 promoter reporter plasmids into papillary thyroid carcinoma cells. We identified a region between –1210 and –1123 bp relative to the transcription start site that is responsive to HGF treatment and harbors a cAMP-responsive element (CRE) at position –1199 (TGACGTCC). Overexpression of various members of the CRE-binding protein family identified activating transcription factor-1 (ATF-1) as the transcription factor responsible for HGF-induced repression of TSP-1 promoter activity. This inhibition was associated with a concomitant increase in the abundance of nuclear ATF-1 protein. Gel shift and antibody supershift studies indicated that ATF-1 was involved in DNA binding to the TSP-1-CRE site. Finally, we utilized small hairpin RNA to target ATF-1 and showed that these small interfering RNA constructs significantly inhibited ATF-1 expression at both the RNA and the protein level. ATF-1 knockdown prevented HGF-induced down-regulation of TSP-1 promoter activity and protein expression and also reduced HGF-dependent tumor cell invasion. Taken together, our results indicate that HGF-induced down-regulation of TSP-1 expression is mediated by the interaction of ATF-1 with the CRE binding site in the TSP-1 promoter and that this transcription factor plays a crucial role for tumor invasiveness in papillary carcinoma of the thyroid triggered by HGF.

Hepatocyte growth factor (HGF) 3 , also known as scatter factor-1, is a mesenchymal-or stromal-derived multifunctional cytokine/growth factor (1, 2) that acts predominantly on cells of epithelial origin in an endocrine and/or paracrine fashion (3,4) through binding the high affinity c-Met tyrosine kinase receptor (5). In vitro, HGF-Met interaction elicits a complex and specific genetic program leading to epithelial cell dissociation ("scattering") (3), migration and extracellular matrix invasion (3,6), proliferation and protection from apoptosis (7,8), acquisition of polarity, and branching morphogenesis (9 -11). In vivo, HGF-Met signaling clearly plays a role in normal cellular processes during embryonic development, and many of these normal activities have been implicated in tumor progression and metastasis. HGF is also a potent inducer of angiogenesis (12) by regulating positively the pro-angiogenic factor vascular endothelial growth factor and negatively the anti-angiogenic protein thrombospondin-1 (TSP-1), an effect that was recently described in a breast cancer cell line (13) and in thyroid carcinoma cells (14).
Papillary thyroid carcinoma represents the most common malignancy of the thyroid gland. Because Met is not expressed or only expressed at low level in the normal thyroid, aberrant expression and autocrine or mutational activation of c-Met receptor in papillary cancer suggest a possible role for the HGF-Met axis in tumor development and progression (15)(16)(17).
TSP-1, the first protein to be recognized as a naturally occurring inhibitor of angiogenesis, is a large multifunctional extracellular matrix glycoprotein that affects tumor growth and metastases through modulation of angiogenesis and other stromal biological functions (18). However, the involvement of TSP-1 in tumor progression remains complex and controversial since both stimulatory (19,20) and inhibitory (21,22) effects in angiogenesis have been described and assigned to up-regulation of matrix-degrading enzymes and their inhibitors.
Thus far, the transcriptional regulation of TSP-1 gene expression by HGF has not yet been investigated. The present study was undertaken to examine the molecular mechanisms by which HGF down-regulates TSP-1 expression in the human papillary thyroid carcinoma cell line TPC-1. We identified activating transcription factor 1 (ATF-1) binding to the ATF/ CREB-responsive element as a negative regulator of TSP-1 expression. Functional analyses revealed that silencing of gene expression at the mRNA and protein level by small hairpin RNAs (shRNA) directed against ATF-1 reversed HGF-dependent down-regulation of TSP-1 expression and inhibited HGFinduced TPC-1 cell invasion.

EXPERIMENTAL PROCEDURES
Cell Line and Materials-The human thyroid papillary carcinoma cell line TPC-1 was kindly provided by Dr. J. A. Fagin (University of Cincinnati, Cincinnati, OH). Cells were routinely grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 2 mM glutamine, 5% fetal bovine serum, and 1% penicillin/streptomycin (Invitrogen) in a humidified CO 2 incubator at 37°C. Human recombinant HGF was from Calbiochem. Rabbit polyclonal anti-TSP-1 was used as described previously (35). Mouse monoclonal antibody raised against recombinant human ATF-1 (clone C41-5.1), which does not cross-react with other members of the ATF/ CREB family, was purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Anti-USF1 and anti-USF2 antibodies were obtained as described (43). Mouse monoclonal anti-␤actin antibody was purchased from Sigma. Goat anti-mouse and goat anti-rabbit IgG-peroxidase conjugates were from Amersham Biosciences (Orsay, France).
Western Blot Analysis-Cultured cells were made quiescent by serum starvation for at least 24 h before incubation with human recombinant HGF. Conditioned media were collected and centrifuged at 5,000 ϫ g for 10 min, and total cell protein was measured using bicinchoninic acid microassay from Pierce. Cells were then washed with ice-cold phosphate-buffered saline, lysed in buffer containing 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1 mM EDTA, 1 mM sodium Na 3 VO 4 , 50 mM NaF, 1 mM 4-(2-aminoethyl)-benzenesulfonylfluoride, and 1 g/ml each aprotinin, leupeptin, pepstatin, and antipain, placed on ice for 20 min, and then centrifuged at 14,000 ϫ g for 15 min at 4°C. Equal amounts of protein (whole cell extract, nuclear extracts, conditioned media) were resolved by SDS-polyacrylamide gel electrophoresis under reducing conditions and then transferred to nitrocellulose membranes. The blot was stained with Ponceau S to confirm equal loading of proteins and then probed with the indicated antibodies. Immunoblots were developed using appropriate secondary horseradish peroxidase-coupled antibodies and an enhanced chemiluminescence (ECL) kit (Amersham Biosciences). To ensure equal loading of proteins from nuclear extracts, the membranes were stripped and reprobed with an anti-␤-actin antibody under the same conditions as described above.
Probes were prepared by labeling 3.5 pmol of the double-stranded oligonucleotides with [␥-32 P]ATP and T4 polynucleotide kinase (Invitrogen). Five micrograms of each nuclear protein was diluted in 20 l of gel retardation buffer (10 mM Tris/HCl, 100 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, 0.5 mM MgCl 2 , 10% glycerol). 32 P-labeled double-stranded oligonucleotide probes (5,000 cpm) were incubated for 20 min with nuclear extracts in the presence of the nonspecific DNA sequence poly(dI-dC)⅐poly(dI-dC).
In competition experiments, the nuclear extract was incubated with a 100-fold molar excess of the appropriate unlabeled specific and nonspecific competitor oligonucleotides. After incubation at room temperature, electrophoresis of the different samples was carried out on nondenaturating polyacrylamide gels for 3 h in 0.5ϫ Tris borate/ EDTA. Gels were dried under vacuum and exposed to Kodak XAR film overnight. In supershift studies, a mouse monoclonal anti-ATF-1 or a nonspecific IgG was preincubated with the crude nuclear extract for 3 h at 4°C before the addition of the labeled probe.
Cell Invasion Assay-Invasion of TPC-1 cells in vitro was investigated using modified Boyden chambers (tissue culturetreated, 6.5-mm diameter, 8-m pore size; Transwell Costar, Brumath, France). Cells were trypsinized, suspended (1 ϫ 10 6 cells/ml) in serum-free Dulbecco's modified Eagle's medium containing 0.2% bovine serum albumin, and seeded onto membranes coated with 30 g/cm 2 Matrigel (extracellular matrix gel, Sigma). Purified TSP-1 or anti-TSP-1 antibody (clone 3F355, 30 g/ml) were added to the upper chamber of the Transwell units. Complete medium containing HGF (50 ng/ml) was A, TSP-1 mRNA expression from serum-starved TPC-1 cells left untreated (CTL) or treated with 50 ng/ml HGF for 24 h was analyzed by RT-PCR. After incubation, total RNA was prepared, and TSP-1 expression was evaluated using PCR analysis as described under "Experimental Procedures." The S26 mRNA was co-amplified as a control. B, TSP-1 protein was analyzed in supernatants from 48-h HGF-treated or untreated (CTL) cells by Western blot using an anti-TSP-1 antibody. C and D, starved TPC-1 cells transiently transfected with the Ϫ1290/ϩ750-bp human TSP-1 promoter upstream of the luciferase gene (LUC) were treated with various concentrations of HGF for 24 h (C) and with 50 ng/ml HGF for the time points as indicated (D). Luciferase activities were determined as described under "Experimental Procedures." The luciferase activity from an unstimulated sample of the Ϫ1290/ϩ750 reporter construct was established as 100%. Each value is the mean Ϯ S.E. of three independent experiments each performed in triplicates. E, plasmids containing various lengths of the 5Ј-flanking sequence of the human TSP-1 gene (Ϫ1290/ϩ750, Ϫ1210/ϩ750, Ϫ1123/ϩ750, Ϫ767/ϩ750, Ϫ267/ϩ750, Ϫ71/ϩ750, or Ϫ43/ϩ750) and a luciferase reporter (LUC) gene are depicted. The transcription start site (arrow) and cisacting elements are indicated. SRE, serum-response element. F, quiescent TPC-1 cells were transiently transfected with each plasmid, treated or not with 50 ng/ml HGF for 24 h, and harvested for luciferase assays. Shown for each plasmid is the ratio (percentage) of relative luciferase activity in HGF-treated cells to that in untreated cells (mean Ϯ S.E. of three independent experiments).
added to the lower compartment of the chamber. After 24 h at 37°C, cells on the upper membrane surface were removed by careful wiping with a cotton swab, and the filters were fixed by treatment with methanol and stained with 0.5% crystal violet solution for 15 min. Invasive cells adhering to the undersurface of the filter were then counted (five high power fields/chamber) using an inverted microscope (Nikon Eclipse).

RESULTS AND DISCUSSION
Identification of Binding Sites That Mediate Down-regulation of the Human TSP-1 Promoter by HGF-HGF-induced downregulation of TSP-1 expression has been reported in various cells including thyroid papillary carcinoma (13,14). We performed RT-PCR analysis on total cellular RNA from HGF (50 ng/ml)-treated TPC-1 cells using specific primers for TSP-1 and S26. As shown in Fig. 1A, the PCR products of TSP-1 and S26 were amplified to the appropriate sizes of 604 and 250 bp, respectively, and HGF incubation induced a down-regulation in TSP-1 mRNA level. To determine the relationship of TSP-1 protein expression to TSP-1 mRNA expression after HGF stimulation, Western blot analysis was performed in culture supernatants (Fig. 1B). These results demonstrated that HGF caused a reduction in TSP-1 production at the protein level. To determine whether HGF was capable of down-regulating TSP-1 promoter activity, we transiently transfected TPC-1 cells with the Ϫ1290/ϩ750-bp human TSP-1 promoter fused to the luciferase reporter gene. HGF treatment resulted in a repression of TSP-1 promoter activity in a dose- (Fig. 1C) and time-dependent (Fig. 1D) manner.
The location of potential binding sites for transcription factors in the human TSP-1 promoter has been published (41,42) and is shown in Fig. 1E. To identify sequences that mediate down-regulation of TSP-1 promoter activity by HGF, functional analysis was carried out using a series of luciferase reporter gene plasmids containing various lengths of the human TSP-1 5Ј-flanking sequences into TPC-1 cells. As shown in Fig. 1F, HGF decreased luciferase activity in the Ϫ1290/ϩ750 and Ϫ1210/ϩ750 plasmid constructs, respectively. When the sequence between nucleotides Ϫ1210/ϩ750 and Ϫ1123/ϩ750 was deleted, the inhibitory effect of HGF was lost, indicating that the region between nucleotides Ϫ1210/ϩ750 to Ϫ1123/ϩ750 contains a negative regulatory element required for transcriptional inactivation of TSP-1. A potential CRE binding site (TGACGTCC) for ATF/CREB transcription factors is located within this sequence. Fig. 1, E and F, also showed that further deletion of the TSP-1 promoter between nucleotides Ϫ1123/ϩ750 and Ϫ767/ϩ750 resulted in a supplementary increase in luciferase activity. This region harbors a putative E-box motif (CAGATG) able to bind USF proteins that might also participate in the negative regulation by HGF.

ATF-1 Transcription Factor Is Involved in the Down-regulation of TSP-1 Expression by HGF-A number of transcription
factors specifically bind to the CRE sequence as homo-and heterodimers via a carboxyl-terminal basic domain leucine zipper (bZip) motif (49). To identify the proteins that could be involved in the regulation of TSP-1 promoter, TPC-1 cells were co-transfected with different plasmids encoding vari-ous members of the ATF/CREB family such as CREB, CREM isoforms, ATF-1, and ICER. Fig. 2A showed that only ATF-1 overexpression was able to inhibit basal TSP-1 promoter activity as well as to enhance the inhibitory effect elicited by HGF, and the observed effect was dose-dependent (data not shown). Expression of all the other members stimulated TSP-1 promoter activity. To test whether ATF-1 phosphorylation is required for HGF repression of the TSP-1 promoter, we used a dominant negative ATF-1 phosphorylation site mutant (S63A). Transfection with this construct had no effect on HGF-induced inhibition of the TSP-1 promoter reporter gene (data not shown). We than tested whether HGF had an effect on the abundance of ATF-1 protein.
Western blot analysis with nuclear extracts from TPC-1 cells showed that HGF treatment caused accumulation of the ATF-1 transcription factor in a time-dependent manner (Fig. 2B). These findings suggest that ATF-1 plays a major role in HGF-mediated down-regulation of TSP-1 promoter activity in TPC-1 cells. Our results corroborate with earlier studies that demonstrated the involvement of ATF-1 in the regulation of TSP-1 in nickel-transformed cells (37).
USF1 and USF2 Do Not Participate in HGF-induced Downregulation of TSP-1 Expression-USF1 and USF2, originally identified as activators of the adenovirus major late promoter (50), are ubiquitously expressed basic helix-loop-helix/leucine zipper transcription factors (51). They recognize and bind to DNA with an E-box motif as either homodimers or heterodimers. USF1 and USF2 regulate the expression of several genes (52, 53) including smooth muscle cell-expressed genes

. ATF-1 transcription factor is involved in HGF-induced inhibition of TSP-1 promoter.
A, TPC-1 cells were co-transfected with the Ϫ1290/ ϩ750-bp human TSP-1-LUC promoter construct and with expression plasmids (25 ng) for different members of the CREB/ATF-1 family including ATF-1, CREB, two isoforms of CREM␣ and CREM␤, and ICER. Cells were incubated in the presence or absence of HGF for 24 h and harvested, and luciferase activity was measured as described previously. CTL, control. B, cells were incubated with 50 ng/ml HGF for the time points as indicated. Nuclear extracts were prepared, and 50 g protein was analyzed by Western blot using an anti-ATF-1 antibody. The amounts of protein were controlled by probing cytosolic extracts from the same cells with an anti-␤-actin antibody. (54). More recently, Wang et al. (38) showed that repression of glucose-induced TSP-1 gene expression by cGMP protein kinase involved down-regulation of USF2 protein levels, resulting in a decrease in USF2 binding to a single region (Ϫ932 to Ϫ915) in the human TSP-1 promoter. To check whether or not USF proteins are involved in the negative regulation of TSP-1 promoter by HGF, co-transfection assays were performed with either the Ϫ1290/ϩ750-bp or the Ϫ1123/ϩ750-bp human TSP-1-LUC promoters and plasmids encoding USF1, USF2, or the dominant negative ⌬bTDU1 and ⌬bTDU2 constructs. Results from Fig. 3, A and B, showed that both USF1 and USF2 stimulated basal and HGF-dependent TSP-1 promoter activity, whereas the dominant negative forms were either without effect (Fig. 3A) or inhibitory (Fig. 3B) when compared with control pCR3 plasmid.
To prove the specificity of the USF plasmids, co-transfection experiments were also performed with a positive control construct that carries three copies of the L4 E-box ligated to the Ϫ96/ϩ54 proximal bp of the L-type pyruvate kinase promoter. Fig. 3C showed that USF1 and USF2 can act as transactivators of the (L4)3-96PK-LUC promoter, whereas both USF1 and USF2 mutants repressed transcription.
We then tested whether HGF had an effect on the amount of USF proteins. Western blot analysis with nuclear extracts from TPC-1 cells showed that HGF treatment had no impact on the amount of USF2 transcription factor level when compared with ␤-actin (Fig. 3D). Identical results were obtained when USF1 protein was analyzed (data not shown). In conclusion to our experimental data, we can rule out the possibility that USF1 and/or USF2 are involved in HGF-dependent repression of TSP-1 gene expression.
Identification of ATF-1 Binding to the CRE-To determine whether the transcription factor ATF-1 was able to bind specifically to the CRE motif, we performed electrophoretic mobility shift assays with labeled double-stranded oligonucleotide probes encoding the CRE consensus, TSP-1-CRE, or TSP-1-CREmut (TSP-1 ⅐ CRE mutated) sequence and nuclear extracts prepared from control and HGF-stimulated TPC-1 cells. Results from Fig.  4A indicate that incubation of nuclear extracts from HGF-treated cells with the CRE consensus probe resulted in an increase in a DNAprotein complex (lane 2) when compared with control (lane 1, arrow). The complex was competed away by an excess of unlabeled CRE consensus oligonucleotide (lane 3) but not by excess of a nonspecific sequence (lane 4). A similar increase in DNA-protein complex was also observed when the TSP-1-CRE oligonucleotide was used as a probe (compare lanes 5 and 6, arrow). In contrast, when an oligonucleotide containing a mutated TSP-1-CRE site was used as a probe, the specific DNA-protein complex disappeared (lanes 7 and 8). These results indicate that in TPC-1 cells, one complex corresponds to a transcription factor that specifically binds to the CRE within the TSP-1 promoter. To identify whether ATF-1 is the transcription factor that binds the promoter, we performed electrophoretic mobility shift assays with nuclear extracts from HGF-stimulated cells, using either CRE consensus or TSP-1-CRE-labeled oligonucleotide probes and an antibody against human ATF-1, which does not cross-react with other members of the ATF/CREB family. Two supershifts were observed with the anti-ATF-1 antibody (Fig. 4B, lane 2, asterisks) but not A and B, TPC-1 cells were co-transfected with the Ϫ1290/ϩ750 (A) or with the Ϫ1123/ϩ750 bp (B) human TSP-1-LUC promoter constructs and with expression plasmids for USF1, USF2, and USF dominant negative mutants ⌬bTDU1 and ⌬bTDU2. Empty expression vector (pCR3) was used as control. Cells were incubated in the presence or absence of HGF for 24 h and harvested, and luciferase activity was measured as described previously. CTL, control. C, TPC-1 cells were co-transfected with a construct carrying three copies of the L4 E-box ligated to the Ϫ96/ϩ54 proximal bp of the L-type pyruvate kinase promoter i.e. (L4)3-96PK-LUC and with expression plasmids for USF1, USF2, ⌬bTDU1, and ⌬bTDU2. As control, cells were co-transfected with the Ϫ96/ϩ11 proximal bp fragment of the L-type pyruvate kinase promoter (Ϫ96PK-LUC) and with pCR3. Cells were harvested after 24 h, and luciferase activity was measured. D, cells were incubated with 50 ng/ml HGF for the time points as indicated. Nuclear extracts were prepared, and 50 g of protein was analyzed by Western blot using an anti-USF2 (amino acids 1-49, domain G) antibody (43). The amounts of protein were controlled by probing cytosolic extracts from the same cells with an anti-␤-actin antibody.
with an irrelevant antibody (lane 3). Similar results were also obtained with the TSP-1-CRE probe (Fig. 4B, lanes 5 and 6), although the migration of the lower supershifted band dif-fered (lanes 2 and 5, compare the asterisks). These data clearly identify ATF-1 as the binding protein for the TSP-1-CRE site in TPC-1 cells. However, we cannot rule out the possibility that heterodimerization occurs with other members of the ATF-CREB family as it has been described by others (55). This can be emphasized by the fact that two bands are supershifted with a specific antibody (Fig. 4B,  lanes 2 and 5). One plausible explanation could be that ATF-1 may be bound to DNA as a homodimer or a complex with additional proteins in the extract and, owing to a difference in size or charge, each band is supershifted to a different position.
Silencing of the ATF-1 Gene by shRNA Significantly Reduced HGF-induced Down-regulation of TSP-1-To further confirm the involvement of ATF-1 during HGF-induced TSP-1 downregulation, a shRNA plasmid targeted at the human ATF-1 mRNA sequence was used in this study. Luciferase reporter assays using the shRNA construct and the Ϫ1290/ϩ750 TSP-1-LUC reporter plasmid were performed. As shown in Fig. 5A, gene silencing of ATF-1 prevented HGF-induced inhibition of TSP-1 promoter activity when compared with control (shCTL) vector. In addition, shATF-1 down-regulated ATF-1 expression at the mRNA (Fig. 5B) and protein level (Fig. 5C). This effect was correlated with the up-regulation of TSP-1 protein expression/secretion independently of HGF treatment. These findings clearly indicate that ATF-1 is an essential transcription factor that regulates TSP-1 expression.
ATF-1-dependent TSP-1 Down-regulation Promotes Tumor Cell Invasion-The effect of HGF on in vitro Matrigel invasion of TPC-1 cells was examined. As shown in Fig. 6A, HGF stimulated Matrigel invasion of TPC-1 cells. To confirm the involvement of TSP-1 in HGF-mediated promotion of tumor invasion, cells were allowed to invade into the Matrigel matrix in the presence or absence of TSP-1 protein. Incubation of cells with TSP-1 protein reduced basal as well as HGF-stimulated Matrigel cell invasion. To further determine the involvement of TSP-1 in TPC-1 cell invasion, we performed assays in the presence of a neutralizing antibody against TSP-1. Anti-TSP-1 antibody was able to significantly enhance basal TPC-1 cell invasion. Taken together, these results demonstrate that HGF-induced cell invasion is mediated at least in part by the regulation of TSP-1 expression. To address the involvement of ATF-1, invasion assays were performed using control and ATF-1 shRNA-expressing cells. Fig. 6B showed that knockdown of ATF-1 inhibited HGF-induced TPC-1 cell invasion. Collectively, our data demonstrate that ATF-1 transcription factor is a major intermediary in the signaling pathway that is responsible for HGF-induced TSP-1 down-regulation, leading to thyroid tumor cell invasion. As reported previously, disruption of ATF-1  activity in human metastatic melanoma cells by using an inhibitory anti-ATF-1 antibody fragment suppressed their tumorigenicity and metastatic potential in nude mice (56). Therefore ATF-1 may represent a target of choice to develop novel strategies for thyroid cancer therapy.