Stable Isotope Labeling with Amino Acids in Cell Culture (SILAC)-based Quantitative Proteomics Study of a Thyroid Hormone-regulated Secretome in Human Hepatoma Cells*

The thyroid hormone, 3, 3′,5-triiodo-l-thyronine (T3), regulates cell growth, development, differentiation, and metabolism via interactions with thyroid hormone receptors (TRs). However, the secreted proteins that are regulated by T3 are yet to be characterized. In this study, we used the quantitative proteomic approach of stable isotope labeling with amino acids in cell culture coupled with nano-liquid chromatography-tandem MS performed on a LTQ-Orbitrap instrument to identify and characterize the T3-regulated proteins secreted in human hepatocellular carcinoma cell lines overexpressing TRα1 (HepG2-TRα1). In total, 1742 and 1714 proteins were identified and quantified, respectively, in three independent experiments. Among these, 61 up-regulated twofold and 11 down-regulated twofold proteins were identified. Eight proteins displaying increased expression and one with decreased expression in conditioned media were validated using Western blotting. Real-time quantitative RT-PCR further disclosed induction of plasminogen activator inhibitor-1 (PAI-1), a T3 target, in a time-course and dose-dependent manner. Serial deletions of the PAI-1 promoter region and subsequent chromatin immunoprecipitation assays revealed that the thyroid hormone response element on the promoter is localized at positions –327/–312. PAI-1 overexpression enhanced tumor growth and migration in a manner similar to what was seen when T3 induced PAI-1 expression in J7-TRα1 cells, both in vitro and in vivo. An in vitro neutralizing assay further supported a crucial role of secreted PAI-1 in T3/TR-regulated cell migration. To our knowledge, these results demonstrate for the first time that proteins involved in the urokinase plasminogen activator system, including PAI-1, uPAR, and BSSP4, are augmented in the extra- and intracellular space of T3-treated HepG2-TRα1 cells. The T3-regulated secretome generated in the current study may provide an opportunity to establish the mechanisms underlying T3-associated tumor progression and prognosis.

The thyroid hormone (TH) 1 , a pleiotropic regulator of growth, differentiation, proliferation and many other physiological processes, acts via interactions with thyroid hormone response elements (TREs) in the regulatory regions of target genes (1). Cheng and co-workers have reported AGGTCA as a putative consensus hexamer half-site sequence of TRE (1). The TREs are arranged as direct repeats (DR), palindromes, and inverted palindromes (IP), and display considerable variations in nucleotide sequences, spacing, number, and orientation of half-sites (1)(2)(3). The amino acid sequences of thyroid hormone receptors (TRs) are highly homologous with those of steroid hormone receptors (2,4). TRs are ligand-dependent transcription factors that consist of modular function domains that mediate DNA binding, hormone binding (ligands), receptor homo-and heterodimerization, and interaction with other transcription factors and cofactors (5,6).
TRs interact with retinoid X receptor (RXR) and form heterodimers that influence target genes via binding to TREs located in the regulatory regions (7,8). TH-bound TRs activate target gene expression. In contrast, gene expression is repressed by non-T 3 -bound TRs. Unliganded TRs act as repressors by recruiting corepressors, such as silencing mediator of retinoic and thyroid receptor (SMRT) and nuclear receptor corepressor (NCoR), as a result of altered conformations upon binding to TREs. Conversely, binding of T 3 to TRs causes conformational changes and subsequent recruitment of multiple coactivator complexes (1,5). Two TR genes, TR␣ and TR␤, have been identified on human chromosomes 17 and 3, respectively (5,9). TR␣1, TR␣2, TR␤1 and TR␤2 isoforms are generated via alternative splicing and promoter usage of the primary transcript (10). The liver is the typical target organ of the thyroid hormone, and equivalent expression levels of TR␣1 and TR␤1 have been reported in human hepatocytes (11).
Previous microarray analysis experiments by our group have demonstrated that numerous genes including coagulation factor system components (12), plasma proteins (12,13), nuclear receptor coactivator (14), antimetastatic proteins (15), proteases (16), and oncogenes (17) are regulated by T 3 . Additionally, we have identified several T 3 -regulated extracellular proteins such as matrix metalloproteinases (MMPs) and cysteine cathepsins. These proteases are involved in cellular processes and intercellular communication during cancer progression and development (18 -20). It implied that T 3 may have a role in the regulation of secreted proteins. Recently, proteomics approach has been successfully applied to investigate the regulatory secreted proteins systemically (secretome) of tumor-associated genes (21)(22)(23)(24)(25). Herein, we applied the stable isotope labeling with amino acids in cell culture (SILAC)-based quantitative proteomic approaches to identify secreted proteins regulated by T 3 and study their underlying physiological significance in hepatoma cell lines stably expressing wild-type TR.
Plasminogen activator inhibitor-1 (PAI-1), a T 3 target, elevated levels in the tumor microenvironment is associated with high mortality and poor prognosis of patients with many forms of cancer (26). In the present study, we focused on the potential role played by PAI-1 after T 3 induction.
Preparation of Conditioned Medium and In-solution Protein Digestion-Cells were passaged and differentiated as described above. HepG2-TR␣1 cells were grown to confluence in 10-cm cell culture dishes. The cells contacting dishes were washed twice with phosphate-buffered saline (PBS) to reduce the amount of contaminating protein in serum (29). Cells were incubated in serum-free medium and either treated with T 3 or left untreated for 24 h. At the end of the treatment period, conditioned media (CM) were collected and centrifuged at 1500 ϫ g to eliminate intact cells following concentration using spin columns with a molecular mass cut-off of 3 kDa (Amicon Ultra, Millipore, Billerica, MA). Equal amounts of proteins were mixed for quantitative proteomic analysis (28).
Preparation of Tryptic Peptides in Solution-Mixed SILAC proteins were reduced and alkylated with 5 mM dithiothreitol for 60 min and 10 mM iodoacetamide for 60 min, followed by digestion with sequencing grade-modified trypsin (1:25, w/w) (Promega, Madison, WI) at 37°C overnight. The digestion reaction was terminated by adding formic acid at a concentration of 0.1%.
One-dimensional Gel Electrophoresis Combined with Nanoliquid Chromatography (GeLC)-Proteins separated using SDS/PAGE were cut out of the gel and divided into 60 fractions for subsequent in-gel digestion (31). Peptides were analyzed with LC-MS/MS using nanoscale RP18 liquid chromatography coupled with LTQ-Orbitrap mass spectrometry. Briefly, peptides were trapped on a Zorbax 300SB RP18 column (0.3 ϫ 5 mm, Agilent Technologies) and separated using a PicoFrit BioBasic C18 capillary column (0.075 ϫ 120 mm; New Objective, Woburn, MA) with an acetonitrile gradient in 0.1% formic acid on a Surveyor HPLC system (Thermo Electron, Bremen, Germany).
Tandem Mass Spectrometry (MS/MS)-MS/MS analysis was performed on a LTQ-Orbitrap mass spectrometer (Thermo Fisher, San Jose, CA) with a nanoelectrospray ion source (Proxeon Biosystem). Full-scan MS spectra (m/z 430 -m/z 2000) were acquired in the Orbitrap mass analyzer at a resolution of 60,000 at m/z 400. The lock mass calibration feature was enabled to improve mass accuracy. The most intense ions (up to 12) with minimal signal intensity of 20,000 were sequentially isolated for MS/MS fragmentation in the order of intensity of precursor peaks in the linear ion trap using collision-induced dissociation energy of 35%, Q activation at 0.25, activation time of 30 ms, and isolation width of 2.0. Targeted ions with m/z Ϯ 30 ppm were selected for MS/MS once (2D-LC) or twice (GeLC), and dynamically excluded for 50 (2D-LC) or 180 s (GeLC).
Protein Identification and Quantification in Shotgun Proteomics-All MS and MS/MS data were analyzed and processed with Quant. exe in MaxQuant environment (version 1.0.13.8) for peptide identification and quantification analyses as described before (32). Top six of fragment ions per 100 Da were extracted for a protein database search using the Mascot search engine (version 2.2.2, Matrix Science) against the concatenated Swiss-Prot version 56 human forward and reverse protein sequence data set with a set of common contaminant proteins (total 45500 entries). The search parameters were set as follows: Carbamidomethylation (C) as the fixed modification, oxidation (M), N-acetyl (protein) and pyro-Glu/Gln (N-term) as variable modifications, 7 ppm for MS tolerance, 0.5 Da for MS/MS tolerance, and 2 for missing cleavage. The SILAC label (K) (R) was set as none, fix or variable modification to generate three Mascot search results. The identified peptides and proteins in all search results were further analyzed in Identity.exe with following criteria: six for minimum peptide length, two for minimum unique peptides for the assigned protein. The posterior error probability of peptides identified in forward and reversed databases was used to rank and determine the false discovery rate for statistical evaluation (32,33). We accepted the peptide and protein identifications with false discovery rate less than 1%. For protein quantification, we considered the protein with at least two ratio counts generated from unique and razor peptides. The median value of the SILAC ratios was calculated as protein abundance (H/L ratio) to minimize the effect of outlier values. The peptides shared (not unique for leading proteins) between multiple leading proteins were assigned to one of them (the first one) as razor peptides (the detail information was summarized in supplementary Table S2). Finally, the global median normalization was applied to recalculate the protein abundance (the normalized protein ratio) to reduce the system error from sample preparation in each experiment.
Network Analysis of Protein Secretion Mechanisms and Functions-Differentially expressed proteins identified from the three experiments were uploaded and analyzed with the GeneGo pathway maps tool of MetaCore™, Version 6.5 build 27009 (GeneGo, St. Joseph, MI). The secretion mechanisms of proteins were analyzed using the SignalP 3.0 program with a hidden Markov model to predict the presence of secretory signal peptide sequences (34), signifying the classical secretory pathway. Additionally, the SecretomeP 2.0 program was employed to predict nonsignal peptide-triggered protein secretion (35), representing a nonclassical secretory pathway, and TMHMM 2.0 to predict transmembrane helices in proteins (36).
Immunoblot Analysis-Total cell lysates and conditioned media were prepared, and protein concentrations determined with the Bradford assay kit (Pierce Biotechnology, Rockford, IL). Equivalent amounts of proteins were fractioned on a 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel. Separated proteins were transferred to a nitrocellulose membrane (PH 7.9, Amersham Biosciences Inc., Piscataway, NJ), blocked with 5% nonfat powdered milk, incubated with a specific primary antibody at 4°C overnight, and hybridized with the respective secondary antibody, HRP-conjugated mouse/rabbit/goat anti-IgG, for 1 h at room temperature. Finally, immune complexes were visualized using the chemiluminescence method with an ECL detection kit (Amersham Biosciences) on Fuji x-ray films, as described previously (37).
Quantitative Reverse Transcription Polymerase Chain Reaction (Q-RT-PCR)-Total RNA was extracted from T 3 -treated HepG2-TR␣1 cells using TRIzol reagent, as described previously (38). Subsequently, cDNA was synthesized via RT-PCR with the SuperscriptIIkit (Invitrogen, Karlsruhe, Germany). Real-time Q-RT-PCR was per-formed on a 15 l reaction mixture containing 750 nM forward and reverse primers, varying amounts of template and 1 ϫ Sybr Green reaction mix (Applied Biosystems, Foster City, CA). Sybr Green fluorescence was determined with the ABI PRISM 7500 detection system (Applied Biosystems). Primers were designed using Primer Express Software (Applied Biosystems). Genes were normalized against the control ribosomal binding protein (RiboL35A) gene. The human PAI-1 oligonucleotides used in this study included the forward primer, 5Ј-GCACAACCCCACAGGAACA-3Ј, and reverse primer, 5Ј-GTC-CCAGATGAAGGCGTCTTT-3Ј.
Cloning and Activities of PAI-1 Promoter Fragments-Fragments of the PAI-1 promoter (positions Ϫ2261 to ϩ1203) were ligated into the pGL2 vector (Promega Corp., Madison, WI), based on the published sequence. Several serial deletion and mutation constructs of the PAI-1 promoter were amplified via PCR and cloned into the pA3TK vector. The constructed promoter sequences were confirmed using automatic DNA sequencing. HepG2-TR␣1 cells treated with 10 nM T 3 for 24 h were cotransfected with 0.6 g DNA/well of pA3TK vector containing the PAI-1 promoter sequence and 0.3 g of SV␤ plasmid, a ␤-galactosidase expression vector (Clontech, Palo Alto, CA), in 24-well plates using the TurboFect in transfection reagent (Fermentas, Glen Burnie, MD) to determine the transcriptional activities of TREs within the PAI-1 promoter. At the end of the treatment period, transfected and non-transfected cells were lysed, and luciferase and ␤-galactosidase activities measured. Luciferase activity was normalized against that of ␤-galactosidase, as described earlier (39).
Chromatin Immunoprecipitation (ChIP) Assay-ChIP assays were performed to examine the interactions between TR and TRE on the PAI-1 promoter (37). HepG2-TR␣1 cells treated with 10 nM T 3 for 24 h or left untreated were harvested and cross-linked with 1% formaldehyde for 10 min at room temperature in medium. The reactions were terminated by adding 0.125 M glycine. Subsequently, cell lysates were washed with PBS three times and resuspended in lysis buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris (pH 8.0), 0.1% SDS and 0.1% sodium deoxycholate) containing three protease inhibitors (1 mM phenylmethylsulfonyl fluoride, aprotinin, and leupeptin). Cell lysates were sonicated with a Misonix Sonicator 3000 Homogenizer (Mandel Scientific Company Inc., Guelph, ON, Canada) to disrupt chromatin. Sonicated DNA was between 200 and 1000 bp in length. The products were precleared with 60 l protein A/G agarose (Sigma Chemicals) for 2 h at 4°C. Complexes were immunoprecipitated with anti-TR (kindly provided by the laboratory of Dr. S-Y Cheng at the National Cancer Institute), anti-RXR␣ (Santa Cruz Biotechnology, Santa Cruz, CA) and anti-IgG antibody (R&D Systems, Inc., Minneapolis, MN). The 100 bp fragments of PAI-1 promoter containing the predicted TRE region were amplified via PCR with the forward primer, 5Ј-CCCAAGCTTC-AGTCAACCTGGCAGGACAT-3Ј, and reverse primer, 5Ј-CCGCTCG-AGGAACAATTGAGCAAACCCCAATA-3Ј.
Zymography Assay for Matrix Metallopeptidase MMP2 and MMP9 -Huh7-PAI-1 and Huh7-control cells (5 ϫ 10 6 ) were cultured in DMEM with 10% FBS. After 24 h of seeding, cells were washed twice with PBS and continuously incubated in serum-free medium for 24 h. Subsequently, conditioned media (CM) were collected and centrifuged at 1500 ϫ g to eliminate intact cells followed by concentration using a spin column with a molecular mass cut-off of 3 kDa (Amicon Ultra, Millipore). Concentrated conditioned media (50 g) were mixed with 50 mM Tris-HCl, pH 8.0, without reducing agent, and fractioned using 10% SDS-PAGE in the presence of 1 mg/ml gelatin. Following electrophoresis, the gel was washed with Zymogram Renaturing Buffer containing 2.5% Triton X-100 for 30 min twice at room temperature and incubated with Zymogram Developing Buffer (40 mM Tris-HCl, pH 8.0, 0.01% NaN3, 10 mM CaCl 2 ) at 37°C overnight. Gels were stained with Coomassie brilliant blue R-250 and destained in 5% methanol and 7.5% acetic acid solution until the appearance of clear bands.
Cloning of PAI-1-Total RNA (1 g) was reverse-transcribed using Superscript II reverse transcriptase (Invitrogen) and Oligo (dT) to synthesize template cDNA. PAI-1 cDNA was amplified via PCR with the forward primer, 5Ј-CCGGAATTCATGCAGA TGTCTCCAGCCCT-3Ј, and reverse primer, 5Ј-CCGCTCGAGTCAGGGTTC CATCACT-TGGC-3Ј, for 30 cycles at 95°C for 1 min, 60°C for 1 min and 72°C for 2 min. The PAI-1 open reading frame was ligated into pcDNA 3.0 expression vector, and the resulting construct sequenced. cm cell culture dishes using Lipofectamine Reagent (Invitrogen). After 24 h, transfected cells were transferred to medium containing G418 (400 g/ml) for selection until the generation of a single-cell clone. Expression of PAI-1 in Huh7 cells was confirmed by Western blotting.
In Vitro Migration and Invasion Assays-The influence of T 3 on PAI-1-mediated invasive activities of HepG2-TR␣1 and Huh7-PAI-1 cells was determined with a rapid in vitro assay (Transwell) (Falcon BD, Franklin Lakes, New Jersey), as described previously (40). Briefly, cell density was adjusted to 2 ϫ 10 5 cells/ml, and 200 l of this suspension seeded on either non-matrigel-coated (migration) or matrigel-coated (invasion) (Becton-Dickinson) upper chambers of the Transwell plate. For both assays, the pore size of the upper chamber was 8 mm. The medium in the upper chamber was serum-free DMEM, whereas the lower chamber contained DMEM supplemented with 20% FBS. After incubation for 24 h at 37°C, cells traversing the filter from the upper to lower chamber were examined via crystal violet staining and cell counting. Experiments were performed at least three times.
Animals-Male Sprague-Dawley (S.D.) rats underwent thyroidectomy (Tx) at 6 weeks of age in accordance with previously reported methods (41). After surgery, each rat was offered drinking water containing 1% calcium lactate. At 2 weeks following surgery, rats were injected peritoneally with T 3 at 10 g/100 g body weight or control vehicle (2.5 mM NaOH in PBS) for an additional 2 weeks. At the end of the experiment, rats were sacrificed, and the concentrations of T 3 and TSH in serum determined. The BALB/c nude mice (Jackson ImmunoResearch Laboratories, West Grove, PA) were injected subcutaneously with Huh7-PAI-1 or Huh7-control cells (each 4 ϫ 10 6 cells), J7-PAI-1 or J7-control cells (each 2 ϫ 10 6 cells) to determine the growth rate of the PAI-1-overexpressed cells. After 1 week of inoculation, the tumor xenografts were measured with two dimensions by caliper once per day. These nude mice were sacrificed at 4ϳ5 weeks after tumor inoculation. Tumor volume was calculated by the following equation: length ϫ height ϫ width. Another part, the severe combined immunodeficiency (SCID) mice were injected intravenously with J7-PAI-1 or J7-control cells (each 2 ϫ 10 6 cells) to examine the invasive ability of PAI-1. These SCID mice were sacrificed at 4 weeks after tumor inoculation, where upon the lung and liver were removed. Similar injections were performed in nude mice or SCID mice with various T 3 conditions (Group A to C) by using J7-TR cells. The SCID mice were divided into 3 groups. Group A (euthyroid) was a control with normal drinking water. Group B (hypothyroid) was treated with 0.02% methymazole and 0.1% sodium perchlorate in the drinking water to repress T 3 synthesis (42). Additionally, Group C (hyperthyroid) was added T 3 in the drinking water (2 mg/L) (Sigma Chem. Co., St. Louis, MO) (43). The sacrifice was performed after about 1 month injection, and the T 3 concentrations were determined.
Tumor volume was calculated using the following equation: length ϫ height ϫ width. All procedures were performed under sterile conditions in a laminar flow hood. Animal experiments were performed in accordance with United States National Institutes of Health guidelines and Chang-Gung Institutional Animal Care and Use Committee Guide for the Care and Use of Laboratory Animals.
In Vitro Neutralizing Assay-The influence of PAI-1 on T 3 -mediated migration of J7-TR␣1 cells was determined using a rapid in vitro assay (Transwell; Falcon BD). J7-TR␣1 cells (5 ϫ 10 4 ) were seeded into the non-matrigel-coated upper chamber of the transwell unit. This chamber contained serum-free DMEM whereas the lower chamber contained DMEM supplemented with 20% (v/v) FBS. J7-TR␣1 cells were pretreated for 1 h with either IgG (control) or an anti-PAI-1 monoclonal antibody (catalog no. 3783; American Diagnostica, Greenwich, CT) after exposure to T 3 (10 nM, 24 h) or not (0 nM). After incubation for 24 h at 37°C, cells traversing the filter from the upper to lower chamber were counted. All experiments were performed at least three times.
Statistical Analysis-Data are expressed as mean values Ϯ S.E. of at least three experiments. Statistical analysis was performed using the Student's t test and One-way ANOVA analysis. p Ͻ 0.05 was considered statistically significant.

RESULTS
Identification and Quantification of the T 3 -regulated Secretome Using SILAC-To investigate the TH-regulated secretome and its physiological significance, we utilized hepatocellular carcinoma (HCC) cell lines stably expressing high levels of wild-type TR␣1 as a model followed by SILAC-based quantitative proteomic strategy. A schematic diagram of the experimental design for exploring the T 3 -regulated secretome in HepG2-TR␣1 cells is shown in Fig. 1A. We collected SILAClabeled conditioned media from HepG2-TR␣1 cells treated with (T 3 ) or without T 3 (Td) for 24 h. Prior to LC-MS/MS analysis, light and heavy proteins were concentrated and examined using Coomassie blue stain (Fig. 1B) and Western blot (Fig. 1C). Western blot analysis clearly revealed the presence of ␤-tubulin in the SILAC-labeled total cell extracts but not conditioned media, indicating that cell death is not the underlying reason for the presence of a high proportion of proteins in conditioned media (Fig. 1C). Equal amounts of heavy and light proteins were mixed and analyzed using GeLC-or 2D LC-MS/MS. In total, 1742 and 1714 unique a Proteins were secreted via the classical secretory pathway using the SignalP software (SignalP probability Ն0.90). b Proteins predicted by the secretomeP program to be secreted via the nonclassical secretory pathway (SignlP probability Յ0.90 and SecretomeP score Ն0.50).
c Proteins predicted via the TMHMM to form trans-membrane proteins that were not predicted to be secreted via the classical and nonclassical secretory pathway. d Proteins predicted were not secreted via the classical pathway, nonclassical pathway and trans-membrane proteins. proteins were identified and quantified, respectively, in three independent experiments, the data from which are shown in Fig. 1D. Of these proteins, 1205 identified and 1176 quantified proteins were consistently present in at least two of the three tests (Fig. 1D). The details of peptide and protein identification and quantification are shown in supplementary Tables S1 and S2, respectively. To elucidate the potential secretion mechanisms of the identified proteins, different bioinformatics programs, including SignalP, SecretomeP, and TMHMM, were employed. The predicted secretion pathways of proteins identified in the three experiments (E1ϩE2ϩE3) are summarized in Table I. Among the 1742 proteins, SignalP predicted that 399 were released via the classical secretion pathway (SignalP probability Ն0.90), the SecretomeP program predicted that another 484 were secreted via the nonclassical secretion pathway (SignalP probability Ͻ0.90 and SecretomeP score Ն0.50), and the TMHMM program estimated that 26 others were not secreted via either the classical or non-classical secretion pathways. Collectively, our data suggest that 52.2% (909 of 1742) of the identified proteins are potentially released into extracellular space via different pathways.
Validation and Pathway Analysis of Protein Candidates-Using twofold change as the criterion for selection of T 3regulated candidates, 50, 83, and 232 up-regulated as well as 20, 41, and 59 down-regulated proteins in T 3 -treated cells were identified in the E1, E2, and E3 experiments, respectively (Fig. 1E). Among these, 61 and 11 proteins that were consis-tently up-regulated and down-regulated, respectively, in two (or more) of the three experiments were selected as T 3 -regulated candidates (Table II). Twenty-three of the 72 candidates (31.9%) have been previously determined as T 3 -associated proteins (13, 16, 37, 44 -59), and the remaining proteins detected for the first time in this study.
To explore the statistically significant biological networks regulated by T 3 , the 72 candidates were analyzed using the GeneGo pathway map tool of MetaCore™ (60). The analysis revealed five significant pathways (p Ͻ 0.0001) with at least three participating candidates (Table III), including blood coagulation, classical and lectin-induced complement pathways, cadherin-mediated cell adhesion, and the TGF-␤-dependent development network. Seven of 15 candidates involved in these pathways have previously been reported as T 3 -regulated target genes, specifically, ANT3, CADH1, CLUS, CO3, FIBG, FINC, and PAI-1 (Tables II and III).
Ten differentially expressed candidates (including 8 increases, 1 decrease and 1 no change) with available antibodies were selected for further Western blot analyses, with a view to validate the quantitative proteomics results. As shown in Fig. 2, all proteins were regulated by T 3 in a dose-dependent manner. Western blotting data were consistent with the MS-based quantitative results. The MS spectra of the representative peptides of proteins are displayed in Fig. 2. For example, fascin1 was not altered following T 3 treatment and thus taken as a reference for the other T 3 -regulated targets. Notably, the levels of fibrinogen (FIBG), fibronectin (FINC),  furin (FURIN), gelsolin (GELS), kininogen-1 (KNG1), plasminogen activator inhibitor-1 (PAI-1), spondin-2 (SPON2), and metalloproteinase inhibitor-2 (TIMP2) were increased, and cadherin-2 (CADH2) decreased in conditioned media after T 3 treatment. Based on the above data, we propose that T 3 is potentially involved in both blood coagulation and progression of the epithelial-mesenchymal transition via regulation of PAI-1 (Table III), which is known to participate in these biological pathways (see the REACT_604 data). Previously, we reported that T 3 plays an important role in blood coagulation by stimulating the expression of fibrinogen and a few blood clotting factors. PAI-1 inhibits the serine proteases tissue-type plasminogen activator (tPA) and urokinase (u)PA/urokinase, and hence inhibits fibrinolysis. Additionally, PAI-1 is particularly associated with the process of metastasis, poor prognosis and high mortality (26). Earlier, Biz et al. showed that thyroid hormones increase the level of PAI-1 mRNA expression in 3T3-L1 adipocytes (59). This finding links T 3 and PAI-1 expression. In view of these findings, PAI-1 was selected for further study.
To further determine the in vivo response of PAI-1 to T 3 , two groups of 6 week-old male S.D. rats (n ϭ 6 in each group) were surgically thyroidectomized. Subsequently, one of the thyroidectomized groups (Tx), used as the sham-operated control (sham), did not receive the T 3 injection, whereas the other group (TxϩT 3 ) was injected with T 3 daily for 2 weeks. After 2 weeks, rats were sacrificed, livers were removed and serum from each group were collected to examine the concentrations of T 3 and TSH, respectively (16). The T 3 -treated group (TxϩT 3 ) displayed enhanced PAI-1 mRNA and protein expression, compared with the Tx group (Fig. 3D). These measurement results support the regulation of PAI-1 expression by thyroid hormones.

FIG. 3. T 3 -regulated PAI-1 expression in HepG2 cells. A, Expression of TR was determined via Western blotting in cell extracts of three
HepG2-TR stable lines and HepG2-neo cells. HepG2 cells were transfected with TR␣1 or TR␤1, as described under "Experimental Procedures." The positions of 47 kDa TR␣1 and 55 kDa TR␤1 are indicated. TR bands are quantified in the right panel. PAI-1 expression was determined in the three HepG2-TR stable lines and HepG2-neo cells at 12-48h in the absence or presence of 1 and 10 nM T 3 using (B) Q-PCR and (C) Western blotting. D, PAI-1 expression was induced by T 3 and determined using Q-PCR and Western blotting in rat liver (sham, Tx, TxϩT 3 ), as described under "Experimental Procedures." Values are shown as fold induction of PAI-1mRNA relative to 0 nM T 3 at each time point or Tx. Differences were analyzed using the One-way ANOVA analysis, **p Ͻ 0.01; *p Ͻ 0.05. struct containing two putative TREs was activated about sixfold by T 3 in HepG2-TR␣1 cells. The two TREs in the -379/ -277 (p10) fragment were sequentially mutated to yield p13 and p14 constructs. After mutation of putative TRE (F2), luciferase activity of the p14 construct was still stimulated by about sixfold by T 3 in HepG2-TR␣1 cells. Conversely, after mutation of putative TRE (DR4), luciferase activity of the p13 construct was completely abolished (Fig. 4A, right). These data strongly suggest that T 3 regulates PAI-1 expression at the transcriptional level, and the putative TRE site exists between positions -379/ -277 (p10) encompassing a DR4-like sequence between positions -327ϳ-312 (AGGTCA AGGG AGGTTC).

TR and RXR Proteins Form a Complex with TRE (Ϫ327ϳ
Ϫ312) Located in the PAI-1 Promoter-To further determine whether PAI-1 TRE (DR4) is directly targeted by TR proteins, the ChIP assay was performed. TR proteins clearly associated with the TRE region of the PAI-1 promoter in vivo (Fig. 4B). TR␣1 and RXR␣ were clearly recruited to the TRE-binding site (Fig. 4B, lanes 12 and 13), whereas control IgG produced only background levels (lane 11). The positive control (human FURIN gene containing TRE) showed detectable bands with antibodies against TRs or RXR␣ (lanes 8 -9). However, no bands were detected using a primer set for the negative control (human GAPDH gene, lanes [3][4][5]. Data from the ChIP

FIG. 4. Regulation of PAI-1 expression by T 3 at the transcriptional level.
A, HepG2-TR␣1 cells were transfected with luciferase reporter plasmid driven by the PAI-1 5Ј-flanking region (positions -2261 to ϩ1203 containing three putative TRE sites) with or without a pA3TK-luc. Promoter activities were calculated, relative to 0 nM T 3 (ϩT 3 /-T 3 ), and further normalized to the pA3TK-luc control as well as ␤-galactosidase activities (T 3 -induced changes were normalized to that of ␤-gal). Columns, mean values obtained from at least three independent experiments performed in triplicate; bars, S.E. Cells were incubated for 24 h in the presence or absence of T 3 (10 nM) before harvesting to determine luciferase activity. Deletions and mutations in the PAI-1 5Ј-flanking region were also generated in pA3TK-luc vector, and the resulting constructs transfected into cells. B, ChIP assay demonstrating that TR is recruited to the PAI-1 5Ј-flanking region, together with RXR. Two sets of primers for PAI-1 TRE, positive control TRE (FURIN) and negative control (GAPDH) were prepared. ChIP assay data were evaluated with PCR and gel electrophoresis. Representative results are shown. assay thus demonstrate that the TR␣1 and RXR␣ complexes bind to the PAI-1 promoter.
PAI-1 is Associated with Cell Motility and Proliferation In Vitro or In Vivo-To determine the function of PAI-1, PAI-I overexpressing (Fig. 5A, Huh7-PAI-1 #1 and #2) or control cell lines (Fig. 5A, Huh7-control #1 and #2) were established. Notably, cell lines with overexpression of PAI-1 (Huh7-PAI-1 #1 and #2) displayed significantly increased (ϳtwo-to threefold) migration and invasion, compared with control cells (Huh7-control #1 and #2) (Figs. 5B, 5C). Images of cells migrating through the upper chamber stained with crystal violet are presented in Figs. 5B and 5C. Fig. 5D shows the proliferation rates of the two cell lines stably overexpressing PAI-1 (Huh7-PAI-1 #1 and #2), which were higher than those of the two control cell lines (Huh7-control #1 and #2). The expression level of PAI-1 in Huh7 cells FIG. 5. Functional assay of PAI-1 in hepatoma cells. The pcDNA3.0-PAI-1 or pcDNA3.0 construct was transfected into Huh7 cells to establish stable Huh7-PAI-1 or Huh7-control lines. Expression of PAI-1 was detected in PAI-1-overexpressing clones (#1, #2) and controls (control#1, #2) with Western blotting (A). B, Migration and (C) invasion abilities were analyzed through a Transwell assay in two PAI-1 over-expressing and two control cell lines. The stable lines (5 ϫ 10 4 ) were added to the upper chamber of Transwell units and incubated for 24 h. The number of cells traversing the filter to the lower chamber was determined and expressed as the total number of cells to provide an index of migration and invasion activity. Transwell filters were stained with crystal violet in the upper panel, and migration and invasion ability quantified in the lower panel. Values are shown as fold increase of Huh7-PAI relative to Huh7-control. Differences were analyzed using the One-way ANOVA analysis, **p Ͻ 0.01. (D) Proliferation ability was analyzed with the MTT assay, as described under "Experimental Procedures." Cell growth rates were determined based on absorption at 570 nm/650 nm up to 7 days. E, Nude mice were injected subcutaneously with Huh7-control (left) and Huh7-PAI-1 (right). Assays were performed 5 weeks after inoculation of tumor cells. The image showed two reprehensive tumor volumes in nude mice (total n ϭ 3) after 5 weeks. The statistics graph indicates that tumor size increased with time, both in control cells and those overexpressing PAI-1 up to 5 weeks. induced with 10 nM T 3 for 24 h was about 40% that of Huh7-PAI-1 cells (data not shown).
To investigate whether the effects of PAI-1 in vitro could be applied in vivo, nude mice were employed. We established a xenograft of stable Huh7-PAI-1 cells in BALB/c nude mice. Nude mice were subcutaneously injected with Huh7-PAI-1 or Huh7-control (Fig 5E), and tumor sizes measured from two to five weeks after injection. Tumor volume and weight are shown in Fig. 5E left (right: Huh7-PAI-1, left: Huh7-control). Additionally, tumor sizes from the two groups (Huh7-PAI-1 and Huh7-control) of mice are illustrated in Fig. 5E right. On average, tumors sizes of mice injected with Huh7-PAI-1 cells were two-to fivefold larger than those of control mice.
To investigate whether the in vitro effect of T 3 was evident in vivo, nude and SCID mice were injected with J7-TR␣1 cells. The animals were subjected to hyperthyroid, euthyroid, and hypothyroid conditions after injection. In the in vivo proliferation assay, tumors in hyperthyroid nude mice were larger (Fig.  6D, left) and heavier (Fig. 6D, right) than was the case with the other two groups (Fig. 6D). Further, the hyperthyroid group of SCID mice, injected with J7-TR␣1 cells, displayed more lung foci, a higher metastatic index (Fig. 6E), and a more elevated level of PAI-1 expression than did the other groups, as shown by hematoxylin and eosin staining (H&E) ( Fig. 6F; a, b, c) and immunohistochemistry (IHC) (Fig. 6F, d, e, f), respectively.
Previously, we showed that cathepsin H (CATH) expression was also directly regulated by T 3 in human hepatoma cell lines and that CATH enhanced the metastatic potential of hepatoma cells by increasing the activities of MMP-2 (61). Therefore, we sought to understand whether MMP2 or MMP9 was a downstream effector of PAI-1 as both MMPs share the same substrate. MMPs are zinc-and calcium-dependent endopeptidases involved in proteolytic processing (62). The activities of MMP2 and MMP9 were determined with the gelatin zymography assay. Both pro-MMP2 (72 kDa) and active MMP2 (67 kDa) activities were increased in PAI-1-overexpressing stable lines (Huh7-PAI-1 #1 and #2), but not in control cells (Huh7-control #1 and #2). In contrast, pro-MMP9 and Values are shown as fold increase relative to J7-control. Differences were analyzed using the Student's t test or One-way ANOVA analysis, *p Ͻ 0.05. active-MMP9 were not detected, even in PAI-1-overexpressing stable cells (Fig. 8A).
MMP2 activity and expression in response to T 3 treatment were assessed by Western blotting (Fig. 8B, upper panel) and zymography (Fig. 8B, lower panel). Such activity and expression were up-regulated by T 3 , in a dose-dependent manner, in HepG2-TR␣1 cells (Fig. 8). The influence of T 3 on MMP2 activity and expression was similar to that in cells overexpressing PAI-1 (Fig. 8B, lower panel versus Fig. 8A). In addition, we used an anti-PAI-1 antibody to explore whether PAI-1 was involved in T 3 -mediated regulation of MMP2 activity. We found that anti-PAI-1 antibody reduced (ϳ32-42%) the MMP2 expression (Fig. 8C, upper panel, lane 4 versus 2) and activity (Fig. 8C, lower panel, lane 4 versus 2) induced by T 3 .
To further explore whether PAI-1 was involved in T 3 -modulated cell migration, we performed an in vitro migration assay featuring addition of anti-PAI-1 (neutralizing) antibody at several concentrations after T 3 treatment. The migration ability gradually decreased in an anti-PAI-1 antibody concentrationdependent manner. Treatment with this antibody (20 g/ml) reduced migration of J7-TR␣1 cells by ϳ20 -30% compared with that of control cells treated with IgG (Fig. 8D), suggesting that PAI-1 plays at least some role in T 3 -mediated cell migration. Collectively, PAI-1 overexpression enhanced tumor growth and migration in a manner similar to what was seen when T 3 induced PAI-1 expression in J7-TR␣1 cells, both in vitro and in vivo.
The Urokinase Plasminogen Activator System is Mediated by the Thyroid Hormone-The uPA system is activated in many cancers (26). In addition to PAI-1, urokinase plasminogen activator surface receptor (uPAR) and brain-specific serine protease 4 (BSSP4), involved in the uPA system (63), quantified in one of the three SILAC experiments, displayed a significant increase (T 3 /Td Ͼ3) in T 3 -treated HepG2-TR␣1 cells (Table II). Protein and mRNA levels of uPAR and BSSP4 in conditioned media and extracts were validated using Western 9C, 9D and 9G, 9H to indicate the confidence of MS-based identification and quantification of these two proteins. DISCUSSION In this study, we have established the T 3 -mediated HepG2-TR␣1 secretome using a SILAC-based proteomics approach. Two protein/peptide separation techniques, GeLC and 2D-LC, were employed. Using the GeLC-MS/MS strategy, 635 proteins were identified, 56% of which were predicted using secretion pathway analyses. These results were comparative to the report of Wu et al. showing that 62.6% of the secreted proteins (485 of 775) identified in HepG2 cells could be predicted (31). In addition, up to 62% of secreted proteins coexisted in the two HepG2 secretome databases analyzed with the GeLC-MS/MS strategy. On the other hand, about 40% secreted proteins were only identified in one study, indicating that limitations of protein identification with the GeLC-MS/MS strategy exist for the HepG2 secretome. Therefore, LC-based peptide separation analysis, 2DLC-MS/MS, was performed. A comparable number of proteins were identified in two independent experiments (E2 and E3), specifically, 1508 proteins in E2 and 1350 in E3. Our results indicate that 2DLC-MS/MS is more effective than GeLC-MS/MS for the identification of proteins from the HepG2 secretome. In total, 1742 and 1714 protein in the HepG2 secretome were identified and quantified, respectively. To our knowledge, this is the largest HepG2 secretome database reported to date.
Recently, more reports have demonstrated the PAI-1 plays a crucial role in cancer cell survival and viability via alterations in the cell signaling pathway (70 -72). PAI-1 has been previously identified, by two dimensional-PAGE, in the secretome of human HepG2 cells (73). PAI-1, one of the T 3 targets, was extensively characterized to elucidate the molecular mechanism of its regulation by T 3 in isogenic HepG2 cell lines. We have shown that T 3 induces PAI-1 mRNA and protein expression in HepG2 and Huh7 cell lines expressing detectable endogenous TR proteins as well as in thyroidectomized rats. Further, it has been reported that thyroid hormones stimulate PAI-1 mRNA expression in 3T3-L1 adipocytes, but the effect was not observed in vivo at either the mRNA or protein level in adipose tissue of T 4 -treated rats (59). These findings suggest that adipocytes can respond in a diverse manner to thyroid hormones in vitro or in vivo. Further studies confirmed that T 3 up-regulates PAI-1 at the transcription level, and TR and RXR␣ complexes directly bind TRE between positions -327/ -312 of the PAI-1 gene 5Ј-flanking region. Additionally, cell lines overexpressing PAI-1 showed higher migration and proliferation abilities, both in vitro and in vivo. Importantly, PAI-1 may aid in accelerating cell migration by abrogating the interactions between vitronectin and uPAR as well as plasmin and uPA, suggestive of vital roles in modulating cell adhesion and motility (74). The PAI-1-mediated migration was parallel with the enhancement of MMP2 activity, which was consistently with the previous report that PAI-1 plays as a crucial role to modulate matrix remodeling in extracellular space (75,76). Collectively, these results suggest T 3 additionally promotes MMP activation, in parallel with PAI-1-mediated migration, by enhancing MMP2 activity. We speculate that PAI-1 is involved in T 3 /TR-mediated liver cancer progression.
Clearly, T 3 -mediated cell migration was significantly reduced upon addition of anti-PAI-1 antibody, but inhibition was not complete. T 3 exerts pleiotropic effects on cell migration and metastasis mediated by several genes including non- FIG. 10. Schematic presentation of the pathway of TH-mediated migration and urokinase plasminogen activator system. Addition of T 3 (10 nM) to HepG2-TR␣1 cells activated PAI-1 expression via direct binding of TR proteins to TRE (-327ϳ-312) of the PAI-1 promoter. MMP2 activity was altered in the presence of PAI-1 in Huh7 cells. PAI-1, uPAR and BSSP4 proteins, involved in the same pathway of the uPA system, were regulated by TH in SILAC-based experiments. metastatic 23 (NM23) (77), FURIN (16), pituitary tumor-transforming 1 (PTTG1) (17), CATH (61), and methionine adenosyltransferase 1 (MATA1) (38); PAI-1 is only one of the known T 3 -regulated targets. Additionally, MMP2 activity and expression were influenced by T 3 treatment in the present study; such an effect is therefore not unique to PAI-1. Extracellular PAI-1 activity, but not intracellular PAI-1 activity, can be abolished by neutralizing antibody. Therefore, T 3 -mediated cell migration was partially inhibited by such an antibody.
In addition to PAI-1, uPAR and BSSP4 involved in the uPA system were also identified and quantified as TH-mediated targets in the SILAC-based analysis (Table II). The uPA system is a serine protease family (including tPA, uPA, uPAR, and PAIs) that is activated in many cancers and are frequently associated with cancer cell progression and high mortality (26). Moreover, Yasuda et al. showed the BSSP4, a novel member of the uPA system, is a serine protease that catalyzes the activation of inactive uPA (63). The sequence of events following activation with T 3 in SILAC-based experiments is depicted in Fig. 10. Taken together, we have successfully identified TH-regulated target genes involved in the uPA system using the SILAC-based strategy. Several genes implicated in cancer cell progression are altered by T 3 , but the mechanism involved in TH-regulated secretion or signaling remains to be established. Our SILAC-based secretome dataset may be employed as a reference for determining the proteins involved in tumor progression and prognosis.