Potential conflict of interest: Nothing to report.
In HCC, inactivation of tumor suppressor genes plays a significant role in carcinogenesis. Apart from deletions and mutations, growing evidence has indicated that epigenetic alterations including aberrant promoter methylation and histone deacetylation are also implicated in inactivation of tumor suppressor genes. The goal of this study was to identify epigenetically silenced candidate tumor suppressor genes in human HCC by comparing the changes in oligonucleotide microarray gene expression profiles in HCC cell lines upon pharmacological treatment with the demethylating agent 5-Aza-2′-deoxycytidine (5-Aza-dC). By analyzing the gene expression profiles, we selected tissue factor pathway inhibitor-2 (TFPI-2), a Kunitz-type serine protease inhibitor, for validation and further characterization. Our results showed that TFPI-2 was frequently silenced in human HCC and HCC cell lines. TFPI-2 was significantly underexpressed in approximately 90% of primary HCCs when compared with their corresponding nontumorous livers. TFPI-2 promoter methylation was detected in 80% of HCC cell lines and 47% of human HCCs and was accompanied by reduced TFPI-2 messenger RNA expression. In addition, TFPI-2 expression in HCC cell lines can be robustly restored by combined treatment with 5-Aza-dC and histone deacetylase inhibitor trichostatin A. These findings indicate that TFPI-2 is frequently silenced in human HCC via epigenetic alterations, including promoter methylation and histone deacetylation. Moreover, ectopic overexpression of TFPI-2 significantly suppressed the proliferation and invasiveness of HCC cells. Conclusion: Our findings suggest that TFPI-2 is a candidate tumor suppressor gene in human HCC. (HEPATOLOGY 2007;45:1129–1138.)
Hepatocellular carcinoma is a major malignancy worldwide1 and is the second most common fatal cancer in Southeast Asia and Hong Kong. Although the risk factors are well established, the molecular mechanisms underlying the development and progression of HCC remain unclear. In addition to genetic alterations such as mutation and deletion, epigenetic alteration has been increasingly recognized as a key event for carcinogenesis. DNA methylation, the most well-characterized epigenetic modification, participates in regulating gene expression and is essential for normal development,2 X chromosome inactivation,3 and gene imprinting.4 Methylation of CpG islands on promoter regions is almost always associated with transcriptional silencing and is implicated in tumor suppressor gene inactivation in cancer cells.5
Conventionally, aberrant promoter hypermethylation has been studied in a candidate gene approach.6 Study of epigenetic silencing of genes on a genome-wide basis has been made feasible using microarray technology. This approach makes use of the pharmacological reversion of promoter methylation upon treatment with demethylating agents such as 5-aza-2′-deoxycytidine (5-Aza-dC). By comparing the gene expression profiles of cancer cell lines treated with vehicle and demethylating agent respectively, epigenetic alterations can then be mapped and characterized.7, 8 In this study, we used this high-throughput, microarray-based epigenetic gene expression profiling technique to screen for genes that are epigenetically silenced in HCC cell lines and sought to identify candidate tumor suppressor genes in HCC.
One of the candidate genes we identified with this approach was tissue factor pathway inhibitor-2 (TFPI-2), a newly identified Kunitz-type serine protease inhibitor. We confirmed that TFPI-2 was frequently underexpressed in human HCC cell lines as well as human HCCs through epigenetic alterations such as promoter hypermethylation and histone deacetylation. Moreover, ectopic overexpression of TFPI-2 significantly suppressed proliferation and invasiveness of HCC cells. Our findings strongly suggest that epigenetic silencing of TFPI-2 plays an important role in hepato-carcinogenesis.
HCC cell lines used in this study were obtained from the American Type Culture Collection (Manassas, VA) and the Shanghai Institute of Cell Biology (BEL7402 and SMMC-7721). Human HCCs and their corresponding nontumorous livers were collected at the time of surgical resection at Queen Mary Hospital, the University of Hong Kong. All specimens were obtained immediately after surgical resection, snap-frozen in liquid nitrogen, and kept at −70°C.
5-Aza-dC and Trichostatin A Treatment.
For epigenetic gene expression profiling analysis, SMMC-7721, BEL7402, and Hep3B cells were split to 1.6 × 105 cells per 10-cm culture dish 24 hours before treatment. Cells were then treated with 10 μM of 5-Aza-dC (Sigma, St. Louis, MO) or vehicle alone (as a control) for 96 hours. For reverse transcription PCR (RT-PCR), 2 × 105 cells were seeded onto 35-mm dishes. Cells were treated with 5-Aza-dC at indicated concentrations for 48 hours, and 0.5 μg/ml of trichostatin A (Sigma) was added to the cells during the last 24 hours of treatment. Drugs and culture medium were refreshed every day during treatment.
Gene expression profiling analysis was performed on Human Genome U133A array (Affymetrix, Santa Clara, CA) containing 18,400 transcripts and representing 14,500 genes. Total RNA was extracted from HCC cells with the RNeasy Mini Kit (Qiagen, Valencia, CA). Biotinylated complementary RNA probes were synthesized from 2 μg of total RNA and hybridized onto the oligonucleotide microarray according to the manufacturer's instructions. Gene expression data of each individual microarray were normalized and analyzed with GeneSpring 7.3 software (Silicon Genetics, San Carlos, CA).
Semiquantitative and Quantitative Real-Time RT-PCR.
Total RNA was extracted with Trizol reagent according to the manufacturer's instructions (Gibco, Grand Island, NY). Complementary DNA was synthesized from 1 μg of total RNA using the GeneAmp RNA PCR Kit (Applied Biosystems, Foster City, CA). The expression of TFPI-2 and GAPDH was detected via semiquantitative RT-PCR (primer sequences and PCR conditions are listed in Supplementary Table 1). Real-time RT-PCR was performed with an ABI Prism 7700 according to the manufacturer's instructions (Applied Biosystems). TFPI-2 expression was normalized against that of the housekeeping gene HPRT. Primers and TaqMan probes for TFPI-2 and HPRT were obtained from Applied Biosystems.
Immunobloting and Immunohistochemistry.
Total protein was harvested by direct lysing of cells with 1.5× SDS sample buffer. The expression of TFPI-2 protein was detected by rabbit polyclonal antibody (H-120) against TFPI-2 at a 1:200 dilution (Santa Cruz Biotechology, Santa Cruz, CA). Immunohistochemistry was performed on formalin-fixed, paraffin-embedded sections as described,9 using the same rabbit polyclonal antibody (H-120) at 1:150 dilution.
Bisulfite Sequencing and Methylation-Specific PCR.
Sodium bisulfite treatment was performed using the CpGenome DNA modification kit (Chemicon, Temecula, CA). Forty nanograms of bisulfite-treated DNA was amplified via PCR (Supplementary Table 1). For bisulfite sequence, PCR product was cloned into pGEM-T Easy vector (Promega, Madison, WI) and at least 5 individual clones were sequenced. Methylation-specific PCR was performed with methylation status-specific primer pairs, which were able to discriminate between methylated and unmethylated alleles of the TFPI-2 gene (Supplementary Table 1). For quality control, bisulfite-treated DNA obtained from normal blood and placenta was used as unmethylation control, and in vitro–methylated DNA (Chemicon) was used as a methylation control in every batch of methylation-specific PCR analysis.
Plasmid Construction and Transfection.
A 0.7-kb fragment of the full-length TFPI-2 coding sequence was amplified from normal liver tissue with primers 5′-GGG GTA CCG CTT TCT CGG ACG CCT TG-3′ (forward) and 5′-CGG GAT CCT GAT TTG TTT CCT CAT GCT GTC-3′ (reverse). PCR product was purified and cloned into the KpnI and BamHI site of pcDNA3.1/Hygro vector (Invitrogen, Carlsbad, CA). The DNA sequence of the recombinant plasmid was confirmed via DNA sequencing.
Colony Suppression Assay.
Two × 105 cells were seeded onto a 35-mm dish 1 day before transfection. Two micrograms of TFPI-2–pcDNA3.1/Hygro or empty vector was transfected into Hep3B using FuGENE 6 Transfection Reagent according to the manufacturer's instructions (Roche Molecular Biochemicals, Indianapolis, IN). After 24 hours, 50% of transfected cells were seeded onto 10-cm culture dishes and grown in culture medium containing 0.2 mg/ml hygromycin (Invitrogen) for 3 weeks. Hygromycin-resistant colonies were fixed with 3.7% formaldehyde and visualized via crystal violet staining.
Establishment of TFPI-2 Stably Expressing Cells.
Two micrograms of TFPI-2–pcDNA3.1/Hygro plasmid was transfected into SMMC-7721 cells. After 24 hours, the transfected cells were trypsinized and split onto 100-mm culture dishes at a density of 5 × 104 per dish. Transfected cells were selected for 3 weeks in culture medium containing 0.2 mg/ml hygromycin. Single hygromycin-resistant clones were isolated from the culture dish using a cloning cylinder obtained from Bellco Biotechnology (Vineland, NJ). Overexpression of TFPI-2 was confirmed via western blotting as described above.
In Vitro Cell Invasion Assay.
In vitro cell invasion assay was performed as described.10 Three × 105 cells were suspended in 300 μl of serum-free Dulbecco's modified Eagle medium and loaded onto the upper compartment of an invasion chamber that contained a polycarbonate membrane with an 8-μm pore size and was coated with a layer of extracellular matrix (ECM) (Chemicon). After 48 hours of incubation, the invasive cells that had migrated through the ECM layer to the complete medium in the lower compartment were stained, and the numbers of invaded cells were photographed and counted under the microscope.
Epigenetic Gene Expression Profiling.
In this study, we employed epigenetic gene expression profiling strategy to screen for methylation-silenced tumor suppressor genes in human HCC. Three human HCC cell lines (SMMC-7721, BEL7402, and Hep3B) were treated with 5-Aza-dC at a condition that achieved maximal re-expression of some known methylation-silenced genes11 without causing obvious cellular toxicity (Fig. 1A,B). The gene expression profiles were then analyzed with an oligonucleotide microarray (Affymetrix U133A) system. Among the 18,400 transcripts analyzed, 341 (1.9%) were found to be significantly up-regulated (≥2.5-fold) by 5-Aza-dC treatment in at least 1 of the HCC cell lines (Fig. 1C). The number of transcripts up-regulated in SMMC-7721, BEL7402, and Hep3B was 159, 81, and 175, respectively. Sixty-five transcripts were found to be up-regulated in more than one HCC cell line, including 9 that were commonly up-regulated in all 3 HCC cell lines (Fig. 1D). These 341 upregulated transcripts represented 266 unique genes. Subsequent in silico analysis revealed that CpG island was absent in the promoter region of 84 genes. These genes were therefore excluded from this study, because we considered that their up-regulation was probably due to secondary regulatory effects instead of direct demethylation of the promoter. On the other hand, it is well documented that DNA methylation plays an essential role in X chromosome inactivation and gene imprinting. For this reason, we further excluded 18 X chromosome genes and 7 well-characterized imprinted genes (Fig. 1E). After filtering genes using the above criteria, 157 candidate genes finally remained for further study. A partial list of the selected genes is shown in Table 1.
Table 1. Partial List of 5-Aza-dC–Induced Genes in HCC Cells
Probe Set ID
Glutathione S-transferase M4
Glutathione S-transferase M2 (muscle)
Glutathione S-transferase M1
Coagulation factor III (thromboplastin, tissue factor)
Growth arrest and DNA-damage–inducible, alpha
Cysteine-rich, angiogenic inducer, 61
Cytochrome P450, family 26, subfamily B, polypeptide 1
v-maf musculoaponeurotic fibrosarcoma oncogene homolog F (avian)
TFPI-2: an Epigenetically Silenced Gene in Human HCC.
Among these candidate genes, TFPI-2 was of particular interest because the epigenetic gene expression profiling revealed a remarkable increase (up to 5-fold) of TFPI-2 mRNA expression after 5-Aza-dC treatment in all SMMC-7721, BEL7402, and Hep3B cell lines. We therefore validated this finding with semiquantitative RT-PCR and western blotting. TFPI-2 mRNA was expressed in SMMC-7721, and a lower basal expression was detected in BEL7402; however, it was not expressed in Hep3B. Consistent with the microarray results, we observed a significant up-regulation of TFPI-2 upon 5-Aza-dC treatment in all 3 HCC cell lines (Fig. 2A-B). In addition, treatment with 5-Aza-dC induced TFPI-2 expression in a dose-dependent and time-dependent manner (Supplementary Fig. 1). These data thus validated the results obtained from epigenetic gene expression profiling analysis and suggested that TFPI-2 was epigenetically silenced in HCC cell lines.
TFPI-2 Promoter Methylation in HCC Cell Lines.
To determine whether epigenetic silencing of TFPI-2 was a common event in HCC cells, we extended our analysis to a panel of 10 HCC cell lines and used real-time RT-PCR for quantitative measurement. In addition, four normal liver tissue samples were included for comparison. Abundant TFPI-2 mRNA expression was found in four HCC cell lines, namely SMMC-7721, HLE, SNU182, and SNU449. On the other hand, TFPI-2 expression was markedly reduced in BEL7402, Huh-7, SNU475, and PLC/PRF/5. No TFPI-2 expression was detected in Hep3B and WRL. The TFPI-2 expression in HCC cell lines was apparently much lower than that of normal liver tissues, indicating frequent down-regulation of TFPI-2 in HCC cell lines (Fig. 2C). Furthermore, we observed a significant (more than 2-fold) increase of TFPI-2 expression after 5-Aza-dC treatment in 8 HCC cell lines, except SNU449 and PLC/PRF/5 (Fig. 2D). To substantiate the role of aberrant promoter hypermethylation in TFPI-2 silencing, we performed bisulfite sequencing and methylation-specific PCR to determine the methylation status of the TFPI-2 promoter. We found that TFPI-2 was completely methylated in Hep3B and WRL, which had no detectable TFPI-2 expression. On the other hand, only an unmethylated allele was detected in SNU449 and PLC/PRF/5, both of which had endogenous TFPI-2 expression and were unresponsive to the 5-Aza-dC treatment. For the remaining HCC cell lines, methylated and unmethylated alleles were both detected, indicating partial methylation of TFPI-2 promoter in these cell lines (Fig. 3A-B). The above findings were consistent with those from demethylation treatment and documented that TFPI-2 was frequently methylated in human HCC cell lines.
Frequent TFPI-2 Hypermethylation in Human HCC.
Next, we sought to determine whether TFPI-2 promoter methylation was also common in primary HCC and analyzed the TFPI-2 promoter methylation status in 34 pairs of HCC samples and 4 normal liver samples. We detected methylated alleles in 47% (16 of 34) of primary HCC samples (Fig. 3A-B). In contrast, no methylated allele could be detected in normal livers and the corresponding nontumorous liver samples, except for one patient who had TFPI-2 methylation detected in both HCC and its corresponding nontumorous but cirrhotic liver. Hence, these findings indicate that TFPI-2 methylation is preferentially found in human HCC.
Histone Deacetylation Contributes to TFPI-2 Inactivation in HCC Cell Lines.
We further examined whether histone deacetylation also contributed to TFPI-2 inactivation. HCC cell lines were treated with trichostain A (TSA), a histone deacetylase inhibitor. Our results indicated that TSA treatment was sufficient to restore TFPI-2 mRNA expression in all HCC cell lines by at least 2-fold (Fig. 4A). These findings suggest that histone deacetylation also contributes to TFPI-2 silencing in HCC cells. Recently, synergistic re-expression of silenced gene by combined demethylating agent and histone deacetylase inhibitor treatment has been reported in cancer cell lines.12 We therefore determined whether combined 5-Aza-dC/TSA treatment had a synergistic effect on TFPI-2 re-expression in HCC cells. SMMC-7712, BEL7402, and Hep3B cells were treated with 5-Aza-dC at a lower concentration (0.1 μM or 0.5 μM of 5-Aza-dC for 2 days), and 0.5 μg/ml TSA was added to the culture medium in the last 24 hours of the treatment. Using real-time RT-PCR, we found that lower concentration of 5-Aza-dC had little effect on TFPI-2 expression. However, pretreatment with 5-Aza-dC resulted in substantial induction of TSA-mediated TFPI-2 re-expression in HCC cell lines, even at subeffective concentrations. This synergistic effect of combined 5-Aza-dC/TSA treatment was particularly drastic in Hep3B, which has no endogenous TFPI-2 expression (Fig. 4B). This observation indicates that DNA methylation and histone deacetylation make a concerted effort to inactivate TFPI-2 in HCC cells.
Frequent Underexpression of TFPI-2 in Primary HCC.
We further investigated the TFPI-2 mRNA and protein expression levels in human HCC samples. Initially, we used semiquantitative RT-PCR to screen for TFPI-2 mRNA expression in primary HCCs. We found that TFPI-2 mRNA was frequently underexpressed in primary HCCs (89%, 16/18) when compared with their corresponding nontumorous livers (Fig. 5A). To quantify TFPI-2 expression level, we performed real-time RT-PCR on a total of 42 pairs of HCC and their corresponding nontumorous liver samples. Four normal liver samples were also included. We observed that TFPI-2 expression was significantly reduced in HCCs. The median TFPI-2 expression in HCCs was 7-fold lower than that of the nontumorous livers (P < 0.0001). In accordance with semiquantitative RT-PCR, 90% (38 of 42) of the patients showed significant (more than 2-fold) underexpression of TFPI-2 in the HCCs compared with their corresponding nontumorous livers (Fig. 5B). Immunohistochemistry showed underexpression of TFPI-2 protein in 10 of 13 HCCs (Fig. 6). Consistently, all 10 of these cases showed significant down-regulation of TFPI-2 mRNA as well. Our results clearly demonstrate that TFPI-2 was frequently underexpressed in human HCC and suggest that down-regulation of TFPI-2 may implicate it in human hepatocarcinogenesis.
TFPI-2 Suppresses Cell Proliferation and Colony Formation Ability in Human HCC Cell Lines.
Frequent epigenetic silencing of TFPI-2 in HCC cell lines and human HCC samples prompted us to further investigate the function of TFPI-2 in human HCC. To assess whether TFPI-2 might possess a tumor-suppressive function, we transiently expressed TFPI-2 complementary DNA in Hep3B. We found that the number of cells was significantly less in TFPI-2–expressing HCC cells than the empty vector control, indicating that TFPI-2 suppressed cell growth in HCC cells (Fig. 7A). This finding was further supported via colony suppression assay. Expression of TFPI-2 significantly suppressed the colony formation ability of HCC cells, as indicated by a significant reduction in both the number and size of colonies formed by TFPI-2–transected cells (Fig. 7B). These findings suggest that TFPI-2 suppresses cell proliferation of HCC cells and may function as a tumor suppressor in human HCC.
TFPI-2 Suppresses Cellular Invasion of Human HCC Cell Lines.
Next, we were interested to investigate whether overexpression of TFPI-2 has an inhibitory effect on HCC cell invasion. For this reason, we established a stable TFPI-2 expression model in SMMC-7721, which has higher cell motility and invasiveness. Stable TFPI-2 expression had no effect on cell motility, as evidenced on transwell assay in the absence of ECM coating (Supplementary Fig. 2). In contrast, stable expression of TFPI-2 significantly abolished the invasion of SMMC-7721 cells through the ECM layer (P < 0.0001) (Fig. 8). Our findings support the notion that TFPI-2 protects the EMC layer from enzymatic degradation and demonstrates that TFPI-2 suppressed cell invasion in human HCC cell lines.
In this study, we sought to identify tumor suppressor genes that are hypermethylated and silenced in HCC by comparing the gene expression profile of HCC cell lines (SMMC-7721, Bel7402, and Hep3B) with or without pharmacological treatment with the demethylating agent, 5-Aza-dC. A total of 341 transcripts were found to be significantly up-regulated by the drug in at least 1 of the 3 HCC cell lines. In silico analysis revealed that these 341 transcripts up-regulated by 5-Aza-dC represented 266 unique genes. We further excluded genes without CpG island on their promoter region, X chromosome genes, and imprinted genes from the study. Eventually, 157 candidate genes remained for further investigation. When we reviewed these genes from the literature, we found that at least 43 of them have been reported to be hypermethylated or possess tumor-suppressive function in various human cancers, including some well-characterized tumor suppressor genes such as E-cadherin, GSTP1, and TIMP3. These findings thus validate that epigenetic gene expression profiling is a powerful and efficient approach for identifying methylation-silenced tumor suppressor genes, which may be potentially implicated in hepato-carcinogenesis.
TFPI-2, also known as PP513 and MSPI,14 is a member of Kunitz-type serine protease inhibitors, which negatively regulate the enzymatic activity of trypsin, plasmin, and VIIa-tissue factor complex.14, 15 It has been hypothesized that inactivation of TFPI-2 was implicated in human carcinogenesis and metastasis. Promoter methylation and underexpression of TFPI-2 is commonly observed in human cancers.16–18 Nevertheless, promoter methylation of the TFPI-2 gene in HCC has not been investigated, and the function of TFPI-2 in human HCC is unclear. We found that upon pharmacological demethylation, the expression of TFPI-2 is upregulated in all three HCC cell lines, suggesting that TFPI-2 may be a common epigenetically silenced gene in human HCC. This finding prompted us to further investigate the roles of TFPI-2 in hepato-carcinogenesis.
In this study, we found that TFPI-2 was frequently underexpressed in HCC and that epigenetic alterations such as promoter methylation and histone deacetylation appeared to be a major underlying mechanism for TFPI-2 gene inactivation. Pharmacological demethylation successfully restored TFPI-2 mRNA expression in most (8 of 10) of the HCC cell lines, and TFPI-2 promoter methylation was validated in those same cell lines. Apart from promoter methylation, histone deacetylation, another common epigenetic alteration, also contributed to the transcriptional silencing of TFPI-2 in HCC cells. Indeed, TFPI-2 expression in HCC cell lines could be up-regulated up to 45-fold after 24-hour treatment with TSA. Although 5-Aza-dC and TSA alone were able to restore TFPI-2 expression to a certain extent, remarkable synergistic restoration was only achieved with combined 5-Aza-dC/TSA treatment. The synergistic effect of combined 5-Aaz-dC/TSA treatment appears to correlate with the methylation status of the TFPI-2 promoter. For example, more dramatic reactivation was found in Hep3B, in which TFPI-2 was completely methylated. Of the 2 partially methylated cell lines, BEL7402 and SMMC-7721, a more promising effect of combined 5-Aza-dC/TSA treatment was found in BEL7402, which had more extensive TFPI-2 promoter methylation than SMMC-7721 as revealed via methylation-specific PCR analysis. Therefore, our data suggest that DNA methylation plays a dominant role in TFPI-2 silencing.
Having demonstrated the epigenetic silencing of TFPI-2 in human HCC cell lines, we further extended our study to human HCC samples. Consistent with our findings in established HCC cell lines, aberrant promoter methylation on TFPI-2 was found in 47% of primary HCC and was accompanied with reduced TFPI-2 gene expression. Quantitative comparison revealed that TFPI-2 mRNA was down-regulated by approximately 7-fold in human HCCs compared with their corresponding nontumorous livers. In fact, 90% of the human HCCs showed a reduced expression of TFPI-2 in both mRNA and protein levels. Nevertheless, promoter hypermethylation may not be the sole causative factor for underexpression of TFPI-2 in human HCCs, because a significant portion of the HCCs had TFPI-2 underexpression even in the absence of promoter methylation. Because chromosomal or allelic deletion on 7q is uncommon in primary HCC, it is unlikely that TFPI-2 gene deletion is the cause for TFPI-2 underexpression in these HCCs.19 According to our findings in established HCC cell lines, histone deacetylation obviously participated in TFPI-2 inactivation, either on its own or by cooperating with DNA methylation. It is therefore reasonable to speculate that histone deacetylation may also contribute to TPFI-2 inactivation in human HCCs, as we observed in established HCC cell lines. Further investigations are therefore required to address this question.
TFPI-2 is a secretory protein predominantly found in ECM.20 It has recently been demonstrated that TFPI-2 counteracted ECM degradation through direct inhibition of plasmin activity or suppression of plasmin-mediated MMP-1 and MMP-3 activation.21, 22 Consistently, ectopic overexpression of TFPI-2 significantly suppressed cellular invasion in different human cancer types, including the lung,23 prostate,24 brain,25, 26 and pancreas.17 In addition, viral transfection of TFPI-2 cDNA into glioblastoma and laryngeal carcinoma cell lines successfully abolished its in vivo tumorigenicity in nude mice.26, 27 In accordance with previous studies, we demonstrated that expression of TFPI-2 significantly inhibited cell proliferation and invasiveness of HCC cells. These pieces of evidence taken together strongly imply that TFPI-2 is a putative tumor suppressor gene and that loss of TFPI-2 is important for hepatocarcinogenesis.
We thank the Genome Research Centre of the University of Hong Kong for collaboration and oligonucleotide microarray support. Special thanks go to Dr. William Mak for technical advice and support.