Supported in part by grants NSC 99-3112-B-016-003, NSC 99-2314-B-016-015-MY2, NSC 100-2320-B-016-011, and NSC 101-2320-B-016-011 from the National Science Council, Taiwan, Republic of China, and by the Liver Disease Prevention and Treatment Research Foundation, Taiwan, Republic of China.
Oncogenic activation of the Wnt/β-catenin signaling pathway is common in hepatocellular carcinoma (HCC). Our recent studies have demonstrated that SRY (sex determining region Y)-box 1 (SOX1) and secreted frizzled-related proteins are concomitantly promoter-hypermethylated, and this might lead to abnormal activation of the Wnt signaling pathway in HCC. SOX1 encodes a transcription factor involved in the regulation of embryonic development and cell fate determination. However, the expression and functional role of SOX1 in HCC remains unclear. In this study, we confirmed via quantitative methylation-specific polymerase chain reaction that SOX1 was frequently downregulated through promoter hypermethylation in HCC cells and tissues. Overexpression of SOX1 by a constitutive or inducible approach could suppress cell proliferation, colony formation, and invasion ability in HCC cell lines, as well as tumor growth in nonobese diabetic/severe combined immunodeficiency mice. Conversely, knockdown of SOX1 by withdrawal of doxycycline could partially restore cell proliferation and colony formation in HCC cells. We used a T cell factor (TCF)-responsive luciferase reporter assay and western blot analysis to prove that SOX1 could regulate TCF-responsive transcriptional activity and inhibit the expression of Wnt downstream genes. Furthermore, we used glutathione S-transferase pull-down, co-immunoprecipitation, and confocal microscopy to demonstrate that SOX1 could interact with β-catenin but not with the β-catenin/TCF complex. Moreover, restoration of the expression of SOX1 induces significant cellular senescence in Hep3B cells. Conclusion: Our data show that a developmental gene, SOX1, may function as a tumor suppressor by interfering with Wnt/β-catenin signaling in the development of HCC. (HEPATOLOGY 2012;56:2142–2153)
The incidence and mortality of hepatocellular carcinoma (HCC) have been increasing rapidly worldwide in recent decades.1 The risk factors associated with hepatocarcinogenesis are numerous and include chronic hepatitis B or C viral infection, alcohol, aflatoxin B1, and others. However, the molecular mechanisms involved in the development of HCC remain unclear. Recent studies have demonstrated that inactivation of tumor suppressor genes (TSGs) through promoter hypermethylation plays an essential role in carcinogenesis.2, 3 Furthermore, methylation profiles have been used as potential biomarkers for early diagnosis, prognostic prediction, and screening in HCC.4 Therefore, exploring the molecular mechanisms of the inactivation of TSGs involved in HCC development could improve the treatment of HCC in the future.
The Wnt signaling pathway is comprehensively involved in cell differentiation, embryonic patterning, proliferation, and adult homeostasis.5 Stabilized β-catenin through nuclear translocation forms a complex with T cell factor/lymphocyte-enhanced factor (TCF/LEF) and triggers the transcription of Wnt target genes such as c-MYC and cyclin D1.6 Abnormal activation of Wnt signaling stemming from mutations in β-catenin, adenomatous polyposis coli (APC), or axins or downregulation of APC occurs in various human cancers.7 Moreover, increasing evidence proposes that aberrant activation of Wnt/β-catenin signaling is involved in the initiation and progression of HCC.8 In addition to mutations in the components of Wnt signaling,9 promoter hypermethylation of Wnt-antagonists such as the secreted frizzled-related protein family, Dickkopf 3 and Wnt inhibitory factor-110, 11 also contribute to abnormal activation of this signaling in HCC.
SRY (sex determining region Y)-box (SOX) family proteins contain a highly conserved high-mobility group DNA binding domain12 and play a role during embryonic and postnatal development.13 Furthermore, SOX2, SOX7, SOX9, and SOX17 have been demonstrated to be tumor suppressors in different types of cancers.14-17 However, some studies have shown that SOX1, SOX2, SOX3, and SOX9 possess oncogene functions.18-21 Until now, the role of the SOX family in cancer development has been unclear. Structurally related to TCF/LEFs, several members of the SOX family have been implied to repress β-catenin activity by either stimulating degradation of β-catenin or by an unknown mechanism.22, 23 SOX1, a member of the SOX family of proteins,24 is evolutionarily conserved in many species and implicated as a key regulator of neural cell fate determination and differentiation.25, 26 Kan et al.26 proposed that SOX1 suppresses β-catenin-mediated TCF/LEF signaling by interacting with β-catenin to promote neurogenesis. These results suggest that SOX family member exertion of their functions through manipulation of Wnt signaling is a common tactic. In previous studies, we identified that SOX1 was hypermethylated in cervical and ovarian cancers.27, 28 Moreover, we recently demonstrated that SOX1 and secreted frizzled-related proteins were concomitantly hypermethylated in HCC tissues by QMS-PCR analysis (unpublished data). These results suggest that Wnt antagonists might be attenuated or shut down simultaneously during the progression of HCC. However, the expression and functional role of SOX1 in the development of HCC are not clear. In this study, our data demonstrated that SOX1 was frequently downregulated through promoter hypermethylation. Furthermore, ectopic expression of SOXl led to significant repression of HCC growth, which is mediated through interaction with β-catenin, thereby interfering with the Wnt signaling pathway. These results indicate SOX1 to be a novel tumor suppressor in hepatocarcinogenesis.
Materials and Methods
Cell Lines and Tissue Specimens.
Eight HCC cell lines (SK-Hep-1, HepG2, Hep3B, Huh6, Huh7, HA22T, TONG, and Mahlavu) were used in this study. Sixty paired HCC samples, including HCC tissues and DNA and RNA samples, were provided by the Taiwan Liver Cancer Network (TLCN). The TLCN is funded by the National Science Council to provide researchers in Taiwan with primary liver cancer tissues and their associated clinical information (Supporting Table 1). The use of the 60 HCC tissues, paired nontumor parts, and hepatic hemangioma tissues (as control livers) in this study was approved by our Institutional Review Board and the TLCN User Committee.
Bisulfite Conversion and Quantitative Methylation-Specific Polymerase Chain Reaction.
Bisulfite conversion and quantitative methylation-specific polymerase chain reaction (QMS-PCR) were performed as described.29, 30 The primer sequence for QMS-PCR has been described.30 All QMS-PCR data were obtained from at least three independent modifications of DNA to ensure reproducibility.
RNA Isolation, Reverse-Transcription PCR (RT-PCR), and Real-time Quantitative RT-PCR.
RNA isolation and RT-PCR were performed according to the manufacturer's protocol. Complementary DNA was amplified via PCR with primers specific for SOX1.27 Quantitative RT-PCR analysis was performed based on our previous report.29 Detailed information is given in the Supporting Information.
In Vivo Tumorigenicity.
HCC cells transfected with vector or SOX1 were injected subcutaneously into the left and right flanks of 6-week-old nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice. For the tet-on system, NOD/SCID mice were injected with Hep3B cells and randomly divided into two groups, with or without 0.2 μg/mL doxycycline (DOX) administration in 5% sucrose drinking water. The tumor volume was calculated as 0.5236L1(L2)2, where L1 is the long axis and L2 is the short axis of the tumor.31 The developing tumors were observed over the next 5 to 6 weeks, and the mice were then sacrificed at the end of follow-up. All animal studies were approved by the Institutional Animal and Committee at the National Defense Medical Center.
Details regarding generation of plasmid constructs, stably or inductively expressing SOX1 clones, cell proliferation, invasion, colony formation, glutathione S-transferase pull-down, co-immunoprecipitation, immunocytochemistry and senescence-associated β-galactosidase staining, and statistical analysis are provided in the Supporting Information.
Frequent Downregulation or Silencing ofSOX1 Is Associated With Promoter Hypermethylation in HCC Cell Lines and Primary Tumors.
First, we examined the messenger RNA (mRNA) and protein expression of SOX1 in eight HCC cell lines. SOX1 transcript and protein was undetectable in 100% of the HCC cell lines, but was expressed in normal liver tissue (Fig. 1A). We then checked the mRNA level of 60 primary HCCs and their corresponding adjacent nontumor tissues using quantitative RT-PCR and found that SOX1 mRNA expression was significantly downregulated in primary HCCs compared with the adjacent nontumor tissues (P < 0.01) (Fig. 1B). There was no significant correlation between SOX1 mRNA expression and clinical characteristics (Supporting Table 2). Based on our previous data, promoter hypermethylation of SOX1 might contribute to downregulation of SOX1 in HCC. Next, we checked the methylation status of the HCC cell lines and clinical HCC tissues by QMS-PCR. Hypermethylation was confirmed in the HCC cell lines (HepG2, Hep3B, Huh7, SK-Hep-1, HA22T, Mahlauv, and Tong) and HCC tissues, which showed downregulated or silenced SOX1 expression, whereas methylation was not found in the nontumor liver tissues (P < 0.01) (Fig. 1C,D). The methylation status in the SOX1 promoter region was then validated by bisulfite sequencing. The bisulfite sequencing results were consistent with QMS-PCR (data not shown). To validate whether promoter methylation is involved in the regulation of SOX1, three HCC cell lines (HepG2, Hep3B, and TONG) with silenced SOX1 expression were treated with 5-AZA-2′-deoxycytidine (5-Aza-CdR) combined with or without trichostatin A. The data showed the decreased methylation status of SOX1 and re-expression of SOX1 mRNA in all cell lines examined (Fig. 1E), further implying that the transcriptional silencing of SOX1 was mediated by promoter methylation and/or histone modification.
Ectopic Expression of SOX1 Represses HCC Cell Growth and Invasion.
To determine whether SOX1 has a tumor-suppressive ability, we checked the growth-suppressive effect through overexpression of SOX1 in HepG2, Huh7, SK-Hep-1, and HA22T, which showed no detectable SOX1 expression. Re-expression of SOX1 in stably expressed HepG2, Huh7, SK-Hep-1, and HA22T was confirmed by RT-PCR (data not shown) and western blot analysis (Fig. 2A). As shown in Fig. 2B and 2C, restoration of SOX1 significantly decreased HCC cell growth and colony formation in HepG2, Huh7, and SK-Hep-1 cells. Restoration of SOX1 in SK-Hep-1 and HA22T cells significantly suppressed the invasion ability (Fig. 2D). The representative photographs of anchorage-independent growth (AIG) and the invasion assay are shown in Supporting Fig. 1A and 1B.
SOX1 Inhibits Tumor formation in NOD/SCID Mice.
The subcutaneous tumor growth of HepG2 or Huh7 stably transfected with SOX1 or empty vector in NOD/SCID mice is shown in Fig. 3A. The tumor volume was significantly smaller in the SOX1-transfected NOD/SCID mice than in the vector control mice (P < 0.05). After 5-6 weeks, the tumors were taken out and weighed. The mean tumor weight was significantly lower in the SOX1-transfected NOD/SCID mice than in the vector control mice (P < 0.05) (Fig. 3B). The SOX1 expression levels in tumors from the SOX1-transfected and vector control groups were checked via western blot analysis (Supporting Fig. 2).
SOX1 Suppresses HCC Cell Growth and Invasion in an Inducible Expression System.
To further validate the tumor suppressor function of SOX1, we used an inducible system to manipulate SOX1 expression. SOX1 was induced by DOX in a dose and time-dependent manner (Supporting Fig. 3A,B). SOX1 can be stably induced by DOX in HepG2, Hep3B, and SK-Hep-1 cells (Fig. 4A). After induction of SOX1 by DOX, SOX1 inhibited cell growth in cell proliferation (MTS) assays (Fig. 4B) and AIG assays (Fig. 4C). Representative photographs of AIG are shown in Supporting Fig. 4A. The invasive ability in SOX1-inducible SK-Hep-1 cells was also significantly inhibited by SOX1 expression compared with parental control cells (Supporting Fig. 4B). These data are concordant with constitutively stable SOX1-transfected cell lines.
Knockdown of SOX1 by Withdrawal of DOX Partially Reverses Malignant Phenotypes.
To further demonstrate the antitumor function of SOX1, we manipulated the SOX1 expression using the tet-on system. First, Hep3B cells were treated with DOX for 7 days to induce SOX1 expression, and then knockdown of SOX1 expression was performed by withdrawing DOX for another 7 days. At the same time, another group of cells were only treated with DOX for 7 days. The SOX1 level in both groups was confirmed via western blot analysis (Supporting Fig. 5A). The detailed manipulation of SOX1 expression is shown in Fig. 4D, and MTS and AIG assays were performed on schedule. The results showed that SOX1 expression significantly suppressed cell growth compared with the control group, whereas knockdown of SOX1 expression partially increased the cell growth compared with SOX1-transfected cells according to the MTS assay (Fig. 4D). Moreover, knockdown of SOX1 expression can restore the malignant phenotype of HCC cells (Fig. 4D, Supporting Fig. 5B). We further investigated the antitumor growth of Hep3B with SOX1 expression by the tet-on system in NOD/SCID mice. After 10 weeks, tumors were taken out and weighed. The mean tumor weight was significantly lower in SOX1-expresssing (DOX+) NOD/SCID mice than in the control (DOX−) mice (P < 0.05) (Fig. 4E).
SOX1 Suppresses β-Catenin-Mediated TCF/LEF Signaling Through Interaction With β-catenin.
To further explore whether SOX1 could also interfere with the Wnt signaling pathway in HCC, we performed a Wnt/TCF-responsive luciferase reporter assay. The results showed that ectopic expression of SOX1 dramatically repressed the relative TCF transcriptional activity compared with control/vector cells (Fig. 5A). The suppressive Wnt/TCF signaling caused by SOX1 was not due to the difference in accumulated nuclear β-catenin (Fig. 5B). Previous studies26 have demonstrated that SOX1 binds to β-catenin in vitro, suggesting that SOX1-mediated repression in HCC cells may involve direct interaction with β-catenin. To test this hypothesis, we first overexpressed a glutathione S-transferase (GST)-SOX1 fusion protein and performed a GST pull-down assay. The results indicated that GST-SOX1 can pull down β-catenin in vitro (Supporting Fig. 6). Next, we overexpressed FLAG-SOX1 and mutant CTNNB1 (β-cateninΔ45)32 proteins in COS7 cells to perform immunoprecipitation. The data showed that SOX1 can interact with β-cateninΔ45 and vice versa and that FLAG-β-cateninΔ45 can immunoprecipitate SOX1 (Fig. 5C). We then used a co-immunoprecipitation assay to test whether the interaction between SOX1 and β-catenin exists in HCC cell lines. SOX1 can be coimmunoprecipitated with β-catenin in SOX1-expressing Hep3B cell extracts and vice versa (Fig. 5D). However, we did not detect the presence of TCF3/4 in the SOX1/β-catenin immunoprecipitation complex. These data suggest that SOX1 might compete with TCFs to interact with β-catenin. Moreover, we examined the cellular localizations of SOX1 and β-catenin using confocal microscopy. As shown in Fig. 5E, the merged images indicated that both SOX1 and β-catenin proteins were colocalized in the nuclei of both Hep3B and Huh7 cells. Taken together, these results demonstrate that SOX1 antagonizes canonical Wnt signaling through interaction with β-catenin.
SOX1 Suppresses the Target Genes of Wnt/β-Catenin in HCC Cell Lines.
To further explore the mechanism responsible for SOX1 functioning as a tumor suppressor, we analyzed the target genes of Wnt/β-catenin, c-MYC, and cyclin D1. As shown in Fig. 6A, Hep3B cells expressing SOX1 showed a noticeable decrease in both c-MYC and cyclin D1 mRNA. SOX1 expression significantly suppressed the c-MYC and cyclin D1 protein levels compared with the control group, whereas knockdown of SOX1 expression restored the c-MYC and cyclin D1 protein levels in Hep3B cells (Fig. 6B). However, the data showed that both the c-MYC mRNA and protein levels were significantly downregulated, but not the cyclin D1 mRNA and protein levels in HepG2 cells expressing SOX1 (Fig. 6A,B). These results indicate that the antitumorigenicity of SOX1 is mediated by transcriptional suppression of a downstream gene of Wnt/β-catenin.
Restoration of SOX1 Modulates the Expression of Molecules Involved in the Cell Cycle and Triggers Cellular Senescence.
To gain insight into the observed SOX1-suppressive cell growth, we evaluated the cell cycle progression by flow cytometry. Our data showed that ectopic expression of SOX1 increased number of cells in G1 while decreasing number of cells in S phase in Hep3B cells (Fig. 6C), but no significant difference was seen in HepG2 cells (data not shown). To further determine the mechanism associated with growth inhibition by SOX1, the molecules involved in the cell cycle were checked using western blot analysis. In Hep3B cells, SOX1 expression significantly enhanced the protein level of p21 and p27 but suppressed expression of CDK4 and CDK6 compared with the control cells. In the SOX1-expressing HepG2 cells, p21 and p27 were also dramatically upregulated. However, there was no significant difference in the protein levels of CDK4 and CDK6 (Fig. 6D). Moreover, SOX1 overexpression did not significantly affect the active forms of caspase-9 and caspase-3 in Hep3B and HepG2 cells (data not shown). Interestingly, we found that expression of SOX1 in Hep3B cells could enhance the signal of SA-β-gal staining, and these data implied that SOX1 could trigger cellular senescence in Hep3B cells (Fig. 6E).
SOX1 Affects the Invasion Phenotype by Regulating Invasion-related Genes.
Overexpression of SOX1 in SK-Hep-1 cells, known as non-β-catenin nuclear accumulation cells, caused a suppression of invasion ability. To elucidate the mechanism, we analyzed the expression of the invasion-related genes CDH1 and SLUG after ectopic expression of SOX1. As shown in Fig. 6F, overexpression of SOX1 could enhance CDH1 expression but repressed SLUG expression in SK-Hep-1 cells.
Over the past 10 years, the SOX family has been proven to regulate the Wnt/β-catenin activity in diverse development and cancer contexts.33 Since the first report of regulation of the canonical Wnt signaling pathway for SOX17 and SOX3 in Xenopus embryos,34 a growing number of SOX proteins have been revealed to interact with β-catenin and TCFs, and several mechanisms have been proposed. In colon cancer cells, SOX7 and SOX17 act through binding to β-catenin and promote its degradation function as tumor suppressors.15, 23, 35 Experiments in a murine osteoblast cell line (OB1) suggest that Sox2 might inhibit osteoblast differentiation by physically interacting with β-catenin and suppressing Wnt target genes.36 There are only a few studies on the SOX1 gene,18, 26 but no study has analyzed the relationship between SOX1 and Wnt/β-catenin signaling in HCC. In this study, we demonstrated that SOX1 inhibited the canonical Wnt signaling in HCC cells and competed with TCFs to bind to β-catenin without affecting the level of nuclear β-catenin accumulation. Interestingly, we also found that SOX1 may suppress HCC invasion through a β-catenin-independent pathway by upregulation of CDH1 and downregulation of SLUG. Taken together, these results demonstrate that SOX1 functions as a tumor suppressor gene in HCC through Wnt pathways.
Indeed, we used HCC cell lines with mutant (HepG2) or wild-type CTNNB1 (Huh7 or Hep3B), and there is a trend toward stronger antiproliferation effects of SOX1 in cell lines with a wild-type CTNNB1 (Fig. 2B,C; Fig. 4B, and Fig. 6). However, in luciferase reporter and xenograft data, it seems that SOX1 could antagonize the Wnt pathway independent of the CTNNB1 mutation. Furthermore, luciferase reporter analysis of mutant SOX1 (with a C terminus truncated region) indicated that they failed to suppress the β-catenin/TCF-dependent transcriptional activity (Supporting Fig. 7). Our data showed that the high-mobility group domain (but not the C terminus) is essential for SOX1 to suppress β-catenin-mediated TCF/LEF signaling. Kan et al.26 showed that SOX1 could bind to β-catenin, and the C terminus of SOX1 is required for this interaction. Transcriptional regulators of SOX proteins generally require the cooperation of partner factors for the regulation of specific target genes in a cell type-specific fashion.37, 38 Although an authentic partner protein associated with SOX1 was not identified, the possible explanation for the conflicting results may result from the putative partner protein influences on the interaction of SOX1 and β-catenin in different cell contexts. Moreover, Mathews et al.18 found that SOX1 promoted invasion of prostate cancers through interaction with STAT3, increasing the IL-6/STAT3 pathway activity. They did not investigate the relationship between SOX1 and Wnt signaling. It has been reported that SOX proteins can play either a tumor suppressor or an oncogenic role owing to variations in the genetic background, signaling network, and cellular context. The controversial results may arise from the property of SOX proteins as transcription factors.
Moreover, we demonstrated that decreased protein levels of c-MYC and cyclin D1 and increased protein levels of p21 and p27 were associated with overexpression of SOX1 in Hep3B cells. In addition, deprivation of SOX1 expression restored the expression levels of both c-MYC and cyclin D1. These results suggest that SOX1 inhibited Wnt signaling and then decreased β-catenin/TCF downstream genes. It has been reported that c-MYC may repress p21 expression through different mechanisms.39, 40 Moreover, van de Wetering et al.41 found that the decreased expression of c-MYC releases p21 (CIP1/WAF1) transcription after disruption of β-catenin/TCF-4 activity, which in turn mediates G1 arrest and differentiation. This master switch mediated by the β-catenin/TCF complex controls proliferation versus differentiation in healthy and malignant intestinal epithelial cells. From our present data, we also found that restoration of SOX1 decreased c-MYC but increased p21 expression. Whether decreased c-MYC can release p21 or whether SOX1 directly regulates the p21 expression still needs further investigation. Furthermore, SOX2 interacts with β-catenin in osteoblasts and inhibits the Wnt-responsive reporter assay in HEK293 cells,36 and plays important roles in growth inhibition through interfering with Wnt signals by downregulation of cyclin D1 and upregulation of p27kip1 level in gastric cancers.14 In accordance with our data, SOX1, like SOX2, belongs to group B of the SOX family, and may affect the cell cycle progression by way of interfering with Wnt signaling in hepatocarcinoma cells. Only one report has demonstrated that SOX6 suppresses cyclin D1 promoter activities by physically interacting with both β-catenin and histone deacetylase 1 in pancreatic beta cells.42 However, how SOX1 reduces the c-MYC and cyclin D1 expression while interacting with β-catenin requires further investigation.
Cell senescence, a state of irreversible arrest of cell proliferation in response to stress, is considered to play critical roles in cancer and aging.43 It has been reported that the key effectors of cellular senescence are regarded as cyclin-dependent kinase inhibitors p16INK4a, p21Cip1, and p27Kip1.44 However, Ye et al.45 reported that downregulation of Wnt signaling triggers cell senescence in primary fibroblasts and epithelial cells, offering an additional mechanism by which Wnt signaling can regulate not only proliferation, differentiation, and apoptosis but also cellular senescence. C-Myc utilizes a variety of mechanisms, including regulation of p16 and p21, to attain modulation of cell senescence.46, 47 In the current study, we surveyed the senescence status triggered by SOX1 in Hep3B, HepG2, and SK-Hep-1 cells and found that only Hep3B cells expressing SOX1 showed significant cellular senescence. Decreased c-MYC and increased p21 expressions were observed in SOX1 overexpressed Hep3B cells. This result was consistent with the notion mentioned above. Nevertheless, why did cellular senescence occur just in Hep3B cells, and not in the other cell lines we tested? We propose that this may be due to SOX1 functioning as a tumor suppressor through a distinct mechanism based on the different genetic backgrounds of HCC cell lines, such as Hep3B being known as a p53 depleted cell line.
In conclusion, SOX1 is frequently downregulated by epigenetic mechanisms in HCC, which may lead to aberrant activation of Wnt/β-catenin signaling. Restoration of SOX1 repressed β-catenin/TCF-responsive transcriptional activity by interacting with β-catenin and restraining the expression of downstream genes. These findings suggest that SOX1 might function as an important tumor suppressor during the development of HCC.
We are grateful to Yu-Ching Chou, School of Pubic Health, National Defense Medical Center, Taipei, Taiwan, ROC, for assistance with statistical analysis. We thank the Taiwan Liver Cancer Network for providing the HCC tissue samples and related clinical data (all are anonymous) for our research work. This network currently includes five major medical centers (National Taiwan University Hospital, Chang-Gung Memorial Hospital-Linko, Veteran General Hospital-Taichung, Chang-Gung Memorial Hospital-Kaohsiung, and Veteran General Hospital-Kaohsiung). TLCN is supported by grants from National Science Council since 2005 till now (NSC 100-2325-B-182-006) and National Health Research Institutes, Taiwan.