Introduction
Hepatocellular carcinoma (HCC), the most common primary liver tumor, is the fifth most common human malignancy and the second leading cause of cancer death (695,900 deaths annually) in men worldwide (1). Owing to difficulty in early diagnosis and a lack of effective chemotherapy and radiotherapy, the 5-year overall survival rate remains poor for patients with advanced HCC (2–4). Changes in growth factors, receptors and their downstream signaling pathway components influence HCC development and malignancy. The signaling pathways involved include IGF-IGF1R, hepatocyte growth factor (HGF/MET), EGF-EGFR and TGFβ-TβR signaling (5). Ligands bind to their corresponding receptors, which are usually overexpressed and/or dysregulated in HCC, and mainly activate the PI3K-PKB (AKT) and MAPK pathways, which affect tumor cell phenotypes (5–8). IGF secreted by liver cells acts in an autocrine and/or paracrine manner by binding to IGF1R (9). This in turn initiates downstream cellular effectors including insulin-receptor substrates and Src homology domain-containing kinases, ultimately contributing to proliferation, anti-apoptosis, invasiveness and metastasis (5,10–13). Growth factors that are dysregulated in tumorigenesis and cancer progression represent rational targets for HCC therapies. Nevertheless, the precise oncogenic mechanisms by which growth factors affect the transcriptional regulation of certain tumor suppressor genes (TSGs) are unclear.
Epigenetic mechanisms play an important role in driving tumorigenesis. DNA methylation catalyzed by DNA methyltransferases (DNMTs), causes changes in chromatin structure, DNA conformation and stability and patterns of DNA-protein interaction. DNA hypermethylation leads to transcriptional silencing of a considerable number of TSGs; thus, hypermethylation has been proposed as the second hit in a model of TSG inactivation (14). Overexpression of DNMTs has been observed in various types of cancers (15–20), and is significantly correlated with poor histological differentiation and prognosis. These findings suggest that aberrant DNA hypermethylation drives hepatocarcinogenesis. It is recognized that many risk factors for HCC may upregulate DNMTs expression, such as HBV-encoded protein X (21). The tobacco-specific carcinogen NNK (nicotine-derived nitrosamine ketone) induces DNMT1 protein upregulation and binding to the promoters of various TSGs, which ultimately leads to tumorigenesis and poor prognosis in lung cancer (22). These findings shed light on the relevance of the expression of DNMTs in carcinogenesis. However, there is little evidence on the delicate mechanism of DNMTs overexpression in HCC. The findings presented here highlight the mechanistic significance of aberrant DNMT1 expression induced by IGF1 in HCC.
Materials and methods
Cell culture
HCC cell lines (HepG2, BEL-7402 and SMMC-7721) were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). The HL-7702 immortalized normal human liver cell line was obtained from the China Center for Type Culture Collection (Wuhan, China). Cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) or RPMI-1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS; HyClone) and 1% penicillin-streptomycin.
Quantitative RT-PCR (qRT-PCR)
qRT-PCR was performed as previously described (23) using an ABI PRISM 7300 detection system. qRT-PCR reactions were performed using SYBR-Green (Roche) technology and were repeated at least three times.
Flow cytometry and cell cycle synchronization assays
HepG2 cells were seeded in DMEM (0.1% FBS) medium at a density of 5×105cells/35-mm dish. The cells were treated with DMSO or 15 μM 5-aza-2-deoxycytidine (5′-Aza; Sigma-Aldrich) in time-course experiments. Cell suspensions in PBS were prepared for analysis using a FlowCellect Bivariate Cell Cycle kit (Millipore) according to the manufacturer’s protocol. At various time-points after release, cells were pulsed with 10 μM BrdU for 2 h immediately before harvesting and fixation. Cell pellets were processed for double staining with an anti-BrdU antibody (Millipore) and propidium iodide (PI, 10 μg/ml; Millipore), followed by analysis on a flow cytometer (Partec CyFlow Space). Gating was performed to focus on G1, S and G2/M populations.
In vivo tumorigenicity assays and 5′-Aza treatment
Female athymic nude mice (Vital River Animal Center, Beijing, China) were housed in a temperature-controlled sterile room in which humidity and light were carefully controlled and monitored. Animal welfare and experimental procedures were in strict accordance with the regulations approved by the Institutional Animal Care Committee of the Medical College at Xiamen University. HepG2 cells (5×106) were injected subcutaneously into the flank of 6-week-old nude mice. When the tumor size reached 0.5 cm3, the mice were randomly divided into two groups and subjected to the following treatments: i) i.p. injection of PBS as a control; ii) 1 mg/kg 5′-Aza was given by i.p. injection every 2 days over a treatment time of 14 days (seven injections). Tumor volume was calculated as volume = width2 × length × 0.52 mm3.
ChIP-on-chip and target gene verification
The protocol provided by NimbleGen was followed. In brief, HepG2 cells were grown, cross-linked with formaldehyde and sheared by sonication. Antibodies (DNMT1, DNMT3B and IgG; Santa Cruz Biotechnology) were used for immunoprecipitation. A HG18 RefSeq Promoter microarray (NimbleGen) was used for hybridization and analysis according to the manufacturer’s instructions. Data extraction was performed using NimbleScan version 2.5 and significant hybridization peaks (FDR <0.05) were identified. For target gene verification we used real-time PCR and a chromatin immunoprecipitation (ChIP) assay.
ChIP and DNA methylation assay
The two experiments were performed as previously described with slight modification (24,25).
Statistical analysis
SPSS 15.0 software (SPSS, Inc., Chicago, IL, USA) was used for all statistical analyses. Data from cell growth and mouse tumorigenesis experiments were analyzed by one-way ANOVA. To analyze differences in DNMT1 protein levels among various cell treatments, a two-tailed Student’s t-test was used. Survival curves were calculated by the Kaplan-Meier method, and comparison was performed using a log-rank test. Results for parametric variables are expressed as mean ± SE. In all cases, P<0.05 was considered to indicate a statistically significant result.
Results
Identification of target genes of DNMT1 and DNMT3B
To screen for genomic DNA methylation profiles and potential target genes in HCC, we carried out ChIP-on-chip coupled with CpG island microarray (MCAM) analysis in HepG2 cells. Extensive genome-wide DNA methylation of CpG islands of target genes was determined, with high co-localization of DNMT1 and DNMT3B; results for chromosome 1 are shown in Fig. 1A. We identified 2,177 genes with promoter hypermethylation in HepG2 cells, and 3,117 and 1,999 genes affected by DNMT1 and DNMT3B, respectively. Of these, 520 and 280 genes shared a methylation chip with DNMT1-ChIP or DNMT3B-ChIP, respectively. Among all three groups, 275 genes were screened out (Fig. 1B). Consistent with the above results, unsupervised hierarchical clustering analysis demonstrated that the top 32 genes with promoter hypermethylation in ChIP-on-chip results are closely related to tumor development and progression (Fig. 1C). However, we also found that DNA methylation of some gene sites does not rely on DNMT1 or DNMT3B (data no shown), implying the potential DNA methylation role of other DNMTs in HCC, such as DNMT3A; this will be the subject of further investigation.
Knockdown of DNMTs inhibit HCC malignant phenotype
We focused on the functional effect of DNMTs in controlling HCC malignant phenotype. MTT assay results show that transient inhibition of either DNMT1 or DNMT3B by siRNA was sufficient to reduce HepG2 cell proliferation (Fig. 2A and B). Strikingly, both siRNAs cooperatively suppressed HepG2 cell proliferation after 6 days and had a more pronounced effect than either siRNA alone (Fig. 2B). Time-course cell cycle analysis revealed that 5′-Aza inhibited HepG2 cell cycle progress from G1 phase to S phase (Fig. 2C). Moreover, 5′-Aza treatment inhibited HepG2 cell migration (Fig. 2D; P=0.003). To delineate the role of DNA methylation in HCC growth in vivo, we performed xenograft experiments in which HepG2 cells were injected subcutaneously into immunodeficient nude mice. The tumor growth rate significantly decreased after treatment with 5′-Aza compared to control animals. The average tumor volume was 1.96 cm3 in the control group on day 28 compared to only 1.23 cm3 in the group treated with 5′-Aza (Fig. 2E; P=0.001). These results argue that DNMTs cooperatively mediate DNA methylation and broadly participate in controlling the malignant phenotype of HCC cells.
DNMT1 protein expression is upregulated by growth factors
Epigenetic regulation is sensitive to the environment, including extracellular environmental factors. We hypothesized that DNA methylation is influenced by growth factor pathways in HCC cells. To dissect the potential relationship between HCC-related growth factors and DNMT1 expression, we determined the effect of IGFs, HGF and EGF on DNMT1 expression in HCC cells. The results from western blotting clearly show that DNMT1 protein expression was upregulated by IGF1 stimulation in a dose-dependent manner in HepG2 cells (Fig. 3A). Time course experiments revealed that 50 ng/ml IGF1 increased DNMT1 expression in HepG2, BEL-7402 and HL-7702 cells (Fig. 3B). Furthermore, HGF and EGF increased DNMT1 protein expression (data no shown). These results point to a novel pathway of HCC-related growth factors that act by upregulating DNMT1, and suggest that DNMT1 has an important biological function in regulating HCC malignant phenotype.
We investigated the involvement of the IGF-IGF1Rβ pathway in the regulation of DNMT1 expression. The PI3K-AKT pathway is a key downstream signal of the IGF-IGF1Rβ activation. To confirm whether this pathway is responsible for DNMT1 upregulation by IGF1, experiments to clarify which signaling pathway was blocked were performed. No obvious inhibition of DNMT1 expression was observed after treatment with 10 μM of LY-294002 (Fig. 3C, lane 3). However, 10 and 50 μM of LY-294002 significantly suppressed IGF1-induced DNMT1 expression, with downregulation of pAKT and pGSK3 (Fig. 3C, lanes 4 and 6). Notably, blocking of the PI3K-AKT pathway by LY-294002 led to a marked increase in IGF1-induced pIGF1Rβ phosphorylation (not necessarily activation) in a dose-dependent manner (Fig. 3C, lanes 2, 4 and 6), indicating the pivotal role of the PI3K-AKT pathway in IGF-IGF1Rβ activation signal transduction in the HepG2 cells. Furthermore, we found that PI3K-AKT phosphorylation was effectively stimulated by IGF2, HGF and EGF (data no shown). Taken together, the results indicate the transmission effects of the PI3K-AKT pathway on IGF-IGF1Rβ-induced DNMT1 protein expression.
IGF1 inhibits ubiquitin-proteasome-mediated degradation of DNMT1 protein by GSK3β-βTrCP complexes
Next, we focused on how the PI3K-AKT pathway affects IGF1-IGF1Rβ axis-induced DNMT1 expression. IGF1 and HGF did not affect DNMT1 mRNA levels (data no shown), so it is likely that its effect is on regulating the protein, including protein stability. It has been reported that AKT phosphorylates and inactivates GSK3β, which recruits βTrCP to degrade substrate protein through a ubiquitin-proteasome pathway and also stabilize DNMT1 and maintains DNA methylation (26,27). To explore this possibility, HepG2 cells were treated with either IGF1 or the GSK3β inhibitor LiCl. Similar to IGF1, 20 M LiCl upregulated DNMT1 expression and GSK3β phosphorylation in a manner independent of IGF1Rβ activation (Fig. 4A). pGSK3β (Ser9), an inactive form, gradually increased with DNMT1 expression in HepG2 cells treated with SB-216763, a GSK3β inhibitor, for 4 h or longer (Fig. 4B). MG132, a proteasome-specific inhibitor, also increased DNMT1 protein levels in HepG2 cells (Fig. 4C). Fig. 4D shows that the synthesis inhibitor cycloheximide (CHX) decreased DNMT1 protein levels in a time-dependent manner; this decrease was rescued by IGF1 in HepG2 cells. These results indicate that IGF1-induced DNMT1 protein upregulation involves the GSK3β-mediated protein degradation pathway.
Next, HA-ubiquitin or DNMT1 plasmids were transiently transfected into HepG2 cells and co-immunoprecipitation (co-IP) was performed with either anti-HA or anti-DNMT1 antibodies, respectively (Fig. 4E and F). As expected, results from western blotting showed that ubiquitin protein interacts with DNMT1, βTrCP and GSK3β (Fig. 4E, lane 5), and this interaction was abrogated by treatment with IGF1 (Fig. 4E, lane 6). The reduced interaction between DNMT1 and the ubiquitin-βTrCP-GSK3β complex induced by IGF1 treatment was confirmed by reverse co-IP with an anti-DNMT1 antibody (Fig. 4F, lanes 5 and 6). Immunofluorescence (IF) staining indicated that the active form of GSK3β co-localized with DNMT1 in the nucleus, which was abrogated by IGF1 treatment, resulting in DNMT1 upregulation (Fig. 4G). Consistent with this observation, Fig. 4H shows that siRNA-mediated IGF1R knockdown (KD) substantially blocked IGF1-induced pI3K-AKT pathway activation and DNMT1 expression. Finally, we evaluated the relationship between pAKT activation and DNMT1 expression in clinical HCC samples. Results from western blotting revealed that DNMT1 expression was markedly increased in 59.4% (19/32) samples compared with adjacent normal tissue. pAKT was noticeably increased in 71.9% (23/32) of HCC samples (Fig. 4I). Among the samples, 16 cases showed upregulation of both DNMT1 and pAKT. These results indicate that IGF-IGF1Rβ activation upregulates DNMT1 expression via an AKT-GSK3β-mediated ubiquitinproteasome pathway in HCC.
IGF1 suppresses TSGs transcription via DNA hypermethylation
Finally, we focused on several potential TSGs (DLC1, CHD5, IGFBP2, ADAM23, LRCC4, LASS2, SCAI, HOXB13, AXIN1 and CSRNP1) commonly methylated in HCC, and performed real-time qPCR to detect their expression in HepG2 cells treated with 5′-Aza. For HepG2 cells exposed to 5′-Aza for 7 days, DLC1, CHD5, IGFBP2, ADAM23 and LRCC4 were markedly upregulated (Fig. 5A). We also found that either DNMT1 or DNMT3B siRNA obviously upregulated DLC1 expression, and siRNAs for both DNMT1 and DNMT3B together dramatically increased DLC1 mRNA levels (Fig. 5B). ChIP assays showed that DLC1 and CHD5 promoters were bound by DNMT1 and DNMT3B, and this binding was abrogated by 5′-Aza (Fig. 5C and D). Bisulfate sequencing assays showed that CpG loci of DLC1 are heavily methylated in wild-type HepG2 cells and the DNA methylation level was obviously reduced by 7 days of treatment with 5′-Aza (Fig. 5E). These results demonstrate that DNMT1 and DNMT3B cooperate in methylation of the DLC1 and CHD5 promoters.
As expected, the mRNA expression of certain DNMTs target TSGs (DLC1, CHD5, IGFBP2, ADAM23, LRRC4 and LASS2) was significantly downregulated in HepG2 cells exposed to IGF1 for 7 days (Fig. 5F). Consistent with this observation, binding of DNMT1 to either DLC1 or CHD5 CpG loci was enhanced in HepG2 cells treated with IGF1 for 7 days (Fig. 5G). Fig. 5H shows that DNA methylation levels for the DLC1 promoter were increased by 7 days of treatment with IGF1. These results clearly show that IGF1 suppresses TSG transcription, at least in part, through DNA methylation mechanisms.
Discussion
Emerging evidence suggests a relationship between DNA hypermethylation and hepatocarcinogenesis (28). However, the mechanistic and biological significance of DNA hypermethylation in human HCC is unclear. DNMTs promote liver cancer cell proliferation by regulating the expression and activity of cell cycle proteins. Moreover, siRNA interference of DNMT1 or/and DNMT3B significantly inhibits liver cancer cell proliferation, and 5′-Aza can inhibit HepG2 cell migration and the growth of established tumors in mice. Overall, DNA methylation is a potential target for effective treatment of HCC. The HCC control potential of chemotherapy is limited by the tolerance of hepatoma cells to drugs. Kurita et al (29) reported that double knockdown of DNMT1 and DNMT3B sensitizes HCC cells to TRAIL-mediated apoptosis, indicating the potential importance of aberrant DNA hypermethylation in the chemotherapy tolerance of HCC. This seems to be related to silencing of certain TSGs by DNA hypermethylation, which controls resistance to drugs in the liver. The function of 5′-Aza in eliminating methylation is explicit, but its clinical use as an anti-neoplastic agent is limited by its toxicity. However, demethylation could be a novel supplementary therapy that might enhance the curative effect of conventional chemotherapy in HCC; this will be an important issue for clinical investigation.
It is now thought that relapse and metastasis severely limit rehabilitation and HCC prognosis after traditional treatments such as surgery and radio/chemotherapy. It has been confirmed that tumor-derived growth factors are associated with recurrence following radical hepatectomy in HCC. Our results reveal a previously unknown epigenetic mechanism: as one component of the tumor microenvironment, growth factors can promote HCC malignant phenotype through DNMT1 upregulation, resulting in nuclear DNMT1 accumulation and hypermethylation of TSG promoters. The mechanism and significance of DNMT1 overexpression induced by growth factors remain poorly understood. We identified a novel growth factor pathway that involves de novo DNA methylation and tumor suppressor silencing via upregulation of DNMT1 protein expression. The significance of IGF-IGF1R in hepatocarcinogenesis has been confirmed (30) and involves multiple stages and genes, as well as numerous signaling pathways, such as MAPK and PI3K/AKT (5). The present study provides direct evidence that activation of the IGF1-IGF1Rβ axis induces DNMT1 protein accumulation in the nucleus, primarily through PI3K-AKT signaling. Co-IP assays revealed that DNMT1 interacts with βTrCP and GSK3β in HepG2 cells, which involves the ubiquitin-mediated protein degradation pathway. IGF1 protects DNMT1 stability through abrogation of the DNMT1-βTrCP-GSK3β complex. These findings are in agreement with the notion that DNMT1 protein ubiquitination and proteasome-mediated degradation depend on a destruction domain mapped to the N-terminal 120 amino acids that is also important for βTrCP interaction (22,31). In clinical samples, high correlation between PI3K-AKT activation and DNMT1 upregulation was observed. This confirms the crucial role of pI3K-AKT in IGF1-induced DNMT1 upregulation. This is similar to the report that NNK activates AKT, inhibiting GSK3β-βTrCP complex formation and thereby attenuating proteasomal degradation of DNMT1 in lung cancer cells (22). We investigated whether a similar mechanism applies to other growth factors and IGF1. The results show that IGF2, EGF and HGF activate AKT phosphorylation. It suggests that protein stability may be the common mechanism in growth factor-induced DNMT1 upregulation.
Although some studies have found increased expression of DNMTs in HCC, the mechanistic significance is unknown at present. Our studies answer this question by identifying a novel mechanism in which DNMT1 overexpression depends on extracellular cytokines, similar to the mechanism for NNK regulation of DNMT1 expression via the ubiquitin protein degradation pathway (22). These findings further confirm that epigenetic modifications such as DNA methylation play an indispensable role in regulating the expression of target genes and controlling cell proliferation, differentiation and functional specialization in hepatocarcinogenesis. We used a promoter array to identify genome-wide methylation profiles in HCC cells and found significant hypermethylation of CpG sites for a series of TSGs (DLC1, CHD5, IGFBP2, LRRC4 and LASS2) with notable biological significance in hepatocarcinogenesis. A putative tumor suppressor gene, DLC1, was identified and mapped to chromosome 8p21.3-22, a region suspected to harbor tumor suppressor genes and recurrently deleted in HCC, prostate, lung and breast cancers. The present study provides evidence that DNMT1 and DNMT3B are directly co-located on DLC1 CpG sites, resulting in promoter hypermethylation. Significant downregulation of DLC1 mRNA expression in HCC tissues was associated with enhanced CpG hypermethylation. DNA methylation is another crucial mechanism for DLC1 inactivation besides genetic regulation. Mapping to 1p36, chromodomain helicase DNA binding protein 5 (CHD5) is a tumor suppressor that serves as a master regulator of a tumor suppressive network controlling proliferation, apoptosis and senescence. The present study provides direct evidence that DNMT1 and DNMT3B bind to the CpG loci of DLC1 and CHD5, leading to promoter hypermethylation and transcriptional silencing. IGF1 significantly suppressed mRNA expression of certain TSGs, accompanied by increased accumulation of DNMT1 at DLC1 and CHD5 CpG loci in HepG2 cells. These results clearly show that IGF1 suppresses TSGs transcription via DNA methylation, at least in part.
In conclusion, our results point to a novel epigenetic mechanism for growth factor-mediated repression of TSG transcription that involves DNA methylation (Fig. 6). IGF1 induces activation of AKT, then promotes GSK3β phosphorylation at Ser9 to form inactive GSK3β, which subsequently attenuates the ability of βTrCP to degrade DNMT1 protein. DNMT1 protein binds to the promoters of various TSGs and results in promoter hypermethylation and transcriptional silencing, which ultimately leads to tumorigenesis and poor prognosis. Our results not only enriched our understanding of a novel epigenetic mechanism of growth factor in the regulation of TSG transcription by DNA methylation but also provided a novel target for HCC diagnosis and prognosis.