MicroRNA-29a induces loss of 5-hydroxymethylcytosine and promotes metastasis of hepatocellular carcinoma through a TET–SOCS1–MMP9 signaling axis

Ten eleven translocation (TET) enzymes convert 5-methylcytosine (5-mC) to 5-hydroxy-methylcytosine (5-hmC) and have crucial roles in biological and pathological processes by mediating DNA demethylation, however, the functional role of this epigenetic mark and the related enzymes in hepatocellular carcinoma (HCC) progression remains unknown. Here, we demonstrated that TET-family enzymes downregulation was one likely mechanism underlying 5-hmC loss in HCC. We found that miR-29a overexpression increased DNA methylation of suppressor of cytokine signaling 1 (SOCS1) promoter was associated with HCC metastasis in vitro and in vivo. Furthermore, miR-29a silenced anti-metastatic SOCS1 through direct TET-family targeting, resulting in SOCS1 promoter demethylation inhibition. Chromatin immunoprecipitation analyses confirmed that TET1 regulated SOCS1 expression through binding to the promoter region of SOCS1. Finally, miR-29a overexpression correlated with poor clinical outcomes and TET–SOCS1–matrix metalloproteinase (MMP) 9 axis silencing in HCC patients. In conclusion, our findings demonstrate that 5-hmC loss is an epigenetic hallmark of HCC, and miR-29a is an important epigenetic modifier, promoting HCC metastasis through TET–SOCS1–MMP9 axis silencing. The results offer a new strategy for epigenetic cancer therapy.

Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide and is among the leading causes of cancerrelated death, especially in China. 1 As HCC is often diagnosed at an advanced stage, a large proportion of HCC patients display symptoms of intrahepatic metastases or experience postsurgical recurrence, and have a 5-year survival rate of 30-40%. 2 Therefore, identifying novel HCC progression molecular markers and gaining insight into the molecular mechanisms responsible for metastasis and postsurgical recurrence will contribute to cancer prevention and increased life expectancy for HCC patients.
Abnormal DNA methylation at 5 position of cytosine (5-mC) is a well-known cancer epigenetic feature. Recent studies demonstrated that promoter hypermethylation and subsequent inactivation of tumor suppressor genes were involved in liver carcinogenesis. 3,4 Therefore, specific DNA methylation patterns alterations are hallmarks of tumors and could represent specific HCC treatment targets. 5 The recent discovery of ten eleven translocation (TET) family members that specifically modify DNA via 5-mC hydroxylation may explain how cells erase existing methylation marks. 6,7 Within the TET proteins family, TET1, 2, and 3 have been shown to convert 5-mC to 5-hydroxy-methylcytosine (5-hmC), which exists at a high level in self-renewing, pluripotent stem cells. 6,8 While, 5-hmC levels are profoundly reduced in many tumors types including breast cancer, liver cancer, and colon cancer. 9,10 These findings identified 5-hmC as a tumor development-associated biomarker. Furthermore, TET2 mutational inactivation and deletion have been associated with decreased 5-hmC levels in various hematopoietic malignancies. 11,12 Taken together, these studies suggest a crucial role for TET enzymes and 5-hmC in malignancy.
MicroRNAs (miRNAs) comprise an abundant class of small, highly conserved noncoding RNAs that bind the 3′-UTRs of protein-coding genes to suppress gene expression. 13 miRNAs can act as oncogenic promoters or tumor suppressors, depending on their target gene's function within a specific cell or tissue type. 14, 15 Song et al. 16 demonstrated that miR-22 exerts its metastatic potential by silencing anti-metastatic miR-200 through direct targeting of the TET protein family, thereby inhibiting miR-200 promoter demethylation. This finding highlights the link between miRNA, TET proteins, and DNA demethylation during tumor progression; however, the role of miRNAs in TET expression and active DNA demethylation in liver carcinogenesis and metastasis remains unclear.
Here, we report that 5-hmC loss serves as an epigenetic hallmark for HCC. Our findings indicate that 5-hmC, TET1, and TET3 significantly impact HCC progression. Importantly, we demonstrate that miR-29a promotes metastasis through SOCS1-expression repression and 5-hmC modulation by directly targeting TET-family members. Ultimately, these findings contribute to our current understanding of cancer epigenetics.

Results
Loss of 5-hmC correlates with HCC metastasis and prognosis. High 5-hmC levels were detected by immunofluorescence (IF) staining in the peritumoral liver tissues, and the 5-hmC signal was absent in corresponding tumor tissues (Supplementary Figure S1A). Using 5-hmC immunohistochemical (IHC) staining, we found that normal and peritumoral liver tissues showed strong nuclear 5-hmC staining, whereas virtually all HCC tumor cells exhibited partial or complete 5-hmC loss (Figures 1a-c). Next, we confirmed higher 5-hmC levels in normal and peritumoral tissues than in HCC tissues using anti-5-hmC dot blot assay (Figure 1d). Moreover, we examined 5-hmC levels by IHC staining a tissue microarray (TMA) comprised of 323 HCC paraffinembedded tissues. Consistent with the individual cases (Figures 1b and c), the TMA confirmed significant 5-hmC loss in tumor tissues compared with corresponding peritumoral liver tissues (Figure 1e and Supplementary Figure S1B). We then determined whether 5-hmC levels were correlated with critical HCC tumor, node, and metastases (TNM) stage. The 5-hmC staining score negatively correlated with TNM stage. Specifically, 5-hmC levels in TNM stage I HCC were significantly higher than in TNM stage II-III HCC (Supplementary Table S1).
5-hmC decrease is associated with substantial TET-family expression reduction in HCC. Next, we investigated cellular factors responsible for 5-hmC loss in HCC. While the TET family is directly responsible for 5-hmC generation, the co-factor α-ketoglutarate (α-KG) is absolutely required 5-mC to 5-hmC conversion, 6 α-KG is mainly produced by isocitrate dehydrogenases (IDHs). 17 To determine whether IDH and /or TET-family enzymes are responsible for 5-hmC loss in HCC, we examined IDH and TET-family genes expressions in HCC tumor tissues (cohort 1, n = 108) using quantitative real-time polymerase chain reaction (qRT-PCR). Strikingly, the expression of all three TET genes were significantly lower in tumor than in adjacent peritumoral liver tissues (all Po0.05, Figure 2a). IDH1 and IDH2 expressions were similar in HCC and peritumoral liver tissues (Supplementary Figure S1C). Furthermore, TET1 and TET3 expressions were decreased in tumor tissues from patients with intrahepatic tumor recurrence or extrahepatic metastasis compared with patients with no tumor recurrence (both Po0.05, Figure 2a). These findings were corroborated at the protein levels using IHC staining (Supplementary Figure S1D).
To further evaluate TET-family enzymes and 5-hmC's association with HCC metastasis, we analyzed TET enzymes and 5-hmC levels in a panel of human HCC cell lines with different metastatic potentials (Figures 2b-f). Higher TET1, 2, 3 and 5-hmC expressions were detected in low metastatic HCC cells lines (SMMC-7721 and HepG2), whereas lower TET1, 2, 3 and 5-hmC expressions were detected in highly metastatic HCC cell lines HCCLM3 and MHCC97H. Taken together, these data suggest that diminished TET-family expression represents a molecular mechanism underlying global 5-hmC loss in HCC.
miR-29a regulates 5-hmC level by directly targeting TET-family members in HCC cells. Recent studies demonstrated that miRNAs (e.g., miR-22), which may contribute to global epigenetic alterations, directly target TET proteins. 16,18 We utilized prediction algorithms, including TargetScan 6.0 (http://targetscan.org) 19 and miRBase (http://www.mirbase. org) 20 to search for miRNAs that potentially regulate TET enzymes in HCC and identified three miRNAs, miR-22, miR-26a, and miR-29a. As miR-29a ranks the highest target score among them, we focused on miR-29a. The miR-29 family have at least three putative target binding sites in TET1, 2 and 3 in the human, mouse genomes and zebrafish, suggesting strong conservation ( Figure 3a). Notably, miR-29a target sites are preferentially located near all ends of the TET family members 3′-UTR (Supplementary Figures S2A and B), implying effective targeting. Moreover, miR-29a coexpression effectively downregulated luciferase expression in the TET family members 3′-UTRs constructs. Seed sequence mutations of the predicted miR-29a binding sites within TET family members abolished miR-29a's inhibitory effects on luciferase expression (Figure 3b; Supplementary Figure S2A). Consistent with previous reports, 21 these results firmly confirmed a direct interaction between miR-29a and the 3′-UTR region of TET genes.
To further evaluate miR-29a's association with HCC metastasis, we analyzed miR-29a levels in a panel of human HCC cell lines with different metastatic potentials. 22 The miR-29a expressions in highly metastatic HCC cell lines HCCLM3 and MHCC97H were considerably higher than in low metastatic potential SMMC-7721 and HepG2 (Supplementary Figure S2C). Moreover, miR-29a upregulation in SMMC-7721 and HepG2 cells led to robust TET enzymes downregulation and global 5-hmC levels reduction (Figures 3c-e). In contrast, miRNA-29a induces loss of 5-hmC and promotes HCC metastasis Q Chen et al Furthermore, RNA interference (RNAi) downregulation of TET1, 2, and 3 in miR-29a-inhibited HCCLM3 cells reduced global 5-hmC levels ( Figure 3f). Taken together, these data suggest that 5-hmC epigenetic marker changes may contribute to miR-29a's impact on HCC progression and metastasis by targeting TET-family members.
The above data suggest that TET-family enzymes regulates SOCS1 expression in HCC cells. Therefore, we assessed the possibility of TETs-dependent regulation of SOCS1 mRNA transcription. Using ChIP-PCR analyses, our results indicated TET1 protein bind to the promoter of SOCS1 in HCC cells. We also found that TET1 can physically bind to multiple common binding sites (#6, #9, and #10) on the SOCS1 promoter in HCCLM3 and HepG2 cell lines (Figure 5f). These findings suggest that TET1 is a critical regulator of SOCS1 gene expression, and specifically, multiple binding sites are essential for TET1-induced SOCS1 transactivation.
In addition, recent studies indicated that miR-29a specifically targets both DNA methyltransferases (DNMT3A and DNMT3B) in cancer. 24,25 The findings of our current study showed that a significant DNMT1, DNMT3A, and DNMT3B protein expressions increased in highly invasive HCCLM3 and MHCC97H cells compared with SMMC-7721 and HepG2 cells (Supplementary Figure S6A). The protein levels of DNMT3A and DNMT3B in SMMC-7721 and HepG2 cells transfected with miR-29a vector, were significantly decreased compared with controls (Supplementary Figure S6B). Then, DNMT3A and DNMT3B knockdown in HCCLM3-anti-miR-29a cells, we observed that 5-hmC levels and HCC cells invasive behavior had no significant alteration. DNMT3B knockdown in HCCLM3-anti-miR-29a cells promoted HCC cells apoptosis (Supplementary Figures S6C-E). In addition, DNMT1, DNMT3A, and DNMT3B knockdown in HCCLM3-anti-miR-29a cells had no significant effect on SOCS1, p-ERK1/2, and MMP9 expressions, except that downregulation DNMT3B slightly decreased p-Stat3 and p-ERK1/2 expressions (Figure 5c). Overall, these findings revealed that miR-29a regulates 5-hmC levels in HCC by directly targeting TET proteins, rather than DNMT3A and DNMT3B. Importantly, these data confirmed that miR-29a suppresses SOCS1 gene promoter demethylation and represses its expression, leading to cognate targets upregulation to promote tumor growth and metastasis.
miR-29a overexpression correlates with poor clinical outcomes and TET-SOCS1-MMP9 axis silencing in HCC patients. Finally, we assessed the association between miR-29a overexpression and HCC patient clinical outcomes. We analyzed miR-29a expression by qRT-PCR in 54 tumor tissues and showed that miR-29a expressions were significantly upregulated in HCC tissues compared to peritumoral liver tissues (Po0.05, Figure 6a). Furthermore, miR-29a levels were increased in tumor tissues from patients with intrahepatic tumor recurrence or extrahepatic metastasis compared with recurrence-free patients (Po0.05, Figure 6a). Critically, HCC patients with high miR-29a expressions had poor survival rates as compared with low miR-29a expression patients (P = 0.0016, Figure 6b). These results indicate that significant miR-29a expression upregulation occurs in HCC and correlates with HCC relapse and metastasis.
Interestingly, we found that SOCS1 expression significantly decreased in tumor tissues compared with corresponding adjacent tumor tissues and correlated with early recurrence (Po0.05, Figure 6a). The HCC patients with high TET1 or TET3 expression levels harbored higher OS rates compared with low TET1 or TET3 expression patients (both Po0.01, Figure 6b). However, the high TET2 expression patients did not have significantly different OS rates than low TET2 expression patients (Figure 6b). More importantly, and consistent with our functional studies, miR-29a expression TET-family downregulation by miR-29a explains its effects on SOCS1 expression in HCC tissues. We found a clear correlation between TET1 and TET3 expression and SOCS1 in HCC patients (n = 108, r = 0.28 and 0.60, respectively, both Po0.01; Figure 6d). Taken together, these data support the notion that miR-29a is a crucial epigenetic modifier and promoter of HCC metastasis via TET family-SOCS1 pathway repression.

Discussion
HCC is one of the most lethal tumors, and few effective therapeutic options are available. Understanding hepatocarcinogenesis molecular mechanisms could provide strategies to arrest, retard, or even reverse the disease in HCC patients. 26 Several recent studies suggested that 5-hmC contents were reduced in solid tumors, including breast, pancreas, liver, and lung tumors. 9,10,27,28 In this study, we found that 5-hmC loss was a distinctive epigenetic event in HCC progression, and that it correlated with shorter OS and RFS, suggesting that 5-hmC loss could potentially be a molecular biomarker with predictive and prognostic value. Taken together, these data suggested that 5-hmC may represent an epigenetic marker for HCC identification and progression. Two mechanisms have been reported that dysfunction of IDH and/or TET enzymes, two key factors in 5-hmC generation, that would lead to greatly decreased 5-hmC levels in human tumors: loss-of-function mutations target TET2 and IDH1/2 mutations that result in TET activity inhibition via α-KG reduction and 2-hydroxyglutarate accumulation. 17,29,30 HCC harbors no IDH1 mutations, 31 recent studies found IDH2 protein decreased in HCC. 9 Herein, we demonstrated IDH1 and IDH2 had similar RNA expressions in both HCC and peritumoral liver tissues. Remarkably, a significant TET1, TET2, and TET3 gene expression decrease was detected in HCC compared with peritumoral liver tissues, suggesting that insufficient TET enzymes may be one mechanism underlying global 5-hmC loss in HCC.
miR-29a has previously been shown to have an effect on the development HCC but the findings have been contradictory. 24,32,33 Two studies supported miR-29 miRNAs directly regulate expression of TET protein in noncancer cells and HCC. 21,34 However, another report revealed that AFP expression transcriptionally regulated miR-29a and upregulated miR-29a suppresses the proliferation of HCC cells by targeting DNMT3A. 24 This observation, which contradicts our study results, may stem from the difference cell lines HLE cells are negative for AFP. While HCC cell lines used in our study all endogenously express AFP. Our findings indicated miR-29a expressions were positively correlated with the metastatic potential of HCC cells. Upregulation of miR-29a also predicted tumor metastasis and poor prognosis for HCC patients, suggesting a tumor-promoting role for this miRNA. We also revealed that miR-29a upregulation significantly reduced TET family members expressions in HCC cells. Conversely, miR-29a downregulation increased TET family members expressions. Moreover, miR-29a overexpression decreased luciferase reporter activity in a wild-type TET family 3′-UTR but not the mutant 3′-UTR. miR-29a upregulation and downregulation in HCC cells yielded distinct global 5-hmC expressions, and the loss of 5-hmC was associated with miR-29a overexpression in HCC cells. More importantly, TET protein inhibition triggers a proliferation and invasion phenotype similar to that induced by miR-29a overexpression in HCC cells. We validated the inverse correlation between miR-29a mRNA levels and TET1, 2, and 3 expression in HCC tissues. Overall, these data demonstrated that TET protein served as a major miR-29a target for HCC progression modulation.
SOCS1 silencing by CpG island hypermethylation in HCC has been demonstrated in several previous studies. 3,23 In addition, due to SOCS1's role as a negative JAK/STAT pathway regulator, it may also be an HCC tumor suppressor. 23 Our findings indicated that miR-29a suppressed SOCS1 mRNA expression by directly targeting TET enzymes, leading to SOCS1 promoter hypermethylation, SOCS1 suppression and STAT3 signaling activation. These effects resulted in MMP9 expression upregulation and carcinogenesis induction, increasing tumor invasiveness and metastasis ( Figure 6e). Importantly, we observed that TET1 protein bind to the promoter of SOCS1 in HCC cells by ChIP-PCR analyses, and three common binding sites of SOCS1 were found on the promoter in HCCLM3 and HepG2 cell lines. Indeed, aberrant SOCS1 CpG islands DNA methylation appears to be closely linked to their silencing during HCC metastasis, whereas epigenetic modifier drugs such as 5′-Aza can reactivate SOCS1 expression, indicating that epigenetic mechanisms have a functional role in controlling expression of this factor in cancers. 4,23 Therefore, miR-29a induces TET protein suppression leads to decreased expression of SOCS1 in HCC, this might be one mechanism for miR-29a promote HCC invasion and metastasis.
In addition, recent studies indicated that miR-29a specifically targets both DNMT3A and DNMT3B in HCC. 24,25 Silenced DNMT3A and DNMT3B protein expression using siRNA, we observed that 5-hmC levels and HCC cells invasive behavior had no significant alteration. Indeed, the absence of DNMT1 and DNMT3A had no significant effect on SOCS1, p-Stat3, p-ERK1/2, and MMP9 expressions compared with control vector, except slightly decreased p-Stat3 and p-ERK1-/2 in downregulation DNMT3B cells. In this study, our findings revealed that miR-29a regulates 5-hmC levels in HCC by directly targeting TET protein, rather than DNMT3A and DNMT3B. A recent report proposed miR-29a regulated the Yin-yang dynamics of DNA methylation in disease and development by targeting both TET family members and DNMTs. 35 These data suggest miR-29a dysregulation in cancer induces a phenotype of DNA methylation instability that could facilitate tumorigenesis. 25 These findings provide important insights into the molecular basis of tumor metastasis but also have important implications for cancer diagnosis, prognosis, and treatment. As miR-29a triggers a metastatic phenotype and its overexpression correlates with poor clinical outcomes in HCC patients, miR-29a might be a useful biomarker to identify metastatic forms of HCC. Our findings suggest that miR-29a inhibition via specific miR-29a inhibitors in highly metastatic cells reduces the metastatic phenotypes as well as elevating TET family members and SOCS1 expression provide a solid rationale for miRNA-29a induces loss of 5-hmC and promotes HCC metastasis Q Chen et al therapeutic miR-29a targeting to prevent HCC metastatic phenotypes.
In conclusion, our findings revealed that miR-29a decreases 5-hmC levels by negatively regulating TET-family expression and the subsequent epigenetic SOCS1 inactivation. Ultimately, SOCS1 dysfunction triggers MMP9 upregulation, which increases HCC cell proliferation, invasion, and metastasis. Therefore, miR-29a modulates HCC progression through the TET-SOCS1-MMP9 axis, which may present a useful clinical target for therapeutic interference of HCC progression.

Materials and Methods
Cell lines and animals. HEK293T, four human HCC cell lines with differing metastatic potentials (HepG2, SMMC-7721, MHCC97H, and HCCLM3), and one human non-transformed hepatic cell line L-02 were used in this study. HCC cell lines with stepwise metastatic potential (MHCC97H and HCCLM3, which have the same genetic background but different lung metastatic potentials) were established at our institute, 36,37 and authenticated by short tandem repeat validation analysis during the study period. The HCC cell lines HepG2 and SMMC-7721 and the normal liver cell line L-02 were purchased from the cell bank of Chinese Academy of Sciences (Shanghai, China). These cell lines were obtained within 6 months before being used in this study. All cell lines were maintained in high glucose Dulbecco's modified Eagle medium, supplemented with 10% FBS and penicillin/ streptomycin, at 37°C under 5% CO 2 in incubator.
Male BALB/c nu/nu mice (4-6 weeks old, Shanghai Institute of Material Medicine, Chinese Academy of Science) were raised in specific pathogen-free conditions. All procedures were performed in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals.
Patients and tissue samples. Two independent cohorts comprised of 431 HCC patients were enrolled in this study. The 108 tumor samples and matched non-tumor liver tissues (cohort 1, snap-frozen tissues) were consecutively collected from patients undergoing curative resection in 2007 between January and December at our institute. The other cohort (cohort 2, n = 323, paraffin-embedded tissues) was randomly collected from patients with HCC undergoing curative resection from 2003 to 2004. Ethical approval for the use of human subjects (No. 2009-002) was obtained from the Research Ethics Committee of Zhongshan Hospital (Shanghai, China), and informed consent was obtained from each patient. Detailed information and follow-up procedures are described in the Supplementary Materials and Methods.
Transduction of lentiviral vectors. Four HCC cell lines (HepG2, SMMC-7721, MHCC97H, and HCCLM3) were transfected using lentiviral vectors (miR-29a expression vector, the control vector for miR-29a, miR-29a inhibitor, and the negative control for the miR-29a inhibitor) according to the manufacture's instruction (GeneChem, Shanghai, China). The detailed protocols are shown in Supplementary Materials and Methods.
TMA and immunohistochemistry assay. TMAs were constructed as previously described. 38 Two core biopsies of 1 mm in diameter were taken from the donor blocks and transferred to the recipient paraffin block at defined array positions. The TMA was constructed from the 323 cases in cohort 2. IHC staining was performed by the avidin-biotin-peroxidase complex method. Briefly, after rehydration and microwave antigen retrieval, polyclonal antibodies against 5-hmC (1:500, Catalog No. 39769, Active Motif, Carlsbad, CA, USA) were applied to slides, incubated at 4°C overnight. Then slides were incubated with secondary antibody (GK500705, Gene Tech, Shanghai, China) at 37°C for 30 min. Staining was carried out with 3,3′-diaminobenzidine (DAB), and Mayer's hematoxylin was used for counterstaining. Negative control slides with the primary antibodies omitted were included in all assays. The scoring system are shown in Supplementary Materials and Methods.
RNA isolation and qRT-PCR. Total RNA was extracted from cell lines and frozen tumor specimens using the Trizol reagent (Invitrogen, Carlsbad, CA, USA). Complementary DNA synthesis was performed using a High Capacity cDNA Reverse Transcription kit or TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems, Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. The mRNA or miRNA expression in HCC cell lines, 108 HCC samples, and adjacent normal liver tissues were measured by RT-qPCR using an ABI7900HT instrument (Applied Biosystems). Levels of TET1, TET2, TET3, SOCS1, IDH1, IDH2, and miR-29a were normalized to GAPDH or U6. All experiments were performed in triplicate. The Taqman 2112 and 1973 probes were used for detection for miR-29a and U6, respectively (Applied Biosystems). The primers used in qRT-PCR assay are listed in Supplementary Table S2.
Luciferase reporter assay. HEK293T cells were seeded in a 96-well plate at 50-60% confluence. After 24 h, cells were transfected with 120 ng of miR-29a mimics or a negative control. Cells were cotransfected with 30 ng of modified pGL3 plasmids containing the wild-type 3′-UTR of TET1, 2, and 3 or a mutant 3′-UTR of TET1, 2, and 3 with all putative target sequences mutated. Transfections were performed using 0.45 μl of Fugene (Promega, Madison, WI, USA). Cells were collected 48 h after transfection, and Renilla luciferase activity was measured using a dual-luciferase reporter system (Promega). Luciferase reporter assays were performed in duplicate and repeated in three independent experiments. Luciferase activity was detected using an Orion II microplate luminometer (Berthold Technologies, Bad Wildbad, Stuttgart, Germany).
Chromatin immunoprecipitation analysis. Chromatin immunoprecipitation assay was performed as described previously. 40 Quantification of precipitated DNA by chromatin immunoprecipitation was performed using RT-PCR amplification. Experimental details are described in the Supplementary Materials. The primers used in the amplification are listed in Supplementary Table S3.
In vivo assays for tumor growth and metastasis. SMMC-7721-miR-29a, SMMC-7721-control, HCCLM3-anti-miR-29a, and HCCLM3-control cells (5 × 10 6 ) were suspended in 100 μl serum-free Dulbecco's modified Eagle medium and Matrigel (BD Biosciences, Franklin Lakes, NJ, USA) at a (1:1 ratio) and then injected subcutaneously into the upper left flank region of nude mice. Liver orthotopic transplantation model were also performed according standard procedures described previously. 41 At the end of study, mice were killed and tumor tissues were collected, photographed and the volume of tumors was calculated in mm 3 as follows: V = ab 2 /2 (with a and b representing the largest and smallest tumor diameters measured at necropsy, respectively). 42 The detailed protocols are shown in Supplementary Materials and Methods. miRNA-29a induces loss of 5-hmC and promotes HCC metastasis Q Chen et al