Hypermethylation-mediated Transcriptional Silencing of lncRNA-SCARF1 Promotes Progression and Metastasis of Hepatocellular Carcinoma

Boyi Liao Fudan University, Fudan University Peiran Huang Fudan University, Fudan University Xiangyu Zhang Fudan University, Fudan University Xinyu Wang Fudan University, Fudan University Kaiqian Zhou Fudan University, Fudan University Feiyu Chen Fudan University, Fudan University Cheng Zhou Fudan University, Fudan University Lei Yu Fudan University, Fudan University Jie Hu Fudan University, Fudan University Zheng Wang Fudan University, Fudan University Jian Zhou (  zhou.jian@zs-hospital.sh.cn ) Fudan University, Fudan University


Background
Primary liver cancer, which accounts for over 840,000 new cases per year, is the third leading cause of cancer-related deaths worldwide (1). Hepatocellular carcinoma (HCC), which is characterized by obscure symptoms and a propensity for hematogenous metastasis, is the most frequently occurring pathological type, accounting for about 90% of all cases. Although great progress has been made in development of multiple therapeutic modalities, such as surgical options, non-ablative treatment and targeted therapy(2), prognosis of HCC patients remains poor due to tumor heterogeneity and high malignancy (3). Previous studies have associated occurrence of HCC with various factors, including virus infection, alcohol abuse and NASH (4). For decades, activation of oncogenes as well as inactivation and mutation of suppressor genes have been the most popular predictors of HCC carcinogenesis (5,6). To date, however, the speci c molecular mechanism underlying HCC occurrence and progression remains elusive despite numerous researches on the subject.
Previous studies have demonstrated that the human genome encodes a spot of protein-coding genes, only representing < 2% of the total genome sequence (7). However, more than 90% of the noncoding RNAs (ncRNAs) are actively transcribed. Long noncoding RNAs (lncRNAs), a class of ncRNAs with a length of at least 200 nucleotides that are extensively expressed during genome transcription,, have been shown to participate in speci c biological processes(8). Therefore, they are considered a generalist due to their involvement in each step of gene regulation. Recent studies have demonstrated that dysregulation of lncRNAs, such as abnormal DNA methylation of promotor (9) and loss of genomic imprinting (10), is strongly associated with development and progression of diverse human diseases, including prostate and nasopharyngeal cancers (11,12). Additional evidences have revealed that some lncRNAs, such as SNHG7 and PDPK2P, can promote progression of HCC and are closely associated with poor prognosis (13,14).
Apart from their regulatory functions on cellular physiology, aberrant expression of lncRNAs via epigenetic regulation has also attracted numerous research attention. Previous studies have indicated that DNA methylation of CpG islands of the promoter regions regulate down-regulation of the tumor suppressor genes (15). Similarly, abnormal DNA methylation of lncRNAs promoters has emerged as a driver for malignancies (16). Although previous studies have shown that DNA methylation is a predictor for poor survival of HCC patients (17), interaction between DNA methylation and expression patterns of HCC-related lncRNAs remains unknown. In the present study, we investigated the mechanisms underlying regulation of HCC-speci c lncRNAs coupled with the changes in DNA methylation levels in promoter regions and the alterations associated with downstream target genes. Our ndings indicate that lncRNAs are a novel molecular marker and therapeutic target for HCC.

Patient recruitment and tissue collection
We recruited 8 HBV-related and pathologically con rmed primary HCC patients, who received radical hepatic resection at the Zhongshan hospital, Fudan university between October and December in 2011. None of the patients had received any auxiliary therapy before surgery. Primary HCC and adjacent normal tissues were obtained from patientsThe study protocol was approved by the Ethics Committee of Zhongshan Hospital, and informed consent was obtained from all subjects prior to inclusion in the study.

Microarray analysis
Expression pro les of LncRNAs from 8 human HCC and matched normal tissues were analyzed using ArrayStar Human LncRNA Microarray (8×60 k, ArrayStar, Rockville, MD, version 2.0). This dataset comprises a total of 33,045 lncRNAs collected from several databases, such as RefSeq, the UCSC Known genes, and Ensembl. Total RNA was extracted, ampli ed, and transcribed into uorescent complementary RNA (cRNA) using the Quick Amp Labeling Kit, One-Color (Palo Alto, CA, USA) according to the manufacturer's instructions. Labeled cRNAs were hybridized onto the human lncRNA arrays, and after washing steps, the arrays were scanned using the Agilent Microarray Scanner (G2565BA) and analyzed with the Feature Extraction software (version 11.0.1.1). Quantile normalization and data processing were performed using the GeneSpring GX software (version 11.5.1, Agilent Technologies). The threshold for screening of differential lncRNAs comprised a fold change >2.0 and P value <0.05. Subsequent processing of DNA methylation microarray data was conducted using ArrayStar Human 2.1M LncRNA Promoter Microarray (Rockville, MD, USA). Brie y, total DNA was extracted using the QIAamp DNA Mini Kit (QIAGEN, Valencia, CA, USA), sonicated, denatured and immunoprecipitated with anti-5-methylcytosine antibodies. The DNA-antibody complex was isolated and puri ed by immunomagnetic beads (Millipore, Bedford, MA, USA), followed by methylated DNA Immunoprecipitation (MeDIP) as previously described (18). After quality control, MeDIP and input DNA fragments were ampli ed using the Sigma WGA kit, and puri ed using the QIAquick PCR puri cation kit (QIAGEN, Valencia, CA, USA). Next, the uorescent-labeled MeDIP (Cy5) was mixed with input DNA (Cy3), denatured and hybridized onto the DNA methylation microarray. The mixture was washed, and the arrays scanned and analyzed. To accurately quantify methylation levels of CpGs, we applied a novel analytical methodology, known as modeling experimental data with MeDIP enrichment (MEDME), which utilizes the absolute methylation score (AMS) as an indicator of DNA methylation to identify differentially methylated regions (DMRs) of lncRNA promoters. Next, we merged differentially methylated probes, based on AMS, into candidate DMRs, then recalculated and re-tested the average AMS of candidate DMRs via t-test. Finally, we selected DMRs with average AMS that were signi cantly different between two groups, and de ned them as AMS DMRs.
RNA isolation, reverse transcription and quantitative real time polymerase chain reaction (qRT-PCR) Total RNA was isolated from tissues or cells using the TRIzol reagent (Invitrogen), then 2 µg of the RNA reverse transcribed to complementary DNA (cDNA) using the High Capacity cDNA Reverse Transcription kit (Ambion Inc., Austin, USA) according to the manufacturer's instructions. qRT-PCR was performed using the Power SYBR Green PCR Master Kit (Applied Biosystems, Foster City, CA) on the ABI 7900HT Fast Real-Time PCR System, targeting speci c genes whose primers are listed in Supplementary Table 1. The housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GADPH) was used as an endogenous control, while relative expression of LncRNAs was evaluated using the 2 −ΔΔCt method. All ampli cations were performed in triplicates.
Bisulphite modi cation combining sequencing PCR (BSP) and Methylation speci c PCR (MSP) Total genomic DNA was isolated from the tissues using the QIAamp DNA Mini Kit (QIAGEN) and 500 ng bisulfate-modi ed using the Methyl Code™ Bisul te Conversion Kit (Invitrogen, Carlsbad, CA, USA). Bisulfate-treated DNA was ampli ed and cloned into the pMD18-T vector. Methylation level for each site is indicated as the mean percentage of the total methylation on the sequencing data obtained from 10 clones. MSP was conducted as previously described (19), targeting genes whose primers are listed in Supplementary Table 2 Cell cultures Human hepatocellular carcinoma cell lines (HepG2, SMMC-7721, MHCC-97L, MHCC-97H, HCC-LM3) and the normal hepatocytes cell line (L02) were obtained from the Liver Cancer Institute, Fudan University (Shanghai, China), while human HCC cell line (Huh-7) was purchased from the Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences (Shanghai, China). Mycoplasma detections revealed negative results in all cell lines. All cells were cultured in Dulbecco's modi ed eagle medium (DMEM, Gibco-BRL, Grand Island, New York, USA), supplemented with 10% fetal bovine serum (FBS, Gibco-BRL, Grand Island, New York, USA), and maintained in a humidi ed incubator with 5% CO 2 and a temperature of 37℃.

Cell proliferation assay
Proliferation of MHCC-97H and HepG2 cells overexpressing lnc-SCARF1 (Lv-lnc-SCARF1) and controls (Lv-GFP) was evaluated using the CCK-8 assay kit (Dojindo Laboratories, Kumamoto, Japan), according to the manufacturer's instructions. Brie y, cell lines were seeded into 96-well plates (Corning Costar Corp., Kennebunk, ME, USA), at a density of 3 × 10 3 per well, and incubated overnight with 200 μl of cell culture medium at 37℃. The rate of cell proliferation was determined at 24, 48, and 72 h after seeding.

Cell invasion assay
The invasive ability of MHCC97H-Lv-lnc-SCARF1 and HepG2-Lv-lnc-SCARF1 cells were also tested using 6.5-mm Transwells with 8.0-μm pore polycarbonate membrane in 24-well plates (Corning Costar Corp., Kennebunk, ME, USA). Brie y, the transwell chambers were rst coated with 100 μl Matrigel (0.8 mg/ml) and incubated for 2 h at 37℃ . Next, 3 × 10 5 cells were inoculated into 100 μl of serum-free medium, and seeded onto the upper compartment of the chamber, while the lower chamber was loaded with 600 μl DMEM supplemented with 10% FBS, followed by a 24-h incubation at 37℃. The cells and Matrigel on the upper surface were then removed, and invasive cells located on the lower surface xed and stained with Giemsa. The invasive cells were numbered in ve microscopic elds (×200 magni cation). All individual experiments were performed in triplicate.

Determination of cell apoptosis
Cell apoptosis was performed using the Annexin V-PE Apoptosis Detection Kit (Becton Dickinson, San Jose, CA, USA) according to the manufacturer's instructions, with apoptotic cells detected via ow cytometry (Becton Dickinson, San Jose, CA, USA).

Western blot assay
Total proteins were extracted from cell lysates using 100 μL pre-cooled lysis buffer (1 ml RIPA buffer + 25 μL PMSF + 110 μL Phosphatase inhibitor cocktail), then quanti ed using the BCA protein quanti cation kit.
Next, 30 μg of denatured proteins were separated on a 10% SDS-PAGE and transferred onto PVDF membranes (Millipore, MA, USA). The membranes were blocked with TBST containing 10% nonfat milk powder for 2 h, and incubated overnight with primary antibodies of CUL9 at 4℃. Next, the membranes were washed with TBST for 30 mins and incubated with horseradish peroxidase (HRP)-labeled secondary antibodies (Jackson ImmunoResearch). The membranes were then exposed on the Pierce ECL Plus (Thermo Fisher Scienti c, MA, USA) in the ChemiDocTM XRS+ system (Bio-Rad, CA, USA), and images acquired and analyzed using the Image Lab Software.
Chromatin isolation by RNA puri cation (ChIRP) and mass spectrometry (MS) MHCC-97H and HepG2 cells were used for the ChIRP-MS experiment (100 million -500 million cells depending on the cell type). Summarily, cell harvesting, lysis, disruption, and ChIRP procedures were performed as previously described (20). Thereafter, protein samples were size-separated on bis-tris SDS-PAGE gels (Invitrogen) then subjected to western blot assay and mass spectrometry.
Establishment of a xenograft tumor mouse model MHCC97H-Lv-lnc-SCARF1, and HepG2-Lv-lnc-SCARF1 cells were subcutaneously inoculated into the left armpits of 6 male BALB/c nude mice (4-6 weeks old), while control cell lines were inoculated into the right armpit. The sizes of subcutaneous tumors were measured at day 0, 7, 14, 21, and 28 after injection, And on the 28 th day, all animals were euthanized, active tumors removed and cut into small pieces (2×2×2 mm 3 ). Orthotopic implantation was conducted on 24 male athymic BALB/c nude mice (4-6 weeks old). Brie y, each mouse was orthotopically implanted with a tumor mass, into the left lobe of their liver (12 mice per group). 6 mice were implanted with lnc-SCARF1 overexpression tumor mass while the other six were implanted with controls. All mice were euthanized by intraperitoneal injection of pentobarbital, after 4 weeks, then tumor sizes, location, and frequency of both intrahepatic and extrahepatic tumors detected and measured. Athymic BALB/c nude mice were housed in laminar-ow cabinets, under speci c pathogenfree conditions, and handled according to the recommendations of the ARRIVE guidelines for the care and use of laboratory animals. All related experimental protocols were approved by the Shanghai Medical Experimental Animal Care Committee.

Statistical analysis
All statistical analyses and illustration were performed using SPSS version 23.0 for Windows (SPSS Inc., Chicago, IL, USA) and Prism.v8.0 for Mac (GraphPad Software, La Jolla, CA, USA). Continuous data were rst subjected to normality tests, then expressed as means ± standard errors of the mean (SE). Differences in tumor recurrence, between control and treatment groups, were determined using the Chi-square test, while cumulative survival and recurrence rates were estimated using the Kaplan-Meier method with a logrank test. Independent prognostic factors were calculated using the Cox proportional hazards regression model. Data followed by P < 0.05 were considered statistically signi cant.  Table 3. To select hypermethylated and down-regulated lncRNAs in HCC, we combined and analyzed data from two microarray pro les. Finally, a total of 22 downregulated lncRNAs, with hypermethylation of the promoter region, were selected as candidates for further validation (Table 1 and Figure 1C). Pro les of expression and methylation of candidate lncRNAs in HCC tissues and cell lines

Differential lncRNAs expression and aberrant methylation pro les in HCC tissues
To verify aberrantly expressed lncRNAs, we performed qRT-PCR to determine their levels of expression in 8 pairs of HCC alongside peritumoral tissues. Results showed that 20 candidate lncRNAs were signi cantly downregulated in HCC, but there was no signi cant difference in expression of SRD5A1P1 and RP1-93C3.1, between HCC and paired nontumor tissues (Figure 2A). Next, we validated expression of candidate lncRNAs in another 52 pairs of HCC tissues, and found that they were signi cantly down-regulated in HCC relative to nontumor tissues ( Figure 2B)  Table 2). Similarly, BSP results con rmed a remarkable hypermethylation of promoters of 4 candidate lncRNAs in HCC tissues ( Figure 2C, Table 3). Table 4). Notably, Lnc-SCARF1 exhibited signi cant differences in promoter methylation between HCC and peritumor tissues (P<0.001), and was ultimately was chosen as a candidate for further explorations.  Hypermethylation in the promoter of lnc-SCARF1 induces its downregulation in HCC tissues Both MSP and BSP results revealed hypermethylation of lnc-SCARF1 promoter in HCC relative to adjacent normal tissues, indicating that most of the CpG sites within the CpG islands of the promoter were hypermethylated in HCC tissues ( Figure 2C). Furthermore, lnc-SCARF1 was signi cantly overexpressed in the normal liver cell line LO2 relative to HCC cell lines ( Figure 3A). Next, we used the inhibitor of DNA methylation, 5-aza-deoxycytidine (5-AZA), to investigate the relationship between hypermethylation in promoter and downregulation of lnc-SCARF1. Two types of HCC cell lines, namely HepG2 and MHCC-97H, which earlier exhibited downregulation of lnc-SCARF1, were chosen and cultured with two different concentrations of 5-AZA (30, and 60µM). After treatment, we observed upregulation of lnc-SCARF1 expression in HepG2 and MHCC-97H cells, in a concentration-dependent manner ( Figure 3B). These results indicated that hypermethylation in the promoter of lnc-SCARF1 induces its downregulation in HCC.

Over-expression of lnc-SCARF1 inhibits tumor proliferation and migration of HCC in vitro
To investigate lnc-SCARF1's biological function, we successfully constructed lentiviral vectors harboring lnc-SCARF1 and stably transfected HepG2 and MHCC-97H cell lines ( Figure 3C). Next, we tested the effect of overexpressing lnc-SCARF1 on proliferation, migration and apoptosis of two transfected cell lines, alongside invalid controls, using a series of assays, and found that its overexpression signi cantly suppresses proliferation of HepG2 and MHCC-97H cells ( Figure 3D). Moreover, we characterized the effects of lnc-SCARF1 overexpression on migration, and as expected, overexpressing lnc-SCARF1 inhibited cell migration and suppressed the number of migrating cells, with upregulation of lnc-SCARF1 was signi cantly lower than that of control cells ( Figure 3E and F). Furthermore, ow cytometry results con rmed that enforced expression of lnc-SCARF1 induced apoptosis of HepG2 and MHCC-97H cells ( Figure 3G).
Lnc-SCARF1 interacts with functional domains of signaling proteins, to act as a class of CUL9 modulators and suppress HCC metastasis Next, we applied chromatin isolation by RNA puri cation (ChIRP) to determine whether lnc-SCARF1 plays a role in inhibiting tumor growth and metastasis by binding to speci c proteins. Results from mass spectrometry (MS) analysis revealed that CUL9 speci cally bound to lnc-SCARF1 ( Figure 4E). To further validate the interaction between lnc-SCARF1 and CUL9, we performed western blot assay to detect CUL9 in both HepG2 and MHCC-97H cell lines. Results revealed a marked upregulation of CUL9 protein in these HCC cell lines after LV-SCARF1 transfection ( Figure 4F).

Discussion
Hepatocellular carcinoma is a lethal disease characterized by high malignancy and an overall ve-year survival rate of 5-30% (21). Accumulating evidences have shown that lncRNAs play a vital role in carcinogenesis of HCC, while aberrant DNA methylation regulates cellular biological function of liver cancer (22). Moreover, previous studies have applied genome-wide DNA methylation analysis to reveal several epigenetically dysregulated lncRNAs in human renal cancer(16) and breast cancer ( DNA hypermethylation modi cation of CpGIs within the promoter of tumor suppression genes (TSGs) has been found to contribute to transcriptional silencing and HCC carcinogenesis (29). These TSGs, including RASSF1A, p16, and APC, have also been shown to play a key role in regulation of cell apoptosis, cell cycle and other tumor-related signaling (30). Results of the present study revealed that lnc-SCARF1 was downregulated in HCC, but the frequency of promoter methylation was high. To ascertain whether inactivation of lnc-SCARF1 was related to hypermethylation of its promoter, we used 5-AZA, an inhibitor of methylation, to prevent promoter hypermethylation in vitro. Results showed that lack or weak expression of lnc-SCARF1 in HCC tissues was closely associated with its hypermethylated promoter CpGIs, consistent with the ndings of a previous study that showed that methyltransferase EZH2 regulated the levels of H3K27me3 at promoters of lncRNAs thereby regulating their expression (31). This result further indicated that lncRNAs regulate promoter methylation in a similar fashion to that of protein coding genes. Moreover, another study found that an inverse correlation between linc-POU3F3 and POU3F3 gene was associated with regulation of EZH2 (32), suggesting that lncRNAs might be associated with chromatin modifying complexes where they modulate their neighboring gene transcription and expression by regulating the methylation status. However, only a handful of studies have described the underlying mechanism of promoter methylation in modulating expression of lncRNAs in HCC. Consequently, we sought to ascertain whether regulation of methyltransferases, such as EZH2, was correlated to the methylation status of lnc-SCARF1 in HCC.
LncRNAs exert their regulatory role by interacting with transcriptional regulatory factors or chromatin protein complexes. For example, Lnc-HEIH was not only associated with EZH2 but competitively recruited it to speci c genomic DNA regions, and repressed the EZH2 target genes(26). Additional evidences have also demonstrated that lncRNAs are directly correlated with functional domains of signaling proteins and they activate the signaling pathway that regulates tumorigenesis and progression. For instance, Lnc-LET was found to play a role in the stabilization of signaling proteins thereby leading to hypoxia-induced cell invasion of HCC (33). On the other hand, Lnc-MEG3 reportedly induced expression of p53 by activating the p53 signaling pathway, further inhibiting cell proliferation and inducing cell apoptosis (34). In order to clarify the potential regulating signaling pathway of lnc-SCARF1, we utilized ChIRP to pull down and purify chromatin protein complexes that interact with lnc-SCARF1, then employed mass spectrometry to further screen target proteins that are involved in relevant tumor signaling. MS results revealed that Cullin 9 (CUL9) speci cally interacted with lnc-SCARF1. CUL9, which is predominantly localized in the cytoplasm, is a member of the cullin family of E3 ubiquitin ligases (35 In summary, we revealed that lnc-SCARF1 plays an important role in HCC suppression. Speci cally, promoter methylation directly regulates expression of lnc-SCARF1, which subsequently modulates proliferation, metastasis and apoptosis of HCC cells by interacting with its target gene CUL9. However, the mechanism of lnc-SCARF1's regulation by promoter methylation, and the interaction between SCARF1 with its target gene CUL9, remain unclear. Elucidating the mechanism through which promoter methylation regulates lncRNAs, and the potential interaction between SCARF1 and CUL9 are imperative to understanding the mechanism underlying the role of lncRNAs in HCC development and progression, and providing new therapeutic signatures and targets.

Conclusion
Taken together, our ndings indicate that LncRNA-SCARF1 plays important roles in HCC progression by interacting with CUL9, thus may serve as a prognostic biomarker or an effective target for future development of HCC therapies. statistical analyses.;Jian Zhou critically revised the manuscript for important intellectual content. Jie Hu and Zheng Wang conceived and designed the study. All authors read and approved the nal manuscript.