MiR-17-92 cluster promotes hepatocarcinogenesis

MiR-17-92 cluster is an oncogenic miRNA cluster that is implicated in several cancers, although its role in hepatocarcinogenesis has not been clearly defined. In this study, we show that the miR-17-92 cluster is highly expressed in human hepatocellular carcinoma (HCC) tissues compared to the non-tumorous liver tissues by RT-PCR and in situ hybridization analyses. Increased miR-17-92 cluster expression in HCC tissues was further confirmed by analysis of the RNA-sequencing data of 319 patients available from the Cancer Genome Atlas (TCGA) Data Portal (https://tcga-data.nci.nih. gov/tcga/). To create an animal model that resembles enhanced miR-17-92 in the liver, we developed liver-specific miR-17-92 transgenic mice and the animals were treated with the hepatic carcinogen, diethylnitrosamine (DEN). We observed that the liver-specific miR-17-92 transgenic mice showed significantly increased hepatocellular cancer development compared to the matched wild-type control mice. Forced overexpression of the miR-17-92 cluster in cultured human hepatocellular cancer cells enhanced tumor cell proliferation, colony formation and invasiveness in vitro , whereas inhibition of the miR-17-92 cluster reduced tumor cell growth. By analyzing the miRNA and mRNA sequencing data from the 312 hepatocellular cancer patients available from the TCGA database, we observed that the expression levels of the miR-17-92 cluster members and host gene in the tumor tissues are negatively correlated with several target genes, including CREBL2, PRRG1, NTN4. Our findings demonstrate an important role of the miR-17-92 cluster in hepatocarcinogenesis and suggest the possibility of targeting this pivotal miRNA cluster for potential therapy. The current study provides key evidence for an important oncogenic role of the miR-17-92 cluster in HCC. We show that miR-17-92 cluster is highly expressed in human HCC tissues and cells. By developing a novel liver-specific miR-17-92 transgenic mouse model, we observe that miR-17-92 expression in the liver enhances hepatocellular cancer development. The oncogenic role of the miR-17-92 cluster is further demonstrated by in vitro studies using cultured human HCC cells with miR-17-92 overexpression or inhibition. Analysis of the available miRNA and mRNA sequencing databases for HCC patients shows that the expression levels of the miR-17-92 cluster members and host gene in HCC tissues are negatively correlated with several target genes, including CREBL2, PRRG1 and NTN4. These findings demonstrate an important role of the miR-17-92 cluster in hepatocarcinogenesis.


Introduction
Hepatocellular carcinoma (HCC) is the fifth most prevalent cancer and third most frequent cause of cancer death (1,2). Common risk factors for HCC include chronic virus hepatitis B and C, alcoholic and non-alcoholic steatohepatitis and other causes of chronic liver diseases (3). The tumorigenic processes are complex and involve deregulation of multiple signaling pathways (4). The tumor is usually diagnosed at advanced stage which prevents curative therapies including surgical resection and liver transplantation, which are only applicable to patients at relatively early stages of tumor development (5). However, a majority of patients have advanced or unresectable disease at presentation which prevents curative treatments. Thus, HCC continues to represent a devastating human malignancy which requires further mechanistic investigations and more effective therapeutic interventions.
The current study provides key evidence for an important oncogenic role of the miR-17-92 cluster in HCC. We show that miR-17-92 cluster is highly expressed in human HCC tissues and cells. By developing a novel liver-specific miR-17-92 transgenic mouse model, we observe that miR-17-92 expression in the liver enhances hepatocellular cancer development. The oncogenic role of the miR-17-92 cluster is further demonstrated by in vitro studies using cultured human HCC cells with miR-17-92 overexpression or inhibition. Analysis of the available miRNA and mRNA sequencing databases for HCC patients shows that the expression levels of the miR-17-92 cluster members and host gene in HCC tissues are negatively correlated with several target genes, including CREBL2, PRRG1 and NTN4. These findings demonstrate an important role of the miR-17-92 cluster in hepatocarcinogenesis.

Materials
Minimum essential medium (MEM), Dulbecco's modified Eagle's medium and heat inactivated fetal bovine serum were purchased from Sigma (St Louis, MO). OPTI-MEM reduced serum medium, Lipofectamine™ 2000 reagent and puromycin were purchased from Invitrogen (Carlsbad, CA). Bronchial epithelial cell basal medium with supplemental growth factors in BEGM SingleQuot Kit was purchased Lonza (Walkersville, MD). The miR-17-92 cluster or miR-92a expressed and scrambled control lentiviral particles with enhanced green fluorescent protein were obtained from GeneCopoeia (Rockville, MD). Total RNA from human hepatocytes was purchased from Zenbio (Research Triangle Park, NC; Catalog -RNA-L10).

Cell culture
Five human HCC cell lines (Huh-7, Sk-Hep-1, Hep3B, HepG2, PLC/PRF/5) were utilized in this study. The HCC cell lines Sk-Hep-1 (ATCC® HTB-52™), Hep3B (ATCC® HB-8064™), HepG2 (ATCC® HB-8065™) and PLC/ PRF/5 (ATCC® CRL-8024™) were obtained from the American Type Culture Collection (Manassas, VA) where the cell line were tested and authenticated by STR DNA profiling, mycoplasma detection, cell viability analysis and/or isoenzyme detection. Cells were cultured minimum essential medium containing 10% fetal bovine serum. Huh-7 (JCRB0403) was obtained from Japanese Cancer Research Resources Bank (Ibaraki City, Japan) where the cell line was tested and authenticated by STR DNA profiling and isoenzyme detection. Huh-7 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. All cells were cultured in a humidified atmosphere of 5% CO 2 incubator at 37°C. The miR-92a or miR-17-92 cluster-overexpressed and scramble control stable cell lines were established by transduction with the corresponding lentiviral vector or miRNA-scrambled control lentiviral vector, followed by selection with media containing puromycin.

In situ hybridization for miRNA
In situ hybridization for miR-92a was performed in the formalin-fixed and paraffin-embedded tissue specimens surgically resected from patients diagnosed with HCC by using the MiRCURY LNA microRNA ISH Optimization Kit (Exiqon, Vedbaek, Denmark) with the approval of the Institutional Review Board. Briefly, 6-μm-thick paraffin sections were deparaffinized and treated with proteinase-K (15 μg/ml) at 37°C for 10 min. After dehydration, slides were incubated with 100 nM miR-92a locked nucleic acid probe (5′-DIG-ACAGGCCGGGACAAGTGCAATA-3′-DIG) at 50°C for 2 h, followed by stringent washes with 5× standard saline citrate, 1× saline sodium citrate and 0.2× saline sodium citrate buffers at 50°C; DIG blocking reagent (Roche) in maleic acid buffer containing 2% sheep serum at room temperature for 15 min; and alkaline phosphatase-conjugated antidigoxigenin (diluted 1:500 in blocking reagent; Roche) at room temperature for 2 h. Enzymatic development was performed by incubating the slides with 4-nitro-blue tetrazolium and 5-brom-4-chloro-3′-Indolylphosphate substrate (Roche) at 30°C for 2 h to allow formation of dark-blue 4-nitro-blue tetrazolium formazan precipitate, followed by nuclear fast red counterstain (Vector Laboratories, Burlingame, CA) at room temperature for 10 min. Slides were then dismantled in water, dehydrated in alcohol solutions and mounted with mounting medium (Vector Laboratories). Scrambled probe and U6 small nuclear RNA-specific probe were used as control. A standard four-point scale method was used to evaluate the staining intensity under microscope and the results were scored as 0 (negative), 1 (+), 2 (++) or 3 (+++) according to established criteria (22). Specifically, 3 (+++) indicates dark staining that is easily visible with a low power objective and involves > 50% of cells; 2 (++) indicates moderate darkly staining areas <50% of cells; 1 (+) indicates weak staining or pale staining in any proportion of cells not easily seen under a low power; 0 (negative) indicates no staining (showing none of the above staining). The same procedure was utilized to analyze the HCC tissue arrays (serial # LV2082, obtained from US Biomax, Rockville, MD). The tissue microarray consisted of 208 tissue core samples, including 94 cases of HCC, 5 cases of cancer adjacent normal hepatic tissue and 5 cases of normal liver tissue, each case with duplicate samples.

miRNA extraction and qRT-PCR
Cellular total RNA was isolated using Trizol reagent (Invitrogen). RNA from macrodissection-based formalin-fixed, paraffin-embedded patient tissues was isolated using an RNeasy formalin-fixed, paraffin-embedded Kit (Qiagen, Valencia, CA) following the manufacturer's protocol. Reverse transcription was performed by using miScript Reverse Transcription Kit (Qiagen). miScript SYBR Green PCR Kit (Qiagen) and miScript Primer Assays for six different miRs were used to amplify individual mature form of miR-17/18/19a/19b/20/92a on Bio-Rad C1000 Thermal Cycler. U6 small nuclear 2 (U6b) was used as the internal control. The Pre-miR-17-92 cluster gene was amplified by using the QuantiFast SYBR Green PCR Kit (Qiagen). GAPDH was used as the internal control. The pre-miR-17-92 cluster forward primer is 5′-CAGTAAAGGTAAGGAGAGCTCAATCTG-3′, reverse primer is 5′-CATACAACCACTAAGCTAAAGAATAATCTGA-3′. miRNA and mRNA expression was normalized to their corresponding internal control genes and relative change was calculated using the 2 −∆∆CT method.

Cell proliferation WST-1 assay
The cell proliferation reagent WST-1 (Roche) was used for quantitative determination of cellular proliferation according to the manufacturer's instruction. Briefly, the 2 × 10 3 cells were seeded onto each well of 96-well plates and cultured for 24 h to allow attachment. Then the medium was replaced with 100 µl fresh medium containing 1% fetal bovine serum or specific reagents as indicated and the cultures were continued for indicated days. For cell growth measurement, the cells were incubated in the presence of 10 μl/well WST-1 at 37°C and 5% CO 2 for 1 h. The absorbance of the samples was measured at 450 nm using an automatic ELISA plate reader (VersaMax Microplate Reader).

Colony-formation assay
About 1 × 10 3 cells were cultured in 10 cm dishes for 14 days to allow colony formation. Colonies were fixed in 100% methanol and stained with 0.1% crystal violet solution (Amersco, Solon, OH) for counting.

Cell invasion assay
The cell invasion assay was performed in matrigel-coated transwell chambers (BD Biosciences Discovery Labware, Bedford, MA). About 500 ul cell suspension (5 × 10 4 cells/ml) was added to each of the upper chambers. Cell culture medium containing 10% fetal bovine serum was added to each of the lower chambers as chemoattractant. After incubating cells at 37°C for 24 h, the cells on the upper surface of the membrane were removed with a cotton swab. The invading cells on the lower surface of Abbreviations   DEN  diethylnitrosamine  HCC  hepatocellular carcinoma  miRNA  microRNA  mRNA  messenger RNA  PBS phosphate-buffered saline qRT-PCR quantitative real-time polymerase chain reaction the membrane were fixed in 100% methanol and stained with Mayer's hematoxylin solution (IHC world, Woodstock, MD). Then the invading cells were counted under a microscope and for each chamber eight fields were randomly selected for counting.

DEN-induced hepatocarcinogenesis model
miR-17-92 liver specific transgenic mice and wild-type control mice were used in this model. At 14 days postnatal age, the entire mouse litter received a single intraperitoneal injection of the genotoxic tumor initiator DEN (25 μg/g body weight; catalog no. N0756, Sigma, St Louis, MO; dissolved in 0.9% saline) which induces hepatocyte DNA damage through DNA adduct formation (23). Groups of seven male mice were used for the experiments. Mice were killed at 4, 8, 10, 12 months after DEN injection. At necropsy, terminal body weights, liver weights and liver morphology were evaluated. Livers were examined for the presence of liver tumors and sections of liver, liver tumors were fixed in 10% formalin and embedded in paraffin for histological examination. About 4 μm sections were made and stained with hematoxylin and eosin (H&E) or anti-Ki67 antibodies.

RNA-Seq data analysis
The Cancer Genome Atlas (TCGA) Data Portal (https://tcga-data.nci.nih. gov/tcga/, last accessed on September 16th, 2014) provides HCC patient mRNA sequence and miRNA sequence result. We utilized all the available level 3 data (319 HCC patient sample data), and imported the data into Subio platform Version 1.17 with Basic and Advanced Plug-ins (Kagoshima, Japan) for database management and data filtering, enrichment, clustering, find similar patterns and miRNA targets analysis.

Statistical analysis
Results are presented as mean ± SD or SEM from a minimum of three replicates. Differences between groups were compared using either Student t test or one-way analysis of variance (ANOVA). P < 0.05 was considered statistically significant. The analysis was performed using by SPSS 13.0 statistical software (IBM, Armonk, NY). For data from human samples, statistical significance between means was determined by the non-parametric Krustal-Wallis test.
We next performed experiments to evaluate the role of miR-17-92 cluster in vivo. We generated liver-specific miR-17-92 cluster transgenic mice by breeding the Loxp-miR-17-92-Loxp mice (which carry a loxP-flanked Neo-STOP cassette with a genomic DNA fragment encoding miR-17-92 gene) with the Alb-Cre mice (which carry Cre recombinase gene under the control of hepatocyte-specific Albumin Promoter) (Figure 2A). The offsprings were genotyped to confirm the removal of Neo-STOP cassette. A 6-25-fold increase of miR-17-92 cluster in liver tissue is confirmed by quantitative real-time polymerase chain reaction (qRT-PCR) at 4-8 weeks as well as at 10 months ( Figure 2B and Supplementary Figure 1, available at Carcinogenesis Online). Thus far, we have successfully established the liver-specific miR-17-92 transgene mice. Next, we subjected the liver-specific miR-17-92 transgene mice and their matched wild-type controls to the hepatic carcinogen diethylnitrosamine (DEN)induced hepatocarcinogenesis. The carcinogenic process was initiated by a single intraperitoneal injection of DEN at 25 µg/g body weight at postnatal day 14. Then the mice were killed at different time periods (4, 8, 10 months). At the 4-month period, only one out of four transgenic mice showed a small dysplastic focus, while the control mice showed no liver lesion. At the 8-month period, only few mice showed dysplastic foci or small tumor nodules, which were too few and two small to tally. At the 10-month period, all liver-specific miR-17-92 transgenic mice develop visible hepatocellular cancer, while control mice showed no tumor or very small tumor nodules ( Figure 2B). As shown in Figure 2C, the average tumor number per mice is 5.29 ± 1.51 in the transgenic groups versus 0.86 ± 0.34 in the wild-type group (P = 0.026); the total tumor volume per mice is 1.86 ± 0.58 cm 3 versus 0.14 ± 0.01 cm 3 (P = 0.029); the maximum size of tumor in each mice is 132.86 ± 26.52 versus 15.51 ± 7.31 mm (P = 0.002). However, the liver/body weight ratio in the TG group is slightly higher compared to the WT group (9.26 ± 2.24 versus 5.66 ± 1.49), the difference is not statistically significant. Taken together, our data indicate that the liverspecific miR-17-92 transgenic mice developed significantly larger and more liver tumors compared to the wild-type mice (at 10 months after DEN injection). Under histological evaluation, the liver tumors in the miR-17-92 liver-specific transgenic mice comprise predominantly of HCC (85.71% HCC, 14.29% adenoma, 0.00% hyperplastic foci), while the liver lesions in the wild-type mice were predominantly of hyperplastic foci (0.00% HCC, 42.86% adenoma, 57.14% hyperplastic foci). The tumor progression status in the liver-specific miR-17-92 mice are generally more advanced than in the wild-type control mice ( Figure 2D and E). No intrahepatic cholangiocarcinoma is identified in these animals. The liver tissue sections from the miR-17-92 transgenic mice show more proliferative cells compared to that of the wild-type mice, as determined by the Ki67 immunostaining ( Figure 2D). These findings demonstrate that overexpression of miR-17-92 cluster in the liver promotes hepatocellular cancer development, in vivo.
We further examined the levels of miR-17-92 cluster members in different human HCC cell lines and examined their effect on tumor cell growth. Compared to the human adult hepatocytes, all of the six miRNAs (miR-17/18/19a/19b/20/92a) are highly expressed in five HCC cell lines (Hep3B, PLC, HepG2, Huh-7 and Sk-hep-1), especially in PLC and Hep3B cells; the miR-17-92  miRNAs are all in very low levels in human adult hepatocyte (HH). Similar to the results in human HCC tissue samples and in the liver tissues from the miR-17-92 transgenic mice, among the six mature miRNAs, miR-92a is the most abundant ( Figure 3). We then constructed human HCC cell lines with stable overexpression of miR-17-92 by infecting the parental cell line (Huh-7 and Sk-hep-1) with lentivirus particles carrying the miR-17-92 gene or the scrambled control miRNA as control [these vectors also carry the enhanced green fluorescent protein gene (EGFP) under the control of the same promoter]. High infection efficiency was confirmed by the expression of EGFP in nearly all transduced cells. Successful overexpression of miR-17-92 cluster in the lentivirus stably transduced cells was verified by qRT-PCR. As shown in Figure 4A, overexpression of the miR-17-92 cluster significantly increased the growth of HCC cells in vitro. MiR-17-92 cluster overexpression also enhanced HCC cell invasiveness and colony-formation efficiency ( Figure 4B and C). In contrast, inhibition of miR-17-92 cluster (via anti-miRs transfection) decreased tumor cell proliferation (Supplementary Figure 2, available at Carcinogenesis Online). In separate experiments, we observed that overexpression of miR-92a also enhanced HCC cell proliferation, colony formation and invasiveness ( Figure 5). Taken together, these data document an important role of miR-17-92 cluster in HCC cell growth.
TCGA data portal contains large set of RNA sequence data of different types of cancer with clinical information. We analyzed all the available level 3 sequencing data of HCC patients from this database, which included the data from 319 HCC patients (all 319 patients had miRNA-seq data; 312 of 319 patients had matched RNA-seq data). Utilizing the Subio platform, we also made comparison between mRNA and miRNA sequencing result combined with seed sequence prediction from several online websites, including TargetScan (http://www.targetscan.org/) (24), PicTar (http://pictar.mdc-berlin.de/) (25)(26)(27) and Miranda (http://www. microrna.org/microrna/) (28). This approach allowed us to identify highly correlated miRNA-Target pairs (listed in Figure 6A and Supplemental Table 2, available at Carcinogenesis Online). CREBL2, EXPH5, NTN4, MYLK, PRRG1, ESR1 are highly possible targets of the miR-17-92 cluster, as they not only are predicted targets based on binding sequence but also show an expression pattern negatively correlated with the respective miRNAs in the tumor and non-tumor samples (Pearson Correlation) ( Figure 6B). Given that MIR17HG is the host gene for the miR-17-92 cluster, we further analyzed its expression level to determine whether there might be a correlation between MIR17HG and the other genes from the RNA-seq database. We observed that the average expression of MIR17HG gene in HCC tumor is significantly higher than nontumor groups (P < 0.0001) ( Figure 6C), which is in accordance with increased miR-17-92 cluster members as shown in Figure 1 by qRT-PCR and ISH analyses. By comparing the expression patterns of MIR17HG and all other genes, we identified a list of genes whose expression pattern is negatively correlated with MIR17HG in the tumor and non-tumorous liver tissues (Supplementary  Table 3, available at Carcinogenesis Online). We noticed that most of the identified genes are also targets of at least one miRNAs in miR-17-92 cluster. The most correlated targets (correlation −1 to −0.3) from that list are shown in Figure 6D; the expression levels of those genes are significantly downregulated (by threshold >1.5fold) in the tumor (compared to the non-tumorous liver tissues). It is notable that several targets identified by miRNA-target pair analysis, including CREBL2, PRRG1 and NTN4, are also among the downregulated targets of the MIR17HG host gene. Based on these findings, we sought to further determine whether the expression of any of these targets might correlate with the patients' clinical outcomes. To this end, we extracted the clinical information from the database and analyzed the survival data; we grouped all of the patients into two groups (low and high) according to each gene's expression level, and draw survival curve for each group. We observed a significant difference in survival curve between the two groups for the CREBL2 gene: the CREBL2-low group has a shorter patient survival time than the CREBL2-high group ( Figure 6E) [P < 0.05, Log-rank (Mantel-Cox) test].

Discussion
MicroRNAs have been reported to be involved in HCCs (29)(30)(31). In the current study, we evaluated the role of the miR-17-92 cluster in hepatocellular cancer by using complementary approaches of tumor induction in mice, human hepatocellular cancer cell studies in vitro, and patient tissue analyses. We observed that target overexpression of the miR-17-92 cluster in the liver enhanced hepatic carcinogen-induced liver tumor development, characterized by more tumor number, larger tumor size and higher tumor volume. These in vivo findings are corroborated by the in vitro studies showing that forced overexpression of the miR-17-92 cluster (or one of its members, miR-92a) enhanced hepatocellular cancer cell proliferation, colony formation and invasiveness, in vitro. We further observed that inhibition of miR-17-92 cluster prevented hepatocellular cancer cell growth. These findings establish a key tumor-promoting role of miR-17-92 cluster in hepatocarcinogenesis. The observation that target overexpression of miR-17-92 cluster in the liver enhances tumor development only in the setting of DEN treatment (but not spontaneously) suggests that the miR 17-92 polycistron is a tumor promoter but not a tumor initiator.
By taking advantage of the available online database (TCGA), we analyzed the mRNA and miRNA sequencing data from HCC patients. RNA-sequencing is a robust approach to quantify overall gene expression level, which provides high quality and reproducible data. Our analysis revealed that several miRNAtargets pairs, including miR-18a-CREBL2, miR-17-NTN4, miR-17-PRRG1, are probably important in hepatocarcinogenesis. We observed that the expression of one of these targets, CREBL2, is positively correlated with patients' survival. These findings suggest that miR-17-92 cluster may enhance hepatocarcinogenesis by downregulating key targets including CREBL2. In this context, it is worth mentioning that CREBL2 (cAMP-response-elementbinding protein-like 2) is a CREB family member which can interact with CREB to regulate lipogenesis (32) and that CREBL2 may be a tumor suppressor as it is frequently deleted in malignancy (33).
Connolly et al. (34) reported increased expression of miR-17-92 cluster and miR-21 in hepatitis B virus-positive human and woodchuck HCCs. In this context, it is interesting that miR-17-92 is known to act with c-myc to promote carcinogenesis (12) and that c-myc can be activated by Hepatitis B virus X (35). Thus, miR-17-92 cluster may regulate carcinogenesis through interaction with other signaling molecules. Additionally, given that miR-21 is known to promote HCC through mechanisms including targeting PTEN, PDCD4, RHOB and other tumor suppressors (36)(37)(38)(39) and that miR-21 is the most overexpressed miRNA in HCC patients (Supplementary Table 1, available at Carcinogenesis Online), it is possible that miR-17-92 and miR-21 may coordinately regulate hepatic carcinogenesis.
A recent study shows that miR-17-92 expression in HCC is regulated by epigenetic modification via histone deacetylation (40). Given that E2F1 and c-myc are well known to transcriptionally activate miR-17-92 cluster (41,42), it is possible that these transcription factors may also be implicated in regulating miR-17-92 expression in HCC. Thus, multiple mechanisms are probably to be involved for upregulation of miR-17-92 during hepatocarcinogenesis.
Recent studies suggest that circulating miRNAs, such as miR-21 and miR-122, may serve as biomarkers for HCC (43,44). Zhou et al. (45) showed that seven miRNAs set (miR-21, miR-26a, miR-27a, miR-122, miR-192, miR-223 and miR-801) can differentiate HCC form healthy, chronic hepatitis B and cirrhosis groups. miR-92a and miR-18a are also implicated in HCC (21,46,47). Given that miR-17-92 cluster, particularly miR-92a, is highly expressed in HCC as shown in the current study, future studies are needed to validate whether miR-17-92 cluster or miR-92a can be used as a valid marker for early diagnosis of HCC. Moreover, given the oncogenic role of the miR-17-92 cluster in HCC as documented in the current study, further investigations are warranted to determine whether miR-17-92 can be targeted for future treatment of human HCC.