Cyclin G1 Is a Target of miR-122a, a MicroRNA Frequently Down-regulated in Human Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) accounts for 80% to 90% of liver cancers and it is one of the most prevalent carcinomas throughout the world (1). Cirrhosis represents the strongest predisposing factor, as 80% of HCCs develop in cirrhotic livers (2). By assaying HCCs with DNA microarrays, several signaling pathways potentially involved in HCC development and progression through the modulation of multiple mRNAs have been recognized (3). In addition, specific molecular signatures have been associated with different etiologic factors, biological characteristics, and clinical evolution (4, 5). In particular, deregulated expression of protein involved in cell cycle regulation and in DNA repair has extensively been described as a crucial event in the carcinogenetic process leading to HCC development (6).

In the recent years, a new class of small noncoding RNAs, microRNAs (miRNA), has been discovered in animals and plants (7–9). miRNAs are 19- to 25-nucleotide-long RNAs, able to bind complementary sequences in 3′-untranslated regions (3′-UTR) of several target mRNAs to induce their degradation or translational repression (10). They are phylogenetically conserved, play important roles in developmental timing, and participate in the regulation of processes, such as cell fate determination, proliferation, differentiation, and cell death (11–14). In the most recent database (miRBase 9.0), ∼500 validated miRNAs are presently identified in the human genome (15), which could regulate thousands of protein-coding genes (16, 17). Most of the protein-coding genes regulated by miRNAs are presently not defined and bioinformatic approaches may help to recognize them.

The development of microarray platforms for the analysis of miRNA expression revealed that multiple miRNAs are aberrantly expressed in human malignancies, suggesting that they may represent a novel class of oncogenes or tumor suppressor genes (18–22). Here, we report a comprehensive analysis of miRNA expression in HCCs arisen on liver cirrhosis (LC). Our data suggest that the aberrant expression of a restricted panel of miRNAs could participate in the molecular events leading to HCC development. Moreover, among deregulated miRNAs, miR-122a, which accounts for 70% of the total liver miRNA population (23), was analyzed as a modulator of cyclin G1 expression.

Patients. Tissues were obtained from 60 patients (45 males and 15 females) undergoing liver resection for HCC on LC. Tissue samples were collected at surgery, immediately snap frozen in liquid nitrogen, and stored at −80°C until RNA extraction. The characteristics of HCC/LC patients included in this study are described in Table 1. Exclusion criteria were a previous history of local or systemic treatment for HCC and the presence of noncirrhotic tissue surrounding the HCC nodule/s. miRNA expression was analyzed by microarray on 17 HCCs and 21 LCs (matched in 13 cases) by Northern blot on 40 cases of matched HCC/LC and by quantitative real-time reverse transcription-PCR (RT-PCR) on 38 cases of matched HCC/LC. Characteristics of each sample and study details are summarized in Supplementary Tables S1 and S2.

miRNA microarray. Total RNA was extracted by using Trizol (Invitrogen) according to the manufacturer's instructions. One-color RNA labeling and hybridization on miRNA microarray chips was done as described previously (18). For the analysis of miRNA expression profile, 17 HCCs and 21 LCs (matched in 13 cases) were hybridized on a miRNA microarray consisting of 381 probes for 238 mature and 143 precursor human miRNAs. Hybridization signals were detected by biotin binding of a streptavidin-Alexa Fluor 647 conjugate using a GenePix 4000B scanner (Axon Instruments). Images were quantified by the GenePix Pro 6.0 (Axon Instruments).

Analysis of microarray data. Raw data from one-color miRNA microarrays were normalized and analyzed by GeneSpring GX software version 7.3 (Agilent Technologies). The GeneSpring software generated a unique value for each miRNA, doing the average of replicate probes present on chip. Samples were normalized using the on-chip median normalization. Then, each tumor was normalized on the cirrhosis of the same patient, when available, or on the average of cirrhosis. Differentially expressed miRNAs were identified by using a filter based on a fold change of 1.3 combined with an ANOVA for HCCs versus cirrhosis comparison (P < 0.05) with Benjamini and Hochberg correction for false-positive reduction. The list of differentially expressed genes was tested for its prediction power with the algorithms prediction analysis of microarrays (PAM; ref. 24) and support vector machine (SVM; ref. 25). Unsupervised hierarchical cluster analysis was done after median centering of each chip using average linkage and standard correlation as measure of similarity. Supervised clusterization was done after the normalization on gene median to highlight differences across samples.

Northern blot analysis. RNA samples (10 μg each) were electrophoresed on 15% acrylamide and 7 mol/L urea Criterion precasted gels (Bio-Rad) and transferred onto Hybond N+ membrane (Amersham Biosciences). Membranes were hybridized with oligonucleotide probes corresponding to the complementary sequences of the following mature miRNAs: mir-221, 5′-GAAACCCAGCAGACAATGTAGCT-3′; let-7a-1, 5′-AACTATACAACCTACTACCTCA-3′; and miR-122a, 5′-ACAAACACCATTGTCACACTCCA-3′. Probes were 5′-end labeled using the polynucleotide kinase in the presence of [γ-³²P]ATP. Hybridization was done at 37°C in 7% SDS/0.2 mol/L Na₂PO₄ (pH 7.0) for 16 h. Membranes were washed at 42°C, twice with 2× standard saline phosphate [0.18 mol/L NaCl/10 mmol/L phosphate (pH 7.4)], 1 mmol/L EDTA [saline-sodium phosphate-EDTA (SSPE)], and 0.1% SDS, and twice with 0.5× SSPE/0.1% SDS. Northern blots were rehybridized after stripping the oligonucleotides used as probes in boiling 0.1% SDS for 10 min. As a control for normalization of RNA expression levels, we hybridized blots with an oligonucleotide probe complementary to the U6 RNA (5′-GCAGGGGCCATGCTAATCTTCTCTGTATCG-3′). Digital images were acquired in the linear range of the scanner Fluor-S MultiImager (Bio-Rad). Intensities of band signals were quantified using the densitometric software Quantity One (Bio-Rad). miRNA amount was normalized with the corresponding U6 RNA in each sample.

Real-time RT-PCR analysis. The expression of mature miRNAs was assayed using the Taqman MicroRNA Assays (Applied Biosystems) specific for hsa-mir-122a (P/N: 4373151), hsa-let-7a (P/N: 4373169), and hsa-mir-221 (P/N: 4373077) on 38 matched HCCs and cirrhosis. The expression level of miR-122a was measured also in HEP3B cells transfected with miR-122a (Ambion). Each sample was analyzed in triplicate. Reverse transcription reaction was done starting from 10 ng of total RNA and using the looped primers. Real-time PCR was done using the standard Taqman MicroRNA Assays protocol on the iCycler iQ Real-Time PCR Detection System (Bio-Rad). The 20 μL PCR included 1.33 μL reverse transcription product, 1× Taqman Universal PCR Master Mix, No AmpErase UNG (P/N 4324018; Applied Biosystems), 0.2 μmol/L Taqman probe, 1.5 μmol/L forward primer, and 0.7 μmol/L reverse primer. The reactions were incubated in a 96-well plate at 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The level of miRNA expression was measured using Ct (threshold cycle). The Ct is the fractional cycle number at which the fluorescence of each sample passes the fixed threshold. The ΔΔCt method for relative quantitation of gene expression was used to determine miRNA expression levels. The ΔCt was calculated by subtracting the Ct of U6 RNA from the Ct of the miRNA of interest. The ΔΔCt was calculated by subtracting the ΔCt of the reference sample (normal liver) from the ΔCt of each sample. Fold change was generated using the equation 2^−ΔΔCt. A pool of three normal livers was used for the standard curve calculation and as reference sample for the ΔΔCt. The Taqman MicroRNA Assays for U6 RNA (RNU6B, P/N: 4373381; Applied Biosystems) was used to normalize the relative abundance of miRNA.

miRNA target prediction. The analysis of miRNA predicted targets was determined using the algorithms TargetScan,⁵

Cell culture. SNU449 [American Type Culture Collection (ATCC) number CRL-2234] and HEP3B (ATCC number HB-8064) cell lines were cultured with Iscove's modified Dulbecco's medium with 10% fetal bovine serum and gentamicin.

Cell transfection with miR-122a. Stability-enhanced miR-122a precursor and negative control 1 ribo-oligonucleotides were from Ambion. The day before transfection, SNU449 and HEP3B cells were seeded in antibiotic-free medium. Transfection of miRNAs was carried out using Lipofectamine 2000 in accordance with the manufacturer's procedure (Invitrogen). The level of miR-122a expression in transfected HEP3B cell line was assayed by real-time RT-PCR (Taqman MicroRNA Assays) 24 h after transfection as described above.

Luciferase activity assay. The human cyclin G1 3′-UTR target site was amplified by PCR using the primers 5′-GCCTCAAACTGAATCCCATC-3′ (CCNG1-3UTR-F) and 5′-ATAAGCTTTTTGGCACAGTAAGGGCATC-3′ (CCNG1-3UTR-R) and cloned downstream of the luciferase gene in pMIR-REPORT luciferase vector (Ambion). This construct, named pMIR-CCNG1, was used for transfection in HEP3B cell line. HEP3B cells were cultured in 24-well plates and each transfected with 0.1 μg of either pMIR-CCNG1 or pMIR-REPORT together with 0.01 μg of pRL-TK vector (Promega) containing Renilla luciferase and 30 pmol of miR-122a or negative control 1. Transfection was done using LipofectAMINE 2000 and Opti-MEM I reduced serum medium (Life Technologies) in a final volume of 0.6 mL. Twenty-four hours after transfection, firefly and Renilla luciferase activity were measured using the Dual-Luciferase Reporter Assay (Promega). Each transfection was repeated twice in triplicate.

Western blot analysis. SNU449 and HEP3B cell lines were transfected in six-well plates with 100 pmol of miR-122a or negative control 1. After transfection, cells were cultured for 72 h and intermediate samples at 24 and 48 h were collected and analyzed by Western blot to assess cyclin G1 expression. A monoclonal antibody (clone 11C8, Novocastra Laboratories) against cyclin G1, diluted at 1:100, was incubated for 16 h at 4°C. A horseradish-conjugated secondary antibody (labeled polymer horseradish peroxidase antimouse, EnVision System, DakoCytomation) was incubated for 45 min at room temperature, and the corresponding band was revealed using the enhanced chemiluminescence method (Amersham). Digital images of autoradiographies were acquired with Fluor-S MultiImager, and band signals were acquired in the linear range of the scanner using a specific densitometric software (Quantity One). Images were calibrated against a reference autoradiography and given in relative density units. After autoradiography acquisition, the membranes were stripped and reprobed for 2 h at room temperature with anti-β-actin antibody (Santa Cruz Biotechnology, Inc.) to normalize protein loading. A ratio between cyclin G1 and β-actin corresponding bands was used to quantify cyclin G1 modulation by miR-122a.

A miRNA expression signature differentiates HCCs from nonneoplastic liver tissue. We analyzed miRNA expression of 17 cases of HCC and 21 cirrhotic liver tissues using a miRNA microarray platform able to assess the expression level of 381 human miRNAs (19, 26). To identify the miRNAs that are differentially expressed between cirrhosis and HCC, we did a statistical comparison between the two groups of samples (combining a 1.3-fold change threshold and ANOVA P < 0.05). Thirty-five miRNAs emerged as differentially regulated in HCC tissue with respect to cirrhosis (Table 2). Based on these differentially expressed miRNAs, supervised hierarchical cluster analysis (by standard correlation) generated a tree showing a good separation between HCCs and LCs (Fig. 1). Two main groups on the opposite sides of the cluster emerged, one predominantly made of HCCs and the other of LCs. Some HCC and cirrhosis samples were included in the wrong group because they tended to cluster based on patient's individual profile instead of tissue type; for example, in the LC group, the HCC sample of patient 29 came together its matched cirrhosis, whereas LCs of patients 23 and 41 in the HCC group clustered together their matched HCCs. Two additional interposed smaller groups, mostly containing HCCs, showed an intermediate variation in the expression of miRNAs, and similarly to above, LCs clustered together their matched HCCs. The HCCs in the different groups did not show significant differences for etiology, focality, size, grading, and α-feto protein expression. To validate their ability to correctly distinguish HCC or cirrhosis tissue types, the 35 deregulated miRNAs were assayed using the algorithms SVM (25) and PAM (24). PAM correctly classified 86% of the samples and SVM classified 91% (Supplementary Table S3). These analyses proved that the miRNA “signature” was able to correctly predict most HCC and cirrhotic tissues and pointed out the more predictive miRNAs (Table 2).

Three of the deregulated miRNAs (miR-221, let-7a-1, and miR-122a) were further assayed by Northern blot and quantitative real-time RT-PCR for validation purposes (Supplementary Table S2). Northern blot analysis was done on a series of 40 HCC-LC patients, which confirmed the results obtained by microarrays. Cases analyzed by both microarrays and Northern blot displayed an identical expression pattern. Overall, Northern blot analysis revealed that miR-221 was up-regulated in 83% of HCCs when compared with matched cirrhotic tissue, let-7a-1 was down-regulated in 70% of HCCs, and miR-122a resulted down-regulated in 70% of HCCs. A selection of results is shown in Fig. 2. A miRNA quantification by RT-PCR was also done using the real-time Taqman assay for miR-122a, let-7a, and miR-221 on 38 paired HCC-LC samples. Results confirmed the down-regulation of miR-122a and let-7a and the up-regulation of miR-221 in tumors compared with cirrhosis (Supplementary Table S4).

CCNG1 is a target of HCC-deregulated miR-122a. Identification of miRNA-regulated gene targets is a necessary step to understand miRNA functions. let-7 and mir-221 were previously shown to be involved in cancer pathogenesis (27–31). Differently, the role of miR-122a, which represents the most abundantly expressed miRNA in human liver, in tumorigenesis remains unclear.

To begin unraveling the potential role of miR-122a in hepatocarcinogenesis, putative human protein-coding gene targets of miR-122a were identified by using miRanda, TargetScan, and PicTar algorithms. Results of the analysis are shown in Supplementary Table S5.

The gene for the cyclin G1, predicted by TargetScan and PicTar (Fig. 3A), was assayed as a target. After testing the HEP3B, SNU182, SNU398, and SNU449 cell lines for miR-122a expression by Northern blot and for cyclin G1 levels by Western blot, the human HCC cell lines SNU449 and HEP3B were chosen for assaying the miR-122a because miR-122a basal expression is undetectable in these cell lines and the 34-kDa cyclin G1 protein is easily detectable. Twenty-four hours after transfection, the level of miR-122a, detected by quantitative real-time RT-PCR, was similar to that of normal livers (fold change was 1.35 in transfected HEP3B versus 1.0 in normal liver; see Supplementary Table S4). Transfection of miR-122a in HEP3B and SNU449 caused a reduction in CCNG1 protein level of 55% and 25%, respectively. Cells transfected with the negative control did not exhibit any change in cyclin G1 levels (Fig. 3B).

To test whether the predicted miR-122a target site in the 3′-UTR of cyclin G1 (CCNG1) mRNA was responsible for its regulation, we cloned the putative 3′-UTR target site downstream of a luciferase reporter gene (pMIR-CCNG1) and cotransfected this vector together with miR-122a or the scrambled Ambion negative control 1 into HEP3B cells. A Renilla luciferase vector (pRL-TK) was used as a reference control. Luciferase activity of cells transfected with miR-122a was decreased ∼2-fold, a statistically significant difference, when compared with vector alone (P = 0.0008, t test) or with negative control (1.5-fold decrease; P = 0.0038, t test; Fig. 3C). Taken together, immunoassay and luciferase data provided strong indications that CCNG1 is a target of miR-122a.

Furthermore, analysis of the expression of miR-122a and cyclin G1 in primary tumors confirmed the existence of an inverse correlation between the expression of miR-122a and cyclin G1 (Fig. 4). These data suggest that miR-122a may play a major role in the control of the level of cyclin G1 in liver tissues.

In the recent years, several studies have shown that expression of miRNAs is deregulated in human malignancies (22, 28, 32). Identification of cancer-specific miRNAs and their targets is critical for understanding their role in tumorigenesis and may be important for defining novel therapeutic targets (18–22).

Here, we report an investigation on miRNA expression in human HCC. We discovered 35 miRNAs differentially regulated in HCC with respect to LC. The expression profile of this miRNA panel was able to discriminate the neoplastic versus the nonneoplastic liver tissue. Several miRNAs, differentially expressed in HCC, were previously found deregulated in other human cancers. Among these, the let-7 family was shown to be down-regulated in various human cancers (22, 30, 31), mir-221 was up-regulated in thyroid carcinomas and glioblastomas (29, 33), and mir-145 was found down-regulated in colon and breast cancers (20, 34–36). Interestingly, miR-122a, a hepato-specific miRNA, resulted down-regulated in the majority of HCCs and in all examined HCC-derived cell lines. The high frequency of aberrant regulation of these miRNAs in HCC versus nontumor liver suggests that they might play an important role in hepatocarcinogenesis.

Other studies investigating the role of miRNA deregulation in human hepatocarcinogenesis have been reported. The microarray-based study by Murakami et al. (37) was done using a two-color approach, with a mixture of five human cell lines as reference. Possibly because of the different approach, none of the above-mentioned cancer-associated miRNAs emerged in their study and there was a limited overlap between our list of 35 and their list of 8 differentially expressed miRNAs: only mir-195 and mir-199a were in common. In addition, because no accession to their raw data was available, it was not possible to compare their primary data with ours. Although based on a limited number of cases, but in agreement with our data, the article by Kutay et al. (38) revealed that miR-122a is significantly down-regulated in human and mouse HCC, confirming that down-regulation of miR-122a may play a role in hepatocarcinogenesis.

Understanding the tumor-promoting mechanism associated with miRNA deregulation remains a difficult task. In fact, although bioinformatic tools may help to reveal putative mRNA targets, experimental procedures are required for their validation. Only few studies have identified oncogenes whose level of expression is regulated by miRNAs: members of the let-7 miRNA family can regulate all three members of the RAS oncogene family (30) and mir-15a/mir-16-1 regulate BCL2 (39). These findings support the idea that miRNA deregulation may be involved in cancer pathogenesis. Here, we show that miR-122a targets the cyclin G1 mRNA, thus revealing a potential mechanism associated with liver tumorigenesis.

In fact, cyclin G1 deregulation is associated with genomic instability (40) and increased cyclin G1 levels have been described in colorectal cancer, breast cancer, and leiomyoma (41–43). Moreover, experimental evidences obtained in cancer cell lines and tumor xenografts have shown that suppression of cyclin G1 results in the inhibition of tumor growth through a reduction of proliferation and induction of apoptosis (44, 45). In experimental hepatocarcinogenesis, loss of cyclin G1 is associated with a significantly lower tumor incidence after carcinogenic challenge and cyclin G1–null hepatocytes enter S phase at a lower rate (46). Cyclin G1 is transcriptionally activated by p53 and p73, and, in turn, it negatively regulates p53 family proteins (47). Taken together, these data suggest that reduced levels of miR-122a in HCC may result in chromosomal instability through deregulation of cyclin G1 and, indirectly, p53-dependent pathways.

The importance of miR-122a for liver physiology is supported by the fact that miR-122a is specifically expressed in normal liver, accounting for 70% of the total liver miRNA population both in human and mouse. Hence, it is conceivable that its deregulation may have a significant effect on various liver functions (23). For example, miR-122a binds to the 5′ noncoding region of hepatitis C virus (HCV) RNA highly conserved in all the six HCV genotypes (23), suggesting that miR-122a is an essential element of the HCV replication-adaptation to the liver. It has been shown that functional inactivation of miR-122a leads to 80% reduction of HCV RNA replication, suggesting that loss of miR-122a in HCC may increase resistance of cancer cells to HCV replication. Furthermore, it has been reported that natural targets of miR-122a include several genes related with the adult liver phenotype and genes involved in cholesterol biosynthesis (48), suggesting that loss of miR-122a expression in HCC may be related to loss of hepatocyte differentiation.

This study helps to define cancer-associated miRNA-deregulated pathways involved in liver diseases, which could help to identify potential therapeutic targets. In addition, because liver can be effectively targeted by in vivo delivery of short RNAs, it represents a good model for the development of anticancer miRNA-based approaches of gene therapy as previously reported for other liver diseases (48, 49). Recognizing the miRNAs that are deregulated in human HCCs represents the first step for the development of experimental therapies of this type.

Grant support: Associazione Italiana per la Ricerca sul Cancro Regional grant, Ministero dell'Università e della Ricerca Scientifica and Comitato Sostenitori Project CAN2006 (M. Negrini), Associazione Italiana per la Ricerca sul Cancro Regional grant and Fondazione CARISBO (L. Bolondi), National Cancer Institute Program Project Grants (C.M. Croce), and Kimmel Foundation Scholar award (G.A. Calin). M. Ferracin is a recipient of a fellowship from Fondazione Italiana per la Ricerca sul Cancro.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.