Hepatocellular carcinoma (HCC) leads to more than 500,000 deaths per year worldwide and has a rising incidence in Western countries despite advanced antiviral therapeutic modalities. More than 85% of HCCs develop because of infections with hepatitis B virus or hepatitis C virus (HCV), chronic ethanol ingestion, or aflatoxin B1 exposure. Various molecular alterations occur in preneoplastic nodules and escalate in HCCs, including dysregulation of well-known molecular pathways in carcinogenesis.1–5 The important role of microRNAs (miRNAs) in those pathways has recently been emphasized. miRNAs are short (∼22 nucleotides) noncoding RNAs that are believed to serve fundamental roles in many biological processes through regulation of gene expression.6–8 miRNAs were first discovered in the nematode Caenorhabditis elegans in 1993.9 Since then, more than 470 human miRNAs have been described, and speculations about the total number of human miRNAs exceed 1000.10–12 miRNAs regulate central physiological cell processes such as apoptosis, proliferation, and differentiation by diverse epigenetic mechanisms and are highly conserved across species.13 Mature miRNAs are formed in a stepwise process from larger transcripts that fold to produce hairpin structures and serve as substrates for the Dicer family of RNase III enzymes.8 One strand of the resulting short double-stranded RNA guides the RNA-induced silencing complex to its target mRNA.13 Strikingly, 50% of the known miRNAs are located inside or close to fragile sites, in minimal regions of loss of heterozygosity, and at common breakpoints associated with cancer.14 Dysregulation of miRNA expression occurs frequently in a variety of carcinomas and has been shown in those originating from the lungs,15–17 colon,16, 18 breast,8, 16, 19 cervix,8 stomach,16 prostate,8, 16 pancreas,16 thyroid,20 and liver.21–23 miRNA expression profiles appear to be tissue-specific and tumor-specific.16 Studies have demonstrated that miRNA signatures may adjust classification, diagnosis, and prediction of outcome in cancer patients.17, 24–26 A comprehensive analysis of the miRNA expression in diverse neoplasms showed a higher accuracy for diagnosis and prediction of tumor behavior for the miRNA genetic fingerprint than for more than 16,000 mRNAs,27 yet much remains unknown about the biological targets of miRNAs. Only a few miRNAs have been functionally annotated, such as miR-372 and miR-373, which neutralize p53-mediated cyclin-dependent kinase inhibition and thus allow tumor growth in the presence of wild-type p53.28 Past miRNA expression studies of liver tumors used small sample sets, which included cell culture material or snap-frozen tissue from a heterogeneous patient population.21, 22 The complex regulatory networks operated by miRNAs29 have thus far not been assessed comprehensively in human HCCs and their corresponding precursor dysplastic nodules (DNs). In the present study, we determined the miRNA expression profiles of HCV-induced HCCs and their precursor lesions in 43 human formalin-fixed paraffin-embedded (FFPE) HCCs and 9 DNs. Furthermore, we elucidated the role of a subset of miRNAs that were dysregulated in tumors herein and might fulfill crucial roles in hepatocarcinogenesis, acting as tumor suppressors or oncogenic factors.
MicroRNAs are small noncoding RNAs that regulate gene expression by targeting messenger RNAs (mRNAs) through translational repression or RNA degradation. Many fundamental biological processes are modulated by microRNAs, and an important role for microRNAs in carcinogenesis is emerging. Because understanding the pathogenesis of viral-associated hepatocellular carcinomas is important in developing effective means of classification, prognosis, and therapy, we examined the microRNA expression profiles in a large set of 52 human primary liver tumors consisting of premalignant dysplastic liver nodules and hepatocellular carcinomas by quantitative real-time polymerase chain reaction. All patients were infected with hepatitis C, and most had liver cirrhosis. Initially, the accessibility of microRNAs from formalin-fixed paraffin-embedded archival liver tissue by real-time polymerase chain reaction assays was shown. Subsequently, target parenchyma from routinely processed tissue was macrodissected, RNA was extracted, and reverse transcription followed by quantitative real-time polymerase chain reaction was performed. Relative quantification was performed by the 2−ΔΔCt method with normal livers as a calibrator. In order to obtain a comprehensive microRNA gene expression profile, 80 microRNAs were examined in a subset of tumors, which yielded 10 up-regulated and 19 down-regulated microRNAs compared to normal liver. Subsequently, five microRNAs (miR-122, miR-100, miR-10a, miR-198, and miR-145) were selected on the basis of the initial results and further examined in an extended tumor sample set of 43 hepatocellular carcinomas and 9 dysplastic nodules. miR-122, miR-100, and miR-10a were overexpressed whereas miR-198 and miR-145 were up to 5-fold down-regulated in hepatic tumors compared to normal liver parenchyma. Conclusion: A subset of microRNAs are aberrantly expressed in primary liver tumors, serving both as putative tumor suppressors and as oncogenic regulators. (HEPATOLOGY 2008.)
Materials and Methods
Hepatoma Cell Lines and Murine Liver Tissues.
The hepatoma cell lines SK Hep-1, Hep 3B, and HepG2 were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany) and maintained in Dulbecco's modified Eagle medium with 10% fetal bovine serum (Sigma Aldrich, Munich, Germany) and penicillin (200 U/mL) at 37°C in a humidified atmosphere with 5% CO2. After approximately 80% cell confluence was reached, cells were harvested for RNA isolation with TRIZOL reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.
In order to test the accessibility of miRNA from FFPE tissues, Friend virus B–type mice were sacrificed, and liver samples of approximately 0.1 to 0.15 cm3 were either immediately snap-frozen in N2 or fixed in 10% neutral buffered formalin for 24 hours and then embedded in paraffin. Samples were used for total RNA isolation by the TRIZOL method or phenol-chloroform extraction, respectively.
Patients and Tumor Characteristics.
All specimens were obtained from the tumor bank of the Institute for Pathology at the University Hospital of Cologne (Cologne, Germany) from 1995 to 2007 and were used in accordance with the policies of the institutional review board of the hospital. The routine glass slides were reviewed, and the diagnoses were confirmed in a blinded fashion and independently by two pathologists (H.V. and U.D.). Only unequivocal cases of DNs and HCCs arising in HCV infection were included in the study in order to obtain a tumor population with comparable biological behavior and little heterogeneity regarding tumor etiology and potential viral interaction with the miRNA expression profiles. DNs are divided into low-grade and high-grade and are defined as nodular regions at least 1 mm in diameter with dysplasia but without definite histologic criteria of malignancy.30
Thus, 52 primary liver tumors from 39 patients induced by HCV infection (Table 1) were included in the study. The distribution of the tumors was as follows: 43 HCCs [8 well differentiated (grade 1), 30 moderately differentiated (grade 2), and 5 poorly differentiated (grade 3)] and 9 DNs (3 low-grade and 6 high-grade).
|Patient No.||Age||Gender||Procedure||Cirrhosis||Normal Tissues and Macrodissected Lesions Taken for miR Profiling||Preoperative Therapy|
|4||70||M||Resection||Yes||HCC G2; cirrhosis|
|6||52||M||Transplantation||Yes||DN LG; DN HG; HCC G1; HCC G2||RAF|
|7||62||M||Transplantation||Yes||DN HG; HCC G2; HCC G3|
|12||80||M||Needle biopsy||Unknown||HCC G2|
|13||70||F||Needle biopsy||Yes||HCC G2|
|17||64||M||Transplantation||Yes||HCC G2; HCC G3|
|18||67||M||Transplantation||Yes||HCC G1; cirrhotic nodule|
|19||80||M||Needle biopsy||Yes||HCC G1|
|20||63||M||Needle biopsy||Yes||HCC G2|
|21||66||M||Needle biopsy||Yes||HCC G2|
|22||64||F||Needle biopsy||Unknown||HCC G2|
|23||74||M||Needle biopsy||Unknown||HCC G2|
|24||49||M||Needle biopsy||Unknown||HCC G2|
|25||63||F||Needle biopsy||Unknown||HCC G2|
|26||77||M||Needle biopsy||Unknown||HCC G2|
|27||78||M||Needle biopsy||Unknown||HCC G2|
|28||57||F||Transplantation||Yes||DN LG; HCC G2; cirrhotic nodule|
|30||49||M||Transplantation||Yes||HCC G2; HCC G3|
|32||61||M||Transplantation||Yes||HCC G1; lung metastasis (28 months later)|
|33||78||M||Resection||Yes||DN HG; HCC G2; kidney metastasis (5 months later)|
|34||62||M||Transplantation||Yes||HCC G1; cirrhotic nodule|
|35||57||M||Transplantation||Yes||DN LG; DN HG; HCC G1|
|36||60||M||Transplantation||Yes||DN HG; HCC G3|
|38||65||M||Transplantation||Yes||DN HG; HCC G3|
|39||54||M||Needle biopsy||Unknown||HCC G2|
|41||74||F||Needle biopsy||Yes||HCC G2|
|42||52||M||Autopsy, 3 years after transplantation||No||HCC G2|
In addition, a metastasis to the lung in a patient with a grade 1 HCC and one to the kidney of a grade 3 HCC were analyzed. One of the HCCs was a hepatic recurrence after liver transplantation. Biopsy material was available from only 13 tumors, and surrounding cirrhosis could therefore not be assessed reliably in all cases. The remaining 39 tumors were removed by liver resection or explantation during orthotopic liver transplantation and arose in the setting of liver cirrhosis in all but one case. In addition, peritumorous cirrhotic liver parenchyma of four random patients was analyzed. Preoperative chemoembolization had been performed in two HCCs, and radiofrequency ablation had been performed in one case.
The mean values of tissue from three different normal livers were used for calibration. All three were cadaveric donor livers that were not used for the intended liver transplantation because of their fat content of 25%-35% in two cases and because of a malignant lung nodule that had not spread to the liver in one case. Patient demographics and details of analyzed tissues are summarized in Table 1.
Tissue Preparation and Macrodissection.
All tissues were FFPE, and sections were prepared and stained with hematoxylin and eosin according to the protocol. Six additional sections of 6-μm thickness were cut, mounted onto glass slides, and used for macrodissection. Tumor areas corresponding to the sample hematoxylin and eosin–stained section were scraped off with a scalpel and collected into plastic tubes.
Total RNA from snap-frozen mouse livers and from hepatoma cells were isolated with the TRIZOL reagent (Invitrogen) according to the instructions of the supplier.
FFPE samples were deparaffinized in xylene by incubation at 65°C for a total of 20 minutes, xylene being substituted twice. After two washes with 100% ethanol, samples were lysed in 200 μL of proteinase K buffer [500 μg/mL proteinase K (Invitrogen), 50 mM trishydroxymethylaminomethane-HCl (pH 7.4), and 5 mM ethylene diamine tetraacetic acid (pH 8)], and total RNA was extracted by phenol/chloroform and subsequently precipitated with 200 mM sodium acetate and isopropanol.31
Reverse-Transcription and Real-Time Polymerase Chain Reaction (PCR).
Extracts of total RNA were measured with the ND-1000 NanoDrop spectrophotometer (NanoDrop, Wilmington, DE), and 35 ng of total RNA was reverse-transcribed in a 10-μL volume with the TaqMan miRNA reverse-transcriptase kit (Applied Biosystems, Foster City, CA) according to the manufacturer's recommendations. Then, 3 μL of the reverse-transcription reaction was used in each of the real-time PCR assays.
For the miRNA screening step, quantitative real-time PCR was performed with the TaqMan miRNA assay–early access kit (Applied Biosystems). Analyses of a subset of miRNAs (miR-10a, miR-100, miR-122, miR-140, miR145, and miR-198) were carried out in triplicates by means of the TaqMan human miRNA assays (Applied Biosystems) on an Stratagene MX3000 thermocycler (Stratagene Europe, Amsterdam, The Netherlands) according to the manufacturer's instructions.
Data Normalization and Statistical Evaluation.
Relative quantification of the miRNA expression was calculated with the 2−ΔΔCt method. In the screening step, data were normalized by global median normalization. During further in-depth analysis of selected miRNAs, data were normalized to the expression of miR-140, which had been found not to be aberrantly expressed in the screening sample set. After normalization, data were transformed as log10 of a relative quantity of the target miRNA with respect to the control sample. In all assays, normal liver parenchyma was used as a calibrator, whereas miRNAs were examined in cirrhotic, dysplastic, and malignant tissue in order to show the stepwise progression of the miRNA profile and to obtain a broad panel of those miRNAs that are involved in the early and late steps of hepatocarcinogenesis.
Statistical analysis was performed with SPSS (Chicago, IL) software. Expression profiles of miR-10a, miR-100, miR-122, miR-145, and miR-198 were presented as box plots. Differences were detected with analysis of variance or Kruskal-Wallis tests when data were not normally distributed. Post hoc multiple comparisons were performed by the Tukey test or Tamhane's T2 test for unequal variances. For all statistical tests, a P value of less than 0.05 was considered significant.
Accessibility of miRNAs from FFPE Tissue.
A large tissue population of 52 FFPE primary liver nodules arising exclusively in HCV infection was available for interrogation of its miRNA expression profile. However, prior to performing real-time PCR assays, we confirmed the accessibility of miRNAs in FFPE liver tissue by comparing snap-frozen mouse liver tissue with FFPE material. As shown in Fig. 1, samples after both conservation methods allow PCR detection of miR-122 in a range of 5 to 50 ng of the required RNA extract. In an additional validation study comparing tissue stored in paraffin blocks for 10, 20, and 30 years, we showed a similar expression level of the ubiquitous miR-16. Thus, PCR validation revealed an adequate quality of miRNA obtained from FFPE tissues regardless of length of storage, which was suitable for quantitative miRNA analyses in this study of HCV-associated primary liver nodules.
Comprehensive Expression Analysis of 80 miRNAs in Primary Liver Nodules.
In order to elucidate miRNA expression in HCCs and their precursor DNs, in a first step, 80 miRNA profiles were analyzed by real-time PCR on a subset of the HCV-positive FFPE liver tissues. The liver tumors consisted of 11 HCCs (n = 2 grade 1, n = 5 grade 2, n = 4 grade 3) as well as 8 DNs (n = 3 low-grade, n = 5 high-grade). Expression levels of the 80 miRNAs were compared to normal liver parenchyma. The subset of samples taken for the screening of putative dysregulated miRNA species is shown in Table 1 (bold type). Aberrant expression of 29 miRNAs in primary liver nodules compared to the normal liver parenchyma was found; 10 of them were more than 2-fold overexpressed, and 19 were less than 0.4-fold underexpressed (Fig. 2 and Table 2).
|miRNA||Mean||Chromosome Location||Potential and Known Targets|
|LG DN||HG DN||HCC G1||HCC G2||HCC G3||Total|
|Increased expression > 2-fold|
|Decreased expression < 0.4-fold|
Extended Analysis of a Subset of miRNAs.
Because the overexpressed miR-122, miR-100, and miR-10a and the underexpressed miR-198 and miR-145 were dysregulated very consistently across all liver nodules, they were further analyzed on an extended sample set of 52 primary tumors (Table 1).
Expression analyses of macrodissected tumor areas showed more than 2-fold overexpression of miR-122 in DNs (P = 0.02) and in HCCs (P = 0.005) compared to normal liver parenchyma (Fig. 3A). Contrary to the primary tumors, metastases of HCCs in the kidney and lung did not show up-regulation. Because previous data have shown down-regulation of miR-122 in hepatoma cells,22 we compared the miR-122 occurrence of the normal livers and primary tumors to expression levels of three different hepatoma cell lines. None of them were derived from HCV-infected liver cells. In Hep3B cells, moderate down-regulation of miR-122 was shown, but in SK-Hep and in HepG2, miR-122 levels were up to 4-fold decreased (Fig. 3A).
In addition to miR-122, miR-100 and miR-10a were up-regulated in primary liver nodules (Fig. 3B,C), although up-regulation was only moderate.
Underexpression of miR-145 was more than 2-fold, and that of miR-198 more than 5-fold; this was highly significant in comparison with normal liver parenchyma (P < 0.001). In addition, we observed a progressive down-regulation of miR-198 and miR-145 from cirrhotic tissue to DNs and further to HCCs of increasing histological grades (Fig. 3D,E). The kidney metastasis of one HCC showed down-regulation similar to that of the primary liver carcinoma samples, but a lung metastasis in a different patient did not reveal aberrant expression of these two miRNAs.
In order to investigate the miRNA expression profile on a comprehensive panel of HCV-associated HCCs and their precursor DNs, we resorted to FFPE archived materials. Because in previous studies limited numbers of frozen primary tumors of different etiologies and hepatoma cells were used,21, 22 the extensive miRNA profiling on a broad tissue panel of consistent etiology is of high impact. However, before we proceeded with real-time PCR assays on human liver tumors, we confirmed the equal accessibility of miRNAs in FFPE material and in snap-frozen tissue by this method (Fig. 1). Xi et al.32 recently demonstrated the stability of miRNAs in colorectal carcinoma specimens for a duration of up to 10 years of archival storage. In addition, it has been shown that formalin fixation of cell culture materials does not affect the quality of different miRNA species.33 Although formalin fixation causes RNA-protein crosslinking, short RNAs including miRNAs may be less affected than other RNA species.32, 33 Therefore, stored FFPE tissues can be used confidently for gene expression profiling studies using real-time PCR, and we were able to expand the miRNA expression profiling on a total of 52 HCV-positive primary tumors.
Our studies of the expression of 80 miRNAs showed an overexpression of 10 miRNAs and underexpression of 19 miRNAs across the spectrum from low-grade and high-grade DNs to well, moderately, and poorly differentiated HCCs. Thus, prior results showing that dysregulation of miRNA expression is a frequent occurrence in diverse types of cancer13 were confirmed. Recently, Murakami et al.22 showed the dysregulation of seven miRNAs in a population of 24 HCCs of different etiologies, but none of those miRNAs matched the ones that we report herein to be overexpressed or underexpressed. One explanation could be the use of real-time PCR in our study as a detection method with high sensitivity, which preferentially recognizes mature and putatively active miRNAs.34 In addition, miRNA was used for normalization of data in the real-time PCR assays described herein, whereas small nuclear RNA or total RNA was used for normalization in hybridization techniques. Furthermore, the Western European descent of our patients (versus Japanese) and the homogeneous population of tumors arising exclusively in HCV infection could have contributed to the differences in the miRNA profiles. Although some authors have compared the gene expression profiles of HCCs with those of normal livers,21 others have chosen to use adjacent nontumorous tissue for this purpose.22, 25 Although a strong genetic similarity between cirrhotic tissue and HCC has been demonstrated in a large set of >19,000 Expressed Sequence Tags (ESTs) before,35 the relationship of miRNA fingerprints in premalignant and malignant liver nodules has not been elucidated thus far. By analyzing selected miRNA expression patterns across the spectrum from normal to cirrhotic, further to dysplastic and finally to malignant liver parenchyma, we were able to show that many of the aberrations of miRNA expression levels in HCCs were already present in peritumorous cirrhotic tissue, albeit to a lesser degree.
The dysregulated miRNAs in our study are located on chromosomal regions that have in the past been linked to hepatocarcinogenesis by comparative genomic hybridization. For example, gains of 17q21-25 and 11q, where the precursor sequences of the up-regulated miRNAs miR-10a or miR-100, miR-125b, and miR-326 are located, have been described36 (Table 2). To the contrary, the down-regulated miR-302b and miR-302b*, which are different miRNAs cleaved from the same stem loop, are located on 4q25, which has been shown to be lost in around 33% of HCCs.36, 37 With the TargetScan and MiRanda databases,38 some of the miRNAs shown to be up-regulated in HCCs and DNs were found to be putative repressors for a panel of targets involved in transcription regulation, such as bromodomain adjacent to zinc finger domain 2A for miR-100 (Table 2). In addition, miR-15a and miR-16 target the antiapoptotic factor B-cell CLL/lymphoma 2 and are overexpressed in HCCs; this supports prior reports that apoptosis is increased in HCCs in comparison with normal livers.39
In a second step of our study, a subset of five miRNAs (miR-122, miR-10a, miR-100, miR-145, and miR-198) were examined in a total of 52 liver nodules arising in HCV infection. miR-122 was strongly up-regulated in dysplastic and malignant liver nodules in the large sample set of our study, and this suggested that it might down-regulate target mRNA of yet to be determined tumor suppressor genes and thus lead to increased tumor growth. miR-122 is specifically and abundantly expressed in hepatocytes and composes approximately 70% of total miRNA in the liver.40–42 Contrary to our findings of miR-122 up-regulation in HCV-associated HCCs, other authors have reported a down-regulation in HCC cell lines42 and rodent HCCs,21 all of which had etiologies other than HCV infection. We confirmed this down-regulation in our HCC cell lines that originated in the non-HCV setting (Fig. 3A). Importantly, miR-122 modulates the expression of HCV RNA by interacting with the 5′-noncoding region of the viral genome,40 and the sequestration of miR-122 in liver cells has been shown to result in a marked loss of autonomously replicating HCV RNAs.41 Because miR-122 closely interacts with the HCV genome and the miR-122 expression pattern in HCV-associated HCCs is directly opposed to non–HCV-infected HCC, we speculate that HCV-infected hepatoma cells are able to circumvent tumorigenic repression of miR-122. Clearly, further studies on the role of miR-122 in HCCs of non-HCV etiologies are needed to fully understand the function of this unique miRNA in the liver. Those data have a potential therapeutic implication, especially because Krützfeld et al.7 have already pioneered a specific down-regulation of miR-122 in a mouse model following injection of chemically engineered anti–miR-122 oligonucleotides (so-called antagomirs).
miR-100 and miR-10a were up to 2-fold up-regulated in primary liver nodules (Fig. 3B,C). This may indicate that miR-100 and miR-10a exert an oncogenic function by the subsequent down-regulation of one or more tumor suppressor genes. The exact target mRNAs or dysregulated downstream proteins for miR-100 and miR-10a are currently not known. miR-10a is located at chromosome 17q21 within the cluster of the homeobox (HOX) B genes and has been postulated to down-regulate HOXA1 by mediating RNA cleavage.24 miR-100 is located on chromosome 11q23-q24-D, which has been shown to be altered in HCCs.36 Its target remains unknown, but it has been shown to be dysregulated in carcinomas originating in the breast, lungs, and ovaries.43, 44
miR-198 and miR-145 were not only underexpressed up to 5-fold but also progressively down-regulated from cirrhosis via DNs to HCCs (low-grade to high-grade) and further to kidney metastasis (Fig. 3D,E). miR-198 and miR-145 may therefore act as tumor suppressors in carcinogenesis. In addition, the expression profile of this subset of miRNAs thus supports a stepwise process of hepatocarcinogenesis from cirrhosis via DNs to HCCs. miR-198 has not been reported to be down-regulated in tumors before. It is located within the 3′ untranslated region of the gene for human follistatin–related protein and targets an unknown mRNA.45 Underexpression of miR-145, which has been postulated to target proteins of the mitogen-activated protein kinase pathway,46 has previously been demonstrated in carcinomas from the lungs, colon, breast, prostate, and cervix.8, 17–19 The progressive down-regulation of miR-145 from lower to higher grade HCCs parallels that seen in breast carcinomas of increasing proliferative index.19 Hence, the loss of miR-145 expression may contribute to the development of carcinomas with increased aggressiveness and proliferative potential.
In conclusion, we have shown herein that miRNA expression analyses by real-time PCR can be performed reliably on routinely processed and stored FFPE human liver tissue and that a subset of miRNAs are aberrantly expressed in HCCs and their precursor nodules, serving as potential oncogenic dysregulators in HCV infections. miR-122 is suggested to play a bivalent role in hepatocarcinogenesis of HCV and non-HCV etiology. Since miR-145 and especially miR-198 levels are prominently decreased in HCC, they have to be considered important tumor suppressors, which might serve as prospective targets of HCC therapy. Further studies on their functional role will shed additional light onto the complex molecular pathways of hepatocarcinogenesis and their diagnostic and therapeutic potential.
We greatly appreciate the excellent technical assistance of Melanie Scheffler, Ali Manav, and Elisabeth Konze.