Molecular profiling of hepatocellular carcinomas by cDNA microarray

Abstract

Hepatocellular carcinoma (HCC) is one of the most common cancers in the world. Conventional diagnosis and treatment of this malignancy have been dismal and should be complemented by novel tools. The development and progress of HCC are believed to be caused by the accumulation of genetic changes resulting in altered expression of thousands of cancer-related genes, which can be measured by globe genetic analysis. Gene expression profiling of HCC has been employed to elucidate hepatocarcinogenesis and disclose molecular mechanisms underlying complex clinical features. Identifying phenotype-associated genes/profiles has impacts on current diagnosis and management strategy of HCC. In spite of some pitfalls of this technology and challenges in improving the research process, scrutinous validation of profiling data of HCC combined with other approaches will eventually benefit the patients.

INTRODUCTION

Liver cancer is one of the most common causes of cancer deaths in the world[1,2]. Last decade statistical data show that HCC is the second major cause of cancer death in men and the third in women in Mainland China[3]. Viral hepatitis B (HBV) and C (HCV) and aflatoxin are major risk factors[4-6]. Chronic HBV infection has a strong association with hepatocellular cancer in China[7]. The early detection, treatment and prevention of this disease are disappointing and most patients are diagnosed in their late stages and management is unsatisfactory with high mortality and recurrent rates[3].

The successful clinical management of this human malignancy requires novel diagnostic and prognostic methods. At present, histopathologic evaluation of a tumor and tumor staging are the mainstays for guiding therapeutic interventions and predicting outcomes. However, the limitations of conventional methods are obvious. Tumors with identical histopathologies may progress differently, respond differently to therapy, and may be associated with different clinical prognosis, suggesting that additional parameters should be identified to predict disease outcomes[8,9]. Gene expression profiling of certain cancers may be able to serve as a complementary tool providing useful information[8].

Like most solid tumors, the development and progression of HCC are believed to be caused by the accumulation of genetic changes resulting in altered expression of cancer-related genes, such as oncogenes or tumor suppressor genes, as well as genes involved in different regulatory pathways, such as cell cycle control, apoptosis, adhesion and angiogenesis[6,10]. Because gene expression profiles provide a snapshot of cell functions and processes at the time of sample preparation, comprehensive analysis of the gene expression patterns of thousands of genes in certain tumor cells and comparison to the expression profile obtained from healthy cells and/or other cancer cells of different phenotype should provide insights into the consistent changes in gene expression that are associated with tumor cellular dysfunction and concomitant regulatory pathways. Current cDNA microarray technology enables investigators to measure the expression of thousands of mRNAs simultaneously in a biological specimen and therefore may provide comprehensive information for diagnosis and therapeutic interventions of tumors in the future[11].

Microarray technology now allows scientists and clinicians to identify qualitative and quantitative changes at RNA level in the development and progress of cancer[11-13]. In general, its HCC-oriented clinical applications can be categorized into three purposes: (1) to define the molecular profile of HCC distinct from non-cancerous liver and other tumor types to identify tumor-specific genes and liver-specific genes; (2) to describe gene expression profiles that correlate with clinical subsets, which will be helpful to disclose underneath mechanisms of HCC development and progress; and (3) to identify tumor-specific and clinical feature-related molecular markers, which will be helpful for cancer diagnosis, prediction of prognosis and response to treatment[8,14-16].

GENERAL GENE EXPRESSION PROFILE OF HCC

Numerous researchers have used microarray to profile gene expression pattern of HCC. Generally, the carcinogenesis and progression of HCC involve thousands of genes. Most of them are either “liver” specific or “cancer” specific. The genes characterizing HCC can mainly be categorized as those correlated with changes of liver function and differentiation, those correlated with dysregulation of signal pathways and those with tumor cell invasiveness.

Change of liver function and differentiation in HCC

HCC cells usually lose their normal function and differentiation. A large number of hepatocyte-specific gene products participating in the metabolism of nutrient factors and those responsible for the liver-synthesized proteins and detoxification enzymes are down-regulated[17-19]. Liver-synthesized functional proteins such as albumin, transferrin, and coagulation factors decreased. Enzymes implicated in biotransformation such as cytochrome P-450, metallothionein families and the glutathione S-transferases also decreased. HCC cells are distinguished for increased glycolysis as seen in many cancers with up-regulated 6-phosphofructokinase-1[17]. All these alterations reflect de-differentiation of cancer cells.

In addition to the overexpression of alpha-feto protein (AFP), dozens of gene markers reminiscent of embryonization of hepatocytes are also noted[17].

Dysregulation of pathways in HCC and non-cancerous liver

Recent studies of cDNA microarrays suggest that heterogeneous hepatocarcinogenetic pathways exist[17-21]. These pathways associated with cell proliferation, cell cycle, apoptosis, and angiogenesis are dysregulated in carcinogenesis, such as p53 in cell cycle regulation, wingless (Wnt) signaling and MAPK pathways in signal transduction, cellular adhesion, and the TGF-beta/ insulin-like growth factor (IGF) axis.

In cell cycle control, negative regulators such as p27, p53, and p10 are less expressed, many cyclins and cyclin-dependent kinases are overexpressed. The combination of these events might drive the hepatocytes into cell proliferation[17]. Activation of Wnt signaling through mutations in beta-catenin (CTNNB1) contributes to the development of HCC and hepatoblastoma[22]. In the Wnt-β-catenin pathway, expressions of β-catenin and Wnt 2b were seen in some patients[17]. Aberrant expression of mitogen-activated protein kinase (MAPK) and associated proteins are also involved in the development of HCC[17,18,23]. The expression of insulin-like growth factor binding protein 3 (IGFBP3) was found to be down-regulated in HCC[18,21]. IGFBP-3 was reported to be a growth suppressor in various pathways. In the IGF receptor-dependent pathway, IGFBP3 mediates a wide variety of growth suppression signals such as TGF-β, retinoic acid, TNF-α and p53. Because retinoid is an accepted therapy to induce differentiation of cells in acute promyelocytic leukemia and is thought to help prevent development of HCC[24], reduced expression of this gene may play a crucial role in hepatocarcinogenesis. Apoptosis-related genes are reduced in HCC and in moderately poorly differentiated tumors, implying that a reduced rate of apoptosis is a major characteristic of tumor progression[17,18]. The abnormality of these pathways, therefore may represent a network required for the multistep process in HCC development.

In non-cirrhotic tissues, the down-regulation of a protein kinase C pathway inhibitor and the up-regulation of a PKC-regulated gene suggest activation of the PKC pathway[21].

Other genes changed in HCC

Other genes also show consistently elevated expression in HCC. These genes are implicated in a variety of cellular processes, including cell signaling, transcriptional regulation, RNA splicing, protein degradation, and cell adhesion[18,19]. Genes regulating the composition of the extra cellular matrix and the cytoskeleton such as fibronectin, tubulin alpha1, matrix metalloproteinase 14, osteonectin SPARC, Rho A are also found to be up-regulated[21]. These genes play important roles in cell motility and invasion.

GENE EXPRESSION PROFILES ASSOCIATED WITH CLINICAL FEATURES

Gene expression profiles generated by microarrays can be used to understand carcinogenesis and to outline clinical features among patients with histologically indistinguishable tumors[16]. Although considerable heterogeneity in gene expression profiles in HCC existed, a link could be seen between certain gene expression patterns and pathological features and some clinical features of HCC, including hepatitis virus status, clonal derivation, vascular invasion, and metastasis[19,21,25,26]. Analysis of gene expression patterns can be used as a classification tool to categorize cancer into various clinically relevant subgroups, which is currently impossible by other methods. These sub-categories often have distinct prognostic significances.

Expression patterns correlated with hepatitis virus status

Comparison of expression profiles between HBV-positive and HCV-positive HCCs implies that hepatitis viruses affect expression of dozens of genes in HCC in a type-specific manner, invoking partly different mechanisms of carcinogenesis. The genes differentially regulated in HBV- and HCV-positive HCCs are related to signal transduction, transcription, metastasis, and immune response[18,21,23].

Expression patterns show that the most differentially expressed genes are enzymes responsible for detoxification, which metabolizes carcinogens and/or anticancer agents[18,23]. One example is aldo-keto reductase family member AKR1C4, which encodes a key molecule for activating chemotherapeutic drugs or detoxifying xenobiotic carcinogens. Some genes exclusively overexpressed in HCV-positive HCCs are the phase I enzymes that convert several pro-carcinogens to activated metabolites. In contrast, expression of some genes of this cluster is preferentially repressed in HBV-positive HCCs while expression levels of the same genes are unchanged in most HCV-positive HCCs, which might reflect enhanced exposure of hepatocytes to activated carcinogens or radicals in HBV-related hepatocarcinogenesis[18].

Some exceptions convey important clinical meanings. Among detoxification related genes, GSTP1 is exceptionally up-regulated in HBV-positive HCC, although its mRNA level is relatively higher in both types of HCCs than in non-cancerous liver. Because GST conjugates reactive oxygen intermediates that are generated by many anticancer agents, the data suggest the efficacy of these anticancer agents in treating HCV-positive HCC and their limitations in treating HBV-positive HCC[18].

HBV- and HCV-positive HCCs exhibit involvement of different cellular pathways. TGF-β-induced encoding gene was found to be up-regulated in 44% of HCV-positive HCCs, suggesting the modulation of the TGF-beta pathway in HCV-positive HCC[21]. Imprinted genes IGF2 were found to be up-regulated specifically in HBV-positive HCC in comparison to both HCV-positive HCC and non-cancerous liver. IGF2 is known to be a mitogen often overproduced in tumors. This result suggests that up-regulation of the IGF-2 pathway may play an important role in the pathogenesis of HBV-positive HCC but not HCV-positive HCC[23].

In general, the distinct expression profiles of HBV- and HCV-positive HCCs suggest different hepatocarcinogenesis mechanisms and provide novel tools for diagnosis and treatment of HBV- and HCV-positive HCCs.

Expression patterns correlated with vascular invasion in HCC

Vascular invasion frequently occurs in HCC patients and seems to have an important role in tumor invasion and metastasis[27,28]. By microarray analysis, we identified genes correlated with the presence or absence of vascular invasion. Metalloproteinase 14 was found to be up-regulated on this list. The association of MMP14 expression with vascular invasion highlights the possible importance of MMPs in the progression of HCC and underscores their potential as therapeutic targets. Most of the genes that are expressed at lower levels in tumors with vascular invasion are “liver specific”, consistent with the classical pathological observation that poorly differentiated HCC tumors tended to be more aggressive and invasive. One of the few genes in this group that does not belong to the liver-specific cluster encode the metalloprotease ADAMTS1, which was recently shown to inhibit endothelial cell proliferation and to have antiangiogenic activity, warrants further investigation[19]. Genes associated with vascular invasion also involve small GTPase-related genes such as Rho C, Rho GAP8 and ARHGEF6, which are preferentially down-regulated in invasive tumors. Because the small-GTPase Rho family plays important roles in controlling cell motility and focal adhesions, alterations of their signaling pathways can enhance the migratory and invasive capacity of tumor cells and induce tumor invasion and metastasis.

Expression patterns correlated with immune response

Cancer-testis (CT) genes are expressed in a variety of human cancers, but not in normal tissues except for testis. These genes are recognized by cytotoxic T lymphocytes (CTLs) and represent promising targets for immunotherapy and gene therapy[29]. The most frequently expressed CT genes in HCC are SSX-1 and GAGE. Most HCC cases express at least one CT gene[30]. Okabe et al[18] found that expression of CT antigens such as MAGEC1, was repressed in poorly differentiated tumors. Reduced expression of genes encoding immune targets may confer a growth advantage by allowing tumor cells to escape from immune surveillance.

Clonal derivation of HCC elucidated by gene expression analysis

In HCC, multifocal growth may be due to intrahepatic metastatic spread or multicentric origin of clonal neoplasms[31]. This issue is of potential clinical and prognostic importance because the survival rate in the independent multicentric occurrence group was significantly better than that in the intrahepatic metastasis group[32]. These two kinds of nodules share similar microscopic features, and reliable differentiation could not be achieved using clinical or morphological criteria alone. Clonality determination using DNA fingerprinting with loss of heterozygosity (LOH) assay, comparative genomic hybridization (CGH), fluorescence in situ hybridization (FISH) and HBV integration pattern are reliable methods[9,33,34]

Microarray can also provide information on whether the tumor is multicentric in origin or single clonality followed by intrahepatic metastasis[35]. Analysis shows that all the tumor samples from each patient shared some similar gene expressions. Multiple clone-related tumor samples in the same patients could show different gene expression patterns due to divergent histories of mutations or chromosomal alterations. Overall, each independently arising tumor could be distinguished from other tumors of the same pathological type, whether they arise in the same patient or different patients, by a distinctive gene expression program that reflects the cell of origin and the unique sequence of genetic events[19,35].

Expression patterns correlated to HCC progress and metastasis

HCC associated with chronic liver disease evolves from pre-cancerous lesions and early HCC to advanced HCC. Nodule-in-nodule-type HCC (advanced HCC within early HCC) represents the transition from early to advanced HCC and, therefore, is useful in molecular genetic analysis of HCC progression during multistage carcinogenesis[36]. By microarray profiling, heat shock protein 70 (HSP70) has been identified as the most abundantly up-regulated gene in early HCC components. Further immunohistochemical examination of HSP70 revealed its significant overexpression in early HCC compared with pre-cancerous lesions and in advanced HCC compared with early HCC. Thus, HSP70 can be a sensitive marker of HCC progress, distinguishing early HCC from pre-cancerous lesion or non-cancerous liver[37].

HCC is one of the most aggressive human malignancies and its high mortality rate is mainly a result of intrahepatic metastases[3]. Ye et al[26] analyzed the expression profiles of HCC samples without or with intra-hepatic metastases by cDNA microarray-based gene expression profiling. Their results showed that the intra-hepatic metastatic lesions were indistinguishable from their primary tumors, regardless of tumor size, encapsulation and age of patients. But primary metastasis-free HCC was distinct from primary HCC with metastasis. These data indicate that changes favoring intra-hepatic metastasis are initiated in primary HCC.

By using a supervised machine-learning classification algorithm known as compound co-variate predictor (CCP), osteopontin (OPN) was identified with an average of 3-fold over-expression in HCC with thrombosis in the portal vein, but not in metastasis-free HCC. OPN is a secreted phosphoprotein highly expressed in patients with metastatic breast tumors and malignant lung, colon, and prostate tumors. An osteopontin-specific antibody can effectively block HCC cell invasion in vitro and inhibit pulmonary metastasis of HCC cells in nude mice[26]. Two recent papers also report OPN in HCC by differential display analysis[38,39] and one of them report that OPN mRNA overexpression is correlated closely with high-grade, late-stage, and early tumor recurrence, which lead to poorer prognosis and suggest OPN overexpression might serve as an unfavorable prognostic factor and a useful marker for predicting early recurrence in early-stage HCC[39]. Thus, osteopontin acts as both a diagnostic marker and a potential therapeutic target for metastatic HCC.

In recent reviews, breast-cancer metastasis suppressor 1 (BRMS1), KAI1, KISS1, NM23, E-cadherin, and TIMPs have been classified as metastasis suppressors, which by definition inhibit metastasis without blocking primary tumor growth[40,41]. In a study by Cheung et al[35], expression profiles of the primary nodules were compared with their corresponding intrahepatic metastatic nodules after determining clonality of all tumor nodules. A total of 90 clones were found to be correlated with intrahepatic metastasis[35]. Among these genes, BRMS1 has been shown to possess the functional capability of decreasing the metastatic potential of breast cancer cells[42] and warrants further investigation in liver cancer. Two other genes are CD53 and the EMP3, membrane proteins consisting of four transmembrane domains possess an important feature of KAI1, an accepted metastasis-related gene associated with lymph node or distant metastases in many cancer types[43,44]. These metastasis suppressors may provide novel therapeutic targets for the treatment of metastasis.

A direct way to identify metastasis-related genes is to compare gene expression profile of human HCC cell lines with different metastatic potentials. Based on gene expression profiles of MHCC97-L and HCCLM3, two HCC cell lines with similar genetic background but different in spontaneous metastatic potentials, were described by Li et al[45,46]. Twenty-five differentially expressed genes were found, including the decreased expression of cell cycle control genes Rb2, mismatch repair gene hMSH2, and PKC beta 2 and 7 and increased expression of MAPK kinase 6, etc[45,46]. These genes can be considered as potential markers to predict metastasis and targets for anti-metastasis intervention.

CLINICAL APPLICATION OF GENE EXPRESSION PROFILING IN HCC

Early detection biomarkers

The development of HCC is generally preceded by chronic liver damage leading to cirrhosis. Screening liver cancer in patients at high risk by AFP and imaging diagnostics are conventional approaches for early detection. However, the cost effectiveness has long been debatable[47]. Half the HCC patients are AFP negative[48]. More appropriate molecular markers are needed to improve better early detection.

A major aim of the studies that profiled RNA expression patterns of human tumors is to identify gene products that may serve as useful biomarkers for cancer diagnosis[8]. Several research groups have focused on identifying subsets of genes that show differential expression between normal tissues or cell lines and their tumor counterparts to identify biomarkers. Xu et al[17] reported overexpression of dozens of gene markers reminiscent of embryonization or dedifferentiation of hepatocytes in addition to the overexpression of AFP in HCC. These genes include CD34, erythropoietin receptor, myeloid cell nuclear differentiation antigen (MNDA), early development regulator 2, and placental protein 15 (PP15), which are associated with hematopoiesis or embryonic development and hence are markers of fetal liver. Although most of these features may reflect the consequences of the cell transformation and thus are unlikely to play an essential role in carcinogenesis, some of them may be used as clinical diagnostic markers.

An alternative is to screen secreting molecules released by cancer. By sequence analysis for transmembrane domain or signal peptides, we were able to distinguish a subset of these genes whose products are likely to be either membrane-bound or secreted[19]. Diehn et al[49] developed a centrifugation-based methodology to separate nuclear and cytosolic mRNA transcripts encoding membrane-associated or secreted versus cytoplasmic proteins. These mRNA pools were then reversely transcribed, PCR amplified, sequenced, and spotted onto glass microarrays. The use of these methods may facilitate the identification of new biomarkers in the form of novel secreted proteins that may be detectable in serum or other body fluids. By raising antibodies against the products of these genes, scientists could identify new serum markers for detection and diagnosis of HCC. They may also facilitate the identification of new therapeutic targets in the form of membrane-bound tumor antigens that may help target chemotherapeutic agents to particular tumor tissues.

Differential diagnosis of liver-occupying lesions

Diagnosing liver-occupying tumors can be difficult in the setting of a poorly differentiated tumor or tumors with no known prior malignancy, especially when only fine needle aspiration biopsy is available. Differential diagnosis is of great importance when treatment is considered. Frequently, AFP and evidence of chronic hepatitis virus infection have been employed by clinicians to distinguish HCC from metastatic adenocarcinoma of neighboring gastrointestinal origin. However, these methods have their limitations. Microarray analysis can tell the origin of cancer cell[25,50].

Microarray analysis shows distinct gene expression profile of HCC from profiles of other metastatic lesions of extrahepatic origin. HCC samples and metastatic cancers can be clustered into two distinct groups, based on differences in their patterns of gene expression. The clustered liver-specific genes are consistently expressed at higher levels in HCC than in tumors of non-liver origin. Metastatic cancers originating from the same tissue are typically clustered together, expressing genes characteristic of the cell type origin. Thus, systematic comparison of the gene expression pattern in a metastatic tumor of uncertain origin observed in a large, diverse sample of tumors and normal tissues will allow reliable recognition of the primary tumor origin[19].

Predicting patients’ survival

Recent advances in gene expression profiling technology have now made it feasible to use microarray technology in the routine management of cancer patients. Gene expression arrays have been demonstrated to be capable of predicting the therapy response, prognosis and survival of patients with various cancer types[16]. However, only few clinical applications of microarray in predicting HCC have been conducted.

Using a newly constructed predictive system consisting of 12 genes, Iizuka et al[51] found that the system correctly predicted early intrahepatic recurrence or non-recurrence with a positive predictive value of 88% and a negative predictive value of 95% in 27 cases within 1 year after curative surgery and suggested this system could serve as a new method for characterizing the metastatic potential of HCC.

In a gastric cancer project, we analyzed gene expression patterns in human gastric cancers and found PLA2G2A, a gene previously implicated as a modifier of the Apc (Min/+) (multiple intestinal neoplasia 1) mutant phenotype in mice, was of especially high variation in the expression level among tumors and significantly correlated with patient survival[52]. We are now attempting a similar study on HCC and hoping to find some new markers.

CHALLENGES

The potential of gene-expression profiling as a novel tool to improve diagnostic and prognostic prediction is very exciting. However, several challenges need to be addressed before routine clinical application. Some of these relate to the technology itself, others concern clinical utility.

The number of microarray-based studies identifying new genes or molecular pathways involved in tumor classification, cancer progression, or patient outcome is growing exponentially. There is a growing need for a standardization of the methodology, so that different datasets may be compared directly and meaningfully[13,53,54]. The technical challenges also include data interpretation. Two systems of databases, one containing expression data and the other containing annotation data are essential to mine the data for the research community[55].

In clinical aspect, major limiting factors for routine use of microarray in a clinical setting at present are cost and access to the microarray technology. It is likely that costs will decrease in the near future and that the technology will become increasingly user friendly and automated[11].

In sample preparation aspect, gene expression analysis using total RNA of bulk tissue usually cannot assign specific messages to particular cell types. Many microarray analyses including ours found distinct patterns of gene expression could provide molecular signatures of diverse cell types, including stromal cells, endothelial cells and infiltrating lymphocytes[19,21]. Cell-specific RNA expression profiling, may be crucial for a better understanding of the role of each distinct cell type within a physiological or pathophysiological setting. RNA profiling based on laser-controlled microdissection (LCM) of defined cells of a tissue now provides a useful tool for studying molecular crosstalk between different cell types within a tissue. LCM technique and suitable RNA amplification will potentially benefit the microarray studies[56].

Finally, validation of candidate genes obtained from genome-wide microarray analysis is critical[57]. Several approaches can be chosen, from basic Northern blotting or semiquantitative reverse transcription-PCR to in situ hybridization (ISH) using tissue microarrays. For most cases, immunohistochemistry (IHC) should be performed if antibodies are available, for only the protein is the ultimate functional form in cells. Two recent examples described flowcharts validating candidate genes dysregulated in HCC[58,59]. Further more, gene expression profiling may need to be combined with other global approaches such as proteomics for optimization of its potential[53].

DISCUSSION

Gene expression profiles obtained by cDNA microarrays may help elucidate initiation, promotion, and progression of HCC. Identifying phenotype-associated genes/profiles has impacts on current diagnosis and management strategy of HCC. Despite pitfalls of technology and some existing challenges in research, microarray can provide reliable and productive profiling that cannot be easily reached with other methods. Clinicians are now approaching the era to initiate “bench to bedside” translation, during which the diagnostic, prognostic, and treatment response biomarker genes identified by microarray screening will be interrogated to provide personalized management of patients[11]. Coupled with more conventional biochemical analyses such as IHC and ELISA, microarrays will result in improvement of diagnostic and prognostic capability in cancer management.

In addition, gene expression profiling analysis also identifies cancer phenotype-associated genes represented by ESTs of unknown function. These genes may be useful as novel tumor markers and drug targets. Their functions warrant further investigations.

ACKNOWLEDGEMENTS

We thank Professor Qin-Hai Ye, Liver Cancer Institute & Zhongshan Hospital, Fudan University, Shanghai, China for his kind assistance in preparing this manuscript.