**4. Global gene expression profiling of liver cancer**

During the past two decades, discoveries on the global gene expression profiling technologies emerged one after the other. Subtractive hybridization, differential display, SAGE, microarrays and more recently next generation sequencing techniques appeared as cutting-edge tools to study genome-wide transcriptional profiling differences in nearly all different types of tissues. With the rapid advances in these technologies, the medicine, particularly cancer genomics, is evolving into numerous dimensions.

The microarrays had a great impact from the way we look at the transcriptome and the way we understand the biology and complexity of it. Microarray expression technologies have allowed the simultaneous analysis of thousands of transcripts that cover nearly the entire genome [25]. Hence, gene expression microarrays, that are providing a comprehensive view of the transcriptional changes that occur during the carcinogenic process, have been applied with great success to the molecular profiling of HCC which has resulted in a much more detailed molecular classification scheme as well as in the identification of potential gene signature sets, molecular biomarkers, prediction of early recurrence and patient survival [26-29].

Over the last decade, numerous studies have applied this technology, and identified a number of candidate genes useful as biomarkers in cancer staging, prediction of recurrence and prognosis, and treatment selection. Considering the complexity of the HCC carcinogenesis many genes may be involved in the initiation and progression of the cancer, and therefore a comprehensive expression analysis using microarray technology has great potential to discover new genes involved in carcinogenesis, as well as may highlight the functional modules and pathways altered in HCC. Indeed, some of the new target molecules that were identified using this technology have been used to develop new serum diagnostic markers and therapeutic targets against HCC to benefit patients.

The first report of cDNA microarray analysis of hepatocellular carcinoma (HCC) by Lau *et al.* [30] studied the gene expression using about 4000 known human genes in 10 pairs of HCC and non-tumorous tissues. Since then numerous studies have been published to date in the context of genome-wide expression profiling of HCC liver. The microarray analyses of HCC highlighted activation of important pathways in liver carcinogenesis, such as winglesstype (WNT), p53, transforming growth factor (TGF)-β, MAPK signalling pathways [31-34] as well as novel genes with altered expression, such as *MARKL1*, *VANGL1*, *PEG10*, *BMAL2*, *HLA-DR*, *GPC3*, and *ROBO1*.

Over the past 10 years, the microarray-based gene expression profiling has been used to identify gene signatures associated with etiological factors, histological phenotypes, and clinical phenotypes, as well as unveiling novel subtypes of HCC previously unrecognized by conventional methods [26, 35-36] . Most cases of HCC originate from chronic liver disease caused by hepatitis viral infection, including hepatitis B virus (HBV) and hepatitis C virus (HCV), exposure to aflatoxin B1 in mold, and alcohol abuse. In this context, gene signatures associated with different etiologies have also been reported [37-39]. Microarray studies indicated that HBV and HCV viral infections lead to the development of liver cancer by different molecular mechanisms [32, 38-39]. Okabe *et al.* analyzed expression profiles of 20 primary HCCs by using cDNA microarrays consisting of 23,040 genes, and compared HBV- with HCV-related HCC [32]. The authors identified a gene signature that is correlated with the infection status, and found that genes that are involved in drug metabolism and carcinogen detoxification were differentially regulated between HCV-based and HBV-based HCC. In another study, Iizuka *et al*. [38] performed genome-wide expression profiling 45 HCC (14 HBV- and 31-HCV-associated) and identified 83 genes whose expression significantly differed between the two types of HCCs. The HBV-associated HCC showed significantly up-regulation of imprinted genes (H19 and IGF2) and genes related to signal transduction, transcription, and metastasis. On the other hand, HCV-associated HCC displayed up-regulation of genes related to detoxification and immune response. Delpuech *et al.* showed that HBV-associated HCC altered different cellular pathways, those controlling apoptosis, p53 signalling and G1/S transition, whereas the HCV-related HCC resulted in an over-expression of the TGF-beta induced gene [31].

Microarray gene expression profiling together with prediction models have been used in numerous studies to identify gene signatures in tumor or surrounding non-tumorus tissues that can predict vascular invasion, metastasis, post-surgical recurrence, survival, and response to therapy. These signatures may aid in identifying patients most likely to benefit from surgery and chemotherapeutic treatment.

Vascular invasion (VI) is an unfavorable prognostic factor for early HCC recurrence. There have been several microarray studies which identified gene signatures that correlated with VI. Ho *et al*. [40] identified 14 genes correlated with VI, which can classify patients with high or low risk of VI development and recurrence after curative hepatectomy. In another study, Budhu *et al*. reported a 17-gene signature expressed in noncancerous hepatic tissues with venous metastasis, capable of predicting recurrence after surgical hepatectomy, with 79% accuracy [41]. Similarly, Wang *et al.* identified a 57-gene signature to predict disease recurrence at diagnosis (84% sensitivity) [42].

Using supervised machine learning methods on the gene expression data, Nam *et al.* identified 240 genes that classified samples into different histological grades, from lowgrade DNs to primary HCC [43]. Kim *et al.* reported 44 genes that can discriminate HBVpositive HCC from non-tumor liver tissues [44]. Iizuka *et al*. [26] reported a gene signature of 12 genes that can predict HCC patients at high risk of early intra-hepatic recurrence (IHR) after curative surgery (93% sensitivity). Similarly, Kurokawa *et al*. [45] identified a 20-gene signature which could predict early IHR after curative resection. In another study, a 3-gene signature (HLA-DRA, DDX17, and LAPTM5) found to be predictor of recurrence after curative hepatectomy, which predicted early IHR with 81% accuracy in the validation group [46].

by conventional methods [26, 35-36] . Most cases of HCC originate from chronic liver disease caused by hepatitis viral infection, including hepatitis B virus (HBV) and hepatitis C virus (HCV), exposure to aflatoxin B1 in mold, and alcohol abuse. In this context, gene signatures associated with different etiologies have also been reported [37-39]. Microarray studies indicated that HBV and HCV viral infections lead to the development of liver cancer by different molecular mechanisms [32, 38-39]. Okabe *et al.* analyzed expression profiles of 20 primary HCCs by using cDNA microarrays consisting of 23,040 genes, and compared HBV- with HCV-related HCC [32]. The authors identified a gene signature that is correlated with the infection status, and found that genes that are involved in drug metabolism and carcinogen detoxification were differentially regulated between HCV-based and HBV-based HCC. In another study, Iizuka *et al*. [38] performed genome-wide expression profiling 45 HCC (14 HBV- and 31-HCV-associated) and identified 83 genes whose expression significantly differed between the two types of HCCs. The HBV-associated HCC showed significantly up-regulation of imprinted genes (H19 and IGF2) and genes related to signal transduction, transcription, and metastasis. On the other hand, HCV-associated HCC displayed up-regulation of genes related to detoxification and immune response. Delpuech *et al.* showed that HBV-associated HCC altered different cellular pathways, those controlling apoptosis, p53 signalling and G1/S transition, whereas the HCV-related HCC resulted in an over-expression of the TGF-beta

Microarray gene expression profiling together with prediction models have been used in numerous studies to identify gene signatures in tumor or surrounding non-tumorus tissues that can predict vascular invasion, metastasis, post-surgical recurrence, survival, and response to therapy. These signatures may aid in identifying patients most likely to benefit

Vascular invasion (VI) is an unfavorable prognostic factor for early HCC recurrence. There have been several microarray studies which identified gene signatures that correlated with VI. Ho *et al*. [40] identified 14 genes correlated with VI, which can classify patients with high or low risk of VI development and recurrence after curative hepatectomy. In another study, Budhu *et al*. reported a 17-gene signature expressed in noncancerous hepatic tissues with venous metastasis, capable of predicting recurrence after surgical hepatectomy, with 79% accuracy [41]. Similarly, Wang *et al.* identified a 57-gene signature to predict disease

Using supervised machine learning methods on the gene expression data, Nam *et al.* identified 240 genes that classified samples into different histological grades, from lowgrade DNs to primary HCC [43]. Kim *et al.* reported 44 genes that can discriminate HBVpositive HCC from non-tumor liver tissues [44]. Iizuka *et al*. [26] reported a gene signature of 12 genes that can predict HCC patients at high risk of early intra-hepatic recurrence (IHR) after curative surgery (93% sensitivity). Similarly, Kurokawa *et al*. [45] identified a 20-gene signature which could predict early IHR after curative resection. In another study, a 3-gene signature (HLA-DRA, DDX17, and LAPTM5) found to be predictor of recurrence after curative hepatectomy, which predicted early IHR with 81% accuracy in the validation group

induced gene [31].

[46].

from surgery and chemotherapeutic treatment.

recurrence at diagnosis (84% sensitivity) [42].

Researchers also used microarrays to identify gene signatures as predictors of survival after surgical resection. Lee *et al.* analyzed the gene expression profile of 91 HCC samples using unsupervised classification approach which divided the patients into two subclasses with significant differences in survival [29]. The authors also identified genes that accurately predicted the length of survival. Functional analyses indicated that genes related to cell proliferation, anti-apoptosis, and cell cycle regulators were found to be predictor of poor prognosis. Other microarray studies reported c-Met- and TGF-beta regulated genes that are highly associated with the length of survival [47-48].

HCC is a challenging malignancy; most cases of HCC are diagnosed in an advanced stage, and, therefore, treatment options are limited. Hence, it is important to diagnose it at early stage. DNA microarray studies attempted to identify markers for early HCC [27, 49]. Recently considerable attention has been placed on global gene expression studies as well as genomic aberrations in order to understand the pathogenesis of HCC, and to look for possible early markers of detection [14, 16-17, 28, 34, 49-51]. Furthermore, combining crossspecies comparative and/or functional genomics approaches from human and animal models of HCC along with genomic DNA copy number alterations enhances the ability to identify robust predictive markers for HCC [13, 36, 52-54]. Thus, characterization of diverse HCC subgroups using the array technologies together with improved analytical approaches are crucial for better management of the disease, especially in the era of personalized medicine approach in HCC treatment.
