**3. Genomic alterations in liver cancer**

Current advances in microarray technologies has resulted in high-dimensional genomic data sets and profoundly improved our understanding of genomic imbalances in the context of its role in carcinogenesis; first, with the introduction of copy number variation (CNV) concept in addition to single nucleotide polymorphisms (SNP), and second, with the improved mapping of such CNVs throughout the whole genome of patients versus normal individuals. While very early observations have identified CNVs as "chromosomal polymorphisms" that are several megabases in size, the lower end of the size range of CNVs continues to drop that is consistent with the pace of technological advancement [6]. With their inclusion of coding genes, it is hardly surprising that CNVs play a role in human health and disease, although their role is only recently being recognized, first in the context of Mendelian disorders and more recently in complex diseases [7-8]. The presence of these polymorphisms, either at small (SNPs or mutations) or large (CNVs and CNAs) scale as well as regions comprising loss of heterozygosity (LOH) blocks are, therefore, likely to contribute to cancer formation [9-11]. The genomic modifications in a tumor represents a structural fingerprint that may include the transcriptional control mechanisms and locally impact gene expression levels [10, 12].

Initial array platforms utilizing either spotted clones inserted in bacterial artificial chromosomes (BACs) or in situ synthesized oligos on chip surfaces have been applied to HCC samples to better understand the role of genomic aberrations at the DNA level. These microarray-based assays are called array comparative genomic hybridization (aCGH)

The advances in high-throughput "omics" technologies such as genomics, transcriptomics, proteomics, and metabolomics combined with the availability of highdensity microarrays and low-cost high-throughput parallel sequencing technologies and their analyses using different bioinformatics tools and algorithms are providing unprecedented biological insights related to HCC. Global molecular profiling studies of HCC are providing a comprehensive view of genomic aberrations and expression changes that occur during the carcinogenic process. Hence, the knowledge gained from continuing research efforts on HCC undoubtedly facilitates the understanding of molecular mechanism of HCC pathogenesis, and to provide the best therapy for each cancer patient and to improve patient management. This approach will create a foundation for personalized therapeutics and treatments and expectantly will be available in the near future alongside the unprecedented advancement of next-generation sequencing technologies. These technologies already began to identify novel genes that may have a driver force for HCC pathobiology [5]. Identification of such driver genes within each tumor will highly likely be a source for the development of novel therapeutic targets for

Our aim in this chapter is to focus on the current advances in the genomics field of HCC as well as recent progress using next-generation deep sequencing technologies, and the current shift towards integrative approaches using data from these advanced technologies that will help better understanding of HCC and for the development of novel biomarkers and cancer

Current advances in microarray technologies has resulted in high-dimensional genomic data sets and profoundly improved our understanding of genomic imbalances in the context of its role in carcinogenesis; first, with the introduction of copy number variation (CNV) concept in addition to single nucleotide polymorphisms (SNP), and second, with the improved mapping of such CNVs throughout the whole genome of patients versus normal individuals. While very early observations have identified CNVs as "chromosomal polymorphisms" that are several megabases in size, the lower end of the size range of CNVs continues to drop that is consistent with the pace of technological advancement [6]. With their inclusion of coding genes, it is hardly surprising that CNVs play a role in human health and disease, although their role is only recently being recognized, first in the context of Mendelian disorders and more recently in complex diseases [7-8]. The presence of these polymorphisms, either at small (SNPs or mutations) or large (CNVs and CNAs) scale as well as regions comprising loss of heterozygosity (LOH) blocks are, therefore, likely to contribute to cancer formation [9-11]. The genomic modifications in a tumor represents a structural fingerprint that may include the transcriptional control mechanisms and locally impact gene

Initial array platforms utilizing either spotted clones inserted in bacterial artificial chromosomes (BACs) or in situ synthesized oligos on chip surfaces have been applied to HCC samples to better understand the role of genomic aberrations at the DNA level. These microarray-based assays are called array comparative genomic hybridization (aCGH)

the malignancies for each HCC-affected individual.

**3. Genomic alterations in liver cancer** 

therapeutics targets.

expression levels [10, 12].

technique since they are a modified version of the comparative genomic hybridization (CGH) approach applied to microarrays. Numerous studies have investigated chromosomal alterations associated with HCC using both CGH and aCGH techniques (as reviewed by Moinzadeh *et al.* [9]). Moreover, two leading microarray companies have developed similar assays containing only SNP probes. This approach was initiated by Affymetrix Inc. then later applied by Illumina Inc. Both companies have come up with different SNP assays comprising different numbers of unique SNP probe sets. While the aCGH approach provides much higher resolution over standard microscope-based banding techniques in terms of cytogenetics analysis, SNP arrays bring two main advantages over the other techniques including aCGH: LOH and uniparental disomy detection, and more diverse applications, such as utilization in association studies based on both SNP as well as CNV calls.

Later, higher density arrays having hundreds of thousands or even currently more than a million unique probe sets targeting CNVs and SNPs have been employed in HCC research and identified critical regions of the genome likely to be involved in molecular carcinogenesis of HCC. Such critical regions commonly exhibit either deletion or increased gene dosage, leading to changes in DNA copy number variations/polymorphisms (CNVs/CNPs), aberrations/abnormalities (CNAs) or contain LOH blocks in various cancers, including HCC [9, 13-15].

It is plausible that such HCC-specific CNVs and LOH blocks spanning from several kilobases to megabases comprise critical driver genes that may play a leading role in hepatocarcinogenesis and contain the genetic factors involved in HCC [14, 16-17]. In one of those early studies, Luo *et al.* utilized an integrated approach of DNA and RNA level analyses for HCC, and investigated overlapping genome-wide transcriptomic and genomic alterations among hepatocellular carcinomas (HCC), hepatoblastomas (HPBL), tissue adjacent to HCC and normal liver tissue derived from normal livers and hepatic resections [14]. In their study, genomic imbalances between 27 HCC samples and matching normal controls were determined using low density oligonucleotide arrays. The results indicated that several regions on chromosome 7, 8, 10 and 12 harbor numerous genomic aberrations. Further investigations revealed that many of these changes do not cause remarkable gene expression alterations. However, among other genes, two genes, GPC3 and TIEG, were found to have significant correlation between their copy numbers and expression changes. Further investigations of these two genes in a larger cohort (484 hepatic tissue and normal samples) confirmed the expression differences in HCC samples. Additional studies investigating the role of GPC3 expression in poorer clinical outcome revealed this gene may have a possible role on HCC aggressiveness and therefore may predict the HCC outcome [14].

In a more recent study, Chen *et al.* employed Affymetrix's 500K SNP arrays with an average of 6 kb distance between its unique SNP probes to examine 13 different HCC cell lines in addition to some other cancer cell lines as well as 45 archived primary HCCs [18]. Numerous common and novel aberrations were observed in multiple cancer lines confirming previously known HCC-related cytogenetic regions detected by lowresolution methods and refining their breakpoints and boundaries, and also introducing previously unknown critical genomic regions associated with HCC. Among 653 amplicons and 57 homozygous deletions (HDs) detected by the arrays using different cell lines, 126 amplicons and 6 HDs were selected and tested to identify novel HCC-related genes. Further analysis of such aberrant regions yielded two genes, FNDC3BB and SLC29A2, consistently up-regulated in multiple HCC data sets. Knock-down studies using short hairpin RNAs targeting both genes showed decreased cell proliferation, tumor formation, and anchorage-independent growth in xenograft models in nude mice confirmed a possible pivotal role of these genes in growth and tumor formation in subsets of HCC samples. Up-regulation of either gene is proposed to be activated through STAT3 signaling pathway which is a well-known phenomenon in HCC progression usually triggered by cytokines such as interleukin-6 [19-22].

In another study, Clifford *et al.* used Affymetrix SNP 6.0 assay comprising probes for detection of CNVs and SNPs, each has more than 900,000 unique oligos, totaling nearly 1.9 million probe sets [23]. In their study, a large number of samples exceeding 1100 cases including histopathologically confirmed HCC and liver cirrhosis (LC) samples as well as normal controls with Korean and Chinese ethnicity were analyzed in two stages; each having different subsets of patients and controls. Based on their analysis, two SNPs were found to diverge significantly between HCC versus LC group and therefore considered a likely factor influencing transitional events from cirrhosis to hepatocellular carcinogenesis. Interestingly, the first SNP, rs2551677, is not within close proximity of any known gene, the closest gene DDX18 being 175 kb upstream of the SNP. The second SNP, rs2880301, is positioned on intron 1 of TPTE2 encoding a homolog of PTEN tumor suppressor gene and is the first time reported to be associated with carcinogenesis [23]. Additionally, three SNPs (rs9267673, rs2647073, and rs3997872) were found to be strongly associated with HCC only and were not presenting any additive/multiplicative effect. The first SNP, rs9267673, is in close proximity of C2 gene unlike the other two SNPs, rs2647073 and rs3997872, associated with SNPs falling into linkage disequilibrium of two different HLA group genes: The rs2647073 with HLA-DRB1, HLA-DRB6, HLA-DRB5, and HLA-DRA whereas the rs3997872 with HLA-DQA1, HLA-DQB1, HLA-DQA2, and HLA-DQB2 loci. The associations were independently confirmed using TAGMAN assays indicating the validity of the SNP study. When they analyzed probes targeting copy number polymorphisms, eight CNV loci including six germline CNVs were identified to be significantly associated with liver carcinogenesis. One of the germline CNVs showing a high level of association with HCC is located on a small region on p arm of chromosome 1 where no gene is known. Five other CNVs found to be linked to HCC involving *KNG1, C4orf29, LARP2, ALDH7A1, PHAX, C5orf48, LMNB1, SRPK2, PUS7*, and *TMPO* genes. Among these CNVs, two involving TRG@ and TRA@ had the strongest association to HCC. Moreover, a functional pathway and network analyses carried out using 1000 most significant SNPs associated with HCC. Among the critical pathways "antigen processing and presentation" is found the most significantly overrepresented pathway with *p*-value of 1x10-11 indicating the strongest association to HCC. Overall, these observations indicate involvement of immune system in constitutional susceptibility to HCC and HCC carcinogenesis which was suggested by clinical observations and animals models previously.

In a recent study, Jia *et al.* searched for critical somatic CNVs in 58 HCC tumor samples with adjacent non-tumor samples using Affymetrix 6.0 assay and identified 1241 regions [24].

previously unknown critical genomic regions associated with HCC. Among 653 amplicons and 57 homozygous deletions (HDs) detected by the arrays using different cell lines, 126 amplicons and 6 HDs were selected and tested to identify novel HCC-related genes. Further analysis of such aberrant regions yielded two genes, FNDC3BB and SLC29A2, consistently up-regulated in multiple HCC data sets. Knock-down studies using short hairpin RNAs targeting both genes showed decreased cell proliferation, tumor formation, and anchorage-independent growth in xenograft models in nude mice confirmed a possible pivotal role of these genes in growth and tumor formation in subsets of HCC samples. Up-regulation of either gene is proposed to be activated through STAT3 signaling pathway which is a well-known phenomenon in HCC progression usually

In another study, Clifford *et al.* used Affymetrix SNP 6.0 assay comprising probes for detection of CNVs and SNPs, each has more than 900,000 unique oligos, totaling nearly 1.9 million probe sets [23]. In their study, a large number of samples exceeding 1100 cases including histopathologically confirmed HCC and liver cirrhosis (LC) samples as well as normal controls with Korean and Chinese ethnicity were analyzed in two stages; each having different subsets of patients and controls. Based on their analysis, two SNPs were found to diverge significantly between HCC versus LC group and therefore considered a likely factor influencing transitional events from cirrhosis to hepatocellular carcinogenesis. Interestingly, the first SNP, rs2551677, is not within close proximity of any known gene, the closest gene DDX18 being 175 kb upstream of the SNP. The second SNP, rs2880301, is positioned on intron 1 of TPTE2 encoding a homolog of PTEN tumor suppressor gene and is the first time reported to be associated with carcinogenesis [23]. Additionally, three SNPs (rs9267673, rs2647073, and rs3997872) were found to be strongly associated with HCC only and were not presenting any additive/multiplicative effect. The first SNP, rs9267673, is in close proximity of C2 gene unlike the other two SNPs, rs2647073 and rs3997872, associated with SNPs falling into linkage disequilibrium of two different HLA group genes: The rs2647073 with HLA-DRB1, HLA-DRB6, HLA-DRB5, and HLA-DRA whereas the rs3997872 with HLA-DQA1, HLA-DQB1, HLA-DQA2, and HLA-DQB2 loci. The associations were independently confirmed using TAGMAN assays indicating the validity of the SNP study. When they analyzed probes targeting copy number polymorphisms, eight CNV loci including six germline CNVs were identified to be significantly associated with liver carcinogenesis. One of the germline CNVs showing a high level of association with HCC is located on a small region on p arm of chromosome 1 where no gene is known. Five other CNVs found to be linked to HCC involving *KNG1, C4orf29, LARP2, ALDH7A1, PHAX, C5orf48, LMNB1, SRPK2, PUS7*, and *TMPO* genes. Among these CNVs, two involving TRG@ and TRA@ had the strongest association to HCC. Moreover, a functional pathway and network analyses carried out using 1000 most significant SNPs associated with HCC. Among the critical pathways "antigen processing and presentation" is found the most significantly overrepresented pathway with *p*-value of 1x10-11 indicating the strongest association to HCC. Overall, these observations indicate involvement of immune system in constitutional susceptibility to HCC and HCC carcinogenesis which was suggested by

In a recent study, Jia *et al.* searched for critical somatic CNVs in 58 HCC tumor samples with adjacent non-tumor samples using Affymetrix 6.0 assay and identified 1241 regions [24].

triggered by cytokines such as interleukin-6 [19-22].

clinical observations and animals models previously.

These regions were then interrogated in search of dysregulated genes and 362 differentially expressed genes were identified. Among these, 20 genes were further evaluated functionally and *TRIM35, HEY1,* and *SNRPE* were confirmed to be involved in HCC by various functional experiments. Involvement of these genes, *TRIM35* as tumor suppressor, and *HEY1* and *SNRPE* as potential oncogenes, in HCC is novel.
