New issues in the pathogenesis of hepatocellular carcinoma: Applying insights from next-generation DNA sequencing technologies to improve therapy for hepatocellular carcinoma

Authors

  • Roongruedee Chaiteerakij M.D.,

    1. Division of Gastroenterology and Hepatology, College of Medicine, Mayo Clinic, Rochester, MN
    Search for more papers by this author
  • Lewis R. Roberts M.B. Ch.B., Ph.D.

    Corresponding author
    1. Department of Medicine, Faculty of Medicine, Chulalongkorn University/King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Bangkok, Thailand
    • Division of Gastroenterology and Hepatology, College of Medicine, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905. E-mail: roberts.lewis@mayo.edu.

    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

Abstract

Watch a video presentation of this article

Watch the interview with the author

Answer questions and earn CME

Abbreviations
ADAM22

A disintegrin and metalloproteinase domain 22

AKAP3

A kinase anchor protein 3

ARID

AT rich interactive domain

AXIN1

axin 1

BAZ2B

bromodomain adjacent to zinc finger domain 2B

BPTF

bromodomain PHD finger transcription factor

BRD8

bromodomain containing 8

BRE

brain and reproductive organ-expressed

C5orf51

chromosome 5 open reading frame 51

CCNG1

cyclin G1

CDK14

cyclin-dependent kinase 14

CPEB4

cytoplasmic polyadenylation element binding protein 4

CSNK1G3

casein kinase 1 gamma 3

CTNNB1

catenin beta 1

CUL9

cullin 9

ELL

elongation factor RNA polymerase II

FGD6

FYVE

HBV

hepatitis B virus

HCC

hepatocellular carcinoma

HCV

hepatitis C virus

HIST1H4B

histone cluster 1 H4b

HOXA1

homeobox A1

IRF2

interferon regulatory factor 2

JAK2

Janus kinase 2

KDM6A

lysine (K)-specific demethylase 6A

KHDRBS2

KH domain containing RNA binding signal transduction associated 2

MAPK

mitogen-activated protein kinase

MLL

myeloid/lymphoid or mixed-lineage leukemia

NEK8

NIMA-related kinase 8

NFE2L2

nuclear factor (erythroid-derived 2)-like 2

NGS

next-generation sequencing

PIK3AP1

phosphoinositide-3-kinase adaptor protein 1

PVTT

portal vein tumor thrombosis

RNF139

ring finger protein 139

RPS6KA3

ribosomal protein S6 kinase 90-kDa polypeptide 3

TERT

telomerase reverse transcriptase

TMEM

transmembrane protein

TP53

tumor protein p53

TRRAP

transformation/transcription domain-associated protein

VCAM1

vascular cell adhesion molecule 1

WES

whole exome sequencing

WGS

whole genome sequencing

Techniques in DNA sequencing have advanced rapidly over the past few years. Next-generation sequencing (NGS) technology is an incredibly powerful tool that is transforming biomedical research, in large part because of technology improvements that have resulted in an astonishingly rapid decline in the cost of sequencing over time. NGS provides high-throughput information on genetic alterations, including single nucleotide variations, small insertions and deletions (indels), copy number variations, chromosomal rearrangements, and viral DNA integration. NGS technology is providing new insights into the highly complex process of liver carcinogenesis and is helping to identify key drivers of hepatocellular carcinoma (HCC) initiation and progression. This article briefly summarizes our current understanding of the pathogenesis of HCC and HCC tumor biology gained from recent studies using NGS technology to analyze HCC tumors. We also discuss potential strategies for harnessing this new information to improve therapy for patients with advanced HCC.

The first report of a comprehensive NGS study of HCC was published by Totoki et al.[1] in 2011 (Table 1). HCC tumor tissue from a hepatitis C virus (HCV)–infected patient was sequenced with both whole genome sequencing (WGS) and whole exome sequencing (WES) approaches. A large number of previously uncharacterized genetic alterations, including 11,731 somatic nucleotide substitutions, 670 small indels, and 22 chromosomal rearrangements, were revealed. The genes harboring nonsynonymous substitutions and small indels in protein-coding regions included two tumor suppressor genes well known to be mutated in HCC [tumor protein p53 (TP53) and axin 1 (AXIN1)] and five genes known to be mutated in other cancer types but identified for the first time in this HCC [A disintegrin and metalloproteinase domain 22 (ADAM22), Janus kinase 2 (JAK2), KH domain containing RNA binding signal transduction associated 2 (KHDRBS2), NIMA-related kinase 8 (NEK8), and transformation/transcription domain-associated protein (TRRAP)]. One nonsense substitution in the tuberous sclerosis 1 gene was identified by WES but was not detected by WGS because of the small number of tumor alleles that harbored this substitution (13%). This suggested that only a small cancer cell population within the tumor contained this mutation and supported the concept of heterogeneity in genetic changes found within an individual tumor.

Table 1. Summary of Recent NGS Publications on HCC
StudyPatient Characteristics and Study DesignMajor FindingsStrengths
  1. aThe WGS approach was applied in the study.
  2. bThe WES approach was applied in the study.
Totoki et al.[1] (2011)a, bSingle HCV-infected HCCA substantial number of heterogeneous genetic alterations occurred within a single tumor (11,731This is the first report of an HCC genome sequenced by NGS.
  substitutions, 670 indels, and 22 rearrangements). 
 Matched normal lymphocyte DNAFive novel mutations were found in ADAM22, JAK2, KHDRBS2, NEK8, and TRRAP. 
Li et al.[2] (2011)†Discovery set: 10 HCV-infected HCCsARID2 is commonly mutated in HCCs of different etiologies.There was a large, heterogeneous validation cohort.
 Validation set: 139 HCCs with different etiologies  
 Matched noncancerous liver DNA  
Tao et al.[3] (2011)a, bSingle HBV-infected HCCMutated CCNG1 and P62 and a C5orf51- CPE4 fusion gene are drivers of HCC recurrence and metastasis.This is the first extensive study of genetic alterations at different HCC stages by NGS.
 Matched noncancerous liver DNA  
Fujimoto et al.[4] (2012)aDiscovery set: 27 HCC tumorsMutations in chromatin remodeling family genes, including ARID1A, ARID1B, and ARID2, were present in 52% of the tumors and may be associated with aggressive HCC.This is the first large-scale study of HCC genomes by WGS.
 Validation set: 120 HCC tumorsHBV DNA integration was found in 8 of 11 HBV-induced HCCs (half occurred within or upstream of TERT).Correlations between genetic changes and clinicopathological features were examined.
 Matched normal lymphocyte DNA  
Guichard et al.[5] (2012)†Discovery set: 24 HCCsMutations in chromatin remodeling genes were found in 24% of HCCs.A novel interaction of oxidative stress and Wnt/β-catenin pathways was identified.
 Validation set: 101 HCCsCell growth regulator IRF2 functions as a tumor suppressor by regulating p53 expression.Novel mutated genes in known HCC pathways were identified.
 Matched noncancerous liver DNAMutations were found in oxidative stress pathway genes, including NFE2L2. 
Huang et al.[6] (2012)†Discovery set: 10 HBV-infected HCCsMutated ARID1A was found in 14 of 110 HCCs (13%) and may drive tumor metastasis.This study addresses genetic changes in advanced HCCs.
 Validation set: 100 HBV-infected HCCsNovel mutations were found in genes regulating cell proliferation and survival, including the cell adhesion molecule encoding gene VCAM1. 
 Matched PVTTs  

Li et al.[2] applied WES to a larger number of HCV-related HCCs (n = 10) in a discovery set and then validated the findings in two independent cohorts of HCCs from patients with different HCC etiologies (n = 23 and n = 106). This study identified the chromatin remodeling gene AT rich interactive domain 2 (ARID2) as a commonly mutated gene in HCC and implied that ARID2 functions as a tumor suppressor gene in HCC on the basis of the predicted function of the mutated protein domains. The prevalences of ARID2 mutations in HCV-related, hepatitis B virus (HBV)–related, and non–viral-related HCCs were 14% (6/43), 2% (1/50), and 5% (2/44), respectively. The different prevalences of ARID2 mutations among HCCs of various etiologies supports the idea that HCCs of different etiologies have different profiles of genetic alterations. Six of nine tumors with mutated ARID2 also contained mutations in the catenin beta 1 (CTNNB1) gene encoding β-catenin, which is the major oncogene of the Wnt signaling pathway. Interestingly, none of the nine tumors had TP53 mutations, and this suggests that the mutation of either ARID2 or the TP53 gene is sufficiently permissive for carcinogenesis.

In a different study aimed at examining the clonal heterogeneity within a tumor and the genetic mechanisms for the acquisition of invasive and metastatic phenotypes in HCC, Tao et al.[3] extensively studied the genomes of nine samples from three resected HCC nodules (seven samples from the primary tumor and one from each of the two recurrent tumors) and seven samples from the noncancerous tissue adjacent to the primary tumor from an HBV-infected patient by WGS, WES, and cell-population genetic analysis. After validation, 214 tumor-specific mutations were identified: 193 were noncoding, and 21 were coding. Nine point mutations and one deletion acquired during critical steps in HCC progression were identified and used to infer the evolution of the tumors. In the primary tumor, major and minor cancer cell populations were differentiated by four mutations, including within the cyclin G1 (CCNG1) encoding gene, and this suggests that these mutations facilitate tumor expansion. Mutations in CCNG1 and P62 and the formation of a chromosome 5 open reading frame 51 (C5orf51)–cytoplasmic polyadenylation element binding protein 4 (CPEB4) fusion gene due to the deletion of chromosome 5q were found to be driver alterations for HCC recurrence and intrahepatic metastasis. These alterations affected the cell cycle control, autophagy, and apoptosis pathways. Eighteen HBV DNA integrations—a key event for HCC development—were also observed. A cancer evolution analysis led to the conclusion that there was a prolonged period of accumulation of background mutations with predominant functions in inflammation, immunity, and cell anchoring. Subsequent to this, the rapid spread of a small number of foreground mutations in genes regulating the cell cycle and apoptosis mechanisms then led to the assertion of the cancer phenotype.

A larger scale study applying WGS to HCC genomes was conducted by Fujimoto et al.[4] Genomes of 27 HCC tumors from 25 patients (HBV, 11; HCV, 14; and nonviral, 2) were sequenced. There were two sets of multicentric, synchronously occurring tumors that had similar whole genome substitution profiles but different mutation profiles; this suggested that the two tumors in each case developed independently within a similar context of genomic instability. Overall, a significant frequency of mutations was found in 15 genes. In addition to the expected mutations in the well-known HCC-associated genes TP53 and CTNNB1, somatic point mutations or indels in chromatin remodeling genes, including ARID1A, ARID1B, ARID2, myeloid/lymphoid or mixed-lineage leukemia (MLL), MLL3, bromodomain adjacent to zinc finger domain 2B (BAZ2B), bromodomain containing 8 (BRD8), bromodomain PHD finger transcription factor (BPTF), brain and reproductive organ-expressed (BRE), and histone cluster 1 H4b (HIST1H4B), were detected in 14 of 27 HCCs (52%). Decreased expression of these genes increased the proliferation of HCC cell lines, and this implied a tumor suppressor function of chromatin remodeling genes in HCC. HBV DNA integrations were found in 8 of the 11 HBV-induced HCCs, and 4 of these occurred within or upstream of the telomerase reverse transcriptase gene (TERT), which probably conferred a clonal advantage to these tumors. Interestingly, one tumor with a somatic nonsense mutation in the mutL homolog 1 DNA mismatch-repair gene had more than 33,000 somatic mutations with a predominance of C>T/G>A transitions at CpGs. A second tumor with an unidentified mismatch-repair defect showed similar characteristics of mismatch-repair deficiency. Overall, the etiological and genetic context of the individual appeared to significantly influence the pattern of somatic alterations that were seen. Significant correlations were found between somatic substitution patterns and habitual alcohol use, and a marginal association was found with the virus type.

The role of the chromatin remodeling pathway in HCC carcinogenesis was strongly supported by the findings from the study of another cohort of 125 HCC patients in which WES was performed on a subset of 24 tumors. In this cohort also, mutations in Wnt/β-catenin and p53 pathway genes were most frequent.[5] Guichard et al.[5] also found at least one mutated chromatin remodeling gene in 24% of HCCs. ARID1A and ARID2 mutations were present in 21 of 125 HCCs (17%) and in 7 of 125 HCCs (6%), respectively. ARID1A mutations were more frequent in alcohol-induced HCCs and were associated with CTNNB1 mutations. In addition, mutations in interferon regulatory factor 2 (IRF2), a cell growth–regulating gene, were identified in 6 of 125 HCCs (5%), which were exclusively HBV-induced. The biological function of IRF2 was studied with in vitro and in vivo experiments. An inactivating mutation of IRF2 decreased TP53 expression, and this suggests that IRF2 functions as a tumor suppressor by regulating the p53 pathway. Novel recurrent mutations were found in the ribosomal protein S6 kinase 90-kDa polypeptide 3 (RPS6KA3) gene, which encodes ribosomal S6 protein kinase 2, a feedback inhibitor of the extracellular signal-regulated kinase 1/2 pathway. RPS6KA3 mutations preferentially occurred in HCCs developing in the absence of cirrhosis and were associated with AXIN1 gene mutations. A potential role of the oxidative stress pathway in hepatocarcinogenesis was also revealed. The mutation of nuclear factor (erythroid-derived 2)-like 2 (NFE2L2), which encodes nuclear erythroid 2 p45-related factor 2, a transcription factor regulating the expression of antioxidant enzyme coding genes, was found in 8 of 125 HCCs (6%). A mutation in NFE2L2 was significantly correlated with CTNNB1 mutation, and this suggests that oxidative stress may cooperate with Wnt signaling pathway activation in HCC pathogenesis. A particularly striking finding was the fact that in contrast to most other solid tumor types, genetic variations in HCCs were characterized by the overrepresentation of G:C>T:A nucleotide transversions; this was particularly true for HCCs developing in the absence of cirrhosis. This observation may be related to a unique susceptibility of the liver to as yet unidentified genotoxic agents.

It is important to note that in most of the aforementioned studies, the HCC genomes were sequenced from tumor tissues obtained from surgical specimens. This approach is prone to a bias toward studying the genomes of early-stage HCCs. This issue was addressed by the most recent publication by Huang et al.[6] The authors reported WES findings for HCC genomes and matched portal vein tumor thromboses (PVTTs), which were manifestations of advanced HCC with a poor prognosis, in 10 HBV-infected patients. The findings were then validated in an independent cohort of 100 HBV-induced HCCs. HCCs and PVTTs shared 94% of the identified somatic mutations. Genes mutated only in PVTTs were identified; they included lysine (K)-specific demethylase 6A (KDM6A), cullin 9 (CUL9), FYVE, RhoGEF and PH domain containing protein 6 (FGD6), A kinase anchor protein 3 (AKAP3), and ring finger protein 139 (RNF139). Mutated ARID1A was found in 14 of the 110 HBV-related HCC cases (13%). ARID1A appeared to be a key driver for HCC metastasis because mutated sequences were detected only in the HCC cell line that had the highest capacity for metastasis. Screening for potential driver gene mutations identified nonsilent mutations in homeobox A1 (HOXA1), vascular cell adhesion molecule 1 (VCAM1), transmembrane protein 35 (TMEM35), and phosphoinositide-3-kinase adaptor protein 1 (PIK3AP1). Finally, a loss of function RNA interference screen of 91 genes in eight HCC cell lines was undertaken to identify genes involved in cell proliferation or survival. Knockdown of VCAM1 and TMEM2 promoted HCC cell proliferation in four of the eight cell lines, whereas knockdown of cyclin-dependent kinase 14 (CDK14), HOXA1, TMEM35, elongation factor RNA polymerase II (ELL), and casein kinase 1 gamma 3 (CSNK1G3) suppressed cell proliferation in at least three of the eight cell lines.

In conclusion, WGS and WES are complementary approaches to the identification of genetic alterations in cancer. Growing evidence from NGS has revealed that the chromatin remodeling pathway is a major contributor to HCC carcinogenesis. Several new genes involved in previously known HCC pathways have also been discovered with these techniques (Fig. 1). This strategy, therefore, is identifying multiple potential new targets for HCC treatment.

Figure 1.

New genes (shown in boxes) identified by NGS technologies contribute to HCC pathogenesis.

Summary, Conclusions, and Anticipated Future Directions

Besides confirming the prevalence of a number of known genetic alterations in HCCs, the new NGS techniques have also identified many novel genetic alterations and a number of pathways that are potentially important in HCC pathogenesis. More importantly, they also provide the means for carefully examining DNA-based mechanisms involved in liver carcinogenesis, invasion, and metastasis. For HBV-induced HCCs, an immediate effect has been to confirm that although there are many integration sites in nonmalignant liver tissue, substantial clonal selection occurs and results in repeated integrations in a much smaller subset of genes in HCCs and most notably in the TERT gene locus; this observation was first reported in 2003.[7, 8] The additional analysis of larger cohorts of HCCs, with attention paid to addressing specific mechanistic hypotheses, should result in substantial new knowledge and translate into improvements in the prevention, diagnosis, and therapy of HCC. In particular, one critical opportunity presented by the improvements in NGS sequencing is the potential to identify the genetic characteristics of patients and their tumors that are associated with clinical response or resistance to particular therapies, such as sorafenib or newer agents currently in clinical trials. The opportunity to pair this strategy with that of implanting freshly resected or biopsied HCC tumors into immunodeficient mice to develop models that show fidelity to human tumors and that can be expanded into groups of mice and used in studies to determine which drug therapies are most effective (so-called xenopatient studies) holds real promise for substantial improvements in therapy for advanced HCC (Fig. 2). Important paradigm shifts in our thinking that will facilitate this potential revolution in therapy include the establishment of clinical patient care grade programs for obtaining tumor biopsy samples from intermediate- and advanced-stage HCC patients that can be implanted fresh into immunodeficient mice (for xenopatient studies) and also snap-frozen (for NGS and other molecular analyses). As a first step in this effort, all new clinical trials should be designed to obtain such samples when this is feasible. These approaches also require a new culture of partnership between basic, translational, and clinical researchers, the pharmaceutical industry, and patient groups and advocates to facilitate and most efficiently accomplish these goals. Just like the advent of the approval of sorafenib in 2007, the recent advances in NGS have opened a new door of opportunity for improving the management of HCC.

Figure 2.

Proposed strategy for improving outcomes of HCC therapy by exploiting NGS technologies in combination with primary tumor xenograft studies with the goal of determining the best candidate drug for each individual patient.

Ancillary