Genomic and transcriptomic somatic alterations of hepatocellular carcinoma in non-cirrhotic livers

Abstract

•Non-cirrhotic HCC genomically resembles cirrhotic HCC•Comprehensive genome- and transcriptome-wide profiling allows detection of novel structural variants, fusions, and undiagnosed viral infections•NR1H4 fusions may represent a novel mechanism for tumorigenesis in HCC•Non-cirrhotic HCC is characterized by genotoxic mutational signatures and dysregulated liver metabolism•Clinical history and comprehensive omic profiling incompletely explain underlying etiologies for non-cirrhotic HCC highlighting the need for further research Worldwide, there are approximately 750,000 new cases of hepatocellular carcinoma (HCC) each year [[1]Ferlay J. Soerjomataram I. Dikshit R. Eser S. Mathers C. Rebelo M. et al.Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012.Int J Cancer. 2015; 136: E359-E386Crossref PubMed Scopus (19883) Google Scholar]. Although HCC has the 5th highest incidence rate in men and 9th highest incidence rate in women, it has the second highest mortality rate of all cancer types [[1]Ferlay J. Soerjomataram I. Dikshit R. Eser S. Mathers C. Rebelo M. et al.Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012.Int J Cancer. 2015; 136: E359-E386Crossref PubMed Scopus (19883) Google Scholar]. HCC is traditionally associated with inflammation-inducing risk factors, which promote liver cirrhosis including: chronic hepatitis infections, such as hepatitis b virus (HBV) and hepatitis c virus (HCV), alcohol abuse, and non-alcoholic fatty liver disease [[2]Fattovich G. Stroffolini T. Zagni I. Donato F. Hepatocellular carcinoma in cirrhosis: incidence and risk factors.Gastroenterology. 2004; 127: S35-S50Abstract Full Text Full Text PDF PubMed Scopus (1835) Google Scholar]. However, approximately 20% of patients present with non-cirrhotic HCC in the absence of these risk factors [[3]Alkofer B. Lepennec V Chiche L. Hepatocellular cancer in the non-cirrhotic liver.J Visc Surg. 2011; 148: 3-11Crossref PubMed Google Scholar]. If diagnosed early, patients with non-cirrhotic HCC maintain adequate liver function, allowing for effective tumor resection with exceptional prognosis when compared to patients with cirrhotic HCC [[4]Maeda T. Shimada M. Harimoto N. Tsujita E. Aishima S.-I. Tanaka S. et al.Prognosis of early hepatocellular carcinoma after hepatic resection.Hepatogastroenterology. 2008; 55: 1428-1432PubMed Google Scholar]. However, late-stage diagnosis of non-cirrhotic HCC typically presents with larger and more aggressive tumors that are prone to metastasis [[5]Llovet J.M. Brú C. Bruix J. Prognosis of hepatocellular carcinoma: the BCLC staging classification.Semin Liver Dis. 1999; 19: 329-338Crossref PubMed Google Scholar]. Even with extensive tumor resection, approximately 50% of patients relapse within three years post-treatment [[6]Shah S.A. Cleary S.P. Wei A.C. Yang I. Taylor B.R. Hemming A.W. et al.Recurrence after liver resection for hepatocellular carcinoma: risk factors, treatment, and outcomes.Surgery. 2007; 141: 330-339Abstract Full Text Full Text PDF PubMed Scopus (307) Google Scholar]. Using high-throughput sequencing, researchers have previously characterized the genomic landscape of cirrhotic HCC [7Schulze K. Imbeaud S. Letouzé E. Alexandrov L.B. Calderaro J. Rebouissou S. et al.Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets.Nat Genet. 2015; 47: 505-511Crossref PubMed Scopus (872) Google Scholar, 8Fujimoto A. Furuta M. Totoki Y. Tsunoda T. Kato M. Shiraishi Y. et al.Whole-genome mutational landscape and characterization of noncoding and structural mutations in liver cancer.Nat Genet. 2016; 48: 500-509Crossref PubMed Scopus (412) Google Scholar, 9Cancer Genome Atlas Research Network. Electronic address: [email protected], Cancer Genome Atlas Research Network. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma.Cell. 2017; 169 (e23): 1327-1341Abstract Full Text Full Text PDF PubMed Scopus (0) Google Scholar, 10Laurent-Puig P. Zucman-Rossi J. Genetics of hepatocellular tumors.Oncogene. 2006; 25: 3778-3786Crossref PubMed Scopus (298) Google Scholar, 11Guichard C. Amaddeo G. Imbeaud S. Ladeiro Y. Pelletier L. Maad I.B. et al.Integrated analysis of somatic mutations and focal copy-number changes identifies key genes and pathways in hepatocellular carcinoma.Nat Genet. 2012; 44: 694-698Crossref PubMed Scopus (961) Google Scholar, 12Kan Z. Zheng H. Liu X. Li S. Barber T.D. Gong Z. et al.Whole-genome sequencing identifies recurrent mutations in hepatocellular carcinoma.Genome Res. 2013; 23: 1422-1433Crossref PubMed Scopus (349) Google Scholar, 13Fujimoto A. Totoki Y. Abe T. Boroevich K.A. Hosoda F. Nguyen H.H. et al.Whole-genome sequencing of liver cancers identifies etiological influences on mutation patterns and recurrent mutations in chromatin regulators.Nat Genet. 2012; 44: 760-764Crossref PubMed Scopus (638) Google Scholar]. These studies included whole genome, whole exome, and/or transcriptome sequencing with a focus on analyzing HCC induced by HBV, HCV, and/or cirrhosis. Prior studies, which have evaluated the genomics of cirrhotic and non-cirrhotic HCC, report that among the most significant and recurrent alterations are TERT mutations which typically occur at the promoter region [[8]Fujimoto A. Furuta M. Totoki Y. Tsunoda T. Kato M. Shiraishi Y. et al.Whole-genome mutational landscape and characterization of noncoding and structural mutations in liver cancer.Nat Genet. 2016; 48: 500-509Crossref PubMed Scopus (412) Google Scholar,[9]Cancer Genome Atlas Research Network. Electronic address: [email protected], Cancer Genome Atlas Research Network. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma.Cell. 2017; 169 (e23): 1327-1341Abstract Full Text Full Text PDF PubMed Scopus (0) Google Scholar,[14]Jhunjhunwala S. Jiang Z. Stawiski E.W. Gnad F. Liu J. Mayba O. et al.Diverse modes of genomic alteration in hepatocellular carcinoma.Genome Biol. 2014; 15: 436PubMed Google Scholar]. Mutations within this region have been observed in a variety of cancer types beyond cirrhotic HCC, suggesting a common role of activating TERT promoter variants in oncogenesis and metastasis [15Mosrati M.A. Malmström A. Lysiak M. Krysztofiak A. Hallbeck M. Milos P. et al.TERT promoter mutations and polymorphisms as prognostic factors in primary glioblastoma.Oncotarget. 2015; 6: 16663-16673Crossref PubMed Google Scholar, 16Huang F.W. Hodis E. Xu M.J. Kryukov G.V. Chin L. Garraway L.A. Highly recurrent TERT promoter mutations in human melanoma.Science. 2013; 339: 957-959Crossref PubMed Scopus (1225) Google Scholar, 17Hosen I. Rachakonda P.S. Heidenreich B. de Verdier P.J. Ryk C. Steineck G. et al.Mutations in TERT promoter and FGFR3 and telomere length in bladder cancer.Int J Cancer. 2015; 137: 1621-1629Crossref PubMed Google Scholar]. TERT expression in terminally differentiated cells promotes telomere maintenance and elongation [[18]Jafri M.A. Ansari S.A. Alqahtani M.H. Shay J.W. Roles of telomeres and telomerase in cancer, and advances in telomerase-targeted therapies.Genome Med. 2016; 8: 69Crossref PubMed Scopus (295) Google Scholar]. Telomere maintenance is required for late stage cancer propagation with TERT misregulation being harnessed by human cancers to evade mitotic catastrophe and apoptosis [[19]Blasco M.A. Telomeres and human disease: ageing, cancer and beyond.Nat Rev Genet. 2005; 6: 611-622Crossref PubMed Scopus (1166) Google Scholar]. Previous studies have recognized that increases in TERT expression could serve as a proxy for telomere maintenance; however, late-stage tumors exhibit shortened telomeres in comparison to their normal counterparts, due to high turnover rates [[20]Saini N. Srinivasan R. Chawla Y. Sharma S. Chakraborti A. Rajwanshi A. Telomerase activity, telomere length and human telomerase reverse transcriptase expression in hepatocellular carcinoma is independent of hepatitis virus status.Liver Int. 2009; 29: 1162-1170Crossref PubMed Google Scholar,[21]Fredriksson N.J. Ny L. Nilsson J.A. Larsson E. Systematic analysis of noncoding somatic mutations and gene expression alterations across 14 tumor types.Nat Genet. 2014; 46: 1258-1263Crossref PubMed Scopus (188) Google Scholar]. While counterintuitive, the presence of shortened telomeres in tumors with TERT overexpression is thought to arise in one of two ways. One manner is when somatic cells with critically short telomeres undergo senescence and selective pressure leading to the acquisition of the TERT promoter mutations and regeneration of telomerase to overcome telomeric crisis [[22]Heidenreich B. Kumar R. TERT promoter mutations in telomere biology.Mutat Res - Rev Mut Res. 2017; 771: 15-31Crossref PubMed Scopus (0) Google Scholar]. Another pathway to shortened telomeres is that a TERT promoter mutation is acquired by the pre-cancerous cell. At first, TERT and telomerase levels are marginal and do not prohibit telomere shortening. Critically short telomeres start accumulating and cells with TERT promoter mutations can then gradually upregulate TERT to stabilize critically short telomeres [[23]Lorbeer F.K. Hockemeyer D. TERT promoter mutations and telomeres during tumorigenesis.Curr Opin Genet Dev. 2020; 60: 56-62Crossref PubMed Scopus (19) Google Scholar]. Among studies specific to cirrhotic HCC, the putative mechanisms of TERT activation can be divided into three categories: 1) HBV integration events in the TERT promoter [[8]Fujimoto A. Furuta M. Totoki Y. Tsunoda T. Kato M. Shiraishi Y. et al.Whole-genome mutational landscape and characterization of noncoding and structural mutations in liver cancer.Nat Genet. 2016; 48: 500-509Crossref PubMed Scopus (412) Google Scholar,[24]Jiang Z. Jhunjhunwala S. Liu J. Haverty P.M. Kennemer M.I. Guan Y. et al.The effects of hepatitis B virus integration into the genomes of hepatocellular carcinoma patients.Genome Res. 2012; 22: 593-601Crossref PubMed Scopus (199) Google Scholar], 2) point mutations (C228T and C250T) in the promoter region mutually exclusive of HBV integration [[9]Cancer Genome Atlas Research Network. Electronic address: [email protected], Cancer Genome Atlas Research Network. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma.Cell. 2017; 169 (e23): 1327-1341Abstract Full Text Full Text PDF PubMed Scopus (0) Google Scholar,[25]Kawai-Kitahata F. Asahina Y. Tanaka S. Kakinuma S. Murakawa M. Nitta S. et al.Comprehensive analyses of mutations and hepatitis B virus integration in hepatocellular carcinoma with clinicopathological features.J Gastroenterol. 2016; 51: 473-486Crossref PubMed Google Scholar], and 3) structural variations of the TERT promoter region [[8]Fujimoto A. Furuta M. Totoki Y. Tsunoda T. Kato M. Shiraishi Y. et al.Whole-genome mutational landscape and characterization of noncoding and structural mutations in liver cancer.Nat Genet. 2016; 48: 500-509Crossref PubMed Scopus (412) Google Scholar,[14]Jhunjhunwala S. Jiang Z. Stawiski E.W. Gnad F. Liu J. Mayba O. et al.Diverse modes of genomic alteration in hepatocellular carcinoma.Genome Biol. 2014; 15: 436PubMed Google Scholar]. This study characterizes biomarkers and elucidates recurrent anomalies in non-cirrhotic HCC. We identified somatic variants in 117 tumor samples whereby 52 samples were cirrhotic, 63 samples were non-cirrhotic, and 2 samples had an unspecified cirrhotic status. Using this cohort, we analyzed single nucleotide variants (SNVs), insertions and deletions (INDELs), structural variation (SV), copy number variation (CNV), loss of heterozygosity (LOH), differential expression, and viral integration events. This comprehensive approach uncovered the genomic features implicated in non-cirrhotic HCC to improve its diagnosis, prognosis, and treatment. The discovery cohort consisted of 30 primary tumor and adjacent matched non-tumor liver samples obtained through surgical resection from adult patients diagnosed with HCC between 2000 to 2011 at the Washington University School of Medicine. Within this cohort, 13 were male and 17 were female. Additionally, 2 were African American and 28 were Caucasian. None of these samples exhibited evidence of hepatocellular adenoma (HCA) and the non-cirrhotic samples did not show signs of advanced fibrosis. 1 sample was HBV positive and 4 samples were HCV positive according to clinical data. All other samples within the discovery cohort had an unknown clinical etiology. The extension-alpha and extension-beta cohorts had 16 HCC tumors with matched non-tumor liver and 71 tumor-only HCC samples, respectively. Discovery and extension-alpha cohort samples were flash-frozen prior to banking and extension-beta samples were derived from formalin fixed paraffin embedded (FFPE) blocks. Across both extension cohorts, 27 were female and 58 were male. Furthermore, 2 were Asian, 13 were African American, and 70 were Caucasian. Within the extension-alpha cohort, two samples were HCV positive, one had chronic cholestasis, and the others had no known clinical etiology. Clinical data for the extension-beta cohort was as follows: 5 had known alcohol use, 8 were HBV positive, 29 were HCV positive, 2 were diagnosed with primary sclerosing cholangitis (PSC), and 6 samples were diagnosed with non-alcoholic steatohepatitis (NASH). From the extension-alpha cohort, 2 patients did not provide information on race and gender (Table S1). All patient samples were acquired after informed consent to an approved study by the Washington University School of Medicine Institutional Review Board (IRB 201106388). DNA and RNA from samples in the discovery cohort were extracted using the QIAamp DNA Mini kit and Qiagen RNeasy Mini kit, respectively. Whole genome sequencing libraries were constructed using Kapa HYPER kits for use on the Illumina HiSeq 2000 platform. The Ovation RNA-seq System V2 (NuGen Inc) kit was used to generate RNAseq libraries. Resulting barcoded libraries were pooled prior to Illumina sequencing. To validate variants identified from WGS, a hybrid capture panel (CAP1) was designed and executed on the Illumina platform to capture fragments from the WGS libraries. The QIAamp DNA Mini kit was used to extract DNA from extension-alpha samples, which was subsequently sequenced using the CAP1 strategy. Finally, CAP1 sequencing was used to identify variants from the DNA extracted from extension-beta samples with the QIAamp DNA FFPE Tissue kit. A second hybrid capture panel (CAP2) utilized Nimblegen and spiked-in IDT probes that hybridized to the TERT promoter locus and HBV genome (designed against a consensus sequence for 10 common HBV strains, see supplementary methods). CAP2 sequencing was employed on all 117 samples. TERT promoter variants were also detected in the discovery and extension-alpha cohorts with Sanger sequencing. cDNA capture was performed on pooled samples from the extension cohorts. WGS and CAP1 data were aligned to GRCh37 via the Genome Modeling System (GMS) using BWA [[26]Griffith M. Griffith O.L. Smith S.M. Ramu A. Callaway M.B. Brummett A.M. et al.Genome Modeling System: A Knowledge Management Platform for Genomics.PLoS Comput Biol. 2015; 11e1004274Crossref Scopus (53) Google Scholar,[27]Li H. Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform.Bioinformatics. 2009; 25: 1754-1760Crossref PubMed Scopus (23536) Google Scholar]. Reads from the CAP2 data were competitively aligned using BWA [[27]Li H. Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform.Bioinformatics. 2009; 25: 1754-1760Crossref PubMed Scopus (23536) Google Scholar] against the human reference genome (GRCh37) along with ten HBV genotypes for which complete genomes were available. RNAseq data were aligned with bowtie/tophat and expression was evaluated with cufflinks [[28]Trapnell C. Williams B.A. Pertea G. Mortazavi A. Kwan G. van Baren M.J. et al.Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation.Nat Biotechnol. 2010; 28: 511-515Crossref PubMed Scopus (9695) Google Scholar,[29]Trapnell C. Pachter L. Salzberg S.L. TopHat: discovering splice junctions with RNA-Seq.Bioinformatics. 2009; 25: 1105-1111Crossref PubMed Scopus (8604) Google Scholar]. All raw RNAseq reads from the discovery cohort were also aligned against the HBV genomes for evidence of HBV expression at the RNA level. The predominant HBV strain was determined using relative coverage for competitive alignments. The precise location of the HBV integration site was identified from discordant read pairs from realigning HBV CAP2 reads to GRCh37 and the predominant HBV strain's genome. A similar procedure was performed for HCV whereby both WGS and RNAseq reads were aligned against six HCV genotypes. The predominant HCV strain was determined using the total read support. To detect AAV1 and AAV2 integration, RNAseq reads were competitively aligned using kallisto [[30]Bray N.L. Pimentel H. Melsted P. Pachter L. Near-optimal probabilistic RNA-seq quantification.Nat Biotechnol. 2016; 34: 525-527Crossref PubMed Scopus (3035) Google Scholar] against AAV1 and AAV2 sequences. Telomeric tumor:normal read ratios were determined from WGS data using the GMS and visualized in R. A Wilcoxon-Mann-Whitney test measured the significance of differences between telomere length in tumor and normal samples. Somatic variant analysis for single nucleotide variants (SNV) and insertions/deletions (INDEL) were performed on all three cohorts while germline variant analysis for these variants was performed on the discovery and extension-alpha cohort. Several computational tools within and outside of the GMS [[31]Griffith M. Griffith O.L. Smith S.M. Ramu A. Callaway M.B. Brummett A.M. et al.Genome Modeling System: A Knowledge Management Platform for Genomics.PLoS Comput Biol. 2015; 11e1004274Crossref Scopus (53) Google Scholar] were employed to facilitate variant calling and subsequent filtering based on variables including variant allele frequency, read count, and predicted pathogenicity. WGS data from samples within the discovery cohort were analyzed for structural variants (SV), copy number variation (CNV), and loss of heterozygosity (LOH). Manta [[32]Chen X. Schulz-Trieglaff O. Shaw R. Barnes B. Schlesinger F. Källberg M. et al.Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications.Bioinformatics. 2016; 32: 1220-1222Crossref PubMed Scopus (589) Google Scholar] was used to identify SV events. Manta-reported breakpoints, along with a 10kb flank were annotated with biomaRt and ensembl (GRCh37.p13). Regions of CNV were identified with the GMS and LOH were identified using VarScan2 [[31]Griffith M. Griffith O.L. Smith S.M. Ramu A. Callaway M.B. Brummett A.M. et al.Genome Modeling System: A Knowledge Management Platform for Genomics.PLoS Comput Biol. 2015; 11e1004274Crossref Scopus (53) Google Scholar,[33]Koboldt D.C. Zhang Q. Larson D.E. Shen D. McLellan M.D. Lin L. et al.VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing.Genome Res. 2012; 22: 568-576Crossref PubMed Scopus (2656) Google Scholar]. The DNAcopy circular binary segmentation algorithm generated segments of LOH and CNV, which served as input for GISTIC [[34]Mermel C.H. Schumacher S.E. Hill B. Meyerson M.L. Beroukhim R. Getz G. GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers.Genome Biol. 2011; 12: R41Crossref PubMed Scopus (1487) Google Scholar] to conduct a recurrence analysis. Fusion detection algorithms identified samples in the discovery cohort harboring gene fusions from RNAseq data. Fusion predictions involving NR1H4 were validated across all 117 samples using a NanoString nCounter® Elements™ TagSets assay. Sequences for predicted transcripts of the fusion calls that met certain read support criteria (≥10 spanning + encompassing reads and ≥1 spanning read) were sent to NanoString for probe design. The R “survival” package [[35]Therneau T.M. Grambsch P.M. Modeling Survival Data: Extending the Cox Model. Springer Science & Business Media, 2013Google Scholar] was used to associate SV-affected genes and CNV/LOH-affected genomic regions with overall survival and recurrence free survival. Only mutated genes and genomic regions occuring in ≥ 4 discovery cohort samples were included in this analysis. A survival analysis was also applied to SNV/INDELs observed in all non-cirrhotic samples from the three cohorts. All Kaplan-Meier survival plots were created in R. Fisher's exact test was used to test for clinical associations with variables: lymphovascular space invasion (LVSI), tumor differentiation status, cirrhosis, and liver disease. Samples without relevant clinical data were excluded. Significance was measured with a multiple test correction using the FDR methodology (q-value < 0.05). Read counts for genes mutated in non-cirrhotic tumors and matched normal samples of the discovery cohort were used by the DEseq2 Bioconductor package [[36]Love M.I. Huber W. Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.Genome Biol. 2014; 15https://doi.org/10.1186/s13059-014-0550-8Crossref Scopus (24864) Google Scholar] to perform differential gene expression analysis using a negative binomial distribution with samples as a blocking factor. Significance was measured with a Wald test and Benjamini & Hochberg multiple test correction (q-value < 0.5). Pathway analysis was performed using log2 differential expression data. There were 30 patients included in the discovery cohort with tumors which were surgically resectable. These surgically resectable tumors were untreated, providing the opportunity to study HCC in the absence of chemotherapeutic intervention, which is normally incorporated in the treatment of cirrhotic HCC. Three of the patients within this cohort developed HCC in the setting of cirrhosis, all of which had been previously diagnosed with HCV. The remaining 27 individuals developed non-cirrhotic HCC, two of these individuals were diagnosed with HBV and another two individuals were diagnosed with HCV. To elucidate the genomic landscape of resected, primarily non-cirrhotic HCC, we performed whole genome sequencing (WGS), hybrid capture sequencing (CAP1), and transcriptome sequencing (RNAseq) on these 30 samples (Table 1). WGS failed for one tumor sample in the discovery cohort, therefore the final data for this cohort included WGS and CAP1 data for 29 samples (26 non-cirrhotic, 3 cirrhotic), and RNAseq data for 30 samples (27 non-cirrhotic, 3 cirrhotic). The sequencing analysis revealed a single previously unknown and undiagnosed HBV case with viral integration occurring at the TERT promoter (Figure S1, Table S1). Median haploid coverage for WGS data was 35.6x (range: 28.5-39.3) and 58.4x (range: 46.8-94.4) for normal and tumor samples, respectively.Table 1Discovery (N=29)Extension-alpha (N=16)Extension-beta (N=71)Sample TypeTumor/Non-tumorTumor/Non-tumorTumorWGSYesNoNoCAP1YesYesYesCAP2YesYesYesRNAseqYesNoNo Open table in a new tab After filtering, we observed a median mutation burden of 1.31 mutations/Mb (range: 0.033-3.28), comprised of 2,633 SNVs and INDELs across all samples (range: 2-200, median: 77.5, mean=87.8) (Figure 1, Table S1). These variants were discovered across 2,245 genes with 258 of these genes mutated in more than one sample. Using WGS data from the 26 non-cirrhotic samples, we identified 6 genes that were significantly mutated above background mutation rates according to MuSiC: ALB, APOB, CTNNB1, TP53, RB1, and RPS6KA3 (Figure 1, Table S1). With regards to all methods of sequencing (WGS, RNAseq, CAP1, and CAP2), the most frequently encountered variant was a SNV in the telomerase reverse transcriptase (TERT) promoter (C228T; G1295228A), which was identified in 17/30 samples and resulted in overexpression of TERT (Figure S2, Table S2). Within the exome, TP53 was the most recurrently mutated gene and was observed in 8/29 of samples (Table S1). Beta catenin 1 (CTNNB1) was also significantly mutated within this cohort (6/29), whereby the majority of variants occurred at amino acids S37 and S45, both of which reside in a putative GSK3B phosphorylation site in exon 3 (ENST00000349496) (Figure S3) [[37]Miyoshi Y. Iwao K. Nagasawa Y. Aihara T. Sasaki Y. Imaoka S. et al.Activation of the beta-catenin gene in primary hepatocellular carcinomas by somatic alterations involving exon 3.Cancer Res. 1998; 58: 2524-2527PubMed Google Scholar]. Frameshift mutations in APOB were observed in 4/29 of samples (Table S1). Mutation signatures using the COSMIC database for the discovery cohort were investigated. Signatures 5 (unknown etiology), 4 (smoking damage association), 16 (unknown etiology), and 12 (liver damage association) were most prevalent and contributed to the overall cohort signature at 23%, 14%, 8%, and 7%, respectively (Figure S4). Differential gene expression analysis performed on the non-cirrhotic samples revealed that 11% of genes, including TERT, were upregulated (4,468/39,392) and 10% of genes, including CTNNB1 and WISP2, were downregulated (4,114/39,392) compared to adjacent non-tumor liver tissue (q-value < 0.1) (Table S1). Comparison of gene log2 fold changes derived from the differential expression analysis revealed the cell cycle pathway as upregulated in the KEGG signaling and metabolism database (q-value ≤ 0.05). Similarly, we observed 16 pathways as down-regulated (q-value ≤ 0.05), most of which are related to metabolic liver processes. Genes such as ADH5 and EHHADH were observed with reduced expression levels and participate in 38% (6/16) of these pathways. Using the Gene Ontology biological process database, we observed 107 pathways as significantly upregulated (q-value ≤ 0.05). The majority of the upregulated pathways were related to cellular division and DNA repair. In addition, 28 pathways were identified as significantly downregulated (q-value ≤ 0.05), many of which were related to liver metabolism (Table S2). When evaluating the samples within the discovery cohort for telomere length at the DNA level, we observed that the majority of tumor samples exhibited shortened telomeres compared to their paired normal sample (p-value = 0.00011) (Figure S2). One exception was seen in sample HCC16_D, which was distinguished by abnormally high expression of TERT (FPKM=36) (Figure S5). We observed recurrent large scale amplification of the q-arm of chromosome 1 in ≥ 50% of the discovery cohort. Similarly, large scale deletions of the p-arms of chromosomes 8 and 17 were found in ≥ 40% of the cohort (Figure 2). In total, analysis with GISTIC and subsequent manual review revealed 75 unique regions across 17 chromosomes as recurrently amplified and 45 unique regions across 17 chromosomes as significantly deleted (q < 0.05) (Table S1). No significant associations with tumor differentiation status were made (α=0.05). Each CNV and LOH event was tested for their association with overall survival and recurrence free survival but no significant association could be made following multiple test correction. A total of 33 genes identified as recurrently deleted by GISTIC showed concordant decreased expression in tumor samples (Table S1). These include genes previously characterized as relevant to HCC development and progression: HEYL [[38]Kuo K.-K. Jian S.-F. Li Y.-J. Wan S.-W. Weng C.-C. Fang K. et al.Epigenetic inactivation of transforming growth factor-β1 target gene HEYL, a novel tumor suppressor, is involved in the P53-induced apoptotic pathway in hepatocellular carcinoma.Hepatol Res. 2014; 45: 782-793Crossref PubMed Scopus (11) Google Scholar] (q-value = 0.032), UQCRH [[39]Park E.-R. Kim S.-B. Lee J.-S. Kim Y.-H. Lee D.-H. Cho E.-H. et al.The mitochondrial hinge protein, UQCRH, is a novel prognostic factor for hepatocellular carcinoma.Cancer Med. 2017; 6: 749-760Crossref PubMed Scopus (19) Google Scholar] (q-value = 0.032), and MUTYH [[40]Krupa R. Czarny P. Wigner P. Wozny J. Jablkowski M. Kordek R. et al.The Relationship Between Single-Nucleotide Polymorphisms, the Expression of DNA Damage Response Genes, and Hepatocellular Carcinoma in a Polish Population.DNA Cell Biol. 2017; https://doi.org/10.1089/dna.2017.3664Crossref PubMed Scopus (13) Google Scholar] (q-value = 0.048). A subset of these genes have also been implicated in tumorigenesis, metastasis, and progression of other cancer types and may prove to be relevant for HCC development and progression: RPL11 [[41]Takafuji T. Kayama K. Sugimoto N. Fujita M. GRWD1, a new player among oncogenesis-related ribosomal/nucleolar proteins.Cell Cycle. 2017; : 1-7Google Scholar] (q-value = 0.048), UBE2D3 [[42]Guan G.G. Wang W.B. Lei B.X. Wang Q.L. Wu L. Fu Z.M. et al.UBE2D3 is a positive prognostic factor and is negatively correlated with hTERT expression in esophageal cancer.Oncol Lett. 2015; 9: 1567-1574Crossref PubMed Scopus (17) Google Scholar] (q-value = 0.032), ARRB1 [[43]Miele E. Po A. Begalli F. Antonucci L. Mastronuzzi A. Marras C.E. et al.β-arrestin1-mediated acetylation of Gli1 regulates Hedgehog/Gli signaling and modulates self-renewal of SHH medulloblastoma cancer stem cells.BMC Cancer. 2017; 17: 488Crossref PubMed Scopus (0) Google Scholar] (q-value = 0.032), ENG [[44]Kokaji E. Shimomura A. Minamisaka T. Nakajima T. Miwa S. Hatta H. et al.Endoglin (CD105) and SMAD4 regulate spheroid formation and the suppression of the invasive ability of human pancreatic cancer cells.Int J Oncol. 2018; 52: 892-900PubMed Google Scholar] (q-value = 0.049), and ABLIM2 [[45]Hwang S.J. Lee H.W. Kim H.R. Song H.J. Lee D.H. Lee H. et al.Overexpression of microRNA-95-3p suppresses