Next-generation sequencing and personalized genomic medicine in hepatobiliary malignancies

Abstract

Liver cancer is a heterogeneous group of tumors characterized by significant molecular and genomic heterogeneity. The advent of powerful genomic technologies has allowed detection of recurrent somatic alterations in liver cancer, including mutations, copy number alterations as well as changes in transcriptomes and epigenomes, with the potential to translate these data into clinically relevant predictive and prognostic factors. In this review, we discuss recent advances in the application of high-throughput genomic technologies in liver cancer and the integration of such cancer genome profiling data, highlighting specific relevant subgroups and explain how this knowledge can be used in translational clinical research, ‘basket trials’, molecular tumor boards, targeted therapy and for personalized genomic medicine applications.

Practice points.

• Hepatobiliary malignancies are common cancers due to multiple etiologies with a poor prognosis.

• There is significant genomic variation between and within subtypes of these malignancies.

• Actionable genomic changes include: detection of a mutation associated with a targeted therapy, alteration in prognosis, assist in diagnosis or alter therapeutic decisions.

• There are rare mutations within hepatocellular carcinoma subtypes that are targetable.

• Additional targeted therapies are becoming available through clinical trials and US FDA clearance, so therapeutic interventions are continuously changing.

Liver cancer is the third leading cause of cancer-related mortality worldwide, including hepatocellular carcinoma (HCC) and cholangiocarcinoma (CC) [1,2]. CC is the second most common primary hepatic malignancy after HCC [1]. CC is divided into extrahepatic and intrahepatic categories; furthermore, intrahepatic cholangiocarcinomas are further subclassified into bile duct and cholangiolar types each with corresponding etiologic, prognostic and molecular correlates. Since the advent of the 21st century, advances in genomic technologies have led to a comprehensive understanding of pathophysiology and natural history of several tumor models. The molecular and ‘omics’ analysis of cancer utilizing whole genome and transcriptome approaches has great promise to advance the current classification system for liver cancer, and it has allowed for identification of substantial distinct genetic and histologic features that correlate with known risk factors, epidemiology, presentation and outcomes [3–5]. Surgical resection represents the only curative modality for patients suffering from liver cancer; however the majority of cases are diagnosed at advanced stages, with median overall survival of less than 2 years from the time of diagnosis despite palliative chemotherapy and other measures [5–10]. In the particular case of HCC, it is typically resistant to systemic chemotherapy, and only sorafenib-targeted therapy has shown to provide a survival advantage over best supportive care [11].

In an effort to improve the outcomes of these patients, several molecular studies have been performed in HCC and CC, focusing on the correlation of cancer-related genomic changes and consequent alterations in molecular signaling pathways, aiming to detect targetable oncogenic driving mutations, as well as predictive and prognostic genomic markers of response. This active research also includes novel fields of bioinformatics, small RNAs and epigenome mapping. More recently, improvements of sophisticated next-generation sequencing (NGS) and high-throughput microarray-based technologies have provided a high-yield perspective on cancer genomics (e.g., coding mutations, DNA insertions and deletions, copy number variations, chromosomal translocations or inversions, gene amplifications and gene expression) at the single nucleotide resolution level, transforming paradigms of the incipient translational genomic medicine field [12–14].

Herein we describe the most commonly associated molecular changes in hepatobiliary malignancies, and also highlight promising targets for pathway-driven personalized cancer therapy.

Genomic variation, carcinogenesis, allelic imbalance & signaling oncogenic pathways

• HCC

Given the natural history of HCC arising in a background of chronic liver inflammation and cirrhosis in the majority of cases, several somatic alterations and structural variation of genes important for cell cycle and apoptosis are associated with primary liver cancer such as c-MYC, RAS and TP53 [15]. Powerful analytical tools such as array comparative genomic hybridization (CGH) have allowed high-yield DNA copy number data, which provides a background on the molecular changes involved in HCC pathogenesis.

• Mutation interpretation relies on annotated genomic databases

Databases such as OncoDB.HCC [16] are currently available and open access validated catalogs of common aberrations from both human and murine HCC models, including loss of heterozygosity (LOH), gene expression, proteomics and CGH data [17]. This website demonstrates the complexity of HCC at the chromosomal level (Figure 1) and allows the user to search regions of interest for more detailed analysis of the involved genomic regions. The challenge remains in the appropriate interpretation of such rich amount of information. In general, variants are reported according to Human Genome Variation Society (HGVS) nomenclature and classified into three categories: pathogenic, variants of uncertain significance and benign. Categorization of variants is dependent upon literature review, and variant existence in other publicly available databases including the Single Nucleotide Polymorphism Database (dbSNP), the Catalogue Of Somatic Mutations In Cancer (COSMIC) [18] and the 1000 Genomes Project. In addition, academic medical centers host genomic medical knowledge resources online that characterize specific genetic mutations, including ‘My Cancer Genome’ [19], which describes genes, mutations and US FDA and clinical trial information. The Broad Institute maintains a research database of somatically mutated genes linked to a clinical action, updated yearly that includes genes, rationale, mutation type (inclusive of translocations) and examples of potential therapeutic agents (tumor alterations relevant for genomics-driven therapy) (TARGET) [20]. Although this database is not curated for clinical or diagnostic use, the site can be used as a springboard for discovery of mutation-directed therapies.

Figure 1. . Common genomic aberration regions of hepatocellular carcinoma (per chromosome).

CGH: Comparative genomic hybridization; HCC: Hepatocellular carcinoma; LOH: Loss of heterozygosity; QTL: Quantitative trait locus.

Data taken from [16,21].

• Genetic & genomic changes detected in HCC

HCCs are a diverse set of tumors whose pathogenesis has been described as a complex process that requires the accumulation of several epigenetic and genetic changes by sequential alterations, which have not been yet completely established [22]. Specific tumor microarray studies have been pursued to evaluate the transition from a ‘dysplastic nodule (DN)’ (a controversial entity that by some investigators is viewed as a precursor to HCC) to early HCC [22,23]. Gene expression profiling of those lesions identified 460 differentially expressed genes, with MYC oncogene activation considered as a potential signaling driver to malignant transformation. Epigenetic analysis of HCC development has also reported progressive CpG island hypermethylation in key genes, as well as distinct methylation patterns based on specific etiology leading to malignant transformation [24–26].

Other efforts have been focused on determining different global gene expression profiles that correlated with prognosis. One of these studies [27] revealed two separate subclasses of HCC via algorithms and Cox regression models that were predictive of survival. Expression of proliferating cell nuclear antigen, and CDK4, CCNB1, CCNA2 and CKS2 (cell cycle regulators) was greater in the poor survival subclass (cluster A), as well as high expression of antiapoptosis, sumoylation and ubiquitination-related genes. These findings may provide the rationale for recent efforts to use proteasome inhibitors and inhibition of other mediators of ubiquitin-dependent protein degradation, particularly in poor prognosis subtypes.

Chromosome aberrations (gains or losses) in human HCC have been demonstrated to correlate with tumor grade and specific etiologic factors (e.g., hepatitis B–HBV vs C–HCV viral infection) [28]. The most commonly reported chromosome losses in HCC per a meta-analysis [29] are in 8p (38%), 16q (35.9%), 4q (34.3%), 17p (32.1%) and 13q (26.2%). HBV infection has been correlated with 4q, 16q, 13q and 8p deletions (in the absence of HCV infection). Gain of 1q appeared to be an early event developing in DN, possibly predisposing affected cells to acquire additional chromosomal aberrations, and gain of 3q22–24 is a potential late genomic event associated with tumor recurrence and poor survival. The most frequent chromosomal amplifications involve 1q (57.1%), 8q (46.6%), 6p (22.3%) and 17q (22.2%). These chromosomal aberrations are commonly associated to gene loci implicated in the WNT signaling pathway (e.g., AXIN2, FZD3, SIAH-1, WISP1), as well as oncogene activation sites (e.g., MYC on 8q24) and as inactivating mutations of tumor suppressor genes (e.g., RB1 on 13q14). Poorly differentiated HCC have chromosomes 13q and 4q losses, and gains of 1q correlate with other high-frequency alterations. In DN, amplifications were most frequent in 1q and 8q, whereas deletions occurred in 8p, 17p, 5p, 13q, 14q and 16q [29].

• Mutations in HCC occur in common signaling pathways

Since the advent of the concepts of ‘oncogene-addiction phenomenon’ and compensatory or escape pathways [30–32], the identification of specific signaling driving pathways and key oncogenes and tumor suppressors regulating hepatocarcinogenesis has become a subject of active investigation [15,33–35]. Signaling pathways and mutations described in HCC include (Figure 2) TP53 [34,36–37], Wnt/β-Catenin (e.g., MYC, Cyclin D1 and E-Cadherin) [38,39], HGF/c-MET [40,41], EGFR (via AKT, STATs, RAS/RAF) [42–44], IGF [45,46], TGF-β [47], Nf-κB [48,49], VEGF [11,50–51] and Hippo [52,53], among other potential candidates.

Figure 2. . Most common gene mutations in hepatocellular carcinoma.

Mutation data taken from [54,55]

Transcriptome analysis of HCC by Boyault et al. [56] identified six different subgroups of HCC (G1-G6) that were characterized by specific clinical and genetic features. G1 tumors had low HBV copy number and increased expression of genes controlled by parental imprinting in hepatoblasts. G2 tumors included HBV-related HCC and PIK3CA and TP53 mutations. G1 and G2 were prominent in AKT pathway activation. G3 tumors had cell-cycle control related gene over expression, as well as TP53 mutations. G4 included HCC and hepatocellular adenomas with TCF1 mutations. G5 and G6 groups had Wnt/β-catenin pathway mutations. This particular study highlighted the fact WNT and AKT pathways were activated in approximately 50% of the analyzed HCC samples, opening up potential therapeutic targets [56].

A meta-analysis of gene expression profiles by Hoshida et al. analyzing data sets from eight independent patient cohorts across the world, including 603 patients with HCC, also reported different HCC subclasses (S1, S2, and S3). S1 was prominent on WNT signaling pathway aberrant activation, S2 had MYC and AKT pathways activation, and S3 was associated with hepatocyte differentiation. S1 tumors-related WNT pathway activation was noted as a result of transforming growth factor-B activation (unlike its common association with β-catenin mutation), thus representing a new mechanism of WNT pathway activation in HCC [57].

A recent study that utilized exome sequencing of 243 hepatocellular carcinomas identified relationships between different types of exposures and mutational signatures [14]. This study identified recurrent mutations in 161 genes and grouped tumors into specific mutational signatures based on the patterns of mutations identified. Further analysis identified 14 genes that were significantly mutated (TP53, CTNNB1, AXIN1, ALB, ARID1A, ARID2, ACVR2A, NFE2L2, RPS6KA3, KEAP1, RPL22, CDKN2A, CDKN1A, RB1), commonly deleted (CFH, IRF2, CDKN2A, PTPN3, PTEN, AXIN1, RPS6KA3) and amplified genes (TERT, VEGFA, MET, MYC, FGF-CCND1 JAK3, CCNE1). Analysis of the 161 commonly mutated genes found that they fell into 11 common pathways-telomerase activating mutations (including TERT promoter mutations [60%]), WNT/β-catenin pathway (54%), PI3K–AKT–mTOR pathway (51%), TP53 and cell cycle alterations (49%), MAPK (43%), hepatic differentiation (34%), epigenetic regulation (32%), chromatin remodeling (28%), oxidative stress (12%), IL-6 and JAK-STAT (9%) and transforming growth factor (TGF)-β (5%) – with analysis showing that in 28% of cases there was at least one potentially actionable gene mutation identified per tumor [14]. These finding suggest that genomic analysis gene mutations and other alterations can offer the benefit of off-label approved therapies or access to clinical trials to a significant portion of patients.

• CC

As with HCC, CC are an even more diverse set of tumors. In an effort to improve the outcomes of these patients, several molecular studies of CC have been performed, aiming to detect targetable oncogenic driving mutations, as well as predictive and prognostic genomic markers of response [58–65]. The substantial genomic variability noted in CC (Figure 3) is a reflection of its diversity but as well provides the rationale for such approaches based on each patient's CC's particular genetic makeup [66,67].

Figure 3. . Most common gene mutations in biliary cancers.

Mutation data taken from [54,55].

• Genetic & genomic changes in CC

A study by Miller et al. [67] investigated changes in gene expression and copy number in biliary cancers and correlated those changes with clinical features and outcomes. Gene expression and CGH analysis on 34 biliary tract cancer specimens found 545 genes with altered expression in extrahepatic cholangiocarcinoma (EHC) and 2,354 in intrahepatic cholangiocarcinoma (IHC).

Despite considerable heterogeneity in the extent of chromosomal instability between patients even within specific cancer subtypes, there appeared to be selected chromosomal regions that were commonly altered. CGH analysis revealed that short segments of chromosomes 1p, 3p, 6q, 8p, 9p, and 14q were commonly deleted across all cancer subtypes. Commonly amplified regions included segments of 1q, 3q, 5p, 7p, 7q, 8q, and 20q. Over-representation analysis revealed an association between altered expression of functional gene groupings and pathologic features, including mutated expression of a large number of cell cycle regulators including UBD, BCL2L2, CDC2, MCM2, and CDKN1C in all subtypes, as well as genes involved with regulation of cellular metabolism and biosynthesis and high pathologic grade. Vascular invasion was associated with mutated expression of genes involved with electron transport and cellular metabolism [67].

Moreover, two main biological classes of CC have recently been described by Sia et al. [66]. A group named the 'proliferation class' (62% of CCs in this study) had specific copy number alterations. These were high-level amplifications in five regions, including 1p13 (9%) and 11q13.2 (4%), and several focal deletions, such as 9p21.3 (18%) and 14q22.1 (12% in coding regions for the SAV1 tumor suppressor). Many features of the poor-prognosis signatures described for HCC were also noted in this class, as well as a worse outcome (p < 0.001) when compared with the other group of CCs. This group was also prominent on DNA amplifications at 11q13.2, deletions at 14q22.1, mutations in KRAS and BRAF, as well as characterized by activation of oncogenic signaling pathways (including RAS, MAP kinase and MET) [65].

The second group described by Sia was the 'inflammation class' (38%), which was characterized by activation of inflammatory signaling pathways, overexpression of cytokines, and STAT3 activation [66]. The STAT-1 has been reported as overexpressed nearly 9-fold in cases of cholangiocarcinoma [18]. These results emphasize the significant genetic diversity of CC and provide specific identifiers for classifying such tumors.

• Common & targetable mutations in CC

In the largest cohort of compiled CC cases to date, Ross et al. performed DNA sequencing of hybridization-captured libraries for 3320 exons of 182 cancer-related genes and 36 introns of 14 genes frequently rearranged in cancer. Two-thirds of patients in this study harbored targetable genomic alterations that have the potential to personalize therapy selection for individual patients [60]. The potential clinical applicability and relevance of next generation sequencing tumor profiling is particularly compelling in cases of rare or resistant tumors such as CC, where treatment guidelines are frequently limited and current clinical trial data is often insufficient. Given evidence of significant pleiotropic and biologic heterogeneity, there is a clear need for appropriate classification of subsets of CC patients during treatment choice selection, as these signatures seem not only prognostic but predictive for response to targeted therapy.

The most commonly described activating mutations in CC genomic studies are KRAS [61,62], HER2 [63], MET [64], as well as hotspot activating missense mutations in genes downstream to EGFR, such as the BRAF [65].

A recent study using Caris Molecular Intelligence™ (Caris Life Sciences, TX, USA) to analyze tumors from 42,000 patients, covering 150 histological subtypes, reported the benefits of tumor profiling of common and rare types of cancer such as CC in order to identify potential treatment targets. Out of these patients, 14,700 were categorized as having rare tumors, defined as a patient segment that is typically underserved by current guidelines. Results from the study included a 16.3% frequency of KRAS mutations in cholangiocarcinoma [68].

Saha et al. [69] further investigated the presence of mutant IDH1 and IDH2 mutations in the liver of genetically engineered mice as a potential biomarker, finding expression of mutant IDH in liver cells to cause impaired hepatocyte differentiation, and markedly elevated levels of cell proliferation. Mutant IDH proteins were associated with activated KRAS, thus promoting malignant transformation and tumor progression to metastatic CC [69]. Interestingly, IDH mutations are present only in intrahepatic CC and enriched in the cholangiolar type.

The MAPK pathway comprises several important targets for targeted therapy. Specific mutations in BRAF can stimulate this pathway, and the presence of a V600E or K mutation predicts responsiveness to BRAF inhibitors or MEK inhibition. BRAF is located on chromosome 7q34. It is a member of the Raf kinase family that plays a critical role in regulating the cell proliferation through the MAPK pathway. BRAF mutation-driven CCs have been reported at a frequency between 0 and 22% [70–73]. Our group recently reported the successful use of dual BRAF and MEK inhibition for the management of a V600E mutant ICC [73].

• Combined hepatocellular-cholangiocarcinoma

Combined hepatocellular-cholangiocarcinoma (CHC) is an intermediate form between HCC and CC, suggesting phenotypic overlap between these tumors, with poorly understood pathogenesis. Woo et al. [74] performed gene expression profiling of human HCC, CHC, and CC, identifying a novel HCC subtype, described as cholangiocarcinoma-like HCC (CLHCC), which expressed cholangiocarcinoma-like traits (CC signature). CLHCC co-expressed embryonic stem cell-like expression traits suggesting a bi-potent hepatic progenitor cell origin. The study suggested that the acquisition of cholangiocarcinoma-like expression traits was a crucial step in progression of HCC, associated to worse prognosis [12,74].

• Promises & pitfalls of next-generation sequencing

Genomic sequencing through the use of targeted gene panels, exome, whole genome and transcriptome sequencing has changed the landscape of medical oncology. The identification of tumors containing mutations with targeted therapeutics available continues to grow, as has detection of mutations with a predictive or prognostic effect (e.g. KRAS in colorectal carcinoma). However, the identification of these mutations can be dependent on the type of methodology and bioinformatics analysis tools utilized.

The majority of tumors sent for sequence analysis are heterogeneous with varying mixtures of neoplastic and non-neoplastic cells. In addition, clonal evolution, loss of heterozygosity and aneuploidy can compound the difficulty in detection and interpretation of results. The annotation of clinical knowledge for less common genes and mutations is often lacking with case reports difficult for treating medical oncologists to utilize for patient treatment. These deficiencies have been recognized and large national and international trials are beginning to address some of these concerns.

• Translating the data: molecular tumor boards & promising targets

Several evidence-based commercial tumor profiling services are fast becoming part of current management of oncology patients, and adoption, knowledge and integration of these genomic data for development of personalized regimens is extremely relevant. Many centers across the globe are incorporating multidisciplinary molecular tumor boards to address those needs, making decisions on the most appropriate regimens based on genomic tumor information [73,75].

The identification of promising targets in hepatobiliary cancer that are being discussed include tyrosine kinase growth factor receptors overexpression such as HER2, EGFR, and MET/HGF, (summarized in Table 1), which as previously described may play important roles in the development and pathogenesis of HCC and CCs [40–44]. Nakasawa et al. [76] reported significant overexpression of ERBB2 (HER2/neu) in extrahepatic bile duct tumors (5.1%). EGFR overexpression (8%) and gene amplification (77%) were also noted. MET overexpression was most frequent in IHC (21.4%) but was not associated with gene amplification. In addition, it has been reported that patients with CC may also harbor gene rearrangements of FGFR2 that could be targetable by agents such as Regorafenib [77–79]. Sia et al. reported that 45% of CC patients harbored at least 1 FGFR2 fusion [66]. Gu et al. confirmed the presence of ROS kinase fusions in CC patients [80]. The expression of ROS fusions confers a transforming ability both in vitro and in vivo and is responsive to its kinase inhibitor, similar to the effect noted in lung adenocarcinoma [81].

Table 1. . Clinical utility for targeted mutation testing in hepatocellular carcinoma.

Gene (ordered by frequency)	Type of alteration (estimated frequency)	Examples of therapeutic agents†	Rationale
AXIN1	LOF (liver: 7%)	None at this time	Wild-type protein inhibits activation of Wnt/β-catenin signaling; may be sensitive to Wnt inhibitors
ARID1A	LOF (CC: 15–21%; liver: 14.3%)	None at this time	ARID1A interacts with EZH2 and inhibition of EZH2 acts as a synthetic lethal manner in ARID1A-mutated cells. EZH2 inhibitors‡
ARID2	LOF (liver: 3%; CC: 2.5%)	None at this time	ARID2 signaling is involved in IFN signaling; may predict insensitivity to IFN if mutated‡
ATM§	Biallelic inactivation (liver: 5.2–19%; CC: 5%)	PARP inhibitors	Biallelic inactivation predicts sensitivity to PARP inhibitors
BAP1	LOF (liver: 0.4%; CC: 17.5%)	None at this time	HDAC inhibitors‡; EZH2 inhibitors‡
BRAF	Mutation, amplification, translocation (liver: 0.4%)	Vemurafenib, dabrafenib, RAF inhibitors, MEK inhibitors	Sensitivity to RAF inhibitors; amplification may predict resistance to MEK inhibitors
BRCA1§	LOF mutation, deletions (liver: 3.6%)	PARP inhibitors	Biallelic inactivation predicts sensitivity to PARP inhibitors
BRCA2§	LOF mutation, deletions (liver: 3.6%)	PARP inhibitors	Biallelic inactivation predicts sensitivity to PARP inhibitors
CDKN2A	LOF mutation, biallelic deletion (liver: 5.6%; CC: 12.5%)	CDK4/6 inhibitors	Biallelic inactivation leads to uncontrolled growth through decreased binding to CDK4/6
CTNNB1	GOF mutation (liver: 14.3–27.4%)	WNT inhibitors	May predict sensitivity to inhibitors of WNT signaling; resistance to EGFR, PI3K and AKT inhibitors
EGFR	Mutations, amplifications, intragenic deletions (liver: 4.2%)	Erlotinib, gefitinib, new-generation EGFR inhibitors	Mutations targetable with TKIs; mutations predict resistance to TKIs. Amplification associated with response to anti-EGFR therapies
FBXW7	Biallelic inactivation (liver: 0.4%; CC: 6.7%)	MTOR pathway inhibitors	May predict sensitivity to MTOR inhibitors; may predict resistance to antitubulin chemotherapies
FGFR2	Amplification, GOF mutations (liver: 1.7%; CC: 10%)	Dovitinib, FGFR inhibitors	May predict response to targeted TKI inhibitors
HNF1A	LOF (liver: 1.6%)	None at this time	Unknown at this time
IDH1	GOF mutations (liver: 1.6%; CC: 20%)	IDH inhibitors, vaccine	May predict sensitivity to IDH inhibitors
IDH2	GOF mutations (liver: 3.7%; CC: 5%)	IDH inhibitors, vaccine	May predict sensitivity to IDH inhibitors
IL6ST	GOF (liver: 3.6%)	None at this time	Unknown at this time
JAK1	GOF mutation (liver: 3.2%)	Ruxolitinib	Activating JAK1 mutations activate STAT signaling; ruxolitinib inhibits JAK1/2 signaling
KRAS	GOF mutations (liver: 1.3%; CC: 37.5%)	None at this time	Mutations can predict resistance to therapies (EGFR, cetuximab); may show sensitivity to combined therapies with MEK inhibitors
MLL3 (KMT2C)	LOF (liver: 9.4%; CC: 25%)	None at this time	May be sensitive to HDAC inhibitors‡
NFE2L2 (NRF2)	Complex: LOF GI (GOF other tumors) (liver: 5.7%)	Tecfidera	Tecfidera mechanism of action is through activation of the NRF2 pathway; mechanism unknown
PBRM1	LOF (liver: 0.9%; CC: 17.5%)	None at this time	Unknown at this time
PIK3CA	GOF mutations (liver: 4.2%; CC: 7.5%)	PI3K/AKT/MTOR inhibitors	Sensitivity to PI3K, AKT, MTOR inhibitors
PTEN	Loss of function mutation, biallelic deletion (liver: 1.7–6.7%; CC: 12.5%)	Inactivation may predict sensitivity to PI3K, AKT, MTOR inhibitors	Biallelic inactivation
RB1	Biallelic inactivation (liver: 5.5–11.3%; CC: 6.1%)	None at this time	Resistance to CDK4/6 inhibitors
RNF43	Biallelic inactivation (CC: 37.5%)	None at this time	Mutations derepress Wnt/β-catenin signaling; may predict sensitivity to WNT inhibitors‡
SMAD4	Biallelic inactivation (liver: 0.9%; CC: 2.5%)	None at this time	Mutation or deletion may lead to sensitivity to TGF-β inhibitors‡; resistance to 5-FU therapy
SMARCA4	LOF (liver: 1.7%; CC: 6.7%)	None at this time	Unknown at this time
TERT	Amplification (liver: 6.2%)	Immunotherapy, telomerase inhibition‡	Mutations lead to increased telomerase expression
TP53§	LOF mutation, deletion (liver: 61.9%; CC: 20%)	Wee1 inhibitors, Chk1 inhibitors, kevetrin, APR-246, nutlins, gene therapy	Biallelic inactivation

Percentages shown here are in whole or part based upon data generated by the TCGA Research Network [82].

^†Targeted therapy may be in research setting only and may not have current trial open.

^‡Speculation based on in vitro data, untested functionally and in model systems/humans.

^§May signal the presence of a germline mutation.

GOF: Gain of function; HDAC: Histone deacetylase; IFN: Interferon; LOF: Loss of function; PARP: PolyADP ribose polymerase.

STAT-1 protein transcription factors are activated by the JAK, and JAK/STAT pathway inhibitors is another potential therapeutic approach [67,77–78]. Similar to HCC, ubiquitin-related genes have been reported as significantly altered in IHC [27,67,79]. Proteasome inhibitors and inhibition of other mediators of ubiquitin-dependent protein degradation are currently under investigation [80,83]. Those pathways are the major subject of current ‘basket’ trial designs involving targeted therapies incorporating the genotype-to-phenotype concept, enrolling patients harboring specific mutations in pertinent Phase II trials exploring molecular targeted therapies or chemotherapeutical regimens, unlike a disease-specific (or phenotype-to-genotype) approach [81,84–85].

Conclusion & future perspective

Recent advances in NGS, transcriptomics, proteomics and epigenomics in liver cancer have exponentially expanded our knowledge about the molecular events involved in the pathogenesis and natural history of HCC and CC [86]. Despite being at an early development phase, continuous efforts in translational oncology will undoubtedly continue to integrate these data into the clinic and subsequent personalized genomic approaches [87–91].

Particularly regarding HCC, and unlike other malignancies, the notion to have tumor biopsies to undergo NGS studies is a relatively novel concept, as the diagnosis of HCC still relies on imaging studies and serum alpha-fetoprotein levels in most cases, except in the rare event where liver cirrhosis is not documented. This has led to a significant gap between the progress of genomic study and its clinical application for patients with liver cancer, which should be overcome by the rapid implementation of diagnostic algorithms that include procurement of tissue for diagnostic, prognostic and therapeutic purposes.

One major challenge remains in the validity of these tests and platforms to guide treatment decisions, given the heterogeneity of the data and lack of prospective proof-of-concept studies. The development and integration of basket clinical trials with such data is critical and warranted for that purpose. Strong bioinformatics support and homogenization of NGS platforms are also essential in achieving that goal [92].

The ever-growing availability and affordability of DNA microarray technologies and profiling, and the advancement of other global initiatives such as the 1000 Genomes Project and the Cancer Genome Atlas will certainly improve our understanding of the pathophysiology of liver cancer and influence future treatment decisions in clinical oncology, with the ultimate goal of improving our patient's outcomes.