Strategies for hepatocellular carcinoma therapy and diagnostics: Lessons learned from high throughput and profiling approaches


  • Potential conflict of interest: Nothing to report.

  • Supported by the German Research Foundation (SFB/TRR77), the Helmholtz Alliance on Immunotherapy of Cancer, and the German Federal Ministry for Education and Research (Virtual Liver Network).


Over the last decade, numerous small and high-dimensional profiling analyses have been performed in human hepatocellular carcinoma (HCC), which address different levels of regulation and modulation. Because comprehensive analyses are lacking, the following review summarizes some of the general results and compares them with insights from other tumor entities. Particular attention is given to the impact of these results on future diagnostic and therapeutic approaches. (HEPATOLOGY 2011;)

Hepatocellular carcinoma (HCC) is a unique tumor entity by several measures. Its causes (chronic viral hepatitis, alcoholic and nonalcoholic steatohepatitis, aflatoxins, and several hereditary diseases [e.g., genetic hemochromatosis]) are much better defined than in other adulthood cancers and are demonstrable in approximately 90% of cases. Consequently, HCC is the most relevant paradigm of virus- and inflammation-associated cancer. This opens a large field for primary (immunization, specific hygienic measures) and secondary (screening programs, therapy of predisposing diseases) prevention, but it is also the reason for another peculiarity: conflicting morbidity. The majority of HCC patients (>80% in highly industrialized countries) have chronic hepatitis or cirrhosis, which influences disease progression and may severely restrict patient prognosis, therapeutic options, and design of clinical trials.1 The morphological sequence of premalignant and early malignant changes is well defined, but the lesions are hardly accessible by diagnostic means,2 which constitutes a significant difference from colon, breast, or skin cancer. Well-characterized model systems are available for mechanistic as well as preclinical analyses,3 and HCC cell lines have been workhorses for biochemical as well as molecular biological and recently even systems biological analyses, providing a wealth of basic research data for current translational approaches.

Because of the obstacles characterized above, HCC has long been an orphan tumor disease with regard to translational research efforts, clinical trial perspectives, and therapeutic options, which stands in sharp contrast to its enormous clinical relevance. HCC is the sixth most frequent cancer and the third most frequent cause of cancer-related death, and numbers of cases are rising, even in industrialized countries.4 Nevertheless, knowledge about molecular pathogenesis of HCC is lagging behind other major tumor diseases, such as breast and colon cancer, where multiple systemic treatment options are starting to convert in many cases previously untreatable metastatic tumors into a chronic disease. Recently, this picture has started to change for HCC and has gained some momentum: The growing Chinese economy has fueled the industry's interest and improved options for clinical trials and novel therapeutics. The successful SHARP (Sorafenib HCC Assessment Randomized Protocol) trial established sorafenib as the first effective and approved systemic treatment for HCC and proved that despite all obstacles, trials employing systemic treatments can be successful.5 Other treatment options such as radioembolization6 and oncolytic approaches7 have entered the field. Meanwhile, more than 150 phase 1 to 3 trials are ongoing, but only some of them are based on rational approaches using knowledge about molecular pathogenesis of human HCC. Comprehensive, large-scale profiling approaches on representative collectives are missing so far, but numerous analyses have been performed at the genomic, epigenetic, and expression level, providing insight into relevant mechanisms, targets, as well as markers, suggesting future strategies for systemic HCC treatment.


CGH, comparative genomic hybridization; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; miRNA, microRNA.

Technical Aspects

Genomic, microRNA (miRNA), and some protein-based assays are more robust and less vulnerable to influences imposed by imperfect sample conditions. These approaches are well applicable to formalin-fixed, paraffin-embedded samples and thus have found their way into even routine diagnostics. In contrast, assays based on larger coding or noncoding transcripts depend highly on material preservation and assay conditions. This does not restrict their potential as exploratory technologies, but it impedes their comparability and restricts meta-analyses and diagnostic applicability. Currently, methylation analyses pose significant challenges in data acquisition as well as interpretation. Broad spectrum proteomic or metabolomic approaches are certainly further away from application and have not been used for significant HCC collectives.

Profiling data analyses can be performed in unsupervised and supervised fashion. Although unsupervised analyses are believed to be less biased, in most cases geographic parameters or the fact that, for example, only resection specimens are used inherently influences data interpretation. However, due to profound knowledge about its etiology, translational HCC research needs hypothesis-driven, supervised analyses guided by epidemiological, clinical, or experimental nominators to identify factors modulating its development or progression. These factors may include viral and nonviral etiology,8, 9 sex,10 tumor recurrence,11 intrahepatic metastasization,12 response to therapy,13 and fetal-type gene expression pattern.14 Knowing and controlling this bias impeding all supervised and unsupervised HCC analyses is of utmost relevance for drawing conclusions and making strategic decisions.

The source of the tissue samples is an important bias, because etiology varies dramatically depending on the geographic region of origin.4 Hepatitis B virus (HBV) etiology is less frequent, and aflatoxin-based effects are usually absent in collectives from Western industrialized countries compared with countries in eastern Asia and southern Africa, whereas the effects of alcohol consumption and metabolic syndrome are more prevalent. Furthermore, significant differences exist in the relative frequency of HBV versus hepatitis C virus (HCV) infection. In addition to geographic differences, collectives based on resection specimens address limited disease and are biased for nonmetastatic, less aggressive tumors of presumably better spontaneous course, and also for a lower frequency of cirrhotic changes in the nontumorous liver. These factors have already been demonstrated to correlate with differences in the results of the respective analyses; thus, we currently have no single analysis in hand that is truly unbiased. Consequently, many array-based analyses have obtained inconsistent and partly contradictory results. One possible way to control this problem is through meta-analyses that integrate as many data from different studies as possible or reflections comparing results from different types of studies. Unfortunately, these approaches suffer from variations in regard to technical platforms, sample preparation and evaluation, normalization, and data analyses; they have successfully been performed with robust genomic data,15 but recently some success has also been achieved using transcriptomic HCC data.16 There is a strong need for comprehensive analyses addressing several levels of regulatory processes in a single collective and for analyses of collectives that are less biased; the results are likely to differ from those obtained so far. Whether ongoing large-scale but still biased efforts for systematic analysis of cancer genomes such as the International Cancer Genome Consortium will improve this specific situation in HCC has yet to be seen.

Screening for Aberrations

Genetic Analyses

Historically, comparative genomic hybridization (CGH) represented the first molecular method to screen tumor tissue for genetic changes in a comprehensive manner. More than 40 single studies in human HCC have elaborated recurrent chromosomal imbalances that correlated with etiology (e.g., losses of 4q, 8q, 13q, and 16q with HBV; losses of 8p in HCV-negative cases) or tumor progression (losses of 4q and 13q).15 Self-organizing tree algorithms identified gains of 1q21-23 and 8q22-24 as early and the gain of 3q22-24 as late genomic events, demonstrating sequential gain of genetic instability.18 In contrast to conventional CGH, array-CGH approaches provide higher genomic resolution and therefore allows one to scale down the correlations of more and smaller aberrations with clinicopathological features such as microvascular invasion and tumor grading.19 Moreover, specific alterations (e.g., 1q32.1, 4q21.2-32.33) discriminate between HBV- and HCV-associated HCCs,9 and the high resolution of this technique allows for the precise delineation of respective candidate oncogenes and tumor-suppressor genes, as demonstrated for Jab1, YAP, and Mdm4.9, 20, 21 In summary, three main conclusions can be drawn from these studies: (1) HCC is a chromosomally instable cancer that, in general, accumulates high numbers of macro- and microimbalances; (2) early chromosomal imbalances precede malignant transformation, because they are detectable in a significant number of premalignant lesions; and (3) etiology matters, because several chromosomal macroimbalances correlate with the underlying cause of the HCC. The reason for this observation has not been clearly defined.

Mutational activation and inactivation of individual genes are frequently observed in most HCCs and represent protumorigenic events independent of genomic instability. Here, especially loss-of-function as well as gain-of-function mutations in TP53 facilitate tumor proliferation, cell migration, and cell survival.22 In addition, several mutations with low or moderate frequency have been described for HCC, for example, in AXIN1/2,23CTNNB1,24, M6P/IGF-2R,25TCF1/HNF1α,26PIK3CA,26K-RAS,27 and p16/CDKN2/INK4A28 (Table 1). Data collected so far demonstrate that few high-frequency mutations and many low-frequency events contribute to the molecular heterogeneity of HCC. The spectrum of mutations will certainly be expanded by high-throughput sequencing technology in the near future.

Table 1. Low- and High-Frequency Mutations in HCC
GeneMutation FrequencyReferences
  • *

    Vinyl chloride exposition.

  • Abbreviations: CDKN2, cyclin-dependent kinase inhibitor 2; CTNNB1, catenin (cadherin-associated protein) beta 1; HNF1α, hepatocyte nuclear factor-1 alpha; IGF-2R, insulin-like growth factor-2 receptor; K-RAS, v-Ki-ras2 Kirsten rat sarcoma viral oncogene; M6P, mannose-6-phosphate; PTEN, phosphatase and tensin homologue; SMAD, small mothers against decapentaplegic; TCF1, transcription factor 1.

Low frequency (<20%)  
 Axin 15%-13%23,26,82
 Axin 23%23
 K-RAS1% (33%*)26,27
High frequency (>20%)  
 TP53 (low aflatoxin exposure)10%-28%100,101
 TP53 (high aflatoxin exposure)48%100

Epigenetic Analyses

In carcinogenesis, global DNA hypomethylation has been associated with activation of oncogenes and genomic instability,29 whereas hypermethylation of CpG (cytosine guanine dinucleotide) islands located especially in gene regulatory sequences (e.g., of the Ras target RASSF1A, the adhesion molecule CDH1, and the cell cycle regulator p16/CDKN2/INK4A) resulted in transcriptional silencing.26, 30 Methylation changes may occur early in the process of cancer development, and CpG island hypermethylation of regulatory regions of tumor-relevant genes is a frequent event accumulating in multistep hepatocarcinogenesis.31 Only a few studies have analyzed the global and promoter-specific levels of DNA methylation in hepatocarcinogenesis. First published data have revealed clear differences in DNA methylation between HCC and surrounding nontumorous tissue based on specific promoter hypermethylation and global hypomethylation.32 In this regard, genomic hypomethylation correlated with genomic instability in HCC, whereas methylation of CpG islands was associated with poor prognosis.33 In addition, DNA methylation status correlated with tumor recurrence after hepatectomy, cancer-free survival, and overall survival.34 Using class comparison analysis, HBV-, HCV-, and alcohol-specific promoter methylation patterns have been described, suggesting etiology-dependent methylation in early stages of hepatocarcinogenesis.32 Knowledge about these modifications in tumorigenesis is certainly fragmentary, but epigenetic analyses may represent valuable tools for diagnosis and classification in the early stages of liver tumor development.

Transcriptional Analyses

Coding Transcripts.

Most transcriptomic studies in HCC have used cDNA or oligonucleotide high-density microarrays. Despite varying technical platforms, biological controls, and mathematical algorithms, these approaches have identified partly novel tumor-relevant genes and networks (e.g., PEG10, insulin-like growth factor-II [IGF-II], Claudin10, RhoC, AP-1, and cell cycle regulators).14, 35-37 Some studies have correlated expression profiling data in HCC with etiology,8 vascular invasion,38 drug response,13 recurrence,12 and survival.36 Unsupervised clustering of transcriptomic data provided subtyping of HCC that was related to tumor-associated inflammation as well as tumor cell proliferation and apoptosis.35, 39 Furthermore, specific expression signatures derived from global gene expression analyses correlated well with the histological classification of premalignant lesions (low- and high-grade Dysplastic Nodules) and HCCs.40 Ye et al.41 also demonstrated that transcriptomic signatures significantly differed between HCCs with and without metastatic spread, whereas expression profiles of respective primary and metastatic tumors varied only by a few genes.

Hierarchical clustering has revealed that HCCs can be divided into subgroups based on transcript profiles. Lee et al.36 described the existence of two distinct groups of HCC characterized by poor survival specifically within the group that showed high expression of genes involved in proliferation and antiapoptosis. This study further demonstrated that the transcriptional pattern of HCCs that shared a signature with fetal hepatoblasts exhibited poor prognosis.14 Yu et al.42 demonstrated an association between their identified subclasses and tumor dedifferentiation (grading G1/2 versus G3/4) as well as overall survival.

Despite all limitations, a recent meta-analysis integrating a high number of HCC data (>600) from independent gene expression profiling analyses was able to demonstrate the existence of three distinct molecular subclasses (S1-S3) and confirmed some important previous findings (e.g., activation of Wingless pathway, existence of a (proliferation) signature). However, it also exemplified some of the difficulties ahead by, for example, showing that activation of typical Wingless-dependent gene expression did not correlate with mutations in CTNNB1.16 In summary, transcriptional signatures have allowed for classification of HCCs according to their molecular and biological characteristics26 and have turned out to be a valuable tool in the identification of tumor-relevant genes and pathways in human HCC.

Noncoding RNA/miRNA.

A steadily increasing number of studies have examined differential expression of noncoding RNAs (especially miRNAs) in HCC. miRNAs bind complementary sequences in the 3′ end of messenger RNAs and therefore represent effective posttranscriptional regulators of mammalian gene bioactivity.43 In addition, miRNAs directly affect promoter activity through binding and/or modifying DNA methylation.44, 45 So far, miRNAs with oncogenic or tumor-suppressing potential have been identified,46, 47 and recent results indicate that different stages of hepatocarcinogenesis as well as liver tissues with HBV or HCV infection can be differentiated from each other based on their miRNA fingerprints.48 This is supported by other studies showing that distinct miRNA signatures were associated with alcohol consumption and HBV infection,49 tumor differentiation and progression,46, 50 metastasis, survival, and relapse.51, 52 Although the key miRNAs identified significantly differed between various studies,48, 49, 53 some miRNAs such as miR-122a48, 49, 54 and miR-22347, 48 have recurrently been identified through independent approaches. Using specific antagomirs and agomirs, it is possible to associate distinct miRNA activity with cellular processes, but because each miRNA potentially interacts with several different targets, it is difficult to define the precise mechanisms and pathways by which miRNAs mediate their biological effects. Recently, reduced miR-26 expression was linked to NF κB and interleukin signaling, shorter survival, and better response to IFN-α therapy.55 In addition, independent studies demonstrated the relevance of other miRNAs such as miR-139,56 miR-125b,57 miR-221,46 and miR-18158 in the regulation of tumor-relevant proteins and processes in hepatocarcinogenesis. Recent data have shown that on the basis of c-MYC–dependent miRNA signatures detected in hepatoblastoma, it is possible to discriminate between HCCs with an invasive phenotype and patients with lower survival probability.59 Based on their high stability even in formalin-fixed, paraffin-embedded tissues, miRNAs represent promising molecular markers for diagnostic HCC classification, prognostic stratification, and drug response prediction even under routine clinical conditions.

Protein Analyses

Proteomic analysis (e.g., by two-dimensional gel electrophoresis, or MALDI-TOF [matrix-assisted laser desorption/ionization time-of-flight] or SELDI-TOF [surface-enhanced laser desorption/ionization time-of-flight] mass spectrometry) is believed to be more informative than other screening approaches because proteins represent the main functionally active principle in cells. Moreover, only moderate correlations between mRNA and protein abundance demonstrate the value of protein assays for biomarker identification in blood and liver tissues.60, 61 For HCC, distinct protein profiles that discriminate between HBV- and HCV-associated tumors have been identified.62 In addition, many differentially expressed proteins (partly derived from patients' sera) in HCCs have been described, including heat shock protein 90 (HSP90)63 and stathmin.64 However, because in the different studies the number of identified proteins is relatively low (normally far below 100), protein-based assays have been of limited value for subtyping of HCCs so far.

Many studies have used immunohistochemistry for the analyses of distinct factors and protein families in HCC. These include signaling pathway constituents (e.g., β-catenin,24 different FZD receptors,65 and c-MET66), cell cycle regulators (e.g., p16/CDKN2/INK4A,27 and survivin67), and transcriptional regulators (e.g., p53,24 c-MYC,24, 68 FBPs,69 YAP,70 and SMADs71, 72). In addition, these analyses discriminate between different subcellular localizations of proteins (e.g., for β-catenin or p53) and provide the possibility of assessing protein expression in a semiquantitative manner using high-throughput imaging systems and respective mathematical analysis algorithms. In conclusion, protein-based assays may provide the most relevant information with regard to levels and location of bioactive units in cells; however, technical limitations currently prevent these approaches from being available for unbiased analyses in larger HCC collectives.

Integrative Approaches

Integrating different types of analyses in HCC has supported the identification of genes and pathways that are often aberrantly regulated by several different low-frequency mechanisms. For example, aberrant activation of pleiotropic growth factors, receptors, and their downstream signaling pathway components represents a central protumorigenic principle in hepatocarcinogenesis. Constitutive dysregulation of these pathways (e.g., hepatocyte growth factor/c-MET, IGF/IGF1 receptor, transforming growth factor β [TGFβ]/epidermal growth factor receptor, and TGFβ/TGF receptor signaling) occurs at different mechanistic levels, such as regulation of expression, cell type–specific and subcellular localization, and mutational (in)activation.73, 74 In addition, crosstalk between pathways and with other tumor-relevant factors (e.g., HSPs, COX-2, and p53) further demonstrate that integrative approaches including genomic, transcriptomic, and protein analyses are necessary to understand the complex and dynamic interplay between different oncogenic modules and pro/antioncogenic mechanisms.

Moreover, integrative approaches have started to support classifying groups of HCCs according to specific overall molecular characteristics, prognostic impact, and even some predictive implications. Laurent-Puig et al.75 identified two molecular subgroups characterized by high chromosomal instability (associated with Axin-1 and TP53 mutations) or more stable conditions (associated with CTNNB1 mutations). HCCs from the first group were less differentiated, more frequently exhibited HBV infection, and in the case of loss of heterozygosity of 9p and 6q, showed poorer prognosis. Based on array-CGH analyses, Katoh et al.76 equally identified genetically homogenous classes of HCC (two clusters and six subclusters). In contrast to the previous study, specific chromosomal alterations (gains of 1q, 6p, and 8q and losses of 8p) in one cluster were associated with high chromosomal instability and poor patient survival. In addition, no correlation of CTNNB1 or TP53 mutations to any of the groups was detectable. Of relevance, some subclusters harbored genomic amplifications of genes involved in mammalian target of rapamycin (mTOR) and vascular endothelial growth factor (VEGF) signaling. Based on the integration of genomic data and gene expression profiles, Woo et al.77 identified 50 potential driver genes in HCC. In fact, tumor class defined by the expression of this signature predicts the prognostic outcome of patients with HCC.

Boyault et al.26 described the existence of six HCC groups (G1-G6) that were characterized by distinct clinical and molecular features. For example, G1-G3 tumors exhibited more chromosomal instability and a tendency for poorer prognosis than G4-G6 HCCs. TP53 mutations accumulate in the subgroups G2 and G3, whereas mutations in CTNNB1 are characteristic for G5 and G6 tumors. Interestingly, this grouping showed some similarities with the molecular classification of Lee et al., such as the existence of groups with chromosomal instability, poor survival, and hepatoblast characteristics.14, 36 By integration of genomic, transcriptomic, and protein information of HCV-associated HCCs, Chiang et al.39 defined five molecular classes of HCC that partly overlapped with previously described groups. These classes were characterized by known molecular features such as activation of the Wingless pathway, proliferation, chromosomal instability,26 and induction of interferon-stimulated genes.35 In addition, this study identified a new subgroup characterized by polysomy of chromosome 7. Overall, there is increasing evidence that robust and clinically relevant subclassification requires integration of different technologies; future analyses will have to elaborate minimal diagnostic patterns in order to allow rational up-front testing.

Lessons Learned from Profiling Analyses

Although comprehensive analyses of all aspects in a representative collective of HCCs are missing and the existing data are either incomplete or biased, several conclusions can be safely drawn.

First, the vast majority of HCCs represent a classical chromosomally instable tumor (CIN-phenotype) carrying multiple genomic imbalances that is comparable to other adulthood malignancies, such as breast and sporadic colon carcinoma or pancreatic cancer. Microsatellite instability (mismatch repair deficiency) and methylator phenotypes seem to have little, if any, overall impact. Chromosomal instability increases during tumor progression, suggesting it as a driving factor.78 Because genomic macroimbalances at some genomic loci (but not global chromosomal instability) can be observed already in premalignant Dysplastic Nodules,15 an important question is: Which processes force chromosomal instability or govern relative chromosomal stability, and at which stage during tumor progression are they altered? There is experimental evidence for p53-dependent senescence surveillance that contributes to chromosomal stability,79, 80 but it is unlikely to be the only factor. Molecular mechanisms of chromosomal stability governance may not only offer insights into mechanisms of HCC progression but may also be valuable therapeutic targets even in progressed, chromosomal-instable HCC.

Second, somatic mutations in HCCs (in addition to chromosomal imbalances) include point mutations, small/medium size insertions and deletions, and balanced chromosomal translocations. Some of the HCC mutations are of high frequency (such as codon 249 mutations in TP53 in aflatoxin-exposed collectives),81 but the vast majority of these mutations affect only small subpopulations of HCCs, as has been demonstrated for Axin mutations.23,82 Mutational analyses of single genes in different HCC populations underscore the molecular heterogeneity of human HCCs as demonstrated also by chromosomal imbalances. Comprehensive mutational data for HCC collectives are lacking and will be collected in the frame of the International Cancer Genome Consortium.17 The first deep sequencing data have been obtained in other adulthood type cancers, such as colon, breast, and pancreatic cancer, where mutations in coding regions of cancer genomes average about 60-80 mutations and in some cases may even reach triple digits.83, 84 Thus, the number of mutations is much higher than previously expected; this is even more evident in light of the fact that these mutations are complemented by epigenetic changes and gross chromosomal imbalances. The situation becomes more complex, considering the impact of these mutations, because data from breast and colorectal cancer suggest that some of them are driver mutations, whereas the vast majority of mutations may be associated only with small fitness advantages.83 There is little doubt that the situation in HCC will be comparable, but the specific impact for tumor therapy is unknown and remains to be analyzed.

Third, there is convincing evidence that etiology leaves its molecular traces in the tumor genome, leading to specific genomic imbalances, mutations, epigenetic changes, and resulting alterations in host gene expression. Whether these effects are direct consequences of the specific carcinogenic mechanisms (e.g., exerted by HBV integrations or direct genotoxic effects of mycotoxins) or represent indirect effects due to functional selection of complementary protumorigenic mechanisms cannot be answered globally. Nevertheless, etiological “fingerprints” offer insight into the stepwise process of molecular carcinogenesis and the interrelation of different oncogenic mechanisms and provide openings for secondary preventive strategies. HCC was one of the first and is certainly one of the best-studied paradigms for molecular cancer epidemiology,81 and this may fuel the search for as-yet undetected etiological mechanisms.

Fourth, comprehensive approaches in other tumor entities, such as breast, colon, and pancreatic cancers have convincingly shown that in common solid cancers of adulthood, a surprisingly high number of pathways (≈12-15) is altered in a protumorigenic manner.83, 84 These different affected pathways seem to cover most of the tumor-relevant functions, but at the same time significant functional overlaps exist between them.74 As outlined above, current evidence for HCC points in the same direction. Frequently affected pathways and effectors include Wnt signaling, growth factor–induced signaling (e.g., IGF and TGFβ), or cellular gatekeepers such as p53. Matching affected pathways and underlying molecular changes shows that these pathways can be altered at different points, which has already been proven for Wnt/Wingless signaling (e.g., Axin-1/Axin-2 and β-catenin mutations, increased cadherin-17).23, 24, 82, 85 Interestingly, growth factor research in HCCs has shown that in each given pathway, frequent typical alterations (nodal points?) exist, but these changes differ between the pathways, varying from aberrant ligand expression (e.g., IGF-II)86 to receptor bioavailability (e.g., c-MET)66 to alterations in intracellular signal transducers (e.g., TGFβ signaling).71, 72 The cause for these observations is unknown, but it may offer some hints about how therapeutic approaches should be designed in order to target essential points of interference. It is reasonable to conclude that the spectrum of molecular changes present in a given HCC also contains stochastic elements largely selected by function. There is no evidence for a dominant driver mechanism and resulting addiction to it, as can be observed in several childhood malignancies and gastrointestinal stromal tumor.

Finally, comprehensive analyses have started and are likely to provide molecular subgrouping of HCC. Initial attempts have been made (e.g., by J. Zucman-Rossi and her group), clearly demonstrating the feasibility of the approach.26 Improvement can be expected from further meta-analyses of existing data and novel comprehensive analyses on well-characterized collectives. There is significant evidence that molecular classification reflects functional aspects and correlates with prognosis. At least some of the subgroups are likely to be relevant for therapy and predictive diagnostics, as exemplified by IGF-IR26,35 and mTOR-associated signaling.87

Consequences for Therapeutic Approaches

What are the consequences for drug development, clinical trials, and molecular (predictive) diagnostics?1, 88

There is certainly sufficient room and need for further (pathway) targeted approaches. Constitutive activation, for example, by mutation or ligand based stimulation of growth factor signaling pathways, is a common theme most likely relevant in every case of HCC.74 On the other side, many different pathways can be affected, and their functional consequences in regard to proliferation, motility, antiapoptosis, and angiogenesis significantly overlap. Thus, response to specific tyrosine kinase–directed approaches may be limited and can be expected only in subgroups of HCCs, and secondary resistance is likely to occur soon, because there is little if any evidence for a specific pathway addiction in HCC. From a mechanistic point of view, approaches to inhibit tyrosine kinase/growth factor signaling pathways should be as broad as possible and should consider complementary and combinatorial settings up front. Identification of patients who may benefit (more) from these approaches requires comprehensive biomarker analyses accompanying the clinical trails. This is state-of-the-art in most other malignancies, but has not been thoroughly respected in HCC, probably due to the fact that HCC is the only relevant tumor entity that does not necessarily require tissue-based diagnosis prior to therapy. Because molecular definition of responsive subgroups is not possible without tissue access, this difference may cause more trial failures than expected or necessary and may turn out to be a negative aspect of HCC in comparison with other tumor entities.

The fact that protumorigenic alterations in relevant pathways in HCCs may occur at different (nodal) points may limit the application of specific inhibitors and has to be respected in predictive diagnostic approaches as well as drug and subsequent trial design.88 A question that must always be addressed is the size of the responsive patient collective and whether it justifies the clinical and commercial effort. Some of these aspects can be better addressed by novel adaptive trial concepts.89, 90 It also has to be considered that drug companies may choose to lower the therapeutic benefit rather than the size of the patient collective amenable to treatment.

Because breakthrough achievements are unlikely to result from specific (pathway) targeted approaches in HCCs, our attention should also focus on mechanisms that are constantly needed by the tumor (that is, the tumors' “Achilles' heels”). These represent either necessarily required cellular functions that support a protumorigenic phenotype or are central mechanisms that allow for tumor persistence or progression. Examples of the first are the chaperone network (e.g., HSP90 and interacting factors)91 as well as all factors that support tumor cell proliferation and cell cycle progression. Tumor-associated neoangiogenesis may represent a double-edged sword: on one hand, it is an indispensable prerequisite for tumor growth; on the other hand, it is required to build up sufficient intratumoral drug concentrations. Recent results indicate that the effect of antiangiogenic approaches may depend on tumor characteristics (e.g., tumor cell biology and stroma content) that may need further attention.92

Examples for central tumor-relevant mechanisms may provide an even more attractive basis for therapeutic concepts. Global down-regulation of miRNAs is found in most tumors and suggests a role for the miRNA processing machinery. There is recent evidence for a critical role of dicer and some link to the p53 family members.93, 94 It will have to be shown whether this holds true in HCC and can be modulated in an antineoplastic manner. Tumor cell aneuploidy, as present in almost all HCCs, is a condition usually not compatible with cell survival under physiological conditions; this may explain the usually higher apoptosis rate of malignant tumors, but tumor cells must also have established mechanisms to prevail and maintain all vital cell functions despite the presence of significant aneuploidy. First screens have demonstrated genes that may provide increased aneuploidy tolerance;95 the future will show whether they may represent valid and innovative drug targets.

These considerations provide different challenges for drug design. Tumor cell specificity may not be achieved by addressing pathways or specific mechanisms that are more or less exclusive to tumor cells; instead, pharmacokinetics and pharmacodynamics may have to be modulated in order to favor tumor cell–associated activity or activation of the drug employing tumor preferential mechanisms.96

Predictive marker analyses do not play a role in current clinical diagnostics in HCC, but it will be necessary to include them in future clinical trials. Even if broader therapeutic approaches are tested, predictive marker analyses may well indicate response as well as primary and secondary resistance to therapy. Paradigms include microsatellite instability testing for chemotherapy response in colorectal cancer and ERCC1 (excision repair cross-complementing 1) testing regarding platinum-based chemotherapy in non–small cell lung cancer.97, 98 Clearly, the value of histological subtyping and molecular predictive diagnostics exceeds target gene evaluation.

Knowledge about molecular pathogenesis of HCC has dramatically improved in recent years, and some progress has been made (or is just ahead) in translation into clinical application,1 but there is room for improvement. In particular, comprehensive molecular analyses and further rationally designed clinical trials based on molecular evidence (e.g., targeting IGF-IR and mTOR) are eagerly awaited.99


The critical discussion and helpful comments of Hendrik Bläker and Federico Pinna are gratefully acknowledged.