Article Text

Download PDFPDF

Moving towards personalised therapy in patients with hepatocellular carcinoma: the role of the microenvironment
  1. Gianluigi Giannelli,
  2. Bhavna Rani,
  3. Francesco Dituri,
  4. Yuan Cao,
  5. Giuseppe Palasciano
  1. Department of Biomedical Sciences and Human Oncology, University of Bari Medical School, Bari, Italy
  1. Correspondence to Professor Gianluigi Giannelli, Dipartimento di Scienze Biomediche ed Oncologia Umana, Padiglione Semeiotica, Policlinico, Piazza G. Cesare 11, Bari 70124, Italy;


The goal of personalised therapy based on hepatocellular carcinoma (HCC) molecular characteristics is still beyond our grasp. Systemic treatments show poor efficacy, mainly because of the great heterogeneity of the tumour. Indeed, differences in aetiology, disease stage and biochemical composition of the fibrotic liver make cirrhosis itself a highly dyshomogeneous disease. Cancer cells grow in a cirrhotic microenvironment, interacting with stromal cells and engaging matrix components that differ from patient to patient, hampering the development of drugs to treat all patients. Growing evidence suggests a role for the cross-talk between HCC and the host stroma in driving disease progression and hence prognosis and survival. Many efforts have been devoted to identifying genes responsible for good or bad prognosis, but no study has yet proven helpful in guiding therapeutic choices and management over time, also taking into account the development of drug resistance. The questions of what to target and in which patient are still unsolved. In the personalised therapy scenario, the patient rather than the disease becomes the target of the therapy. However, this still requires an evidence-based medical approach. Herein, we will discuss how individual differences in terms of quality and quantity of the tissue microenvironment components affect progression of HCC. Then, we will highlight potential druggable pathways, also considering ongoing clinical trials. The development of biomarkers will be discussed in the light of new experimental research conducted with the aim of moving towards personalised therapy in patients with HCC.

View Full Text

Statistics from


Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide, with approximately 750 000 new cases diagnosed in 2008, and the second most frequent cause of cancer-related death.1 It is interesting to note that from 1990 to 2005, in the USA, the overall cancer death rate decreased by more than 20%, whereas the rate for HCC and bile ducts neoplasms increased by approximately 27%.2 These epidemiological data point out the relevance of HCC as a major healthcare problem worldwide, whose incidence is expected to increase progressively in the upcoming years in Western countries and North America. This expected rise is related to HCV infection and lifestyles, including dietary errors and intravenous drug addiction.3 ,4 By contrast, the vaccination programme for HBV carried out in Taiwan in all newborns dramatically reduced HCC onset in children and adults, further confirming the importance of primary prevention.5

Although all the international and national scientific associations, including the American Association for the Study of Liver Diseases and the European Association for the Study of the Liver, strongly recommend surveillance programmes for patients at higher risk of developing HCC, most cases are still diagnosed at too late a stage of the disease, when treatment options are usually limited. Nevertheless, a few years ago the natural history for patients with an advanced stage of disease changed after the approval of sorafenib, a multikinase inhibitor, as drug-based therapy.6 These researchers pioneered the field of systemic treatments for patients ineligible to receive so-called ‘curative therapies’, but this therapeutic approach is still in its infancy.

HCC is a unique prototype of cancer in the sense that in Western countries and North America, it develops in chronically damaged tissue due to liver cirrhosis, whereas the other most common malignancies develop on otherwise healthy tissue, as in the cases of breast, lung and colon cancer. This makes HCC an extraordinarily heterogeneous disease because, in addition to the common intrinsic cancer heterogeneity resulting from genome instability, a number of epigenetic alterations occur, induced by the host microenvironment. Since several aetiologies can trigger the onset of liver cirrhosis, including HBV, HCV, alcohol intake, hemochromatosis and nonalcoholic steatohepatitis, molecular differences can underlie the different microenvironment composition, affecting the growth and spread of HCC cells.7 Indeed, the current classification of liver cirrhosis is only suitable for strictly clinical purposes because it is too wide and does not take into account biomolecular aspects. This poor classification of liver cirrhosis makes it impossible to assess which alterations lead to which disease form, further complicating the issue of HCC heterogeneity and hampering the development of new therapies that could be successfully adopted in all patients.

In this review, we will refrain from focusing upon the involvement of each cell type in tumour progression since many other papers, including some very recent ones, have already dealt with this issue. The main goal of the present work is to highlight the complexity of the tumour/stroma interactions and, based on examples in the literature, to outline microenvironment modifications that may drive therapeutic choices, thereby moving towards a more personalised medical approach.

Remodelling of the tissue microenvironment

Recent close research into the molecular mechanisms responsible for cancer development, progression or chemoresistance has led scientists to focus on extremely detailed genetic or biomolecular mutations or alterations, thereby running the risk of losing sight of the wider scenario. Although by definition the onset of a neoplasm is the consequence of monoclonal expansion of a single cancer cell, this does not mean that other normal neighbouring cells are entirely extraneous to the disease progression process. This understanding has induced scientists to extend their interest to several different cell types and a number of molecules that together constitute the microenvironment. This further complicates the issue of the molecular mechanisms underlying carcinogenesis or tumour progression, so that at present the important question of what to target in the microenvironment appears purely rhetorical.

For instance, in the 1990s, proteolytic enzymes such as matrix metalloproteases (MMPs) were considered crucial in cancer progression because of their ability to break down tissue boundaries and so to allow cancer cells to spread to the surrounding tissue.8 MMPs, in particular MMP-2 and MMP-9, are activated from the latent form by the membrane type 1 metalloprotease (MT1-MMP)/tissue inhibitor of matrix metalloprotease-2 complex, a physiological inhibitor that serves to prevent too extensive and uncontrolled a degradation of the extracellular matrix (ECM) proteins and to maintain a balance.9–11 Several companies have synthesised MMPs inhibitors, and the most famous, including marimastat, have been tested in clinical trials but failed to show any therapeutic efficacy, while some were actually toxic, such as batimastat, for which British Biotech had to stop the experimentation.12 A more updated view considers these enzymes relevant not just as molecular busters of ECM proteins, but as finely regulated shears that cleave ECM components or growth factors stored in the microenvironment and change their biological activities.13 ,14 This is the case of laminin (Ln)-332, an isoform belonging to the Ln family, originally named Ln-5, formed by an α3, β3, γ2 chain linked with disulfide bonds.15 In the liver, once activated, hepatic stellate cells are the main source of ECM production. In HCC, but not in the surrounding normal liver, they produce and secrete Ln-332 that stimulates the migration and invasion of HCC cells in preclinical experimental models. Ln-332 ensures a strong adhesion to epithelial cells; in fact, it is also a constituent of the hemidesmosomes that anchor the basal layer cells of the epidermis to the basement membrane.16 ,17 However, MMP-2 or MT1-MMP cleave the γ2 chain, unveiling a cryptic site that then becomes accessible to cancer cells, including HCC cells, so that they acquire a migratory and invasive phenotype (figure 1).18–20 It should be noted that MMP-2 and MMP-9, as also other common proteases, are commonly expressed along the advancing tumour edge, where Ln-332 is also localised and where cancer cells need to negotiate their penetration into the surrounding tissues.21 In this context, the cleavage of Ln-332 could facilitate the spread of HCC cells.

Figure 1

(A) Diagram of Ln-332. Ln-332 consists of an α3β3γ2 chain, three β chains and two γ chains coiled together, ending up in the LG domain that consists of five LG molecules. LG 1–3 interact with integrin binding, LG 3 and 4 are involved in proteolytic processing and LG 4 and 5 in heparin or proteoglycan binding. (B) Diagram of Ln-332-mediated cell migration in carcinoma. This figure depicts the normal arrangement of the cell basal membrane and then cancerous events, which increase Ln-332 expression and early invasion processes, where cells lose their contacts due to the loss of E-cadherin. During cancer progression, metalloproteinases such as MMP-2 and MT1-MMP are secreted. MMP-2 and MT1-MMP cleave the γ chain of Ln-332 and hence facilitate cancer cell migration. Ln, laminin; MMP, matrix metalloprotease; MT1-MMP, membrane type 1 metalloprotease.

This same molecule, Ln-332, also promotes proliferation, likely because the G2 domain contains different epidermal growth factor (EGF)-like sequences that can mimic the effect of EGF, a potent mitogen factor.22 In HCC experimental clinical models, Ln-332 stimulates cell proliferation by activating Erk, as observed in the case of EGF.23 Also, expression of the Ln-332 γ2 chain translocates β-catenin into the nuclei, inducing a more aggressive phenotype of colorectal cancer cells.24 Moreover, Ln-332 has a part in triggering the epithelial mesenchymal transition (EMT) of HCC cells in combination with transforming growth factor (TGF)-β1, upregulating the Snail and Slug zinc finger protein. As a downstream consequence, E-cadherin is downregulated, the cell morphology changes and β-catenin is translocated into the nuclei, while HCC cells become more migratory and invasive.25 All these data explain why the presence of Ln-332, and in particular of the γ2 chain, is correlated with a more aggressive behaviour of tumours, including HCC, and worse survival.26 Whether Ln-332 may affect cancer stem cell biology is still unknown, but a strong association between Ln-332 and karatin-19, a marker of cholangiocytes, was reported to be correlated in a prospective cohort of 242 consecutive patients, with a worse prognosis in HCC, and also seems to identify a different subset of patients.27 This is consistent with another study reporting a strong upregulation of the Ln-332 γ2 chain in cholangiocarcinoma.28

As regards Ln-332 receptors, their activities have been reported to be dependent upon integrin α3β1, but in other malignancies the other main receptor, α6β4, has been reported to drive similar effects.29 These data are not conflicting because both receptors can display different functions in the same tissue, as in the skin, for instance, or their function can be tissue-related.30 ,31

In conclusion, according to the proteolytic remodelling of Ln-332 and to its engagement with α3β1 or α6β4, different cell functions and behaviours will arise and affect the clinical outcome. However, the role of the tissue microenvironment cannot be reduced to the effect of just one single component, even if it is a very important one like Ln-332. The goal of this discussion is to propose the view that it is not just a matter of an increased or decreased production of a single component present in the tissue microenvironment that influences the biological behaviour of the tumour, but rather the structural rearrangement of some of these finely regulated components that makes a difference.

Therapies targeting microenvironment components

For several decades, systemic therapy for patients with HCC has been based on empirical approaches, including chemotherapy, interferon and tamoxifen. In 2006, Liu and collaborators reported that a multikinase inhibitor directed against several molecular components of the tissue microenvironment, such as Raf kinase, vascular endothelial growth factor receptor (VEGFR)-2, platelet derived growth factor receptor, FMS-like tyrosine kinase-3, the Ras/MEK/ERK pathway, Ret and c-kit, displayed an antitumor activity in HCC preclinical models.32 These results offered the scientific rationale for conducting a single-arm clinical phase II study in 137 patients receiving sorafenib, which demonstrated an overall survival (OS) of 9.2 months.33 This encouraging result stimulated a further, larger, phase III study, a double-blind placebo-controlled trial enrolling 602 patients with advanced stage HCC according to the BCLC classification, and Child–Pugh A or B. Most of the patients enrolled in this study were Caucasian, and the main aetiologies, HCV, HBV and alcohol abuse, were similarly distributed. In this study, sorafenib showed an OS of 10.7 versus 7.9 months and a 1-year survival rate of 44% as opposed to 33% in the best supportive care group; this difference was statistically significant.6 At the same time, a similar study designed to assess the efficacy and tolerability of sorafenib versus best supportive care was conducted in the Asia–Pacific trial. Again, sorafenib significantly prolonged the overall patient survival as compared with placebo (6.5 vs 4.2 months), although the expected higher prevalence of HBV and the worse ECOG performance status likely affected the survival of both groups.34 Based on these studies, sorafenib is now the standard of care for patients with advanced stage HCC, being effective and well tolerated.6 ,34 However, the safety profile of sorafenib is an important issue. In the SHARP and Asia–Pacific trials, the percentages of adverse events were similar, being 80%–82% grade 1, 20%–22% grade 3 and 0% grade 4. The rate of treatment discontinuation was 38% in the SHARP and 19.5% in the Asia-Pacific trial, but in both studies the percentage was similar for the drug and placebo groups. However, this percentage was higher in a multicentre study including 296 patients with similar characteristics to those reported in the SHARP and Asia-Pacific trials, in which 91% of patients experienced adverse events and 45% of the patients with grade 3/4 toxicity had to reduce or interrupt treatment.35

The incidence of adverse events and disease progression occurring during treatments is still a major drawback. In fact, after promising initial results in a clinical phase II study, sunitinib, a potent inhibitor of VEGFR, was investigated in various clinical phase III trials. In these studies, sunitinib failed to improve OS as second-line treatment in patients who were intolerant or progressed while in treatment with sorafenib. Moreover, in April 2010, Pfizer announced discontinuation of a study because of side effects and failure to achieve the primary end point.

Brivanib, an inhibitor of both VEGFR and fibroblast growth factor receptor, was reported in a single-arm clinical phase II trial to offer clinical benefit in those patients who failed to respond while in treatment with sorafenib, and although adverse events were quite frequent only 26.1% had to drop out because of adverse events.36 These results pushed the drug into clinical phase III, where it failed to achieve the primary end points in patients who were intolerant or had disease progression while in treatment with sorafenib.37 ,38 Also disappointing was the Novartis announcement in August 2013 that everolimus did not meet the primary end point, failing to extend OS in patients who were intolerant or had disease progression with sorafenib, despite the fact that in phase II the treatment had been well tolerated although displaying only a modest antitumor activity.

Recently, it has been reported that an oral selective mesenchymal epithelial transition (MET) factor receptor inhibitor showed a favourable safety profile in phase I studies. Based on this result, a larger phase II multicentric, randomised, double-blind, placebo-controlled study was conducted as a second-line treatment in Child–Pugh A patients with advanced HCC who were intolerant or resistant to sorafenib. In this study, patients receiving tivantinib had longer TTP weeks compared with placebo. Stratifying patients by c-MET expression, the difference between the two groups became much more evident, being 11.7 versus 6.1 weeks, with an OS of 7.2 versus 3.8 months in patients receiving drug or placebo, respectively. Serious adverse events were mainly neutropenia, in 16% of the patients receiving higher doses and in 6% of those receiving lower doses.39 Based on these results, a larger phase III study is currently underway.

A phase III clinical study has recently been concluded, assessing the therapeutic efficacy of ramucirumab, a humanised monoclonal antibody directed against VEGFR-2, and the results are expected in the next few months. This study is based on promising results from a phase II study of 42 patients in which this molecule showed a disease control rate of approximately 50% and a median progression-free survival of 4.3 months, with grade >3 adverse events in 26% and one treatment-related death due to bleeding (table 1).40

Table 1

Molecular classification of HCC

At this point some considerations need to be made to balance the pros and cons of systemic therapies. In particular, arguments against systemic therapies are the relative short improvement of OS and the frequency of adverse events. The main elements supporting the use of systemic therapies are an already advanced disease stage and the dose-dependency of adverse events that are, therefore, manageable in most cases.

Microenvironment pathways driving HCC progression

The most disappointing aspect is that although the drugs used so far are directed against different targets of the microenvironment, none has achieved a profound change in the natural history of the disease, even though some therapeutic effectiveness has been reported with different drugs. This observation supports the contention that the heterogeneity of HCC cannot be simply addressed by using selective drugs mainly directed against only one microenvironment component, such as angiogenesis. A possible strategy could be to combine sorafenib with other agents, but it is not yet clear which subsets of patients are more likely to benefit from certain drug associations. An interesting example is the combination of erlotonib, an epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor very similar to gefitinib, and sorafenib; theoretically this should be beneficial. This proof of concept was assessed in a phase III study, enrolling 720 patients with advanced HCC, randomised to a double arm of sorafenib and placebo versus sorafenib plus erlotonib. The results of this trial, SEARCH, presented in 2012 at the European Society for Medical Oncology conference, showed no improvements in terms of OS and TTP of the combination of the two drugs versus a single drug plus placebo. It is surprising to note that failure of this trial was reported in the same year as Sieghart et al published an interesting work demonstrating that the combination of sorafenib and erlotonib did not display a synergistic effect on HCC progression as compared with a single drug treatment in a Morris hepatoma model.41 This is a good example of how preclinical research can assist in the development of new drugs, accelerating knowledge in the field and saving money.

In the case of sorafenib, the preexisting data on using the drug in renal cell carcinoma preclinical research were used to demonstrate a therapeutic efficacy in HCC. In that study, pErk was suggested for use as a tissue biomarker to select those patients most responsive to therapy and in developing future clinical trials.33 Nevertheless, this encouraging result in the phase II clinical trial was not confirmed in the SHARP trial.6 Nor was any positive result gained in the investigation conducted in the same patients to determine reliable biomarkers for assessing therapeutic effectiveness, although some of them, such as c-kit, vascular endothelial growth factor (VEGF) or VEGFRs, are directly targeted by sorafenib; the most reliable biomarker remained AFP.42

The hepatocyte growth factor (HGF)/c-Met pathway has long been recognised to be involved in liver regeneration following different injuries. Binding of HGF to the c-Met receptor induces phosphorylation of the receptor downstream pathway involving phosphatidylinositide 3-kinases (PI3K) and mitogen-activated protein kinase (MAPK), affecting the regulation of cell proliferation, apoptosis, survival and differentiation, and leading to a worse prognosis. This is the scientific rationale for targeting c-Met with tivantinib, and for stratification of patients according to c-MET expression.

PI3K is recognised to play a relevant role in cancer; this wide family of enzymes includes three classes based on substrate specificity and sequence homology. Class I has been most closely investigated and further divided into two subclasses: PI3K-IA and PI3K-IB. PI3K-IA includes the p110α, p110β and p110δ catalytic isoforms, while subclass IB includes only the p110γ isoform, but the specific contribution of each class to tumour progression is still unknown.43 So far, p110α is the isoform most commonly related to cancer, mainly showing an altered or mutated expression.44 ,45 In particular, the highest percentages of cases with p110α gene (PIK3CA) mutations have been reported in liver and breast cancers, together with downregulation of onco-suppressor proteins, such as PTEN or PIK3IP1 (a negative regulator of PI3K).46 However, it has recently been reported that the p110γ isoform is overexpressed in HCC issues, where it is directly correlated with Ki-67 and inversely with p21.47 The downstream mediator of the PI3K pathway is mammalian target of rapamycin, hence the scientific rationale for targeting this isoform in HCC, but the trial results failed to support the hypothesis.

There is no doubt that all the pathways discussed above are fundamental for cancer progression, justifying the development of targeted therapy, but it is unlikely that tumour progression can depend on just a single pathway (figure 2). An imbalance of the cytokine network can also affect responsiveness to therapy, but the mechanisms are still unclear. For instance, the proinflammatory cytokines tumour necrosis factor (TNF)-α or interluekin-1 activates nuclear factor (NF)-κB, being related to tumour progression and therapy resistance.48 TNF-related apoptosis-inducing ligand (TRAIL) has a potent apoptotic effect, but many HCC cells are resistant to TRAIL. Sorafenib has been shown to sensitise such resistant cells to TRAIL-induced apoptosis inhibiting signal transducers and activators of transcription (STAT3) activation, while sorafenib and TRAIL display a synergistic activity.49–51 Biological redundancy is further enhanced as a consequence of the pharmacological inhibition of one or more molecular pathways, that this could remove the selective pressure regulating physiological tissue homeostasis. This observation is supported by marginal results of clinical trials and could explain the heterogeneity of cancer and in particular of HCC, and also the partial results achieved so far. In fact, all the different drugs attempt to interfere with the progression of HCC, without taking into account the underlying cirrhosis that induces a constant hepatocarcinogenic stimulus.

Figure 2

Targeted therapy in HCC. Mechanisms of resistance to sorafenib in HCC often rely on pathways that are activated after long-term sorafenib therapy (eg, mTOR/AKT-dependent cytoprotective autophagy or HIF pathways). In some patients, subsets overexpression (↑) of sorafenib-sensitive molecular targets (green labelled) might be predictive of a better response to sorafenib. Other pathways can become dysregulated in HCC and thus are potential novel candidates for targeted therapy (green labelled). The drug-induced upregulation of mediators of sorafenib resistance (eg, AKT, HIF-1α, TGFα/ EGFR) suggests a strategy for overcoming this problem by blocking such mediators using specific inhibitors (GDC0068, EF24, gefitinib). Patients showing downregulation (↓) of these drug pathways (red labelled) are prone to sorafenib resistance. In addition, sorafenib can inhibit tumour-suppressive pathways, so raising general issues concerning drug-based therapies in HCC. HCC, hepatocellular carcinoma; mTOR, mammalian target of rapamycin; HIF, hypoxia inducible factor; TGF, transforming growth factor; EGFR, epidermal growth factor receptor; HDACs, histone deacetylases; BCRP, breast cancer resistance protein; EMT, epithelial mesenchymal transition; VEGFR, vascular endothelial growth factor receptor; VEGF, vascular endothelial growth factor.

Inhibiting TGF-β pathways as a paradigm of therapy directed against the tumour microenvironment

In the liver, TGF-β1 seems to be an interesting target, being involved in the liver fibrogenesis leading to HCC development and, therefore, orchestrating the whole microenvironment organisation and the consequent tumour/stroma cross-talk. However, TGF-β1 has been widely described as a tumour suppressor owing to its capacity to inhibit cell proliferation52 and hence suppress tumoral development.53 On the other hand, other studies have proposed TGF-β1 as a tumour promoter because of its effect on suppressing the immune response or else on stimulating cancer dissemination.54 ,55 TGF-β is a potent stimulator of the EMT, leading to a modification of cell polarity, of E-cadherin and β-catenin distribution, and so to acquisition of a motile and invasive phenotype.56 ,57 What mechanism switches on/off the activity of TGF-β as tumour suppressor versus tumour promoter is unclear and is unlikely to be elucidated in the near future since it is difficult to follow the dynamic process in humans and experimental models. Nevertheless, this dual function of TGF-β has been thought to relate to the hepatocarcinogenesis stage, whereby in the early phase TGF-β acts as tumour suppressor and in the late phase as tumour promoter. This has been demonstrated in an experimental mouse model and also in a large cohort of patients. The effect in the late phase is likely related to the strong capability of TGF-β to stimulate the EMT.58 Many other later studies reinforced this idea so that the whole process seems to be a bottleneck, being responsible for converting the HCC phenotype and perhaps for the unexplained cancer dormancy. However, the main interest in targeting TGF-β is because of its involvement in the multiple activities affecting HCC progression, including the downregulation of E-cadherin allowing cells to spread in the surrounding tissue.59 ,60 Also in liver transplant patients, downregulation of E-cadherin is highly predictive of HCC recurrence, in up to 88% of cases.61 ,62 Furthermore, TGF-β also has a role in the intravasation of HCC cells, phosphorylating the intracellular tail of integrin β1 that engages with the main ECM present on the blood vessels, such as fibrinogen and fibronectin.63–65 Macrovascular invasion is recognised as one of the most important prognostic factors for the possible development of portal thrombosis and for the clinical outcome of patients after liver transplantation.66–69 TGF-β also stimulates the production of VEGF, accelerating tumour growth and is mainly responsible for neovascularisation.70 ,71 Increased circulating levels of VEGF are correlated with disease progression and tumour recurrence after surgery.69 ,72 ,73 This provides the rationale for targeting VEGF with antiangiogenic drugs.

The stromal microenvironment has recently been considered a fundamental component for cancer metastasis, although the desmoplastic reaction in HCC is not quantitatively diffuse, as in other malignancies like pancreatic cancer. Nevertheless, it has a crucial role in view of the underlying liver disease where HCC commonly develops. Consistently, in a distinct data set from a different patients' cohort, the upregulation of α-smooth muscle actin supported the role of fibrosis in HCC progression, and likely in developing drug resistance.74 In such a context, connective tissue growth factor (CTGF) is responsible for the fibrotic peritumoral reaction and finally inhibiting the T-cell response, leading to an immune tolerant response to HCC development.75 TGF-β directly regulates CTGF expression via Smad phosphorylation, while in HSCs the MPAK and Stat3 pathways are also involved.76 In addition, the effect on CTGF could be responsible for aggravating the underlying liver fibrosis, being a risk factor for the development of new tumours, but also for peritumoral fibrotic tissue, the biological scaffold modulating the different activities of HCC cells.77 ,78 Thus, the biochemical composition of the stroma, enriched by one particular ECM or another, could affect different cell functions depending on the different integrin receptors expressed on the cell surface of HCC cells.79 TGF-β inhibition by means of LY2157299, an inhibitor of TGFβRI kinase, shown in a number of preclinical experimental models, was the scientific rationale for a clinical trial that is still underway in patients with advanced stage HCC who failed or are intolerant to sorafenib. Preliminary results indicate some therapeutic activity, like other drugs reported above, in a subset of patients. However, unlike in other studies, LY2157299 was evaluated in patients with elevated AFP, and about 20% of the patients showed a reduction of AFP levels by more than 20%, which was correlated with a survival of more than 17 months. Also, in the same patient population, plasma TGF-β1 levels were elevated and in half of the patients the reduction of plasma TGF-β1 by more than 20% was associated with median OS of about 12 months, while patients who had no such reduction had a median OS of 4 months. Based on these findings, the community awaits the start of a phase III study. The main challenge for the future in cancer biology and, in particular, in HCC is to set up a platform that can individuate the level of sensitivity of which patient to which drug. In this case, TGF-β itself, or related or affected molecules, could be investigated as a potential biomarker.80

HCC classification according to microenvironment molecular pathways

So far, we are used to treating cancer according to the histological classification, but in recent years growing evidence based on molecular studies has suggested new molecular classifications based on genetic characteristics of the tissue microenvironment. The general idea is that heterogeneity may affect few pathways rather than being responsible for a variety of pathways involved in tumour progression. This is the case of KRAS and EGFR, for instance, whose mutations drive the therapeutic choice in colorectal and non-small cell lung cancer. In HCC, we are still far from this goal, both because sorafenib is the only drug recommended for systemic treatment while no second-line treatments have been approved although many clinical studies are underway and because no clear association between a specific mutation or biomarker expression and a therapeutic agent has been reported in a phase III clinical study. Also in this case, there are many interesting phase II/III studies with c-MET and TGF-β inhibitors, tivantinib and LY2157299 respectively, but conclusive data are not yet available. It must also be borne in mind that the limited availability of tissue specimens hampers rapid development in the field. Although a classification based on molecular/genetic characteristics orienting the therapeutic choice is lacking for HCC, pathogenic stratifications of patients according to transcriptomic data, with prognosis as the primary end point, are becoming available.81 ,82 This approach is a step towards analysing the most relevant pathways with a significant effect on clinical outcome and facilitating the design of new drugs, as well as offering the rationale for combining several drugs.

In this regard, a classification of HCC according to few molecular pathways affecting the clinical outcome, and therefore the prognosis, has been proposed. According to this study, the patient's subset with the worst prognosis was the group with involvement of the WNT or MYC/AKT pathway, also correlated without/with increased AFP levels. Interestingly, WNT was not regulated by the canonical β-catenin pathway but rather by TGF-β, thus stimulating cell proliferation and providing a new mechanism for WNT regulation in HCC.83 In this study, a pathogenic pathway is linked to the biological mechanism and hence to the clinical outcome.

Experimental tools guiding therapeutic choice based on tissue microenvironment heterogeneity

There is no doubt that advances in preclinical research, together with the use of biomarkers, will represent important achievements in the roadmap towards a more personalised medicine. This will promote faster advances in the field and also save financial resources by preventing unnecessary clinical trials. However, some technical obstacles strictly related to disease management limit rapid development in the field. For instance, all genomic analyses have been performed on surgical tissues, posing an obvious bias in patients' selection, which does not take account of other stages of disease. On the other hand, the small amount of tissue obtained with needle biopsy is an insoluble limitation. A further problem is the availability of biological samples collected during follow-up of the disease. In some haematological malignancies, where repeated tissue biopsies are mandatory criteria for assessing therapeutic response, significant improvements of survival have been achieved, whereas in HCC this is obviously impossible.

To overcome these limits, experimental models dealing with human samples could be very helpful. The detection of circulating tumour cells (CTC), commonly known as liquid biopsy, is becoming a hot topic in the cancer biology field. It allows genomic analysis to be performed on cancer cells while avoiding liver biopsy. The main limitation remains the low number of cells detectable in the blood, although different techniques are currently under investigation to enrich such preparations. Another bias is the ability to capture CTC according to the expression of specific antigens, mostly EpCam, on the cell surface. The advantages are evident, for following the natural history of the disease and modifications that can occur during systemic treatment (Chiappini review).

Another powerful tool is the patient-derived xenograft model (PDX). This model is based on subcutaneous or orthotopic implant of tissues obtained from patients without any passage on plastic culture dishes, mimicking the tumoral microenvironment also in terms of the histological architecture.84 ,85 The protocol consists of implanting a small tissue specimen of about 3–4 mm3 freshly derived from surgery in non-obese diabetic/severe combined immunodeficiency mice that appear to be the most suitable for hosting tumour growth. Once the tumour has grown, it is transplanted into a second and then a third generation of animals. At this point, there are many animals carrying the human-derived tumour available for use either in experimental trials or for further expansion. It is not yet known through how many passages these animals will maintain the original genetic and histological characteristics, but of course earlier passages are recommended. Also in this model, a number of issues could be raised, such as the elevated costs, the long time required to establish the full model and, of course, the degree of similarity with the original tumour in humans, which cannot be measured.86 ,87 On the other hand, the second-generation and third-generation PDX ensure a rearrangement of tumour/host interactions that overcomes the initial alteration so as to ensure tissue heterogeneity (figure 3). Orthotopic engraftment, although more complicated, is considered the most reliable model for mimicking the histological and genomic properties of the original tumour.88 However, we must remember that each model is, by definition, a surrogate of the reality, even if each represents a further step towards xenograft injection of cell culture models.

Figure 3

Diagram of experimental models for use as an important tool in personalised therapy. (A) This figure depicts the patient-derived xenograft mouse model system. Tumour tissue specimens collected from (hepatocellular carcinoma) surgery could be sectioned into 2–3 mm fragments and implanted subcutaneously into immunodeficient mice. After engraftment, the tumour is allowed to grow and expand and then surgically collected for subsequent extension to other experimental cohorts. Patient-derived xenografted nude mice must be treated with drug X and drug Y, and controls with placebo. Timely measurement of tumour expansion volume is necessary in both placebo and drug-treated mice. A typical tumour regression response to the drugs can be expected. Therefore, biological assays could be performed, such as molecular analyses for mutational status, gene expression and ploidy analysis, along with drug efficacy assays. Based on these integrative genome analyses and biochemical drug efficacy assays, novel biomarkers for targeted therapies will likely emerge. From a broader standpoint, these novel biomarkers could be used in early clinical phase trials for screening and patient selection strategies, and eventually personalised treatments. (B) This figure illustrates the new method for the early detection of cancer and recurrence. Liquid biopsy is a non-invasive method of screening for signs of cancer in the circulation. Blood from cancer patients could be drawn to retrieve and grow circulating tumour cells or to collect plasma DNA for mutational analysis and molecular analysis. This technique may help promote early detection of cancer and recurrence, as well as guide a better selection of personalised therapy.

In the future, further preclinical research is highly necessary to avoid unnecessary clinical trials, save financial resources and accelerate the development of drugs that could be administered to patients with specific molecular characteristics, offering the best treatment to each patient.


The authors are grateful to Mary V. Pragnell for language revision. This work was supported by the Italian Association Cancer Research (AIRC grant number 11389) awarded to GG. BR and YC are supported by the EU-Marie Curie Initial Training Network (ITN), FP7-PEOPLE-2012-ITN 2012, grant agreement number 316549 to GG.


View Abstract


  • Contributors All the authors have contributed to the conception or design of the work, or the acquisition, analysis or interpretation of data. All authors have approved the final submitted version of the article.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.