Sorafenib is currently the medical treatment of reference for hepatocellular carcinoma (HCC), but it is not known whether sorafenib is equally active in all HCC. Here, our aim was to explore intrinsic differences in the response of HCC cells to sorafenib, to identify potential mechanisms leading to primary resistance to this treatment. We analyzed a panel of six human HCC cell lines and compared the activity of the main oncogenic kinase cascades, their clonogenic potential, proliferation and apoptosis upon exposure to sorafenib. We report that HCC cells present important differences in their response to sorafenib, and that some cell lines are more resistant to the actions of sorafenib than others. We identify the activated epidermal growth factor receptor (EGFR) as a parameter that promotes the resistance of HCC cells to sorafenib. In resistant cells, the efficacy of sorafenib was increased when EGFR was inhibited, as was demonstrated using two chemical inhibitors (erlotinib or gefitinib), a monoclonal antibody directed against EGFR (cetuximab), and RNA interference directed against EGFR. A combination of EGFR inhibitors and sorafenib affords a better control over HCC proliferation, most likely through an improved blockade of the RAF kinases. Our findings therefore confirm the importance of RAF kinases as therapeutic targets in HCC, and identify EGFR as a determinant of the sensitivity of HCC cells to sorafenib. Our findings bear possible implications for the improvement of the efficacy of sorafenib in HCC, and might be useful for the identification of predictive biomarkers in this context.
Hepatocellular carcinoma (HCC) is the most frequent form of primary liver tumor.1, 2 Sorafenib, an inhibitor of oncogenic kinases, is to date the only treatment proven to extend the survival of patients with advanced stages of this tumor.3 However, its clinical benefits remain modest and most often consist of temporary tumor stabilization.3, 4 In the absence of validated predictive biomarkers, identification of the individuals that are most likely to benefit from this treatment is not possible.5 It is currently not known whether HCCs with distinct genotypes differ in their response to sorafenib, and could eventually present primary resistance to the antioncogenic action of sorafenib.
Sorafenib was originally identified as an inhibitor of the RAF kinases, a family of three members (A-RAF, B-RAF and C-RAF/Raf-1) that play a key role in the transduction of mitogenic signals.6 These kinases are often activated in HCC, and it is usually accepted that the oncogenic transduction cascade RAS-RAF-MEK-ERK plays a role in liver transformation.7, 8 However, whether the RAF kinases are direct therapeutic targets of sorafenib in HCC remains partially unclear. In HCC patients treated with sorafenib, high activation levels of the RAF kinases, measured through the phosphorylation of their downstream effector ERK, are associated with a longer time to progression.4 Although this observation suggests that the RAS-RAF-MEK-ERK cascade could be an important therapeutic target of sorafenib, the overall weak potency of sorafenib as a blocker of preactivated RAF kinases9–12 and its lack of specificity toward these kinases have recently led to the questioning of the role of other kinases as targets.13 In particular, several receptor tyrosine kinases (RTKs), such as the vascular endothelial growth factor receptor (VEGFR), are also blocked by sorafenib in vitro and could be therapeutic targets in this setting.13–15 The role of RAF kinases as targets of sorafenib in HCC therefore requires further investigation.
To examine these questions, we decided to compare a panel of human hepatoma cell lines. We found large differences in the sensitivity of HCC cells toward sorafenib. We explore the mechanisms that account for these differences and discuss their potential clinical implications.
The human hepatoma cell lines Huh7, Hep3B, HepG2 are model cell lines used to analyze the response of HCC cells to therapeutic agents.16 The cell lines SNU-182, SNU-398 and SNU-449 were described previously.17 An extensive characterization of the genotype of all cell lines used here has been reported elsewhere (Ref.18 and http://www.sanger.ac.uk/genetics/CGP/cosmic/). Cell provenance, authentication and culture protocols are described in the Supporting Information section.
Drugs and reagents
Sorafenib, gefitinib and erlotinib were purchased from Selleck chemicals and were resuspended in DMSO. Cetuximab was purchased from Merck. Amphiregulin was obtained from Sigma and was used as indicated by the provider. In all experiments, conditions with solvent alone were applied at least once to ensure the specificity of the observations presented.
Clonogenicity assays were performed as previously described,19 with cells seeded at a concentration of 200 cells per dish with 30 mm diameter (500 for HepG2 cells). Cells were incubated for 2 weeks under the indicated conditions, stained with Giemsa and the clones were finally counted.
The Proteome Profiler Human Phospho-RTK Array Kit was purchased from R&D and used according to the provider's instructions.
RAF kinase activity assay
Kinase assays were performed using a commercial kit from Millipore. The kit was used as indicated by the provider, using recombinant MEK1 as substrate and detection by chemiluminescence for the measurement of RAF kinase activity.
Silencer select validated siRNA directed against EGFR (s564 and s565), C-RAF (s11749), B-RAF (507) and Silencer negative control (am4635) were purchased from Applied Biosystems. Transfections were performed using the siPORT-neoFX reagent (Applied Biosystems) and Optimem transfection medium (Invitrogen), according to the manufacturer's instructions.
All experiments presented were performed at least three times. Student's t test was used as indicated, and a value of p < 0.05 was considered as threshold for significance. Linear regressions analyses were performed and the values of R2 were calculated using the software R version 2.12.0 (http://www.r-project.org/).
HCC cell lines exhibit different sensitivities to sorafenib
To examine the effect of sorafenib on HCC cells, sorafenib was applied on six hepatoma cell lines with different genetic backgrounds (Huh7, Hep3B, HepG2, SNU-182, SNU-398 and SNU-449). Cells were exposed to increasing concentrations of sorafenib (0.5–10 μM) chosen to encompass the average plasma concentrations measured in patients receiving 400 mg of sorafenib bi-daily, i.e., ∼5 μM.4 Although the clonogenic potential of some cell lines was inhibited at concentrations of sorafenib in the micromolar range (HepG2, Huh7 and SNU-398), higher concentrations of sorafenib (above 5 μM) were required to achieve inhibition of clonogenic growth in other cell lines, such as Hep3B and SNU-449 (see Table 1 and Figure 1a for a comparison between Huh7 and HepG2 and the more resistant cell lines Hep3B and SNU-449). Strikingly, low concentrations of sorafenib even paradoxically stimulated the growth of individual clones of Hep3B (Fig. 1a). The existence of differences between the hepatoma cell lines was independently shown with an MTT assay that confirmed the heterogeneity of their response to sorafenib (Supporting Information Fig. 1).
Table 1. IC50 of sorafenib measured in the clonogenicity assay and on RAF kinase cascade activity (p-ERK) in HCC cells
To explore the basis for the different sensitivities of HCC cells toward sorafenib, we compared the effect of sorafenib on the two main oncogenic kinase cascades, RAF-MEK-ERK and PKB-mTOR, in these hepatoma cells (Fig. 1b). In all cell lines, we noticed a concentration-dependent activation of PKB phosphorylation by sorafenib, suggesting that the control of this kinase was unlikely to be an important target for sorafenib in HCC cells. On the contrary, great differences in the half maximal inhibitory concentrations (IC50) of sorafenib were observed with the RAF-MEK-ERK cascade among the different cell lines (Table 1). A close to 10-fold difference in IC50 was noted in ERK phosphorylation between the most sensitive cell line, HepG2 and the most resistant, SNU-449 (Table 1, p < 0.05 between cell lines). These differences in the sensitivity of the RAF-MEK-ERK kinase cascade to sorafenib overlapped with those previously noted on clonogenic growth (Table 1), and a correlation could be established between the IC50 measured with the control of clonogenic growth and the inhibition of the RAF-MEK-ERK cascade (Fig. 1c).
Interestingly, ERK phosphorylation was paradoxically activated at low micromolar concentrations in Hep3B cells (Fig. 1b). To further explore the regulation of RAF kinases in Hep3B cells upon sorafenib application, we measured the activity of B-RAF or C-RAF kinases with an in vitro assay using MEK1 as a substrate (Fig. 1d). When sorafenib was applied at a concentration of 5 μM, corresponding to the plasma concentration measured in the clinical context,4 we observed an increase in MEK1 kinase activity associated with C-RAF immunoprecipitate (Fig. 1d). In agreement with previous findings reporting the paradoxical activation of RAF kinases by their inhibitors in a context of heterodimerization,9–12 coimmunoprecipitation experiments revealed the appearance of heterodimers formed between B-RAF and C-RAF upon exposure of Hep3B to sorafenib (Fig. 1e). These findings led us to postulate that some intrinsic defect(s) in the RAF kinase pathway might partially explain the resistance of hepatoma cells to sorafenib.
High EGFR activation levels correlate with resistance to sorafenib in HCC cells
RTKs are important upstream activators of the RAF kinases, and their activation has been reported to promote resistance to RAF inhibitors.12 To comprehensively examine their implication in the resistance to sorafenib, we used an antibody array to analyze the activation status of all 42 RTKs in basal conditions in each hepatoma cell (Fig. 2a). Strikingly, the two resistant HCC cell lines Hep3B and SNU-449 were characterized by higher activity levels of EGFR compared to the other cell lines examined (Figs. 2a and 2b). Immunoblotting of complete cellular extracts prepared from all cell lines confirmed these observations, because the cell lines with the highest levels of EGFR activity were also found to display the highest expression levels of EGFR (Supporting Information Fig. 2). A clear correlation was observed between EGFR activity measured using RTK array and the values of IC50 presented in Table 1 (Fig. 2c). These observations suggest that higher basal activation levels of EGFR are associated with a reduced effect of sorafenib on clonogenic growth.
To examine the role of the EGFR in the response of HCC cells to sorafenib, we used gefitinib and erlotinib, two well-characterized inhibitors of EGFR active at concentrations as low as 500 nM20 (Supporting Information Fig. 3). Gefitinib and erlotinib were used in the clonogenicity assay to examine the possibility that EGFR activation might modulate efficacy of sorafenib. Although neither gefitinib nor erlotinib alone significantly inhibited the clonogenic potential of Hep3B or SNU-449 cells at a concentration ≤1 μM (data not shown), we found that both inhibitors greatly increased the efficacy of sorafenib in this assay (Figs. 2d and 2e). In Hep3B, this increased efficacy of sorafenib was reflected by a shift in the IC50 of sorafenib under control conditions (IC50 = 7.5 ± 0.5 μM) compared to sorafenib combined with either of the two blockers of EGFR (IC50 = 2.0 ± 0.5 μM with erlotinib, and IC50 = 2.5 ± 0.5 with gefitinib, p < 0.01 compared to control for both inhibitors) (Fig. 2e). Interestingly, gefitinib and erlotinib did not only reduce the IC50 of sorafenib, but they also abolished the paradoxical increase in the clonogenic potential of Hep3B that we had previously noticed at low, micromolar concentrations of sorafenib (Fig. 2d). This effect of gefitinib and erlotinib was specific: we found no sensitization to sorafenib in the EGFR-negative Huh7 or HepG2 cells (Fig. 2e).
To support the specificity of our findings, we used a different approach relying on an extracellular inhibitor of EGFR and a ligand of the EGFR. We used cetuximab, a monoclonal antibody that specifically interacts with the extracellular domain of the EGFR, thereby blocking its ligand binding and activation.21 Cetuximab, applied at a concentration of 100 nM in the clonogenicity assay, sensitized Hep3B, but not Huh7 cells, to sorafenib (Fig. 3a), thereby confirming our previous results obtained with chemical inhibitors. Conversely, the extracellular application of amphiregulin, a specific and physiological ligand of EGFR in HCC cells22 conferred a protection against sorafenib in the clonogenicity assay (Fig. 3b). These observations further confirmed the role of EGFR activity as a determinant of the antitumor activity of sorafenib in HCC cells. Finally, we applied a genetic approach to examine the role of the EGFR in HCC resistance to sorafenib. The first approach relied on the use of RNA interference directed against EGFR: Hep3B cells were transfected with two independent siRNA targeting EGFR or a control, nonrelevant siRNA, and used in our clonogenicity assay (Fig. 3c). We found that EGFR depletion increased the inhibitory effect of sorafenib in these conditions. Although sorafenib (5 μM) applied in the control siRNA condition reduced only 34.8% of the total number of clones, with the use of two EGFR siRNAs, we observed a striking inhibition amounting to 59.8 and 59.7% (Fig. 3c, p < 0.05 between control and each siRNA targeting EGFR). Reciprocally, we examined the effect of the intracellular expression of a constitutively active mutant of EGFR, the variant EGFRvIII, in the Huh7 cell lines, previously found by us to be more sensitive to sorafenib and to express low levels of the EGFR. This mutant consists of a partial deletion of the extracellular domain of EGFR and displays constitutively active, ligand-independent, kinase activity.23 In Huh7 cells, the intracellular expression of the EGFRvIII protein promoted a resistance of the RAF-MEK-ERK kinase cascade to sorafenib (Supporting Information Fig. 4). Based on our combined findings obtained with chemical inhibitors, neutralizing antibody, EGFR ligand, transgene expression and RNA interference, we concluded that EGFR is an in vitro determinant of the response of HCC cells to sorafenib.
Synergistic effects of sorafenib and EGFR inhibitors on the RAF-MEK-ERK cascade in EGFR-positive cells
Based on these findings, we decided to explore the impact of EGFR activity on the control of oncogenic kinases of the RAF family in HCC cells treated with sorafenib. Although sorafenib (5 μM) or either EGFR inhibitor alone had a modest impact on ERK phosphorylation, we noticed a potent synergy between these drugs in the control of the RAF kinase cascade, as determined by immunodetection of p-ERK in Hep3B cells (Fig. 4a). To explain this synergy, we turned our attention to the regulation of C-RAF, because of our previous observation that sorafenib can paradoxically activate this kinase in Hep3B cells (Figs. 1d and 1e). We analyzed the phosphorylation status of the main residues controlling this kinase. Phosphorylation of Ser259 and Ser621 regulate the binding of C-RAF to proteins of the 14-3-3 family, whereas Ser338 is located in the so-called N-region and has an important role as a direct positive regulator of C-RAF kinase activity.6 Interestingly, the most noticeable variation was observed at the level of Ser338, whose phosphorylation levels were increased on average threefold upon exposure to sorafenib (Fig. 4a). Gefitinib and erlotinib, as well as cetuximab prevented the phosphorylation of C-RAF on Ser338 induced by sorafenib (Figs. 4a and 4b). In line with these observations, EGFR inhibitors were found to prevent the paradoxical increase in the kinase activity of C-RAF induced by sorafenib on Hep3B cells (Fig. 4c). We concluded that inhibitors of the EGFR and sorafenib afford a synergistic control over the activity of the RAF-MEK-ERK kinase cascade in EGFR-positive HCC cells.
Sorafenib and EGFR inhibitors synergistically control the proliferation of EGFR-positive cells
Clonogenicity assays reflect simultaneously the ability of cancer cells to survive and to proliferate, so we directly examined these facets of sorafenib's activity in HCC cells. Sorafenib, applied alone at a concentration of 5 μM, only modestly reduced the expression of the Proliferating Cell Nuclear Antigen (PCNA), a marker of cell proliferation (Fig. 5a). However, gefitinib and erlotinib synergized with sorafenib to decrease the expression of PCNA, in the same experimental conditions under which ERK phosphorylation was prevented (Fig. 5a). Because our previous results suggested that EGFR activation modulates the ability of sorafenib to block RAF kinases, we decided to genetically examine the role of RAF kinases in Hep3B cells. A selective knock-down approach was used, using RNA interference directed against B-RAF, C-RAF, or B-RAF and C-RAF (Fig. 5b). Although the near complete inhibition of either B-RAF or C-RAF did not alter the expression of PCNA, a combined knock-down of B-RAF and C-RAF strongly decreased PCNA expression in Hep3B cells (Fig. 5b). Under these conditions, we only observed a small increase in apoptosis of Hep3B cells, as evaluated through the microscopic examination and counting of cells with chromatin condensation (Supporting Information Fig. 5).
To establish the synergistic antiproliferative effect of the combination of sorafenib and EGFR inhibitors, we performed cell counting experiments on the two most resistant HCC cell lines, i.e., Hep3B (Fig. 5c and Supporting Information Fig. 6) and SNU-449 (Fig. 5d). Clearly, both chemical inhibitors of EGFR (Figs. 5c and 5d) and the monoclonal antibody cetuximab (Supporting Information Fig. 6) synergistically prevented the proliferation of EGFR-positive cells exposed to low concentrations of sorafenib. In contrast to the spectacular synergy noted in terms of control of cell proliferation, the inhibition of the EGFR did not significantly increase apoptosis in EGFR-positive cells exposed to sorafenib (Supporting Information Fig. 7). To verify the specificity of these observations, the combination of EGFR inhibitors and sorafenib was examined in the EGFR-negative cell line Huh7 (Supporting Information Fig. 8). Using this cell line, no synergy was observed in the control of RAF-MEK-ERK kinase cascade, inhibition of cell proliferation or apoptosis induction (Supporting Information Fig. 8). We concluded that the synergistic inhibition of the proliferation afforded by sorafenib and EGFR inhibitors occurred specifically in EGFR-positive cells and was most likely the result of an improved inhibition of RAF kinases in these cells.
Although the recent introduction of sorafenib as a treatment for HCC has been a major clinical breakthrough, several questions regarding the mode of action of this drug as well as the development of resistance have been raised. Our study highlights the existence of important differences in the sensitivity of HCC cells to sorafenib, and suggests that some HCC cells might be spontaneously resistant to sorafenib when applied at clinically relevant concentrations. Our findings highlight the heterogeneity of HCC, and argue in favour of the potential interest of patient stratification before the prescription of sorafenib. At this stage however, an important limitation of the present work must be underlined. Our study is based on a limited number of HCC cell lines, most of which have been established in the context of viral carcinogenesis. These cell lines represent a biased estimation of the variety of clinical situations under which HCC cells can arise. Clearly, a major aim of future research should be to individually explore the activity of sorafenib in a more clinically relevant setting.
Our observations point to the importance of the control of the activity of oncogenic kinases in the treatment of HCC. The role of RAF kinases as therapeutic targets of sorafenib was suggested in the seminal study of Abou-Alfa et al., reporting an association between increased immunolabeling of phosphorylated ERK in HCC and a better efficacy of sorafenib.4 Since then, however, this association has not been fully explored. Both the incomplete specificity of sorafenib and its weak potency as an inhibitor of the RAF kinases have led some authors to question the role of RAF kinase inhibition in the effects produced by sorafenib.13 Our observations showing that RAF kinases escape inhibition in the context of sorafenib resistance highlight the importance of these kinases as therapeutic targets in HCC. Interestingly, a recent study exploring the paradoxical ability of sorafenib to promote skin carcinogenesis also reported the formation of heterodimers between B-RAF and C-RAF and the activation of C-RAF.24 The biochemical rationale for sorafenib-induced skin carcinogenesis therefore seems to present intriguing similarities with the mechanisms that we report for sorafenib resistance in HCC.
Finally, we show that EGFR is a potential determinant of the resistance of HCC cells to sorafenib. EGFR is a well-documented member of the RTK family and a potent regulator of the activity of RAS-RAF-MEK-ERK cascade.20 EGFR is overexpressed and found activated in more than half of HCCs.25, 26 During hepatocarcinogenesis, the activation of the EGFR does not depend on activating mutations, contrary to what is found in other solid tumors.27 Instead, autocrine loops implicating natural ligands of the EGFR, such as amphiregulin, activate this RTK.22 Here, we used several different approaches that block either EGFR expression, the kinase activity of EGFR, or its autocrine activation. We found that all of these approaches lead to the sensitization of EGFR-positive cells to sorafenib. Interestingly, both the chemical agents and the monoclonal antibody that block EGFR are already being used in clinical practice, and they have been tested as single agents in HCC in small clinical trials.28–32 Although the efficacy of these EGFR inhibitors as single agents is controversial and most likely modest, it is currently unknown if these agents could be used in combination with sorafenib in HCC. Our observation that EGFR activity can modulate the sensitivity to sorafenib suggests that EGFR neutralization might be useful as an adjunct to sorafenib in the treatment of HCC. Recent clinical phase II trials performed on various solid tumors already indicate that EGFR inhibitors can safely be associated with sorafenib in humans.33, 34 Whether EGFR inhibitors will improve the efficacy of sorafenib in HCC is currently investigated in a large phase III clinical trial (NCT00901901, Nexavar-Tarceva Combination Therapy for First Line Treatment of Patients Diagnosed With HCC, http://clinicaltrials.gov). In addition, our results open another interesting perspective related to the identification of predictive biomarkers. Indeed, they suggest the possibility that biological analyses of EGFR activation might be useful to predict the efficacy of sorafenib. Whether the direct measurement of EGFR expression, activity or the detection of its ligands, will help to anticipate the efficacy of sorafenib is an interesting possibility that awaits future studies. Exploring this possibility might ultimately facilitate the emergence of individually adapted medical strategies for patients with HCC.
The authors thank Paul Mischel, Cédric Boudot and Zuzana Saidak for kind gift of reagents, technical help and comments on the manuscript.