Sorafenib-Mediated Targeting of the AAA⁺ ATPase p97/VCP Leads to Disruption of the Secretory Pathway, Endoplasmic Reticulum Stress, and Hepatocellular Cancer Cell Death

Hepatocellular carcinoma (HCC) is globally the third cause of cancer-related death, with the majority of patients dying within 1 year of diagnosis (1). Recently sorafenib, a multikinase inhibitor of the Raf/MEK/ERK pathway and of tyrosine kinase receptors, was shown to induce apoptosis as well as inhibiting tumor cell proliferation and angiogenesis in a variety of tumors (2). The molecular involvement of the Raf-1 and tyrosine kinase signaling pathways is well established in the pathogenesis of HCC and provided a rationale for the use of sorafenib in HCC treatment. Indeed, sorafenib was the first targeted therapy to improve the outcome of patients with HCC (3) and has become a standard treatment of such disease. In preclinical experiments, sorafenib reduced tumor angiogenesis and increased tumor apoptosis in a mouse xenograft model of human HCC (4). However, the precise mechanisms through which sorafenib induces cell death in HCC remain to be identified.

Reports have indicated that sorafenib treatment may induce cell death signaling pathways through the activation of endoplasmic reticulum (ER) stress (5–10). The ER is the first compartment of the secretory pathway and is composed of a system of interconnected membranous tubules and vesicles in which proteins and lipids are synthesized and intracellular calcium levels are regulated. In addition, the ER functions in the maintenance of protein homeostasis (proteostasis) of secreted and transmembrane proteins. Misfolded or toxic proteins present in the ER are retrotranslocated to the cytosol and are subsequently degraded by the proteasome, in a process known as ER-associated degradation (ERAD). When the levels of misfolded protein overwhelms the ER folding and degrading capacity, an evolutionarily conserved cellular adaptive response, named the unfolded protein response (UPR), is triggered through the activation of 3 ER stress sensors phosphorylated extracellular signal–regulated kinase (PERK), ATF6α, and IRE1α. The UPR initially compensates for damage and restores ER homeostasis, however, when ER stress is not alleviated it can activate apoptosis (11, 12). Induction of apoptosis is a prominent mode of cytotoxicity for many chemotherapeutic drugs, of which some use ER stress–mediated cell death pathways.

In the present study, we found that sorafenib treatment selectively activates the PERK and IRE1α arms of the UPR but prevents activation of the ATF6α arm in liver cancer–derived cells. We show that sorafenib alters critical organelles of the secretory pathway exemplified by the fragmentation of the Golgi complex. We identify that tyrosine phosphorylation of the AAA⁺ ATPase p97/VCP is prevented upon sorafenib treatment, in turn leading to the disruption of the secretory pathway. Finally, we show that treatment with the vasolin containing protein (VCP) inhibitor N2,N4-dibenzylquinazoline-2,4-diamine (DBeQ) enhances sorafenib toxicity. Our results highlight for the first time an unanticipated impact of sorafenib on the secretory pathway in HCC cells and point toward p97/VCP as a potential therapeutic target.

Antibodies to calnexin were described previously (13). Antibodies against ERK1, C/EBP homology protein (CHOP), and c-jun-NH2-kinase (JNK) were obtained from Santa Cruz Biotechnology. Anti-alpha-1 antitrypsin (α1AT) antibodies were purchased from DAKO and Invitrogen. Mouse monoclonal anti-ATF6α was purchased from BioAcademia. Anti-LC3 antibody was from Novus Biologicals. Anti-phospho-ERK, anti-phospho-JNK (thr183/tyr185), and anti-eIF2α were from Cell Signaling Technology. Anti-LC3 antibodies were from Abnova. Rabbit polyclonal anti-Giantin and mouse monoclonal anti-p97 antibodies were purchased from Abcam. Mouse monoclonal anti-FLAG M2 and dithiothreitol (DTT) were obtained from Sigma. Fluorescently conjugated secondary antibodies were from Molecular Probes (Invitrogen). Tunicamycin (Tun) was from Calbiochem (EMD Biosciences Inc.). GFP-LC3 was kindly given by Dr. P. Codogno (INSERM U984, University Paris-Sud 11, Châtenay-Malabry, France). Autophagy was monitored as recommended by Klionsky and colleagues (14). siRNA to ATG5 was from Qiagen and used at 20 nmol/L final concentration in all experiments. NH₄Cl was from Sigma and used at 40 mmol/L final concentration. BpVPhen was from Enzo Life Sciences and used at 5 μmol/L final concentration. HepG2 and HuH7 cell lines were cultured in Dulbecco's Modified Eagle's Medium supplemented with 10% FBS (Hyclone) and antibiotics in a 37°C incubator containing 5% CO₂.

Total RNA was extracted using Trizol reagent (Invitrogen). Semiquantitative real-time PCR (RT-PCR) analysis was conducted as previously described (13). PCR products were applied to 1% to 3% agarose gels. For quantitative RT-PCR (qRT-PCR) analysis, all reactions were conducted using the SYBR Green PCR Core Reagents Kit (Bio-Rad) using Stratagene X4000 thermocycler (Stratagene). Each sample was normalized to Gapdh mRNA. The primers used are listed in Supplementary Table S1.

Characterization of the cell lines was achieved by genotyping specific polymorphic markers in the past 4 months.

Sorafenib (Fig. 1A) has previously been described to induce cell death in various cellular systems, and so we first tested for a similar response in liver tumor–derived HepG2 cells. Following 24 hours of treatment with 10 μmol/L sorafenib, cells showed toxicity. We quantified this toxicity by analyzing cellular viability using sulforhodamine B staining. As the incubation time with sorafenib increased, the percentage of viable cells decreased to 60% at 24 hours and to approximately 30% at 48 hours, thereby confirming our visual observations (Fig. 1B). To investigate whether the cell death observed in our system occurred via apoptosis, we incubated HepG2 cells with 10 μmol/L sorafenib for 24 or 48 hours followed by staining with Annexin V/propidium iodide (PI). As observed in Fig. 1B, the proportion of apoptotic cells increased to 20% at 24 hours and then to 60% at 48 hours of treatment. Finally, and similar to what was previously shown (15), sorafenib treatment of serum-starved cells largely prevented ERK1/2 phosphorylation upon exposure to medium containing 10% FBS (Fig. 1C). Together these results confirmed that sorafenib-mediated cell death in hepatocellular cells correlated with apoptosis induction and inhibition of the ERK/MAPK pathway. However the molecular mechanisms underlying this phenomenon still remained poorly characterized.

Besides the inhibition of the ERK/MAPK pathway, sorafenib-mediated cell death has been linked to the induction of apoptosis through ER stress (9), and so we sought to investigate whether this mechanism was occurring in the sorafenib-exposed HepG2 cells. We first analyzed whether the IRE1 arm of the UPR was activated. IRE1 is a transmembrane protein that serves as an ER stress sensor and can mediate the UPR. Activation of IRE1 leads to the unconventional splicing of the Xbp-1 mRNA, a target for the IRE1 endoribonuclease (16). Treatment of HepG2 cells with sorafenib led to a detectable splicing of Xbp-1 after 16 hours of incubation. This splicing was significantly reduced in HepG2 cells expressing a dominant negative form of IRE1 [IRE1α dominant negative mutant (IRE1-DN); ref. 17; Fig. 2A], indicating that sorafenib treatment induced Xbp-1 splicing through an IRE1-mediated mechanism. Similar results were obtained with tunicamycin, a known inducer of ER stress (Fig. 2A; Supplementary Fig. S1A). IRE1 has been implicated in apoptotic signaling through the activation of JNK via TRAF2, which leads to the phosphorylation of different cellular targets culminating in apoptosis (16). Because HepG2 cells were shown to undergo apoptosis upon incubation with sorafenib (Fig. 1C), we investigated whether JNK phosphorylation occurred after treatment with sorafenib. In wild-type cells, JNK phosphorylation was clearly detectable after 2 hours of treatment and increased with time. In contrast, no detectable JNK phosphorylation was observed in response to sorafenib in IRE-DN cells (Fig. 2B). Another potential link between ER stress and apoptosis is mediated by PERK, a transmembrane protein that senses ER stress, in turn phosphorylating and thereby inactivating eIF2α, eventually leading to a reduction in protein synthesis (12). In wild-type HepG2 cells, sorafenib stimulated eIF2α phosphorylation, which was significantly stronger in the IRE1-DN cells (Fig. 2C; Supplementary Table S2). This indicated that the PERK arm is activated in HepG2 cells upon treatment with sorafenib. To further study the induction of ER stress, we analyzed the expression of other UPR-associated genes, namely Chop and Gadd34. The expression of Chop, a potent transcription factor involved in the promotion of apoptosis upon ER stress, and which accumulates in the cells after eIF2α phosphorylation (12), increased in wild-type HepG2 cells after 8 hours of treatment with sorafenib (Fig. 2D). Incubation with imatinib, another tyrosine kinase inhibitor, did not affect Chop expression. Moreover, in IRE1-DN cells, there was also an increase in the expression of Chop, but to a lesser extent than that seen in wild-type cells, thereby confirming the contribution of IRE1 signaling to Chop mRNA induction. The expression levels of Gadd34, whose protein product contributes to eIF2α dephosphorylation, underwent a marked increase in its expression upon sorafenib treatment (Fig. 2D). Gadd34 induction ultimately mediates translation recovery and is a downstream target of ATF4, whose translation is triggered upon eIF2α phosphorylation (12). Gadd34 mRNA expression increased followed both Chop mRNA expression pattern and eIF2α phosphorylation (Fig. 2D). Incubation with imatinib did not lead to any discernible increase in Gadd34 expression.

The involvement of the third arm of the UPR in HepG2 cells exposed to sorafenib was also investigated. ATF6 is a transmembrane protein that serves as a proximal ER stress sensor, and that upon ER stress induction migrates from the ER to the Golgi complex where it becomes activated by proteolysis. The cytosolic portion of this protein translocates to the nucleus where it functions as a transcription factor, which induces the expression of various chaperones (such as BiP and GRP94), CHOP, and XBP1 (18, 19). To investigate the role of ATF6α, we transfected HepG2 cells with a FLAG-ATF6α construct and conducted immunocytochemistry followed by fluorescence microscopy. Transfected cells treated for 2 hours with the ER stress inducer DTT, displayed a nuclear localization of ATF6α (Fig. 3A, top). However, when we treated the cells for 2 hours with sorafenib before DTT treatment, ATF6α remained in the ER (Fig. 3A, bottom). Quantification of the images revealed that 15% ± 3% of the DTT-treated cells contained nuclear-localized ATF6α, compared with only 2% ± 0.1% of the cells incubated with sorafenib only (Fig. 3B), indicating that sorafenib treatment prevented the activation of the ATF6α pathway. This inhibition of ATF6α activation was further confirmed by measurement of the expression of Orp150, Herpud1, and BiP, which did not vary when compared with untreated cells (Fig. 3C), in contrast to the sharp increase in expression seen in cells treated with DTT. Interestingly, when we coincubated HepG2 cells with sorafenib and DTT, Herpud1 and BiP mRNA showed expression levels similar to those observed in the untreated cells, seemingly as if ATF6α was not activated (Fig. 3C). Activation of the ATF6α pathway was independent of IRE1 activity as illustrated by the results obtained in IRE1-DN cells (Fig. 3B). Together this indicates that although the PERK and IRE1α arms of the UPR are activated upon sorafenib treatment, the ATF6α arm is impaired. This further suggests that the secretory pathway might be affected in sorafenib-treated cells. To test the functionality of the secretory pathway, we examined α1AT, a serpin family protease inhibitor secreted by hepatocytes. By monitoring the α1AT content in the culture medium and in the cell lysate by immunoblotting, we found that α1AT secretion was reduced by more than 60% upon sorafenib treatment, whereas the intracellular content remained identical (Fig. 3D). Imatinib treatment had only a modest effect on inhibition of secretion (Fig. 3D).

We next sought to establish whether these effects correlated with phenotypic changes in secretory pathway organelles. We therefore studied the morphology of the Golgi complex using immunofluorescence. We observed that sorafenib-treated cells displayed a time-dependent increase in Golgi fragmentation (Fig. 4A). A more pronounced effect was observed in IRE1-DN cells, with a higher proportion of cells presenting fragmented Golgi. This was not associated with defects at the level of the cytoskeleton, as shown by the presence of intact microtubule and actin networks (Supplementary Fig. S1B and S1C). To analyze whether the observed Golgi fragmentation could be reversed upon sorafenib withdrawal, we incubated HepG2 cells for 4 hours with sorafenib, which was then removed and incubation continued for a further 2 hours in the presence of complete medium. Incubation with sorafenib led to approximately 60% and 100% of wild-type displaying fragmented Golgi (Fig. 4B). Incubation with sorafenib led to approximately 100% of IRE1-DN cells displaying fragmented Golgi (Supplementary Fig. S2). In wild-type cells, when sorafenib was removed, the majority of the cells were able to recover, resulting in only 10% of the cells displaying fragmented Golgi (Fig. 4B). In the IRE1-DN cells, the effect of sorafenib on Golgi morphology could also be reversed, albeit to a lesser extent than in the wild-type cells, with approximately 60% of the cells still displaying a fragmented organelle (Supplementary Fig. S2). This was most likely due to the fact that the absence of IRE1 in the IRE-DN cells results in a lower capacity of the secretory pathway to restore its homeostasis.

Taking into account the results presented earlier about the disruption of the secretory pathway and the induction of ER stress, we investigated whether autophagy was occurring in the sorafenib-treated HepG2 cells as a clearance mechanism for damaged organelles. Sorafenib, in conjunction with the histone deacetylase inhibitor vorinostat, has previously been shown to induce autophagy, a cellular adaptive response to stress (7). We transfected HepG2 cells with a GFP-tagged LC3 construct and analyzed its subcellular localization upon sorafenib treatment. GFP-LC3 was evenly distributed throughout the cell under basal conditions (Fig. 4C). However, when cells were incubated for 8 hours with sorafenib, GFP-LC3 distribution became punctate with many of the punctate structures colocalizing with the Golgi complex (Fig. 4C, bottom). This phenomenon was further amplified in IRE1-DN cells (not shown). Finally, HepG2 (or HuH7, not shown) cells were transfected with a siRNA to ATG5 to prevent autophagosome maturation or with NH₄Cl to attenuate phagolysosomal fusion in the presence of sorafenib. The presence of the lipidated form of LC3 (LC3-II) was monitored using immunoblot. This revealed that ATG5 silencing prevented sorafenib-induced LC3 maturation, whereas preventing phagolysosomal fusion led to the accumulation of LC3-II, thereby showing an increased autophagic flux upon sorafenib treatment (Fig. 4D).

So far our experiments have established that sorafenib, a molecule defined as a multikinase inhibitor, acts on HepG2 (or in HuH7) cells by disrupting the secretory pathway, thereby promoting the induction of ER stress and autophagy. In addition, treatment with sorafenib induces the cells to undergo apoptosis. We hypothesized that the sorafenib-induced effects on the secretory pathway may be mediated by the altered phosphorylation of a target protein normally involved in one of the earlier cellular processes. To identify such a candidate an in-depth literature search was carried out, revealing that p97/VCP, an AAA⁺ ATPase involved in vesicle transport and fusion, presented these characteristics (20, 21) Furthermore, we have previously shown that the status of p97/VCP tyrosine phosphorylation is important for the assembly of the ER thereby serving as a regulatory step in the early secretory pathway (20, 21). We first analyzed the phosphorylation status of p97/VCP in cells treated with sorafenib for 4 hours. We found that p97/VCP was tyrosine-phosphorylated in the untreated controls, but this phosphorylation was abolished in sorafenib-treated cells (Fig. 5A). As a control, we incubated HepG2 cells with BpVphen, a tyrosine phosphatase inhibitor, and observed that under these conditions p97/VCP phosphorylation increased when compared with controls. When cells were coincubated with sorafenib and BpVphen, p97/VCP phosphorylation was greatly reduced as compared with the negative control (Fig. 5A). This suggests that sorafenib acts on p97/VCP by preventing its tyrosine phosphorylation. Using a cell-free system, we have previously shown that nonphosphorylated p97/VCP associates with membranes, where it participates in transitional ER membrane remodeling (22). Because we observed that sorafenib impaired the functionality of the early secretory pathway by promoting Golgi fragmentation, we analyzed the localization of p97/VCP in our system. To this end, we prepared membrane fractions from control or drug-treated HepG2 cells and conducted immunoblotting against p97/VCP. In the presence of sorafenib, the membrane association of p97/VCP was increased, whereas in the presence of BpVphen, it showed only residual membrane association (Fig. 5B), consistent with our previous report (22). When both inhibitors were combined, p97/VCP displayed lower membrane association when compared with the untreated control, but membrane association was greater than when only BpVphen was used (Fig. 5B). One explanation for this would be the higher degree of p97/VCP phosphorylation that was observed under these conditions, as shown in Fig. 5A. To test whether p97/VCP tyrosine phosphorylation was altered upon inhibition of its activity, the same experiment as in Fig. 5A was carried out, but when BpVphen was replaced by the p97/VCP reversible inhibitor DBeQ (23), it revealed that DBeQ did not alter the phosphorylation of this protein (Fig. 5C). We then further characterized the impact of these molecules on Golgi fragmentation. HuH7 cells treated with 1 or 10 μmol/L sorafenib showed an incremental fragmentation of the Golgi apparatus (Fig. 5D, 4 and 7). BpVphen did not impact significantly on the fragmentation and DBeQ treatment led to the perinuclear concentration of the Golgi apparatus (Fig. 5D, 2 and 3). The cotreatment of HuH7 with both sorafenib and BpVPhen partially reversed the sorafenib effect (compare Fig. 5D, 4, 5, 7, and 8). Incubation with the combination with sorafenib and DBeQ led to the formation of large intracellular punctate (Fig. 5D, 6), likely to constitute autophagosomes as illustrated in Fig. 4.

Because sorafenib is toxic to HepG2 cells (Fig. 1), and as it affects the phosphorylation and the localization/function of p97/VCP, we postulated that targeting p97/VCP activity might enhance sorafenib cellular effects. To this end, HepG2 cells were treated with the recently described p97/VCP reversible inhibitor, DBeQ (23, 24), either in the presence or absence of sorafenib. Incubation of HepG2 or HuH7 cells for 48 hours with increasing concentrations of sorafenib showed that cell death increased with the concentration of the inhibitor, resulting in calculated IC₅₀ of 4.3 ± 0.22 and 2.7 ± 0.11 μmol/L (Fig. 6A). DBeQ alone was not toxic to HepG2 cells at the concentrations tested in this study (Fig. 6A). However, when sorafenib and DBeQ (20 μmol/L) were combined, a synergistic effect was observed, leading to a statistically significant decrease in the sorafenib IC₅₀ to 0.6 ± 0.12 μmol/L and 0.4 ± 0.07, P < 0.03 (Fig. 6A). These results were also supported by the quantification of sorafenib-induced apoptosis in HepG2 (Fig. 6B, left) and HuH7 (Fig. 6B, right) cells, which was increased when cells were incubated with the combination of sorafenib (2 μmol/L) and DBeQ (20 μmol/L). Finally, LC3 maturation was assessed using immunoblot in HuH7 cells treated with sorafenib (2 μmol/L) or DBeQ (20 μmol/L) or a combination of both (Fig. 6C). This revealed that as previously shown DBeQ induced an accumulation of LC3-II (23), phenomenon that was amplified by cotreatment with low dose of sorafenib, thereby indicating the amplification of the autophagic process.

Altogether, our data indicate that sorafenib-mediated cell death occurs in part through the disruption of the secretory pathway, which could occur due to the altered tyrosine phosphorylation of p97/VCP. Moreover, we show that this phenomenon can be further enhanced when cultured cells are exposed to a reversible inhibitor of p97/VCP.

In this study, we show that sorafenib induces cell death in a HCC model cell line, at least in part through the activation of the UPR, which results from sorafenib-mediated impairment of secretory pathway function. In addition to disrupting the secretory pathway, sorafenib treatment promotes autophagy, most likely as a compensatory mechanism. We also show that upon sorafenib treatment, tyrosine phosphorylation of the AAA⁺ ATPase p97/VCP is prevented, thereby stabilizing its membrane association. Together, these results led us to propose that inhibition of p97/VCP might lead to enhance sorafenib-mediated toxicity. This hypothesis was validated using the reversible pharmacologic inhibitor of p97/VCP, DBeQ, which further sensitized the cells to sorafenib.

Sorafenib primarily inhibits VEGF receptor (VEGFR), platelet-derived growth factor receptor (PDGFR), and Raf kinases, although other kinases are also potential targets (4, 5, 25). It was previously reported that p97/VCP is phosphorylated (26), and that its phosphorylation is regulated by the tyrosine phosphatase PTPH1 in cultured cells (27) and the tyrosine kinases JAK2 in vitro systems (22), or upon ligation of the T-cell receptors (TCR) in T cells (28). Other tyrosine kinases may also be involved, as p97/VCP presents conserved tyrosine phosphorylation sites for Abelson murine leukemia viral oncogene homolog (ABL) and lymphocyte-specific protein tyrosine kinase (LCK) kinases (data not shown). Sorafenib has been shown to inhibit JAK2 activation (29), which could explain its effect on p97/VCP phosphorylation. However, the effects of sorafenib on JAK2 remain controversial (30), and more work is needed to identify the kinases responsible for p97/VCP tyrosine phosphorylation in the liver.

The functions of p97/VCP in the liver have been described in both physiologic and pathologic mechanisms. Proteomics studies have suggested that p97/VCP might play some role in the liver regeneration and in fetal liver development (31), and it has also been shown that p97/VCP contributes to the formation of Mallory bodies (32). Moreover, p97/VCP autoantibodies have been found in cases of autoimmune hepatitis (33), primary biliary cirrhosis (34, 35), and in HCC (36). Elevated expression of p97/VCP in HCC has also been associated with increased incidence of tumor recurrence (37). Our results show that in the context of HCC, p97/VCP might play an important role in maintaining protein secretion homeostasis, through a tightly regulated tyrosine phosphorylation-dependent mechanism (Fig. 6D). Together, this information suggests that p97/VCP might play an instrumental role in liver physiology and pathophysiology and consequently represents an interesting potential therapeutic target.

In our study, we also show that sorafenib treatment leads to the disruption of the secretory pathway, thereby uncoupling the PERK/IRE1 branches of the UPR from the ATF6 branch, which requires functional ER-to-Golgi transport. This relatively unique observation also highlights the role of sorafenib-inhibited kinases in the maintenance of the secretory pathway and might provide novel insights into a kinase-mediated control of protein secretion. Moreover, we identify p97/VCP as a central player coordinating the balance between protein secretion and clearance/autophagy through phosphorylation-dependent mechanisms. This study therefore shows that p97/VCP could represent a relevant therapeutic target to improve HCC treatment through alteration of proteostasis as previously suggested for other models (38). As a consequence in HCC, alteration of the secretory pathway by sorafenib, leading to partial UPR activation would be greatly enhanced by inhibition of p97/VCP, thus leading to proteostasis imbalance and ultimately to cell death. As cancer cells are particularly sensitive to such alteration (38), this type of strategy would be particularly adapted for HCC, which arise in most cases on cirrhotic livers.

In conclusion, we propose that in HCC sorafenib prevents p97/VCP tyrosine phosphorylation, thereby leading to the disruption of the secretory pathway and ER stress–mediated cell death. This mechanism represents a novel aspect of sorafenib-mediated cell death in HCC and points toward p97/VCP as a potential therapeutic target in this disease.

No potential conflicts of interest were disclosed.

Conception and design: E. Chevet

Development of methodology: S. Taouji, D. Arma, E. Chevet

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Yi, A. Higa, E. Marza, C. Castain, B. Le Bail, C. Balabaud, J.-F. Blanc, E. Chevet

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Yi, A. Higa, S. Taouji, E. Marza, J.-F. Blanc, E. Chevet

Writing, review, and/or revision of the manuscript: P. Yi, S. Taouji, M.G. Bexiga, E. Marza, D. Arma, C. Castain, B. Le Bail, J.C. Simpson, J. Rosenbaum, P. Bioulac-Sage, J.-F. Blanc, E. Chevet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Balabaud, E. Chevet

Study supervision: E. Chevet

Sorafenib was kindly provided by Bayer. DBeQ was kindly provided by Dr. Franck Schoenen (Molecular Libraries Probe Production Centers Network Specialized Chemistry Center, Lawrence, KS). The authors thank Gaelle Cubel for technical help.

This work was funded by a grant from Avenir (INSERM)—Institut National du Cancer (INCa), Ligue contre le Cancer, a Programme d’Action Intégré de Recherche sur le CHC to E. Chevet and a Ulysses/IRCSET grant to E. Chevet and J.C. Simpson. E. Chevet was supported by a contrat d'interface from Inserm—Centre Hospitalo-Universitaire (CHU) Pellegrin, Bordeaux. P. Yi was funded by a “poste vert” INSERM, E. Marza was funded by a grant from the Association de Recherche contre le Cancer (ARC). M.G. Bexiga was funded by an European Science Foundation (ESF) grant.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.