The pancreatic niche inhibits the effectiveness of sunitinib treatment of pancreatic cancer

Current treatments for pancreatic ductal adenocarcinoma (PDA) are ineffective, making this the 4th leading cause of cancer deaths. Sunitinib is a broad-spectrum inhibitor of tyrosine kinase receptors mostly known for its anti-angiogenic effects. We tested the therapeutic effects of sunitinib in pancreatic cancer using the Ela-myc transgenic mouse model. We showed that Ela-myc pancreatic tumors express PDGFR and VEGFR in blood vessels and epithelial cells, rendering these tumors sensitive to sunitinib by more than only its anti-angiogenic activity. However, sunitinib treatment of Ela-myc mice with either early or advanced tumor progression had no impact on either survival or tumor burden. Further histopathological characterization of these tumors did not reveal differences in necrosis, cell differentiation, angiogenesis, apoptosis or proliferation. In stark contrast, in vitro sunitinib treatment of Ela-myc– derived cell lines showed high sensitivity to the drug, with increased apoptosis and reduced proliferation. Correspondingly, subcutaneous tumors generated from these cell lines completely regressed in vivo after sunitinib treatments. These data point at the pancreatic tumor microenvironment as the most likely barrier preventing sunitinib treatment efficiency in vivo. Combined treatments with drugs that disrupt tumor fibrosis may enhance sunitinib therapeutic effectiveness in pancreatic cancer treatment.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDA) is the most common pancreatic tumor (accounting for more than 90% of cases). PDA is aggressive and difficult to detect at an early stage, with few effective treatments and low survival rates, making it a major challenge for biomedical research. Indeed, this adenocarcinoma is predicted to become the second leading cause of cancer-related deaths in the United States by 2020 [1,2]. Current therapies for unresectable tumors, which are the most common ones, include chemotherapy administration based on gemcitabine, folifirinox or nab-paclitaxel, but these are very inefficient and only minimally improve patient survival, such that new therapeutic strategies with improved efficiency are urgently needed. In contrast, other types of less-frequent pancreatic tumors, such as acinar cell carcinomas [3] and pancreatic neuroendocrine tumors [4], are associated with better prognosis and longer survival rates.
Sunitinib is a broad-spectrum receptor tyrosine kinase (RTK) inhibitor whose targets include VEGFR-1,-2,-3, PDGFR-α,-β [5], stem cell factor receptor (c-KIT), Fmslike tyrosine kinase-3 receptor, the glial cell line derived neurotrophic factor receptor (RET) and colony-stimulating factor type 1 receptor (CSF-1R) [5][6][7]. Sunitinib inhibits endothelial cell proliferation and therefore is considered Research Paper www.impactjournals.com/oncotarget as an anti-angiogenic drug [8]. However, the expression of sunitinib-targeted RTKs in other stromal cells, as well as in the tumor epithelium, suggests that it could have a more extended mechanism of action with potential multiple effect in these different cells [5,9,10]. Sunitinib treatment has given impressive results in neuroendocrine pancreatic tumors, and its use for this pathology was recently approved by the FDA [11,12]. In contrast, phase II clinical trials to test sunitinib for treating PDA have consistently shown its failure, either when sunitinib was combined with gemcitabine for patients with advanced or metastatic PDA [13] or when it was given as a second-line therapy after gemcitabine failure [14].
The reasons behind this inefficacy of sunitinib treatment in PDA patients are not well understood. Several in vitro and in vivo reports on pancreatic ductal tumor cell lines and xenografts showed encouraging results, suggesting that sunitinib could potentially be used to improve standard chemotherapeutical treatments [10,[15][16][17][18][19][20]. In vivo sunitinib treatment in transgenic mouse models of pancreatic cancer was ineffective for k-ras-driven PDA yet successful for a neuroendocrine preclinical model (RIP-Tag) [21]. Understanding these apparently contradictory results thus requires further studies to elucidate the precise mechanistic roles and potential uses of sunitinib in PDA therapy.
Here, we have tested sunitinib effects in the Ela-myc transgenic mouse model of pancreatic cancer, both in vitro and in vivo. Ela-myc mice develop acinar tumors that can undergo acinar-to-ductal metaplasia and progress to ductal adenocarcinomas, mimicking human PDA progression [22][23][24]. In contrast to xenograft models, transgenic mice can better recapitulate tumor onset, tumor progression and tumor-stroma crosstalk, the latter of which is particularly important in pancreatic cancer. This study using the Ela-myc model complements the only one previously published preclinical report of sunitinib in pancreatic cancer that used transgenic mice (with a k-ras-based model) [21]. Using the Ela-myc model, we were able to analyze sunitinib effects in ductal as well as, for the first time, in acinar tumors. Our in vivo and in vitro data provide new insights into the therapeutic use of sunitinib for exocrine pancreatic cancer.

Ela-myc tumors express sunitinib-targeted tyrosine kinase receptors in both stromal and cancer cells
The Ela-myc transgenic mouse model overexpresses the oncogene c-myc under the control of the elastase promoter, which leads to acinar pancreatic tumors. Importantly, approximately 50% of these tumors progress to ductal tumors, which are associated with abundant stroma-one of the principle hallmarks of human PDA [22,23]. Although mutation of K-Ras is the most frequent alteration driving pancreatic cancer, c-Myc appears to have a key role in PDA development and progression [25][26][27] and has also been linked with K-Ras in this pathology [28,29]. To test the effects of sunitinib in this model of pancreatic cancer, we first analyzed the expression of sunitinib-targeted RTK in acinar and ductal tumor areas from Ela-myc mice ( Figure 1). Immunohistochemical analyses showed that VEGFR2, PDGFR-α and PDGFR-β were expressed in pancreatic tumors, both acinar ( Figure  1a-1f) and ductal areas (Figure 1g-1l). Importantly, these receptors were expressed not only in blood vessels and tumor stroma (Figure 1, arrows) but also in epithelial cancer cells (Figure 1, arrowheads). These data suggested a potential sunitinib sensitivity of Ela-myc tumors, which could be due not only to its well-known anti-angiogenic effect but also to its effects on the RTK that are expressed in pancreatic tumor cells.

Sunitinib treatment effects on survival, tumor burden, differentiation and necrosis in Ela-myc mice
Next, we analyzed the effects of sunitinib treatment in tumor progression and survival of Ela-myc mice. These mice develop pancreatic cancer with 100% penetrance between 2 and 7 months of age. Ela-myc mice with advanced tumors (4.5 months old) were treated by oral administration of sunitinib or control vehicle for 15 days. After this treatment, animals were followed until endpoint disease as determined by strict ethical guidelines ( Figure  2A). Sunitinib treatment had no significant effect on Elamyc mice survival, as determined by Kaplan-Meier curve analyses ( Figure 2B). Similarly, Ela-myc tumors from sunitinib-treated mice did not display differences in tumor mass ( Figure 2C). Finally, further histopathologic analyses of tumor samples to evaluate tumor differentiation, as determined by the percentage of acinar versus ductal component in each tumor ( Figure 2D) and by necrosis ( Figure 2E), failed to detect significant differences between sunitinib-treated and control mice. These data indicate that sunitinib therapy is inefficient in well-established pancreatic tumors in the Ela-myc cancer model.
To determine if mice at earlier stages of pancreatic tumorigenesis were sensitive to sunitinib therapy, 2-month-old Ela-myc mice (which usually have incipient tumors) were treated with sunitinib in a similar manner as the older mice ( Figure 2F). However, sunitinib treatment did not significantly affect animal survival ( Figure 2G) tumor weight ( Figure 2H), differentiation ( Figure 2I) or necrosis ( Figure 2J).

Sunitinib treatment effects on tumor angiogenesis in Ela-myc mice
As sunitinib treatment inhibits blood vessel formation in cancer, we tested for an anti-angiogenic effect on Ela-myc mice by analyzing the number of blood vessels in pancreatic tumors using von Willebrand factor immunohistochemical staining ( Figure 3). Elamyc pancreatic ductal tumor areas displayed higher vascularization as compared to acinar ones ( Figure 3; compare a with c), suggesting increased susceptibility to sunitinib anti-angiogenic effects. However, quantification of the number of blood vessels did not show any differences between vehicle-treated (control) and sunitinib-treated mice, either in acinar or ductal tumor areas ( Figure 3, right panel). These data indicate that Ela-myc pancreatic tumors are not responsive to the antiangiogenic effects of sunitinib.

Sunitinib treatment effects on cancer cells in Ela-myc mice
Considering the high expression of sunitinibtargeted RTKs found in Ela-myc pancreatic cancer cells (Figure 1), we next aimed to determine whether sunitinib treatment may directly impact cancer cells, by either inducing apoptosis or blocking tumor cell proliferation. To evaluate sunitinib effects on cancer cell apoptosis, we analyzed cleaved caspase 3 staining of both acinar and ductal tumor areas ( Figure 4A). As previously reported [30], acinar tumors showed increased cell apoptosis compared to ductal ones ( Figure 4A, compare a with Figure 1: Ela-myc tumors express tyrosine kinase receptors that can be targeted by sunitinib. Immunohistochemical analyses of Ela-myc mice pancreatic tumors were performed to detect tyrosine kinase receptors known to be sunitinib targets, such as VEGFR2, PDGFR-α and PDGFR-β. Scale bars for a-c. and g-i., 50 μm; for d-f. and j-l., 20 μm. www.impactjournals.com/oncotarget c), although quantification of cleaved caspase 3 positive areas showed no significant differences between control and sunitinib-treated animals ( Figure 4A, right panel). Subsequently, we tested whether sunitinib treatment could inhibit tumor cell growth by measuring cell proliferation using P-Histone H3 for acinar tumors ( Figure 4B, a and b) and Ki67 immunostaining for ductal tumors ( Figure  4B, c and d) (see Material and Methods). No significant differences in the proliferative rate of acinar or ductal pancreatic tumor lesions were detected between vehicletreated and sunitinib-treated animals ( Figure 4B, right panels).
Altogether, these data indicate that in vivo sunitinib treatment in the Ela-myc model has no effect on tumor angiogenesis, cell proliferation or apoptosis, with no impact on tumor progression or animal survival.

Sunitinib increases cell death and inhibits cell proliferation in Ela-myc tumor-derived cell lines
Fibrotic stroma is one of the major hallmarks of pancreatic cancer, and its role in hindering efficient drug delivery and promoting therapy resistance is well known [31]. To determine if the lack of effects in vivo of sunitinib treatment in Ela-myc mice could be due to impaired drug delivery because of this pancreatic fibrotic barrier, we tested sunitinib in vitro on two distinct Ela-myc pancreatic tumor-derived cell lines (Emyc-1 and Emyc-10). Similar to Ela-myc pancreatic tumors in vivo, these cells expressed the sunitinib-targeted RTKs VEGFR2, PDGFR-α and PDGFR-β ( Figure 5A). Next, we tested the susceptibility of these cells in vitro to different doses of sunitinib. Emyc-1 cells treated with sunitinib (1 to 4 μM) showed impaired viability as compared to vehicle-treated cells, as determined by bright field microscopy ( Figure  5B). Kinetic quantification of cell viability by MTT detection clearly showed that in vitro sunitinib treatment of Emyc-1 cells was strongly cytotoxic, in a specific and dose-dependent manner, as compared to vehicle treatment ( Figure 5C). We then analyzed whether the sunitinibinduced cytotoxicity was a consequence of increased apoptosis, decreased proliferation, or both, by detecting the active form of caspase 3 and P-Histone H3, respectively. In vitro treatment of Emyc-1 cells with sunitinib resulted in increased apoptosis, shown by immunofluorescence and Western blot detection of cleaved caspase 3 ( Figure 5D), and in a dose-dependent reduction of cell proliferation, shown by immunofluorescence staining and quantification of P-Histone H3 ( Figure 5E). Similar dose-dependent susceptibility to sunitinib treatment was observed in Emyc-10 cells (data not shown).
Thus, in vitro, sunitinib increases cell death and reduces cell proliferation of Ela-myc-derived pancreatic cancer cell lines. This suggests that the lack of effects observed in vivo may be due to impaired drug delivery to the pancreatic tumor, rather than cell insensitivity to the drug.

Sunitinib treatment affects subcutaneous tumors established from Ela-myc tumor-derived cell lines
We next tested the efficiency of sunitinib in vivo on tumors from Emyc-1 cells injected subcutaneously in the dorsal flanks of immunosuppressed mice. Once tumors reached 0.5 cm 2 , animals were orally treated with sunitinib or control vehicle, and evaluated for drug effects. After a daily treatment with sunitinib or vehicle for 15 days mice were sacrificed and tumors analyzed ( Figure 6A). Interestingly, sunitinib administration caused complete tumor regression ( Figure 6B, 6C). H&E and histopathological analysis revealed subcutaneous tumors only in mice treated with control vehicle, whereas in mice treated with sunitinib we only observed an inflammatory abscess with no presence of tumor cells ( Figure 6D).
These data suggest that the ineffectiveness of sunitinib treatment in Ela-myc transgenic mice was due to the pancreatic tumor localization. This most likely due to its abundant desmoplastic stroma, as tumors derived from this model in a non-pancreatic niche were highly responsive to the drug in vivo.

DISCUSSION
Gemcitabine was already in use as a first-line PDA therapy by 1997, but without high success rates [32]. Patient survival has slightly improved since then through the use of novel chemotherapy combinations [33,34], but these new treatments are still unable to reach one year as patient median overall survival, leaving PDA as one of the tumors with the worst prognosis. Thus, new effective therapies are urgently needed to treat this pathology. The multi-target inhibitor of RTKs, sunitinib, recently emerged as a possible hope for PDA therapy. However, although sunitinib therapy has improved treatment of tumors like renal cell cancer and gastrointestinal stromal tumors, it has not led to improvements for PDA patients [13,14]. Using a preclinical mouse model of pancreatic cancer, we now provide insights to the failure of sunitinib treatment in this pathology as well as potential ways in which this can be overcome. Specifically, we have reported a preclinical study that addresses the impact of sunitinib in pancreatic ductal and acinar tumors using a transgenic mouse model of pancreatic cancer. We found no significant differences on tumor progression or hallmarks or on survival between sunitinib-treated animals and control ones (Figure 7).
However, tumor derived cell lines did respond to sunitinib in vitro in a dose-dependent manner. Further, allografts from subcutaneous injections of Ela-myc tumor-derived cells regressed after sunitinib treatment, suggesting that the complex tumor microenvironment present in the pancreatic niche impairs sunitinib effects in the Ela-myc model (Figure 7).
One of the best characterized effects of sunitinib as an anti-cancer therapy is its ability to inhibit angiogenesis by targeting VEGFR on tumor endothelial cells. As angiogenesis is a key contributor to tumor progression and metastasis in most solid tumors, a large number of preclinical and clinical studies have focused on targeting tumor vasculature as an anti-cancer therapy [35]. However, tumor vessels are sparse and poorly functional in PDA, and whether angiogenesis is important for its progression is controversial. On the one hand, hypovascularization of PDA should render these tumors very dependent on this limited vascular network and therefore highly sensitive to anti-angiogenic therapies. Accordingly, several preclinical studies have gathered data showing the benefits of antiangiogenic treatments in PDA [15,36,37]. On the other hand, recent data have shown that increasing blood perfusion and normalizing the vasculature development favor drug delivery and survival in PDA animal models [38]. Moreover, in PDA patients, clinical trials using antiangiogenic drugs have repeatedly failed [39][40][41].
An important advantage of sunitinib as compared to standard anti-angiogenic treatments is that it exerts additional anti-tumor effects [42,43]. Thus, sunitinib can impair tumor desmoplasia by targeting PDGF-dependent fibroblasts proliferation [44] and can block tumor growth by inhibiting several RTKs in cancer cells [5,6,42]. Indeed, sunitinib has shown excellent results both in preclinical and clinical settings for different tumors [5,8,42,45,46] and has gained FDA approval for several neoplasms [11,47].
However, despite its multi-bullet targeting, sunitinib has not proven effective for PDA. In preclinical mouse models using pancreatic xenografts, the overall survival rate was not improved by treatment of sunitinib alone [10] but was improved when sunitinib was combined with chemotherapy [10,[15][16][17][18][19][20], radiotherapy [19,48] or miRNA strategies [49]. In contrast, sunitinib failed to increase survival rates in either the k-ras or the c-myc genetically engineered mouse models ( [21] and this manuscript), even in combination with gemcitabine [21]. Here, it is important to note that k-ras-and c-myc-driven pancreatic cancer mouse models, in contrast to xenograft models, recapitulate histopathology and progression of human pancreatic cancer, including acinar-to-ductal metaplasia and a strong desmoplastic reaction [23,38,50]. In fact, fibrotic stroma has been reported to act as a mechanical barrier hampering chemotherapy delivery in PDA [31], which could explain the lack of response after sunitinib treatment in these transgenic models. In support of this hypothesis, we found that even though sunitinib treatment had no effects on tumor volume or histopathological hallmarks in Ela-myc mice in vivo, cancer cells derived from these tumors were highly sensitive to the drug in vitro. Further, after injecting cells Ela-myc pancreatic acinar and ductal tumors does not show any significant effect in vivo (left). In contrast, Ela-myc-derived tumor cell lines are highly sensitive to sunitinib in vitro, with increased apoptosis and decreased proliferation (center). Importantly, subcutaneous injection of these cells in SCID Beige mice forms tumors that are highly sensitive to sunitinib treatment (right). These data may help understand sunitinib failure for PDA patients and suggest that a combination of sunitinib and an anti-stromal drug may be a promising strategy for successfully treating this pathology in the clinics. derived from these tumors into a non-pancreatic niche in the mouse, sunitinib was highly effective in vivo, suggesting that sunitinib exerts its anti-cancer effects if it can efficiently reach the tumor (Figure 7). Importantly, therapeutic targeting of the tumor stroma has given encouraging results in preclinical studies and is now being tested in clinical trials [38,51,52]. Moreover, one of the few improvements in managing human PDA in the last 20 years has been by treating with a combination of gemcitabine and nab-paclitaxel, which compromises the stromal architecture to increase perfusion [34]. Preliminary data from mice xenograft PDA models combining nab-paclitaxel with sunitinib are positive [17], although further future preclinical experiments using sunitinib in combination with stromal disrupting drugs are required to better understand the anti-tumor effects of sunitinib in PDA therapy.
In agreement with a lack of sunitinib effectiveness in the k-ras and c-myc genetically engineered mouse models ( [21] and this manuscript), sunitinib failed in human PDA treatments when administered as a secondline therapy [14] or in combination with gemcitabine [13]. These data reinforce the idea that xenograft animal models have a low predictive value as compared to genetically engineered models in preclinical studies [53] and highlight the necessity of obtaining extensive preclinical data before moving to patients. In fact, an important meta-analysis of preclinical studies with sunitinib showed that only 2% of these had used genetically engineered mouse models [54]. These data encouraged us to replenish the lack of preclinical information with transgenic models for PDA pathology, using Ela-myc mice as a different transgenic pancreatic tumor model from the previously published k-ras model [21].
The molecular complexity of PDA tumors, which are genetically heterogenous, can explain why sunitinib gave positive results in maintenance therapy in metastatic PDA that had achieved disease control with first-line therapy (a very rare subset of PDA patients) [55]. Likewise, a case report has documented an impressive extended survival after treating with gemcitabine and sunitinib [56]. Thus, the lack of patient selection in published clinical trials may have underestimated the sunitinib potential for PDA treatment, highlighting the importance of personalized medicine in identifying patients who can benefit from sunitinib.
Finally, even though PDA is the most common type of pancreatic cancer, there are patients with other less frequent pancreatic neoplasms who respond poorly to conventional treatments and who might benefit from sunitinib therapy. For instance, patients with pancreatic neuroendocrine tumors, which represent around 1.3% of pancreatic cancers, have shown a very good response to sunitinib in preclinical models [21,43,57] and clinical trials [11,58,59]. Indeed, sunitinib has been recently approved by the FDA for the treatment of these tumors [11]. These differences in sunitinib response between neuroendocrine and PDA tumors might be explained by the high vascularization of the former and/or by the dense desmoplasic reaction that hampers drug delivery in PDA. Another uncommon pancreatic cancer is acinar cell carcinoma, which accounts for 1% to 2% of all pancreatic tumors in adults and up to 15% in infancy. Surgical resection is the only chance of cure for acinar cancer, and there is a lack of systemic successful therapies for advanced non-resectable tumors [60]. Research on acinar cancer treatment has been handicapped by the low number of acinar cell lines available, which has limited the in vitro laboratory tools, compromised the use of xenografts in vivo and resulted in a lack of preclinical studies addressing acinar tumor therapy. Moreover, as this pancreatic cancer type is very rare [61], clinical trials are unfeasible, limiting studies to small sample sizes and case reports. The transgenic Ela-myc mouse model, which develops mixed acinar and ductal tumors, now provides the first tool for studying sunitinib treatment of pancreatic acinar cell carcinomas. Our data show that, similar to PDA tumors, acinar tumors do not respond in vivo to sunitinib in this preclinical model. Unfortunately, we cannot test in vitro sunitinib effects on acinar cells derived from Ela-myc tumors, as these cells undergo acinar-to-ductal metaplasia in culture. Therefore, whether this lack of sunitinib efficacy in Ela-myc acinar tumors in vivo is due to inefficient drug delivery into the tumor, or to insensitivity of acinar cells to the drug, requires further investigation.
The data reported here fit with previous preclinical results with sunitinib in PDA transgenic models and, more importantly, with unselected clinical trials for this pathology. Overall, these results stress the importance of using genetically engineered mouse models rather than xenograft mouse models in preclinical studies. We hypothesize that combining sunitinib with agents targeting the stroma (rather than with the conventional gemcitabine) and segregating patient cohorts using personalized medicine could result in the successful use of this drug for PDA patients.

Mice and sunitinib treatments
Animal procedures were approved by the PRBB Ethical Committee for Animal Experimentation. Founder Ela-myc mice (C57Bl/6 genetic background) were kindly provided by E. Sandgren (University of Wisconsin-Madison, WI). Mice were housed and fed ad libitum as previously described [22,62]. Genotyping primers: c-myc (5′-CAC CGC CTA CAT CCT GTC CAT TCA AGC-3′ and 5′-TTA GGA CAA GGC TGG TGG GCA CTG-3′), resulting in a 200 bp band. SCID Beige mice were obtained from the PRBB Animal Facility. Emyc-1 cells (3 × 10 6 ) were injected subcutaneously in the two dorsal lateral flanks of 20 SCID Beige mice. Sunitinib (kindly provided by Pfizer) was prepared in a vehicle solution (0.5% carboxymethylcellulose, 1.8% NaCl, 0.4% Tween-80, 0.9% benzyl alcohol) and kept in the dark. Mice (n = 10 for each condition) were treated for 15 days by oral administration daily with 80 mg/Kg of sunitinib malate (corresponding to 60 mg/Kg of the active ingredient) or vehicle solution, according to the manufacturer instructions. Ela-myc mice were treated when they were 2-or 4.5-months old, for incipient or advanced tumors, respectively; SCID Beige mice were treated when tumors reached 0.5 cm 2 .
Ela-myc mice were sacrificed when tumors compromised animal well-being, as judged by the protocol of Morton and Griffiths [63]. SCID Beige mice were treated with sunitinib or the control vehicle for 15 days and tumor size in all mice was measured weekly with a caliper. Mice were sacrificed, tumors were weighed, fixed in buffered formalin for 24 h, dehydrated and embedded in paraffin.

Histopathology and immunohistochemistry
For histopathological analysis, tumor sections from mice treated with sunitinib or control vehicle were stained with H&E and evaluated by two expert pathologists (M.I. and J.M-C.) to determine acinar ductal differentiation and the necrotic index [30,62].
For quantification, 10 images were acquired at 10× for each tumor and analyzed with ImageJ software. To quantify angiogenesis and apoptosis, the area stained positive for vWF or active caspase 3, as indicated, was related to total area. Due to intrinsic differences in their proliferation, ductal and acinar cells were immunostained with Ki67 or P-Histone H3, respectively, as previously described [62]. For quantification, the percentage of Ki67 (in acinar regions) or P-Histone H3 positive cells (in ductal regions) were obtained by relating the area corresponding of nuclei positive for Ki67 or P-Histone H3 to the total nuclei area (positive for hematoxylin).

Western blots
Whole cell extracts were prepared with Laemmli buffer, and samples were boiled at 95°C for 5 min. Proteins were resolved by SDS-PAGE and transferred to nitrocellulose filters for Western blot analysis. The primary antibodies used were: anti-VEGFR2 (Cell Signaling Technology), anti-PDGFR-α and anti-PDGFR-β (Santa Cruz), anti-cleaved caspase 3 (Cell Signaling Technology), anti-P-Histone H3 (Ser10) (Millipore) and anti-tubulin (Sigma). Specie-specific secondary antibodies conjugated to HRP (Dako) and ECL detection (Amersham) were used for band visualization.

Immunofluorescence to detect apoptosis and proliferation
Cells seeded over sterile coverslips and grown in 10% FBS were fixed in 4% PFA. Cells were permeabilized with 0.2% Triton in PBS and blocked with 5% BSA, 0.1% Tween-20. Coverslips were incubated overnight with rabbit anti-P-Histone H3 (Ser10) (Millipore) to determine proliferation or with rabbit anti-cleaved caspase 3 (R&D) to determine apoptosis. Negative controls were incubated with an irrelevant rabbit IgG (data not shown). An antirabbit Alexa Fluor 488 (Invitrogen) was used as the secondary antibody. Coverslips were mounted with DAPI Fluoromont-G (Southern Biotech), and IF was detected with an Olympus BX61 Microscope. Ten fields were quantified per coverslip at 10×. The number of P-Histone H3 positive cells was manually counted and compared www.impactjournals.com/oncotarget to the total number of cells (DAPI positive) for each experiment, which was quantified using an ImageJ macro developed in the CRG Advanced Light Microscopy Unit facility (CRG, Barcelona).

Statistical analysis
Statistical analyses were performed with SPSS version 12.0. Statistical significance cut-off was considered as p < 0.05. Kaplan-Meier analyses were used to establish survival curves, and comparisons were made using the log-rank test. The Student's t-test was used with normally distributed data (in vitro) and the Mann-Whitney test was used in other occasions (in vivo).