Angiogenic and signalling proteins correlate with sensitivity to sequential treatment in renal cell cancer.

BACKGROUND: We aimed to study key signalling proteins involved in angiogenesis and proliferation on the response to inhibitors of tyrosine kinases and mammalian target of rapamycin in first- and in second-line treatment of renal cell carcinoma (RCC).
METHODS: In a panel of human RCC tumours, in vitro and in nude mice, we evaluated the effect of sunitinib, sorafenib and everolimus, alone and in sequence, on tumour growth and expression of signalling proteins involved in proliferation and resistance to treatment.
RESULTS: We demonstrated that, as single agents, sunitinib, sorafenib and everolimus share similar activity in inhibiting cell proliferation, signal transduction and vascular endothelial growth factor (VEGF) secretion in different RCC models, both in vitro and in tumour xenografts. Pre-treatment with sunitinib reduced the response to subsequent sunitinib and sorafenib but not to everolimus. Inability by sunitinib to persistently inhibit HIF-1, VEGF and pMAPK anticipated treatment resistance in xenografted tumours. After first-line sunitinib, second-line treatment with everolimus was more effective than either sorafenib or rechallenge with sunitinib in interfering with signalling proteins, VEGF and interleukin-8, translating into a significant advantage in tumour growth inhibition and mice survival.
CONCLUSION: We demonstrated that a panel of angiogenic and signalling proteins can correlate with the onset of resistance to sunitinib and the activity of everolimus in second line.

Renal cell carcinoma (RCC) is the most common type of kidney cancer in adults, its incidence is increasing and it is often diagnosed when already metastatic (Lipworth ; Jonasch ).

Treatment of metastatic RCC has dramatically changed in the past 6 years, with the approval of several agents targeting the vascular endothelial growth factor (VEGF), such as bevacizumab; its receptors (VEGFRs), such as the multiple tyrosine kinase inhibitors (TKIs) sorafenib, sunitinib and pazopanib; the mammalian target of rapamycin (mTOR), such as everolimus and temsirolimus (Hudes ; Motzer ; Escudier , 2007b; Motzer ; Sternberg ). More recently, another VEGFRs TKI, axitinib, gained a pre-eminent place in the therapeutic arena (Rini ).

The availability of so many active agents requires a therapeutic algorithm to optimise their use and to enable patients to take advantage of multiple lines of treatment (Escudier ). However, a biological rationale and validated biomarkers predictive of response to single agents and, more in general, to antiangiogenic drugs, are still lacking.

Several studies analysed the predictive value of a variety of angiogenesis-related biomarkers, including VHL mutations, soluble VEGF and VEGFRs in the serum, polymorphisms of VEGF and interleukin-8 (IL-8), and IL-8 expression and secretion (Sonpavde and Choueiri, 2012). For instance, IL-8 overexpression was demonstrated in RCC xenografts resistant to sunitinib (Huang ) and IL-8 polymorphisms were associated with survival in patients treated with pazopanib (Xu ).

In spite of the fact that, to date, sunitinib in first line, and sorafenib or everolimus in second line, are the drugs mostly used to treat RCC patients, little information is available on their differential biological effect, as single agents and in sequential manner, on the expression of signalling effectors controlling angiogenesis and tumour growth.

To this aim, we used a panel of human RCC cell lines with different VHL status and studied the effect of sunitinib, sorafenib and everolimus, alone and in sequence, in vitro and in nude mice, on tumour growth and on the expression and function of a variety of signalling proteins critical for RCC proliferation, angiogenesis and development of resistance to treatment.

Materials and methods

Compounds

Everolimus, sunitinib and sorafenib were purchased by Selleck Chemicals (Houston, TX, USA). Cobalt chloride (CoCl2) was purchased from Sigma-Aldrich (Milan, Italy).

Cell cultures

Human ACHN, 769-P, 786-O, and Caki-2 RCC lines were obtained from the American Type Culture Collection (Manassas, VA, USA). Cells were maintained in RPMI or in McCoy's medium supplemented with 10% heat-inactivated fetal bovine serum, 20 mM HEPES, pH 7.4, penicillin (100 IU ml−1), streptomycin (100 mg ml−1) and 4 mM glutamine (ICN, Irvine, UK) in a humified atmosphere of 95% air and 5% CO2 at 37 °C.

Cell survival assay

Cells (104 cells per well) were grown in 24-well plates and exposed to increasing doses of the drugs. The percentage of cell survival was determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay according to the manufacturer's instructions.

Western blot analysis

Total cell lysates were obtained from cell cultures. Protein extracts were resolved by 8% SDS–PAGE and probed with anti-human, polyclonal pEGFR and EGFR, monoclonal pMAPK, MAPK, HIF-1, VEGF (Santa Cruz, Santa Cruz, CA, USA), polyclonal pAkt, Akt, pp70S6K, p70S6K (Cell Signaling Technologies, Beverly, MA, USA) and monoclonal actin (Sigma-Aldrich). Immunoreactive proteins were visualised by enhanced chemiluminescence (Pierce, Rockford, IL, USA).

ELISA

Human VEGF (hVEGF) concentrations in conditioned media from tumour cells and in mice sera were determined by ELISA as previously reported (Bianco ). The absorbance was measured at 490 nm on a microplate reader (Bio-Rad, Hercules, CA, USA) and the hVEGF concentrations were determined using linear regression analysis. Human IL-8 (hIL-8) concentrations in sera from nude mice were determined by using the human CXCL8/IL-8 Quantikine ELISA Kit (R&D Systems, Minneapolis, MN, USA), according to the manufacturer's instructions.

Nude mouse cancer xenograft models

Five-week-old Balb/cAnNCrlBR athymic (nu+/nu+) mice (Charles River Laboratories, Milan, Italy) maintained in accordance with guidelines of the Institutional Animal Care and Use Committee and in accordance with the Declaration of Helsinki were injected subcutaneously (s.c.) with 786-O human RCC cells (107 cells per mice) resuspended in 200 μl of Matrigel (Collaborative Biomedical Products, Bedford, MA, USA). To compare the effects of the different drugs, tumour-bearing mice were treated with sunitinib 40 mg kg−1, sorafenib 30 mg kg−1 or everolimus 5 mg kg−1, all administered per os, by gavage, once daily for 3 weeks following the schedule indicated in detail in Results.

Tumour diameter was assessed with a Vernier caliper, and tumour volume (cm3) was measured using the formula π/6 × larger diameter × (smaller diameter)2 as previously reported (Rosa ). Humane end points for removing animals from the study or for their euthanasia have been approved by the Institutional Animal Care and Use Committee before the study: (1) a tumour burden of >10% body weight (maximum allowed tumour size in an adult mouse, 2 cm3); (2) tumours that ulcerate, become necrotic or infected. Efforts have been made to minimise the pain and distress experienced by animals used in research.

Statistical analysis

The Student's t-test was used to evaluate the statistical significance of the in vitro results. The statistical significance of differences in tumour growth was assessed by one-way ANOVA and Dunnett's multiple comparison post test, and the statistical significance of differences in survival was evaluated by a log-rank test. All reported P-values were two-sided. All analyses were performed with the BMDP New System statistical package version 1.0 for Microsoft Windows (BMDP Statistical Software, Los Angeles, CA, USA).

Results

Characterisation of a panel of human RCC cell lines

We used a panel of clear cell RCC cell lines including ACHN, 769-P, 786-O and Caki-2 (Karam ). Based on the catalogue of somatic mutations in cancer from Sanger Institute (*COSMIC database, Catalogue of Somatic Mutations in Cancer, http://www.sanger.ac.uk/), ACHN cells are wild type, while the other cell lines are mutant for the VHL gene. Moreover, as reported for many RCC cell lines, ACHN and 786-O do express VEGFRs (Liu ; Ji ). We found that, consistently with their VHL status, ACHN cells secreted lower VEGF levels compared with the other cell lines, when cultured in complete medium (Figure 1A). Moreover, ACHN cells showed HIF-1α expression only when treated with the hypoxia-mimetic agent cobalt chloride (CoCl2), while 769-P and Caki-2 cells showed basal expression of HIF-1α. Although 786-O cells showed no expression of HIF-1α, this result is consistent with the report that 786-O is a VHL-deficient RCC cell line that constitutively expresses only HIF2α (Shinojima ) (Figure 1B).

Figure 1

Characterisation of a panel of human RCC cell lines. (A) Human VEGF (hVEGF) secretion in conditioned media from ACHN, 769-P, 786-O and Caki-2 cells cultured for 24 h in complete medium, as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments performed in triplicate. Bars, s.d. (B) Western blot analysis of HIF-1α expression on ACHN, 769-P, 786-O and Caki-2 total cell lysates. Cells were cultured in complete medium and stimulated for 3 h with CoCl2 (100 μM) before protein extraction.

Sensitivity of human RCC cell lines to sunitinib, sorafenib and everolimus

We analysed the sensitivity of ACHN, 769-P, 786-O and Caki-2 RCC cells to sunitinib, sorafenib and everolimus by performing dose–response MTT assays. As shown in Figure 2A, all the three agents were able to inhibit the survival of RCC cells, with everolimus resulting more effective than the other agents, particularly on 769-P and Caki-2 cells.

Figure 2

Sensitivity of human RCC cell lines to sunitinib, sorafenib and everolimus. (A) Percent of survival of ACHN, 769-P, 786-O and Caki-2 cells treated with increasing doses of sunitinib, sorafenib or everolimus (0.01–1 μM), as measured by MTT assays. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. *Two-sided P<0.005 vs control. Bars, s.d. (B) hVEGF secretion in conditioned media from ACHN, 769-P, 786-O and Caki-2 cells treated with sunitinib, sorafenib or everolimus (1 μM), as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. *Two-sided P<0.005 vs control. Bars, s.d. (C) Western blotting on total cell lysates from ACHN, 769-P, 786-O and Caki-2 cells treated for 24 h with sunitinib, sorafenib or everolimus (1 μM).

We analysed the capability of the three agents to reduce VEGF secretion in RCC cells by measuring hVEGF levels in conditioned media of 24 h-cultured cells with ELISAs. Everolimus was more efficient than sunitinib or sorafenib in causing a statistically significant reduction in VEGF secretion by tumour cells (two-sided P<0.005 vs control; Figure 2B).

We studied the effect of sunitinib, sorafenib and everolimus on signal transduction. Sunitinib showed no or poor effect on Akt, p70S6K and MAPK phosphorylation in all tested RCC lines. Sorafenib showed no effect or a slight induction on Akt or MAPK activation. Everolimus was able to inhibit mTOR effector p70S6K in all the cell lines, while the effects on Akt and MAPK phosphorylation were cell line dependent (Figure 2C).

Effect of sunitinib, sorafenib and everolimus on tumour growth, survival and signal transduction of athymic mice bearing subcutaneous 786-O RCC tumours

To evaluate the sensitivity of RCC cell lines to sunitinib, sorafenib and everolimus also in vivo, we xenografted the VHL mutant 786-O cells s.c. in nude mice. When tumours reached about 0.5 cm3 (day 21), mice were randomly assigned (n=10 per group) to receive the following: sunitinib 40 mg kg−1, sorafenib 30 mg kg−1 or everolimus 5 mg kg−1 administered orally once daily for 3 weeks. After an initial inhibition by the three agents, all tumours had resumed their growth by day 63. However, mice treated with everolimus reached the allowed size of 2 cm3 2 weeks after mice treated with sunitinib and 1 week after mice treated with sorafenib (Figure 3A). The one-way ANOVA test was used to compare tumour sizes among different treatment groups at the median survival time of the control group (6 weeks), which showed statistically significant differences for everolimus vs control (P<0.0001). Median survival of mice treated with everolimus was 90 days compared with 56 and 76 days of mice treated with sunitinib or sorafenib, respectively (Figure 3B). The difference was statistically significant for everolimus vs sunitinib (log-rank test, P<0.05). Western blot analysis on tumours lysates from mice killed on day 45, at the end of treatment, showed that everolimus caused an almost total suppression of pp70S6K, pMAPK, HIF-1 and VEGF levels (Figure 3C). Moreover, ELISAs on sera collected at the same time point showed a more than 50% reduction in circulating hVEGF following everolimus treatment (two-sided P<0.005 vs control; Figure 3D).

Figure 3

Effect of sunitinib, sorafenib and everolimus on tumour growth, survival and signal transduction of athymic mice bearing s.c. 786-O RCC tumours. (A) After 21 days following s.c. injection of 786-O RCC cells, mice were randomised (10 per group) to receive sunitinib, sorafenib or everolimus, as described in Materials and methods. The one-way ANOVA test was used to compare tumour sizes among different treatment groups at the median survival time of the control group (6 weeks). The differences resulted to be statistically significant for everolimus vs control (P<0.0001). Bars, s.d. (B) Differences in median survival resulted to be statistically significant for each treatment vs control and for everolimus vs sunitinib (log-rank test, P<0.05). (C) Western blot analysis on total lysates from tumour specimens of two mice killed on day 45, at the end of treatment. (D) hVEGF levels in the serum of two mice killed on day 45, at the end of treatment, as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments performed in triplicate. *Two-sided P<0.005 vs control. Bars, s.d.

Effect of sunitinib, sorafenib or everolimus on RCC cells pre-treated with sunitinib

786-O and Caki-2 cells, either naïve or pre-treated with increasing doses of sunitinib for 4 weeks, were treated for 3 days with sunitinib, sorafenib or everolimus and their effects were compared with cells pre-treated or not with sunitinib using an MTT assay. Compared with sunitinib-naïve cells, 786-O and Caki-2 cells pre-treated with sunitinib showed a reduced response to sorafenib or rechallenge with sunitinib. Conversely, we found that in both cell lines the sensitivity to everolimus was not affected by sunitinib pre-treatment (Figure 4).

Figure 4

Effect of sunitinib, sorefenib or everolimus on RCC cells pre-treated with sunitinib. Percent of survival of 786-O (A) and Caki-2 (B) cells pre-treated or not with increasing doses of sunitinib (0.01–2.5 μM) for 4 weeks, and then treated for 3 days with sunitinib, sorafenib or everolimus (1 μM). Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. *Two-sided P<0.005 vs cells not pre-treated with sunitinib. Bars, s.d.

Effect of long-term sunitinib on the expression of signalling proteins in 786-O RCC tumour xenografts

To evaluate the activity of everolimus and sorafenib after sunitinib treatment also in vivo, we xenografted s.c. the VHL mutant 786-O cells in nude mice. When tumours reached about 0.5 cm3 (21 days), 30 mice were treated with sunitinib 40 mg kg−1 per os. Then, when tumours reached about 1 cm3 (56 days, after 5 weeks of sunitinib treatment), mice were randomised (n=10 per group) to receive sunitinib 40 mg kg−1, sorafenib 30 mg kg−1 or everolimus 5 mg kg−1 (all administered per os) once daily for 3 weeks.

Western blot analysis on 786-O tumour lysates and ELISAs on mice sera from mice treated for 5 weeks with sunitinib were performed on day 56, before the switch to second treatment. At this time point, sunitinib started to lose its ability to inhibit p70S6K and MAPK phosphorylation and hVEGF secretion (Figures 5A and B) while, conversely, induced a 5-fold increase in IL-8 levels (Figure 5C).

Figure 5

Effect of sunitinib long-term treatment on signal transduction of athymic mice bearing s.c. 786-O RCC tumours. (A) Western blot analysis on total lysates from tumour specimens of two mice killed on day 56, at the end of sunitinib treatment. (B) hVEGF levels in the serum of two mice killed on day 56, at the end of sunitinib treatment, as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. Bars, s.d. (C) Human IL-8 levels in the serum of two mice killed on day 56, at the end of sunitinib treatment, as measured by ELISAs. *Two-sided P<0.0001 vs control. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. Bars, s.d.

Effect of sunitinib, sorafenib or everolimus in second line after sunitinib on growth, survival and signal transduction of 786-O tumours

We analysed the effect of sunitinib, sorafenib or everolimus administered after 5 weeks of sunitinib on 786-O tumours xenografted in nude mice. In second line, sorafenib was initially very effective, then, starting from day 84, everolimus effect was more sustained. Conversely, continuing treatment with sunitinib (rechallenging) was ineffective. In fact, tumours treated with the sequence sunitinib>everolimus reached the 2 cm3 target volume 2 weeks after mice treated with sunitinib>sorafenib, and 4 weeks after mice continuing with sunitinib (Figure 6A).

Figure 6

Effect of sunitinib, sorafenib or everolimus after a sunitinib long-term treatment on tumour growth, survival and signal transduction of athymic mice bearing s.c. 786-O RCC tumours. (A) On day 56, after 5 weeks of sunitinib treatment, 786-O tumour-bearing mice were randomised again (10 per group) to receive sunitinib, sorafenib or everolimus, as described in Materials and Methods. The one-way ANOVA test was used to compare tumour sizes among different treatment groups at the median survival time of the sunitinib chronic treatment group (8 weeks). They resulted to be statistically significant for sunitinib>sorafenib and sunitinib>everolimus vs sunitinib chronic treatment (P<0.0001). Bars, s.d. (B) Median survival resulted to be statistically significant for sunitinib>everolimus vs sunitinib chronic treatment (log-rank test, P<0.05). (C) Western blot analysis on total lysates from tumour specimens of two mice killed on day 80, at the end of second treatment. (D) hVEGF levels in the serum of two mice killed on day 80, at the end of second treatment, as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. *Two-sided P<0.005 vs control. Bars, s.d. (E) Human IL-8 levels in the serum of two mice killed on day 80, at the end of second treatment, as measured by ELISAs. Data represent the mean (±s.d.) of three independent experiments, each performed in triplicate, and are presented relative to untreated control cells. *Two-sided P<0.005 vs control. Bars, s.d.

Moreover, compared with the median survival of 42 days in untreated mice and 61 days in mice receiving continuing sunitinib, the sequence sunitinib>everolimus and sunitinib>sorafenib determined a median survival of 97 and 72 days, respectively. The difference in median survival resulted to be statistically significant for sunitinib>everolimus compared with chronic sunitinib (P=0.0029; Figure 6B). The effects on tumour growth were paralleled by biological evidence. In fact, the sequence sunitinib>everolimus caused an almost total suppression of pp70S6K, HIF-1 and VEGF levels in tumour lysates (Figure 6C) and a strong reduction in circulating hVEGF in mice sera (Figure 6D). Both sunitinib>sorafenib and sunitinib>everolimus sequences caused a statistically significant reduction in IL-8 levels (Figure 6E).

Discussion

The recent availability of several effective agents allowed a significant prolongation of survival of RCC patients. In fact, up to 70% of patients after obtaining a benefit from a first-line treatment, generally sunitinib, initiate a second-line treatment and, following a further progression, a considerable fraction of them maintains a performance status good enough to receive a third-line treatment. However, in spite of such abundance of therapeutic options, following the first-line treatment with a TKI-like sunitinib, it is still unclear the biological basis to choose another TKI or an mTOR inhibitor in second line. To date, either the TKI sorafenib or the mTOR inhibitor everolimus is the preferred second line in the majority of cases, while no direct comparison is still available (Escudier ; Ravaud and Gross-Goupil, 2012). More recently, the AXIS study showed that the novel TKI axitinib is effective and superior to sorafenib in second line following the TKI sunitinib (Rini ). Unfortunately, no comparative studies have been conducted on the effect of TKIs and mTOR inhibitors, used in first and in second line, on proliferative and angiogenic signalling. Notably, the issue of treatment sequence optimisation may be relevant also to delay, or overcome, the inevitable onset of resistance to targeted agents (Carbone ).

In the attempt to address in part this issue, we conducted a series of studies with different RCC models, in vitro and in nude mice, evaluating the activity of sunitinib, sorafenib and everolimus, either as single agents in front line or as sequential treatments, on tumour growth and on the expression and function of a panel of key effectors involved in cell proliferation and angiogenesis. We tested sunitinib, sorafenib and everolimus on RCC lines either VHL wild type or deficient. It has been reported that RCC cell lines express VEGFRs (Liu ; Ji ); therefore, an autocrine VEGF/VEGFRs loop may be activated. All the three agents resulted to be active in inhibiting cell proliferation, with a major effect observed after everolimus treatment (Figure 2A). In fact, even if both TKIs sunitinib and sorafenib were moderately effective in inhibiting proliferation of RCC cells in vitro at doses comparable to those used in our study (Juengel ; Liu ; Takeuchi ; Huang ; Kim ), the target of everolimus, mTOR, is directly implicated in tumour cell proliferation and survival via PI3K, Akt and MAPK. The doses of drugs chosen by us and in similar studies are expected to be in the therapeutic range achieved in human studies. Interestingly, the effect of everolimus was observed also in 786-O cells, which express only HIF-2 (Shinojima ). It has been demonstrated that HIF-2α translation is completely dependent upon the activity of TORC2 and not of TORC1 (Toschi ). Consistently, Cho reported that a dual PI3K/mTOR inhibitor, able to bind directly to the mTOR ATP-binding domain and thereby inhibiting both TORC1 and TORC2, decreased the level of HIF-2α, while rapamycin did not. However, we recently reported that everolimus is able to inhibit proliferation, signal transduction and VEGF production also in 786-O cells (Damiano ). Sunitinib and sorafenib had no or poor effect on Akt, p70S6K and MAPK phosphorylation and on VEGF secretion in conditioned media, while everolimus inhibited the mTOR-dependent effector p70S6K and reduced hVEGF levels (Figures 2B and C). This result does not conflict with the increase in VEGF levels found in clinical measurements after VEGFR inhibition. In fact, VEGF upregulation observed in patients treated with TKIs may be considered as a mechanism of resistance to these agents, as previously demonstrated (Casanovas ; Rini and Atkins, 2009). This event is not reasonably expected by measuring hVEGF levels in conditioned media of cells after a single, 24-h treatment with the drug.

We observed that 4 weeks in vitro pre-treatment with sunitinib may significantly affect the sensitivity to further treatment with sunitinib or sorafenib, while everolimus retains its antiproliferative activity (Figure 4). This may be due to the fact that these TKIs share a large part of the targets.

We translated this preliminary in vitro information in nude mice bearing established 786-O tumours and compared the activity of TKIs and everolimus in first line. All treatments were effective with a moderate advantage in tumour growth inhibition for everolimus over sunitinib and sorafenib. Analysis of tumour lysates at the end of treatment revealed, already at this early time point, an efficient inhibition of expression of pMAPK, mTOR-specific effector pp70S6K, HIF-1 and VEGF by everolimus, accompanied by over 50% inhibition of hVEGF secretion in the serum (Figure 3). In this study, we did not analyse the drugs effect on the levels of murine VEGF (mVEGF). Actually, the functional significance of murine, stromal VEGF on the growth of different tumours has been a controversial issue. However, we and others reported that VEGF production by tumour stroma has a modest role in tumour angiogenesis, with some variations among different tumours (Viloria-Petit ; Liang ; Damiano ). Moreover, the contribution of host stroma-derived VEGF on the growth of different tumours may be a confounding factor in the interpretation of the effects of various pharmacological blockers or genetic modulation of tumour VEGF expression.

To investigate the effect of the TKIs and everolimus in second line, after a 5-week treatment with sunitinib, when tumour growth was still apparently inhibited by this drug, we rechallenged the tumours with sunitinib (chronic treatment) or switched to either sorafenib or everolimus. Further treatment with sunitinib was ineffective since tumours rapidly increased their size and, at an earlier time point, analysis of tumour lysates showed no more ability to inhibit HIF-1 and VEGF expression accompanied by a marked increase in IL-8 in mice sera (Figure 5C). This finding is consistent with previous studies that have reported high tumour expression of IL-8 in sunitinib-resistant tumours (Huang ). Moreover, studies on polymorphisms also show a correlation between variant genotype of IL-8 and reduced progression-free survival in patients treated with TKI inhibitor pazopanib (Xu ). Treatment with sorafenib produced a short but relevant antitumour activity resulting in a statistically significant 2 weeks gain in survival when compared with mice rechallenged with sunitinib. This transient but significant inhibitory effect may be due to the different spectrum of kinases inhibited by sorafenib as compared with sunitinib. The sequence sunitinib>everolimus significantly delayed tumour growth with 2 weeks gain in survival as compared with sunitinib>sorafenib, and 4 weeks gain compared with mice treated with long-term sunitinib (P=0.0029; Figure 6B). These results are consistent with those obtained by Larkin in an orthotopic model of mouse RCC. The effect on tumour growth was anticipated by western blot analysis of tumour specimens, demonstrating that the sequence sunitinib>everolimus caused an almost total suppression of pp70S6K, HIF-1 and VEGF levels (Figure 6C) mirrored by a marked reduction in circulating hVEGF as measured by ELISAs (Figure 6D). Remarkably, both sequences, sunitinib>sorafenib and sunitinib>everolimus, produced a significant reduction in circulating IL-8 levels in mice serum (Figure 6E). Once again, IL-8 seems to be a general indicator of the sensitivity to TKIs and everolimus, while the lack of sustained inhibition of main angiogenic effectors HIF-1 and VEGF seems to be an early predictor of treatment failure. Interestingly, IL-8 overexpression has been recently associated also with resistance to antiangiogenic therapy in pancreatic cancer (Carbone ). These results are the biological counterpart of recent clinical studies with micro-bubble ultrasonography (DCE-US) showing that, when treatment with sunitinib is interrupted or suboptimal treatment schedule is used, increased tumour vascularisation occurs predicting treatment failure (Bjarnason . It was previously reported that the PI3K/Akt/mTOR pathway is linked to angiogenesis and resistance of tumour cells via Akt and HIF, due to their role in cell response to hypoxia and energy depletion (Faivre and Raymond, 2008; Dancey, 2010). In the same manner, in hepatocellular carcinoma models, it has been shown that the PI3K/Akt/mTOR pathway is involved in resistance to sorafenib (Chen ). Our results are in agreement with these observations, suggesting that lack of persistent inhibition of VEGF and HIF-1, two main inducers of vessel formation, may predict resistance to treatment and regrowth of tumour.

Our results also suggest that, following first-line sunitinib, everolimus more than sorafenib efficiently inhibits a series of critical effectors of tumour proliferation and angiogenesis involved also in the onset of resistance, causing a significant inhibition of tumour growth and increased survival of treated mice.

Recent data suggest that VEGFR-targeted therapies may still hold clinical activity in third line beyond an mTOR inhibitor (Di Lorenzo ; Grünwald ) and spur the notion that resistance to VEGFR inhibitors remains transient at least in a subgroup of patients. Several ongoing clinical trials take into account this observation and will test this hypothesis in a phase III setting.