RAS oncogenes are among the most frequently mutated genes in human cancer, but effective strategies for therapeutic inhibition of the RAS pathway have been elusive. Sprouty1 (SPRY1) is an upstream antagonist of RAS that is activated by extracellular signal-related kinase (ERK), providing a negative feedback loop for RAS signaling, and other evidence suggests that SPRY1 may have a tumor suppressor function. Studies of RAS status in the human childhood tumor rhabdomyosarcoma (RMS) indicated mutations in approximately half of the tumors of the embryonal rhabdomyosarcoma subtype (ERMS) but not the alveolar subtype (ARMS). ERMS tumors also showed overexpression of SPRY1, which was indeed upregulated by mutant RAS. However, we found that, in the presence of mutant RAS, the function of SPRY1 was changed from an antagonist to an agonist of RAS signaling. Thus, SPRY1 supported formation of activated ERK and mitogen-activated protein/ERK kinase and was essential for ERMS cell proliferation and survival. Conversely, silencing of SPRY1 in ERMS cells (but not ARMS cells) abolished their tumorigenicity in mice. Moreover, silencing of SPRY1 caused regression of established ERMS tumors (but not ARMS tumors) formed in xenograft settings. Our findings argue that SPRY1 inhibition can offer a therapeutic strategy to treat childhood RMS and possibly other tumors carrying oncogenic RAS mutations. Cancer Res; 70(2); 762–71
About one third of all human cancers are thought to carry a mutated RAS gene (1). Mutations are concentrated in codons 12, 13, and 61 of the three RAS family members (HRAS, KRAS, and NRAS) and result in activation of the RAS downstream pathway. Extensive research in this field has revealed much of the downstream signaling pathways that execute the programs initiated by mutant RAS, but therapeutic inhibition of the RAS pathway has not yet been successful (1, 2). This may indicate that much redundancy exists downstream of oncogenic RAS or vital targets have escaped our attention.
Sprouty proteins are among the downstream mediators of RAS signaling. The mammalian SPRY family has four members (SPRY1, SPRY2, SPRY3, and SPRY4), which differ in tissue distribution, activity, and interaction partners (3). Sprouty proteins antagonize RAS signaling during development and postnatal tissue growth and maintenance. SPRY expression is transcriptionally upregulated by increased extracellular signal-regulated kinase (ERK) activity, providing a negative feedback loop for RAS signaling (4). Decreased expression of SPRY genes in human cancer, including tumors from breast, prostate, and liver tissue, suggested a tumor suppressor function (5–7). Indeed, several studies showed that SPRYs, when overexpressed in cell lines, inhibit cell proliferation, migration, and anchorage-independent growth in vitro (8, 9) and limit tumor growth in vivo (10–12). The decreased SPRY expression in human cancer has been attributed to loss of heterozygosity or increased promoter hypermethylation (13, 14). However, SPRY genes are upregulated in some other cancers, suggesting alternative mechanisms at play. In some cases, gene mutation of SPRY targets may obliterate the tumor-suppressive effect of SPRY. For instance, SPRY2 has been shown to bind and inhibit wild-type, but not mutant, BRAF (15, 16). Consequently, melanoma cells, wild-type for BRAF, showed low SPRY2 levels, whereas melanoma cells with mutant BRAF had no decreased SPRY2 levels. Alternatively, phosphorylation of SPRY2 (i.e., by constitutive growth factor signaling) also prevents association with wild-type BRAF (15), allowing increased SPRY expression in these tumors. These latter studies describing elegant mechanisms of tumor escape further enforce the suggested tumor-suppressive role for SPRY proteins in diverse tissues.
Here, we have analyzed SPRY1 in rhabdomyosarcoma (RMS), a pediatric malignancy expressing markers of myogenic differentiation. We found that SPRY1 levels were consistently higher in the embryonal subtype (ERMS) compared with the alveolar subtype (ARMS). We show that the elevated SPRY1 expression in ERMS is caused by oncogenic RAS mutations. They are frequent (∼50%) in ERMS but absent in ARMS. We further show that in the presence of oncogenic RAS, the function of SPRY1 changed from an antagonist to an agonist of RAS/ERK signaling. In this context, SPRY1 is essential for proliferation and survival. RMS tumors grown in mice and harboring a RAS mutation completely regressed after SPRY1 silencing. Silencing SPRY1 in RMS tumors without RAS mutation had no effect.
Materials and Methods
Tumor specimen for microarray data and sequence analysis
Fresh tumor and normal skeletal muscle tissue were obtained and handled as previously described (17). The generation and processing of expression profile data is described elsewhere.1 Informed consent was obtained from all volunteers, from each patient, or from the legal guardian of each patient.
RMS cell lines (see Supplementary Data) were cultured in DMEM containing 10% FCS (Life Technologies) supplemented with glutamine and penicillin/streptomycin (Life Technologies). Stably transfected cell lines were selected in medium containing blasticidin (200 μg/mL) and zeocin (2.5 μg/mL). Cells were cultured in a 5% CO2 humidified atmosphere.
HRAS, KRAS, NRAS, BRAF, and SPRY1 were sequenced using BigDye Terminator v3.1 chemistry (Applied Biosystems). Sequencing was performed on an ABI 3100 capillary sequencer (Applied Biosystems). Forward and reverse sequences on electropherograms were analyzed using Codon Code aligner. Primer sequences and analysis of promoter methylation are described in Supplementary Data.
Generation of inducible cell lines
To generate RMS cells with doxycycline-inducible transgene expression, we used the Tet repressor system as previously described (18). Full-length human SPRY1 cDNA was cloned into pCDNA4/TO/Myc-HisA (Invitrogen). shSPRY1 oligonucleotides (Supplementary Table S1; Eurogentec Nederland b.v.) were cloned in the pTER vector followed by transfection into cells expressing the Tet repressor. To induce SPRY1 or shSPRY1 expression, 100 ng/mL doxycycline was added to the culture medium.
Transfection and transduction
For each transfection, 8 μg/dish of empty vector (pHAPAK1crib) or KRAS-G12D (KRASmt; based on NM_004985) were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. KRAS-G12D was cloned in BamHI and PstI of pHAPAK1crib. The lentiviral short hairpin RNA (shRNA) constructs (Supplementary Table S2) against SPRY1 and NRAS were derived from the TRC library (Sigma-Aldrich Chemie BV). Replication-defective viral particles were produced in HEK293T cells as described (19).
Cell proliferation and viability assay
The modified NIH/3T3 proliferation protocol was described previously (18). Cell viability was determined using the MTT assay as previously described (20). The mitogen-activated protein/ERK kinase (MEK) inhibitor PD98059 was from Sigma.
Western blot analysis
Cells/tissues were processed for Western blot analysis as described (21). Primary antibodies raised against hemagglutinin (HA) were obtained from Santa Cruz Biotechnology; ERK1/2, phosphorylated ERK (P-ERK) 1/2 (Thr202/Tyr204), MEK, phosphorylated MEK (P-MEK; Ser217/221), RAS, and poly(ADP-ribose) polymerase (PARP) antibodies were from Cell Signaling; and SPRY1 antibodies were from Abcam.
RMS cells were harvested by trypsin treatment and counted. Cells were suspended at 1 × 106/20 μL in PBS supplemented with 10% mouse serum and injected in the tibialis anterior muscles of 4- to 8-wk-old nonobese diabetic/severe combined immunodeficient (NOD/SCID) male mice (Charles River). To silence SPRY1, animals were i.p. injected with doxycycline (40 μg/mL in PBS) and kept on 200 μg/mL doxycycline in drinking water, which was refreshed weekly. Control animals were injected with sterile PBS and received regular drinking water ad libitum. In total, eight pairs (treated and untreated) of animals were included in the experiments. Animals were screened regularly to follow tumor formation and sacrificed before tumor reached 2 cm3.
Tumor formation was determined by measuring diameter (d) and tumor length (l). The average diameter of a healthy hind leg was 0.4 cm, so that the corrected tumor diameter (dc) was d − 0.4. Tumor volume (V in mm3) was calculated by V = (dc)2 * (l/2) adapted from Jacob and colleagues (22).
To investigate protein expression in the xenografts, tumors were resected at indicated time points after adding doxycycline, snap frozen, and homogenized in radioimmunoprecipitation assay buffer (described above) using an Ultra-Turrax. Samples for gel electrophoresis were prepared as described above. All animal experiments were conducted under the institutional guidelines and according to the law and approved in DAG101025 by the local animal ethics committee.
RAS mutations occur frequently in ERMS but are absent in ARMS
RAS genes are frequently mutated in RMS (6, 7), but it is unknown how these mutations are distributed over the ERMS and ARMS subtypes. We therefore sequenced HRAS, KRAS, and NRAS (codons 12, 13, and 61) and BRAF (codon 600) in 10 RMS cell lines and 23 primary RMS tumors. No RAS or BRAF mutations were found in 2 of 2 ARMS cell lines and in 10 of 10 tested primary ARMS samples (Table 1). In contrast, 3 of 8 (37.5%) ERMS cell lines and 6 of 13 (46%) primary ERMS samples harbored activating mutations in HRAS, KRAS, or NRAS. RAS mutations were therefore exclusively found in the ERMS subtype.
Elevated SPRY1 expression in ERMS is associated with hyperactive ERK signaling
We used gene expression profiles from ARMS and ERMS tumors (23) and cell lines to search for differential expression of RAS signaling–related genes and found that the levels of SPRY1 and its homologues SPRY2 and SPRY4, feedback inhibitors of ERK signaling (4, 24), were consistently higher in ERMS than in ARMS (Fig. 1A; Supplementary Fig. S1). We decided to study the role of SPRY1 in RMS.
SPRY1 is known as a tumor suppressor gene, and mutational inactivation might allow ERMS cells to tolerate elevated levels of SPRY1. Sequencing of the coding region of SPRY1 in 29 RMS cell lines and tumors identified a single mutation (Y29C) in only one ERMS sample (Table 1). The significance of this mutation is currently unknown and requires additional study, but this shows that mutational inactivation of SPRY1 in RMS is unlikely.
Western blot analyses detected SPRY1 protein in four of six ERMS cell lines but not or only at low levels in ARMS cell lines (Fig. 1B). Sequence analysis excluded promoter hypermethylation as a major cause of the low SPRY1 levels in ARMS (Supplementary Fig. S2). Alternatively, the differential SPRY1 expression could be explained by different levels of RAS/ERK signaling because it is known that SPRY proteins are induced by activated ERK (25). Therefore, we analyzed ERK phosphorylation in RMS cells. Figure 1B shows that elevated SPRY1 expression correlated with increased levels of P-ERK. To directly test the association between ERK activity and SPRY1 expression in RMS cells, we inhibited ERK activity in the ERMS cell line RD with the MEK inhibitor PD98059. Consequently, phosphorylation of ERK was abrogated and SPRY1 levels were reduced (Fig. 1C). These results suggest that elevated SPRY1 levels in ERMS are caused by activated ERK signaling. Interestingly, three of four ERMS cell lines (indicated with an asterisk in Fig. 1B) with elevated P-ERK and SPRY1 levels contain a mutated RAS gene (Table 1). In the fourth ERMS cell line with high P-ERK and SPRY1 levels, TE617T, we could not detect RAS or BRAF mutations.
SPRY1 levels are increased by oncogenic RAS
To further substantiate the correlation between mutant RAS, ERK signaling, and SPRY1 expression, we transiently transfected the ERMS cell line RUCH2 (wild-type for all three RAS genes and low P-ERK and SPRY1 levels) with a construct expressing constitutively active KRAS (KRASmt). Mutant KRAS not only significantly increased P-ERK levels but also increased SPRY1 levels (Fig. 2A).
Inversely, we applied lentivirally mediated shRNA to silence the mutated NRAS gene in the ERMS cell line RD. As a control, wild-type NRAS was also silenced in the ARMS cell line RH30. Two of the five shRNA constructs were very efficient in silencing NRAS expression (A3 and A4; Fig. 2B). In RD cells, this resulted in reduced P-ERK and SPRY1 levels. In RH30, there was no correlation between NRAS knockdown and SPRY1/P-ERK levels (Fig. 2B).
Furthermore, we used panRAS oligonucleotides to knock down all three members of the RAS family (HRAS, KRAS, and NRAS). Transfection of these oligonucleotides into RD cells resulted in potent NRAS knockdown and concomitant decrease of SPRY1 mRNA levels (Supplementary Fig. S3). Transfection into RH30 cells also reduced NRAS levels, but the low basal SPRY1 levels did not change. In summary, oncogenic RAS induces ERK activity and SPRY1 expression levels in ERMS cells.
SPRY1 expression is a RAS/ERK agonist in ERMS cells
Literature data have dubbed SPRY genes as tumor suppressor genes (4). We therefore asked which role SPRY1 exerts in RMS cells. We silenced endogenous SPRY1 expression in both RD and RH30 cells using four different lentivirally mediated shRNAs. In RH30 (ARMS cell line), silencing of SPRY1 had no effect on MEK or ERK phosphorylation (Fig. 3A). In contrast, in RD (ERMS cell line), SPRY1 knockdown resulted in an attenuation of both MEK and ERK phosphorylation (Fig. 3A). Similar results were found when SPRY1 was silenced in TE381T cells (Fig. 3B), another ERMS cell line harboring oncogenic RAS (Table 1). In ERMS cells with an activated RAS gene, SPRY1 expression is apparently required to maintain high P-MEK and P-ERK levels. These data were confirmed when SPRY1 was upregulated in the RD and RH30 cell lines. Doxycycline-induced overexpression of ectopic SPRY1 resulted in decreased ERK phosphorylation in RH30 cells, but it strongly increased ERK phosphorylation in RD cells (Fig. 3C). The latter effect was attenuated by preincubation with the MEK inhibitor PD98059, indicating that SPRY1 acts upstream of MEK. The potent effects of SPRY1 silencing on P-MEK levels (Fig. 3A) substantiate this conclusion. These data fully contradict a tumor suppressor role of SPRY1 in ERMS cells. To test whether the two RMS cell lines depend on RAS signaling, we have measured cell viability after incubation with increasing concentrations PD98059. RD cells seemed to be more sensitive to this MEK inhibitor than RH30 cells (Fig. 3D), showing that in RD ERMS cells RAS signaling is not only activated but the cells also depend on it.
SPRY1 is required for proliferation and survival of ERMS cells harboring oncogenic RAS
To further investigate the role of SPRY1 in RMS cells in vitro and in vivo, we constructed RD and RH30 cell clones (two clones per cell line) expressing a doxycycline-inducible shRNA specific for SPRY1 (RD-shSPRY1 and RH30-shSPRY1). The induction of shRNA resulted in both RD and RH30 cells in a strongly reduced SPRY1 expression (Fig. 4A). Strikingly, SPRY1 knockdown in vitro resulted in a virtually complete growth arrest in RD cells but only slightly reduced proliferation in RH30 cells (Fig. 4B and C). Also in TE381T cells, another ERMS cell line harboring oncogenic RAS (Table 1), SPRY1 silencing completely inhibited cell proliferation (Fig. 4C). However, silencing SPRY1 in ERMS RUCH2 cells (wild-type RAS) did not affect cell proliferation (Fig. 4C). The same results for all these cell lines were also found with a second doxycycline-inducible SPRY1 shRNA construct (Supplementary Fig. S4). Moreover, SPRY1 silencing by the four different lentivirally mediated shRNAs used in Fig. 3 also resulted in a strongly decreased viability in ERMS cell lines harboring oncogenic RAS (RD and TE381T) but not in ARMS or ERMS cells without oncogenic RAS (RH30 and Hs729T) or in VH10 primary human fibroblasts (Supplementary Fig. S5).
Fluorescence-activated cell sorting analyses showed that SPRY1 silencing in RD cells, but not in RH30 cells, resulted in a strong increase in the sub-G1 fraction, suggesting apoptotic cell death (Supplementary Fig. S4D). To show that the shSPRY1-mediated loss of cell survival was indeed due to SPRY1 depletion and not caused by off-target effects, we used our doxycycline-inducible SPRY1 RD cell line (see Fig. 3C). Silencing of SPRY1 in these cells by three independent lentiviral shRNAs resulted in a strong inhibition of cell proliferation, which could in all three cases be rescued by the concomitant doxycycline-induced expression of ectopic SPRY1 (Fig. 4D; Supplementary Fig. S6). Finally, to show that this SPRY1-specific effect on survival is indeed mediated at the level of or upstream from MEK, we transiently expressed constitutively active MEK1 (CA-MEK) in RD and RH30 cells before SPRY1 knockdown. In RD cells, CA-MEK partially rescued the SPRY1-dependent loss of cell survival (Supplementary Fig. S7A). In contrast, in RH30 cells, SPRY1 knockdown did not affect cell survival and no effect of ectopic CA-MEK expression was observed (data not shown). Western blot analysis showed that CA-MEK expression prevented attenuation of P-MEK following silencing of SPRY1, indicating that the effects of SPRY1 in RD ERMS cells are (partially) MEK dependent (Supplementary Fig. S7B). All these results show that RAS-mutated RMS cells not only strongly depend on MEK1 activity but that, in the context of oncogenic RAS, these cells require SPRY1 for proliferation and survival.
SPRY1 is essential for RMS formation and maintenance in vivo
Next, we analyzed whether SPRY1 was also necessary for in vivo tumorigenicity of the RD and RH30 tumor cell lines. We grafted the RH30-shSPRY1 and RD-shSPRY1 cells orthotopically in the tibialis anterior muscle of the hind leg of immunodeficient (NOD/SCID) mice. Each mouse received the RH30 cells and RD grafts in opposite legs. Xenografts formed in all control animals (n = 5, not treated with doxycycline) both at the RH30-shSPRY1 and at the RD-shSPRY1 injection sites (Fig. 5A). Both cell lines formed large tumors of small round blue cells within 6 weeks (Fig. 5A and B). A second group of mice was treated with doxycycline 1 week after grafting to silence SPRY1 expression in the xenografts. In the doxycycline-treated mice, RH30-shSPRY1 still formed tumors that were only slightly retarded in their growth (note that tumors in control mice were harvested 1 week before the tumors in the doxycycline-treated mice). Strikingly, RD-shSPRY1 cells completely failed to form tumors when mice were treated with doxycycline to silence SPRY1 (Fig. 5A). Histologic analysis of the xenografts from the doxycycline-treated animals showed a small patch of cells remaining at the site of injection of RD-shSPRY1 cells, surrounded by otherwise normal skeletal muscle (Fig. 5B, right, bottom row).
As human tumors are, at diagnosis, usually well established, treatment is more difficult. We therefore also analyzed the effect of SPRY1 silencing in established tumors. We injected RH30-shSPRY1 and RD-shSPRY1 as described above. However, in these experiments, doxycycline treatment was not initiated until clearly palpable tumors were formed (2–4 weeks after injection; note that RH30 tumors have a slightly slower take). Addition of doxycycline to the drinking water did not affect the growth kinetics of RH30 tumors (Fig. 5C). However, already a few days after adding doxycycline, RD tumors stopped growing and eventually completely regressed (Fig. 5C). The effect was observed in all five animals in the doxycycline group but not in any of the five control animals.
To verify in vivo the effects of SPRY1 silencing, we repeated the xenograft experiment with RH30-shSPRY1 and RD-shSPRY1 cells and resected tumors before complete regression. At 8 and 12 days of treatment with doxycycline, RD tumors were regressing but were still at ∼85% and ∼50% of their maximal tumor size, respectively. Western blot analyses clearly showed that adding doxycycline indeed induced SPRY1 silencing both in RH30 and in RD tumors. However, only in RD tumors, SPRY1 silencing resulted in reduced P-MEK and P-ERK levels, which is in line with the in vitro experiments (Fig. 3A). The decrease of P-ERK was attenuated at the 12-day time point possibly due to apoptotic cells (26). Indeed, PARP levels were strongly reduced in the RD tumors, suggesting massive apoptosis (Fig. 5D). This was not observed for the RH30 tumors.
In conclusion, our results show that SPRY1 expression in RD ERMS cells, which is increased by oncogenic RAS signaling, is required not only in vitro for cell proliferation and survival but also in vivo for formation and maintenance of the tumors.
Data presented in this article show that endogenous levels of SPRY1 seem to protect oncogenic RAS-positive ERMS cells from the induction of cell death. Silencing SPRY1 in these cells irreversibly and very potently inhibited cell proliferation and survival both in vitro and in vivo. Even well-established tumors grown in mice completely regressed when we silenced SPRY1 expression. Silencing SPRY1 in ERMS or ARMS cells or primary fibroblasts without RAS mutations had no or only minimal effects on proliferation.
SPRY proteins are transcriptionally upregulated by the activated RAS pathway, as we have also shown in this article. Most studies reporting on the function of SPRY proteins in normal and diseased tissue show an antagonistic role for SPRY proteins, including SPRY1 (27), in the RAS/ERK pathway. However, in a recent study, Pratilas and colleagues (28) showed that, in tumor cells with mutated BRAF, this negative feedback loop is disabled. Although the expression of SPRY or other RAS pathway feedback inhibitors, such as DUSP proteins, was elevated, they were no longer able to inhibit the RAS pathway. Brady and colleagues (15) showed that differences in, for instance, SPRY2 phosphorylation may play a role in this. However, it is unlikely that this is the case for SPRY1 in ERMS, as we did not detect any mobility shifts of SPRY1 compared with ARMS or other cell lines. Moreover, our data presented here show that SPRY1 in an oncogenic RAS background in fact supports ERK signaling and is even essential for proliferation and survival of cells. SPRY1 overexpression in RD ERMS cells harboring oncogenic NRAS increased P-ERK levels, whereas silencing SPRY1 in these cells decreased P-ERK levels (Fig. 3). In RMS cells without RAS mutations, SPRY1 still antagonized RAS/ERK signaling.
Activating RAS mutations occur frequently in RMS, but we show that they are only present in ERMS but not in ARMS tumors. About a third of all human cancers are thought to carry a mutated RAS gene (29). Elevated levels of SPRY1 or SPRY2 have indeed also been reported in non–small cell lung cancer and melanoma (12, 30), tumors that frequently harbor RAS or BRAF mutations. In line with our observations, Sutterlüty and colleagues (12) showed that induction of SPRY2 was also unable to antagonize RAS/ERK signaling in non–small cell lung cancer cells expressing oncogenic RAS. Whether the other SPRY homologues, which we also found to be upregulated in ERMS versus ARMS tumors (Supplementary Fig. S1), behave similarly as SPRY1 remains to be investigated. Recent data from Lito and colleagues (31, 32) showed that SPRY2 expression was required for xenograft formation of HRAS-V12–transformed fibroblasts. However, silencing SPRY2 in fibrosarcoma cells expressing oncogenic NRAS only slightly inhibited tumor formation (32). As SPRY2 was silenced in those experiments by stable expression of shRNAs in the cultured cell lines, selection for resistance to SPRY2 silencing might have played a role in those results. In our study, we used doxycycline-induced silencing, avoiding selection of resistant clones. The use of doxycycline-inducible shRNA allowed us to study the role of SPRY1 even in established xenografts from RMS cell lines.
The striking results we obtained by silencing SPRY1 in vivo in ERMS cells harboring oncogenic RAS (i.e., complete tumor regression) strongly suggest that a significant population of ERMS patients (∼50% of ERMS patients or ∼40% of all RMS patients) might benefit from therapeutic approaches (small-molecule inhibitors, antibodies) that specifically target either SPRY1 or other members of the RAS/ERK pathway, particularly MEK. The study by Pratilas and colleagues (28) showed that tumor cells with mutational activation of the RAS pathway were sensitive to MEK inhibitors, whereas cells activated by upstream receptor tyrosine kinases were insensitive. We show by multiple approaches that in ERMS cells with oncogenic RAS, SPRY1 supports survival at the level of MEK. Moreover, we show that RD ERMS cells were indeed also sensitive for the growth-inhibiting effects of the MEK1/2 inhibitor PD98059, whereas the RH30 ARMS cells were not. However, clinical use of PD98059 is limited due to its toxicity. Other MEK inhibitors have now been developed that show inhibition of tumor growth of, for example, thyroid xenografts harboring mutant RAF or RAS (33, 34). Clinical trials with these inhibitors or similar agents are ongoing (35, 36). Until now, SPRY1 has not been reported to antagonize oncogenic RAS function. The tumor-supporting function described in this study may not be restricted to RMS but may represent a common event in human cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Anne Vroon and Johan van Nes for critically reading the manuscript, Kees Fluiter for providing panRAS oligos, and Alvin Chan and Richard Volckmann for technical support.
Grant Support: Koningin Wilhelmina Fonds (Dutch Cancer Society fund) grant 2001-2558, Stichting Kindergeneeskundig Kankeronderzoek (Childhood Cancer Fund), Netherlands Organisation for Scientific Research Venigrant 016.056.165, and Ter Meulen Fund (Royal Netherlands Academy of Arts and Sciences).
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.
- Received July 7, 2009.
- Revision received November 4, 2009.
- Accepted November 6, 2009.
- ©2010 American Association for Cancer Research.