DDX31 Regulates the P53-HDM2 Pathway and Rrna Gene Transcription Through Its Interaction with NPM1 in Renal Cell Carcinomas

Published OnlineFirst September 27, 2012; DOI: 10.1158/0008-5472.CAN-12-1645

Cancer Molecular and Cellular Pathobiology Research

DDX31 Regulates the p53-HDM2 Pathway and rRNA Gene Transcription through Its Interaction with NPM1 in Renal Cell Carcinomas

Tomoya Fukawa1,2, Masaya Ono4, Taisuke Matsuo1, Hisanori Uehara3, Tsuneharu Miki5, Yusuke Nakamura6, Hiro-omi Kanayama2, and Toyomasa Katagiri1

Abstract Studies of renal cell carcinoma (RCC) have led to the development of new molecular-targeted drugs but its oncogenic origins remain poorly understood. Here, we report the identiﬁcation and critical roles in renal carcinogenesis for DDX31, a novel nucleolar protein upregulated in the vast majority of human RCC. Immu- nohistochemical overexpression of DDX31 was an independent prognostic factor for patients with RCC. RNA interference (RNAi)-mediated attenuation of DDX31 in RCC cells signiﬁcantly suppressed outgrowth, whereas ectopic DDX31 overexpression in human 293 kidney cells drove their proliferation. Endogenous DDX31 interacted and colocalized with nucleophosmin (NPM1) in the nucleoli of RCC cells, and attenuation of DDX31 or NPM1 expression decreased pre-ribosomal RNA biogenesis. Notably, in DDX31-attenuated cells, NPM1 was translocated from nucleoli to the nucleoplasm or cytoplasm where it bound to HDM2. As a result, HDM2 binding to p53 was

reduced, causing p53 stablization with concomitant G1 phase cell-cycle arrest and apoptosis. Taken together, our ﬁndings deﬁne a mechanism through which control of the DDX31–NPM1 complex is likely to play critical roles in renal carcinogenesis. Cancer Res; 72(22); 5867–77. 2012 AACR.

Introduction recommended as treatment options for metastatic RCC and – Renal cell carcinoma (RCC) is the most common cancer of the have been shown to prolong survival (10 13). However, such – kidney (1, 2), and up to 30% of patients with RCC present with treatments have also resulted in severe side effects (14 16). The metastatic disease (3, 4). At an early stage, RCC can be cured by prognostic factors that determine the treatment outcome in fi surgical resection, which is the most effective treatment for patients with metastatic RCC have been identi ed according to localized RCC tumors (5). Moreover, cytokine therapies, such as the status of the patient (17). These markers have been widely interleukin-2 or IFN-a, are widely used as a first-line treatment used to determine the best treatment approach and, as a result, for metastatic disease. However, these treatments can be asso- may improve survival. At present there are no new relevant fi ciated with severe toxicity (6–9). In addition to these anticancer molecular markers that have been identi ed to predict the drugs, several molecular target drugs, such as anti-VEGF anti- prognosis of RCC (18). Therefore, it is highly important to body (bevacizumab) and small-molecule inhibitors (sunitinib, develop a new molecular target agent(s) against RCC as well sorafenib, axitinib, temsirolimus, and everolimus) have been as prognostic molecular markers. The "omics" technologies, including transcriptomics, geno- mics, and proteomics, provide a detailed characterization of Authors' Affiliations: 1Division of Genome Medicine, Institute for Genome individual cancers that should help improve clinical strategies 2 Research, The University of Tokushima; Department of Urology and for neoplastic diseases through the development of novel 3Molecular and Environmental Pathology, Institute of Health Biosciences, The University of Tokushima Graduate School, Tokushima; 4Chemother- drugs, as well as providing the basis for personalized treatment apy Division and Cancer Proteomics Project, National Cancer Center (19). Hence, these technologies have helped identify the diag- Research Institute, Tokyo; 5Department of Urology, Kyoto Prefectural University of Medicine, Kyoto; and 6Laboratory of Molecular Medicine, nostic and therapeutic targets for cancer and understand the Human Genome Center, Institute of Medical Science, The University of detailed molecular mechanism of carcinogenesis. To this end, Tokyo, Tokyo, Japan. we analyzed the expression profiling of clear cell RCC (ccRCC), Note: Supplementary data for this article are available at Cancer Research which is a major histologic type of RCC (the Gene Expression Online (http://cancerres.aacrjournals.org/). Omnibus database accession number: GSE39364; ref. 20), and fi M. Ono and T. Matsuo contributed equally to this work. identi ed and characterized DEAD (Asp-Glu-Ala-Asp) box polypeptide 31 (DDX31), a novel member of the DEAD box Corresponding Author: Toyomasa Katagiri, Division of Genome Medicine, Institute for Genome Research, The University of Tokushima, 3-18-15, protein family. Kuramoto-cho, Tokushima 770-8503, Japan. Phone: 81-88-633-9477; DDX31 is the human homologue of Saccharomyces cerevisiae Fax: 81-88-633-7986; E-mail: [email protected] DEAD box protein 7 (Dbp7), which was originally identified in a doi: 10.1158/0008-5472.CAN-12-1645 Basic Local Alignment Search Tool (BLAST) search of the 2012 American Association for Cancer Research. Saccharomyces cerevesiae Genome Database (Stanford

www.aacrjournals.org 5867

Fukawa et al.

University, Stanford, CA) as a new ATP-dependent RNA heli- monoclonal antibody (3F10, Roche 11867423001; 1:1,000), anti- case (21). Dbp7 has been shown to play an important role in the FLAG M2 monoclonal antibody (M2, Sigma-Aldrich F3165; assembly of pre-ribosomal particles during the biogenesis of 1:5,000), b-actin (AC-15, Sigma-Aldrich A5441; 1:5,000), anti- the 60S ribosomal subunit (21). However, to date, no reports p53 (DO-7, Dako M7001; 1:200), anti-p53 (FL-393, Santa Cruz have characterized the biologic function of DDX31, the human sc-6243; 1:1,000), anti-p21 (Abcam, ab-7860; 1:200), anti-HDM2 homologue of Dbp7, and the signiﬁcance of its transactivation (2A10, Abcam, ab16895; 1:500), or anti-PARP (Cell Signaling in clinical RCC progression. Technology #9542; 1:200). Here, we identify and characterize DDX31 as a clinical prognostic factor and a potential therapeutic target, and Western blot analysis provide evidence that DDX31 is involved in regulating ribo- SDS-PAGE and immunoblotting were conducted as somal RNA (rRNA) gene transcription and the p53–HDM2 described previously (23). Full description for this analysis is pathway in renal carcinogenesis. described in Supplementary Materials and Methods.

Materials and Methods Immunohistochemical staining Cell lines and specimens Preparation of frozen sections (5 mm) of RCCs and normal ACHN, Caki-1, Caki-2, 769-P, HK-2, HEK293, and COS-7 kidney tissues and immunohistochemical staining were con- cells were obtained from the American Type Culture Collec- ducted as described previously (2), with anti-DDX31 antibody tion between 1994 and 2012. OS-RC-2 was provided by the (1:500), and classified the DDX31 expression levels on the RIKEN BioResource Center (Tsukuba, Japan) in 2005. KC-12, stained tissues based on a scoring method as follows; The KMRC-1, KMRC-20, and YCR-1 cells were kindly provided by percentage of DDX31-positive cancer cells was scored on a Dr. Taro Shuin (Department of Urology, Kochi Medical scale of 0to 2 (0: <5%, 1: 5–49%, and 2: >50%). Moreover, the School, Kochi University, Kochi, Japan) in 2006. All cells were intensity of DDX31 or NPM1 staining was semiquantitatively cultured under conditions recommended by their respective evaluated by 3 independent experienced pathologists without depositors. We monitored the cell morphology of these cell prior knowledge of clinicopathologic data based on the fol- lines by microscopy and confirmed that they had maintained lowing scoring system: 0: no staining, 1: low, 2: moderate, and 3: their morphologic states in comparison with the original high. Cases were accepted as positive only if all reviewers morphologic images. No mycoplasma contamination was independently defined them as such. In addition, the product detected in cultures of all of these cell lines using a Myco- of the DDX31 staining score was divided into the following 2 plasma Detection Kit (Roche) in 2011. Surgically resected groups: 3 and 4. RCC tissue samples and their corresponding clinical infor- mation were acquired from The University of Tokushima RNA interference experiments Hospital (Tokushima, Japan) after obtaining written We conducted RNA interference (RNAi) experiments as informed consent. This study, aswellastheuseofallclinical previously described (24, 25). The detailed Materials and materials described above, was approved by the Ethical Methods of this experiment are described in Supplementary Committee of The University of Tokushima. Materials and Methods.

Plasmids Establishment of stable transfectants Plasmids used in this study are described in Supplementary Mock (no insert) or pCAGGS-DDX31-HA expression vectors Materials and Methods. were transfected into HEK293 cells as previously described (20, 26). Following experiments are described in Supplemen- Quantitative and semiquantitative reverse transcription tary Materials and Methods. PCR The extraction of total RNA from cultured cells and RCC Measurement of pre-rRNA biogenesis clinical tissues, reverse transcription, and subsequent PCR The amount of 45S pre-ribosomal RNA (pre-rRNA) in total ampliﬁcation were conducted as previously described (20). RNA extracted was analyzed as described previously (27). Full b-Actin was used as a quantitative control. The sequences of description of this analysis is described in Supplementary each primer set are described in Supplementary Table S1. Materials and Methods.

Generation of a specific anti-DDX31 polyclonal antibody Immunocytochemical staining analysis The recombinant DDX31 protein was expressed and purified The methodology for this analysis is described in Supple- as previously reported (22) and subsequently inoculated into mentary Materials and Methods. rabbits (Scrum Inc.). The immune sera were then purified on antigen affinity columns using Affi-gel 10 gel (Bio-Rad). Immunoprecipitation and mass spectrometry DDX31-9 and Mock-3 stable transfectant cells were lysed in Antibodies 0.5% NP-40 IP-lysis buffer at 4C for 30 minutes. The detailed Antibodies used are described as follow; the anti-DDX31 Methods are described in Supplementary Materials and Meth- polyclonal antibody (see above; 1:100), anti-NPM1 monoclonal ods. Liquid chromatography/mass spectrometry and data antibody (FC-61991, Invitrogen 32-5200; 1:1,000), anti-HA rat acquisition were conducted as previously described (28).

5868 Cancer Res; 72(22) November 15, 2012 Cancer Research

DDX31 Regulates the p53 Activity in RCC Cells

Coimmunoprecipitation assay tionship between the clinical outcome and DDX31 expression Coimmunoprecipitation assay was conducted as described was assessed using the trend log-rank test. Cox proportional in Supplementary Materials and Methods. hazards analysis was used to identify significant prognostic clinical factors and to test for an independent contribution of p53 In vivo ubiquitination assay DDX31 expression to progression-free survival. To examine The methodology for this assay is described in Supplemen- potential confounding factors, age was treated as a continuous tary Materials and Methods. variable, and gender, nuclear grade, T stage, M stage, and the DDX31 expression score were treated as ordinal variables. Cell-cycle analysis Statistical significance was calculated using Student t test to We conducted flow cytometry with analysis gates to remove evaluate cell proliferation or gene expression. A difference of P debris and doubles after transfection of siDDX31 or siEGFP, as < 0.05 was considered statistically significant. previously described (23). The DNA content of 10,000 cells was analyzed by FACS Calibur and CellQuest software (BD Results Biosciences). Upregulation of DDX31 in renal cell carcinomas We verified that DDX31 was upregulated in 7 of 9 ccRCC Statistical analysis cases compared with the corresponding normal kidney, and in Progression-free survival curves were estimated using the 8 of 9 ccRCC cell lines compared with normal human kidney by Kaplan–Meier method. The statistical significance of a rela- quantitative RT-PCR (qRT-PCR; Fig. 1A and B). To investigate

Figure 1. DDX31 expression in RCCs. A and B, expression of the DDX31 gene in 9 RCC tissues (T), the adjacent normal kidney (N), and 9 RCC cell lines and a normal kidney as determined by quantitative RT-PCR (qRT-PCR). b-Actin served as a quantitative internal control. C, DDX31 protein expression in RCC cell lines and HK-2 cell line as determined by Western blot. W, p53 wild-type; M, p53 mutant. D, genomic structure of 3 DDX31 splicing variants (V1, V2, and V3). White and black triangles indicate the initiation and stop codons, respectively. Arrows indicate the primer set that detects each of the 3 variants. The number above each box indicates the exon number. E, expression pattern of each of the three DDX31 splicing variants in three RCC cell lines and a normal human kidney examined by semiquantitative RT-PCR.

www.aacrjournals.org Cancer Res; 72(22) November 15, 2012 5869

Fukawa et al.

DDX31 expression at the protein level, we generated an anti- the anti-DDX31 antibody, and classified the DDX31 expression DDX31 polyclonal antibody and confirmed that this antibody levels on the stained tissues based on a scoring method (see specifically recognized exogenously expressed DDX31 protein Materials and Methods). DDX31 was observed in the nuclei of in HEK293 cells as well as endogenous DDX31 in Caki-1 cells in many RCC specimens, although the staining patterns and Western blot analyses (Supplementary Fig. S1A). The Western intensities varied across individual cases (Fig. 2A), whereas blot results accorded well with those of the qRT-PCR results in no staining was detected in normal kidney tissues (Fig. 2B). most of the tested RCC cell lines and a normal human kidney Next, to investigate the clinicopathologic significance of epithelial cell line, HK-2 (Fig. 1C). Furthermore, northern blot DDX31 expression in RCC tissues, we analyzed the relationship analysis showed that the DDX31 transcript was weakly or between the calculated immunohistochemical DDX31 scores undetectably expressed in normal human organs (data not and the clinicopathologic variables of 70 ccRCC specimens shown). In the NCBI database, cDNA sequences corresponding (Supplementary Table S2). We used a Cox proportional to 3 transcript variants, denoted DDX31V1, DDX31V2, and hazards regression analysis to examine the association DDX31V3, consist of 2555, 2328, and 2337 nucleotides that between DDX31 expression and other clinical and pathologic encode 851, 775, and 778 amino acids, respectively. These 3 variables to predict progression-free survival. Both univariate transcriptional variants contain 20, 20, and 18 exons, respec- and multivariate analyses revealed that DDX31 expression was tively, as shown in Fig. 1D. The V2 variant has a unique exon 19, a significant and independent prognostic factor for patients which is 58 bp longer at the 50 end than exon 19 of V1 and V3, with RCC (P ¼ 0.0062 and P ¼ 0.014, respectively; HR: 1.8 per resulting in a frameshift within exon 19 and an early stop score; Table 1). A Kaplan–Meier survival analysis with the log- codon. The V3 variant lacks exons 17 and 18 and has a unique rank significance test also showed that high DDX31 expression exon 16, which is 35 bp shorter at the 50 end than exon 16 of V1 (score 4) was associated with a significantly worse progres- and V2 (Fig. 1D). Subsequent semiquantitative RT-PCR con- sion-free survival than those with low DDX31 protein expres- firmed that the DDX31V1 variant had the highest expression in sion (Score 3; Fig. 2C, log-rank, P ¼ 0.00002). These results ccRCC cells (Fig. 1E). Therefore, we focused on elucidating the suggested that DDX31 overexpression was indicative of worse biologic roles of the DDX31V1 transcript due to its predom- outcomes in RCC. inant expression in cancer cells. Effects of DDX31 expression on cell growth Association of higher DDX31 expression with poorer To investigate the growth promoting role of DDX31 in RCC prognosis cells, we knocked down the expression of endogenous DDX31 To investigate the detailed expression pattern of the DDX31 in Caki-1 (Fig. 3A and B) and KMRC-1 (Fig. 3C and D) using protein and characterize its biologic functions, we conducted RNAi techniques. qRT-PCR showed significant DDX31 knock- immunohistochemical staining of 70 ccRCC specimens with down in cells transfected with DDX31 shRNAs or siRNAs

Figure 2. Expression of DDX31 in RCC and normal kidney tissue sections. A and B, immunohistochemistry analysis of DDX31 in ccRCC and normal kidney tissue sections. Representative images of RCC tissues evaluated as score 2 (1109), score 4 (1077), and score 6 (993) as well as normal kidney tissues (sample no. 993 and 1109) are shown in A and B, respectively. Enlarged images of the rectangular region of each RCC tissue in A. Scale bar, 50 mm. C, Kaplan–Meier survival curves of progression-free survival in patients with RCC according to DDX31 expression. High DDX31 expression was signiﬁcantly associated with poor progression- free survival (log-rank test, P ¼ 0.00002).

5870 Cancer Res; 72(22) November 15, 2012 Cancer Research

DDX31 Regulates the p53 Activity in RCC Cells

Table 1. The association of DDX31 expression with other clinical and pathologic variables

Univariate Multivariate

Variables P HR (95% CI) P HR (95% CI) Age 0.61 1.02 (0.95–1.10) Gender 0.89 1.13 (0.218–5.83) Nuclear grade 0.036 3.51 (1.09–11.36) 0.13 3.53 (0.71–17.69) Score 0.0062 1.72 (1.17–2.52) 0.014 1.8 (1.13–2.87) T stage 0.041 2.45 (1.04–5.79) 0.16 1.98 (0.76–5.15) N stage Not calculated M stage 0.0008 16.13 (3.16–82.30) 0.029 7.40 (1.23–44.3)

NOTE: Cox proportional hazard regression analysis was used to test for prediction of the progression-free survival. Univariate analysis showed that nuclear grade, T stage, M stage, and DDX31 scores were signiﬁcant predictors of progression (P ¼ 0.036, 0.041, 0.0008, and 0.0062, respectively).

(#1 and #2), but not the controls (shEGFP or siEGFP). In NPM1 in KMRC-1 cells (Fig. 4B). Next, we examined the effects concordance with the knockdown effect, MTT assays clearly of knocking down NPM1 on the growth of RCC cell line KMRC- revealed significant growth suppression in the 2 cell lines by 1, and found that NPM1 depletion resulted in a significant both #1 and #2, compared with the controls (shEGFP or decrease in cell viability of KMRC-1 cells (Fig. 4C), although this siEGFP; Fig. 3B and D). In addition, a colony formation assay effect was not comparable with siDDX31. These findings also confirmed that introducing both shRNA-DDX31 con- suggest that NPM1 also has important roles in the growth of structs remarkably suppressed the growth of Caki-1 cells ccRCC cells. compared with shEGFP-transfected cells (shEGFP: sh#1 or ¼ fi sh#2 100%: 56% or 18%; Fig. 3E). To further con rm the DDX31 is involved in rRNA gene transcription DDX31 growth promoting effects of overexpression, we estab- Dbp7 (Dead-box protein 7), the yeast homologue of DDX31, lished HEK293 cell derivatives that stably express exogenous reportedly plays an important role in the assembly of preri- DDX31 (DDX31-1, -8, -9, and -10). Western blot analysis with an bosomal particles during the biogenesis of the 60S ribosomal fi anti-HA antibody con rmed the high expression of DDX31 subunit (21). In addition, NPM1 is also a nucleoprotein that has protein in these 4 clones (Fig. 3F). Subsequent MTT assays been implicated in ribosome biogenesis (27, 30). Particularly, a showed that all 4 DDX31-stable derivative cells (DDX31-1, -8, -9, previous report showed that NPM1 depletion reduced the fi and -10) grew signi cantly faster than the mock-transfected transcription rate of the ribosomal RNA (rRNA) gene (27). P ¼ controls (Mock-1, -2, and -3; Fig. 3G: 0.022). Together, these Therefore, to examine the effects of DDX31 or NPM1 knock- fi ndings strongly imply that DDX31 plays an important role in down on the transcription rate of rRNA, we monitored the cell proliferation of RCC. amount of pre-rRNA by qRT-PCR. DDX31 expression did not affect NPM1 knocking down in KMRC-1 cells (Fig. 4D), but DDX31 interacts with NPM1 knocking down of DDX31, NPM1, or both significantly reduced Because the biologic functions of DDX31 are unknown, we the expression of the pre-rRNA RNA45S subunits (Fig. 4D and searched for DDX31-interacting protein(s) using immunopre- E). Conversely, we confirmed that pre-rRNA synthesis was cipitation and 2-dimensional image converted analysis of significantly increased in DDX31 stable transfectants (Fig. liquid chromatography and mass spectrometry (2DICAL) anal- 4F). These findings suggest that the DDX31–NPM1 complex yses (28). Nucleophosmin (NPM1), a member of the nucleo- plays a crucial role in rRNA gene transcription. plasmin superfamily (29), was identified as a candidate DDX31- interacting protein. We first investigated the expression levels DDX31 regulates the p53 stabilization through its of NPM1 in RCCs by qRT-PCR, Western blot, and immuno- interaction with NPM1 in the nucleoli histochemistry analyses and found that NPM1 was highly It has been reported that the redistribution of NPM1 upon expressed in most of the RCC specimens and cell lines (Sup- DNA damage, such as UV damage, leads to an interaction with plementary Fig. S1B–S1D). To validate this interaction, HA- HDM2, and that this binding could affect p53 stabilization and tagged DDX31 (DDX31-HA) and FLAG-tagged NPM1 (NPM1- activity by inhibiting HDM2–p53 complex formation, indicat- FLAG) were co-transfected into COS-7 cells, and then the ing that NPM1 acts as a negative regulator of HDM2–p53 proteins were immunoprecipitated with an anti-FLAG or interaction (31). Moreover, we found that a reduced level anti-HA antibodies. Subsequent Western blot analysis with of pre-rRNA biogenesis (Fig. 4E) weakly correlated with the indicated antibodies showed that exogenous NPM1 co- decreased cell growth (Fig. 3) in DDX31- or NPM1-depleted precipitated with DDX31 (Fig. 4A). In addition, we confirmed RCC cells. Therefore, we hypothesized that DDX31 regulates that endogenous DDX31 co-precipitated with endogenous p53–HDM2 pathway through its interaction with NPM1 in RCC

www.aacrjournals.org Cancer Res; 72(22) November 15, 2012 5871

Fukawa et al.

Figure 3. Growth promoting effects of DDX31 in Caki-1 (A and B) and KMRC-1 (C and D). A and C, DDX31 knockdown by both vector-based shRNA (sh#1 and sh#2) and DDX31-speciﬁc siRNA oligonucleotides (si#1 and si#2) was validated by qRT-PCR. siRNA targeting EGFP (siEGFP) was used as a negative control. b-Actin was used as a quantitative control. B and D, MTT assay showing a decrease in cell number upon DDX31 knockdown in Caki-1 (B) and KMRC-1 (D) cells. E, colony formation assay showing decreased colony numbers in DDX31-depleted Caki-1 cells. F, Western blot analysis of cells expressing exogenous DDX31 (DDX31) or those transfected with a Mock vector. Endogenous (left) and exogenous DDX31 (right) expression was veriﬁed with the indicated antibodies. G, In vitro growth of HEK293–DDX31 and -Mock stable transfectant cells.

cells. We first examined the effects of knocking down DDX31 other hand, p21, a p53 downstream target, was clearly on the nuclear localization of NPM1 in Caki-1 cells by immu- upregulated at the protein and transcriptional levels at 72 nocytochemistry with anti-DDX31 or anti-NPM1 antibodies. hours after siDDX31 transfection (Fig. 5B, Supplementary We observed that DDX31 and NPM1 colocalized in the nucleoli Fig. S3B and S3C). Moreover, we confirmed that DDX31 of Caki-1 cells transfected with siEGFP as a control (Fig. 5A; depletion did not affect NPM1 protein expression levels (Fig. siEGFP), whereas translocation of endogenous NPM1 from the 5C). Subsequently, we investigated whether NPM1 could bind nucleoli to the nucleoplasm or cytoplasm was observed in to endogenous HDM2 when DDX31 expression was knocked DDX31-depleted cells (Fig. 5A; siDDX31). On the other hand, down by siRNA, and found that endogenous NPM1 and HDM2 NPM1 knockdown did not affect the nucleolar localization of formed a protein complex in DDX31-depleted cells, whereas DDX31 (Supplementary Fig. S2A). Moreover, translocation of the formation of p53–HDM2 complex was clearly decreased endogenous nucleolin, which is also an abundant nucleopro- (Fig. 5D). tein, from the nucleoli to the nucleoplasm or cytoplasm was Because HDM2 regulates p53 turnover via its E3 ligase not observed in DDX31-depleted KMRC-1 cells (Supplemen- activity, we hypothesized that DDX31 deficiency inhibits p53 tary Fig. S2B). These findings suggest the possibility that degradation by reduction of its HDM2-mediated ubiquitina- DDX31 specifically interacts with NPM1, and thereby inhibits tion. We transfected KMRC-1 cells with plasmids expressing its translocation from the nucleoli to the nucleoplasm or ubiquitin and p53 proteins and transfected these cells with cytoplasm in RCC cells. siDDX31 or siEGFP as a control. Subsequent immunoprecip- Next, we examined the expression level of p53 protein upon itation and immunoblotting analyses revealed that the p53 siRNA-mediated knockdown of DDX31. The expression of p53 polyubiquitination was clearly decreased in DDX31-depleted was clearly upregulated at the protein level upon DDX31 cells compared with that in siEGFP-transfected cells under the knockdown in both Caki-1 and KMRC-1 cells (Fig. 5B), but presence of MG132, which is a 26S proteasome inhibitor (Fig. the steady-state levels of p53 mRNA remained constant in cells 5E). These results show that DDX31 inhibition possibly pre- treated with DDX31-siRNA (Supplementary Fig. S3A). On the vents HDM2-mediated p53 polyubiquitination.

5872 Cancer Res; 72(22) November 15, 2012 Cancer Research

DDX31 Regulates the p53 Activity in RCC Cells

Figure 4. Interaction of DDX31 with NPM1. A, coimmunoprecipitation of DDX31 and NPM1 proteins. Cell lysates from COS-7 cells that were transfected with DDX31-HA and NPM1-FLAG were immunoprecipitated (IP) and subsequently immunoblotted (IB) with indicated tag antibodies. B, conﬁrmation of an endogenous interaction between DDX31 and NPM1 in KMRC-1 cells. C, growth inhibitory effects of NPM1 knockdown on RCC cells. qRT-PCR showing the suppression of endogenous NPM1 expression by NPM1 siRNA (left). MTT assay showing a decrease in cell number upon NPM1 knockdown in KMRC-1 cells (right). D, effects of DDX31, NPM1, or both knockdown on pre-rRNA synthesis in KMRC-1 cells. qRT-PCR showing suppression of endogenous DDX31 (left) and NPM1 expression (right) by DDX31, NPM1, and both siRNA, respectively. E, effects of DDX31, NPM1, or both knockdown on pre-rRNA expression in KMRC-1 cells. F, enhancement of pre-rRNA synthesis in HEK293 cells overexpressing DDX31.

DDX31 deficiency induced p53-dependent G1 arrest and manner (Fig. 6F). On the other hand, in p53-mutant KMRC-20 apoptosis cells, which do not express p53 protein due to a 5 bp insertion Next, to examine the effects of DDX31 ablation on the within exon 5, DDX31 knockdown did not affect p21 expression perturbation of cell cycle of RCC cells, we conducted fluores- level and growth reduction (Supplementary Fig. S4B and S4C) cence-activated cell sorting (FACS) analysis. A clear increased compared with that in p53-wildtype KMRC-1 cells (Fig. 6B, C). population of G0–G1 cells at 96 hours after siDDX31 transfec- Thus, DDX31 levels are relevant to the p53-mediated cellular tion (siEGFP: siDDX31 ¼ 69.71%: 79.01%) was observed (Fig. responses in RCC cells. 6A). We detected upregulation of p53 at protein level, but not at mRNA, whereas upregulation of p21 occurred at both protein and mRNA levels (Fig. 6B and C). We also observed an Discussion increased sub-G1 population from 96 hours (siEGFP: siDDX31 Through an expression profiling analysis of ccRCC, we ¼ 3.45%: 16.52%; Supplementary Fig. S4A) to 120 hours identified the upregulation of DDX31 in the majority of ccRCC (siEGFP: siDDX31 ¼ 8.52%: 28.54%; Fig. 6D). In addition, we cases but its minimal expression in normal human tissues detected a cleaved PARP at 120 hours (Fig. 6E), indicating an (data not shown). RNAi-mediated knockdown of DDX31 dras- induction of G1 arrest and subsequent apoptosis by DDX31 tically suppressed RCC cell growth, whereas introducing depletion. More importantly, siRNA-mediated p53 knockdown DDX31 into HEK293 cells enhanced cellular proliferation. reversed the effect of DDX31 silencing on cell growth, indicat- Moreover, clinicopathologic evidence through immunohis- ing that DDX31 deficiency led to apoptosis in a p53-depedent tochemistry using 70 ccRCC specimens showed that RCC cases

www.aacrjournals.org Cancer Res; 72(22) November 15, 2012 5873

Fukawa et al.

Figure 5. DDX31 regulates p53 stabilization through its interaction with NPM1. A, effects of DDX31 knockdown on the subcellular localization of NPM1. Forty- eight hours after transfection with siDDX31 or siEGFP, immunocytochemical staining was conducted to assess the localization of NPM1 in Caki-1 cells.B, effects of DDX31 knockdown on p53 and p21 protein levels in Caki-1 and KMRC-1. C, effects of DDX31 or NPM1 knockdown on HDM2, p53, and NPM1 expression. D, HDM2–NPM1 and HDM2–p53 interactions in DDX31-depleted KMRC-1 cells. E, KMRC-1 cells were transfected with the p53 expression vector. After 24 hours, cells were transfected with siDDX31 or siEGFP for 48 hours, followed by treatment with MG132 (20 mmol/L) for 3 hours before lysis. Lysates were immunoprecipitated with anti-p53 antibody (FL-393) and then immunoblotted with antibodies to p53 (DO-7) and ubiquitin (Ub). Relative polyubiquitination level was estimated by the intensity ratio of polyubiquitinated p53/total p53 after immunoprecipitation with anti-p53 antibody. Representative results are shown from one of 3 independent experiments.

strongly expressing DDX31 had shorter progression-free sur- Accumulating evidence indicates that NPM1 is implicated in vival periods than those with negative or weak DDX31 expres- multiple functions, including ribosomal biogenesis (28, 30), sion. These results strongly suggest that DDX31 overexpression centrosome duplication (33), and regulation of p53 stabiliza- is critically involved in renal carcinogenesis. tion (34–38). Therefore, we first examined whether depleting The DDX31 gene encodes a putative 851 amino acid protein DDX31 or NPM1 affects pre-rRNA synthesis and found that with a predicted ATP-dependent RNA helicase domain that is there was a significant reduction in cell growth in DDX31- or conserved among the DEAD-box protein family. It has been NPM1-depleted RCC cells, whereas a significant enhancement reported that Dbp7, the Saccharomyces cerevesiae homologue of pre-rRNA synthesis was observed in DDX31-overexpressed of DDX31 (21% identity and 72% similarity), localizes to the cells. These findings suggest that DDX31–NPM1 complex nucleolus and is required for optimal vegetative growth (21). formation in the nucleoli plays a crucial role in rRNA gene Dbp7 depletion results in reduced amounts of the 60S ribosome transcription in renal carcinogenesis. subunits and the accumulation of halfmer polysomes (21). Previous reports have indicated that cellular stress causes Thus, Dbp7 is likely involved in the assembly of pre-ribosomal NPM1 to undergo nucleoplasmic translocation and associate particles during the biogenesis of the 60S ribosomal subunit with HDM2, and the formation of this complex leads to the (21, 32), but the pathophysiologic functions of human DDX31 stabilization of the p53 protein (31). In addition, NPM1 also in RCC cells are unknown. Here, we showed that DDX31 regulates basal levels of the p53 protein in unstressed cells interacts and co-localizes with the nucleophosphoprotein through its interaction with HDM2 (31), indicating that NPM1 NPM1 in the nucleoli of RCC cells. Moreover, siRNA-mediated negatively regulates the p53–HDM2 interaction. Thus, the knockdown of DDX31 or NPM1 expression led to a significant nucleoli to nucleoplasmic translocalization of NPM1 resulted reduction in cell growth, suggesting that this protein complex in p53 activation after cellular or nucleolar stress, but it is plays an important role in the cell growth of RCC. unclear how NPM1 is anchored in the nucleoli of RCC cells,

5874 Cancer Res; 72(22) November 15, 2012 Cancer Research

DDX31 Regulates the p53 Activity in RCC Cells

Figure 6. DDX31 silencing induces p53-dependent G1 arrest and subsequent apoptosis. A, FACS profiles of KMRC-1 cells that were treated with siDDX31 or siEGFP and then stained with propidium iodide (PI) and analyzed by FACS analysis at 96 hours after transfection. B, Western blot analysis of p53 and p21 protein expression in DDX31-depleted KMRC-1 cells. C, qRT-PCR analysis of p53 (left) and p21 (right) expression at the transcriptional level in DDX31- depleted KMRC-1 cells. ns; not significant. D, FACS profiles of KMRC-1 cells that were treated with siDDX31 or siEGFP. Sub-G1 cells were measured by FACS analysis at 120 hours after transfection. E, Western blot analysis of caspase-specific PARP cleavage in DDX31-depleted KMRC-1 cells. F, KMRC-1 cells were cotransfected with siEGFP, siDDX31, and/or sip53 followed by MTT assay. G, schematic model of the regulation of the p53–HDM2 pathway and rRNA gene transcription by DDX31–NPM1. DDX31 depletion causes the failure of rRNA gene transcription, resulting in NPM1 translocation to nucleoplasm or cytoplasm. Subsequently, NPM1 binds to HDM2 and results in TP53 accumulation (depletion of DDX31). On the other hand, DDX31 plays an important role in rRNA gene transcription and regulates the p53–HDM2 pathway through its interaction with NPM1 in the nucleoli of RCC cells (overexpression of DDX31). which highly expresses DDX31, even after doxorubicin- translocation of NPM1 to the nucleoplasm of normal kidney induced cellular stress (Supplementary Fig. S5A). On the other cells. hand, in normal epithelial kidney cell HK-2, which does not Next, we investigated the effects of DDX31 depletion on the show DDX31 protein at detectable levels, NPM1 translocated p53–HDM2 pathway in RCC cells. Decreased rRNA gene tran- to the nucleoplasm with treatment of doxorubicin, but was scription caused by DDX31 depletion resulted in translocation localized in the nucleoli without doxorubicin (Supplementary of NPM1 from the nucleoli to the nucleoplasm, and NPM1 Fig. S5B). Nucleolar stress is often caused by disruption of bound HDM2, which prevented HDM2 from interacting with ribosomal biogenesis, which in turn can be caused by mal- p53 leading to p53 stabilization and p21 upregulation, p53- – function of nucleolar proteins (39 41) and chemotherapeutic dependent G1 cell-cycle arrest, and subsequent apoptosis (Fig. drugs such as doxorubicin (42). As shown in Fig. 4E, DDX31 6F, depletion of DDX31). In addition, we also observed NPM1 in depletion caused a reduction in rRNA gene transcription, the cytoplasm of DDX31-depleted RCC cells (Fig. 5A). Accu- which is a key step of ribosome biogenesis. Our results suggest mulating evidence indicates that NPM1 shuttles between the the possibility that nucleolar stress such as malfunction of nucleus and cytoplasm during the cell cycle (38, 43, 44) and rRNA gene transcription induced by DDX31 depletion resulted translocates to the cytosol and mitochodria in response to in translocation of the NPM1 from the nucleoli to the nucle- apoptotic stressors (45). Therefore, our findings suggest the oplasm or cytoplasm in DDX31-depleted RCC cells, whereas possibility that cytoplasmic NPM1 translocated from the nuclear stress such as doxorubicin treatment caused the nucleoli may influence the cell cycle or apoptosis in DDX31-

www.aacrjournals.org Cancer Res; 72(22) November 15, 2012 5875

Fukawa et al.

depleted RCC cells, although further analysis of the DDX31– Authors' Contributions NPM1 functions will be necessary. Together, we showed that Conception and design: T. Fukawa, T. Katagiri – Analysis and interpretation of data (e.g., statistical analysis, biostatistics, DDX31 captures NPM1 in the nucleoli, which prevents NPM1 computational analysis): T. Fukawa, M. Ono, T. Matsuo, H. Uehara, T. Katagiri HDM2 complex formation in the nucleoplasm and subsequent Writing, review, and/or revision of the manuscript: T. Fukawa, T. Katagiri Administrative, technical, or material support (i.e., reporting or orga- enhancement of p53 ubiquitination and degradation, resulting nizing data, constructing databases): T. Miki, H. Kanayama in p53 downregulation in RCC cells (Fig. 6F, overexpression of Study supervision: M. Ono, Y. Nakamura, T. Katagiri DDX31). To date, mutations in the p53 gene have only been rarely observed in RCC specimens and cell lines (46, 47). Hence, Acknowledgments our results may propose a new mechanism of p53 inactivation The authors thank Dr. Taro Shuin for kindly providing the RCC cell lines, Drs. Hirofumi Arakawa and Takeshi Imamura for kindly providing expression vectors, in RCC cells. Drs. Tetsuro Yoshimaru and Masato Komatsu for helpful discussions, Ayako In conclusion, our ﬁndings show that DDX31 likely plays a Ikarashi, National Cancer Center Research Institute, for 2DICAL analysis, and ﬁ – Drs. Junko Stevens and Hirofumi Aruga, Life Technologies, for the development signi cant role in renal carcinogenesis by regulating the p53 of Multiplex-quantitative Real Time-PCR system. HDM2 pathway and rRNA gene transcription via its interaction with NPM1. We believe that DDX31 could be a useful prog- Grant Support nostic marker for RCC and inhibiting DDX31–NPM1 complex This work was supported by future advanced research project from The formation is a promising therapeutic target for RCC, especially University of Tokushima (T. Katagiri). The costs of publication of this article were defrayed in part by the payment of in patients with P53 protein-intact RCC. page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Disclosure of Potential Conﬂicts of Interest Received April 27, 2012; revised August 16, 2012; accepted September 8, 2012; No potential conﬂicts of interest were disclosed. published OnlineFirst September 27, 2012.

References 1. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T, et al. Cancer and temsirolimus and their management in patients with metastatic statistics. CA Cancer J Clin 2008;58:71–96. renal cell carcinoma. Euro Urol 2008;53:917–30. 2. American Cancer Society: cancer facts and figures 2010. Atlanta, GA: 16. Bellmunt J, Szczylik C, Feingold J, Strahas A, Berkenblit A. Temsir- American Cancer Society, 2011. olimus safety profile and management of toxic effects in patients with 3. Lam JS, Leppert JT, Belldegrun AS, Figlin RA. Novel approaches in the advanced renal cell carcinoma and poor prognostic features. Ann therapy of metastatic renal cell carcinoma. World J Urol 2005;23: Oncol 2008;19:1287–392. 202–12. 17. Motzer RJ, Bacik J, Murphy BA, Russo P, Mazumdar M. Interferon- 4. Motzer RJ, Bander NH, Nanus DM. Renal-cell carcinoma. N Engl J Med alfa as a comparative treatment for clinical trials of new therapies 1996;335:865–75. against advanced renal cell carcinoma. J Clin Oncol 2002;20: 5. Cohen HT, McGovern FJ. Renal-cell carcinoma. N Engl J Med 289–96. 2005;353:2477–90. 18. Zhao H, Ljungberg B, Grankvist K, Rasmuson T, Tibshirani R, Brooks 6. Stadler WM, Huo D, George C, Yang X, Ryan CW, Karrison T, et al. JD. Gene expression profiling predicts survival in conventional renal Prognostic factors for survival with gemcitabine plus 5-fluorouracil cell carcinoma. PLoS Med 2006;3:115–24. based regimens for metastatic renal cancer. J Urol 2003;170:1141–5. 19. Petricoin EF III, Hackett JL, Lesko LJ, Puri RK, Gutman SI, Chumakov 7. Fyfe G, Fisher RI, Rosenberg SA, Sznol M, Parkinson DR, Louie AC. K, et al. Medical applications of microarray technologies: a regulatory Results of treatment of 255 patients with metastatic renal cell carci- science perspective. Nat Genet 2002;32:474–9. noma who received highdose recombinant interleukin-2 therapy. J Clin 20. Hirota E, Yan L, Tsunoda T, Ashida S, Fujime M, Shuin T, et al. Oncol 1995;13:688–96. Genome-wide gene expression profiles of clear cell renal cell carci- 8. Negrier S, Escudier B, Lasset C, Douillard JY, Savary J, Chevreau C, noma: Identification of molecular targets for treatment of renal cell et al. Recombinant human interleukin-2, recombinant human interfer- carcinoma. Int J Oncol 2006;29:799–827. on alfa-2a, or both in metastatic renal-cell carcinoma. N Engl J Med 21. Daugeron MC, Linder P. Dbp7p, a putative ATP-dependent RNA 1998;338:1272–8. helicase from Saccharomyces cerevisiae, is required for 60S ribosom- 9. Negrier S, Maral J, Drevon M, Vinke J, Escudier B, Phillip T. Long-term al subunit. RNA 1998;4:566–81. follow-up of patients with metastatic renal cell carcinoma treated with 22. Kanehira M, Harada Y, Takata R, Shuin T, Miki T, Fujioka T, et al. intravenous recombinant interleukin-2 in Europe. Cancer J Sci Am Involvement of upregulation of DEPDC1 (DEP domain containing 1) in 2000;6:93–8. bladder carcinogenesis. Oncogene 2007;26:6448–55. 10. Lane BR, Rini BI, Novick AC, Campbell SC. Targeted molecular 23. Park JH, Lin ML, Nishidate T, Nakamura Y, Katagiri T. PDZ-bindingki- therapy for renal cell carcinoma. Urology 2007;69:3–10. nase/T-LAK cell-originated protein kinase, a putative cancer/testis 11. Escudier B, Eisen T, Stadler WM, Szczylik C, Oudard S, Siebels M, et al. antigen with an oncogenic activity in breast cancer. Cancer Res Sorafenib in advanced clear-cell renal-cell carcinoma. N Engl J Med 2006;66:9186–95. 2007;356:125–34. 24. Dobashi S, Katagiri T, Hirota E, Ashida S, Daigo Y, Shuin T, et al. 12. Hudes G, Carducci M, Tomczak P, Dutcher J, Figlin R, Kapoor A, et al. Involvement of TMEM22 overexpression in the growth of renal cell Temsirolimus, interferon alfa, or both for advanced renal-cell carcino- carcinoma cells. Oncol Rep 2009;21:305–12. ma. N Engl J Med 2007;356:2271–81. 25. Endo A, Kitamura N, Komada M. Nucleophosmin/B23 regulates ubi- 13. Motzer RJ, Hutson TE, Tomczak PP, Michaelson MD, Bukowski RM, quitin dynamics in Nucleoli by recruiting deubiquitylating enzyme Rixe O, et al. Sunitinib versus interferon alfa in metastatic renal-cell USP36. J Biol Chem 2009;284:27918–23. carcinoma. N Engl J Med 2007;356:115–24. 26. Togashi A, Katagiri T, Ashida S, Fujioka T, Maruyama O, Wakumoto Y, 14. Wu S, Chen JJ, Kudelka A, Lu J, Zhu X. Incidence and risk of et al. Hypoxia-inducible protein 2 (HIG2), a novel diagnostic marker for hypertension with sorafenib in patients with cancer: a systematic renal cell carcinoma and potential target for molecular therapy. Cancer review and meta-analysis. Lancet Oncol 2008;9:117–23. Res 2005;65:4817–26. 15. Bhojani N, Jeldres C, Patard JJ, Perrotte P, Suardi N, Hutterer G, et al. 27. Murano K, Okuwaki M, Hisaoka M, Nagata K. Transcription regulation Toxicities associated with the administration of sorafenib, sunitinib, of the rRNA gene by a multifunctional nucleolar protein, B23/

5876 Cancer Res; 72(22) November 15, 2012 Cancer Research

DDX31 Regulates the p53 Activity in RCC Cells

Nucleophosmin, through Its histone chaperone activity. Mol Cell Biol 37. Li J, Sejas DP, Burma S, Chen DJ, Pang Q. Nucleophosmin suppresses 2008;28:3114–26. oncogene-induced apoptosis and senescence and enhances onco- 28. Ono M, Shitashige M, Honda K, Isobe T, Kuwabara H, Matsuzuki H, genic cooperation in cells with genomic instability. Carcinogenesis et al. Label-free quantitative proteomics using large peptide data sets 2007;28:1163–70. generated by nanoflow liquid chromatography and mass spectrome- 38. Grisendi S, Mecucci C, Falini B, Pandolfi PP. Nucleophosmin and try. Mol Cell Proteomics 2006;5:1338–47. cancer. Nat Rev Cancer 2006;6:493–505. 29. Frehlick LJ, Eirin-Lopez JM, Ausio J. New insights into the nucleo- 39. Fumagalli S, Di Cara A, Neb-Gulati A, Natt F, Schwemberger S, Hall J, phosmin/nucleoplasmin family of nuclear chaperones. Bioessays et al. al. Absence of nucleolar disruption after impairment of 40S 2007;29:49–59. ribosome biogenesis reveals an rpL11-translation-dependent mech- 30. Savkur RS, Olson MO. Preferential cleavage in pre-ribosomal RNA by anism of p53 induction. Nat Cell Biol 2009;11:501–8. protein B23 endoribonuclease. Nucleic Acids Res 1998;26:4508–15. 40. Zhang Y, Lu H. Signaling to p53: ribosomal proteins find their way. 31. Kurki S, Peltonen K, Latonen L, Kiviharju TM, Ojala PM, Meek D, et al. Cancer Cell 2009;16:369–77. Nucleolar protein NPM interacts with HDM2 and protects tumor 41. Warner JR, McIntosh KB. How common are extraribosomal functions suppressor protein p53 from HDM2-mediated degradation. Cancer of ribosomal proteins? Mol Cell 2009;34:3–11. Cell 2004;5:465–75. 42. Burger K, Muhl B, Harasim T, Rohrmoser M, Malamoussi A, Orban M, 32. Bernstein KA, Granneman S, Lee AV, Manickam S, Baserga SJ. et al. Chemotherapeutic drugs inhibit ribosome biogenesis at various Comprehensive mutational analysis of Yeast DEXD/H Box RNA heli- levels. J Biol Chem 2010;285:12416–25. cases involved in large ribosomal subunit biogenesis. Mol Cell Biol 43. Borer RA, Lehner CF, Eppenberger HM, Nigg E. Major nucleolarpro- 2006;26:1195–208. teins shuttle between nucleus and cytoplasm. Cell 1989;56:379–90. 33. Okuda M, Horn HF, Tarapore P, Tokuyama Y, Smulian AG, Chan PK, 44. Szebeni A, Herrera JE, Olson MO. Interaction of nucleolar protein B23 et al. Nucleophosmin/B23 is a target of CDK2/cyclin E in centrosome with peptides related to nuclear localization signals. Biochemistry duplication. Cell 2000;103:127–40. 1995;34:8037–42. 34. Wu MH, Chang JH, Chou CC, Yung BY. Involvement of nucleophos- 45. Lindenboim L, Blacher E, Borner C, Stein R. Regulation of stress- min/B23 in the response of HeLa cells to UV irradiation. Int J Cancer induced nuclear protein redistribution: a new function of Bax and Bak 2002;97:297–305. uncoupled from Bcl-x(L). Cell Death Differ 2010;17:346–59. 35. Yao Z, Duan S, Hou D, Wang W, Wang G, Liu Y, et al. B23 acts as a 46. Stickle NH, Cheng LS, Watson IR, Alon N, Malkin D, Irwin MS, et al. nucleolar stress sensor and promotes cell survival through its dynamic Expression of p53 in renal carcinoma cells is independent of pVHL. interaction with hnRNPU and hnRNPA1. Oncogene 2010;29:1821–34. Mutant Res 2005:578:23–32. 36. Li J, Zhang X, Sejas DP, Bagby GC, Pang Q. Hypoxia-induced 47. Gurova KV, Hill JE, Razorenova OV, Chumakov PM, Gudkov AV. p53 nucleophosmin protects cell death through inhibition of p53. J Biol pathway in renal cell carcinoma is repressed by a dominant mecha- Chem 2004;279:41275–9. nism. Cancer Res 2004;64:1951–8.

www.aacrjournals.org Cancer Res; 72(22) November 15, 2012 5877

DDX31 Regulates the p53-HDM2 Pathway and rRNA Gene Transcription through Its Interaction with NPM1 in Renal Cell Carcinomas

Tomoya Fukawa, Masaya Ono, Taisuke Matsuo, et al.

Cancer Res 2012;72:5867-5877. Published OnlineFirst September 27, 2012.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-12-1645

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2012/09/27/0008-5472.CAN-12-1645.DC1

Cited articles This article cites 46 articles, 12 of which you can access for free at: http://cancerres.aacrjournals.org/content/72/22/5867.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/72/22/5867.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/72/22/5867. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.