Identification of a MicroRNA Panel for Clear-cell Kidney Cancer

Identification of a MicroRNA Panel for Clear-cell Kidney Cancer Cancer David Juan, Gabriela Alexe, Travis Antes, Huiqing Liu, Anant Madabhushi, Charles Delisi, Shridhar Ganesan, Gyan Bhanot, and Louis S. Liou OBJECTIVES METHODS RESULTS CONCLUSIONS To identify a robust panel of microrna signatures that can classify tumor from normal kidney using microrna expression levels. Mounting evidence suggests that micrornas are key players in essential cellular processes and that their expression pattern can serve as diagnostic biomarkers for cancerous tissues. We selected 28 clear-cell type human renal cell carcinoma (ccrcc), samples from patientmatched specimens to perform high-throughput, quantitative real-time polymerase chain reaction analysis of microrna expression levels. The data were subjected to rigorous statistical analyses and hierarchical clustering to produce a discrete set of micrornas that can robustly distinguish ccrcc from their patient-matched normal kidney tissue samples with high confidence. Thirty-five micrornas were found that can robustly distinguish ccrcc from their patientmatched normal kidney tissue samples with high confidence. Among this set of 35 signature micrornas, 26 were found to be consistently downregulated and 9 consistently upregulated in ccrcc relative to normal kidney samples. Two micrornas, namely, MiR-155 and mir-21, commonly found to be upregulated in other cancers, and mir-210, induced by hypoxia, were also identified as overexpressed in ccrcc in our study. MicroRNAs identified as downregulated in our study can be correlated to common chromosome deletions in ccrcc. Our analysis is a comprehensive, statistically relevant study that identifies the micrornas dysregulated in ccrcc, which can serve as the basis of molecular markers for diagnosis. UROLOGY 75: 835 841, 2010. 2010 Elsevier Inc. Renal cell carcinoma (RCC) is the most common neoplasm in the adult kidney, accounting for 3% of all malignancies in the United States. This year, more than 50 000 men and women will be diagnosed with RCC and about 12 000 people will die because of this disease. 1 In recent years, studies on the biological mechanisms that contribute to clear-cell RCC (ccrcc) have focused on mutations to the genome, expression of protein coding genes, and epigenetic changes. Increasing evidence indicates, however, that dysregulation of a class of noncoding RNA genes, micrornas, is also David Juan and Gabriela Alexe are joint first authors. Gyan Bhanot and Louis S. Liou are joint corresponding authors. From the Department of Pathology, Boston University, Boston, Massachusetts; The Broad Institute of MIT and Harvard, Cambridge, Massachusetts; System Biosciences, Mountain View, California; BioMaPS Institute, Rutgers University, Piscataway, New Jersey; Departments of Molecular Biology and Biochemistry, and Physics, Rutgers University, Piscataway, New Jersey; Department of Biomedical Engineering, Boston University, Boston, Massachusetts; the Cancer Institute of New Jersey, New Brunswick, New Jersey; The Simons Center for Systems Biology, Institute for Advanced Study, Princeton, New Jersey; and Cambridge Health Alliance, Harvard Medical School, Cambridge, Massachusetts Reprint requests: Louis S. Liou, M.D., Ph.D., Department of Pathology, Boston University, 670 Albany Street Room 441, Boston, MA 02118. E-mail: Louis. liou@bmc.org Submitted: September 30, 2008, accepted (with revisions): October 19, 2009 associated with cancer and their expression profiles can be correlated with diseases pathogenesis and diagnosis. 2 Dysregulation of micrornas has been observed in many cancers, including solid tumors and hematological malignancies. 3 MicroRNAs are a class of naturally occurring, noncoding RNAs that regulate protein expression by targeting the messenger RNA of protein coding genes for either translation repression or transcriptional modulation. 3 They are a class of gene regulators that are endogenously produced to play important roles in a wide range of biological functions, including cellular differentiation, development, and apoptosis. 4 Connections between microrna expression and cancer have been made on several levels. Altered patterns of microrna expression have been shown to be associated with a variety of tumors. 2 Also, micrornas have been implicated to function as either putative tumor suppressors or oncogenes, and multiple micrornas are located within cancer-associated chromosomal fragile sites, which are susceptible to point mutation, amplification, deletion, or translocation. 5 Indeed, with micrornas estimated to regulate 30% of all gene transcripts, it is quite possible that their aberrant expression might contribute to 2010 Elsevier Inc. 0090-4295/10/$34.00 835 All Rights Reserved doi:10.1016/j.urology.2009.10.033

ccrcc formation by altering the balance in favor of oncogenes while inhibiting tumor suppressor genes. 6 In this report, we used a real-time quantitative polymerase chain reaction (qpcr) procedure designed to identify the global expression patterns of micrornas in ccrcc and their patient-matched normal kidney tissues. Our goal was to identify a robust panel of microrna signatures that can classify normal kidney from ccrcc using microrna expression levels. To accomplish this, we first applied a stringent analysis to find a suitable endogenous control, and then applied principal component analysis and consensus clustering analysis to identify a panel of 35 micrornas out of 384 profiled that can accurately (with high specificity and sensitivity) and robustly (independent of data perturbation and method of clustering) distinguish normal from ccrcc tissue samples. MATERIAL AND METHODS Tissue Specimens A total of 28 clear-cell RCC tissue specimens, along with their patient-matched normal kidney tissue, were obtained from patients at Boston Medical Center and Cleveland Clinic immediately after radical nephrectomy. Each specimen was then examined by a pathologist at the respective institution and histologically classified. Twelve of the specimens were classified as high-grade RCC (Fuhrman grade 3 or 4), whereas 16 of the specimens were classified as low-grade RCC (Fuhrman grade 1 or 2). Institutional Review Board approved informed consent for the collection of specimens was obtained from all patients. These specimens were then put into 5 volumes of RNA preservatives, set at 4 C for 1 hour, and stored at 80 C until RNA extraction. RNA Extraction and cdna Synthesis Total RNA was extracted by homogenizing approximately 0.2 g of frozen tissue with a Powergen 35 tissue homogenizer (Fisher scientific, Pittsburgh, PA), followed by isolation using the mir- VANA mirna Isolation Kit (Ambion, Austin, TX). The purified RNA was subsequently assessed for concentration using a Nanodrop ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, DE). Two micrograms of the purified RNA were reverse-transcribed into first-strand cdna using the QuantiMir RT Kit (System Biosciences, Mountain View, CA), which converts all small RNAs simultaneously into detectable cdnas for real-time PCR. Quantitative Real-time PCR MicroRNA expression profiling was performed using real-time PCR in either a 96-well using the cancer microrna qpcr Array (System Biosciences), 384-well format using the HT384 microrna Profiler Array (System Biosciences), or in a Custom 60 microrna array panel using a 384-well reaction plate. Each real-time PCR reaction in the 96-well format contained 5.0 ng of synthesized cdna, 300 nm Universal reverse primer (System Biosciences), and 500 nm microrna-specific forward assay primer in a total reaction volume of 20 L. For the 384-well real-time PCR reactions, the volume was performed at 5 L while keeping the reagent ratios the same. The sequence of the microrna-specific forward primer was identical to the target microrna being measured. All reactions were subjected to standard real-time PCR thermocycling conditions and performed in triplicate using an ABI 7900HT Sequence Detection System with PowerSYBR reagents (Applied Biosystems, Foster City, CA). Expression levels of the micrornas were measured using the comparative ct method and normalized to an endogenous control. Of the 384 microrna measurements on the HT 384 microrna profiler, 12 micrornas were undetectable and 372 micrornas provided reliable measurements across all the patient RNA samples. The raw datasets produced on the cancer microrna qpcr 96-well array as well as the HT384 microrna Profiler array are presented in Supplementary Tables S1 and S2 (published online), respectively. Supplementary Tables S3 and S4 (published online) list the relative fold changes in microrna expression for each sample profiled on the cancer microrna qpcr 96-well array and the HT384 microrna Profiler array, respectively. Endogenous MicroRNA Reference Control For relative quantification, the normalization of each microrna expression is defined relative to endogenous control (housekeeping) microrna with respect to each sample, by dividing the raw microrna intensity value to the value of the endogenous control. Ten samples profiled on the 96-well cancer microrna qpcr Array were used to identify the endogenous control. The variation in each microrna for all tumor (t) and normal/reference (n) samples was estimated using scores and permutation P values for several tests: Mann Whitney (Wilcoxon rank sum for paired T/N samples), Kolmogorov Smirnov, F test, variance, and rank invariant test. 7 These scores were normalized and combined into an average global score, with the lowest score determining the endogenous control. Statistical Analysis Data from the comparative ct method were evaluated using the signal-to-noise ratio (SNR) test to identify the micrornas able to distinguish between tumor and normal phenotypes. The P values for the SNR scores were determined using 1000 permutation tests with multiple hypothesis correction using the False Discovery Rate (FDR) and q values. 8 We then used principal component analysis to further reduce the dimensionality of the data. Furthermore, to evaluate the predictive power of the micrornas, several classification models, including Weighted Voting (WV) and k-nearest Neighbor (knn), were applied to the above identified markers to assess their predictive accuracy based on the leave-one-out (LOO) cross validation test. These tests were conducted using the class prediction module of BRB Tools (http://linus.nci.nih.gov/brb-arraytools.html). RESULTS We investigated the microrna expression profile in ccrcc by initially profiling 10 patient-matched samples (patients 9-18) for 384 micrornas using real-time PCR. This initial survey detected 86 micrornas with significant fold changes in expression level, of which 84 were found to be downregulated ( 1.2- to 50-fold), whereas 2 micrornas were found to be upregulated (mir592 2.27, mir210 6.23-fold). 836 UROLOGY 75 (4), 2010

Table 1. A description of the samples used in the study and the platforms they were analyzed on Number Patient ID Grade Array Objective Fuhrman 1 YR00-11 3 96-Well cancer qpcr array Endogenous control & profile establishment 2 YR00-17 4 96-Well cancer qpcr array Endogenous control & profile establishment 3 YR01-19 2 96-Well cancer qpcr array Endogenous control & profile establishment 4 YR01-37 3 96-Well cancer qpcr array Endogenous control & profile establishment 5 YR05-27 3 96-Well cancer qpcr array Endogenous control & profile establishment 6 YR06-29 3 96-Well cancer qpcr array Endogenous control & profile establishment 7 YR06-37 2 96-Well cancer qpcr array Endogenous control & profile establishment 8 YR00-11 2 96-Well cancer qpcr array Endogenous control & profile establishment 9 YR07-79 2 96-Well array & HT384 microrna profiler Endogenous control & profile establishment 10 YR01-21 3 96-Well array & HT384 microrna profiler Endogenous control & profile establishment 11 YR04-002 2 HT384 microrna profiler Profile establishment 12 YR05-34 4 HT384 microrna profiler Profile establishment 13 YR06-38 2 HT384 microrna profiler Profile establishment 14 YR06-119 2 HT384 microrna profiler Profile establishment 15 YR07-002 2 HT384 microrna profiler Profile establishment 16 YR07-004 2 HT384 microrna profiler Profile establishment 17 YR07-24 4 HT384 microrna profiler Profile establishment 18 YR06-31 3 HT384 microrna profiler Profile establishment 19 YR00-02 3 Rencamir panel Refinement of profile 20 YR00-18 3 Rencamir panel Refinement of profile 21 YR00-20 3 Rencamir panel Refinement of profile 22 YR01-02 2 Rencamir panel Refinement of profile 23 YR01-05 1 Rencamir panel Refinement of profile 24 YR01-06 2 Rencamir panel Refinement of profile 25 YR01-23 1 Rencamir panel Refinement of profile 26 YR01-78 1 Rencamir panel Refinement of profile 27 YR01-84 2 Rencamir panel Refinement of profile 28 YR05-14 2 Rencamir panel Refinement of profile The first set of 10 matching normal/tumor samples were analyzed for 96 microrna expression levels to identify an endogenous control and a preliminary discriminatory panel. Two of these sample pairs and a set of 8 new sample pairs were analyzed in a 384-well format to validate and establish the panel. Finally, a new set of 10 samples pairs were analyzed on the Rencamir panel, which included the best set of microrna from first 2 experiments augmented with a set chosen from published data. The final step resulted in a list of 35 micrornas that are our final optimal set of discriminatory micrornas, which can distinguish between tumor and normal with perfect (100%) accuracy in leave-one-out validation. To further refine the microrna test set, we investigated a select panel of 60 micrornas. This panel consisted of micrornas that we found dysregulated from the HT384 microrna Profiler array experiment, 96-well cancer microrna qpcr Array experiment, and a literature search. The 60 microrna panel was then tested on 10 additional ccrcc patients (patients 19-28). What emerged was a robust ccrcc microrna signature panel consisting of 35 micrornas that can discriminate accurately between ccrcc and normal patient-matched tissue. Details of the patient samples used for each part of the profiling are given in Table 1. We used distinct patient cohorts for panel identification and refinement (training and test samples). Identification of the Endogenous Control Statistical analysis of the dataset generated from the 96-well cancer microrna qpcr Array identified mir- 106 b as having the least variation in expression between ccrcc and patient-matched normal tissue across all 10 samples. The complete analysis, which describes why we chose this endogenous control, is presented in Supplementary Table S5 (published online). MicroRNA Expression Patterns Stratify ccrcc From Normal Kidney We initially profiled a set of 384 human micrornas in 10 ccrcc and their paired normal kidney tissues to identify differences in microrna expression patterns. The SNR test was applied to rank the micrornas with respect to the Tumor vs Normal class membership. Of the 372 detectable micrornas from the HT384 qpcr array data, 86 micrornas were found to be at a significance level of P.01 and FDR 0.05, with 2 micrornas consistently upregulated and 84 micrornas downregulated in ccrcc. The entire dataset is provided in Supplementary Table S6 (published online). We then used principal component analysis on the full set of microrna expression data and found that only a few ( 10) principal components were needed to separate tumors from normal. By looking at the coefficients of the eigenvectors of the principal components with largest eigenvalues, we identified the micrornas responsible for 85% of the variation in the data distinguishing tumors from normal. These had excellent overlap with the microrna identified by SNR above. Figure 1 shows the projection of the tumor and normal samples on the first 2 principal components. UROLOGY 75 (4), 2010 837

Figure 1. Projection of the Tumor and Normal log 2 quantified and outlier corrected expression ratios on the first 2 principal components using Mir-106 b as baseline for 10 tumor and 10 normal samples in 3 replicates. These 2 principal components represent 85% of the variation in the data. Identification of a MicroRNA Signature in ccrcc To identify a focused but robust panel of micrornas to distinguish normal kidney tissue from ccrcc, we selected 60 micrornas for analysis. This collection was an overlap of the top 36 microrna markers identified in the previous 384 microrna dataset, 8 micrornas identified in the 96 microrna dataset, and 10 micrornas identified in the literature as potentially relevant to ccrcc, but not identified by initial experimental analysis. We added the following set of literature curated micrornas, which were believed to target genes involved in kidney cancer: mir-195, mir-768-5p, mir-584, mir-371, mir-212, mir-449, mir-421, mir-185, mir-34a, and mir-34b (Supplementary Table S10 [published online]). The custom panel contained mir-199 (a b) twice as a check for noise level and also 2 additional controls, snorna 43 (RNU43), and mir-106 b. Details and references of the collection of micrornas are presented in Supplementary Table S10 (published online). The ability of this 60-microRNA panel to differentiate ccrcc from normal kidney was then tested on additional samples of ccrcc and normal adjacent kidney from 10 new patients (patient 19-28). The raw dataset and the results after applying the comparative ct method are listed in Supplementary Tables S7 and S8 (published online), respectively. Supervised analysis using the SNR test identified 35 micrornas (26 downregulated and 9 upregulated micrornas) as an optimal set to discriminate normal kidney from ccrcc. These are listed in Supplementary Table S9 (published online) and correspond to a significance of P.05 and FDR 0.05. We call this microrna panel RencaMir. Three of the literature curetted microrna (mir-34a, mir-34b, mir-185) made our final panel. The fold change in these 35 micrornas is depicted in Figure 2 and a composite heat map of all the measurements (normal tumor) for the expression of these 35 micrornas is shown in Figure 3. Again, the triplicate samples cluster together at the bottom of the sample dendrogram showing minimal experimental noise. We applied several classification methods to refine the microrna in our RencaMir panel using 10 tumor samples and 10 reference samples created on the Custom 60 microrna qpcr array. LOO cross validation was performed with the classifiers constructed over a grid of significance levels (t test) for microrna selection based on the most discriminative individual microrna markers. At 0.5% significance level, we achieved 100% accuracy from Diagonal Linear Discriminant predictor and 1-Nearest Neighbor, Compound covariate Predictor, 3-Nearest Neighbor, and Nearest Centroid and Support Vector Machine. The following 10 micrornas were consistently selected by each of the 4 classifiers in all LOO experiments: mir-200c, mir-185, mir-34a, mir-142-3p, mir-21, mir-155, mir-224, mir-210, and mir-592. COMMENT In this study, we set out to identify whether microrna expression signatures could be used to distinguish ccrcc from normal kidney tissue in patient-matched samples. Our first task established microrna-106 b as the best endogenous control and all subsequent measurements were normalized to this reference. Although published data on microrna-106 b have identified it as overexpressed in several types of cancers and functionally linked to proliferation and antiapoptotic pathways, 9 there is also evidence that this microrna is invariant in many tissues. 10 Previous studies have established the use of oncogenic micrornas as endogenous controls including let-7a and microrna-16 in breast cancer. 11 Therefore, we are confident in our analysis that establishes microrna-106 b as a suitable ccrcc endogenous control. However, the possibility exists that the normal tissue used in our paired normal-tumor sample set could harbor alterations in the microrna-106-b regulation as a first step in the process of malignant transformation. A comparison between the tumor samples and normal tissue from nonmalignant kidney specimens of nephrectomies (nonmatched) could resolve this issue, although it is beyond the scope of this article. We then selected 10 patients with various grades of ccrcc and profiled 384 micrornas in triplicate using qpcr for each tumor and normal RNA sample, for a total 22 980 individual real-time qpcr data points. From this initial study, we discovered a subset of micrornas that can robustly and accurately distinguish between normal and tumor in ccrcc patients. We then independently tested the performance of this microrna set on samples from an additional 10 patients, for a total of 20 unique ccrcc patients, to better define a more discrete set of differentially expressed micrornas. What emerged from our studies is a robust set of 35 micrornas that will be used for future validation in larger numbers of ccrcc patients. Within this set of 35 838 UROLOGY 75 (4), 2010

Figure 2. Graphical representation of the average fold change levels for the 9 upregulated and 26 downregulated microrna found across 10 ccrcc samples, with mir-106 b as the endogenous control. Figure 3. Heat map of the 35 micrornas that can be used to distinguish between ccrcc and patient-matched normal kidney tissue. This panel of 35 micrornas corresponds to 9 upregulated and 26 downregulated micrornas found through SNR test using microrna 106 b as the endogenous control. Red signifies upregulation while green signifies downregulation. Note that replicate samples cluster together, showing that measurement noise is well controlled (Color version available online). ccrcc signature micrornas, 26 were downregulated and 9 were upregulated (Fig. 2). This is consistent with other findings that numerous micrornas in tumor samples are expressed at lower levels than in normal tissues. 2 To date, we know of one other study 12 that demonstrated differential microrna expression patterns in RCC. In UROLOGY 75 (4), 2010 839

this study, Gottardo et al 12 reported 4 micrornas (mir- 28, mir-185, mir-27, let-7f-2) significantly upregulated in RCC. Our data show agreement with the increased expression of mir-185 in RCC. Here, we present a thorough analysis using a larger patient-matched tissue sampling population along with a more comprehensive microrna assay set measured, (384 in our study vs 161 human micrornas in their study), which reveal additional dysregulated micrornas. Four of the upregulated micrornas identified in our study (mir-21, mir-155, mir-34a, mir-210) have strong correlations with other tumorigenic states and are commonly dysregulated in other solid tumors. For example, the upregulation of mir-21 has been documented in at least 6 solid tumors and linked to a variety of cancerrelated processes including proliferation, invasion, and metastasis. 13 MiR-155 is also overexpressed in cancers of the breast, colon, lung, pancreas, and those of hematologic origin. 13-15 Thus, our study provides another example of the apparent connection that these 2 commonly upregulated micrornas are associated with carcinogenesis in general. In addition to these 2 micrornas, we have identified an additional and unique set of micror- NAs that are dysregulated in ccrcc malignancy, suggesting that this set of micrornas could contribute to the pathogenesis of ccrcc. Two other upregulated micrornas of great interest in ccrcc identified in our study are mir-34a and mir-210. Recently, overexpression of mir-34a was associated with cell proliferation in an oxidative stress induced rat renal carcinogenesis model, 16 while mir-210 was associated with oxidative stress and hypoxia. 17 Our study reveals that mir-210 is overexpressed in ccrcc at 10-fold higher levels compared with normal kidney. We believe there could be a correlation between the overexpression of mir-210 and common genomic alterations already characterized in ccrcc. It has been well documented that most ccrcc cases are associated with Von Hippel- Lindau (VHL) gene deletions and mutations. VHL is a known regulator of the transcription factor HIF-1A and in cases where VHL is missing or mutated, elevated levels of HIF-1A have been identified. 18 Importantly, HIF-1A has been demonstrated to induce the expression of mir- 210 through direct binding to the promoter region of mir-210. 17 Thus, our study supports the growing evidence that hypoxic conditions can induce mir-210 through HIF-1A mediated transactivation and we provide a further connection to human ccrcc. The overall decrease in microrna expression is a common theme in cancer. Sites of chromosomal deletions in tumors are often associated with the loss of tumor suppressor genes, 19 and key micrornas downregulated in our study map to known sites of chromosomal deletions associated with ccrcc. The most common deletion known to be involved in renal tumor development is located at 3p21, 20 where the RAS association Family 1 gene (RASSF1A) has been postulated to be the candidate tumor suppressor gene deleted and associated with ccrcc. 21 Also, located within this 3p21 locus is the gene for mir-135a. A potential tumor suppressor role for mir-135a may be in its putative regulatory binding site within the 3=-untranslated region of HIF-1A. 22 Our study shows a consistent downregulation of mir-135a in ccrcc, potentially alleviating the translational repression of and allowing for higher expression of HIF-1A. Five additional micrornas (mir-136, mir-154, mir- 337, mir-377, mir-411), also identified to be downregulated in our study, map to 14q32, which is another common deletion site observed in ccrcc. 23 Deletions within this microrna gene cluster region have also been found in human epithelial ovarian cancer 24 and bladder cancer, 25 suggesting a common loss of tumor suppressor microrna genes in at least 3 genitourinary organ tumors. CONCLUSIONS In summary, this study contributes to the growing understanding of the role that micrornas play in cancer and describes the global expression patterns of micrornas in patient-matched ccrcc and normal kidney tissues. Both the micrornas identified and the signaling pathways they regulate may be important therapeutic targets in human ccrcc. However, further studies are needed in larger samples of ccrcc and also in other malignant and benign kidney tumors. References 1. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2007. CA Cancer J Clin. 2007;57:43-66. 2. Lu J, Getz G, Miska E, et al. MicroRNA expression profiles classify human cancers. Nature. 2005;435:834-838. 3. Cummins J, Velculescu V. Implications of micro-rna profiling for cancer diagnosis. Oncogene. 2006;25:6220-6227. 4. Alvarez-Garcia I, Miska E. MicroRNA functions in animal development and human disease. Development. 2005;132:4653-4662. 5. Calin G, Sevignani C, Dumitru C, et al. Human microrna genes are frequently located at fragile sites and genomic regions involved in cancers. Proc Natl Acad Sci USA. 2004;101:2999-3004. 6. Chen C. MicroRNAs as oncogenes and tumor suppressors. N Engl J Med. 2005;353:1768-1771. 7. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Statist Soc Ser B. 1995;57:289-300. 8. Storey J, Tibshirani R. Statistical significance for genome-wide studies. Proc Natl Acad Sci USA. 2003;100:9440-9445. 9. Ivanovska I, Ball A, Diaz R, et al. MicroRNAs in the Mir-106b family regulate p21/cdkn1a and promote cell cycle progression. Mol Cell Biol. 2008;22:22. 10. Liang Y, Ridzon D, Wong L, et al. Characterization of microrna expression profiles in normal human tissues. BMC Genomics. 2007; 8:166. 11. Davoren PA, McNeill RE, Lowery AJ, et al. Identification of suitable endogenous control genes for microrna gene expression analysis in human breast cancer. BMC Mol Biol. 2008;9:76. 12. Gottardo F, Liu C, Ferracin M, et al. Micro-RNA profiling in kidney and bladder cancers. Urol Oncol. 2007;25:387-392. 13. Volinia S, Calin G, Liu C, et al. A microrna expression signature of human solid tumors defines cancer gene targets. Proc Natl Acad Sci USA. 2006;103:2257-2261. 840 UROLOGY 75 (4), 2010

14. Marton S, Garcia M, Robello C, et al. Small RNAs analysis in CLL reveals a deregulation of mirna expression and novel mirna candidates of putative relevance in CLL pathogenesis. Leukemia. 2007;8:8. 15. Gironella M, Seux M, Xie M, et al. Tumor protein 53-induced nuclear protein 1 expression is repressed by Mir-155, and its restoration inhibits pancreatic tumor development. Proc Natl Acad Sci USA. 2007;104:16170-16175. 16. Dutta K, Zhong Y, Liu Y, et al. Association of microrna-34a overexpression with proliferation is cell type-dependent. Cancer Sci. 2007;98:1845-1852. 17. Kulshreshtha R, Ferracin M, Wojcik S, et al. A microrna signature of hypoxia. Mol Cell Biol. 2007;27:1859-1867. 18. Maynard M, Ohh M. Von Hippel Lindau tumor suppressor protein and hypoxia-inducible factor in kidney cancer. Am J Nephrol. 2004;24:1-13. 19. Dong JT. Chromosomal deletions and tumor suppressor genes in prostate cancer. Cancer Metastasis Rev. 2001;20:173-193. 20. Lubinski J, Hadaczek P, Podolski J, et al. Common regions of deletion in chromosome regions 3p12 and 3p14.2 in primary clear cell renal carcinomas. Cancer Res. 1994;54:3710-3713. 21. Dreijerink K, Braga E, Kuzmin I, et al. The candidate tumor suppressor gene, RASSF1A, from human chromosome 3p21.3 is involved in kidney tumorigenesis. Proc Natl Acad Sci USA. 2001; 98:7504-7509. 22. Dalmay T, Edwards DR. MicroRNAs and the hallmarks of cancer. Oncogene. 2006;25:6170-6175. 23. Mitsumori K, Kittleson J, Itoh N, et al. Chromosome 14q LOH in localized clear cell renal cell carcinoma. J Pathol. 2002;198:110-114. 24. Zhang L, Volinia S, Bonome T, et al. Genomic and epigenetic alterations deregulate microrna expression in human. Proc Natl Acad Sci USA. 2008;105:7004-7009. 25. Saito Y, Liang G, Egger G, et al. Specific activation of microrna- 127 with downregulation of the proto-oncogene BCL6 by chromatin-modifying drugs in human cancer cells. Cancer Cell. 2006;9: 435-443. APPENDIX SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.urology.2009.10.033. UROLOGY 75 (4), 2010 841