Search tips
Search criteria 


Logo of epigLink to Publisher's site
Epigenetics. 2011 June; 6(6): 703–709.
Published online 2011 June 1. doi:  10.4161/epi.6.6.16158
PMCID: PMC3230543

Polycomb group genes are targets of aberrant DNA methylation in renal cell carcinoma


The combined effects of genetic and epigenetic aberrations are well recognized as causal in tumorigenesis. Here, we defined profiles of DNA methylation in primary renal cell carcinomas (RCC) and assessed the association of these profiles with the expression of genes required for the establishment and maintenance of epigenetic marks. A bead-based methylation array platform was used to measure methylation of 1,413 CpG loci in ~800 cancer-associated genes and three methylation classes were derived by unsupervised clustering of tumors using recursively partitioned mixture modeling (RPMM). Quantitative RT-PCR was performed on all tumor samples to determine the expression of DNMT1, DNMT3B, VEZF1 and EZH2. Additionally, methylation at LINE-1 and AluYb8 repetitive elements was measured using bisulfite pyrosequencing. Associations between methylation class and tumor stage (p = 0.05), LINE-1 (p < 0.0001) and AluYb8 (p < 0.0001) methylation, as well as EZH2 expression (p < 0.0001) were noted following univariate analyses. A multinomial logistic regression model controlling for potential confounders revealed that AluYb8 (p < 0.003) methylation and EZH2 expression (p < 0.008) were significantly associated with methylation class membership. Because EZH2 is a member of the Polycomb repressive complex 2 (PRC2), we next analyzed the distribution of Polycomb group (PcG) targets among methylation classes derived by clustering the 1,413 array CpG loci using RPMM. PcG target genes were significantly enriched (p < 0.0001) in methylation classes with greater differential methylation between RCC and non-diseased kidney tissue. This work contributes to our understanding of how repressive marks on DNA and chromatin are dysregulated in carcinogenesis, knowledge that might aid the development of therapies or preventive strategies for human malignancies.

Key words: EZH2, DNA methylation, renal cell carcinoma, polycomb, microarray


There are approximately 58,000 new cases and 13,000 deaths from kidney cancers each year in the US and incidence has risen steadily over the last three decades.1,2 Arising from the epithelium of the renal tubule, renal cell carcinoma (RCC) accounts for greater than 85% of all malignancies of the kidney, with clear cell (conventional) RCC being the most common histological subtype. With an overall 5-year survival rate of around 60%, the prognosis associated with metastatic disease is less than 10%.3 While classical pathological methods based on histological assessment are still used predominantly for the diagnosis and prognostication of RCC, an understanding of the molecular pathology of the disease holds great promise for enhanced detection as well as more refined and targeted treatment approaches.

Appreciation of the molecular character of RCC was initiated by the recognition of the association with von Hippel-Lindau (VHL) tumor suppressor, the somatic inactivation of which is the most frequent genetic event leading to the development of sporadic conventional RCC.4,5 Since then, other genetic biomarkers, many of which are directly related to the VHL defect, have been identified.6

Stable control of transcription by promoter hypermethylation is now recognized as playing an important role in carcinogenesis in a wide variety of malignancies, including RCC. Several groups have demonstrated frequent inactivation of VHL and other tumor suppressor genes by promoter methylation in RCC, some of which have been associated with poorer patient prognosis.79 It has also been suggested that subsets of RCCs may display a CpG Island Methylator Phenotype (CIMP); though, as yet, this does not associate with a known RCC subtype or other clinicopathologic characteristics.10

Several proteins are known to coordinate the process of promoter methylation. While DNA methyltransferases DNMT3A and DNMT3B are responsible for establishing de novo patterns of methylation, DNMT1 is the major maintenance methyltransferase, maintaining these patterns in adult cells. DNMTs are modestly overexpressed in many cancers11 and recent studies have shown a mechanistic link between the DNMTs and enhancer of zeste homolog 2 (EZH2). EZH2, the catalytic subunit of Polycomb repressive complex 2 (PRC2), is a histone methyltransferase that can direct DNA methylation via recruitment of DNMTs.12 Increased EZH2 expression has been observed in many cancers including RCC and has been seen to promote cell proliferation and inhibit apoptosis in RCC cell lines.13 Additionally, a prospective study of RCC found that high tumor EZH2 expression was associated with significantly lower disease recurrence.14 Lastly, recent work on the vascular endothelial zinc finger 1 (VEZF1) protein has suggested that its abundance is linked with the expression of DNMT3B, and its loss is associated with widespread loss of methylation both globally and in distinct CpG islands.15

Utilizing high throughput measurement of CpG methylation in 1,413 autosomal loci contained in 773 cancer-associated genes, this study aimed to test the hypothesis that the expression of critical DNA methylation-associated genes is associated with patterns of genomic methylation in these tumors.


Non-diseased kidney tissues cluster together based on methylation profiles.

The extent of methylation of 1,413 CpG loci in ~800 cancer-related genes was determined for 67 RCC samples and six non-diseased kidney samples using the Illumina GoldenGate Methylation BeadArray. Demographic and tumor stage and grade data for cases are shown in Table 1. A recursively partitioned mixture model (RPMM) was used to cluster samples into classes according to similarities in their methylation profiles. When applied to methylation data from all autosomal loci in tumors and normal samples, RPMM yielded six distinct methylation classes (a–f). Figure 1A shows a heat map resulting from the RPMM analysis, depicting the average methylation beta values across samples in each class (columns) at each locus (rows). Five out of six normal samples fell into one class (Class c), while the sixth sample fell into a neighboring class (Class b), which originated from the same node in the hierarchy, indicating that normal kidney samples had similar methylation profiles and were generally distinct from the majority of tumors (Chi-squared p = 0.002).

Figure 1
Results of a recursively partitioned mixture model. In heat maps, columns represent samples, with column widths proportional to the number of samples in each class, and rows represent individual CpG loci. The color of rows in each class represents the ...
Table 1
Patient demographics and clinicopathologic characteristics

EZH2 expression is associated with methylation class in RCC tumors.

When analyzing only RCC samples, RPMM resulted in separation of tumors into three distinct methylation classes (1–3) (Fig. 1B). Overall methylation extent progressively decreases from Class 1 to Class 3 (Fig. 1C). After Bonferroni correction, methylation of 87 genes was found to vary significantly across the three RPMM classes (Sup. Table 1). RPMM classes were analyzed for associations with expression of four genes whose altered expression has been associated with changes to DNA methylation, specifically EZH2, VEZF, DNMT1 and DNMT3B. In a univariate analysis, EZH2 was found to have significantly higher expression in tumors in methylation Class 3 compared with the other two classes (Kruskal Wallis, p < 0.0001; Fig. 2A). A multinomial logistic regression was used to model the association between the RPMM classes and the expression of the aforementioned genes while controlling for potential confounders. Adjusting for age, gender, tumor stage and methylation of both LINE-1 and AluYb8, EZH2 expression was significantly associated with methylation class membership (Wald p = 0.008; Sup. Table 2). Figure 2D illustrates the results of the model: as expression of EZH2 increases for any given sample, the probability of being a member of Class 1 or 2 decreases while, simultaneously, the probability of being in Class 3 increases.

Figure 2
Univariate and multivariable associations with methylation class. Plots depicting variation in samples across RPMM classes of (A) EZH2 expression, (B) AluYb8 methylation and (C) LINE-1 methylation (Kruskal Wallis p < 0.0001, for each case). Plots ...

Classes of promoter methylation are associated with global methylation extent and histological stage.

Univariate analyses were performed on RPMM classes to examine if the resulting methylation classes were associated with tumor characteristics. A significant association among RPMM classes was found with tumor stage Chi-squared, p = 0.05). While Class 3 was comprised exclusively of tumors with low stage, Class 2 was comprised of stage I, II and III tumors and Class 3 was comprised of tumors of all stages (Table 2).

Table 2
Chi-Squared analysis of stage by RPMM class

Pyrosequencing was used to determine methylation at repetitive elements LINE-1 and AluYb8, two common markers of global methylation. Samples belonging to Class 1 (the class with the highest average extent of gene-associated methylation) had significantly lower methylation of both AluYb8 and LINE-1 (Kruskal Wallis, p < 0.0001; Fig. 2B and C, respectively) when compared to Classes 2 and 3. The association between class membership and AluYb8 remained significant in a multinomial logistic regression model controlling for potential confounders (Wald, p = 0.003; Sup. Table 1) and probabilities of class membership with increasing AluYb8 methylation were plotted (Fig. 2E).

Classes of CpG loci with least correlation of methylation between tumor and normal are enriched for polycomb group targets.

Given the association between DNA methylation in these tumors and the expression of EZH2, a PRC2 component, we investigated the distribution of PcG target genes among loci with similar methylation extents in RCC tumors. All 1,413 CpG loci (as opposed to samples) were subjected to clustering using RPMM to arrive at eight individual classes of methylation. Presence or absence of PcG targets was plotted coordinately for each locus. A logistic regression was used to calculate the odds of containing a PcG-target locus for each methylation class as compared to the referent class, Class 8, which demonstrated the highest extent of CpG methylation. The classes with lower extent of methylation (Classes 1–5) all showed significantly higher odds of containing PcG loci compared to the referent class (p < 0.0001 in all comparisons; Table 3).

Table 3
Logistic regression: Odds of containing PcG-target locus by methylation class

We next sought to investigate how the particular groups of CpG loci in each class differed between RCC tumor and non-diseased kidney epithelia. Correlation plots were generated and Spearman's rank correlation coefficients calculated for each individual class. Methylation classes 2–4, which contained the most PcG target gene loci, also showed the lowest correlation in methylation between tumor and normal tissue (Fig. 3).

Figure 3
Classes with poor correlation in methylation status of tumor and normal tissue are enriched for PcG target-associated loci. Top strip shows methylation status of all CpG loci separated into 8 classes derived by using RPMM to cluster the CpG loci based ...


The utility of genome-wide methylation profiles in defining biologically and clinically relevant sub-classes of cancer has been established for multiple cancer types and is attributed to high correlation between epigenetic alterations across CpG loci.16 The definition of methylation-based classes has revealed associations between these distinct classes and clinicopathologic tumor characteristics1720 as well as environmental exposures.17,20,21 The goal of the present study was to understand how methylation-based profiles are associated with expression of the genes responsible for establishing and maintaining methylation marks in the context of RCC. The use of an array for identifying methylation profiles provides a broad genomic assessment of epigenetic alterations, reducing the bias of a candidate gene approach and allowing for a robust statistical assessment for finding correlations with a range of covariates. Our method for clustering samples, RPMM, allows for data-driven clustering based on methylation, and subsequent examination of the biological or clinical meaning of these clusters.

While much is known about the genetics of RCC, only limited work has been done toward understanding the contributions of epigenetics to the disease. For example, the von Hippel Lindau (VHL) tumor suppressor, which had long been recognized as playing a central role in development of sporadic clear cell RCC due to decreased expression by mutational inactivation or loss of heterozygosity, has also been shown to be silenced by DNA methylation in a significant proportion of tumors.7 As appreciation for the role of epigenetics in renal carcinogenesis has grown, several other epigenetically altered genes have been identified in RCC including SFRP1 and Wnt-antagonist DKK2.9,2225

Initially, RPMM was applied to methylation data from both tumor and normal samples, and demonstrated that the pattern of methylation in cancer-related genes varies between tumor and normal samples. This analysis resulted in six classes, two of which contained all the normal samples, suggesting that normal tissues have similar methylation profiles to one another and are distinct from the majority of tumors. The observation that normal tissues fell into classes along with some tumor samples suggests an increased presence of non-malignant cells, a less aggressive phenotype, or both for those tumors. RPMM on RCCs alone resulted in only three classes, suggesting a decrease in heterogeneity of methylation profiles upon removal of normal kidney tissues. Mean methylation extent across all loci decreased progressively from Classes 1–3.

An altered epigenetic landscape can originate from dysregulation in the expression of genes that control DNA methylation or regulate epigenetic processes by direct or indirect mechanisms. Therefore, we evaluated the expression of key epigenetic regulatory genes DNMT3B, DNMT1, VEZF1 and EZH2 in tumors. An association was noted between class membership and EZH2 expression, where methylation Class 3 tumors demonstrated higher expression of EZH2 than tumors in the other two classes. Class 3 was also comprised of only tumors of low stage (I and II), whereas the prevalence of higher stage tumors (III and IV) was seen to increase progressively from Class 2 to Class 1. With the lower extent of aberrant promoter hypermethylation observed in Class 3 tumors, our data suggest that tumors overexpressing EZH2 are less advanced and demonstrate fewer epigenetic abnormalities. In fact, this finding is consistent with recent work showing that while EZH2 expression is significantly elevated in primary RCC tumors compared with histologically normal kidney samples, higher EZH2 expression among tumors was characteristic of less aggressive tumors and a more favorable disease prognosis.14 Taken together, these data suggest that EZH2 may be more important in tumor initiation than in progression of RCC.

Our analysis of global methylation markers revealed that the extent of methylation at both LINE-1 and AluYb8 was significantly higher in Classes 2 and 3 as compared to Class 1. Thus, methylation of these repetitive elements was inversely related to the extent of gene-associated methylation across classes, a pattern consistent with the gene-specific regional hypermethylation and concomitant global genomic hypomethylation that is a hallmark of neoplasia.26 Our study is limited in its sample size to more precisely quantify this relationship, and further studies focused on the relationship between LINE-1 and AluYb8 methylation and gene-specific methylation in low-grade tumors is warranted.

Considering that methylation profiles of samples were associated with EZH2 expression, and EZH2 is the major catalytic subunit of the histone methyltransferase complex PRC2, the distribution of PcG target genes across loci with differential methylation was of great interest. The significant contribution of aberrant chromatin structure in RCC was highlighted in a recent finding, demonstrating that mutations of the SWI/SNF chromatin remodeling complex gene PBRM1 existed in 41% of RCC tumors in the study.27 Our classification of CpGs based on methylation state revealed that classes of CpG loci with relatively low methylation extent in tumors were significantly enriched for loci associated with PcG target genes. Expanding this examination to include a comparison with non-diseased kidney tissue led to the observation that these same classes were relatively less correlated between tumor and normal tissue, and likely include genes targeted for alteration in carcinogenesis. Interestingly, this suggests that while there is a broad range of methylation degree among cancer-related genes in RCC, the genes that are most highly methylated among tumors are not necessarily those that drive the cancer phenotype, which is characterized by gene-specific hypermethylation in tumors compared to normal tissue. In addition, PcG target genes associate with a large proportion of the loci that are differentially methylated (specifically hypermethylated) in tumors compared to normal tissue. This phenomenon may be explained by a new body of literature describing the epigenetic landscape of cancer stem cells. These studies present a model whereby PcG target genes, specifically regulatory targets of EZH2, which in normal cells carry the distinctive mark of trimethylated lysine 27 of histone 3, are coordinately regulated with multiple genes that become aberrantly hypermethylated in cancer.2830 Collectively, these data suggest that EZH2 pre-marks genes for ultimate “deep silencing” by DNA methylation and epigenetic silencing in the context of carcinogenesis.

The pivotal role that EZH2 plays in RCC carcinogenesis exemplifies its newly understood function as a mechanistic connector between the two major modes of epigenetic gene repression, namely polycomb group repression via histone remodeling and DNA methylation.12 The present work strongly supports the existing model that these two mechanisms, which act together to determine the accessibility of chromatin to transcriptional machinery, are inextricably linked. In addition to showing that patterns of tumor methylation are associated with the expression of a gene, EZH2, which is indispensable for establishing these marks, our work demonstrates that these methylation patterns may be reflective of important and early gene-specific alterations which have shaped the tumor epigenome.


Study population.

The cases examined in this analysis are drawn from the tumor bank of the Molecular Pathology Core Laboratory at Rhode Island Hospital. All sample collections were performed in accordance with procedures approved by the Institutional Review Boards of Brown University and Rhode Island Hospital. Sixty-seven fresh-frozen RCC tumor tissues were available for examination. Data on histology and clinical course were abstracted from the medical record. A study pathologist confirmed >75% tumor content in each of the RCC samples used in these analyses. Non-diseased kidney samples were procured from the National Disease Research Interchange.

RNA isolation.

Total RNA was isolated from tumors using the mirVANA RNA Isolation Kit (Ambion, Inc., Austin, TX) according to the manufacturer's protocol. RNA was quantified using the Nanodrop ND-1000 spectrophotometer (Nanodrop, Wilmington, DE), aliquoted and stored at −80°C briefly until used for laboratory analysis.

Quantitative reverse transcription-PCR.

A cDNA library was generated using the SuperScript II Reverse Transcription Kit (Invitrogen, Gaithersburg, MD). 2 µg RNA were used for each reverse transcription (RT) reaction along with other RT components, per manufacturer's specifications. cDNA was diluted 1:4 before use in real-time PCR reaction. TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA) were used to determine expression of DNMT1, DNMT3B, EZH2, VEZF1 and GAPDH, which was used as an endogenous control. All reactions, excluding no-template controls and non-reverse-transcribed controls, were run in triplicate on an ABI 7500 Fast Real Time PCR Detection System. All real-time PCR data were quantified by calculating the difference in Ct value between each gene and GAPDH for each sample.

DNA extraction and methylation analysis.

DNA was extracted from fresh-frozen RCC samples using the QIAamp DNA mini kit according to the manufacturer's protocol (Qiagen, Valencia, CA). Sodium bisulfite modification of the DNA was performed using the EZ DNA Methylation Kit (Zymo Research, Orange, CA) following the manufacturer's protocol, with the addition of a 5 min initial incubation at 95°C prior to addition of the denaturation reagent. Illumina GoldenGate® methylation bead arrays (Illumina, San Diego, CA) were used to simultaneously interrogate 1,505 CpG loci associated with 803 cancer-related genes. Bead arrays were run at the University of California-San Francisco Institute for Human Genetics, Genomics Core Facility according to the manufacturer's protocol and as described in reference 31. Results from this array platform have been validated multiple times by alternative approaches for measuring methylation.32,33

Following sodium bisulfite modification, determination of LINE-1 and Alu Yb8 region methylation extent was performed as previously described in reference 34 and 35, using pyrosequencing analysis.

Statistical analysis.

Data were assembled using BeadStudio Methylation software (Illumina). All array data points are represented by fluorescent signals from both methylated (Cy5) and unmethylated (Cy3) alleles and methylation level is given by β = [max (Cy5,0)]/(|Cy3| + |Cy5| + 100), the average methylation (β) value is derived from the ~30 replicate methylation measurements for each locus. As a quality control measure, CpG loci with median detection p-values > 0.05 (n = 8, 0.5%), computed by comparing to a background model generated from negative controls, were eliminated from the analysis. X-linked CpG loci were also removed from analysis leaving 1,413 autosomal CpG loci in the final data set.

Statistical analyses were carried out using R v8.0. We utilized a recursively partitioned mixture model (RPMM) approach to cluster samples based on their methylation information, as previously described in reference 36. Class membership was determined from RPMM and subsequent univariate associations were tested via permutation test with 10,000 permutations each. For continuous variables, the Kruskal-Wallis test statistic was used, whereas for categorical variables, a permutation Chi-square test was used. PcG-target status was mapped coordinately for each CpG locus. A locus was defined as a PcG-target locus if the gene with which it was associated was identified as a PcG-target in 1 or more of 4 publications containing gene lists of PcG-targets identified in embryonic cells.28,3739 A penalized multinomial logistic regression was used to model methylation class membership while controlling for potential confounders. Class 1, from the RPMM used to cluster samples on the basis of their methylation profiles, was held as the referent. Due to the large number of potential methylation classes, logistic regression coefficients were regularized using a ridge (L2) penalty, with coefficients for a common (non-intercept) covariate across outcome levels shrunken toward zero, and the tuning parameter was selected by minimizing Bayesian information criterion.40,41 Continuous variables used in the model were mean-centered and standardized. Stage data were dichotomized into low (I and II) and high (III and IV) categories.


We thank Dr. Murray Resnick at the Rhode Island Hospital and the Center for Biomedical Research Excellence in Cancer Research (P20RR017695) for ascertainment of tumor samples. This work was supported by the Superfund Basic Research Program (P42ES013660) and Flight Attendant Medical Research Institute (YCSA 052341).

Supplementary Material

Supplementary Material:


1. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ. Cancer statistics, 2009. CA Cancer J Clin. 2009;59:225–249. [PubMed]
2. Ljungberg B, Hanbury DC, Kuczyk MA, Merseburger AS, Mulders PF, Patard JJ, Sinescu IC. Renal cell carcinoma guideline. Eur Urol. 2007;51:1502–1510. [PubMed]
3. Lohse CM, Cheville JC. A review of prognostic pathologic features and algorithms for patients treated surgically for renal cell carcinoma. Clin Lab Med. 2005;25:433–464. [PubMed]
4. Gnarra JR, Tory K, Weng Y, Schmidt L, Wei MH, Li H, et al. Mutations of the VHL tumour suppressor gene in renal carcinoma. Nat Genet. 1994;7:85–90. [PubMed]
5. Latif F, Tory K, Gnarra J, Yao M, Duh FM, Orcutt ML, et al. Identification of the von Hippel-Lindau disease tumor suppressor gene. Science. 1993;260:1317–1320. [PubMed]
6. McGuire BB, Fitzpatrick JM. Biomarkers in renal cell carcinoma. Curr Opin Urol. 2009;19:441–446. [PubMed]
7. Herman JG, Latif F, Weng Y, Lerman MI, Zbar B, Liu S, et al. Silencing of the VHL tumor-suppressor gene by DNA methylation in renal carcinoma. Proc Natl Acad Sci USA. 1994;91:9700–9704. [PubMed]
8. Morris MR, Gentle D, Abdulrahman M, Clarke N, Brown M, Kishida T, et al. Functional epigenomics approach to identify methylated candidate tumour suppressor genes in renal cell carcinoma. Br J Cancer. 2008;98:496–501. [PMC free article] [PubMed]
9. Morris MR, Ricketts C, Gentle D, Abdulrahman M, Clarke N, Brown M, et al. Identification of candidate tumour suppressor genes frequently methylated in renal cell carcinoma. Oncogene. 2010;29:2104–2117. [PMC free article] [PubMed]
10. McRonald FE, Morris MR, Gentle D, Winchester L, Baban D, Ragoussis J, et al. CpG methylation profiling in VHL related and VHL unrelated renal cell carcinoma. Mol Cancer. 2009;8:31. [PMC free article] [PubMed]
11. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nature Reviews Genetics. 2002;3:415–428. [PubMed]
12. Vire E, Brenner C, Deplus R, Blanchon L, Fraga M, Didelot C, et al. The Polycomb group protein EZH2 directly controls DNA methylation. Nature. 2006;439:871–874. [PubMed]
13. Wagener N, Holland D, Bulkescher J, Crnkovic-Mertens I, Hoppe-Seyler K, Zentgraf H, et al. The enhancer of zeste homolog 2 gene contributes to cell proliferation and apoptosis resistance in renal cell carcinoma cells. Int J Cancer. 2008;123:1545–1550. [PubMed]
14. Hinz S, Weikert S, Magheli A, Hoffmann M, Engers R, Miller K, Kempkensteffen C. Expression profile of the polycomb group protein enhancer of Zeste homologue 2 and its prognostic relevance in renal cell carcinoma. J Urol. 2009;182:2920–2925. [PubMed]
15. Gowher H, Stuhlmann H, Felsenfeld G. Vezf1 regulates genomic DNA methylation through its effects on expression of DNA methyltransferase Dnmt3b. Genes Dev. 2008;22:2075–2084. [PubMed]
16. Marsit CJ, Houseman EA, Christensen BC, Eddy K, Bueno R, Sugarbaker DJ, et al. Examination of a CpG island methylator phenotype and implications of methylation profiles in solid tumors. Cancer Res. 2006;66:10621–10629. [PubMed]
17. Christensen BC, Kelsey KT, Zheng S, Houseman EA, Marsit CJ, Wrensch MR, et al. Breast cancer DNA methylation profiles are associated with tumor size and alcohol and folate intake. PLoS Genet. 2010;6:1001043. [PMC free article] [PubMed]
18. Christensen BC, Marsit CJ, Houseman EA, Godleski JJ, Longacker JL, Zheng S, et al. Differentiation of lung adenocarcinoma, pleural mesothelioma and non-malignant pulmonary tissues using DNA methylation profiles. Cancer Res. 2009;69:6315–6321. [PMC free article] [PubMed]
19. Marsit CJ, Houseman EA, Christensen BC, Gagne L, Wrensch MR, Nelson HH, et al. Identification of methylated genes associated with aggressive bladder cancer. PLoS One. 2010;5:e12334. [PMC free article] [PubMed]
20. Wilhelm-Benartzi CS, Koestler D, Houseman EA, Christensen BC, Wiencke JK, Schned AR, et al. DNA methylation profiles delineate etiologic heterogeneity and clinically important subgroups of bladder cancer. Carcinogenesis. 2010;31:1972–1976. [PMC free article] [PubMed]
21. Marsit CJ, Christensen BC, Houseman EA, Karagas MR, Wrensch MR, Yeh RF, et al. Epigenetic profiling reveals etiologically distinct patterns of DNA methylation in head and neck squamous cell carcinoma. Carcinogenesis. 2009;30:416–422. [PMC free article] [PubMed]
22. Awakura Y, Nakamura E, Ito N, Kamoto T, Ogawa O. Methylation-associated silencing of SFRP1 in renal cell carcinoma. Oncol Rep. 2008;20:1257–1263. [PubMed]
23. Dalgin GS, Drever M, Williams T, King T, DeLisi C, Liou LS. Identification of novel epigenetic markers for clear cell renal cell carcinoma. J Urol. 2008;180:1126–1130. [PubMed]
24. Hirata H, Hinoda Y, Nakajima K, Kawamoto K, Kikuno N, Kawakami K, et al. Wnt antagonist gene DKK2 is epigenetically silenced and inhibits renal cancer progression through apoptotic and cell cycle pathways. Clin Cancer Res. 2009;15:5678–5687. [PubMed]
25. Seliger B, Handke D, Schabel E, Bukur J, Lichtenfels R, Dammann R. Epigenetic control of the ubiquitin carboxyl terminal hydrolase 1 in renal cell carcinoma. J Transl Med. 2009;7:90. [PMC free article] [PubMed]
26. Esteller M. Epigenetics in cancer. N Engl J Med. 2008;358:1148–1159. [PubMed]
27. Varela I, Tarpey P, Raine K, Huang D, Ong CK, Stephens P, et al. Exome sequencing identifies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma. Nature. 2011;469:539–542. [PMC free article] [PubMed]
28. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Zimmerman J, et al. Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer. Nat Genet. 2007;39:232–236. [PubMed]
29. Yasuda H, Soejima K, Watanabe H, Kawada I, Nakachi I, Yoda S, et al. Distinct epigenetic regulation of tumor suppressor genes in putative cancer stem cells of solid tumors. Int J Oncol. 2010;37:1537–1546. [PubMed]
30. Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L, et al. A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat Genet. 2007;39:237–242. [PMC free article] [PubMed]
31. Bibikova M, Lin Z, Zhou L, Chudin E, Garcia EW, Wu B, et al. High-throughput DNA methylation profiling using universal bead arrays. Genome Res. 2006;16:383–393. [PubMed]
32. Christensen BC, Houseman EA, Marsit CJ, Zheng S, Wrensch MR, Wiemels JL, et al. Aging and environmental exposures alter tissue-specific DNA methylation dependent upon CpG island context. PLoS Genet. 2009;5:1000602. [PMC free article] [PubMed]
33. Marsit CJ, Houseman EA, Christensen BC, Gagne L, Wrensch MR, Nelson HH, et al. Identification of methylated genes associated with aggressive bladder cancer. PLoS One. 2010;5:12334. [PMC free article] [PubMed]
34. Wilhelm CS, Kelsey KT, Butler R, Plaza S, Gagne L, Zens MS, et al. Implications of LINE1 methylation for bladder cancer risk in women. Clin Cancer Res. 2010;16:1682–1689. [PMC free article] [PubMed]
35. Choi SH, Worswick S, Byun HM, Shear T, Soussa JC, Wolff EM, et al. Changes in DNA methylation of tandem DNA repeats are different from interspersed repeats in cancer. Int J Cancer. 2009;125:723–729. [PMC free article] [PubMed]
36. Houseman EA, Christensen BC, Yeh RF, Marsit CJ, Karagas MR, Wrensch M, et al. Model-based clustering of DNA methylation array data: a recursive-partitioning algorithm for high-dimensional data arising as a mixture of beta distributions. BMC Bioinformatics. 2008;9:365. [PMC free article] [PubMed]
37. Bracken AP, Dietrich N, Pasini D, Hansen KH, Helin K. Genome-wide mapping of Polycomb target genes unravels their roles in cell fate transitions. Genes Dev. 2006;20:1123–1136. [PubMed]
38. Lee TI, Jenner RG, Boyer LA, Guenther MG, Levine SS, Kumar RM, et al. Control of developmental regulators by Polycomb in human embryonic stem cells. Cell. 2006;125:301–313. [PMC free article] [PubMed]
39. Squazzo SL, O'Geen H, Komashko VM, Krig SR, Jin VX, Jang SW, et al. Suz12 binds to silenced regions of the genome in a cell-type-specific manner. Genome Res. 2006;16:890–900. [PubMed]
40. Hastie T, Tibshirani R. Efficient quadratic regularization for expression arrays. Biostatistics. 2004;5:329–340. [PubMed]
41. Cawley GC, Talbot NLC, Girolami M. Sparse multinomial logistic regression via Bayesian L1 regularisation. In: Scholkopf B, Platt J, Hofmann T, editors. Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press; 2007. pp. 209–216.

Articles from Epigenetics are provided here courtesy of Taylor & Francis