ABSTRACT: We assessed DNA methylation and copy number status of 27,000 CpGs in 149 urothelial carcinomas and integrated the findings with gene expression and mutation data. Methylation was associated with gene expression for 1,332 CpGs, of which 26% showed positive correlation with expression, i.e., high methylation and high gene expression levels. These positively correlated CpGs were part of specific transcription factor binding sites, such as sites for MYC and CREBP1, or located in gene bodies. Furthermore, we found genes with copy number gains, low expression and high methylation levels, revealing an association between methylation and copy number levels. This phenomenon was typically observed for developmental genes, such as HOX genes and tumor suppressor genes. In contrast, we also identified genes with copy number gains, high expression and low methylation levels. This was for instance observed for some keratin genes. Tumor cases could be grouped into four subgroups, termed epitypes, by their DNA methylation profiles. One epitype was influenced by the presence of infiltrating immune cells, two epitypes were mainly composed of non-muscle invasive tumors, and the remaining epitype of muscle invasive tumors. The polycomb complex protein EZH2 that blocks differentiation in embryonic stem cells showed increased expression both at the mRNA and protein levels in the muscle invasive epitype, together with methylation of polycomb target genes and HOX genes. Our data highlights HOX gene silencing and EZH2 expression as mechanisms to promote a more undifferentiated and aggressive state in UC. The study was conducted on genomic DNA from 150 primary fresh-frozen urothelial carcinoma tumors selected to have equal representation of stages Ta, T1 and T2>. The design also included four fully methylated and four unmethylated hybridization controls as well as 7 Normal urothelium samples. One tumor sample (UC_0556_1) was included as a technical replicate in each of four bisulfite treatment batches. The samples were bisulfite treated and hybridized to Illumina Human Methylation27 Bead Chip arrays according to manufacturers instructions at the SCIBLU Genomics Centre at Lund University, Sweden. Note: Illumina methylation data set comprises total of 168 samples (raw data), but 156 samples (processed data), derived from primary urothelial carcinoma samples. See README.txt for details of the 12 samples not included in the final, processed data.
