Summary Since DNA methylation is considered an important mechanism for silencing of retroelements in the mammalian genome, hypomethylation in human tumours may lead to their reactivation. The methylation status of LINE-1 retroposons was determined in 73 samples of urinary bladder cancers, 34 specimens of renal cell carcinoma and in the corresponding normal tissues by Southern blot analysis. LINE-1 sequences were strongly methylated in normal tissues and were significantly hypomethylated in 69 (95%) urothelial carcinomas, but in none of the renal carcinomas. Hypomethylation in bladder cancers was independent of stage and tended to increase with grade. The methylation status of HERV-K proviral DNA in normal and transformed urothelial cells paralleled that of LINE-1 sequences (r 2 = 0.87). It was shown by ligation-mediated polymerase chain reaction that hypomethylation also extended to the LINE-1 promoter sequence located at the 5′-ends of full-length elements which is repressed by methylation in somatic tissues. Accordingly, full-length LINE-1 transcripts were detected by Northern blot analysis in two urothelial carcinoma cell lines. In contrast, transcripts from HERV-K proviruses were restricted to teratocarcinoma cell lines. Our data indicate that genome-wide DNA hypomethylation is an early change in urothelial carcinoma, but is absent from renal cell carcinoma. The coordinate changes of LINE-1 and HERV-K DNA methylation suggest that hypomethylation in urothelial cancer affects a variety of different retroelements to similar extents. We speculate that decreased methylation of LINE-1 retroelements, in particular, may contribute to genomic instability in specific human tumours such as urothelial carcinoma by rendering these normally repressed sequences competent for transcription and recombination.

Retroelements constitute approximately 45% of the human genome. Long interspersed nuclear element (LINE) autonomous retrotransposons are predominantly represented by LINE-1, nonautonomous small interspersed nuclear elements (SINEs) are primarily represented by ALUs, and LTR retrotransposons by several families of human endogenous retroviruses (HERVs). The vast majority of LINE and HERV elements are densely methylated in normal somatic cells and contained in inactive chromatin. Methylation and chromatin structure together ensure a stable equilibrium between retroelements and their host. Hypomethylation and expression in developing germ cells opens a “window of opportunity” for retrotransposition and recombination that contribute to human evolution, but also inherited disease. In somatic cells, the presence of retroelements may be exploited to organize the genome into active and inactive regions, to separate domains and functional regions within one chromatin domain, to suppress transcriptional noise, and to regulate transcript stability. Retroelements, particularly ALUs, may also fulfill physiological roles during responses to stress and infections. Reactivation and hypomethylation of LINEs and HERVs may be important in the pathophysiology of cancer and various autoimmune diseases, contributing to chromosomal instability and chronically aberrant immune responses. The emerging insights into the pathophysiological importance of endogenous retroelements accentuate the gaps in our knowledge of how these elements are controlled in normal developing and mature cells

In prostate carcinoma (PCa) increased DNA methylation ('hypermethylation') occurs at specific genes such as GSTP1. Nevertheless, overall methylation can be decreased ('hypomethylation') because methylation of repetitive sequences like LINE-1 retrotransposons is diminished. We analysed DNA from 113 PCa and 36 noncancerous prostate tissues for LINE-1 hypomethylation by a sensitive Southern technique and for hypermethylation at eight loci by methylation-specific PCR. Hypermethylation frequencies for GSTP1, RARB2, RASSF1A, and APC in carcinoma tissues were each 470%, strongly correlating with each other (Po10 À6 ). Hypermethylation at each locus was significantly different between tumour and normal tissues (10 À11 oPo10 3 ), although hypermethylation, particularly of RASSF1A, was also observed in noncarcinoma tissues. ASC1 hypermethylation was observed in a subgroup of PCa with concurrent hypermethylation. Hypermethylation of CDH1, CDKN2A, and SFRP1 was rare. LINE-1 hypomethylation was detected in 49% PCa, all with hypermethylation at several loci. It correlated significantly with tumour stage, while hypermethylation was neither related to tumour stage nor Gleason score. Coordinate hypermethylation of several genes may occur early in PCa, with additional hypermethylation events and LINE-1 hypomethylation associated with progression. Hypermethylation allows detection of 482% of PCas. PCa may fall into three classes, that is, with few DNA methylation changes, with frequent hypermethylation, or with additional LINE-1 hypomethylation.

Aberrant DNA methylation of CpG sites is among the earliest and most frequent alterations in cancer. Several studies suggest that aberrant methylation occurs in a tumour type-specific manner. However, large-scale analysis of candidate genes has so far been hampered by the lack of high throughput assays for methylation detection. We have developed the first microarray-based technique which allows genome-wide assessment of selected CpG dinucleotides as well as quantification of methylation at each site. Several hundred CpG sites were screened in 76 samples from four different human tumour types and corresponding healthy controls. Discriminative CpG dinucleotides were identified for different tissue type distinctions and used to predict the tumour class of as yet unknown samples with high accuracy using machine learning techniques. Some CpG dinucleotides correlate with progression to malignancy, whereas others are methylated in a tissue-specific manner independent of malignancy. Our results demonstrate that genome-wide analysis of methylation patterns combined with supervised and unsupervised machine learning techniques constitute a powerful novel tool to classify human cancers.

To elucidate the relationship between genomewide DNA hypomethylation and chromosome instability, 55 prostate carcinoma specimens were analyzed for extent of hypomethylation by Southern blot analysis of LINE-1 sequence methylation and for loss or gain of chromosomal material by comparative genomic hybridization. Seventeen (31%) tumors showed strong hypomethylation of DNA, whereas four (7%) displayed slight hypomethylation and the rest of the tumors normal-level methylation. Chromosomal aberrations were observed in 34 carcinomas. The most frequent chromosomal alterations were loss of 13q in 18 cases and aberrations in 8p (loss) or 8q (gain) in 16 cases. The presence of chromosomal loss or gain was significantly associated with the presence of strong hypomethylation. A striking correlation (P = 0.00001) was observed between aberrations on chromosome 8 and hypomethylation, whereas no association was seen between DNA hypomethylation and loss of 13q. The association between DNA hypomethylation and the presence of metastases was statistically significant (P = 0.044), and both chromosomal alterations and DNA hypomethylation tended to be more frequent in higher-stage tumors. In conclusion, the data indicate that hypomethylation is associated with chromosomal instability in prostate cancer. Specifically, a surprisingly strong association between alterations on chromosome 8 and genomewide hypomethylation was found. This association suggests that DNA hypomethylation and alterations in chromosome 8 may be mechanistically linked to each other in prostate carcinoma.