seminars in CANCER BIOLOGY, Vol. 12, 2002: pp. 389–398 doi:10.1016/S1044–579X(02)00059-7, available online at http://www.idealibrary.com on DNAmethylationandgenomicimprinting:insightsfrom cancerintoepigeneticmechanisms Andrew P. Feinberg a,b,c,∗ , Hengmi Cui a,b and Rolf Ohlsson d Sincethediscoveryofepigeneticalterationsincancer20years ago by Feinberg and Vogelstein, a variety of such alterations have been found, including global hypomethylation, gene hypomethylation and hypermethylation, and loss of imprinting (LOI). LOI may precede the development of cancer and may thus serve as a common marker for risk, but also as a model for understanding the developmental mechanism for normal imprinting. Key words: epigenetics / DNA methylation / genomic imprinting © 2002 Published by Elsevier Science Ltd. Introduction Epigenetics is defined as stable alterations in the genome, heritable through cell division, that do not involve the DNA sequence itself. Epigenetic alter- ations are reversible, at least in the germline, and they often act over a distance. The distance can be relatively small, in the order of kilobases, as in telom- ere silencing in yeast, or the distance can be large, in the order of megabases, as in position effect var- iegation in Drosophila. For these reasons, epigenetic alterations are all thought to involve modifications of chromatin, and one of the most intriguing questions in this field is whether common types of modification account for the diverse examples of epigenetic effects. As will be discussed in this review, some of the lessons gained from the study of imprinting in cancer do have From the a Institute of Genetic Medicine, b Departments of Medicine, c Departments of Molecular Biology & Genetics, and Oncology, Johns Hopkins University School of Medicine, 1064 Ross, 720 Rutland Avenue, Baltimore,MD21205,USAand d DepartmentofAnimalDevelopmentand Genetics, Uppsala University, Uppsala, Sweden. *Corresponding author. E-mail: afeinberg@jhu.edu © 2002 Published by Elsevier Science Ltd. 1044–579X / 02 / $ –see front matter general implication, not only in the understanding of cancer biology but chromatin in general. The first indication that epigenetics played a role in cancer was the discovery in 1983, by Feinberg and Vogelstein, 1 of altered methylation of genes in col- orectal tumors. These alterations were found at the time to occur in all cancers and adenomas, 2 mak- ing them by far the most common type of genetic change in cancer, which is still true in light of current knowledge. While these first observations were of col- orectal cancer, they have been generalized to virtually all types of neoplasia. 3 Many genes in cancers lose methylation, many gain methylation, and at least in colon cancer, there is also a generalized loss of total methylation content in the genome. 3–5 Both gains and losses are likely important. Gains of methylation include promoters of tumor suppressor genes, and it has been hypothesized that the methylation alteration itself is responsible for gene silencing. 6 While that may be true, it is also possible that other chromatin al- terations play a primary role in gene silencing. For ex- ample, it has recently been shown in Neurospora that methylation is dependent upon histone modification. 7 Similarly, losses of methylation likely lead to chro- mosome instability. This has been shown directly by treatment with 5-aza-2 ′ -deoxycytidine, 8 and by direct observation of tumors. 9 The epigenetic alteration that is the main focus of this chapter is loss of genomic imprinting, which is linked to alterations in methy- lation. Indeed epigenetic and genetic alterations are interrelated. 10 One of the most important ideas we wish to commu- nicate in this review is that the study of human disease is an extremely powerful tool to understand normal biology. This is an old idea in genetics, of course, first promulgated by Garrod in his studies of inborn errors of metabolism, and has been true from the studies of alkaptonuria, through the great insights in cholesterol metabolism. 11 However, in the study of epigenetics this idea is particularly important, since outbred populations may reflect a more normal epi- genetic milieu than an inbred laboratory strain. This 389