DNA methylation and cancer.

The main thrusts of the arguments that aberrant DNA methylation is involved in the generation of tumor heterogeneity and progression can be summarized as follows. The methylation of specific cytosine residues in DNA is certainly an important component in multilevel gene control in eukaryotes. The discovery of CpG clusters in the flanking regions of genes and their under-methylation on housekeeping genes, except those located on inactive X-chromosomes, strongly suggests a controlling function for modification in these regions. Since methylation plays an important role in controlling normal cellular development, it follows that aberrations within this mechanism may be implicated in the abnormal gene control which characterizes cancer. Methylation patterns are not copied rigorously in rapidly dividing cells. This may be because there is normally a close coordination between DNA synthesis, DNA methylation, and DNA packaging, and changes in the timing of these processes could conceivably result in hypomethylation at some sites and de novo methylation at others. Since the greatest variability of methylation patterns is seen in nonexpressed genes, it is possible that there is a tendency for cells to activate genes when dividing in an inappropriate growth environment. The constant evolution and shuffling of methylation patterns which occur during division might play a role in the development of new phenotypes within cell populations. One might predict that selective pressures within the host would select for those cells with specific new methylation patterns allowing for the expression of genes necessary for survival in a particular environment. Many experiments have in fact shown that methylation levels and patterns and indeed methyltransferase levels (57) are altered in cancer cells. Thus, there is considerable heterogeneity within tumor populations with regard to this fundamental biological control mechanism. The fact that direct intervention by the use of 5-aza-Cyd can result in dramatic alterations in malignant potential allows this hypothesis to be tested more critically. Hopefully, the use of 5-aza-Cyd in defined systems will allow us to isolate genes which might become activated by drug treatment and which might contribute to metastatic potential. An understanding of the fundamental aspects of the enzymology and control of DNA methylation might therefore allow us to make significant inroads into understanding how heterogeneity is generated and what we might do about it.