A slow overall decrease in DNA methylation occurs during aging of several mammals, with shorter-lived animals experiencing faster decreases in methylation. The hypothesis of this proposal is that the loss occurs because repair patches are not efficiently remethylated in adult animals. Most of the proposed experiments will test whether or not demethylation is non-randomly distributed over the mouse genome in ways corresponding to DNA repair sites. One experiment will evaluate an alternative hypothesis, that an enzyme removes all or part of the methylated nucleotide. Tests for demethylation non-randomness will involve two levels of testing. The fine level will involve sequencing both strands of genes using ligation-mediated PCR after Maxam-Gilbert sequencing reactions. This method will show whether transcribed genes are preferentially demethylated and whether the transcribed strand is preferentially demethylated. These preferences are established for DNA repair. The coarse level involves examination of demethylation in nuclear matrix-associated versus looped domains of DNA in general, as well as examination of clustering of demethylated regions along single looped domains. The patterns for repair suggest preference for matrix-associated regions and clustering of repair sites along loops. To provide the most critical distinction between expressed and unexpressed genes, a myelin basic protein transgene linked to a metallothionein promoter will be used to obtain transgenic mouse lines with and without liver expression. Thus, the same gene in the same tissue can be compared for demethylation patchiness holding all constant except for transcription (and resultant repair). An enzyme acting as a glycosidase or an endonuclease to remove part or all of 5methyldeoxycytidylate (d5mCTP) will be looked for during transient genome-wide demethylation of induced Friend cells. A successful assay will be applied to aging mouse liver.
