Gilbert's conjecture: the search for DNA (cytosine-5) demethylases and the emergence of new functions for eukaryotic DNA (cytosine-5) methyltransferases.

In 1985 Walter Gilbert challenged members of the DNA methylation community assembled at a National Institutes of Health meeting organized by Giulio Cantoni and Ahron Razin with the following words: "The most exciting aspect about the methyl groups on DNA is the thought that they might provide a locally inherited change in a DNA structure. However, for that to be interesting, those changes have to be different in different cells. Furthermore, the alterations in methylation have to be freely imposable and have to be maintained. It is not yet clear that all these properties are true. So I don't think one will find that methylation ever is one of the primary, top-level controls on gene expression."In essence, Gilbert's conjecture, that DNA methylation is not one of the top-level controls on gene expression, assumes that evidence in favor of both of its testable propositions will not be obtained. Evidence for the first proposition, that alterations in methylation status associated with gene-expression states have to be maintained, was already available in 1985 and has been strengthened by a number of very recent experiments. However, the extensive effort to obtain evidence for the second proposition, that alterations in methylation status be freely imposable, has not been successful in its original intent. The effort has, on the other hand, resulted in the emergence of new functions for 5-methylcytosine and the cytosine methyltransferases in eukaryotic DNA repair, recombination and chromosome stability. Copyright 2000 Academic Press.