Enzyme erases DNA's molecular coating.

For masons that still perplex scientists, clusters of atoms called methyl groups blanket much of the DNA in humans and other vertebrates. Several years ago, investigators uncovered enzymes that could attach these methyl groups to DNA, a process known as methylation. Now, a Canadian research group reports identifying a demethylase, an enzyme able to strip those methyl groups from DNA.

The discovery has attracted considerable attention because shifting patterns in methylation seem to regulate the activity of genes, particularly during the growth of embryos. Furthermore, the genomes of cancer cells exhibit abnormal methylation, an anomaly that may fuel the runaway growth of the cells.

Moshe Szyf of McGill University in Montreal and his colleagues reasoned that any demethylating enzyme would contain an amino acid sequence capable of binding methylated DNA. By scanning a database for gene fragments encoding such a sequence, the researchers identified several candidate human demethylase genes.

As hoped, the protein encoded by one of those genes proved able in test-tube experiments to detach methyl groups from DNA. Moreover, when the gene was slipped into cells, demethylation occurred in several stretches of DNA, the team reports in the Feb. 18 NATURE.

"We have shown by a number of different assays that [the DNA] gets demethylated," says Szyf. Additional experiments by his group suggest that the enzyme produces a reaction between methylated DNA and water that results in unmethylated DNA and methanol.

Several other research groups have reported discovering enzymes with demethylation activity, but the claims haven't proved compelling to many scientists. The latest announcement has also drawn some skepticism.

"They have to have genetic evidence that this [enzyme] is involved in reshaping methylating patterns," says Timothy H. Bestor of Columbia-Presbyterian Medical Center in New York. To do that, he says, Szyf's team should create mice that have a mutation in the gene for the putative demethylase and observe whether that mutation alters methylation in the animal.

"Is it involved in generating methylation patterns in real life? We don't know yet, but God doesn't put these things here for fun," counters Szyf, who has started to generate mice lacking the enzyme.

The researchers have some evidence suggesting that the protein plays a role in cancer. Its gene is active in tumor cells. When the scientists used so-called antisense technology (SN: 8/6/94, p. 88) to thwart the gene's activity, they produced a dramatic result. "If you knock [the gene] out, cancer cells don't grow anymore," says Szyf.