"This is a big step forward in beginning to understand the quite mysterious mechanisms that are involved in memory-associated DNA methylation," wrote David Sweatt, University of Alabama at Birmingham, in an e-mail to Alzforum (see full comment below). Sweatt was not involved in the study.

As everyone over 40 will admit, advancing age clearly takes a toll on cognition. Researchers have long known that DNA methylation also falls over time (see Wilson et al., 1987), and more recent studies suggest this modification is crucial for memory formation (see ARF related news story on Miller and Sweatt, 2007). In the current study, first author Ana Oliveira and colleagues wanted to know if the expression of enzymes that cause DNA methylation also sags with age, and if propping them up could rescue memory.

Oliveira analyzed the DNA methyltransferases in aged (18-month-old) C57Bl/6 mice and found that Dnmt3a1 and Dnmt3a2 fell with age. Dnmt3a1 is normally associated with the tightly wound, inactive form of DNA, while Dnmt3a2 attaches to the genetically active, uncoiled form of DNA called euchromatin. When NMDA receptors became activated by learning, levels of Dnmt3a2 shot up in young mice, meaning the enzyme was acting like an immediate early-on gene. (Levels rose less with learning in aged mice.) Since previous studies suggested that Dnmt changes with aging center on euchromatin (see Chen et al., 2002), and since Dnmt3a2 rises with neuronal activity, the researchers focused on that enzyme.

Would a boost in Dnmt3a2 fix cognitive deficits in aged mice? To find out, the researchers used gene therapy. They injected an adeno-associated virus that expressed Dnmt3a2 into the hippocampi of mice cognitively impaired by age. Global methylation levels rose in the hippocampi of these mice and hippocampus-dependent memory improved. In trace fear conditioning, old treated mice froze more than old controls during tone presentation 24 hours after training (though less so than young adults), implying that their memory for a conditioned stimulus bested that of aged controls. Further, the treated mice beat controls at identifying a object's changed location a day after training, finding the moved item about as well as did young adult mice. That suggested that restoring DNA methylation in the aged mouse hippocampus rescued long-term, hippocampus-dependent memory.

Showing similar memory deficits in much younger Dnmt3a2-deficient mice would further solidify the enzyme's role in cognition. Oliveira used a virus to ferry short hairpin RNAs into cognitively normal young adult mouse hippocampi, where they decreased Dnmt3a2 levels, then administered the same two behavior tests. Treated mice spent less time freezing in response to a stimulus 24 hours, but not one hour, after training. In addition, mice with Dnmt3a2 knocked down recognized a new object location worse than untreated controls after 24 hours, suggesting that a drop in methylation impaired long-term memory even in young mice.

"The study links the biochemical process causally to cognitive function," said Bading. "It is now clear that in order to form memories, you need this biochemical process. If it doesn't work, then memory doesn't work."

What could Dnmt3a2 be doing to bolster memory? It is not clear yet from this study, but in cultured hippocampus cells, this shRNA treatment prevented levels of Arc and BDNF—plasticity-related immediate-early gene proteins—from reaching those of controls. That means, without DNA methylation, expression of some learning-related genes is suppressed. It seems counterintuitive, but scientists are accepting now that DNA methylation is tied to activating genes as well as silencing them, said Bading (see ARF related news story on Chahrour et al., 2008).

The next steps will be to figure out what causes the reduction in Dnmt3a2 with age in animals, and to see if human Dnmt3a2 expression declines with age, said Bading. Which genes Dnmt3a2 methylates is a other key question, he said. Further down the road, a compound may be able to enhance Dnmt3a2 activity, though no such compound exists yet.

"The distinction of actions of Dnmt3a2 and Dnmt3a1 represents an important advance," said Paul Coleman, Banner Sun Health Research Institute, Sun City, Arizona, who was not involved in the study. Researchers might also consider whether or not methylation has subtle effects on synaptic vesicles or dendritic spines, and what the gene expression profiles look like in these aging animals, he said. Aging might affect methyltransferase gene expression differently in various areas of the brain, Coleman added, since he has found that gene expression varies widely between brain areas (see Berchtold et al., 2008).—Gwyneth Dickey Zakaib.

Reference:
Oliveira AM, Hemstedt TJ, Bading H. Rescue of aging-associated decline in Dnmt3a2 expression restores cognitive abilities. Nat Neurosci. 2012 Jul 1. Abstract