24 March 2007. They may only tip the scales at a measly 15 grams per mole, but methyl groups carry considerable weight when added to larger macromolecules such as nucleic acids and protein—methylation silences gene expression and can dramatically alter protein activity. So it is not surprising that aberrant methylation patterns are linked to a variety of neurologic disorders, including Alzheimer disease. Some recent studies strengthen the connection between methylation and the inner workings of the brain. Papers in Neuron and PNAS reveal how reversible methylation of DNA plays a crucial role in learning and memory, while a report in the Journal of Neuroscience shows that the methylation status of protein phosphatase 2A (PP2A) may be intimately linked to phosphorylation patterns of amyloid-β precursor protein (APP) and tau that are associated with Alzheimer pathology. The latter study also links PP2A methylation to levels of homocysteine (Hcy), which has been studied as a potential biomarker for AD.
Doing a Number on Protein Phosphatase 2A
Protein phosphatase 2A is a heterotrimer that dephosphorylates a wide range of substrates in the cell, including tau. Estelle Sontag and colleagues have shown that efficient dephosphorylation of tau by PP2A (see Sontag et al., 1996) requires that the phosphatase be methylated by a specific PP2A methyl transferase (PPMT). Methylation leaves the basal catalytic activity of PP2A unchanged, but alters its substrate specificity. Because the activity of PPMT is sensitive to levels of S-adenosylmethionine (SAM) and S-adenosylhomocysteine (SAH), Sontag and colleagues at the University of Texas Southwestern Medical Center in Dallas tested how these amino acid metabolites might affect the PPMT, PP2A, and tau relationship. Their findings are reported in the March 14 Journal of Neuroscience.
Sontag first tested the effects of both SAM and SAH on N2a neuroblastoma cells. She found that SAM treatment causes elevated expression of PPMT and an increase in PP2A methylation of about 25 percent. SAH, on the other hand, had no effect on PPMT expression, but it did decrease PP2A methylation. To test the consequences of these changes, the researchers looked at levels of phospho-tau. They found that addition of SAH to cell cultures causes an increase in levels of phosphorylation at serine 396/serine 404. This motif is recognized by the monoclonal antibody PHF-1, which binds paired-helical fragments, the building blocks of the neurofibrillary tangles. In contrast, SAM decreased tau phosphorylation, an effect that could be blocked by the PP2A inhibitor okadaic acid, suggesting that the SAM effect is mediated via PP2A and not some other pathway.
Next the researchers turned to APP. Because phosphorylation of the precursor protein at threonine 668 spurs cleavage to generate Aβ, any change in PP2A activity that impacts Thr-668 phosphorylation could have a major effect on pathology. Sontag and colleagues found that SAH and SAM also had opposite effects on APP, the former causing an increase in Thr-668 phosphorylation and the latter decreasing it. Overexpression of a PP2A demethylase, PME-1, also caused an increase in phosphorylated APP. These effects seemed directly linked to APP processing, since SAM, which induced PP2A methylation, also enhanced non-amyloidogenic processing of APP, as judged by changes in the soluble α-secretase cleavage product sAPPα. In contrast, expression of a methylation incompetent PP2A or PME-1 led to higher levels of sAPPβ and were associated with increased Aβ40 secretion into the culture medium.
The results suggest that decreased methylation of PP2A could have a dramatic impact on AD pathology by allowing increased phosphorylation of both tau and APP and leading to elevated paired-helical fragments and Aβ. To test the physiological relevance of this, the authors turned to a model of hyperhomocysteinemia—mice lacking the homocysteine metabolizing enzyme cystathionine-β-synthase that were fed a diet high in methionine and low in folate. Methionine is given because it drives homocysteine production via SAM and SAH, while in the absence of folate the other major means of metabolizing homocysteine, which is to methylate it back to methionine, is also blocked. Under these conditions, homocysteine accumulates.
In these animals PPMT expression was reduced by about 40 percent and PP2A methylation was decreased by about 75 percent. These changes were accompanied by about a 2.5-fold increase in phosphorylated tau in the brain and about a 3.5-fold increase in phospho-APP compared to wild-type mice on a normal diet. The findings “support the hypothesis that impaired Hcy metabolism and deregulation of critical methylation reactions can trigger the accumulation of phosphorylated tau and APP in the brain, a process that may favor neurofibrillary tangle formation and amyloidogenesis,” write the authors.
Heavy Methyl Encore—Protein Phosphatase 1 Promoter
Silencing DNA by methylation has generally been viewed as a permanent modification, ensuring that patterns of gene activity are passed from one cell generation to the next during mitosis. But the other two papers support the idea that in terminally differentiated neurons, reversible DNA methylation, including that of the protein phosphatase 1 (PP1) promoter, is linked to synaptic plasticity and the formation of memory. Writing in the March 15 Neuron, Courtney Miller and David Sweatt of the University of Alabama, Birmingham, report that DNA methylase activity is boosted in animals when new memories are being formed and that this leads to silencing of PP1, which can suppress memories. They also report that activation of the gene for reelin, a protein that helps remodel synaptic connections (see ARF related news story) and that marks neurons lost early in AD (see ARF related news story), is increased during memory formation by none other than demethylases—enzymes that remove methyl groups from DNA. The findings suggest that methylation and demethylation play a key role in how memories are formed and stored.
That idea is supported by data from Erminio Costa and colleagues at the University of Illinois, Chicago, who also found that demethylation of the gene for reelin and another protein, the 67 kDa glutamic acid decarboxylase, can be induced in mice by administering small molecules that interfere with the packaging of DNA in the nucleus. That finding is reported in the March 11 PNAS online. These new studies may not only change how we think about memory formation, but they suggest that DNA methylation, once considered permanent, is dynamic in neurons and might be exploited for therapeutic benefit.
The idea that covalent modification of the chromatin structure of DNA is involved in memories is not new. It has been established that acetylation of histone proteins, which form the chromatin scaffold upon which DNA is tightly wrapped, is linked to synaptic signal transduction. Sweatt and colleagues previously showed that activation of NMDA-type glutamate receptors leads to acetylation of histone H3 (Levenson et al., 2004), a modification that weakens the affinity of histones for nucleic acids and allows other proteins, such as those involved in gene activation, to access DNA. In fact, the histone acetyl transferase (HAT) activity of CREB binding protein, a key neuronal transcription factor, has been linked to that protein’s effects on memory (see ARF related news story), while boosting histone acetylation by inhibiting histone deacetylases enhances long-term memory as well.
Because of these links among histone acetylation, gene silencing, and memory, Miller and Sweatt wondered if methylation of DNA might have similar effects. DNA methylase activity is high in the adult mammalian brain and DNA methylation silences genes, in part by recruiting histone deacetylases. To specifically test this idea, Sweatt and colleagues inhibited DNA methyl transferases (DNMTs) in hippocampal slices. They found that this prevents induction of long-term potentiation, the activity-dependent strengthening of synapses that is crucial for learning and memory. They also found that these inhibitors reduced methylation of reelin DNA, an indication that methylation is reversible. These experiments were described last year (see Levenson et al., 2006). Now, Miller and Sweatt advance those observations by looking at methylation patterns in live mice during contextual fear conditioning, a paradigm where animals lean to associate a particular environment with an unpleasant stimulus, such as a mild shock.
The researchers report that levels of mRNA for methylases DNMT3A and DNMT3B, which are believed to be involved in de-novo methylation, are significantly increased in the hippocampus following contextual fear training. Furthermore, mice given DNMT inhibitors seemed to have trouble making memories, because when placed back into the fear context they froze in place much less frequently than did control animals.
How might DNA methylation affect mouse memories? Any number of methylation-prone DNA regions could be involved, so to narrow things down Miller and Sweatt looked at methylation of genes known to play key roles in memory. First they looked at the memory suppressor, protein phosphatase 1 (PP1), on the premise that silencing that gene might boost memory. Indeed, the researchers found that 1 hour after contextual fear training, methylation of the PP1 promoter region was increased by over 100-fold and mRNA levels of PP1 in the CA1 region of the hippocampus were slightly, though significantly reduced. For this effect the animals had to experience both the new context and the mild foot shock; alone, neither had any effect on methylation, indicating that a true memory must be formed for PP1 methylation to take place. Interestingly, DNMT inhibitors dramatically increased the fraction of PP1 promoters that were not methylated, again suggesting that demethylation may be just as important for regulation as methylation.
To address the role of demethylation, Miller and Sweatt measured how reelin DNA is altered by the learning paradigm. They found that after 1 hour of contextual fear training, reelin promoter methylation was decreased and reelin mRNA levels increased almost twofold. DNMT inhibitors led to an even greater demethylation of reelin DNA. Though DNA methylation has generally been considered a permanent modification, these results suggest that in neurons, at least, the process may be more dynamic.
The Rolling Histones
In the second paper, Costa and colleagues describe a slightly different approach to study reelin gene demethylation. They studied downregulation of reelin and the 67 kDa glutamic acid decarboxylase (GAD67) in mice treated chronically with the methyl donor methionine. The suppression of the two genes under these conditions is attributed to enhanced methylation, leading to the recruitment of histone deacetylases (HDACs), which in turn increase histone affinity for DNA, leading to gene silencing. First author Erbo Dong and colleagues wondered what might happen if they prevented that histone deacetylation.
After treating mice with methionine for a week, Dong and colleagues then gave the animals HDAC inhibitors and followed the pattern of DNA methylation. The researchers discovered that in the presence of these inhibitors demethylation of the two genes was accelerated as judged by the reduction in number of promoters that immunoprecipitated with MeCP2, a protein that binds to methylated DNA. The researchers suggest that the rapid demethylation could be due to either inhibition of a methylase or stimulation of putative demethylase activity, but they favor the latter scenario because a methylase inhibitor had no effect on the rate of demethylation.
All told, these findings point to a dynamic methylation/demethylation process that is linked to synaptic plasticity and memory formation. “An as yet unknown signaling pathway targets the nucleus and activates demethylases and DNMTs. This results in the demethylation of positive regulators of memory, such as reelin. HATs are then free to acetylate demethylated genes, releasing them from the transcriptional silencing induced by methylation. This leads to transcriptional activation of reelin and, likely, other memory-enhancing genes. Simultaneously, DNMTs target negative regulators of memory, such as PP1, for transcriptional silencing,” write Miller and Sweatt.
This new, though poorly understood regulatory mechanism may also yield new clues to various neurologic diseases such as fragile X mental retardation, Rett syndrome, and autism, which have been linked to DNA methylation. It could also lead to a better understanding of the memory losses associated with AD and other dementias.—Tom Fagan.
Sontag E, Nunbhakdi-Craig V, Sontag J-M, Diaz-Arrastia R, Ogris E, Dayal S, Lentz SR, Arning E, Bottiglieri T. Protein phosphatase 2A methyltransferase links homocysteine metabolism with tau and amyloid precursor protein regulation. J. Neuroscience. 2007, March 14;27:2751-2759. Abstract
Miller CA, Sweatt JD. Covalent modification of DNA regulates memory formation. Neuron. 2007, March 15;53:857-869. Abstract
Dong E, Guidotti A, Grayson DR, Costa E. Histone hyperacetylation induces demethylation of reelin and 67-kDa glutamic acid decarboxylase promoters. PNAS. 2007, March 13;104:4676-468. Abstract