21 November 2012. Franklin’s genius notwithstanding, death and taxes are not the only things that are certain. It’s pretty clear you can’t evade your genetics, either. But what about epigenetics? While tweaking the tax code might improve your fortune, changes that alter the outcome of your genetic code could improve your health. At the 42nd annual meeting of the Society for Neuroscience, held 13-17 October in New Orleans, Louisiana, researchers pored over epigenetic marks predisposing to AD and cognitive decline, and discussed how those might be targeted therapeutically. While the work seems promising, scientists cautioned that these are still early days, and many questions remain difficult to answer. “Epigenetics is a newly developing field, and much remains to be worked out,” said Paul Coleman, Banner Sun Health Research Institute, Sun City, Arizona. “Nevertheless, I believe it will turn out to be at least as important as genetics itself,” he predicted. With Suzana Petanceska from the National Institute on Aging, Bethesda, Maryland, Coleman co-chaired a mini-symposium on the role of epigenetics in the development and maintenance of human cognition.
Many research labs, including Coleman’s, investigate links between disease and DNA methylation, which typically occurs on cytosine-guanine dinucleotide (CpG) sequences. Methylation silences genes and can profoundly alter biology (see ARF related news story). However, faced with upwards of 20 million CpG sites in the human genome, how are scientists to finger those that are most important to the brain? At the SfN meeting, David Bennett, Rush University Medical Center, Chicago, Illinois, outlined a quantitative trait strategy to identify some that influence AD pathology and/or cognitive decline.
Bennett and colleagues used samples from the Religious Orders Study and the Rush Memory and Aging Project. These two longitudinal observational studies correlate dementia incidence with epidemiological and pathological data. They probed DNA samples taken from the postmortem dorsolateral prefrontal cortex with an Illumina beadset that detects methyl groups at ~485,000 sites in the human genome. Bennett reported that methylation on many of those sites is coordinated, i.e., when a given site is methylated, so are others nearby. The number of such methylation blocks, or mBlocks, in the whole genome could top 300,000 he said.
Turning to AD, Bennett reported significant correlations between methylation at 168 sites in the genome and the presence of neuritic plaques seen on autopsy of 759 subjects. The methylation site that most tightly associated with plaques does not lie near any known genes, but the next two strongest sites are close to each other and to the AGPAT6, ANK1, and NKX6-3 genes. None of those have been previously linked to AD. Two of the 168 methylation sites turned up in AD susceptibility loci indentified in recent genomewide association studies: BIN1 and ABCA7, which rank second and fourth in the AlzGene database, respectively. Looking at pathologic cases and controls, Bennett predicted that about 45 percent of the variance in plaque burden can be explained by methylation among the 168 sites.
Are the genes near any of these methylation sites up- or downregulated in AD? Using RNA microarrays to correlate gene expression with plaque burden in the Religious Orders and the Rush Memory and Aging samples, Bennett found in eight of the 168 loci there were eight genes with expression patterns that correlated with neuritic plaque burden. Assessing those eight genes against an independent set of samples from the Mayo Clinic, where four of the genes—AP3M2, ACACB, DDB1, and HSPB2—were differentially expressed in the hippocampus of AD samples compared to controls. HSPB2 expression was linked to astrocytes in the vicinity of neuritic plaques before, but the other genes had not been previously associated with AD. None appears to be linked to AD; however, a separate analysis that looked for correlation with cognitive decline revealed a methylation site near the α-2 macroglobulin gene, which has been genetically linked to AD in some studies (see AlzGene entry).
Dana Dolinoy and colleagues at the University of Michigan, Ann Arbor, adopted a similar approach but with smaller sample sets. These researchers probed postmortem frontal cortex samples from 12 late-onset AD patients and 12 cognitively normal age- and sex-matched controls. They used a smaller Illumina array that surveys ~28,000 CpG sites spanning almost 15,000 genes. At the SfN meeting, Dolinoy reported that, overall, she found no dramatic differences in methylation status between the AD and control samples. In both, there was a cluster of sites that was between 75 and 100 percent methylated, and a second cluster with less than 10 percent methylation. Dolinoy said that 948 CpG sites might be potentially associated with AD, while over 2,400 associated with age. She showed a “heat map” for the top 25 that indicated both hypo- and hypermethylated sites in the AD samples. Honors for being most strongly associated with AD went to a CpG in the promoter of the transmembrane protein 59 gene (TMEM59), which is reported to modulate expression and processing of APP (see Ullrich et al., 2010). The researchers found that TMEM59 methylation associated with AD in an additional three matched sample pairs. RNA and protein levels also correlated to methylation status (see Bakulski et al., 2012).
Researchers agreed that these unbiased approaches for scanning the genome have their shortcomings. For example, the arrays only scan a fraction of the number of potential methylation sites. They need a large amount of starting material, which is a problem because it means scientists have to use homogenates rather than defined cell types or single cells, said Coleman.
Philip Landfield’s group at the University of Kentucky, Lexington, has used laser capture microscopy to isolate and study single cells for transcriptional changes in AD tissue. Up- or downregulated gene expression could reflect epigenetic changes that predispose to disease, he said. Previously, Landfield’s group used microarray analysis to look for correlations between hippocampal gene expression and cognition or AD pathology. The scientists reported that several thousand genes may be regulated up or down as Mini-Mental State Exam (MMSE) scores declined, 89 of which correlated with MMSE score and neurofibrillary tangle burden (see ARF related news story). While that analysis relied on whole tissue sampling, Landfield reported at the SfN meeting that his group has revisited that analysis using laser capture to isolate grey matter from the CA1 region of the hippocampus (see Blalock et al., 2011). This time the scientists did not detect correlations with glial and growth factors that had been positive in the previous study, suggesting that those AD-related changes are specific to white matter. “The functional relevance of these findings remains a major question,” said Landfield. Genes involved in chromatin assembly and organization, including histone acetyltransferases and histone deacetylases, were upregulated in samples from people with incipient AD. The authors found that similar changes to the same epigenetic regulators occurred with normal aging. In a variation on the classic "who will guard the guards," Landfield noted that "the altered expression of epigenetic regulators in the hippocampus with aging and/or AD indicates that they are, in turn, modified by other unknown factors."
Other research groups are also taking more direct approaches. Farah Lubin and colleagues at the University of Alabama at Birmingham have looked at histone methylation to see if it plays a role in memory consolidation. Transcription drives long-term memory, and histone modifications can activate or repress genes (see ARF related news story on Gupta et al., 2010). Lubin decided to examine the involvement of the histone lysine methyltransferase G9a. This enzyme methylates lysine 9 on histone H3, which in turn recruits other proteins that repress transcription, including histone deacetylases (HDACs). There are indications that G9a may be important in some forms of plasticity associated with cocaine addiction (Maze et al., 2010), but whether it more broadly influences learning and memory was not clear.
Hints came from studying changes elicited by contextual fear conditioning, a type of learning that involves crosstalk between the hippocampus and the entorhinal cortex. At SfN, Lubin showed that a foot shock drove methylation on lysine 9 of H3 (H3K9) in this brain area. To test if the methylation was important for learning, she blocked the methyltransferase by injecting an inhibitor (BIX01294) into the brain an hour before fear conditioning. Animals given CA1 injections of BIX01294 froze half as often as did untreated mice when reintroduced into the conditioning environment a day later. This suggests that the methylase supports this type of learning, Lubin said. In contrast, injecting the methylase inhibitor into the entorhinal cortex boosted the animals’ memory of the foot shock. These opposite effects revealed a complex relationship between this epigenetic modification and memory, said Lubin. In keeping with this, the methylase inhibitor suppressed long-term potentiation (LTP) in the Schaeffer collateral pathway of the hippocampus. Interestingly, blocking G9a in the ERC raised methylation in the hippocampal CA1.
How might methylation work in this context? Lubin showed that BIX01294
infused into the ERC elevated histone methylation at the promoter for the catechol-O-methyltransferase (COMT) gene in the CA1. Despite its name, COMT is not involved in DNA or histone modification, but instead metabolizes catecholamines, including dopamine, which support long-term memory. This finding shows that epigenetic changes to histones, which could influence many different genes, can directly modulate genes involved in memory.
Stephen Haggarty, Massachusetts General Hospital, Boston, also spoke to the role of epigenetics in plasticity. Working with Li-Huei Tsai, at MIT, Haggarty and colleagues work to tease out the roles of HDACs in learning and memory. The brain contains many different HDACs. The researchers found earlier that HDAC2 suppresses memory in mice (see ARF related news story), and that this deacetylase is activated in mouse models of neurodegeneration (see ARF related news story). The researchers are looking for compounds to block HDAC2, though its similarity to HDAC1, which is neuroprotective, makes this difficult, Haggarty pointed out (see ARF related news story on Kim et al., 2008). Using combinatorial chemistry, Haggarty’s collaborators built a family of macrocyclic compounds (see Marcaurelle et al., 2010) and tested them for HDAC inhibition in an unbiased screen. Of 14 hits, none was selective against HDAC2. Haggarty reported some success in screens for HDAC1 activators. Prototypes also activated HDAC2, but only at higher concentrations, said Haggarty, raising hopes that specific HDAC1 activators can be developed.
Looking to a different therapeutic approach, Claes Wahlestedt from the University of Miami, Florida, talked about using antagoNATs, or antagonists to natural antisense transcripts. Wahlestedt pointed out that for many human genes, both sense and antisense transcription goes on at the same time; hence, the antisense transcript can potentially repress transcription of the sense strands. Scientists are still working out how this repression occurs, but it seems to involve epigenetic regulation, namely, modification to chromatin histones. Blocking antisense transcripts could, therefore, relieve this type of transcriptional repression, said Wahlestedt.
As an example, the Miami group focused on regulation of brain-derived neurotrophic factor. BDNF protects neurons from degeneration, and its loss with age may put people at risk for AD (see ARF related news story). The brain makes antisense BDNF strands, and Wahlestedt reported that antagoNATs to those transcripts boost expression of the sense transcript.
Working with collaborators, he generated several BDNF antagoNATs using locked nucleic acids. These are chemical analogs that resist degradation by nucleases. When the scientists added these antagoNATs to N2a neuroblastoma cells, they saw a robust increase in sense BDNF transcripts (see Modarresi et al., 2012). The strategy worked in vivo, too. Delivered by mini-osmotic pumps into the brain ventricles over four weeks, the antagoNATs drove up BDNF mRNA and protein across the forebrain. This increased neuronal survival and proliferation.
Wahlestedt said this antagoNAT strategy could be useful for neurodegenerative diseases, claiming the gains may go beyond BDNF. Of 250 different genes Wahlestedt looked at, 132 have natural antisense transcripts.
Overall, the symposium highlighted the angles researchers are taking to examine the role of epigenetics in normal brain function and in disease, from genomewide to candidate approaches. How do the pieces fit into a comprehensive model of learning, memory, and neurodegeneration? “That is a key challenge in the field,” said Coleman.—Tom Fagan.