In recent years, genome-wide association studies (GWAS) have identified 28 genomic regions that associate with risk for late-onset Alzheimer’s disease. Most of these hits are in non-coding regions, leaving scientists wondering how they influence disease. Now, researchers led by David Bennett, Rush University Medical Center, Chicago, report specific DNA methylation patterns in five of these loci in people who died with pathological AD. The postmortem analysis was published in the November 3 JAMA Neurology online. “These data clearly reinforce the implication that these five loci are involved in AD,” Jean-Charles Lambert, a geneticist at INSERM, Lille, France, wrote to Alzforum in an email. Lambert was not involved in the study. Though the results need to be replicated, this type of data might ultimately help determine which genes are responsible for GWAS signals, he wrote.

In DNA, methyl groups are added to cytosine (C) bases, often ones that lie next to a guanine (G) at a CpG site. The modification generally silences transcription, though sometimes it enhances it. Recently, Bennett’s group reported that methylation changes at 71 CpG sites across the genome tracked with amyloid plaque burden in postmortem brains (see Aug 2014 news story). In general, greater methylation accompanied more plaques. Interestingly, CpGs from the ABCA7 and BIN1 regions—which contain reported AD-risk variants—were among those found. The researchers wondered if other AD risk genes are abnormally methylated as well.

First author Lei Yu and colleagues analyzed tissue samples of 740 brain donors from the Religious Orders Study and the Rush Memory and Aging Project. On average, people were 88 years old when they died. Aβ plaque and tau pathology in 447 justified a pathological diagnosis of Alzheimer’s. The researchers extracted DNA from the gray matter in the dorsolateral prefrontal cortex—which includes neurons, microglia, endothelial cells, and other cell types. Using a commercial methylation array, Yu determined which of 500 or so sites within 100 kilobases of the AD-linked loci sported a methyl group.

The researchers first looked to see whether differences in methylation of CpGs in those loci associated with a diagnosis of AD. At each of five loci—SORL1, ABCA7, HLA-DRB5, SLC24A4, and BIN1—between two and 19 sites significantly correlated with disease. For the most part, more methylation correlated with a higher risk of AD. Next, the researchers asked if there was any relationship between methylation and plaques and tangles. Overall methylation in the five loci varied both with Aβ load and density of tau tangles, though not necessarily at the same nucleotides. To find out whether and how these changes in methylation translated to altered gene transcription, they looked at the level of mRNAs for each gene in the brain extracts. For most, transcription was correlated weakly, if at all.

The study is unique in that it combines knowledge about genetic influences on Alzheimer’s with methylation data to begin to nail down some of the key suspects’ pathology, said Marilyn Miller, National Institute on Aging, Bethesda, Maryland. “It’s pretty clear from these data that the disruption of methylation is a part of the pathology of Alzheimer’s disease,” she told Alzforum.

The findings do not totally mesh with GWAS data. In several cases, the methylation site that associated with AD pathology lay near, but not in, a gene implicated by GWAS. That was true for sites near HLA-DRB5, ABCA7, and BIN1. This finding fits with the idea that GWAS just approximate the area of the genome associated with disease, said Bennett. Further studies can help pinpoint the exact genes responsible he said. Co-author Philip De Jager, Brigham and Women’s Hospital, Boston, emphasized that these methylation effects occur independently of the previously reported AD-associated SNPs. The group disregarded CpGs in those SNPs to avoid technical artifacts in the data.

Taken together, the results suggest that altered methylation plays a role in Alzheimer’s disease, said Bennett. Since only some loci experience changes, the epigenome appears to affect AD in specific and complex ways, the authors wrote. An accompanying editorial by Bryan Traynor and Alan Renton, National Institute on Aging, praised the work, saying it provided compelling evidence implicating DNA methylation in the pathogenesis of Alzheimer’s. However, they wondered about causality. Since Yu and colleagues examined postmortem brains, methylation changes might have accumulated during disease, rather than initiating it. Bennett agreed that issue remains unresolved, but noted their previous study found that the methylation changes, including those in ABCA7 and BIN1, correlated with neuritic plaque burden in people who remained cognitively normal. That the new methylation hits lie near previously identified GWAS polymorphisms implies that they could drive disease, added De Jager.

Methylation usually alters gene transcription, so why did the methylation patterns that associated with AD have little to no impact on RNA expression? The researchers are unsure, but suspect that RNA degradation in the postmortem sample may have led to spurious results. In addition, many other factors, such as histone modification, feedback regulation, and microRNAs, contribute to levels of gene expression, said Bennett. Methylation is just one small piece of that puzzle. In future studies, he aims to figure out how methylation fits into the bigger picture of RNA expression by testing how other factors affect transcription.

Traynor and Renton also point out that the researchers were examining methylation in cells that had survived until the patient died. For that reason, their methylation status may not represent that of the cells that succumbed. Bennett agreed, and said future research will have to sort that out.—Gwyneth Dickey Zakaib


  1. The paper by Yu et al. is a complementary study of a very recent work published by the same team in Nature Neuroscience (De Jager et al., 2014). In the first study, the authors studied the epigenome in the brains of 704 autopsied samples from two very well-characterised cohorts. The authors established that levels of methylation in specific genomic regions can be significantly associated with neuritic plaque burden in the brain.

    In this new study, the authors reanalysed their epigenomic data to determine whether methylation levels may be associated with the risk of developing AD. Rather than describing a genome-wide analysis of their complete epigenomic data, they restricted this paper to the study of the known genes/loci reported to be potentially involved in the AD monogenic or sporadic forms. They observed that the global level of methylation in five loci (out of 28 tested) was associated with AD risk. Remarkably, for these five loci, this global level of methylation was also associated with Aβ load and tau tangle density.

    These data clearly reinforce the implication that these five loci are involved in AD. Indeed, even if this work utilizes a candidate gene approach (with the risk of “tautological reasoning”), the data point out that the expression level of the gene(s) located within these loci may be particularly important for AD pathophysiology because the epigenomic signal is likely fully independent of genetic variations associated with AD risk in genome-wide association studies (GWAS). This is particularly true for the BIN1 locus, for which some evidence suggests that its expression might be important for tau pathology (Chapuis et al., 2013).

    Yu et al.'s work may be also very useful at another level: A GWAS-defined locus can encompass a large number of genes. On a pure genetic basis, it can be difficult to determine which one is responsible for the GWAS signal, all the more since the functional variants can sometimes be quite far away from the real causal gene. With this in mind, we notice two interesting points concerning the HLA-DRB5 and SLC24A4 loci. The most relevant and dense methylation signals are not in the genes exhibiting the highest level of association with AD risk in GWAS. This is particularly true for the signal in the HLA region: The epigenomic data mainly pointed out HLA-DRA, which is also within the GWAS-defined locus. Notably, transcriptomic analysis suggests this gene may be differentially expressed in the brains of AD cases compared to controls (Chapuis et al., 2009). Extending the epigenetic analysis to the complete HLA region would help decipher this highly complex genetic region. Similarly, whereas the best GWAS signal was close to SLC24A4, methylation variations associated with AD risk seemed to point out RIN3. At a biological level, SCL24A4 is involved in iris and skin pigmentation, whereas RIN3 would be a more convincing candidate since it is involved in neuronal function, as noticed in our initial GWAS paper (Lambert et al., 2013) and by Yu et al. 

    Taken as a whole, these data are clearly interesting and represent a new piece in the complex puzzle of AD genomics. However, as mentioned in the companion editorial in JAMA neurology, “Like all good science, the article by Yu et al. has limitations and raises compelling questions” (Traynor and Renton, 2014).  Importantly, these data of course need to be replicated: Using a powerful technology and an appropriate design does not, unfortunately, avoid the classic methodological biases in genomic epidemiology, e.g., sampling variations and underpowering. Since a candidate approach was used, it is difficult to assess the possibility that this study generated false positives. Though the authors took some precautions by performing permutation analyses, knowing the value of a genomic control parameter such as the λ inflation factor, which reflects potential false positives, would have been of interest. 

    As for all observational studies, it is not possible to answer to the crucial question: cause or consequence? For instance, it cannot be excluded that these differences in methylation are late responses to earlier dysregulation of gene expression.

    Finally, statistical approaches cannot replace biological evidence and it will be necessary to understand the real impact of these methylation fluctuations on gene expression in different types of cells and in a pathophysiological context.


    . Alzheimer's disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci. Nat Neurosci. 2014 Sep;17(9):1156-63. Epub 2014 Aug 17 PubMed.

    . Increased expression of BIN1 mediates Alzheimer genetic risk by modulating tau pathology. Mol Psychiatry. 2013 Feb 12; PubMed.

    . Transcriptomic and genetic studies identify IL-33 as a candidate gene for Alzheimer's disease. Mol Psychiatry. 2009 Nov;14(11):1004-16. PubMed.

    . Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet. 2013 Dec;45(12):1452-8. Epub 2013 Oct 27 PubMed.

    . Exploring the epigenetics of Alzheimer disease. JAMA Neurol. 2015 Jan;72(1):8-9. PubMed.

    View all comments by Jean-Charles Lambert

Make a Comment

To make a comment you must login or register.


News Citations

  1. Alzheimer’s Brains Mottled with Epigenetic Changes

External Citations

  1. SORL1
  2. ABCA7
  3. HLA-DRB5
  4. BIN1

Further Reading


  1. . Elevation of peripheral BDNF promoter methylation links to the risk of Alzheimer's disease. PLoS One. 2014;9(11):e110773. Epub 2014 Nov 3 PubMed.
  2. . Epigenetic modulation of Cdk5 contributes to memory deficiency induced by amyloid fibrils. Exp Brain Res. 2015 Jan;233(1):165-73. Epub 2014 Sep 19 PubMed.
  3. . Alzheimer disease: AD-susceptible brain regions exhibit altered DNA methylation. Nat Rev Neurol. 2014 Oct;10(10):548. Epub 2014 Sep 2 PubMed.
  4. . Epigenetic modifications in Alzheimer's disease: cause or effect?. J Alzheimers Dis. 2015;43(4):1169-73. PubMed.
  5. . 24-hour rhythms of DNA methylation and their relation with rhythms of RNA expression in the human dorsolateral prefrontal cortex. PLoS Genet. 2014 Nov;10(11):e1004792. Epub 2014 Nov 6 PubMed.

Primary Papers

  1. . Association of Brain DNA methylation in SORL1, ABCA7, HLA-DRB5, SLC24A4, and BIN1 with pathological diagnosis of Alzheimer disease. JAMA Neurol. 2015 Jan;72(1):15-24. PubMed.
  2. . Exploring the epigenetics of Alzheimer disease. JAMA Neurol. 2015 Jan;72(1):8-9. PubMed.