Aging is the strongest risk factor for Alzheimer’s disease, but exactly what it is about aging that underlies the pathology of AD, or cognitive decline in general, is a question that still confounds researchers. One theory, for example, suggests that neurons, which are terminally differentiated and so don’t undergo cycles of DNA replication and proofreading, are slouches when it comes to DNA repair and accumulate a disproportionate burden of genetic mutations over time (see ARF recent live discussion ). Now, in the June 9 online version of Nature, Bruce Yankner and colleagues at Children’s and Brigham and Women’s hospitals, both in Boston, provide experimental evidence to back up this hypothesis. Going further, they offer a new conceptual framework that links DNA damage in gene promoters to altered expression of genes in the aging brain. They also articulate a new concept of “vulnerable genes” that contain sequences which predispose to DNA damage. Some of these vulnerable genes are known to play critical roles in learning and memory.

First author Tao Lu and colleagues examined postmortem frontal cortex samples taken from normal donors aged 26 to 106 years. Lu analyzed these samples, 30 in all, with Affymetrix gene arrays to measure how the expression of approximately 11,000 genes changed with age. Compared with people younger than 42, Lu found that in people older than 76, about 4 percent of genes are differentially expressed in the frontal cortex—some up- and some downregulated. The latter included genes involved in synaptic plasticity (e.g., NMDA and AMPA receptor subunits), long-term potentiation (e.g., calmodulin CaM kinase II and protein kinase C), vesicular transport (e.g., synaptobrevin, dynein, and Rab GTPases), and microtubular function (MAP1B, MAP2, and tau), while upregulated genes included those that mediate stress responses, such as chaperones and antioxidant enzymes. DNA repair genes, such as the 8-oxoguanine DNA glycosylase, which excises damaged nucleotides, were also upregulated. The paper includes a table of differentially expressed genes representating functional categories; a complete list is provided in the Supplementary Information.

The induction of DNA repair mechanisms led Yankner and colleagues to wonder if this pattern of gene expression may be tied to DNA damage. To test this, Lu et al. devised an assay that could resolve DNA damage in specific gene sequences. They isolated genomic DNA under conditions that prevent in-vitro oxidation and incubated it with enzymes that specifically excise damaged bases, particularly 8-oxoguanine. Then, using the real-time polymerase chain reaction, they were able to identify and quantify stretches of damaged DNA by their failure to amplify (excision of the damaged base creates a single strand break that inhibits PCR).

Lu and collegues then applied this approach to brain samples spanning the entire adult age range and found that DNA damage was present in every gene examined in aging brain. The twist is that damage was greatest in promoter regions. These regions may be hardest hit because they are not subject to transcription-coupled repair, the major repair mechanism in mature neurons. (Promoter regions are usually repaired by a process that requires transit through the cell cycle.) Of 30 promoters examined, many showed an age-related increase in DNA damage by age 40, and all genes showed damage by 70. The clincher came when the researchers compared promoter damage with age-related changes in gene expression: They found that those genes that were downregulated had significantly more damage than genes that were upregulated or unchanged. The results suggest an association between promoter DNA damage and age-related changes in gene expression, the authors note.

To determine whether there is a causal relationship between damage and gene expression, Lu and colleagues examined DNA damage in cultured human neuroblastoma cells and primary human cortical neurons treated with a mild oxidative stress (hydrogen peroxide and ferric chloride) that did not kill the cells. Here again, genes that were downregulated in aging brains e.g. the tau and calmodulin 1 genes, were damaged and transcriptionally repressed to a greater extent than genes that did not change in the aged brain.(e.g. GAPDH and beta tubulin). To validate this correlation between vulnerable genes in vitro and in vivo, the authors examined gene promoters in human cortical neurons subjected to oxidative stress. Transfection of the DNA repair enzyme reversed both the damage and transcriptional downregulation .

Why do some promoters appear more vulnerable than others? To approach this question, Lu et al. cloned the promoters from brain genomic DNA into luciferase reporter plasmids and damaged them in vitro by treatment with hydrogen peroxide or irradiation with ultraviolet light. The promoters that show increased damage and reduced transcription in the aged brain also showed increased damage and reduced transcription in vitro. Furthermore, when these promoter constructs were transfected into neuronal cells they showed reduced base excision DNA repair. These experiments suggest that the vulnerability of these genes is a function of their DNA sequence, as opposed to some difference in signaling pathways or cellular responses to stress.

Other experiments in the paper examined the reduction in expression of some mitochondrial genes that the authors observed in the aging brain. They mimicked these gene expression changes in cultured neuronal cells using siRNA, and found that this increased DNA damage to vulnerable nuclear genes. These findings suggest that one source of DNA-damaging free radicals in the aging brain might be dysfunctional mitochondria.

The “findings suggest that accelerated DNA damage may contribute to reduced gene expression in the human brain after age 40,” and that “genome damage may compromise systems that subserve synaptic function and neuronal survival” write the authors. The also authors suggest that this could be a starting point for trying to understand why aging of the brain is the major risk factor for AD —Tom Fagan

Q&A with Bruce Yankner

Q: How does this view of aging fit in with the amyloid hypothesis of AD?
A: The amyloid hypothesis does not provide a clear explanation for why age is the major risk factor for Alzheimer's disease. This also belies our basic ignorance about the initial events that underlie sporadic AD. This gap in our understanding of AD reflects, in my view, a major gap in our understanding of the molecular basis of aging in the human brain.

Q: Do you think that the very nature of the samples, i.e., postmortem, may factor into the reduced expressions and DNA damage you observed?
A: We also examined expression and DNA damage in intracortical biopsy samples from elective neurosurgical procedures. Although the number of these samples was limited, the results were quite similar to those obtained for age-matched postmortem samples. To directly assess the role of postmortem interval, we performed linear regression analysis among all the postmortem intervals and expression changes in the two age-related gene clusters overall, or for 20 individual age-downregulated and 20 individual age-upregulated genes. Neither of these analyses showed a statistically significant relationship between postmortem interval and expression level. This may reflect the fact that we did not use tissue from brains with long postmortem intervals. For the DNA damage assays, it was important to isolate DNA using conditions that prevent in-vitro oxidation, e.g., including a free-radical scavenger and purging buffers with nitrogen.

Q: Are similar changes occurring in non-brain tissues?
A: We plan to examine blood cells and skin fibroblasts, but do not yet have results. We have examined other brain regions, specifically hippocampus, and find similarities that suggest to us that there may be a global program of brain aging that is superimposed on region-specific changes.

Q: Will it ever be possible to repair age-related damage to DNA?
A: We found that DNA repair enzymes can restore the expression of vulnerable genes damaged in cell culture, raising the whimsical possibility that some aspects of brain aging may be reversible.

Q: What about the genes that are upregulated with aging? Could DNA damage explain this, too?
A: Some genes are upregulated in the aging brain with a time course that parallels the time course of DNA damage. One possible mechanism could involve inactivation of motifs for transcriptional repressors. Alternatively, DNA damage could affect histone modification and interfere with gene silencing, although we don’t have evidence for this.

Q: What about AD and other diseases. Will you look at tissue samples to estimate how DNA damage may correlate with neurodegeneration?
A: Experiments along those lines are ongoing.

Q: Have you any theories as to why some promoters are affected and not others?
A: This is a fundamental mechanistic question. Our preliminary studies raise the possibility that specific guanine-rich motifs may be “hot spots” of oxidative DNA damage. This may be a function of the specific sequence as well as the GC content. For example, the beta tubulin promoter, which is relatively resistant to damage, and the CaM1 promoter, which is more sensitive, have similar overall guanine content, but the spatial distribution of GC-rich sequences relative to the transcription start site differs.


  1. Yankner and his colleagues have done an excellent job of showing that aging is a cause of cognitive decline. Although this has been thought for a long time, it has not been directly supported until very recently with the latest technology. The study shows that the neurodegenerative process begins early in life. This suggests that aging—that is, time—may underlie may problems in the AD brain i.e., apoptosis, radicals, ion inbalance.

    Of great interest to me is that, among others, the genes involved in Ca2+ signaling are down-regulated. These include: Ca2+ channels, pumps, calmodulins, CaM kinases, PKC, calcineurin, etc. Some of them have been previously shown on an individual basis, but the beauty of this study is to demonstrate this point in a comprehensive and non-biased way. If these key elements are down, then what about the long-held dogma that Ca2+ signaling is going up and up throughout aging and AD?

    Another important finding of the study is age-related down-regulation of several protein kinases, especially CDK5. This is in contrast to current enthusiasm that activation of this kinase causes tau hyperphosphorylation. Now, if CDK5 is actually down, then tau hyperphosphorylation would have to be explained by some other models. One of them may be by a decrease in phosphatase activity. The down-regulation of calcineurin observed here fits in with the idea.

    Unfortunately, missing from the study is the fate of many proteases. Half of us make a living by working on these enzymes, as they must be important in AD. My particular interest is protease calpain. Why? As we argued, it may be difficult to explain the origin of plaques and tangles without focusing on this protease (1).

    Hopefully, future studies will also tell us the fate of APP and presenilins. Presenilin mutations cause familial AD, but whether they do so through gain or loss of function is a key issue. If gene arrays can tell us their changes in aging, this will help greatly in figuring out their mechanism in the disease.

    Overall, Lu et al. have raised several important, but also painful, questions for us. Answering these questions is necessary if we are to understand AD scientifically, even though research can go a long way by ignoring them.


    . Stimulation of beta-amyloid precursor protein alpha-processing by phorbol ester involves calcium and calpain activation. Biochem Biophys Res Commun. 2004 Apr 2;316(2):332-40. PubMed.

  2. This recent study provides novel insights into the biology of aging in human brains. However, the comment by Dr. Ming Chen posted July 12 is somewhat misleading. Dr. Chen states that the “age-related downregulation of several protein kinases, especially Cdk5” is very important, because this finding is in “contrast to current enthusiasm that activation of this kinase causes tau hyperphosphorylation.” As such, Dr. Chen concludes that, “if Cdk5 is actually down, then tau hyperphosphorylation would have to be explained by some other models.”

    In fact, the study by Lu et al. does not demonstrate downregulation of Cdk5. Instead, the Cdk5 activator p35 is significantly downregulated in aging brains. The misunderstanding of this point confounds the interpretation of this data by Dr. Chen. Furthermore, Lu et al. suggest that tau is also downregulated, so hyperphosphorylation of tau is unlikely to be relevant, according to Dr. Chen’s argument.

    To date, there is no evidence to suggest that Cdk5/p35 is neurotoxic. In fact, physiological Cdk5/p35 activity has been established as an important regulator of synaptic plasticity, learning, and memory (Li et al., 2001; Fischer et al., 2002; Fischer et al., 2003). Therefore, downregulation of p35 in aging brains is in agreement with the current knowledge about the role of Cdk5/p35 in the adult brain.

    It should be noted that tau is a very poor substrate of Cdk5/p35 in vivo and in vitro (Patrick et al., 1999; Van den Haute et al., 2001; Hashiguchi et al., 2003). On the other hand, the Cdk5/p25 kinase has been shown to be a potent tau kinase. P25 is produced via calpain-mediated cleavage of p35. P25 causes deregulation of Cdk5 activity, as p25 is more stable and displays altered subcellular localization. Thus, the levels of p25 in vivo cannot be assessed by transcriptional profiling, and decreased p35 levels do not necessarily lead to reduced Cdk5 activity in human brains. Notably, we have shown that low levels of p25 production are able to partially compensate the phenotype of p35-deficient mice (Patzke et al., 2003), suggesting that p25 can compensate for loss of p35. This raises the possibility that downregulation of p35 in aging brains may initially induce p25 production as a compensatory effect, which under certain circumstances leads to p25 accumulation and neurodegeneration. This possibility warrants further investigation.


    . Cyclin-dependent kinase 5 is required for associative learning. J Neurosci. 2002 May 1;22(9):3700-7. PubMed.

    . Regulation of contextual fear conditioning by baseline and inducible septo-hippocampal cyclin-dependent kinase 5. Neuropharmacology. 2003 Jun;44(8):1089-99. PubMed.

    . Truncation of CDK5 activator p35 induces intensive phosphorylation of Ser202/Thr205 of human tau. J Biol Chem. 2002 Nov 15;277(46):44525-30. PubMed.

    . Regulation of NMDA receptors by cyclin-dependent kinase-5. Proc Natl Acad Sci U S A. 2001 Oct 23;98(22):12742-7. PubMed.

    . Gene regulation and DNA damage in the ageing human brain. Nature. 2004 Jun 24;429(6994):883-91. PubMed.

    . Conversion of p35 to p25 deregulates Cdk5 activity and promotes neurodegeneration. Nature. 1999 Dec 9;402(6762):615-22. PubMed.

    . Partial rescue of the p35-/- brain phenotype by low expression of a neuronal-specific enolase p25 transgene. J Neurosci. 2003 Apr 1;23(7):2769-78. PubMed.

    . Coexpression of human cdk5 and its activator p35 with human protein tau in neurons in brain of triple transgenic mice. Neurobiol Dis. 2001 Feb;8(1):32-44. PubMed.

    View all comments by Li-Huei Tsai

Make a Comment

To make a comment you must login or register.


Other Citations

  1. ARF recent live discussion

Further Reading

No Available Further Reading

Primary Papers

  1. . Gene regulation and DNA damage in the ageing human brain. Nature. 2004 Jun 24;429(6994):883-91. PubMed.