How do Aβ oligomers cause synapses to wither in Alzheimer’s disease? In the April 10 Neuron, researchers led by Franck Polleux at the Scripps Research Institute, La Jolla, California, propose a pathway that leads from Aβ oligomers via two kinases, including AMP-activated kinase (AMPK), to tau phosphorylation and eventually to synapse loss. In a mouse model of AD, blocking either kinase or preventing tau phosphorylation at a single site prior to development of plaques preserved dendritic spines. The data suggest that this pathway plays a critical role in the synaptotoxic effects of Aβ, Polleux told Alzforum. In addition, “It is one of the first really clear kinase pathways to directly link Aβ oligomers to tau phosphorylation,” he said. It remains to be seen whether blocking this pathway would stop other aspects of AD pathology, such as amyloid accumulation and neuron loss, he noted. The findings, all from mouse studies, could have implications for proposed AD trials of metformin, an AMPK activator.

Other researchers found the data intriguing. “The paper makes a compelling case linking AMPK activation to spine degeneration in vivo,” Sung Yoon at Ohio State University, Columbus, told Alzforum.

AMPK is a key metabolic sensor in many cell types. It drew Polleux’s attention because it has been shown to be highly activated in neurons of AD brains, as well as in several other tauopathies, and other neurodegenerative diseases, including amyotrophic lateral sclerosis, and Huntington’s (see Vingtdeux et al., 2011). A previous study by Yoon found that Aβ oligomers switch on AMPK, which in turn ramps up Aβ production in an apparent feed-forward loop (see ARF related news story). Meanwhile, researchers led by David Carling at Imperial College London, U.K., reported that after Aβ activates AMPK, the kinase phosphorylates tau at several sites (see Thornton et al., 2011). While these findings implicate AMPK in AD pathology, other work contradicts this, showing that the kinase stimulates autophagy and helps degrade Aβ in APP/PS1 mice (see Vingtdeux et al., 2010; Vingtdeux et al., 2011).

To study the role of AMPK, first author Georges Mairet-Coello prepared synthetic Aβ oligomers using the method developed at William Klein's lab at Northwestern University, Evanston, Illinois (see Lambert et al., 1998). They confirmed that this preparation activated AMPK in mouse hippocampal and cortical neuronal cultures. As expected, after 24 hours of exposure to 1 μM Aβ, the neurons lost dendritic spines. The authors prevented this loss by inhibiting calcium/calmodulin-dependent kinase kinase 2 (CAMKK2), which normally turns on AMPK. Likewise, neurons from CAMKK2 or AMPK knockout mice kept their synapses when exposed to Aβ. This suggested that both kinases are necessary for Aβ to poison synapses. The authors then demonstrated that activation of this pathway even in the absence of Aβ sufficed to wither spines. Overexpression of CAMKK2 or AMPK, or activation of AMPK by several pharmacological means, mimicked the toxic effects of Aβ, they report.

Polleux and colleagues wondered if they could prevent synapse loss in vivo by blocking this pathway. They used in utero electroporation to express a dominant-negative, kinase-dead version of either CAMKK2 or AMPK in hippocampal neurons of embryos of the APP transgenic J20 line. After the embryos were born, the scientists counted spines in young mice at three months of age. While untreated transgenics had lost a significant fraction of their spines compared to wildtype controls, the mice that expressed the “dead” kinases, which block the endogenous active ones, had normal numbers of spines. At this age, the animals do not yet have amyloid plaques but they do overexpress mutant human APP and generate oligomeric Abeta.

How might AMPK be damaging spines? The authors linked AMPK activation to tau pathology. First they confirmed previous findings that AMPK phosphorylates tau at serine residue S262, in tau’s microtubule-binding region. Phosphorylation at this location makes the protein detach from microtubules, and has been hypothesized to prime pathological hyperphosphorylation (see Biernat et al., 1993; ARF related news story on Nishimura et al., 2004). When the authors expressed a mutated form of tau, S262A, which cannot be phosphorylated at this site, spines stayed healthy after exposure to Aβ both in vitro and in vivo. Tau phosphorylation at S262 therefore seems to mediate the synaptotoxic effects of Aβ, the authors conclude. Previous studies reported that early in AD, free tau wanders into dendrites, where it overexcites and poisons synapses through Fyn kinase (see ARF related news story on Ittner et al., 2010; ARF related news story on Zempel et al., 2010; Hoover et al., 2010).

In future work, Polleux will cross AMPK knockout mice with several different AD mouse models to examine the effect of chronically suppressing the kinase. Besides protecting synapses, would this approach prevent other features of AD, such as amyloid accumulation, neuron loss, and cognitive decline? Polleux will also test whether AMPK acts through other downstream mechanisms besides tau phosphorylation. If blocking AMPK can preserve synapses, does this have therapeutic potential? Polleux does not believe AMPK itself would make a good therapeutic target because it does so many things in different cell types. CAMKK2 might make a better drug target, as it is highly enriched in neurons, he suggested.

While these findings paint activated AMPK as a cause of AD pathology in J20 mice, previous studies found that turning on this kinase clears amyloid in APP/PS1 mice, suggesting the kinase can have divergent effects on synapses and amyloid. Sylvain Lesné at the University of Minnesota, Minneapolis, told Alzforum it will be interesting to test whether AMPK activation also curbs amyloid plaques in the J20 mice. For her part, Yoon suggested that the effects of AMPK may depend on environmental triggers. She pointed to a recent paper that found that the AMPK activator metformin affects worms differently depending on what bacteria the animals eat (see Cabreiro et al., 2013).

The question of what AMPK does has important clinical implications, as metformin is a widely-used treatment for Type II diabetes that is being considered for AD. Led by Jose Luchsinger at Columbia University, at least one phase 2 trial in amnestic MCI has already been completed, although no results have been reported (see also Luchsinger, 2010). “Before metformin goes into large clinical trials in the U.S., I think we need to test whether AMPK activation is beneficial or detrimental,” Yoon said. Polleux agreed, noting that, “In our hands, any drug, including metformin, that activates AMPK is as synaptotoxic as Aβ.” The use of metformin in people should be re-evaluated to make sure it is not harming cognition, Polleux suggested.

Lesné told Alzforum that the current results need repeating to find out how central the proposed mechanism is to AD pathogenesis. “There is pretty good evidence that there’s probably not just one pathway connecting Aβ oligomers to tau dysfunction and phosphorylation, but multiple pathways,” he said. He also would have liked to see the authors more extensively characterize the Aβ oligomers they used in vitro. Researchers in the field recommend using multiple rigorous techniques and oligomer-specific antibodies to determine the exact species of Aβ used in experiments (see Alzforum webinar). This allows other labs to reproduce findings, which has been lacking in Aβ oligomer research. Likewise, Lesné said, it would be interesting to know precisely what mixture of Aβ species is present in the three-month-old J20 mice, and what effect each type has on tau. “The challenge ahead lies in determining which Aβ oligomer is acting in which pathway, and what consequences each has on tau phosphorylation, oligomerization and aggregation,” he told Alzforum.—Madolyn Bowman Rogers.

Mairet-Coello G, Courchet J, Pieraut S, Courchet V, Maximov A, Polleux F. The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Aβ oligomers through tau phosphorylation. Neuron. 2013 Apr 10. Abstract


Make a Comment

To make a comment you must login or register.

Comments on News and Primary Papers

  1. I would like to point out what I believe constitutes misinterpretations/misconceptions related to oligomeric Aβ found in this article.

    In this report, the authors claimed to detect amyloid-β in brain lysates of APP transgenic J20 mice using the antibody 6E10 (Fig. 4). This antibody is specific to the human form of Aβ as well as its precursor molecule APP and some of the proteolytic cleavage products of APP, namely CTF-β. The authors claim that they detected Aβ by Western blot analyses as a band running at ~13 kDa. In fact, monomeric Aβ should be detected at 4 kDa and not at 13 kDa as reported in Figure 4B (unless this represents trimers, but that would require using specific Aβ antibodies or samples which do not contain APP-CTFs to be certain). It is likely that the 6E10 antibody detected the carboxyl terminal fragment of β-secretase-cleaved soluble APP (sAPPβ), which has a molecular weight of 12-13 kDa.

    In addition, the histogram (Fig. 4C) illustrating the quantification of these results depicts the levels of human APP and alleged Aβ (which I believe should be labeled as CTF-β) as the percentage of the signal detected in WT mice. Since the non-transgenic mice do not express human proteins, the signal detected with 6E10 in those animals must correspond to noise or non-specific signals, making it impossible to conclude that expression of APP is ~100-fold greater in J20 compared to WT mice. Using 4G8 would be a better choice since it detects both mouse and human APP and its derivatives.

    In the absence of liquid phase assays, it is difficult to prove that oligomeric Aβ, and in particular which form oligomeric of Aβ, exists in the animals used in this study. I believe this is a problem that plagues our field, as illustrated by the editorial published in Nature Neuroscience in April 2011 and by the Webinar hosted by Alzforum, which I had the privilege to be part of (ARF Webinar).

    Furthermore, the authors show that the AMPK pathway is activated in four-month-old J20 mice compared to wild-type controls. At eight to 10 months, when Aβ deposition is already prominent (Mucke et al., 2000) and when the Aβ concentration rises exponentially, the activation level of the AMPK remained similar to that detected in younger J20 mice (roughly a ~150 percent activation) (Fig. 4F, G). This result indicates that the AMPK activation observed does not depend on age or Aβ in vivo.

    Finally, the authors did not address whether the AMPK pathway activated in vivo was modulating the metabolism of APP/Aβ, as suggested earlier by the Marambaud group (Vingtdeux et al., 2010; Vingtdeux et al., 2011).

    In summary, I'm not convinced that the presence of oligomeric Aβ assemblies was established in vivo in these animals, or that they cause activation of AMPK (this is particularly important to establish in order to draw a parallel with the studies done in vitro). While APP and its C-terminal fragments may be responsible for the activation of AMPK-CaMKK2 in vivo, that does not appear to be age dependent—a cardinal feature of Alzheimer's disease.

    View all comments by Sylvain Lesne
  2. We would like to thank Dr. Sylvain Lesne for his interest in our work and for his comments. We would like to clarify some of the points he raised concerning our report, since we disagree with some of his statements.

    Regarding Aβ detection, we used the 6E10 antibody to monitor the level of APP transgene expression at the age we performed our analysis of spine density in vivo (i.e., three months postnatal), but our intention was not to characterize the presence of Aβ1-42 oligomers at this age in this mouse strain. They have been extensively characterized previously by others (see below). The 6E10 antibody is known to recognize human APP, CTF-β, and Aβ. We acknowledge that the band running below 15 kDa should be considered as a mixture of CTF/Aβ since tissues were homogenized in RIPA buffer. Regarding the APP signal at ~100 kDa, we could clearly detect a faint signal in the wild-type mice (after overexposure of the membrane), suggesting that the antibody can cross-react with mouse APP. This is further supported by the detection of endogenous CTF/Aβ in wild-type control animals with this antibody (Fig. 4B).

    Dr. Lesne suggests that liquid phase assays are essential to prove the presence of oligomeric Aβ. The composition of Aβ oligomeric forms in the J20 mouse line was characterized previously by Dr. Lennart Mucke’s group (Cheng et al., 2007; Meilandt et al., 2009). In fact, Dr. Lesne is a coauthor on one of these two papers. Mucke's group showed that, in addition to CTF-β fragments, these mice present high levels of Aβ*56 at three to four months, Aβ1-42 dimers were detected from 17 months, and Aβ1-42 trimers were assessed in five- to six-month-old animals.

    We agree that AMPK activation does not seem to correlate with increased Aβ levels in vivo, but this doesn’t mean that Aβ does not activate AMPK in vivo. The absence of further increase in AMPK activation at later ages could simply reflect a threshold effect where AMPK activation peaks when the Aβ oligomers (or other APP cleaved products) reach a certain concentration. Other important pT172-AMPK changes, for example, in localization at the synapse, could go undetected using this simple Western blot approach.

    We cautiously mentioned in our report that inhibiting the CAMKK2-AMPK pathway provides protective effects from spine loss in the J20 mouse line at an early age (three to four months), when loss of synapses are detected but prior to the appearance of plaques. However, we acknowledge that further experiments will be needed to determine if inhibiting this pathway is relevant at later stages of the disease in this hAPP-J20 mouse model as well as in other AD mouse models.

    We did not address the question of whether the AMPK pathway activated in vivo was modulating the metabolism of APP/Aβ, since this was not the main point of our study. Recently, several reports, besides our own, indicated that Aβ42 oligomers activate AMPK in vitro (Yoon et al., 2013; Son et al, 2011; Thornton et al., 2011). Our goal was to determine if the CAMKK2-AMPK pathway plays a functionally important role in mediating the synaptotoxic effects of Aβ in vivo, paralleling our in-vitro findings, which demonstrated that inhibition of this pathway protects against the synaptotoxic effects of Aβ42 oligomers. Therefore, we employed a strategy to analyze the role of the pathway at the single-cell level and determine if the protection against Aβ42 conferred by blocking the CAMKK2-AMPK pathway is cell autonomous.

    The role of AMPK in APP/Aβ metabolism in vivo is definitely another interesting question that implies non-cell-autonomous/paracrine effects on dendritic spine maintenance, and is indeed still controversial. Studies from Marambaud’s group have suggested that AMPK activation might decrease APP cleavage/Aβ production or increase Aβ clearance through autophagy (Vingtdeux et al., 2010; Vingtdeux et al., 2011). On the other hand, other reports indicate that AMPK overactivation increases Aβ generation through β-secretase transcriptional upregulation (Chen et al., 2009) and/or mTOR-dependent protein synthesis inhibition (Yoon et al., 2012). Solving this discrepancy will be interesting and challenging. A definitive answer will be obtained by careful analyses of mouse models of AD in which the catalytic activity of AMPK is genetically deleted. Since there are two isoforms of AMPK-α in mammals, generation of double conditional knockouts for AMPK-α1 and -α2 crossed with an appropriate Cre-driver and the hAPP-J20 transgenic mice would be required (quadruple transgenic mice). This was clearly beyond the scope of our report. However, we agree that this question is interesting, and we (and others) are currently performing some of these experiments.

    In summary, the presence of oligomeric forms of Aβ42 as well as other forms of Aβ (Aβ*56 and CTF-Aβ) have been characterized previously in this hAPP-J20 transgenic mouse strain (see Cheng et al., 2007; Meilandt et al., 2009) and are very likely to contribute to AMPK activation in vivo, as we and others have shown in vitro. We combined in-vitro and in-vivo approaches to 1) ascertain the specific effects of Aβ42 oligomers (in vitro), and 2) verify in vivo that in the well-characterized hAPP-J20 model that produces Aβ42 oligomers (Cheng et al., 2007; Meilandt et al., 2009), blocking CAMKK2 or AMPK function protected neurons from the synaptotoxic effects detected at early stages of disease progression. It will indeed be interesting to explore if this CAMKK2-AMPK pathway also mediates the potential synaptotoxic effects of other forms of Aβ (Aβ*56 and CTF-Aβ).

    View all comments by Georges Mairet-Coello


News Citations

  1. Vicious Cycle? Aβ Triggers Cell Stress, Generating More Aβ
  2. MARK Homologue Sparks Tau Terror in Fruit Fly
  3. Honolulu: The Missing Link? Tau Mediates Aβ Toxicity at Synapse
  4. The Plot Thickens: The Complicated Relationship of Tau and Aβ

Webinar Citations

  1. Clearing the Fog Around Aβ Oligomers

Paper Citations

  1. . AMPK is abnormally activated in tangle- and pre-tangle-bearing neurons in Alzheimer's disease and other tauopathies. Acta Neuropathol. 2011 Mar;121(3):337-49. PubMed.
  2. . AMP-activated protein kinase (AMPK) is a tau kinase, activated in response to amyloid β-peptide exposure. Biochem J. 2011 Mar 15;434(3):503-12. PubMed.
  3. . AMP-activated protein kinase signaling activation by resveratrol modulates amyloid-beta peptide metabolism. J Biol Chem. 2010 Mar 19;285(12):9100-13. PubMed.
  4. . Novel synthetic small-molecule activators of AMPK as enhancers of autophagy and amyloid-β peptide degradation. FASEB J. 2011 Jan;25(1):219-31. PubMed.
  5. . Diffusible, nonfibrillar ligands derived from Abeta1-42 are potent central nervous system neurotoxins. Proc Natl Acad Sci U S A. 1998 May 26;95(11):6448-53. PubMed.
  6. . Phosphorylation of Ser262 strongly reduces binding of tau to microtubules: distinction between PHF-like immunoreactivity and microtubule binding. Neuron. 1993 Jul;11(1):153-63. PubMed.
  7. . PAR-1 kinase plays an initiator role in a temporally ordered phosphorylation process that confers tau toxicity in Drosophila. Cell. 2004 Mar 5;116(5):671-82. PubMed.
  8. . Dendritic function of tau mediates amyloid-beta toxicity in Alzheimer's disease mouse models. Cell. 2010 Aug 6;142(3):387-97. PubMed.
  9. . Abeta oligomers cause localized Ca(2+) elevation, missorting of endogenous Tau into dendrites, Tau phosphorylation, and destruction of microtubules and spines. J Neurosci. 2010 Sep 8;30(36):11938-50. PubMed.
  10. . Tau mislocalization to dendritic spines mediates synaptic dysfunction independently of neurodegeneration. Neuron. 2010 Dec 22;68(6):1067-81. PubMed.
  11. . Type 2 diabetes, related conditions, in relation and dementia: an opportunity for prevention?. J Alzheimers Dis. 2010;20(3):723-36. PubMed.
  12. . The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Aβ oligomers through Tau phosphorylation. Neuron. 2013 Apr 10;78(1):94-108. PubMed.

Other Citations

  1. APP transgenic J20

External Citations

  1. APP/PS1
  2. Cabreiro et al., 2013
  3. phase 2 trial

Further Reading


  1. . The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Aβ oligomers through Tau phosphorylation. Neuron. 2013 Apr 10;78(1):94-108. PubMed.

Primary Papers

  1. . The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Aβ oligomers through Tau phosphorylation. Neuron. 2013 Apr 10;78(1):94-108. PubMed.