The widely used psychiatric drug lithium, and other agents that inhibit glycogen synthase kinase-3 (GSK3), have been mentioned as possible Alzheimer's disease therapies for some time, primarily because GSK3 is one of several kinases known to phosphorylate tau. An article in today's Nature reinforces more recent speculation that interfering with GSK3 could also reduce the production of the Aβ peptide, putting before drug developers the tantalizing prospect of hitting two birds (i.e., the two major AD pathologies) with one stone.

The suggestion that GSK3 could be involved in Aβ production stems from the kinase's interaction with presenilins, though there has been no demonstration of direct involvement of GSK3 in presenilin-mediated γ-secretase cleavage of amyloid precursor protein. Last year, Akihiko Takashima's group reported that high doses of the GSK3 inhibitor lithium interfere with in-vitro production of Aβ40 and 42 (Sun et al., 2002). The current report by Peter Klein and associates at the University of Pennsylvania in Philadelphia confirms and extends this finding, showing that more clinically relevant doses of lithium chloride reduce Aβ production from full-length APP in cultured neurons, as well as in the brains of a mouse model transgenic for mutated APP and containing a "knock-in" of a presenilin mutation. Klein and colleagues also identified the target of lithium responsible for this effect, namely GSK3α. The fact that C-terminal fragments of AβPP pile up in the in-vitro models indicates that this effect of lithium occurs before or during the γ-secretase cleavage of AβPP. Inhibitor experiments in AβPP-transgenic CHO cells using kenpaullone (which inhibits GSK and, less strongly, CDKs) and roscovitine (which inhibits CDKs but not GSK), indicated that lithium inhibits Aβ generation via GSK inhibition, not via CDK inhibition.

The researchers provide several lines of evidence to show that it is the GSK3α isoform—and not GSK3β—that facilitates Aβ production. For example, RNAi-mediated depletion of GSK3α, but not β, reduces Aβ production. Conversely, moderate overexpression of the α isoform increases Aβ production.

Unlike most γ-secretase inhibitors, lithium did not inhibit Notch processing by γ-secretase. This would be an important specificity criteria for a drug candidate. The researchers suspect that GSKα might specifically regulate γ-secretase activity toward AβPP, or access of AβPP to the enzyme complex. NSAIDs that modulate γ-secretase activity also do not affect Notch cleavage (see ARF related news story), though they appear to act by a different mechanism.

Lithium targets both the α and β isoforms, making an agent that targets only GSK3α preferable, write the authors. In an accompanying News and Views article, Bart de Strooper of KU Leuven, Belgium, and James Woodgett at Ontario Cancer Institute in Toronto add that GSK inhibition might carry the risk of tumor-causing side effects through the GSK target β-catenin. Lithium itself is not associated with increased risk of cancer, but new, more potent GSK-3 inhibitors might, so it is important to keep that possibility in mind, noted Klein. On the up side, however, de Strooper and Woodgett write that the effective dose of lithium chloride in the present experiments falls within the range of the accepted therapeutic dose for this drug. They write that some Alzheimer's patients might benefit from lithium, but recommend that any potential effect of this drug on dementia be assessed in a clinical trial designed for that purpose, since measuring this outcome in psychiatric patients who currently receive this drug will be difficult.—Hakon Heimer and Gabrielle Strobel


  1. This describes an as-yet unknown and unexpected effect of lithium ions as inhibitors of γ-secretase, the intramembrane proteinase responsible for the cleavage of AβPP and Notch, among other targets. Thereby, lithium reduces the Aβ levels (40 and 42!) in transfected CHO-cells in primary neurons and in mouse brain in vivo.

    This effect and the data being what they are, one is left to explain them, and the authors go to great lengths to identify GSK3, but, surprisingly, not GSK3β, but GSK3α as the target in this setting. So far, the latter has not been regarded with much interest in AD, while the GSK3β variant has been a favorite of some tauists among us, if not Baptists. The results are most certainly interesting, and when confirmed, will raise many questions and follow-up studies.

    First, the "unspecific" inhibition of γ-secretase by lithium ions, and by some NSAIDs, certainly makes understanding the control of γ-secretase’s specificity even more pivotal and central, and not just for AD. The "selectivity-control" mechanisms operate with regard to differentiating between distinct substrates, e.g., AβPP and Notch, at different times, e.g., in the developing and adult brain. Moreover, cleavage of the AβPP substrate renders both amyloid peptides of 40 or 42 residues (and some others); that appears to be controlled by yet other mechanisms, as implied by the current data. The multicomponent proteinase that PS1-dependent γ-secretase is will certainly show more Janus faces in the future.

    Second, lithium ions appear to invoke a pronounced accumulation of AβPP-CTF, mainly of the C83 or α-stubs of AβPP (judging from Fig 1c, about a 100-fold increase). This poses a biochemical problem with pathophysiological consequences. Although adult neurons in transgenic mouse brain can do without PS1 altogether and without the larger fraction of their γ-secretase activity, the ensuing reduction in amyloid peptide levels and the elimination of amyloid plaques did not improve cognition any behavior. In fact, the opposite happened, which was attributed to the accumulated AβPP-CTF (Dewachter et al., 2002). It will be worthwhile to study the effect of treatment with lithium ions on the cognitive performance of AβPP transgenic mice!

    Third, the data imply an unexpected difference, indeed an opposite effect of GSK3α and GSK3β on amyloid peptide levels, and hence on γ-secretase. Despite the soothing words in the conclusion, this must be a blow for believers in the tau-GSK3 connection. If inhibition of GSK3 had any effect on tangle pathology by reducing tau hyperphosphorylation, which is not certain (Spittaels et al., 2000) and yet to be demonstrated in vivo, its overall effect on AD pathology would be negative by increasing amyloid peptide production, as demonstrated here.

    Certainly a hard one to crack. It leaves me wondering why lithium ions do what they are claimed here to do in transgenic mouse brain. Since lithium ions evidently inhibit not just GSK3α but also GSK3, this must have opposite effects on Aβ production. Does α override β ?

    More questions emerge from studying these data, and only time and more experimentation will bring us answers.


    . Glycogen synthase kinase-3beta phosphorylates protein tau and rescues the axonopathy in the central nervous system of human four-repeat tau transgenic mice. J Biol Chem. 2000 Dec 29;275(52):41340-9. PubMed.

    . Neuronal deficiency of presenilin 1 inhibits amyloid plaque formation and corrects hippocampal long-term potentiation but not a cognitive defect of amyloid precursor protein [V717I] transgenic mice. J Neurosci. 2002 May 1;22(9):3445-53. PubMed.

  2. I have never understood why GSK3α was neglected in studies of Alzheimer's neurodegeneration for the past two decades. The enzyme has over 90 percent homology with GSK3β, is also inhibited by lithium, and both isoforms are abundant in brain. The possible differences in their regulation in neurons are particularly interesting, especially if one may antagonize the other. I am happy to see this paper bring GSK3α into the fray in AD. This should lead to some exciting new papers on the role of these kinases in brain and in AD pathology.

  3. It was with surprise that I read these results about the role of LiCl in GSK3 inhibition that confirmed and extended our results (reported in Neuroscience Letters, 2002). Recent technological advances in methodology that uses RNAi clearly showed that GSK3 is involved in Aβ production. While we have successfully shown a direct association between PS1 and GSK3β, the role of GSK3α had not been investigated in our lab. GSK3α may be associated with PS1 and regulate Aβ production. We previously predicted that the phosphorylation of a substrate by GSK3 may be a factor in altering the metabolism of AβPP, but not of Notch. Therefore, an effective inhibitor of GSK3 could lead to therapeutic treatments for AD. To this end, the development of an AβPP/tau double-transgenic Tg mouse model would be ideal to test potential GSK3 inhibitors’ effectiveness in controlling Aβ and neurofibrillary tangles.

  4. Interesting but was already suggested in Neuroscience.


    . Lithium inhibits amyloid secretion in COS7 cells transfected with amyloid precursor protein C100. Neurosci Lett. 2002 Mar 15;321(1-2):61-4. PubMed.

  5. Phiel et al. (2003) show that Lithium (Li), an inhibitor of GSK3β and GSK3α, can decrease Aβ via inhibition of γ-secretase and of GSK3α. These new findings indicate that inhibition of GSK3α is an important therapeutic target in Alzheimer’s disease (AD). This adds a new dimension to the earlier data showing that GSK3β inhibition is also able to reduce the release of Aβ in vitro (Sun et al., 2002).

    GSK3β inhibitors (direct or receptor-mediated) can be useful therapies in AD, inter alia, as inhibitors of neuronal apoptosis and as agents that can block/prevent the accumulation and toxicity of Aβ and hyperphosphorylation of tau. Aβ-dependent neurotoxicity is related to a loss of function of Wnt signaling components (Garrido et al., 2003). The neurotoxic effects of inactivating Wnt signaling by Aβ may result from uncontrolled GSK3β activity. This further implicates GSK3β as a central component of AD pathophysiology and provides additional support for the development of GSK3β inhibitors as therapeutic options.

    Selective M1 muscarinic agonists from the AF series improve cognitive dysfunctions in several animal models, decrease Aβ levels (in vitro and in vivo) and inhibit tau hyperphosphorylation (in vivo and in vitro) (Fisher et al., 2002; Fisher, ADPD, 2003, Seville, May 8-12, 2003). These M1 muscarinic agonists, via activation of M1 receptors, protect hippocampal neurons from Aβ-induced neurotoxicity, and this effect involves the Wnt signaling pathway (Inestrosa, ADPD, 2003, Seville). Notably these M1 agonists show effects similar to Li on Aβ, Wnt components, GSK3β, and tau, yet by different mechanisms (e.g., activation of M1 mAChR for M1 agonists and direct inhibition of GSK for Li). This may be leveraged in the treatment of AD with these highly selective M1 agonists, as these compounds have a very wide safety margin vs. Li.—Abraham Fisher, Israel Institute for Biological Research, POB 19, Ness Ziona 74100, Isreal, and Nibaldo C. Inestrosa, Center for Cell Regulation and Pathology, Faculty of Biological Sciences, P. Catholic University of Chile, Santiago, Chile.


    . Impact of muscarinic agonists for successful therapy of Alzheimer's disease. J Neural Transm Suppl. 2002;(62):189-202. PubMed.

    . Protein kinase C inhibits amyloid beta peptide neurotoxicity by acting on members of the Wnt pathway. FASEB J. 2002 Dec;16(14):1982-4. PubMed.

    . GSK-3alpha regulates production of Alzheimer's disease amyloid-beta peptides. Nature. 2003 May 22;423(6938):435-9. PubMed.

    . Lithium inhibits amyloid secretion in COS7 cells transfected with amyloid precursor protein C100. Neurosci Lett. 2002 Mar 15;321(1-2):61-4. PubMed.

  6. We were the first to show that lithium fully prevents the activation of GSK-3 alpha in vivo. Indeed, we demonstrated that 7 mM lithium in drinking water prevented Abeta-induced translocation of GSK-3 alpha and beta into the nucleus in rabbit brain. This effect was accompanied by inhibition of the Abeta-induced apoptosis but not inhibition of p-tau.


    . Lithium inhibits Abeta-induced stress in endoplasmic reticulum of rabbit hippocampus but does not prevent oxidative damage and tau phosphorylation. J Neurosci Res. 2003 Mar 15;71(6):853-62. PubMed.

  7. The study by MacAulay et al. finding that lithium inhibited GSK3 expression, which was associated with glycogen synthase activation, has me wondering what to expect in Alzheimer's disease.

    Allaman et al. report glycogen accumulation in AD.

    The fact that Skurat and Dietrich report that DYRK1A inactivates glycogen synthase made me think that lithium may be a useful treatment for those with Down's syndrome. It has been shown to reduce the elevated levels of myoinositol in the mouse model Huang et al., 2000).

    A recent study by Hill et al. reporting increased VIP in neonates with Down's syndrome throws some doubt into this suggestion in view of the role of VIP in glycogen synthesis.

    Is Down's syndrome a glycogen storage disorder?


    . Use of lithium and SB-415286 to explore the role of glycogen synthase kinase-3 in the regulation of glucose transport and glycogen synthase. Eur J Biochem. 2003 Sep;270(18):3829-38. PubMed.

    . Protein targeting to glycogen mRNA expression is stimulated by noradrenaline in mouse cortical astrocytes. Glia. 2000 Jun;30(4):382-91. PubMed.

    . Phosphorylation of Ser640 in muscle glycogen synthase by DYRK family protein kinases. J Biol Chem. 2004 Jan 23;279(4):2490-8. PubMed.

    . Brain myo-inositol level is elevated in Ts65Dn mouse and reduced after lithium treatment. Neuroreport. 2000 Feb 28;11(3):445-8. PubMed.

    . Vasoactive intestinal peptide in the brain of a mouse model for Down syndrome. Exp Neurol. 2003 Sep;183(1):56-65. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Anti-inflammatory Drugs Side-Step COX Cascade to Target Aβ

Paper Citations

  1. . Lithium inhibits amyloid secretion in COS7 cells transfected with amyloid precursor protein C100. Neurosci Lett. 2002 Mar 15;321(1-2):61-4. PubMed.

Further Reading

Primary Papers

  1. . GSK-3alpha regulates production of Alzheimer's disease amyloid-beta peptides. Nature. 2003 May 22;423(6938):435-9. PubMed.
  2. . Alzheimer's disease: Mental plaque removal. Nature. 2003 May 22;423(6938):392-3. PubMed.