Biochemical Journal

Research article

AMP-activated protein kinase (AMPK) is a tau kinase, activated in response to amyloid β-peptide exposure

Claire Thornton , Nicola J. Bright , Magdalena Sastre , Phillip J. Muckett , David Carling

Abstract

Hyperphosphorylation of tau is a hallmark of Alzheimer's disease and other tauopathies. Although the mechanisms underlying hyperphosphorylation are not fully understood, cellular stresses such as impaired energy metabolism are thought to influence the signalling cascade. The AMPK (AMP-activated protein kinase)-related kinases MARK (microtubule-associated protein-regulating kinase/microtubule affinity-regulating kinase) and BRSK (brain-specific kinase) have been implicated in tau phosphorylation, but are insensitive to activation by cellular stress. In the present study, we show that AMPK itself phosphorylates tau on a number of sites, including Ser262 and Ser396, altering microtubule binding of tau. In primary mouse cortical neurons, CaMKKβ (Ca2+/calmodulin-dependent protein kinase kinase β) activation of AMPK in response to Aβ (amyloid-β peptide)-(1–42) leads to increased phosphorylation of tau at Ser262/Ser356 and Ser396. Activation of AMPK by Aβ-(1–42) is inhibited by memantine, a partial antagonist of the NMDA (N-methyl-D-aspartate) receptor and currently licensed for the treatment of Alzheimer's disease. These findings identify a pathway in which Aβ-(1–42) activates CaMKKβ and AMPK via the NMDA receptor, suggesting the possibility that AMPK plays a role in the pathophysiological phosphorylation of tau.

  • Alzheimer's disease
  • AMP-activated protein kinase (AMPK)
  • amyloid-β peptide
  • microtubule-associated protein
  • N-methyl-D-aspartate (NMDA)
  • tau

INTRODUCTION

The physiological role of the microtubule-associated protein tau is to bind to tubulin heterodimers and promote microtubule assembly. Phosphorylation of tau is required for efficient neurite outgrowth and the regulation of axonal transport [1,2]. However, hyperphosphorylation of tau is believed to underlie the formation of NFTs (neurofibrillary tangles), a neuropathological characteristic of Alzheimer's disease and other ‘tauopathies’ resulting in neurodegeneration [3]. This abnormal phosphorylation allows the release of tau from microtubules, its aggregation into paired helical filaments, a major component of NFTs, and promotes microtubule disruption [4]. Events surrounding the initiation of tau hyperphosphorylation in vivo are not clearly understood. At least 30 protein kinases have been reported to phosphorylate tau in vitro, and of these a number, including GSK3β (glycogen synthase kinase 3β), cdk5 (cyclin-dependent kinase 5), CK1 (casein kinase 1), PKA (cAMP-dependent protein kinase) and MARKs (microtubule-associated protein-regulating kinase/microtubule affinity-regulating kinases), have been suggested to play a role in tau phosphorylation in vivo [5]. Despite the complexity of the system, only a limited number of the potential tau kinases have been shown to phosphorylate tau at Ser262 within the KXGS motif of the microtubule-binding domain. Phosphorylation of Ser262 is thought to influence the ability of tau to bind to microtubules [6].

MARKs are members of the AMPK (AMP-activated protein kinase)-related kinase family which also includes two characterized brain-specific kinases BRSK1 and BRSK2 [7]. All kinases within the group share amino acid sequence homology within their catalytic domains and require phosphorylation of a conserved threonine residue within the activation loop region. LKB1, which was identified as an upstream kinase in the AMPK cascade, phosphorylates and activates all of the AMPK-related kinases, apart from MELK (maternal embryonic leucine-zipper kinase) [7]. In addition to LKB1, AMPK is phosphorylated and activated by CaMKKβ (Ca2+/calmodulin-dependent protein kinase kinase β), and this has been shown to mediate activation of AMPK in response to increased intracellular Ca2+ concentrations. [8]. AMPK is also activated in response to an increase in AMP, following a fall in ATP levels within the cell. AMP allosterically activates AMPK as well as protecting the enzyme against dephosphorylation [9]. AMPK is a heterotrimer comprising a catalytic α subunit and regulatory β and γ subunits. Isoforms of all three subunits exist (α1, α2, β1, β2, γ1, γ2 and γ3) and all twelve heterotrimeric complex combinations have been identified in vivo. The physiological significance of the different AMPK isoform complexes is only partly understood, although it seems likely that they allow tissue-specific regulation of AMPK as well as tissue-specific function (reviewed in [10]).

AMPK is expressed ubiquitously and is a key regulator of metabolic pathways such as fatty acid and cholesterol synthesis [10]. However, delineating the function of neuronal AMPK is still in its infancy. The most detailed studies have taken place in the arcuate nucleus of the hypothalamus where AMPK is involved in co-ordinating the regulation of appetite and satiety [11]. AMPK is active in other brain regions but, as yet, it is unclear whether activation of AMPK plays a neuroprotective or neurodegenerative role with results supporting either position [1215]. More recently, AMPK has been implicated in both beneficial and deleterious modulations of APP (amyloid precursor protein), proteolytic cleavage of which generates Aβ (amyloid β-peptide)-(1–42), a key component of disease-characteristic amyloid plaques [16]. In response to metformin treatment, activation of AMPK was reported to promote the formation of both intracellular and extracellular Aβ from APP [17]. Conversely, resveratrol, a plant polyphenol with potential neuroprotective properties, is reported to negatively regulate the production of Aβ-(1–42) through an AMPK-mediated pathway [18].

Recent research in both Drosophila and mice suggest that the MARKs and, to a lesser extent, the related BRSKs may play a role in early priming phosphorylation of tau [1921]. However, to date the physiological regulation of these kinases and the signals that modulate their activity remain unknown [22]. Currently, there has been little examination of the role of AMPK in the phosphorylation of tau. In the present study, we provide results implicating AMPK as a physiological kinase in a pathway phosphorylating tau. We have identified a number of sites in tau phosphorylated by AMPK, many of which are clustered around the microtubule-binding domain. Intriguingly, phosphorylation of tau by AMPK regulates its ability to bind to microtubules. In addition, we show that exposure of primary neurons to Aβ-(1–42) increases AMPK activity via a CaMKKβ-dependent pathway. Aβ-(1–42)-mediated activation of AMPK is blocked by memantine, an NMDA (N-methyl-D-aspartate) receptor antagonist licensed for treating Alzheimer's disease. These findings place the NMDA receptor upstream of AMPK in a CaMKKβ-mediated activation pathway resulting in the phosphorylation of tau.

EXPERIMENTAL

Preparation of primary neurons

Animal use was in accordance with local rules and with the regulations and guidance issued under the Animals (Scientific Procedures) Act (1986). Primary cortical neurons were prepared from WT (wild-type) and CaMKKβ−/− E15 (embryonic day 15) animals as described previously [23]. Briefly, cortices from embryos in a single litter were dissected, meninges were removed and tissue was pooled. Cortices were roughly chopped before incubation in 0.25% trypsin/EDTA followed by trituration. Cells were pelleted by centrifugation, resuspended and plated in Neurobasal medium supplemented with B27, 100 units/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B, 300 μM glutamine and 25 μM 2-mercaptoethanol. Treatments were carried out on neurons grown for a minimum of 7 days in vitro, and all media were obtained from Invitrogen.

Neuronal cell treatment and lysis

Primary cortical neurons were treated with the following compounds for the times indicated: H2O2 (0.5 mM for 15 min; Sigma), ionomycin (2 μM for 5 min; Tocris), Aβ-(1–42) (20 μM for 30 min; Invitrogen), NMDA (50 μM for 15 min; Sigma), STO609 (10 μg/ml for 3 h; Tocris) and memantine (10 μM for 3 h; Sigma). In cases in which a 3 h pre-treatment was required (memantine and STO609), activators were subsequently added in the presence of the pre-treating compound. Aggregated Aβ-(1–42) peptide was prepared according to the manufacturer's instructions (Invitrogen); freeze-dried peptide was first resuspended to 6 mg/ml in sterile distilled water, vortex-mixed thoroughly and then diluted to a final concentration of 1 mg/ml in PBS. Aggregation was achieved by incubation at 37 °C for 24 h. Treated cells were harvested in Hepes buffer A containing Triton [50 mM Hepes (pH 7.5), 50 mM sodium fluoride, 5 mM sodium pyrophosphate, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM PMSF, 157 μg/ml benzamidine, 10% (v/v) glycerol and 1% (v/v) Triton X-100] and the protein concentration was determined using a Bradford assay (Bio-Rad Laboratories).

Immunoprecipitation

Total protein lysate (50 μg) was incubated for 2 h at 4 °C in the presence of Protein A– or Protein G–Sepharose (for rabbit and mouse antibodies respectively) and antibody as follows: pan-AMPKβ [24], BRSK2 [22], His6 and MARK3 (Abcam), and AMPKα1 and AMPKα2 (Cell Signaling Technology). Immobilized immune complexes were washed thoroughly in PBS containing 1% (v/v) Triton X-100 and then used for kinase assays or Western blot analysis.

In vitro phosphorylation

Recombinant bacterially expressed WT (α1/β1/γ1; 1 μg) or kinase-inactive (α1D157A/β1/γ1; 1 μg) AMPK was purified as described previously [25]. GSK3β (1 μg) was from New England Biolabs, MARK4 (1 μg) was from Cell Signaling Technology, and recombinant tau (2N4R, longest human tau isoform, 0.5 μg) was from Sigma. Purified protein kinases and recombinant tau (0.5 μg) were incubated in phosphorylation buffer {50 mM Tris/HCl (pH 7.5), 10 mM MgCl2, 5 mM dithiothreitol and 3 mM [32P]ATP} for 1 h at 37 °C. Phosphorylation reactions were resolved by SDS/PAGE, transferred on to PVDF membrane and detected by fluorography overnight at −80 °C. For Western blots of AMPK-phosphorylated recombinant tau, [32P]ATP was replaced with 3 mM unlabelled ATP and 20 ng of phosphorylated tau per lane was analysed by phospho-tau-specific antibodies.

Kinase assays

AMPK, BRSK2 and MARK3 activities were measured using the AMARA substrate peptide (AMARAASAAALARRR) as described previously [26]. Assays were carried out for 20 min at 37 °C.

Western blot analysis

Total protein lysate or immobilized immune complexes (25–50 μg) was boiled in SDS sample buffer and resolved on 10% (w/v) NuPage BisTris gels (Invitrogen). Proteins were transferred on to PVDF and analysed by Western blotting with the following antibodies: phospho-tau (Ser214) (Abcam), 12E8 [27], PHF-1 (paired helical filament-1) [28], phospho-tau (Ser199), phospho-tau (Thr231), phospho-tau (Ser262), phospho-tau (Ser356), phospho-tau (Ser396), phospho-tau (Ser400), phospho-tau (Thr404), phospho-tau (Ser422) and tau5 (Biosource), phospho-AMPK (Thr172) and phospho-ACC (acetyl Co-A carboxylase) (Cell Signaling Technology), and actin (Sigma). Membranes were washed and incubated with secondary antibodies conjugated to Alexa Fluor® 680 (Invitrogen) or IRDye800 (Li-Cor) and were scanned on the Li-Cor Odyssey Infrared Imaging System using Odyssey software 2.0 for band quantification. Typically, primary antibody incubations took place overnight at 4 °C and secondary antibody incubations were for 2 h at room temperature (22 °C).

Tau overexpression

cDNA encoding 2N4R tau was subcloned into pcDNA3. Mutation of Ser262 to alanine was carried out by site-directed mutagenesis (Invitrogen). pcDNA3-TauWT and -TauS262A were expressed in CCL13 cells by calcium phosphate transfection of 10 μg of purified plasmid DNA (BD Clontech). Precipitates were removed after overnight incubation and cells were harvested 48 h later. Expressed protein was isolated by immunoprecipitation with anti-His antibodies.

Mass spectrometry

Recombinant AMPK-phosphorylated tau (0.5 μg) was resolved by SDS/PAGE, excised and sent for analysis to the Taplin Mass Spectrometry Facility, Harvard Medical School, Boston, MA, U.S.A.

Microtubule binding

Taxol-stabilized microtubules were generated according to the manufacturer's instructions (Cytoskeleton). Briefly, a stable population of microtubules was generated by polymerizing bovine tubulin (100 μg) at 37 °C for 20 min in the presence of GTP (1 mM) and taxol (20 μM). AMPK-phosphorylated or non-phosphorylated tau was incubated in the presence of taxol-stabilized microtubules at room temperature for 30 min. The reaction was fractionated by ultracentrifugation through a taxol-stabilized cushion buffer [50% (v/v) glycerol and 20 μM taxol] at 100000 rev./min (TLA-100 rotor; Beckman Coulter) for 40 min at room temperature. The resulting fractions were resolved by SDS/PAGE and were analysed by Western blot using a phosphorylation-independent tau antibody.

Statistical analyses

Results are expressed as means±S.D. from three or more independent experiments. Statistical analysis was performed using Prism (Graphpad). Statistical significance between two conditions was established with a two-tailed Student's t test, whereas significance among multiple datasets was determined by ANOVA, followed by the post-hoc Tukey multiple comparison test. Results was considered significant at *P<0.05 or **P<0.01.

RESULTS

AMPK is a tau kinase

In transient transfection studies, we have observed previously [22] that BRSK1 and BRSK2 were insensitive to activation by a range of pathophysiological stimuli, including those that might occur early in Alzheimer's disease. In the present study, we have extended this to primary cortical neurons and have found that the activities of endogenous MARK3 and BRSK2 are not significantly altered in response to ionomycin (increasing intracellular Ca2+ concentrations) or H2O2 (to mimic oxidative stress; Figure 1A). Therefore we examined the hypothesis that AMPK itself was a physiological tau kinase, as AMPK is activated in response to these and other cellular stresses. We carried out in vitro phosphorylation experiments using active and inactive (mutation of D157A in the ATP-binding pocket of AMPKα) AMPK, GSK3β, MARK4 and recombinant tau (Figure 1B). In addition to GSK3β and MARK4, which have been shown previously to phosphorylate tau [20,29], AMPK caused a robust phosphorylation of tau (Figure 1B). An identical experiment analysed by Western blotting determined that AMPK increased the phosphorylation of Ser262 (Figure 1C, upper panel), a site implicated previously as a major phosphorylation site for BRSKs and MARKs [29,30]. Only weak phosphorylation of Ser262 was observed for GSK3β, although the migration of tau by SDS/PAGE was retarded, suggesting phosphorylation at other sites (Figure 1C, lower panel). To determine whether Ser262 was the predominant phosphorylation site for AMPK on tau, we generated His-tagged WT tau or a construct harbouring the mutation of Ser262 to alanine (S262A) and expressed the proteins in CCL13 cells, a mammalian cell line lacking LKB1 [26]. These cells therefore have the advantage that the basal activity of AMPK is low and overexpression of tau in this cell line is likely to remain unphosphorylated by AMPK. Tau was isolated from cell lysates by immunoprecipitation and subjected to phosphorylation by AMPK or GSK3β. As expected, mutation of S262A had no detectable effect on the ability of GSK3β to phosphorylate tau. However, tau phosphorylation by AMPK also remained relatively unchanged in the S262A mutant compared with WT, indicating the presence of additional AMPK phosphorylation sites within tau (Figure 1D).

Figure 1 AMPK is a tau kinase

(A) Primary neurons were treated with vehicle control (DMSO; Con), H2O2 or ionomycin (Iono), lysed and immune complexes of BRSK2, MARK3 or AMPK were assayed for kinase activity. The average fold activity compared with the control is shown for three independent experiments (**P<0.01). (B) Recombinant tau was phosphorylated (ptau) by GSK3β, MARK4 and active or inactive (D157A) AMPK, and incorporation of [32P]ATP into tau was determined by fluorography (upper panel). Equivalent amounts of tau was confirmed by Commassie Brilliant Blue (CBB) gel staining (lower panel). Additional bands present in lane 1 are from recombinant GSK3β. (C) Phosphorylated recombinant tau was analysed by Western blotting for phospho-tau Ser262 (phospho-tau Ser262 antibody; upper panel) or total tau (lower panel). (D) His-tagged cDNA constructs of WT or mutated (S262A) tau were transfected into a mammalian cell line. Expressed tau was purified by immunoprecipitation with anti-His antibodies and subjected to phosphorylation by GSK3β (left-hand panels) or AMPK (right-hand panels). Phosphorylation was determined by fluorography as above (upper panels) and equivalent expression of tau was confirmed by Western blotting (lower panels). Blots and fluorographs are representative of three independent experiments. ut, untransfected

AMPK phosphorylates multiple tau epitopes

In order to study further the phosphorylation of tau by AMPK, we compared phosphorylation of tau by MARK4 or AMPK using a panel of phospho-tau antibodies (Figure 2A). Five phosphorylation sites common to both kinases were identified (Thr231, Ser262, Ser356, Ser396 and Ser422). In addition, AMPK, but not MARK4, was capable of phosphorylating Ser214 (Figure 2A). In blots in which there was no phosphorylated signal detected (Ser199, Ser400 and Ser404), a positive control phosphorylation was carried out using GSK3β to confirm antibody recognition of the phospho-specific epitope (Supplementary Figure S1 at http://www.BiochemJ.org/bj/434/bj4340503add.htm). As Western blotting revealed that AMPK phosphorylates a number of sites in tau, we used MS to comprehensively identify as many AMPK sites as possible. Briefly, AMPK-phosphorylated tau was subjected to MS analysis, achieving 81% coverage of 2N4R tau (Supplementary Figure S2 at http://www.BiochemJ.org/bj/434/bj4340503add.htm). Using this approach, we were able to assign with confidence over 20 AMPK phosphorylation sites within tau (Figure 2B). Interestingly, nine were identified within the microtubule-binding domain and seven out of eight conserved serine residues within the four repeated tubulin-binding motifs were identified as AMPK sites (Figure 2B, top panel).

Figure 2 AMPK phosphorylates multiple tau sites and interferes with microtubule binding

(A) MARK4 (M)- or AMPK (A)-phosphorylated recombinant tau was analysed by Western blotting using phospho-specific tau antibodies (epitope indicated above each panel in single-letter amino-acid notation) and total tau (lower panels). (B) A schematic diagram of the longest isoform of tau (2N4R; 441 amino acids) showing domains and known phosphorylation sites of serine/threonine kinases or sites that are phosphorylated in vitro or in Alzheimer's disease brain post mortem [5,55]. AMPK sites identified from MS analysis are shown in bold. (C) Unphosphorylated (tau) or AMPK-phosphorylated (ptau) tau was incubated in the presence or absence of microtubules (MT). Microtubule bound (pellet) or microtubule unbound (supernatant) fractions were generated by ultracentrifugation. Fractions were analysed by Western blotting for total tau (upper panel). The lower panel shows sedimented microtubules by Coomassie Brilliant Blue staining. The histogram shows the bound compared with unbound tau (**P<0.01) or phospho-tau (##P<0.01) as a percentage of total tau. For (A) and (C), images are representative of four individual experiments.

AMPK phosphorylation inhibits tau binding of microtubules

As the bulk of AMPK phosphorylation sites are clustered around the microtubule-binding domain, we investigated the effect of AMPK phosphorylation on binding of tau to microtubules in vitro. Tau, before and after phosphorylation by AMPK, was incubated with microtubules and subjected to ultracentrifugation to pellet microtubule-binding proteins. The majority (75%) of non-phosphorylated tau was detected in the high-speed pellet, indicating efficient binding to the microtubules (Figure 2C). In contrast, only 20% of the AMPK-phosphorylated tau was detected in the microtubule pellet, with 80% remaining in the supernatant fraction.

AMPK is activated in response to Aβ-(1–42) exposure

The deposition of extracellular plaques, of which a short Aβ [Aβ-(1–42)] is a key component, is a pathological characteristic of Alzheimer's disease [16]. Exposure of primary hippocampal neurons to aggregated Aβ-(1–42) has been reported to induce tau phosphorylation [31]. To determine whether there is a role for AMPK in the Aβ-(1–42) pathway, a mixed cortical/hippocampal culture of primary neurons was prepared and treated with aggregated Aβ-(1–42) (20 μM) and AMPK activity in immune complexes was determined. As shown in Figure 3(A), there was a significant increase in activity from AMPK isolated from Aβ-(1–42)-treated cell lysates compared with control cells. Phosphorylation of AMPK at Thr172, another direct measure of AMPK activation, was also increased and this correlated with an increase in phospho-ACC, a well-characterized downstream target of AMPK (Figure 3B). As both BRSK and MARK have been implicated in tau phosphorylation [19,32], the activity of these kinases following Aβ-(1–42) treatment was also determined. In contrast with AMPK, no change in either BRSK2 or MARK3 activity after Aβ-(1–42) treatment was detected (activities for control compared with Aβ-(1–42) treatment: BRSK2, 0.0537±0.0103 compared with 0.0597±0.0083 nmol/min per mg of protein; and MARK3, 0.0464±0.0064 compared with 0.0463±0.0089 nmol/min per mg of protein; n=4). Acute exposure of primary cortical neurons to aggregated Aβ-(1–42) induces a rapid increase in intracellular Ca2+ concentrations [33]. Therefore we tested the hypothesis that the regulation of AMPK by Aβ-(1–42) was mediated through a CaMKKβ signalling pathway. We isolated primary neuronal cultures from CaMKKβ−/− mice and treated these cells with Aβ-(1–42). In marked contrast with WT neurons, Aβ-(1–42) treatment had no effect on AMPK activity in the cells derived from the CaMKKβ−/− mice (Figure 3C), establishing that CaMKKβ is required for AMPK activation by Aβ-(1–42). Previous studies by us [34,35] and others [36] have identified a substrate preference in CaMKKβ for activation of AMPKα1-, rather than AMPKα2-, containing complexes. To determine which isoform mediates the response to Aβ-(1–42), isoform-specific depletion of AMPK was repeatedly performed on the same lysate until all isoform complexes were removed. AMPKα1 complexes were significantly activated in response to Aβ-(1–42), whereas there was no significant increase in AMPKα2 activity (Figure 3D). These results are similar to a previous study [35], where we only observed significant activation of AMPKα1 complexes following activation of CaMKKβ.

Figure 3 AMPK is activated by Aβ-(1–42)

Primary neurons from WT (A, B and D) and CaMKKβ−/− (C) mice were treated with 20 μM Aβ-(1–42). Treated (Aβ) and untreated (Con) neurons were harvested after 30 min and kinase activity was determined in immune complexes of AMPK (A and C). Box plots show the minimum, median and maximum activities obtained for neurons from between three and six individual experiments (**P<0.01). (B) Lysates from WT neurons treated with Aβ-(1–42) were analysed by Western blotting with antibodies recognizing phosphorylated (active) AMPK (pAMPK) and a known AMPK substrate (pACC). Western blotting using a pan-AMPKβ antibody for total AMPK and anti-actin antibodies demonstrated there was no change in the amount of total protein. Representative images shown are neuronal lysates from two of four individual experiments. The molecular mass in kDa is indicated on the left-hand side. (D) Lysates from treated (Aβ) and untreated (Con) neurons were subjected to sequential immunoprecipitation with either AMPKα1- or AMPKα2-specific antibodies to isolate all available endogenous isoform-specific AMPK complexes. Results shown represent the total isoform-specific activity from neuronal lysates of three individual experiments (*P<0.05).

Activation of AMPK by Aβ-(1–42) increases tau phosphorylation at disease-relevant epitopes

Our initial findings suggested that AMPK phosphorylates a number of residues on recombinant tau. As Aβ-(1–42) treatment activates AMPK in primary neurons, we determined the downstream effect on the phosphorylation of endogenous tau. We obtained 12E8 and PHF-1 antibodies, commonly used to detect pathologically phosphorylated tau [27,28], and examined the phosphorylation state of tau in Aβ-(1–42)-treated WT and CaMKKβ−/− neurons. We observed a marked increase in signal for both antibodies in the WT neurons, indicating that phosphorylation had occurred at Ser262/Ser356 (12E8) and Ser396/Ser404 (PHF-1; Figure 4A), consistent with our earlier observations identifying Ser262, Ser356 and Ser396 as candidate AMPK phosphorylation sites in vitro (Figure 2). Conversely, there was little difference in tau phosphorylation between control and Aβ-treated CaMKKβ−/− neurons after Western blot analysis with 12E8. More surprisingly perhaps, was that phosphorylation of Ser396/Ser404, as detected by the PHF-1 antibody, was reduced following treatment with Aβ-(1–42) in CaMKKβ−/− cells (Figure 4B). To determine whether this difference was due to reduced total tau or reduced phosphorylation after Aβ treatment, we compared the levels of total tau in WT and CaMKKβ−/− lysates. Although there was a difference in total tau protein expression in the CaMKKβ−/− cells compared with WT (32% less tau in CaMKKβ−/− neurons), there was no change in tau expression between control and Aβ-(1–42)-treated neurons (Figure 4C). These results suggest that tau phosphorylation at Ser396/Ser404 in response to Aβ-(1–42) is significantly influenced by active CaMKKβ.

Figure 4 Increases in tau phosphorylation in response to Aβ-(1–42) are muted in cells lacking CaMKKβ

Primary cortical neurons from WT (A and C) and CaMKKβ−/− (B and C) mice were treated with Aβ-(1–42) as before and analysed by Western blotting for phosphorylated tau (ptau) using 12E8 (Ser262/Ser356 phosphorylation) and PHF-1 (Ser396/Ser404 phosphorylation) antibodies (A and B) or for phosphorylation-independent total tau (C). Quantification of phosphorylation (A and B) and total tau (C) normalized relative to actin is shown in the histograms (*P<0.05). Blots shown are primary cell lysates from two of four (WT) or two of three (CaMKKβ−/−) individual experiments. The molecular mass in kDa is indicated on the left-hand side.

AMPK activation by Aβ-(1–42) is inhibited by memantine

One mechanism by which Aβ-(1–42) may exert its intracellular effects is through potentiating glutamate-mediated excitotoxicity [37]. Neurodegeneration of the glutamatergic pyramidal neurons of the hippocampus and cortex is an early event in Alzheimer's disease and cultured neurons can be protected from Aβ-(1–42) toxicity by treatment with glutamatergic NMDA receptor antagonists [38]. Interestingly, we found that AMPK was robustly activated in primary neurons after NMDA treatment, and that this activation was abolished by pre-treatment with STO609, a pharmacological inhibitor of CaMKKα and CaMKKβ (Figure 5A). Memantine is a non-competitive NMDA receptor antagonist and recently its use for treatment in moderate-to-severe Alzheimer's disease has been approved [39], highlighting the role of this pathway in the pathology of cognitive decline. We found that treatment of primary neurons with memantine prior to exposure to NMDA prevented the activation of AMPK (Figure 5A). Memantine is reported to reduce both the excitotoxic effects of Aβ-(1–42) in rat cortical neurons and the phosphorylation of tau [40]. Therefore we repeated the Aβ-(1–42) treatment on primary neurons with or without memantine pre-treatment. As with NMDA, memantine successfully prevented the activation of AMPK by Aβ-(1–42) (Figure 5B), implying that Aβ-(1–42) acts on AMPK through activation of the NMDA receptor increasing intracellular Ca2+ concentrations and activating CaMKKβ.

Figure 5 Activation of AMPK is prevented by the NMDA receptor antagonist memantine

(A) Primary neurons were left untreated (Con), pre-treated with either STO609 or memantine alone, or treated with NMDA. Neurons were harvested and AMPK activity was determined from the immune complexes. (B) Primary neurons either untreated (Con) or pre-treated with memantine were treated with Aβ-(1–42). Neurons were harvested and assayed for AMPK activity as above. Box plots show the minimum, median and maximum activities obtained for neurons from four [**P<0.01 in (A)] or three [*P<0.05 in (B)] individual experiments.

DISCUSSION

The hyperphosphorylation of tau is a key event in the progression of Alzheimer's disease yet the mechanisms underlying this event remain unclear. Our present results provide the first evidence that AMPK activation may act as a link between exposure of neurons to Aβ and subsequent tau phosphorylation.

MS analysis identified a cluster of candidate AMPK phosphorylation sites around the microtubule-binding domain of tau including all four conserved serine residues (Ser262, Ser293, Ser324 and Ser356) that lie within the four KXGS microtubule binding motifs (Figure 2B, expanded panel). This suggested a role for AMPK phosphorylation in the regulation of microtubule binding. In addition, phosphorylation at Ser214 and Thr231 was also detected by Western blotting and MS. In line with previous research suggesting that phosphorylation of these epitopes induces maximal disruption of binding [19,41], the interaction of tau with microtubules was greatly reduced following phosphorylation by AMPK in vitro. Neither MS nor Western blot analysis identified Ser404 as a potential AMPK site, suggesting that Ser396 provides the epitope for recognition by the PHF-1 antibody. MS also highlighted Thr71 and Thr403 as AMPK phosphorylation sites. An online database of tau phospho-epitopes (http://cnr.iop.kcl.ac.uk/hangerlab/tautable) lists both sites as phosphorylated in brain tissue isolated from patients with Alzheimer's disease, but as yet no kinase has been identified. The N-terminal sequence of tau containing Thr71 is referred to as the projection domain as it projects away from the microtubule to interact with cytoskeletal proteins linking it to the plasma membrane and may also regulate tau aggregation [42,43]. Further work is currently underway to determine whether AMPK phosphorylation at Thr71 regulates these properties of tau.

Exposure of primary neurons to Aβ-(1–42) rapidly increases AMPK activity, but is also reported to alter the activities of other kinases. Exposure of primary rat neuronal cultures to Aβ for 6–8 h increased cdk5 and p38 activities [44] and, in a similar study, Song et al. [40] demonstrated increased ERK1/2 (extracellular-signal-regulated kinase 1/2) activity after a 1 h Aβ exposure. In animals, a study describing lentiviral injection of Aβ into rat motor cortex revealed prolonged activity of both cdk5 and GSK3β [45]. In all cases, tau phosphorylation was reported to increase. In contrast, a very recent study [46] found that Aβ exposure of primary rat hippocampal neurons for 3 h resulted in a synaptically localized increase in MARK, BRSK, p70S6K (p70 S6 kinase) and cdk5 activities, whereas GSK3β and MAPK activities did not change. An increase in tau phosphorylation at 12E8 sites (Ser262 and Ser356 phosphorlyation) was also observed and ascribed to the activities of AMPK-related kinases, although the activity of AMPK itself was not determined. These findings suggest that further work is required to distinguish the effects of short- and long-term Aβ exposure and intracellular localization on kinase activities.

In our present study, we were unable to detect changes in the activities of BRSK2 or MARK3 in response to 30 min of Aβ treatment. The MARK family were originally identified through their ability to regulate the interaction of tau with microtubules [19]. Studies of the MARK orthologue in Drosophila PAR-1 show that overexpression of APP up-regulated PAR-1 activation and phosphorylation of tau [21]. In the present study, we have shown that short-term Aβ-(1–42) exposure exerts its effects through CaMKKβ activation, consistent with the hypothesis that Aβ renders the neuron more susceptible to glutamate receptor neurotoxocity by destabilizing Ca2+ homoeostasis [33]. These findings provide an explanation for the lack of effect of Aβ-(1–42) treatment on the activity of the AMPK-related kinases tested, since a previous study reported that these kinases were not regulated by Ca2+ signalling pathways, including CaMKKβ [22]. AMPK-related kinases are activated by LKB1 phosphorylation [7] and, although the upstream stimulus for this pathway remains unclear, this may be the mechanism by which they play a role in tau phosphorylation.

Antibodies recognizing both early- (12E8) and late- (PHF-1) stage tau phosphorylation were used to determine the effect of Aβ-(1–42) treatment on primary neurons. Treatment conditions which activated AMPK also caused a substantial increase in tau phosphorylation at a number of disease-relevant epitopes in WT neurons. Conversely in CaMKKβ−/− neurons, we did not detect any change in Ser262/Ser356 phosphorylation following treatment. We did, however, observe a reduction in phosphorylation of Ser396/Ser404 following Aβ-(1–42) treatment in CaMKKβ−/− neurons relative to untreated cells. This is not due to reduced levels of tau as the 12E8 epitopes are unaffected and there is no change in the amount of total tau protein expression in CaMKKβ−/− control compared with CaMKKβ−/−Aβ-(1–42) lysates (Figure 4C compare lanes 5 and 6 with lanes 7 and 8). As the actions of kinases and phosphatases are tightly balanced, one possible explanation is that removing CaMKKβ disturbs this balance. Following influx of Ca2+ after Aβ-(1–42) treatment, it is possible that Ca2+-dependent phosphatases are activated and, in the absence of competing activation of CaMKKβ, the equilibrium is pushed in favour of dephosphorylation. Interestingly, Ser396 and Ser404 have been identified as substrates for Ca2+-dependent PP (protein phosphatase) 2B (calcineurin) [47]. Although the activity of a number of major tau phosphatases has been reported to decrease in Alzheimer's disease hence promoting tau hyperphosphorylation [48], it has been suggested that the activity of PP2B may increase [47].

Our present study demonstrates robust AMPK activity in response to pharmacological activation of the NMDA receptor, an activation that can be inhibited by the NMDA receptor antagonist memantine. A recent study [49] identified another NMDA receptor-mediated pathway for AMPK activation. AMPK is involved in the development of excitotoxic apoptosis through a mechanism requiring the BH3-only protein Bim. Prolonged activation of AMPK leads to increased JNK (c-Jun N-terminal kinase) phosphorylation, increased bim mRNA expression and neuronal death [49]. Other groups report both JNK activity and Bim are up-regulated on exposure to Aβ-(1–42) [50,51]. Taking these studies together, it is interesting to speculate that these effects may be exerted in part by Aβ-(1–42) activation of AMPK.

AMPK activity is reported to be altered in the APP/PS2 animal model of Alzheimer's disease which exhibits amyloidosis, cognitive deficits, impaired synaptic plasticity and perturbed cerebrovascular development [52]. In a study carried out in young APP/PS2 animals (less than 8 months old), where the phenotype is mild, AMPK is abnormally activated compared with control mice [53]. This activation, in association with IGF (insulin growth factor)-1 imbalance, contributes to increased VEGF (vascular endothelial growth factor) expression and decreased formation of blood vessels. Overexpression of Aβ-(1–42) in these animals is apparent as early as 5 months even though cognitive impairment is not detectable until 8 months. These findings fit with our present results from primary neurons in which there is a rapid activation of AMPK in response to the pathological up-regulation of Aβ-(1–42).

Currently there is little information regarding the alteration of AMPK activity in human neurological disease. Indeed, a practical limitation of assessing AMPK activity in neural tissue is the rapidity needed for dissection and homogenization to avoid spurious increases in AMPK activity caused by anoxia and such samples are rarely available. However, a study [50] published while the present manuscript was in review carried out a comprehensive analysis of human tauopathy samples post mortem (including Alzheimer's disease) with phospho-AMPK antibodies. They observed an increase in phospho-AMPK in both tangle- and pretangle-bearing neurons and confirmed our identification of Ser396 as a tau epitope phosphorylated by AMPK, using the same PHF-1 antibody [54]. Taken together with our present results, it suggests that AMPK may play a key role in governing the early phosphorylation of tau.

In summary, we present results suggesting that AMPK is rapidly activated in primary neurons on exposure to Aβ, resulting in an increase in tau phosphorylation at residues critical for microtubule binding. On the basis of the properties of Aβ published previously, we have identified a key role for the NMDA receptor and the Ca2+-activated upstream AMPK kinase CaMKKβ in this pathway. Our present work suggests that AMPK deserves close examination as a candidate in vivo kinase in the regulation of tau hyperphosphorylation and further studies are required to determine whether AMPK has a clinically relevant role in neurodegeneration.

AUTHOR CONTRIBUTION

Claire Thornton and David Carling designed the research; Claire Thornton, Nicola Bright and Magdalena Sastre performed the research; Philip Muckett contributed new analytical tools; and Claire Thornton analysed the data and wrote the paper.

FUNDING

This work was supported by the Medical Research Council (to N.J.B., P.J.M. and D.C), Imperial College (to M.S.) and a 30th Anniversary Research into Ageing Fellowship Award (to C.T.)

Acknowledgments

We are very grateful to Professor Michel Goedert (MRC Laboratory of Molecular Biology, University of Cambridge, Cambridge, U.K.) for the pRK172-htau46 plasmid, Professor Peter Davies (Department of Pathology, Albert Einstein College of Medicine, Bronx, NY, U.S.A.) for the PHF-1 antibody and Peter Seubert (Elan Pharmaceuticals, San Francisco, CA, U.S.A.) for the 12E8 antibody.

Abbreviations: ACC, acetyl Co-A carboxylase; AMPK, AMP-activated protein kinase; APP, amyloid precursor protein; Aβ, amyloid β-peptide; BSRK, brain-specific kinase; CaMKK, Ca2+/calmodulin-dependent protein kinase kinase; cdk5, cyclin-dependent kinase 5; GSK3β, glycogen synthase kinase 3β; JNK, c-Jun N-terminal kinase; MARK, microtubule-associated protein-regulating kinase/microtubule affinity-regulating kinase; NFT, neurofibrillary tangle; NMDA, N-methyl-D-aspartate; PHF-1, paired helical filament-1; PP, protein phosphatase; WT, wild-type

References

View Abstract