Calcium-mediated Transient Phosphorylation of Tau and Amyloid Precursor Protein Followed by Intraneuronal Amyloid-β Accumulation*

Intraneuronal accumulation of hyperphosphorylated protein tau in paired helical filaments together with amyloid-β peptide (Aβ) deposits confirm the clinical diagnosis of Alzheimer disease. A common cellular mechanism leading to the production of these potent toxins remains elusive. Here we show that, in cultured neurons, membrane depolarization induced a calcium-mediated transient phosphorylation of both microtubule-associated protein tau and amyloid precursor protein (APP), followed by a dephosphorylation of these proteins. Phosphorylation was mediated by glycogen synthase kinase 3 and cyclin-dependent kinase 5 protein kinases, while calcineurin was responsible for dephosphorylation. Following the transient phosphorylation of APP, intraneuronal Aβ accumulated and induced neurotoxicity. Phosphorylation of APP on Thr-668 was indispensable for intraneuronal accumulation of Aβ. Our data demonstrate that an increase in cytosolic calcium concentration induces modifications of neuronal metabolism of APP and tau, similar to those found in Alzheimer disease.

The coexistence of neurofibrillary tangles and senile plaques in the brain confirms the clinical diagnosis of Alzheimer disease (AD) 2 (1). Intraneuronal neurofibrillary tangles are made of paired helical filaments containing the hyperphosphorylated microtubule-associated protein tau (2). The amyloid core of extracellular senile plaques contains fibrils of amyloid ␤ peptide (A␤) (3,4), which results from the cleavage of the amyloid precursor protein APP (5). The most compelling evidence that A␤ is the causative agent of AD comes from observations on genetic mutations that cause familial forms of AD. Mutations of APP or presenilin genes alter the processing of APP (6), giving rise to increased production of A␤, in particular of A␤ containing 42 amino acids (A␤1-42), which is more prone to aggregation (7). These genetic data founded the amyloid cascade hypothesis (8) that implies that plaques should develop before tangles. However, neuropathological studies have indicated that the initial development of tangles precedes the development of amyloid plaques by at least two decades (9). To reconcile these controversial observations, it has been proposed that the initial neurofibrillary changes are independent of A␤ (10), while this process is accelerated by the presence of A␤, which can stimulate AD-like phosphorylation of tau in neuronal cultures (11). For many years, the amyloid cascade hypothesis maintained that memory failure in AD derived from neuronal death induced by insoluble extracellular deposits of amyloid fibrils. Newer findings, however, demonstrate that accumulation of intraneuronal A␤ is neurotoxic (12,13). A common alteration of the neuronal metabolism of APP and tau, leading to the production of both hyperphosphorylated tau and A␤, remains elusive. Here we show that in cultured neurons membrane depolarization induced a calcium-mediated transient phosphorylation of APP and tau by the Cdk5 and GSK3 protein kinases, followed by a dephosphorylation of both proteins. Following transient phosphorylation of APP, intraneuronal A␤1-42 accumulated and induced important neurotoxicity. This intraneuronal accumulation of A␤1-42 was not observed following expression of the APPT668A mutant, indicating that phosphorylation of APP on Thr-668 was indispensable. Thus, an increase of neuronal cytosolic calcium concentration may lead to both neuronal AD-like phosphorylation of tau and intraneuronal accumulation of A␤, which induces neuronal death in AD.

EXPERIMENTAL PROCEDURES
Neuronal Culture-Primary cultures of cortical neurons were prepared from embryonic day 17 rat brains as described (14). All cell culture reagents were purchased from Invitrogen. Cortices were dissected and incubated in Hanks' Buffered Salt Solution medium (1 mM sodium pyruvate, 10 mM HEPES without Ca 2ϩ and Mg 2ϩ ), dissociated in the same medium containing 1.26 mM Ca 2ϩ , 0.50 mM MgCl 2 , and 0.41 mM MgSO 4 . Cells were plated in 6-, 12-, or 96-well culture dishes (4 ϫ 10 5 cells/ cm 2 ), pretreated with 10 g/ml poly L-lysine, and cultured for 6 days in vitro in Neurobasal TM medium ϩ B-27 supplement and 0.5 mM L-glutamine prior to infection with recombinant adeno-viruses. Under these conditions, neuronal cultures (up to 98% neurons) display high differentiation and survival rates (15).
Viral Infections-The construction and purification of recombinant adenovirus encoding wild-type human APP695 (AdRSVAPP) was performed as described previously (16). Thr-668 of wild-type human APP695 was mutated to alanine (APPT668A) by PCR, and the mutation was confirmed by sequencing. AdRSVAPPTA was constructed and purified using the AdEasy TM XL adenoviral vector system (Stratagene, La Jolla, CA). After 6 days in vitro, neurons were infected at a multiplicity of infection of 100 for 4 h in a minimal volume of culture medium. The infection medium was then replaced by fresh culture medium for 4 days. Under these conditions, at least 75% of neurons express the proteins encoded by recombinant adenoviruses (14).
Neuronal Treatments-Four days after infection, cells were depolarized in neuronal culture medium containing 35 mM KCl (final concentration) for increasing periods of time up to 8 h. In some experiments, neurons were depolarized during 10 min before reincubation for different times up to 7 h in a culture medium without KCl. Depolarization did not induce any modification of APP or tau expression. In some experiments, neurons were pretreated for 1 h with 25 M SB415286, an ATPcompetitive inhibitor of GSK3 kindly provided by A. Goffinet (Université Catholique de Louvain, Brussels, Belgium), with 10 M roscovitine (Calbiochem) for 1 h, with 1 M nimodipine (Bayer, Leverkusen, Germany) for 10 min, with 4 mM EGTA (Merck) for 2 h or with 250 nM N-[N- (3,5-difluorophenylacetyl-L-alanyl)]-S-phenylglycine t-butyl ester (DAPT), a functional inhibitor of ␥-secretase kindly provided by L. Mercken (Sanofi-Aventis, France) for 2 h, with 10 M calmidazolium (Calbiochem) for 1 h, with 10 M cyclosporine A (Calbiochem) for 1 h before a further depolarization in the presence of these components. All these treatments did not affect APP or tau expression or neuronal survival.
Immunoprecipitation of APP-Neurons (4 ϫ 10 6 cells plated in each well of a 6-well plate) were scraped and pelleted in cold phosphate-buffered saline. Cells were solubilized in radioimmune precipitation assay buffer (RIPA 1 ϫ 25 mM Tris, pH 7.4, 0.5% Triton X-100, 0.5% Nonidet P-40) containing protease inhibitors (1 g/ml pepstatin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride) and phosphatase inhibitors (vanadate 1 mM, nafamostat mesylate 20 mM, EGTA 1 mM, ␤-glycerophosphate pentahydrate 5 mM, and sodium pyrophosphate 5 mM) purchased from Roche Applied Science. The samples were immunoprecipitated with 1.5 g of WO-2 antibody (described above) overnight at 4°C. Protein A-Sepharose (5 mg/ml; Amersham Biosciences) was then added for 4 h. The samples were centrifuged at 15,800 ϫ g for 2 min at 4°C, and the pellets were washed twice with RIPA 1 ϫ and once with Tris-buffered saline (10 mM Tris, pH 7.5). Pellets were incubated for 5 min at 96°C in sample buffer (see above). The samples were then centrifuged at 15,800 ϫ g for 2 min at 4°C, and the supernatants were analyzed by Western blotting on a 4 -12% Nupage TM gel using the WO-2 or anti-PT668 antibodies as described previously.
Extraction and Quantification of A␤-A␤ was extracted from cell lysates as described previously (20) with slight modifications. Neurons (12 ϫ 10 6 cells) plated in three wells of a 6-well plate were solubilized in 150 l of formic acid (70%), cleared by centrifugation (55,000 ϫ g, 4°C, 20 min), and supernatant fluids were vacuum dried. The resulting pellet was resuspended in 1 ml of alkaline carbonate buffer (2% Na 2 CO 3 , 0.1 N NaOH) and centrifuged (15,800 ϫ g, 2 min, 4°C). Protein concentration was measured on 50 l of the resulting supernatant fluid, called protein solution, using the BCA protein assay (Pierce). For intracellular A␤ quantification, 100 l of the remaining protein solution was neutralized and diluted 1:3 in H 2 O containing 10% (v/v) fetal calf serum, 0.5% Triton X-100 (v/v), and 0.5% (v/v) Nonidet P-40 (final concentration). For extracellular A␤ quantification, 100 l of culture medium, collected after neuronal treatments, was treated with protease inhibitors (see above) and cleared by centrifugation (15,800 ϫ g, 5 min, 4°C). The quantification of A␤ 1-40 and A␤ 1-42 isoforms was performed by fluorescent sandwich enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's instructions (BIOSOURCE). Standard curves of synthetic A␤ 1-40 and A␤ 1-42 were performed in the same medium as the samples. Control experiments demonstrated that there is no cross-reaction between A␤ 1-40 and A␤ 1-42 detection. Fluorescence emission was measured on an HTS 7000 Plus plate reader (PerkinElmer) at excitation/emission wavelengths of 485/535 nm, respectively. A␤ 1-42 intracellular concentrations were expressed as pg/mg of cellular protein.
Statistical Analysis and Presentation of the Results-The number of samples (n) in each experimental condition is indicated in the figure legends. Statistical analysis was performed by one-way analysis of variance followed by Bonferroni's multiple comparison post-test. Differences were considered as significant when p Ͻ0.05 (**, p Ͻ0.01; ***, p Ͻ0.001). Each Western blot presented in the figures is representative of at least three independent experiments.

RESULTS
Neuronal Depolarization Affects Tau Phosphorylation-Accumulation of phosphorylated microtubule-associated protein tau in affected neurons in AD has been attributed to increased kinase or decreased phosphatase activity (21,22). Many neuronal protein kinase and phosphatase activities are calcium dependent, and an increase in cytosolic calcium concentration triggers the hyperphosphorylation of tau protein (23). We investigated whether tau phosphorylation would be affected by neuronal depolarization. Primary cultures of rat cortical neurons were incubated in the presence of 35 mM KCl, and the phosphorylation of tau was measured by increased immunoreactivity toward the PHF1 antibody directed against phosphotau Ser-396/404 (18). Although depolarization did not affect tau expression, a small mobility shift of tau was observed together with an increase in tau phosphorylation after 30 min of depolarization (Fig. 1A). In depolarizing conditions, activation of GSK3␤ was demonstrated using a phospho-specific antibody recognizing active GSK3␤ phosphorylated on Tyr-216 ( Fig. 1, B and C). In the same experimental conditions, activation of Cdk5 was demonstrated by the processing of p35 to p25 (Fig. 1D). In a kinetic study conducted for up to 8 h, an increase in tau phosphorylation was followed by a dephosphorylation corresponding to a decrease in tau phosphorylation below its initial state (Fig. 1, E and F). A non-toxic concentration of SB415285 completely inhibited GSK3␤-mediated phosphorylation of tau but did not affect dephosphorylation of the protein (Fig. 1, E and F).
Neuronal Depolarization Affects APP Phosphorylation-Phosphorylation of APP affects the production of A␤ (24,25). The Thr-668 of APP695 can be phosphorylated by Cdk5 and GSK3 protein kinases (26,27), which are activated in depolarizing conditions (Fig. 1, B-D) and can be modulated by calcium (23,28). We therefore investigated whether neuronal depolarization could modify the phosphorylation of APP695 on Thr-668. Following adenoviral expression of human APP695, rat cultured neurons were incubated for different times in the presence of 35 mM KCl, and the phosphorylation of both transfected Asterisk indicates the significant difference between each condition as determined by paired t test (*, p Ͻ0.05) D, under the same experimental condition processing of p35 to p25 was analyzed by Western blotting using the C19 antibody. Neurons were depolarized and incubated for different times, up to 8 h, in the presence (ϩ) or the absence (Ϫ) of the GSK3 inhibitor SB415286. E and F, the P-tau/tau signal ratio, detected in each experimental condition with an electrophoresis gel imaging system, was quantified and expressed (mean Ϯ S.E.) as relative values; n ϭ 4. DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 and endogenous APP was measured using an anti-phospho-APP Thr-668 antibody. Depolarization did not affect APP expression but induced a transient phosphorylation of APP, followed by a dephosphorylation of the protein (Fig. 2, A and B), with kinetics similar to that observed for the phosphorylation of tau (Fig. 1, E and F). This effect on APP phosphorylation was calcium dependent, because no modification of APP phosphorylation was observed when neurons were depolarized in a culture medium without calcium (Fig. 2, A and B). Roscovitine, an inhibitor of Cdk5 (29), or SB415286, an inhibitor of GSK3 (30), partly inhibited the phosphorylation of APP. Co-treatment of neurons with non-toxic concentrations of both inhibitors completely inhibited the phosphorylation of APP but did not affect the dephosphorylation of the protein (Fig. 2, A and B). Because depolarization induced a rapid phosphorylation of APP, we investigated whether a short preincubation of neurons in the presence of 35 mM KCl followed by their reincubation for different times without KCl (Fig. 3A) was sufficient to induce APP phosphorylation. In these experimental conditions, an increase in APP phosphorylation was measured during the first 2 h of reincubation. A longer reincubation did not induce APP dephosphorylation below its initial phosphorylation state (Fig.  3, B and C), indicating that dephosphorylation of APP only occurred when neurons were maintained in 35 mM KCl for longer periods of time.

One Cellular Mechanism for the Two AD Lesions
Dephosphorylation of APP was inhibited by calmidazolium, indicating the contribution of a calmodulin-dependent phosphatase activity (Fig. 4, A and B). Calcineurin is a calcium/calmodulin neuronal phosphatase that is inhibited by cyclosporine A. Dephosphorylation of APP was indeed inhibited by cyclosporine A, which did not modify phosphorylation of the protein (Fig. 4, A and B). Cyclosporin A also inhibited dephosphorylation of tau by maintaining tau phosphorylation for a longer period of time (Fig. 4, C and D).
Transient Phosphorylation of APP Induces Intraneuronal Accumulation of A␤-Following adenoviral expression of human APP695 in rat cultured neurons, the production of human A␤ was measured by ELISA. Intraneuronal A␤ 1-42, but not A␤ 1-40, was detected after 2 h of depolarization and progressively accumulated in neurons up to 8 h (Fig. 5A). The amount of extracellular A␤1-40 was not affected by depolarization (13). Because intraneuronal A␤ 1-42 accumulated during the dephosphorylation period of APP, we investigated whether phosphorylation or dephosphorylation of APP was responsible for A␤ production. Nimodipine, a specific antagonist of L voltage-sensitive calcium channels (31), completely inhibited the phosphorylation of APP but did not affect the dephosphorylation of the protein (Fig. 5, B and C). When neurons were depolarized in the presence of nimodipine, intraneuronal A␤ was not detectable by ELISA (not shown), indicating that phosphorylation of APP rather than its dephosphorylation was needed for intraneuronal A␤ production. In the presence of roscovitine, the initial increase in APP phosphorylation was still observed (Fig. 2, A and B) and was sufficient to produce intraneuronal A␤, because roscovitine did not significantly decrease

the amount of intraneuronal A␤ recovered after an 8-h depolarization (not shown).
Phosphorylation of APP on Thr-668 Is Indispensable for the Neurotoxic Accumulation of Intraneuronal A␤-To confirm that phosphorylation of APP695 on Thr-668 is needed for intraneuronal A␤ accumulation, a human APPT668A mutant was expressed in rat cultured neurons using recombinant adenoviruses. Similar levels of APP and APPT668A were detected by immunoblotting using the WO-2 antibody, which specifically detects human APP (Fig. 6A). Following immunoprecipitation with the WO-2 antibody, the anti-phospho-APP Thr-668 antibody detected human APP, but not human APPT668A (Fig. 6B). When neurons expressing human APPT668A were depolarized in the presence of 35 mM KCl, intraneuronal A␤ was not detected by ELISA (Fig. 6C), confirming that the phosphorylation of APP on Thr-668 was indispensable for intraneuronal A␤ 1-42 accumulation. The neurotoxicity induced by intraneuronal accumulation of A␤ 1-42 (12,32) is prevented by inhibition of A␤ production in depolarized neurons (13). Because APPT668A did not produce any intraneuronal A␤, we investigated whether differences in toxicity could be observed following depolarization of neurons expressing either human APP or human APPT668A. In control neurons, Ͼ80% of cell survival was measured following a 72-h depolarization (Fig. 6D). On the contrary, in neurons expressing APP only 60% of cell survival was measured, and a significant recovery was observed in the presence of 250 nM DAPT, a functional inhibitor of ␥-secretase (Fig. 6D). In neurons expressing APPT668A, no toxicity related to A␤ production was observed (Fig. 6D).

DISCUSSION
In 1985, the major constituents of senile plaques and neurofibrillary tangles, which coexist in AD, were identified as A␤ and tau, respectively (2,4). Since then, investigators studying the metabolism of tau have provided strong evidence that hyperphosphorylated protein tau plays a key role in the development of AD. In addition, neurofibrillary tangles precede the formation of senile plaques (9,33). Otherwise, strong evidence that A␤ is the causative agent of AD was provided by the study of familial cases in which gene mutations give rise to an increased production of A␤ (6). The challenge is to identify common cellular mechanisms that could account for the formation of both typical AD lesions. Transgenic mice expressing human APP and presenilin gene mutations show amyloid deposits in their brain but do not display any neurofibrillary tangles (34). These latter lesions are found in the brain of transgenic mice carrying mutations of the human tau gene (11). The

. The use of calcineurin inhibitors suggests that dephosphorylation of APP triggered by an increase of intracellular calcium concentration is performed by the calcium/calmodulin-dependent protein phosphatase calcineurin.
Rat cortical neurons expressing human APP695 were depolarized with 35 mM KCl and incubated for increasing periods of time, up to 8 h, in the presence (ϩ) or the absence (Ϫ) of calcineurin inhibitors (calmidazolium, cal, or cyclosporine A, CyA). A, the expression of APP or phospho-APP (P-APP) in cell extracts was analyzed by Western blotting using the anti-Cter (upper panel) or the anti-PT668 (lower panels) antibody. B, the P-APP/APP signal ratio, detected in each experimental condition, was quantified and expressed (mean Ϯ S.E.) as relative values; n ϭ 4. C, the expression of phospho-tau (P-tau) in cell extracts was analyzed by Western blotting using the PHF1 antibody. D, the P-tau signal, detected in each experimental condition, was quantified and expressed (mean Ϯ S.E.) as relative values; n ϭ 3. DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 only way to reproduce both typical AD lesions in vivo is to create transgenic animals expressing human APP, presenilin, and tau gene mutations (35), a situation that never occurs in AD.

One Cellular Mechanism for the Two AD Lesions
In cultured cells, extracellular A␤ neurotoxicity requires fibril formation (36,37) and is prevented by molecules inhibiting the polymerization of A␤ (38). Such neurotoxicity could be related to tau phosphorylation, because extracellular A␤ fibrils stimulate AD-like phosphorylation of tau (11). In contrast, extracellular monomeric A␤ is not neurotoxic (12,39). In cultured neurons constitutively producing A␤, neurotoxicity is only observed when intraneuronal A␤ accumulates (12,32). In transgenic mice as well as in AD patients, intraneuronal accumulation of A␤ precedes the formation of extracellular amyloid deposits (40) and a structural organization of intraneuronal A␤ as small soluble oligomers induces important neurotoxicity (41).
Here, we identified a common cellular mechanism leading to transient phosphorylation of both tau and APP followed by intraneuronal accumulation of A␤. Following depolarization, an increase in cytosolic calcium concentration can activate several signaling pathways and different protein kinases and phosphatases. Consequently, phosphorylation of APP and tau was observed, followed by dephosphorylation of the proteins. Transient phosphorylations of tau and APP showed similar kinetics, indicating that phosphorylation of one protein is not needed for the subsequent phosphorylation of the other one. The use of phospho-specific antibodies recognizing active GSK3␤ and the analysis of the processing of p35 to p25 demonstrated the activation of GSK3␤ and Cdk5 in depolarizing conditions. The use of inhibitors identified Cdk5 and GSK3 as protein kinases involved in phosphorylation of APP on Thr-668, confirming previous observations indicating that Thr-668 of APP is indeed a substrate of these protein kinases (26,42). Cdk5 and GSK3␤ are also involved in AD-like phosphorylation of tau (43,44). Following neuronal depolarization, a transient phosphorylation of tau was demonstrated by an increased immunoreactivity toward the PHF1 antibody. Concomitantly, a mobility shift of the protein was observed, although rather small because tau from primary cultures of embryonic neurons is highly phosphorylated. A transient increase in intracellular calcium concentration was previously demonstrated to induce a GSK3␤-  mediated phosphorylation of tau (23), and a biphasic effect of calcium influx on tau phosphorylation was also observed (45). Although we have no evidence that tau phosphorylation is associated with intraneuronal A␤ accumulation, the transient phosphorylation of APP on Thr-668 was indispensable to induce the accumulation of intraneuronal A␤1-42. Inhibition of phosphorylation of APP on Thr-668 was previously demonstrated to influence the amyloidogenic catabolic pathway of APP (24,25).
Cdk5 and its regulatory subunit p35 play a key role in the development of the mammalian central nervous system (46). A calcium-dependent proteolytic cleavage of p35 generates p25, leading to aberrant Cdk5 activation (47,48). In transgenic mice overexpressing p25 in the postnatal forebrain, endogenous tau was hyperphosphorylated and neurofibrillary tangles developed in the brain (49). The absence of amyloid deposits in these animals could result from the absence of human APP expression, because amyloid deposits were found only in transgenic animals expressing mutated human APP (50,51).
Because a modification in intraneuronal calcium concentrations can induce both phosphorylation of tau and the intraneuronal accumulation of A␤, neuronal calcium homeostasis has to be precisely regulated. Presenilins were recently demonstrated to form endoplasmic reticulum calcium leak channels, a function disrupted by familial AD-linked mutations (52). This calcium-signaling function for presenilins provides support for the calcium hypothesis of AD linked to presenilin mutations. In sporadic AD, alterations in cellular calcium homeostasis are associated with neurodegenerative processes (53,54). A major protein involved in the control of neuronal calcium homeostasis is the N-methyl-D-aspartate ionotropic glutamate receptor. A decreased glutamate uptake increases extracellular glutamate levels in aging and in AD fibroblasts (55), and an excessive stimulation of the N-methyl-D-aspartate receptor triggers neurotoxic intracellular calcium accumulation. Memantine, a noncompetitive antagonist of the N-methyl-D-aspartate receptor, is the only anti-glutamatergic drug currently approved for the treatment of moderate to severe AD (56,57). The search for molecules able to control calcium homeostasis as potential drugs is actively underway and might prove to be effective in AD in view of the present results.