Inhibition of protein phosphatase 2A overrides tau protein kinase I/glycogen synthase kinase 3 beta and cyclin-dependent kinase 5 inhibition and results in tau hyperphosphorylation in the hippocampus of starved mouse.

Hyperphosphorylated tau is the major component of paired helical filaments in neurofibrillary tangles found in Alzheimer's disease (AD) brain. Starvation of adult mice induces tau hyperphosphorylation at many paired helical filaments sites and with a similar regional selectivity as those in AD, suggesting that a common mechanism may be mobilized. Here we investigated the mechanism of starvation-induced tau hyperphosphorylation in terms of tau kinases and Ser/Thr protein phosphatases (PP), and the results were compared with those reported in AD brain. During starvation, tau hyperphosphorylation at specific epitopes was accompanied by decreases in tau protein kinase I/glycogen synthase kinase 3 beta (TPKI/GSK3 beta), cyclin-dependent kinase 5 (cdk5), and PP2A activities toward tau. These results demonstrate that the activation of TPKI/GSK3 beta and cdk5 is not necessary to obtain hyperphosphorylated tau in vivo, and indicate that inhibition of PP2A is likely the dominant factor in inducing tau hyperphosphorylation in the starved mouse, overriding the inhibition of key tau kinases such as TPKI/GSK3 beta and cdk5. Furthermore, these data give strong support to the hypothesis that PP2A is important for the regulation of tau phosphorylation in the adult brain, and provide in vivo evidence in support of a central role of PP2A in tau hyperphosphorylation in AD.

marks called senile plaques and neurofibrillary tangles. The former are deposits of the ␤-amyloid peptide (A␤) (1), whereas neurofibrillary tangles consist of hyperphosphorylated tau protein assembled in paired helical filaments (PHF) (2). Hyperphosphorylation refers to the state that tau is phosphorylated at more sites than tau from adult brain and that, for a given site, a higher than normal percentage of tau molecules is phosphorylated (3). Tau is a microtubule-associated protein, and its normal physiological function is to bind and stabilize microtubules. In vitro studies have shown that PHF-tau fails to promote microtubule assembly (4 -9), and thus it has been proposed to lead to microtubule destabilization, appearance of neurofibrillary tangles, and neurodegeneration in AD brain (10). Phosphorylation of tau can be regulated by many protein kinases and phosphatases in vitro (11)(12)(13). These findings, and the changes in kinases and phosphatases observed in AD, suggest that tau hyperphosphorylation in AD brain is likely to be due to an imbalance of the protein phosphorylation and dephosphorylation systems. But to date the mechanism of conversion from normal adult tau to hyperphosphorylated tau, the significance of tau hyperphosphorylation in PHF formation, and its relationship to A␤ deposition remain largely elusive.
Tau hyperphosphorylation is a physiological reversible response of the brain to stressful conditions like cold water stress 2 (14), heat-shock (15), or starvation (16). Starvation induces decreases in circulating glucose, insulin, and leptin, and increases in corticosterone (17), and results in a large decrease of glucose in many parts of the brain (18). Reduced glucose metabolism is associated with certain regions of the brain of probable and preclinical AD subjects (19,20), and is considered by some as aggravating, if not causative in the development of AD (21). These data, combined with the observations that starvation induces tau hyperphosphorylation at many PHF-tau epitopes, and with a regional selectivity analogous to that in AD, suggest that a common mechanism might be mobilized in starvation and AD for tau hyperphosphorylation (16).
Here, we studied the mechanism of tau hyperphosphorylation in the starved mice in terms of protein kinases and phosphatases which, according to previous studies, are either implicated in the generation of tau pathological epitopes in vitro or exhibit altered activity in AD brain, and evaluated the relevance of our findings to AD. The most striking changes concomitant with tau hyperphosphorylation at specific sites were the inhibition of PP2A, concurrent with the inhibition of TPKI/GSK3␤, and cdk5 activities toward tau. These results demonstrate that the activation of TPKI/GSK3␤ and cdk5 is not necessary to obtain hyperphosphorylated tau in vivo, and indicate that tau site-specific hyperphosphorylation in the starved mouse hippocampus involves a complex mechanism in which PP2A inhibition plays a dominant role, overriding the inhibition of key tau kinases such as TPKI/GSK3␤ and cdk5.

EXPERIMENTAL PROCEDURES
Mice-Eight to ten-week-old C57BL/6NJcl male mice (Clea Japan, Tokyo) were singly housed in cages with grid floors to deny coprophagy. Food was removed for up to 3 days, but mice were allowed free access to water. Room temperature was 23°C, and the light period was 08:00 -20:00. Animals were handled according to the procedures approved by the Animal Care and Use Committee of the Mitsubishi Kasei Institute of Life Sciences.
Protein Extraction and SDS-PAGE-Mice were sacrificed by cervical dislocation, brains immediately removed, and hippocampi were dissected from brains in ice-chilled saline. The tissues were quickly weighed and homogenized in 10 times volume/weight of Oϩ buffer, modified from O'Farrell's buffer O (22) (62.5 mM Tris-HCl, pH 6.8, 10% (w/v) glycerol, 5% (v/v) 2-mercaptoethanol, 2.3% (w/v) SDS, 100 M orthovanadate, 1 M okadaic acid (OA), 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM EGTA, and 1 mM EDTA). The samples were then placed in boiling water for 5 min, centrifuged for 15 min at 20,000 ϫ g at 4°C, and the protein content of the supernatants was determined with the Bio-Rad Protein Assay (after proper dilution to ensure reagent compatibility) using bovine serum albumin in equivalent buffer as standard. Eighteen g of protein (determined to fit in the linear range for quantification, calibration data not shown) were separated by 10% SDS-polyacrylamide gel electrophoresis, and electrotransferred to nitrocellulose membrane (Protran BA 85, 0.45 m, Schleicher & Schuell).
Western Blotting and Analysis of Results-Membrane blocking and antibody incubations were done according to New England Biolab MAPK Immunoblotting Protocol (New England Biolabs number 9910), with appropriate primary and secondary Ig dilutions. Anti-mouse, antirabbit, and anti-goat horseradish peroxidase-conjugated secondary antibodies were purchased from Santa Cruz Biotechnology. The bands were developed with SuperSignal West Pico Chemiluminescent Substrate (Pierce) or with 4-chloro-1-naphthol and H 2 O 2 . The images obtained with the LAS-1000plus Luminescent Image Analyzer (Fuji Film), or scanned from stained membranes, were quantified using a Macintosh version of Fuji Films Science Lab 99 Image Gauge. Statistical analysis of the results was performed by one-way analysis of variance (ANOVA). Significant ANOVA (p Ͻ 0.05) was followed by the Tukey/Kramer test of pairwise multiple comparisons. In all the figures, error bars represent the standard deviation (S.D.), while * and ** indicate significant difference with p Ͻ 0.05 and p Ͻ 0.01, respectively.
Lithium Administration-LiCl experiments were conducted on 6 groups of 3 mice. At 10:00 on day 1 through day 3, mice of groups 2 and 5 received intraperitoneal injection with 400 l of 0.3 M NaCl, and mice of groups 3 and 6 were injected with 400 l of 0.3 M LiCl. Mice of groups 4 to 6 were deprived of food at 10:00 on day 2. The mice were sacrificed in the morning of day 4, and hippocampi were taken and analyzed as described above.
Immunoprecipitation was carried out at 4°C. Five l of fresh brain extract were mixed with complexes of T1.7 or Cdk5 (C-8) antibodies coupled to IgG beads (Dynabeads, rat anti-mouse IgG1 M-450 or sheep anti-rabbit M-280, prepared according to the manufacturer's instructions) suspended in 50 l of cell extraction buffer (10 mM Tris-HCl, pH 7.4, 50 mM NaCl, 50 mM NaF, 1 mM EDTA, 1 mM EGTA, 50 mM ␤-glycerophosphate, 3 mM benzamidine, 0.05% NaN 3 , 100 M Na 3 VO 4 , 1 g/ml each of leupeptin, aprotinin and pepstatin, 1 mM PMSF, 1 M okadaic acid) and incubated for 1 h at 4°C on rotating agitator. After washing 3 times with cell extraction buffer, and twice with kinase activity buffer (KAB; 100 mM MES-NaOH, pH 6.2, 1 mM Mg acetate, 1 mM EGTA, 10% glycerol, 0.02% Tween 20, 1 mM PMSF, 1 M okadaic acid), the samples were resuspended in 10 l of KAB. Kinase activity was measured by phosphorylation of human recombinant tau-441 (Panvera). Ten-l aliquots of suspension of immunoprecipitated kinases in KAB were mixed with 10 l of KAB containing 200 M ATP and 100 g/ml tau, and incubated at 30°C for 20 min (for TPKI) or 2 h (for cdk5). The reaction was stopped by adding 20 l of Oϩ buffer and boiling for 5 min. Five l of this mixture were loaded per lane on 10% SDS-PAGE gels, followed by immunoblotting and band quantification with phospho-tau-specific antibodies as described above.
Phosphatase Assays-To prepare phosphatase substrate, 1 l of fresh RIPA buffer extract of hippocampus (10 ϫ v/w) was added to 25 l of tau phosphorylation mixture (100 g/ml recombinant tau, 1 mM ATP, 10 mM Tris-HCl, pH 7.5, 2 mM 2-mercaptoethanol, 2 mM Mg acetate, 1 mM PMSF, 1 M okadaic acid, 1 g/ml aprotinin, leupeptin, and pepstatin) and incubated for 4 h at 37°C. This procedure allows incorporation of 6 -7 mol of phosphate/mol of recombinant tau, as determined by a standard radiometric assay using [␥-32 P]ATP (data not shown). The reaction was terminated by boiling for 5 min followed by centrifugation for 15 min at 20,000 ϫ g and 4°C. After addition of trichloroacetic acid (1/9, w/v), the supernatant was incubated for 30 min on ice, and centrifuged for 10 min at 20,000 ϫ g and 4°C. The pellet was washed 3 times with 20% trichloroacetic acid, redissolved in the same volume of substrate solubilization buffer (50 mM Tris-HCl, pH 8.5, 0.1 mM EDTA, 5 mM 2-mercaptoethanol, 0.01% Tween 20) as used during phosphorylation, and dialyzed twice overnight at 4°C against 1 liter of dialysis buffer (25 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 5 mM 2-mercaptoethanol, 0.01% Tween 20), to remove any traces of OA. Phosphatase activity was measured by decrease in phosphorylation of tau by the brain extracts as analyzed by Western blotting with anti-phospho-tau antibodies. Mouse hippocampi were homogenized in 5 ϫ volume/weight of ice-chilled phosphatase sample buffer (50 mM Tris-HCl, pH 7.0, 0.25 M sucrose, 10 mM 2-mercaptoethanol, 0.1 mM EDTA, 1 mM benzamidine, 1 g/ml each of aprotinin, leupeptin, and pepstatin, 1 mM PMSF) modified from Gong et al. (31), centrifuged for 30 min at 20,000 ϫ g and 4°C, and the protein contents were determined as above. One l of phosphatase extract was added to 20 l of phosphatase activity mixture (50 g/ml phosphorylated tau, 37.5 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 5 mM 2-mercaptoethanol, 0.01% Tween 20, 1 mM PMSF, 1 g/ml each of aprotinin, leupeptin, and pepstatin) and incubated at 30°C for various periods of time. The reaction was stopped by addition of an equal volume of Oϩ buffer and boiling for 5 min. Five-l aliquots were loaded on 10% SDS-PAGE gels and analyzed as above. PP1 and PP2A activities in the brain extracts toward phosphorylated recombinant tau were assessed after 10 min of incubation at 30°C by analyzing the dephosphorylation of AT8 epitope in the presence or absence of 5.0 M OA (inhibitory to PP2A and PP1, Calbiochem), or 1.0 M inhibitor-2 (PP1specific inhibitor, Sigma) (32-34). Phosphorylated tau was incubated in 20 l of phosphatase activity mixture for 1 h at 30°C in the presence or absence of 62.5 microunits/l of recombinant catalytic subunit of PP1 (New England Biolabs).
Neuronal Culture-Hippocampal neurons were prepared from 18day-old embryonic rat brains and cultured as described previously (35). These cultures were largely free of non-neuronal cells. Seven days after plating, OA was added to the medium to 100 nM. After 1 h of exposure, the cells were lysed in 5 volumes of RIPA buffer, and the extracts centrifuged for 15 min at 20,000 ϫ g and 4°C. The supernatants were analyzed by immunoblotting using anti-tau and anti-protein kinase antibodies. One g of protein was used for Tau-C, 10 g for p35 and TPKI PS9, and 2 g for other antibodies. Immunoblots were reacted with anti-rabbit or anti-mouse secondary antibodies conjugated to alkaline phosphatase and developed with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (Promega).

Starvation Up-regulates Activating Phosphorylation of MAPK and JNK, and Inhibitory Phosphorylation of TPKI/
GSK3␤-Batches of three or five young adult mice were starved for up to 3 days and some were refed for up to 1 week after 2 days of starvation. Hippocampal extracts of these mice were individually analyzed by immunoblotting with antibodies to protein kinases and phosphatases (Figs. 1-4). As described qualitatively (16), tau undergoes phosphorylation and dephosphorylation, illustrated here by phosphorylation-dependent anti-tau antibody AT8 (Fig. 1, A1 and B1). The two isoforms of rat GSK3, GSK3␣ and TPKI/GSK3␤, are encoded by different genes and share a 85% homology at the amino acid level (36). TPKI/GSK3␤ is a leading candidate protein kinase responsible for tau hyperphosphorylation in AD brains (11). This enzyme phosphorylates tau at PHF sites Ser 199 , Thr 231 , Ser 396 , Ser 404 , and Ser 413 (23) (numbering of amino acids according to the longest human tau (37)), but can also phosphorylate other sites in combination with other kinases (38). Phosphorylation of TPKI/GSK3␤ at Tyr 216 is essential for its activity, while phosphorylation at Ser 9 leads to partial inhibition (39). Phosphorylation at Tyr 216 did not change significantly (Fig. 1, A2 and B2), but starvation induced a dramatic elevation of phospho-Ser 9 level (Fig. 1, A3 and B2). PKB phosphorylates TPKI/GSK3␤ at Ser 9 (39), but the active-PKB epitope was increased only during the first day of refeeding (Fig. 1, A7 and B6). Interestingly, the activating phosphorylation of MAPK was transiently increased after 1 day of starvation (Fig. 1, A4 and B3). Activating phosphorylation of JNK, a kinase reported to phosphorylate tau at Thr 181 , Ser 202 /Thr 205 , Thr 231 , Ser 396 , and Ser 422 epitopes (40), rose during starvation (Fig. 1, A5 and B4). Levels of phospho-CaMKII immunoreactivity during starvation were not significantly different from control values (Fig. 1, A6 and B5). These observations indicate that, during starvation, MAPK is transiently activated, JNK activated, while TPKI/GSK3␤ is inhibited.
Decrease in p35 Immunoreactivity-Tau protein, as detected with anti-Tau-C antibody, occurred in normally fed mice as a strong band at 56 kDa and a faint one at 60 kDa, with a smear between 60 and 70 kDa (Fig. 2, A1, lanes 2-6). After 2 days of starvation, the 60-kDa band became prominent, the 56-kDa band weakened while 2 additional bands appeared at 64 and 70 kDa (Fig. 2, A1, lanes 12-16). These mobility shifts reflecting tau phosphorylation (41) were not accompanied by significant changes in total tau protein levels (Fig. 2, B1). Protein levels of TPKI/GSK3␤, MAPK, PKB, JNK, or catalytic subunits of PKA, did not show changes that correlated with tau phosphorylation levels ( Fig. 2, A2 -5, 8 and B2-5, 8). Tau protein kinase II (TPKII) is a heterodimer of a cdk5 catalytic subunit and p25, a 25-kDa regulatory subunit derived proteolytically from p35 (25,42). TPKII phosphorylates tau in vitro at PHF sites Ser 202 , Thr 205 , Ser 235 , and Ser 404 (23). Cdk5 immunoreactivity did not change significantly during starvation (Fig. 2, A7 and B7). Interestingly, p35 displayed a slight upward mobility shift upon starvation, followed by a decrease in intensity, and the strengthening of two lighter bands at 33 and 34 kDa, which are likely to be partial degradation products (Fig. 2, A6). One day of refeeding restored the upper band to its original intensity, and 1 week canceled the band shift (Fig. 2, A6 and B6). The level of p25 was undetectable. Thus, during starvation, the protein levels of tau or the kinases studied did not change, except for cdk5 activator p35 which displayed a slight mobility shift followed by a decrease.
Lithium Reduces Tau Hyperphosphorylation at Specific Sites-As starvation induced a sharp rise in TPKI/GSK3␤ phospho-Ser 9 , the role of this enzyme in tau hyperphosphorylation was further studied with an inhibitor of this enzyme, lithium, which reduces tau phosphorylation by inhibiting GSK3 both in vitro and in vivo (43,44). Daily injections of solutions of LiCl or control NaCl were given to fed or starved mice, and hippocampal extracts were analyzed for tau and TPKI/GSK3␤ phosphorylation (Fig. 3). Two days of starvation induced increases in tau phosphorylation levels by ϳ50 fold at AT8 and PT231, ϳ5 fold at Tau-1 and PS262, and ϳ65% at PS396, but not at PS199 and PS404 epitopes when compared with fed animals (Fig. 3, A2-8 and B2-8). Note that Tau-1 is a dephosphorylation-dependent epitope. NaCl injections did not produce a significant effect in the fed animals for all the antitau antibody tested when compared with fed controls (Fig. 3, A1-8, lanes 4 -6 and B1-8), and in the starved animals for Tau-1, PS199, PS396 ,and PS404, when compared with starved controls (Fig. 3, A4, 7, and 8 lanes 13-15 and B4, 7 and 8). On the other hand, NaCl injections in starved mice induced a decrease in immunoreactivity at AT8 and PT231 epitopes, and an increase at PS262 when compared with starved controls (Fig. 3, A3, 5 lanes 13-15 and B3, 5). Injections of either NaCl or LiCl in the fed mice induced a rise of phospho-Ser 9 immunoreactivity (Fig. 3, A10, lanes 1-9 and B9). LiCl injections did not produce a significant effect in the fed animals for all the anti-tau antibody tested when compared with fed controls (Fig.  3, A1-8, lanes 7-9 and B1-8), but dramatically reduced the starvation-induced tau hyperphosphorylation at the Tau-1 site (Fig. 3, A2, lanes 10 -12 versus 16 -18 and B2), along with the reappearance of the basal 56-kDa band as revealed by antibodies Tau-C, Tau-1, PS199, PS404, and PS396 (Fig. 3A, arrows at  right). Although of lesser statistical significance, LiCl seemed to have caused a further reduction in phosphorylation at PT231 site over the effect of NaCl (Fig. 3, A5, lanes 16 -18 versus  13-15, and B5). Phosphorylation level of TPKI/GSK3␤ Tyr 216 was not different over the entire experiment (Fig. 3, A9 and  B9), and phospho-Ser 9 levels were not different after the three conditions during starvation (Fig. 3, A10 and B9). Thus, starvation resulted in tau hyperphosphorylation at AT8, PT231, to a lesser extent at PS262 and Tau-1, and to a minor or negligible extent at PS199, PS396, and PS404 epitopes. Lithium injec- tions specifically reduced tau hyperphosphorylation at the Tau-1 site, and possibly also at PT231, along with the reappearance of the 56-kDa band. This probably reflects the fact that, while being partially inhibited by phosphorylation at Ser 9 during starvation, TPKI/GSK3␤ still contributes to tau hyperphosphorylation and can be further inhibited by lithium.
PP2A Catalytic Subunit Accumulates during Starvation-Our results above fail to explain the extent of tau hyperphosphorylation occurring during starvation in terms of kinase activation. Serine/threonine protein phosphatases (PP) were then studied. PP are classified into four types, PP1, PP2A, PP2B, and PP2C, on the basis of their specificity toward certain substrates, and sensitivity to specific activators and inhibitors (45). Biochemical studies have demonstrated that PP1, PP2A, and PP2B (calcineurin) can dephosphorylate tau while PP2C cannot (46 -48). PP2B and PP1 catalytic subunit immunoreactivities did not show significant changes during the course of the starvation and refeeding experiment (Fig. 4, A2, 3 and B2,  3). On the other hand, PP2A catalytic subunit (PP2Ac) immunoreactivity rose during 2 and 3 days of starvation (Fig. 4, A1  and B1).
Activities of total hippocampal PP were assessed by measuring the dephosphorylation of recombinant tau previously phosphorylated in vitro by a brain extract of normally fed mouse. After 2 h of incubation, AT8 and PT231 sites were dephosphorylated by ϳ90% by the hippocampal extract from fed mouse, and during the same time, PS262 was dephosphorylated by ϳ60%, PS199 and PS396 by ϳ15%, and PS404 by ϳ25% (Fig. 5, B1). The level of Tau-C immunoreactivity did not change, indicating negligible degradation of the substrate during the assay. Hippocampal extract from the starved mouse was much less efficient in dephosphorylating tau: after 15 min of incubation, tau phosphorylation at the AT8 epitope was still ϳ85% of the control, while it was down to ϳ25% with the extract from fed mouse (Fig. 5, B2).
Tau can be dephosphorylated by PP1, PP2A, or PP2B in vitro (46 -48). However, the effect of PP inhibitors on metabolically competent rat brain slices has demonstrated that PP2A is likely to be the main regulator of tau phosphorylation in the mammalian brain while PP2B is only minimally involved (34). To identify the type of PP which is involved and decreased in tau hyperphosphorylation in the brain of starved mice, the effect of 5.0 M OA (a concentration inhibitory to both PP1 and PP2A) or of 1.0 M inhibitor-2 (specific to PP1) on the in vitro dephosphorylation of tau by hippocampal extracts was studied. Five M OA completely inhibited the dephosphorylation of tau at AT8 epitope by fed mouse hippocampal extracts (Fig. 5, C1), indicating that the dephosphorylating activity is due to either PP1 or PP2A. The addition of 1.0 M inhibitor-2 did not have a significant effect on tau dephosphorylation by either fed or starved extracts (Fig. 5, C2, n ϭ 4 mice, quantitative data not shown), indicating that the AT8 epitope dephosphorylation is mainly due to PP2A and that, in the hippocampal extracts, PP1 has negligible activity toward tau. In vitro susceptibility of AT8 epitope to PP1 was confirmed by a preparation of recombinant PP1 catalytic subunit: PP1 completely dephosphorylated the AT8 epitope, and this dephosphorylation was prevented by 1.0 M inhibitor-2 (data not shown). These results indicate that the activities of hippocampal TPKI/GSK3␤, cdk5, and PP2A are reduced during starvation, and suggest that the dephosphorylating activity of hippocampal extracts toward tau is predominantly due to PP2A.
Okadaic Acid Reproduces in Cultured Neurons the Changes Observed during Starvation for Tau, p35, and TPKI/GSK3␤-The changes in tau phosphorylation at specific sites, and immunoreactivities of TPKI/GSK3␤ phospho-Ser 9 and p35 during starvation were compared with the effects of OA on cultured neurons, which have the advantage of being readily exposed to phosphatase inhibitors. OA was used at 100 nM, a concentration sufficient to inhibit PP2A and partially PP1, but not PP2B in vitro (49). One hour of OA application induced the electrophoretic retardation and sharpening of tau bands (Fig. 6A), and a robust elevation of phosphorylation at Tau-1, AT8, PT231, and PS262 epitopes (Fig. 6, B, C, E, and F). In contrast, the level of phosphorylation was almost the same at PS199, PS396, and PS404 (Fig. 6, D, G, and H, quantification data not shown, but compare with Fig. 6A). Analysis of the samples with anti-TPKI PS9 or anti-p35 antibodies revealed a sharp increase in TPKI/GSK3␤ phosphorylation at Ser 9 (Fig. 6I), and the upward mobility shift and decrease of p35 (Fig. 6J). The band for p25 was not detected. Treatment of cultured neurons with 1.0 M cyclosporin A, a specific inhibitor of PP2B, did not affect tau phosphorylation, TPKI/GSK3␤ phosphorylation at Ser 9 or p35 (data not shown). Thus, treatment of neuronal cultures with OA reproduced the changes observed during starvation in tau hyperphosphorylation at respective sites, increased TPKI/ GSK3␤ phosphorylation at Ser 9 , and degradation of p35.

DISCUSSION
The purpose of this study was to clarify the involvement of tau kinases and PP in the starvation-induced tau hyperphosphorylation, and to compare the findings with those reported for AD brains. We hypothesized that the plausible tau kinases such as TPKI/GSK3␤ and cdk5 would be activated during starvation. Contrary to our expectations, they were inhibited and hyperphosphorylation of tau at specific sites was accompanied by a decrease in PP2A activity toward tau. These results demonstrate that the activation of TPKI/GSK3␤ and cdk5 is not obligatory to obtain hyperphosphorylated tau in vivo. Here we propose a mechanism of tau hyperphosphorylation during starvation (Fig. 7) and discuss the relevance of our findings to AD. The early activation of MAPK and late activation of PKB, combined with the inhibition of PP would cause the phosphorylation of TPKI/ GSK3␤ at Ser 9 and a decrease in its activity. The inhibition of TPKI/ GSK3␤ could lead to further inhibition of the PP constituting an amplification loop. PP inhibition induces the phosphorylation-mediated degradation of p35 by the proteasome pathway, and thus results in cdk5 inhibition. PP2A inhibition leads to tau hyperphosphorylation, overriding TPKI/GSK3␤ and cdk5 inhibition, and triggers an autoregulatory mechanism leading to accumulation of its catalytic subunit.
Putative Mechanism of Tau Hyperphosphorylation during Starvation-Treatment of neuronal cultures with OA, an inhibitor of PP leading to tau hyperphosphorylation both in vitro and in vivo (reviewed in Ref. 12), reproduced the changes observed during starvation for tau phosphorylation: tau was hyperphosphorylated at Tau-1, AT8, PT231, and PS262 epitopes, but not or to a minor extent at PS199, PS396, and PS404 (Fig. 6). In hippocampal extracts the activity of PP2A toward AT8 epitope was inhibited in starved compared with fed mice, and the extent of tau hyperphosphorylation at the respective sites in the starved mouse hippocampus or in OA-treated neuronal cultures correlated with the extent of recombinant tau dephosphorylation by hippocampal extracts (AT8, PT231, PS262, Tau-1 Ͼ PS199, PS396, PS404). On the other hand, application of PP2B inhibitor to neuronal cultures, or addition of PP1 inhibitor to hippocampal extracts did not affect tau phosphorylation. These results advocate the idea that PP2A inhibition is chiefly accountable for tau hyperphosphorylation at AT8, PT231, Tau-1, and PS262 epitopes in the adult starved mouse (Fig. 7), consistent with reports suggesting that PP2A is the main regulator of tau phosphorylation in the mammalian brain (34, 50 -52), and with the hypothesis that PP activity primarily determines the level of tau phosphorylation in the developing and adult brains (53,54). The inhibition of PP2A may have caused the gradual increase of PP2Ac over 3 days of starvation (Fig. 4, A1 and B1), through an autoregulatory mechanism (Fig. 7) similar to the one leading to accumulation of PP2Ac upon inhibition of PP2A in fibroblasts (55). Starvation induces a reduction of glucose in the hippocampus, and glucose deprivation elicits tau neurofibrillary tangles-like antigenic changes in hippocampal neurons (56). It is thus tempting to suggest that the inhibition of PP2A in the hippocampus was a consequence of the decrease in hippocampal glucose. Other factors like changes in brain insulin could play a role in tau phosphorylation (43), but whether changes in circulating insulin alter the level of brain insulin remains a controversial issue. Actual causal chain from food deprivation to PP inhibition remains to be clarified (Fig. 7).
TPKI/GSK3␤ is thought to be one of the most plausible candidates responsible for tau hyperphosphorylation in AD (11). During starvation, however, the kinase activity was decreased, and this inhibition is most likely to be due to Ser 9 phosphorylation. The transient activation of MAPK after 1 day of starvation may have been partly responsible, through the 90-kDa ribosomal S6 kinase (p90 RSK) activation (39), for the early phosphorylation of Ser 9 . Similarly, the transient activation of PKB after 1 day of refeeding may have been partly responsible for the late phosphorylation of Ser 9 . During starvation, the inhibition of TPKI/GSK3␤ was likely further assisted by PP inhibition, because (i) PP2A activates TPKI/ GSK3␤ by dephosphorylating Ser 9 (57), (ii) inhibition of both PP1 and PP2A in rat brain slices leads to a decrease in GSK3 activity (52), and (iii) treatment of hippocampal cultures with OA increases Ser 9 phosphorylation. However, the maximal level of Ser 9 phosphorylation observed during starvation did not result in complete inhibition of TPKI/GSK3␤ toward tau, and this residual activity probably contributed to tau hyperphosphorylation at Tau-1 and possibly at PT231 epitopes because a GSK3 inhibitor, lithium, caused a decrease in phosphorylation at these sites. Hence, the inhibition of TPKI/GSK3␤ is probably the result of a network of kinases and phosphatases, with the inhibition of PP playing a prominent role in enhancing phosphorylation at Ser 9 during starvation (Fig. 7).
Noteworthy is the fact that TPKI/GSK3␤, which was initially studied as a protein phosphatase activator (58), can phosphorylate a protein substrate and, via the activation of PP1 by phosphorylation of inhibitor-2, cause the dephosphorylation of the same substrate (45). Thus, inhibitions of TPKI/GSK3␤ and PP can constitute a loop of amplification (Fig. 7), leading to a rapid build up of tau phosphorylation.
Cdk5 is also a candidate kinase for tau hyperphosphorylation in AD brains, particularly through pre-phosphorylation of tau facilitating further phosphorylation by TPKI/GSK3␤ (23). Starvation induced the electrophoretic retardation of the p35 band, followed by a gradual decrease in intensity, and the inhibition of cdk5 activity. These changes were recapitulated in cultured neurons by OA treatment, and confirm previous reports indicating that inhibition of PP stimulates the phosphorylation-dependent degradation of cdk5 activator p35 by the ubiquitin-proteasome pathway (59), and results in inhibition of cdk5 activity (52). Thus, inhibition of PP during starvation probably enhanced p35 phosphorylation, inducing its degradation by the proteasome and the decrease of cdk5 activity. Starvation did not completely inhibit the kinase activity toward tau, and cdk5 probably contributed to tau hyperphosphorylation in the hippocampus. Similarly, JNK, CaMKII, MAPK, or other kinases also probably contributed to tau hyperphosphorylation (Fig. 7).
The stress-activated protein kinase JNK was activated during starvation. Starvation induces decreases in leptin and increases in corticosterone (17), two hormones contributing to the activation of neuroendocrine stress system, raising the possibility that stress might play a role in tau hyperphosphorylation. Cold water stress produces rapid and reversible tau hyperphosphorylation (14) accompanied by TPKI/GSK3␤ Ser 9 phosphorylation in the rodent brain. 2 Thus tau hyperphosphorylation with TPKI/GSK3␤ phosphorylation at Ser 9 may be an integral part of neural stress reactions. Presumably, the stress of injection triggered Ser 9 phosphorylation in the hippocampus of the fed mice (Fig. 3).
Thus, tau hyperphosphorylation in the starved mouse hippocampus is the result of a complex mechanism in which inhibition of PP probably plays a central role, inducing and "overriding" the inhibition of key tau kinases such as TPKI/GSK3␤ and cdk5 (Fig. 7). In future experiments, it will be important to determine the pathways leading to TPKI/GSK3␤ and PP2A inhibition.
Relevance of Our Findings to AD-Reduced brain glucose metabolism is well documented in AD and has been suggested to be the result of brain insulin receptor desensitization (60), but whether the hypometabolism is causative of AD or the result of neuronal hypofunction or death remains to be determined. As reduction in brain glucose might have induced the inhibition of PP2A observed during starvation, similarly, reduction of glucose metabolism might be responsible for the reported inhibition of PP in AD brain (31,61).
In AD brains, TPKI/GSK3␤ and cdk5 immunoreactivities have been found to be associated with hyperphosphorylated tau (24,(62)(63)(64). Increase in cdk5 activity (65), and accumulation of p25 (66) have also been reported, although this last result has been recently challenged (67). To date no study has succeeded in demonstrating that GSK3 activity is enhanced in AD brains, which led to the hypothesis that in a context of reduced PP activity, normal levels of GSK3 activity might be sufficient for the hyperphosphorylation of tau (68). At seeming variance with AD, TPKI/GSK3␤ and cdk5 are inhibited in the starved mouse, whereas inhibition of PP2A occurs both during food deprivation and in AD. The seven phospho-tau epitopes studied here are all hyperphosphorylated in PHF-tau (69), while during starvation tau was hyperphosphorylated at Tau-1, AT8, PT231, and PS262, but not or to a limited extent at PS199, PS396, and PS404 epitopes. The latter are the sites less readily dephosphorylated by brain PP (Fig. 5B), and are phosphorylation sites for TPKI/GSK3␤ or cdk5 (23). These comparisons suggest that in AD brain, the inhibition of PP2A would chiefly account for tau hyperphosphorylation at Tau-1, AT8, PT231, and PS262, and that relatively unimpaired activities of TPKI/GSK3␤ and cdk5 may account, in combination with PP2A inhibition, for tau hyperphosphorylation at PS199, PS396, and PS404.
In vitro studies showed that A␤ activates TPKI/GSK3␤, enhances tau phosphorylation at Ser 199 , Ser 396 , and Ser 404 (70,71), and induces conversion of p35 to p25 (72). The presence of A␤ in AD brain might interfere with the PP-mediated inhibition of TPKI/GSK3␤ and cdk5, and thus lead to tau hyperphosphorylation at all PHF-tau sites. This hypothesis could be substantiated by starving transgenic mice displaying accumulation of A␤ and examining the pattern of tau phosphorylation.
Another explanation for the discrepancies in tau phosphorylation at the above sites between the starved mouse and AD may be possible considering that much of tau in AD brain occurs in insoluble fractions and PHF-tau phosphorylation sites were determined with respect to such tau preparations (69). Thus the AD brain contains an additional pool of tau, possibly aggregated, that is subject to a different phosphorylation-dephosphorylation dynamics, such as relative inaccessibility to PP. Substantial postmortem intervals inevitable for human brain samples also need to be taken into account.
Thus tau hyperphosphorylation in starved mice exhibits features that appear to be analogous to those of AD as well as distinct aspects. Our findings indicate that PP2A inhibition is likely to be responsible, in combination with the presence of other factors, for the formation of tau hyperphosphorylation at all the PHF-tau sites in AD brains. This simple and accessible system could be further exploited to reveal mechanisms of AD pathogenesis.
Concluding Remarks-For the first time, it is demonstrated that an AD-like tau hyperphosphorylation can occur in vivo despite inhibition of the two main tau kinases, with a residual lithium inhibitable activity contributing to hyperphosphorylation at certain sites. We conclude that inhibition of PP induces and overrides the inhibition of TPKI/GSK3␤ and cdk5, and leads to tau hyperphosphorylation in vivo, thus giving strong support to the idea that PP plays a primary role in regulation of tau phosphorylation in the adult brain. Furthermore, these data provide in vivo evidence in support of a central role of PP2A in tau hyperphosphorylation in AD, and advocate the idea that in the context of reduced PP activity in AD brain, even normal levels of GSK3 activity are sufficient for the hyperphosphorylation of tau at many PHF sites. Extending analyses to other forms of stress, and use of genetically modified mice will facilitate delineation of cumulative and chronic effects of stress, physiological variables including aging, and identification of molecules or conditions that may compromise the reversibility of tau phosphorylation and lead to PHF formation. Finally, this study suggests that in developing pharmaceuticals for tauopathies, those specifically targeted at a particular kinase may not be effective, and that those that affect the outcome of the network of protein kinases and phosphatases as a whole need to be sought.