Role of tau phosphorylation by glycogen synthase kinase-3β in the regulation of organelle transport

Tau is a major neuronal microtubule associated protein (MAP). One of its established functions is the promotion of assembly and maintenance of microtubule structure (Weingarten et al., 1975; Drubin et al., 1986). One of the most important post-translational modifications of tau is phosphorylation, because the degree of phosphorylation regulates its microtubule-binding and tubulin-polymerizing activities (Lindwall and Cole, 1984). Normal adult tau, which binds to microtubules and promotes microtubule assembly in vitro, contains at least two to three phosphate groups per molecule (Köpke et al., 1993). Glycogen synthase kinase-3 (GSK-3), a microtubule-binding protein, which is expressed abundantly in neurons of adult brain (Woodgett, 1990), is one of the kinases that may be involved in such physiological tau phosphorylation. For example, in most embryonic or neonatal neuronal cultures, endogenous tau is already phosphorylated by GSK-3 to some extent and can be downregulated by the GSK-3 inhibitors, lithium (Klein and Melton, 1996; Muñoz-Moñtano et al., 1995; Hong et al., 1997) and valproic acid (Chen et al., 1999). We have previously shown that in adult rat hippocampal progenitor cells (AHPs) GSK-3β phosphorylates adult tau isoforms mostly at the Tau-1 site (Ser 195/198/199/202) and that fibroblast growth factor-2 (FGF-2) upregulates the expression of both the substrate and the kinase dose dependently with increasing phosphorylation at the Tau-1 site but without decreasing tau's microtubule-binding ability (Tatebayashi et al., 1999) (Y.T. and I.G.-I., unpublished data). We have also observed that phosphorylation of tau at the Tau-1 and other sites during differentiation of SY5Y cells does not affect its microtubule-binding activity (Haque et al., 1999). However, the physiological role of the tau phosphorylation by GSK-3β has been elusive, but of major interest and importance because abnormal hyperphosphorylation of this protein is believed to cause neurodegeneration in Alzheimer disease (AD) and related tauopathies (Grundke-Iqbal et al., 1986a; Grundke-Iqbal et al., 1986b; Iqbal et al., 1986; Alonso et al., 1996).

In the present study, we show that in PC12 cells differentiated by nerve growth factor (NGF) or FGF-2, the expression of GSK-3β and tau and the phosphorylation of tau are upregulated. Inhibition of GSK-3β by lithium (LiCl), sodium valproate (VA) or alsterpaullone, three structurally independent GSK-3β inhibitors induced reversible inhibition of anterograde transport of mitochondria in association with dephosphorylation of tau but without any change in the ratio between microtubule-bound and -unbound tau. Similar association between anterograde transport and the phosphorylation state of tau was found in transfected CHO cells overexpressing tau. These data raise the intriguing possibility that the phosphorylation of tau by GSK-3β might be involved in the regulation of anterograde organelle transport.

The following phosphorylation-dependent site specific monoclonal antibodies to tau were employed: Tau-1 (to unphosphorylated Ser 195/198/199/202; numbers according to the longest human brain tau isoform tau₄₄₁; 1:50,000) (Binder et al., 1985; Grundke-Iqbal et al., 1986b; Szendrei et al., 1993); PHF-1 (to phosphorylated Ser 396/404; 1:1000) (Greenberg et al., 1992; Liu et al., 1993; Otvos et al., 1994); 12E8 (to phosphorylated Ser 262/356; 1:500) (Seubert et al., 1995); polyclonal phosphorylation-dependent antibodies to tau phosphorylated at T212, S396, or S404 (all 1:1000, BioSource International, Hopkinton, MA, USA). The phosphorylation-independent antibodies to recombinant human tau₄₁₀, R134d (1:5000) (Tatebayashi et al., 1999) were raised in a rabbit. Polyclonal antibodies to GSK-3 were 127d (to C-terminal residue 353-364 of rat GSK-3β; 1:500) (Pei et al., 1997) and GSK-3β Tyr 214/216 phosphospecific antibody to PTyr 214/216 (1:500, Biosource International, Hopkinton, MA, USA). Other primary antibodies employed were monoclonal antibody DM1A to α-tubulin (1:2000, Sigma, St. Louis, MO, USA) and monoclonal antibody 3G5 to MAP1B (1:1000) (Tucker et al., 1988). NGF and FGF-2 were purchased from Alamone Lab. (Israel) and Sigma, respectively. PD 98059, butyrolactone I, LiCl, valproic acid (VA), alsterpaullone and myo-inositol were purchased from Calbiochem (La Jolla, CA), Biomol (Plymouth Meeting, PA), Alexis Biochemicals (San Diego, CA) and Sigma, respectively.

PC12 cells were grown in 10 cm culture dishes (Corning, NY) which had been precoated with 0.01% poly-D-lysine (Sigma, St Louis, MO) and maintained in RPMI medium 1640 with L-glutamine (Invitrogen, Carlsbad, CA) supplemented with 5% horse serum and 10% fetal bovine serum. For differentiation, cells were transferred to BIOCOAT® poly-D-lysine 60 mm culture dishes or slides (Becton Dickinson, Bedford, MA) and grown overnight in regular medium. Cells were differentiated either with 50 ng/ml NGF or 20 ng/ml FGF-2 in medium containing 2% horse serum (differentiation medium). The medium was changed every 2 days.

CHO cells were maintained in RPMI medium 1640 with L-glutamine, supplemented with 10% fetal bovine serum, 100 μg/ml penicillin G and 100 μg/ml streptomycin. Cells were plated on Lab Tek™ slides (Nunc, Naperville, IL, USA) coated with poly-D-lysine 24-48 hours prior to use. For the experiments, either cell lines stably expressing tau (Haque et al., 1995) or cells transiently transfected with tau were employed. For transient transfection, 5-10 μg of pcDNA3-T₁₀ (tau₄₄₁ cDNA) generated as described previously (Haque et al., 1995), was complexed with 20 μg of Lipofectamine (Invitrogen) and overlaid on cells containing medium without serum and antibiotics. After 15 hours the cells were replenished with fresh, regular culture medium. Studies were carried out 24-48 hours post-transfection.

Neural progenitor cells from adult (3-month old) rat hippocampus (AHPs) were isolated and cultured as described previously (Tatebayashi et al., 1999) with some modifications. Briefly, cells isolated from tissue after enzymatic dissociation were further separated using a 4-ml step density gradient [Optiprep, at 7, 9.4, 11.7 and 16.4%, 1 ml each, in Hibernate A/B27 (Invitrogen), centrifuged at 800 g at room temperature for 15 minutes]. The resulting lowest 3 ml supernatant, including a dense band of cells and the pellet, were collected, resuspended in Neurobasal A (12.5 mM NaCl + Neurobasal, Life Technologies) containing 2% B27 supplement and 0.5 mM glutamine and plated on 60 mm culture dishes that had been precoated with 0.01% poly-D-lysine (135 kDa, Sigma). After 30 minutes, cells were washed and further cultured with Neurobasal A/B27 containing 0.5 mM glutamine, 100 IU/ml penicillin, 100 μg/ml streptomycin and 10 ng/ml FGF-2 (Life Technologies). Medium was changed every 2-3 days. Proliferating cultures, maintained for about 6 months through more than 15 passages, were used for this study. Tau expression and phosphorylation and its upregulation by FGF-2 (Tatebayashi et al., 1999) were monitored throughout this period.

To study the effect of GSK-3 and cyclin-dependent kinase 5 (cdk5) on different phosphorylation sites of tau, PC12 cells differentiated with FGF-2 for 7 days were incubated for 3 hours with medium alone or with 20 mM LiCl (GSK-3 inhibitor) or 10 μM Bu1 (cdk5 inhibitor). The phosphorylation of tau was determined by ¹²⁵I western blots developed with phosphodependent and site-specific tau antibodies.

For harvesting, the culture medium was first replaced by cold (4°C) PBS and the cells were then detached from the plate with a cell-scraper and centrifuged at 800 g at 4°C for 5 minutes. The cell pellets were lysed in 0.4% sodium dodecyl sulfate (SDS) and 0.4% β-mercaptoethanol, probe sonicated and boiled for 5 minutes. The resulting cell lysates were aliquoted and stored at –85°C until used. The protein concentrations were determined by the modified Lowry method (Bensadoun and Weinstein, 1986). Indicated amounts of protein samples were electrophoresed at least in triplicate on 10% SDS-polyacrylamide gels, transferred to Immobilon membrane (Millipore, Bedford, MA) and probed with primary antibodies. To assay the degree of phosphorylation at the Tau-1 site, the membrane was untreated or treated with alkaline phosphatase (196 U/ml) at 37°C for 8 hours in dephosphorylation buffer (50 mM Tris, pH 8.2, 2 mM MgCl₂, 1 mM PMSF, 5 μg/ml leupeptin and aprotinin, 2 μg/ml pepstatin A, 50 μg/ml phosphoramidon) prior to application of Tau-1 antibody (Tatebayashi et al., 1999). This procedure fully dephosphorylates even the hyperphosphorylated tau from Alzheimer brain (Khatoon et al., 1992). Bound antibodies were probed with ¹²⁵I-conjugated anti-mouse or anti-rabbit IgG (0.1 μg/ml; Amersham, Arlington Heights, IL, USA). The radio-immunoblots were scanned with a Fuji BAS 1500 Bio Image analyzer (Raytest USA Inc., Wilmington, DE). Images were processed using the Tina software and the strength of immunostaining was expressed as pixels per square length (PSL). Phosphorylation at the Tau-1 epitope (% phosphorylation) was calculated by the following formula: (PSL of staining by Tau-1 in alkaline phosphatase treated blot, i.e. total tau minus PSL of staining by Tau-1 in the untreated blot/PSL of staining by Tau-1 in alkaline phosphatase treated blot) × 100.

GSK-3β activity was assayed according to the method of Tanaka et al. (Tanaka et al., 1998) with some modifications. Briefly, cells were lysed on ice for 30 minutes in lysis buffer containing 50 mM Tris (pH 7.4), 0.1% Triton X-100, 20 mM NaCl, 1 mM PMSF, 5 μg/ml leupeptin, 5 μg/ml aprotinin, 2 μg/ml pepstatin A, 50 μg/ml phosphoramidon and centrifuged at 200,000 g for 30 minutes. The resulting cell extract protein (50 μg) was diluted to 250 μl with lysis buffer, followed by addition of 250 μl of immunoprecipitation buffer (Tris-buffered saline (TBS) with 300 mM NaCl and 1 mM PMSF) and then half of this solution was mixed with 2 μl of 127d antibody to GSK-3β and the other half served as the control. After incubation at 4°C overnight, 20 μl of immobilized protein G (Pierce, Rockford, IL) was added to the above mixture, incubated at 4°C for 2 hours and then centrifuged. The pellets were washed three times, suspended in 50 μl of 50 mM Tris, pH 7.4, 10 mM MgCl₂, 1 mM PMSF, and employed for assaying GSK-3 activity towards its specific substrate phosphoglycogen synthase peptide 2 (Upstate Biotechnology, Lake Placid, NY, USA) for 30 minutes at 30°C as described previously (Tatebayashi et al., 1999). The immunoprecipitation reaction in which the antibody to GSK-3β was omitted was used as a negative/background control.

The effect of GSK-3 inhibitors on the GSK-3 enzyme activity was measured in undifferentiated PC-12 cells. Cells were plated on poly-D-lysine-coated dishes and after 5 days in culture treated with the GSK-3 inhibitors LiCl (20 mM) or alsterpaullone (20 μM) for 1.5 hours. As controls the cells were treated with medium alone or with 10 μM Bu1, an inhibitor of cyclin-dependent kinases (cdk), including cdk5. Cells were then lysed and the GSK-3 activity determined as above. Three independent experiments were carried out with each inhibitor and GSK-3 activity assayed in duplicate.

Cells were fixed with 4% paraformaldehyde in PBS for 10 minutes, permeabilized with 0.5% Triton X-100 in PBS for 5 minutes, blocked in 5% bovine serum albumin in TBS (BSA/TBS) for 10 minutes and then incubated with primary antibodies in BSA/TBS at 4°C overnight. With this fixative the intensity of tau staining is optimal. However, tau binding to microtubules is eliminated and cannot be observed. Bound antibodies were probed with a combination of secondary antibodies (Molecular Probes, Eugene, OR): Oregon Green™ 488-conjugated goat anti-mouse antibody (1:1000) with or without Texas Red®-conjugated goat anti-rabbit antibody (1:2000), or Oregon Green™ 488-conjugated goat anti-rabbit antibody (1:1000) together with biotin-XX goat anti-mouse antibody (1:250) followed by avidin NeutrAvidin™-Cascade Blue® conjugate (1:1000). Images were captured either with a Nikon PCM 2000 epifluorescent microscope or a Zeiss Axiophot fluorescent microscope. Both were equipped with a SPOT-RT digital camera (Diagnostic Instruments, Sterling Heights, MI, USA).

For observation of mitochondria in live PC12 cells the fluorescent dyes Rhodamine 123 or Mito Tracker Red CMXRos (both from Molecular Probes) were employed. Since Mito Tracker Red is stable after fixation with paraformaldehyde this dye was used when the mitochondrial staining was followed by immunocytochemistry. Both dyes were stored as 1 mg/ml DMSO stock solutions at 4°C. To study the effect of GSK-3 inhibitors on the mitochondrial movement, PC12 cells differentiated by NGF or FGF-2, or tau-transfected CHO cells were treated with the GSK-3 inhibitors LiCl (20 mM), VA (0.6 mM) and alsterpaullone (20 μM). As controls, cells were treated with medium alone or with Bu1 (10 μM), an inhibitor of cyclin-dependent protein kinases for 1.5 to 3 hours. Cells were then further incubated in addition to the inhibitors with Mito Tracker Red (1 μg/ml) for 30 minutes, washed in PBS three times, fixed and immunostained. In PC12 cells, both Rhodamine 123 and Mito Tracker Red usually stained cell bodies more intensely than the cell periphery and processes, probably because of their small (10-20 μm) and round cell bodies. To study the reversal of the lithium effect on mitochondria, PC12 cells differentiated with NGF or FGF-2 were treated with 20 mM LiCl for 30 minutes, or left untreated (control), and then Rhodamine 123 was added in the medium to a final concentration of 10 μg/ml. After a further 30 minutes incubation at 37°C, the cells were washed gently three times in differentiation medium (with or without lithium). Images were captured with either a Nikon PCM 2000 epifluorescence microscope or a Zeiss Axiophot fluorescence microscope using a SPOT-RT digital camera (Diagnostic Instruments, Sterling Heights, MI).

For reversal of the LiCl effect, cells that had been incubated with LiCl were washed three times in differentiation medium without LiCl, followed by incubation for 1 hour at 37°C. Accumulated dye was observed with a 20× objective of a Nikon inverted fluorescence microscope and the same field observed before and after the washout period.

PC12 cells differentiated with NGF for 4 days or FGF-2 for 7 days were treated with or without 20 mM LiCl for 3 hours, and tau binding to microtubules determined as described by Hong et al. (Hong et al., 1997). Briefly, cells were washed once with PBS at 37°C, and lysed at 37°C in RAB buffer (0.1 M 4-morpholineethanesulfonic acid, 0.5 mM MgSO₄, 1 mM EGTA and 2 mM dithiothreitol, pH 6.8) supplemented with 0.1% Triton X-100, 20 μM taxol, 2 mM GTP, 5 μg/ml leupeptin, 5 μg/ml aprotinin, 2 μg/ml pepstatin A and 50 μg/ml phosphoramidon. Cell lysates were homogenized with 20 strokes in a Dounce homogenizer and centrifuged for 20 minutes at 50,000 g, 25°C. The supernatants from treated and untreated cells were adjusted to the same protein concentrations as determined by the Bradford method (Pierce, Rockford, IL). The pellets containing the cytoskeleton were resuspended in ice-cold RAB buffer to the same volume as the corresponding supernatants. Equal volumes of pellets and supernatants were then subjected to 10% SDS-PAGE and analyzed by ¹²⁵I western blots.

NGF and FGF-2 are known to induce neuronal differentiation of PC12 cells probably by sustained activation of the mitogen-activated protein kinase (MAPK) pathway (Cowley et al., 1994; Young et al., 1994; Maher, 1999). We found that PC12 cells differentiated with NGF had a larger proportion of thicker neurites than those differentiated with FGF-2 (Fig. 1A). During differentiation with NGF both tau and MAP1B levels increased gradually and were, at 10 days, about five- and nine-times higher, respectively, than in the untreated cells. In contrast, upon differentiation with FGF-2, only tau expression was increased in a time-dependent manner, whereas MAP1B levels remained low (Fig. 1B,C). Neither MAP1A nor MAP2 were detected in FGF-2 or NGF-differentiated cells (data not shown). Thus, of the four major MAPs, only tau seems to be upregulated upon differentiation of PC12 cells.

Western blots of the PC12 cells differentiated with NGF or FGF-2 and developed with monoclonal antibody Tau-1, which recognizes unphosphorylated tau, showed maximal staining when the blots were treated with alkaline phosphatase (Fig. 2A, Tau-1 dp and Tau-1). These findings indicate that during differentiation tau is phosphorylated at the Tau-1 epitope. The Tau-1 epitope is known to be a preferred site for GSK-3β. Like tau, the expression of GSK-3β was also found to be increased in NGF- and FGF-2-differentiated PC12 cells (Fig. 2A, 127d). The phosphorylation of GSK-3β at tyrosine 216, which activates this kinase, increased in parallel with the total level of the enzyme in both NGF- and FGF2-differentiated PC12 cells (Fig. 2A, Y216).

The activity of GSK-3β in NGF- and FGF-2-differentiated PC12 cells was assayed by immunoprecipitating the enzyme from the cell extracts and using phosphoglycogen synthase peptide 2 as a substrate. Both growth factors decreased GSK-3β activity to about 60–80% of that of the control during the first 20 minutes after the stimulation (Fig. 2B). This decrease in GSK-3β activity was probably due to transient activation of the phosphoinositide 3-kinase pathway (Raffioni and Bradshaw, 1992; Kleijn et al., 1998). The GSK-3β activity in the treated cells remained low for about 2 days and then increased almost simultaneously with the increase in the expression of tau and GSK-3β. In NGF-differentiated PC12 cells, the activity peaked at day 4 of the differentiation, and decreased at days 7 and 12. Since the phosphorylation of GSK-3β at Tyr 216 (Fig. 2A) was not decreased in total cell lysates at day 7, one of the reasons for the decrease of immunoprecipitated GSK-3β activity might be a change in distribution of the enzyme from the cytosol to the membrane pool. In contrast, in FGF-2-differentiated PC12 cells, the increased activity was sustained as long as 10 days (Fig. 2B). The increased GSK-3β activity was associated with increased phosphorylation of tau at the Tau-1 site (Fig. 2A,C). The degree of phosphorylation at the Tau-1 site was about 35% in undifferentiated PC12 cells at the start and increased to 65-70% at day 4 of differentiation (Fig. 2C). This increased phosphorylation was followed by a slight decrease in NGF-differentiated cells and by sustained hyperphosphorylation in FGF-2-differentiated PC12 cells.

Besides GSK-3β, cdk-5 and mitogen-activated protein kinase (MAPK) have been shown in vitro to phosphorylate tau at the Tau-1 site. To identify which one(s) of these kinases might phosphorylate tau in vivo at this site, NGF- and FGF-2-differentiated PC12 cells were incubated with different specific kinase inhibitors. Western blot analysis demonstrated that the phosphorylation of tau at the Tau-1 site was inhibited by the GSK-3β inhibitors LiCl and VA, but not by the cdk inhibitor Bu1, or by the MAPK kinase inhibitor, PD98059, indicating that GSK-3β is most probably the major kinase that phosphorylates tau at the Tau-1 site in differentiated PC12 cells (NGF; Fig. 2D, FGF-2; data not shown).

To confirm the involvement of GSK-3 activity in the phosphorylation of tau at the Tau-1 site, we determined in PC12 cells the activity of this enzyme in cells treated with LiCl, a GSK-3 inhibitor, alsterpaullone, a GSK-3/cdk5 inhibitor (Leost et al., 2000) and as a control Bu1, a cdk5 inhibitor. Since a reversal of LiCl-mediated inhibition of tau phosphorylation in cultured differentiated neurons has been observed previously after withdrawal of the inhibitor (Hong et al., 1997; Takahashi et al., 1999) (this study), care was taken to minimize any potential reversal of the inhibition by performing all washes, lysing of the cells and immunoprecipitations at 4°C. Only the phosphorylation of the substrates was carried out at 30°C for 30 minutes. Despite these precautions the degree of inhibition of GSK-3 in situ might be higher than measured here in immunoprecipitates. Treatment of the cells for 1.5 hours with LiCl or alsterpaullone, but not with butyrolactone, decreased the GSK-3 activity significantly (Fig. 2E). These findings suggested that the phosphorylation of tau at the Tau-1 site observed in the PC12 cells was most likely catalyzed by GSK-3.

In vitro tau can be phosphorylated by the proline-dependent kinases GSK-3 and cdk5 at a number of sites. To elucidate which of these sites are phosphorylated by these two kinases in differentiated PC12 cells, we determined the degree of inhibition of phosphorylation of tau at specific sites in cells treated with LiCl and Bu1. PC12 cells that had been differentiated for 7 days with FGF-2 and then incubated for 2 hours with the inhibitors were analyzed by ¹²⁵I western blots using site-specific phosphodependent tau antibodies. We found that inhibition of GSK-3 led to a decrease of phosphorylation in addition to the proline-dependent Tau-1 epitope (Fig. 2D), at T212, S396 and S404 sites (Fig. 3). Like LiCl, Bu1 also inhibited the phosphorylation of tau at S404 (Fig. 3). These results indicate that in differentiated PC12 cells the Tau-1 epitope (see Fig. 2D) as well as the T212 and S396 sites are phosphorylated preferentially by GSK-3, whereas at S404 tau is phosphorylated both by cdk5 and GSK-3. The proline-independent site 12E8 was not affected by either GSK-3 or cdk5.

The effect of phosphorylation on the binding of tau to microtubules was studied in differentiated PC12 cells incubated without or with 20 mM LiCl for 3 hours. Although in both NGF- (Fig. 3C) and FGF-2- (data not shown) differentiated PC12 cells LiCl decreased the phosphorylation at the Tau-1 epitope by ∼70%, tau bound to endogenous taxol-stabilized microtubules with the same efficacy regardless of its phosphorylation state (Fig. 3C).

Since tau has been shown to affect organelle transport, especially the kinesin-mediated anterograde transport (Ebneth et al., 1998) and its levels during differentiation were increased dramatically (Fig. 2A), we studied the distribution of mitochondria in the differentiated cells. Both in the control (Fig. 4Aa-c,m-o) and the Bu1-treated (Fig. 4Ad-f) differentiated cells we found an almost complete blockage of the Tau-1 epitope associated with an apparently normal distribution of mitochondria in cell bodies and processes. In contrast, when GSK-3β was inhibited with either LiCl, VA (Fig. 4Ag-i,j-l), or alsterpaullone (not shown in the figure) the Tau-1 epitope was dephosphorylated. Alsterpaullone inhibits GSK-3 and, to a lesser degree, cyclin-dependent kinases/cdk5 but neither kinases of the ERK family nor protein kinase C (Leost et al., 2000). ERKs and PKC have been reported to be affected by LiCl and VA (Manji and Chen, 2002). The reduced phosphorylation of tau was associated with a reduction of mitochondria in the neurites and clustering around the nucleus. Higher magnification phase-contrast images revealed that the inhibition of GSK-3 not only affected the transport of mitochondria but also of most cellular organelles (Fig. 4Al,o). While the effect of GSK-3 inhibition was observed in PC12 cells that had been differentiated with FGF-2 at all stages of differentiation up to the 10 days studied, in the case of NGF the mitochondrial clustering was only observed in cells differentiated for up to 4 days. In cells differentiated for 7 days or longer, concomitant with the increased expression of MAP1B (see Fig. 1B), no mitochondrial clustering was observed (data not shown).

It has been shown previously that the LiCl-induced inhibition of tau phosphorylation is reversible in situ upon removal of LiCl (Hong et al., 1997; Takahashi et al., 1999). Similarly, in the present study, the effect of LiCl on the mitochondrial aggregation was reversible, because the deletion of lithium from the medium reversed the distribution of the Rhodamine 123-positive mitochondria into the neurites and the periphery of the cells (Fig. 4B). These data raise the possibility that GSK-3β might regulate the anterograde organelle transport including that of mitochondria in differentiated PC12 cells concomitant with the alteration of the phosphorylation state of tau on microtubules.

To confirm the possible role of the phosphorylation of tau in mitochondrial transport, we studied CHO cells transfected with the longest human brain tau isoform, tau₄₄₁, with respect to tau phosphorylation and mitochondrial distribution (Fig. 5). Immunocytochemical staining with phospho-independent polyclonal antibody R134d to total tau (green in Fig. 5) and the phospho-dependent antibody Tau-1 to unphosphorylated tau (blue) was employed for this study. In CHO cells stably or transiently transfected with tau we observed cells in which tau was phosphorylated (Fig. 5a,d, arrowheads) and others in which it was not (Fig. 5a,d, arrows). In most cells with phosphorylated tau (Tau-1 negative, arrowheads) the distribution of mitochondria (red) was similar to that in untransfected cells (asterisk), while in the cells with unphosphorylated tau (Tau-1 positive, arrows), the mitochondria were clustered around the nucleus. In the latter cells we also observed, by phase contrast microscopy, loss of organelles from the periphery (Fig. 5g,h, arrow); a similar phenomenon was also observed in PC12 cells treated with GSK-3 inhibitors (Fig. 4Al). These data, showing the association between the phosphorylation state of tau and change in distribution of mitochondria, are consistent with the possibility that the phosphorylation of tau might be involved in the regulation of the mitochondrial transport in the cell.

The effect of LiCl on the initial neurite outgrowth and the distribution of mitochondria was examined in PC12 cells in the presence of NGF or FGF-2. LiCl (20 mM) or VA (0.6 mM) induced mitochondrial clustering (after 1 day in NGF or FGF-2) and almost completely inhibited initial neurite outgrowth (up to 6 days examined) of both NGF- and FGF-2-treated PC12 cells (data not shown). Another prominent known effect of LiCl and VA is their inhibition of inositol monophosphatase, which might result in inositol depletion (Klein and Melton, 1996; Hallcher and Sherman, 1980). However, excess myo-inositol administration failed to rescue the inhibition of neurite outgrowth by LiCl (Fig. 6A) or VA (data not shown), suggesting that inositol depletion might not be involved in this phenomenon.

To learn whether the regulation of the neurite outgrowth through mitochondrial transport by the tau-GSK-3β phosphorylation system is similar in immature cells, we carried out the above studies on neural progenitor cells derived from adult rat hippocampus (AHPs). In these cells, FGF-2 is known to upregulate the expression of adult isoforms of tau and GSK-3β-mediated phosphorylation at the Tau-1 site (Tatebayashi et al., 1999). We found that AHPs do not express MAP1B (data not shown). AHPs were plated with FGF-2 in the presence or absence of LiCl for 1 day. Lithium caused mitochondrial accumulation (data not shown) and blocked initial neurite outgrowth of AHPs almost completely (Fig. 6Ba-c). These data are consistent with the possibility that the tau-GSK-3β phosphorylation system might be involved in the neurite generation of PC12 cells and of neural progenitor cells derived from the adult central nervous system.

Tau is a major MAP of neurons. To date the most established function of tau is its promotion of assembly and maintenance of microtubules (Weingarten et al., 1975; Drubin and Kirschner, 1986). That tau might also play a role in the regulation of organelle transport was shown by Ebneth et al. (Ebneth et al., 1998) in CHO and differentiated Neuro 2A cells where upon transfection with tau, the anterograde kinesin-mediated mitochondrial/organelle movement was disrupted. The present study shows that the phosphorylation of tau by GSK-3β during differentiation might be involved in the regulation of the intracellular anterograde mitochondrial/organelle transport in small caliber neurites where tau represents a large proportion of total MAPs.

In the present study in PC12 cells differentiated by NGF or FGF2, the levels of tau and GSK-3β were upregulated simultaneously and the GSK-3β-mediated tau phosphorylation was increased at several sites canonic for proline-dependent kinases including the Tau-1 site. Previously most of these sites had been shown in vitro to be phosphorylated by ERK 1/2, cdc2, GSK-3 and cdk5 (for a review, see Johnson and Hartigan, 1998). In situ, in differentiated neurons, GSK-3 clearly seems to have a role in tau phosphorylation as was shown by inhibition of GSK-3 with LiCl (Hong et al., 1997; Muñoz-Montaño et al., 1997). Of the cyclin-dependent kinases in differentiated neurons, cdc2, which is activated in dividing cells (Meyerson et al., 1992) is most probably of less importance than cdk5, the activity of which is increased in differentiating neurons (Fu et al., 2002). However, relatively little is known about tau phosphorylation by cdk5. That this kinase might play a role in tau phosphorylation in situ is indicated by the finding that transfection of primary neurons with its activator p25 induced phosphorylation of tau at the AT8 (PS202/T205) epitope (Patrick et al., 1999). In the present study, in differentiated PC12 cells, treatment with LiCl and Bu1 revealed that the Tau-1 epitope, T212 and S396 were almost exclusively phosphorylated by GSK-3, whereas both GSK-3 and cdk5 phosphorylated tau at S404. As expected, the proline-independent 12E8 site, which is phosphorylated by PKA, MARK and CaMKII (Drewes et al., 1997; Sironi et al., 1998) was not affected by inhibition of either GSK-3 or cdk5.

The inhibition of GSK-3 by LiCl for 3 hours and the concomitant decrease in phosphorylation of tau at various sites in PC12 cells did not significantly affect the binding of tau to microtubules. This finding is consistent with a previous study in which LiCl treatment of differentiated SY5Y cells (for 24 hours) did not change microtubule binding (Xie et al., 1998). In contrast, in differentiated NT2 neurons treated overnight with LiCl, the binding of tau to microtubules and the degree of polymerization of tubulin was found to be enhanced (Hong et al., 1997). The exact cause of this discrepancy between different neuronal cell lines is, at present, not understood.

The present study indicated that in differentiated PC12 cells the GSK-3-mediated increase in phosphorylation of tau at defined sites might be required for organelle/mitochondrial transport. When the GSK-3β activity was inhibited, tau was dephosphorylated at both the proline-rich N-terminal and the C-terminal domain and the anterograde organelle transport was inhibited. In contrast, inhibition of the cyclin-dependent kinases/cdk5, which also phosphorylated tau at Ser-404, but not at the GSK-3-preferred sites studied, resulted in flattening of the cells but no apparent disturbance of the mitochondrial transport. Similarly, in tau-transfected CHO cells, the mitochondrial/organelle transport was disrupted in those cells in which tau was not phosphorylated, but was largely intact in cells containing phosphorylated tau. Also it has been shown previously that phosphorylated tau can affect the mitochondrial transport. In this case conditions favoring the phosphorylation of tau at sites inhibitory to its microtubule binding, such as phosphorylation with A-kinase led to enhanced and less staggered transport of organelles in the axons of primary sensory neurons (Sato-Harada et al., 1996). The present study shows that in differentiated PC12 cells, endogenous tau, when physiologically phosphorylated by GSK-3, is not detached from the microtubules but seems to support kinesin-mediated mitochondrial anterograde transport. Inhibition of GSK-3 resulted in dephosphorylation of tau and inhibition of the anterograde transport, possibly because of phosphorylation-dependent conformational changes of tau or its effect on the dynamics of microtubule assembly. It has been shown previously that the microtubule nucleation activity of tau is reduced by dephosphorylation (Morita-Fujimura et al., 1996). The present study supports the notion that a balance between the levels of GSK-3 and tau phosphorylation is required to maintain the anterograde transport. As was previously shown, overexpression of tau in cultured cells led to disturbance of the axonal transport (Ebneth et al., 1998; Trinczek et al., 1999), and to axonopathy and motor deficiencies in tau transgenic mice (Spittaels et al., 1999). The deficient anterograde transport together with the axonopathy was restored to almost normal when GSK-3 levels were increased by crossing these mice with GSK-3 transgenic animals (Spittaels et al., 2000; Nuydens et al., 2002). In addition to tau, GSK-3 has also been shown, in vitro, to phosphorylate kinesin light chains, resulting in a release of cargo organelles and thus disturbance of the axonal transport (Morfini et al., 2001). It is not known whether tau and kinesin may be connected functionally and what is the effect of GSK-3 inhibition on kinesin.

The role of phosphorylated tau in organelle transport was observed when tau was the major MAP. At 4 days of NGF-but not FGF-2-induced differentiation, coincident with the expression of MAP1B, this role of tau was eliminated. A similar functional relationship between tau and MAP1A has been speculated in the normally tau-rich small-caliber axons of adult tau-knockout mice (Harada et al., 1994). In these structures decreased microtubule stability and altered microtubule organization was observed. However, it is not known whether the expression of MAP1B in NGF-differentiated PC12 cells observed in the present study causally or incidentally affected the Li-mediated disturbance of the axonal transport.

That GSK-3 might play a role in regulating the neurite outgrowth has been shown previously (Burstein et al., 1985) and is also indicated by the findings in the present study that exposure to lithium for 3 days and 1 day, respectively, inhibited almost completely the initial neurite outgrowth of differentiating PC12 cells (Fig. 6A) and AHPs (Fig. 6B). This is in agreement with the data of Goold et al. (Goold et al., 2001) who observed inhibition of neurite formation and MAP1B expression during prolonged lithium treatment. Exposure to lithium for several days has also been reported to result in the formation of shorter than normal axons in cultures of hippocampal and cerebellar neurons (Lucas et al., 1998; Takahashi et al., 1999). However, it should be kept in mind that besides tau, GSK-3 phosphorylates a number of other proteins such as MAP1B (Goold et al., 2001), and several transcription factors (Welsh et al., 1996). Especially upon chronic administration of lithium, besides GSK-3, other factors and signaling cascades such as the PKC and MAPK cascades (Manji and Chen, 2002) are affected that might also modulate the regulation of neurite outgrowth and retraction.

Expression of MAP1B is upregulated during embryogenesis where it peaks at 2-3 days and decreases within 2-3 weeks after birth (Riederer et al., 1986; Schoenfeld et al., 1989). Thus, the effect of the phosphorylation of tau by GSK-3 on neuritic extension might be only minor during development, and major during adulthood, especially in axons. This role of GSK-3 might be one of the critical functions lost in adult tau-knockout mice, which were reported to exhibit some neurological defect (Ikegami et al., 2000), and in the human, in certain cases with frontotemporal dementia where tau protein is not expressed (Zhukareva et al., 2001).

Activated GSK-3β has been shown to be localized to paired helical filaments/neurofibrillary tangles in AD brain (Yamaguchi et al., 1996; Pei et al., 1997; Pei et al., 1999) and is believed to be a major protein kinase involved in the abnormal hyperphosphorylation of tau (Mandelkow et al., 1992; Ishiguro et al., 1993; Lovestone et al., 1994; Lucas et al., 2001). Moreover, FGF-2 levels are elevated in AD brains (Stopa et al., 1990) and strong immunoreactivities of FGF-2 also have been found in the tau lesions (Cummings et al., 1993). The FGF-2-induced upregulation of both the expression of tau and GSK-3β and the GSK-3β-mediated tau phosphorylation in PC12 cells in the present study and in AHPs shown previously (Tatebayashi et al., 1999) raise the possibility that this growth factor might be involved in the process of neurofibrillary degeneration in an environment favorable to phosphorylation, such as the AD brain in which the phosphatase activity is compromised (Gong et al., 1993).

In conclusion, in differentiated PC12 cells, GSK-3β through phosphorylation of tau at a number of sites seems to regulate anterograde transport of mitochondria in tau-dominated small caliber neurites. Since in situ the GSK-3β activity is regulated by several kinases such as protein kinase B via the receptor-mediated phosphoinositide 3-kinase pathway (Proud and Denton, 1997), it is possible that the GSK-3β-tau phosphorylation pathway might play a role in the regulation of neurite outgrowth and the plasticity of tau-dominated thin neurites. Although the role of these neurites in the central nervous system is still unclear, their abnormal upregulation can be involved, via upregulation of hyperphosphorylated tau, in the pathophysiology of AD and other tauopathies.

We are grateful to Lester I. Binder (University of Alabama, Birmingham, AL, USA) for the monoclonal antibody Tau-1, Peter Davies (Albert Einstein College of Medicine, Bronx, NY, USA) for PHF-1, and Dale Schenk (Athena Neurosciences, San Francisco, CA, USA) for 12E8 antibody. We also acknowledge Janet Biegelson and Sonia Warren for secretarial assistance. These studies were supported in part by funds from the New York State Office of Mental Retardation and Developmental Disabilities and NIH grants NS18105, AG05892 and AG19158.