Lithium Reduces Tau Phosphorylation by Inhibition of Glycogen Synthase Kinase-3*

Lithium is one of the most widely used drugs for treating bipolar (manic-depressive) disorder. Despite its efficacy, the molecular mechanism underlying its action has not been elucidated. One recent study has proposed that lithium inhibits glycogen synthase kinase-3 and thereby affects multiple cellular functions. Because glycogen synthase kinase-3 regulates the phosphorylation of tau (microtubule-binding protein that forms paired helical filaments in neurons of the Alzheimer’s disease brain), we hypothesized that lithium could affect tau phosphorylation by inhibiting glycogen synthase kinase-3. Using cultured human NT2N neurons, we demonstrate that lithium reduces the phosphorylation of tau, enhances the binding of tau to microtubules, and promotes microtubule assembly through direct and reversible inhibition of glycogen synthase kinase-3. These results provide new insights into how lithium mediates its effects in the central nervous system, and these findings could be exploited to develop a novel intervention for Alzheimer’s disease.

Paired helical filaments are the major building blocks of neurofibrillary lesions in Alzheimer's disease brains. Paired helical filaments are composed of hyperphosphorylated central nervous system tau protein (1)(2)(3). Predominantly expressed in axons, tau is a group of microtubule-associated proteins that normally promote and stabilize the assembly of microtubules (4 -6). Paired helical filament-tau differs from normal tau by its abnormal hyperphosphorylation, which results in the decreased tau binding to microtubules. It has been shown that paired helical filament-tau does not bind to microtubules unless it is enzymatically dephosphorylated (7,8). This decreased affinity of paired helical filament-tau for microtubules, coupled with the reduced amount of normal tau, may lead to the destabilization of microtubules and result in impairment of axonal transport, neuronal degeneration, and the aggregation of paired helical filaments. Therefore the hyperphosphorylation of tau is believed to be a key event in the pathogenesis of Alzheimer's disease.
The phosphorylation of tau is regulated by kinases and phosphatases. We have shown previously that glycogen synthase kinase-3 (GSK-3), 1 a serine/threonine protein kinase, plays an important role in the regulation of tau phosphorylation (9,10). When overexpressed in cultured human neurons, GSK-3 induces an Alzheimer's disease-like hyperphosphorylation of tau. When GSK-3 activity is down-regulated by insulin or insulinlike growth factor I, the phosphorylation state of tau is decreased (10).
GSK-3 was originally identified as a kinase that phosphorylates and inactivates glycogen synthase, which catalyzes a regulated step in insulin-mediated glycogen synthesis (11,12). Insulin induces a phosphorylation-dependent down-regulation of GSK-3, which leads to the activation of glycogen synthase and therefore increased glycogen synthesis.
GSK-3 was later found to have many other cellular functions. For example, the homologs of GSK-3 play a central role in the development of diverse organisms including Drosophila, Dictyostelium, and Xenopus (13)(14)(15)(16)(17)(18). During dorso-ventral axis formation in Xenopus embryogenesis, inhibition of GSK-3 can lead to increased dorsal tissue (17). Interestingly, lithium, a drug that has been widely used for the treatment of bipolar (manic-depressive) disorder, induces similar effect on Xenopus development, i.e. causing an expansion of the dorsal mesoderm leading to duplication of the dorsal axis (19,20). Indeed, a recent study showed that lithium directly inhibits GSK-3, suggesting that this may be a mechanism whereby lithium exerts its effect on cell fate determination as well as other cellular functions (21).
Additionally, lithium has been reported to activate glycogen synthase and stimulate glycogen synthesis in rat hepatocytes (22). This has also been proposed (21) to be a result of the inhibition of GSK-3 by lithium because the inhibition of GSK-3 abrogates the phosphorylation-dependent inactivation of glycogen synthase, which in turn increases glycogen synthesis.
Despite the remarkable efficacy of lithium in treating bipolar disorder, the molecular mechanisms underlying its therapeutic actions have not been fully elucidated. Previous studies have shown that lithium inhibits the phosphorylation of the mid sized neurofilament subunit and alters cytoskeletal organization in growing chick sensory neurons (23,24). These studies suggest that lithium may exert its effect on the central nervous system by affecting the neuronal cytoskeleton. As a neuronal microtubule-associated protein, tau is a substrate of GSK-3, and the phosphorylation of tau by GSK-3 could be directly inhibited by lithium and microtubule assembly thereby affected.
To test this hypothesis, we examined the effects of lithium on tau phosphorylation and microtubule assembly in cultured NT2N neurons, which are derived from a human embryonal carcinoma cell line (NTera2/D1 or NT2) following retinoic acid treatment (25,26). We demonstrate that lithium reduces the phosphorylation state of tau by a direct and reversible inhibition of GSK-3 and that this results in an increase in the binding of tau to microtubules, as well as an augmentation of microtubule assembly in NT2N neurons.

EXPERIMENTAL PROCEDURES
Materials-[␥-32 P]ATP and 125 I-labeled secondary antibodies were purchased from NEN Life Science Products and the Semliki Forest virus (SFV) gene expression system from Life Technologies, Inc. Taxol was obtained from Dr. V. Narayanan of NCI. Other chemicals were purchased from Sigma. GSK-3␤ S9A cDNA and the cDNA construct to make human recombinant tau were obtained from Dr. M. Goedert. GSK-3␤ WT cDNA was from Drs. X. He and H. Varmus.
Cell Culture-To generate NT2N neurons, a human embryonal carcinoma cell line (NTera2/D1, or NT2) was grown and maintained as described (26). Briefly, NT2 cells were treated with retinoic acid for 5 weeks and then replated at reduced density. After 10 days, neuron-like NT2N cells were mechanically and enzymatically separated from the parent NT2 cells and replated at a density of 1.2 ϫ 10 6 /well on six-well plates previously coated with poly-D-lysine (10 g/ml) and Matrigel. The NT2N neurons were maintained for up to 4 -6 weeks in Dulbecco's modified Eagle's medium with high glucose (DMEM-HG), supplemented with 5% fetal bovine serum, penicillin/streptomycin, and mitotic inhibitors (1 M cytosine arabinoside, 10 M fluorodeoxyuridine, and 10 M uridine). 2-4-week-old NT2N neurons were used for experiments, and each experiment was repeated at least three times.
Overexpression of GSK-3␤ in NT2N Neurons Using the SFV Gene Expression System-Two GSK-3␤/SFV viral constructs were used. GSK-3␤ WT is the wild type GSK-3␤; and GSK-3␤ S9A is a constitutively active form of GSK-3␤, in which the serine 9 residue was mutated to alanine. A 10-amino acid c-Myc tag (EQKLISEEDL) was introduced at the COOH termini of both constructs. Another SFV construct expressing ␤-galactosidase (LacZ) was used as a control. To overexpress GSK-3␤ in NT2N neurons, cells were washed once with DMEM-HG without supplements, and viral stocks diluted in DMEM-HG (multiplicity of infection ϭ 10) were applied to the cells. After 1 h, the medium containing viruses was replaced by DMEM-HG with 5% fetal bovine serum. The infected NT2N neurons were then incubated overnight (18 h) before they were harvested. For experiments on the effects of lithium or calyculin A in infected cells, cells were treated with 10 mM LiCl for 2 h or with 0 -50 nM calyculin A for 30 min after overnight infection.
Treatment of NT2N Neurons-To determine the effects of lithium on tau phosphorylation, NT2N neurons were treated with 0 -25 mM LiCl for 2 h before being harvested. To examine the phosphorylation sites in tau affected by GSK-3 and lithium, NT2N neurons were either infected with GSK-3␤/SFV overnight or treated with 10 mM LiCl for 2 h. To test whether calyculin A blocks the effects of lithium on tau phosphorylation, cells were treated with or without 10 mM LiCl for 30 min in the presence of 0 -50 nM calyculin A. To determine whether lithium reversibly decreases tau phosphorylation, NT2N neurons were treated with 10 mM LiCl for 2 h and subsequently washed three times with lithiumfree medium and incubated for another 0.5 or 2 h. As controls, cells were treated with lithium for 2, 2.5, and 4 h without wash.
Immunoblot Analysis of Tau Phosphorylation-NT2N neurons were washed once with phosphate-buffered saline and lysed in ice-cold high salt RAB buffer (0.1 M Mes, 0.5 mM MgSO 4 , 1 mM EGTA, 2 mM dithiothreitol, and 0.75 M NaCl, pH 6.8) supplemented with 0.1% Triton X-100 and a mixture of protease inhibitors. Cell lysates were incubated on ice for 10 min, sonicated, and centrifuged for 20 min at 50,000 ϫ g, 4°C. The supernatants were collected and protein concentrations determined using the bicinchoninic acid method (Pierce). An equal amount of total protein was resolved on 10% SDS-PAGE. Immunoblotting was performed with a panel of site and phosphorylation-sensitive anti-tau antibodies. The blots were developed by the enhanced chemiluminescence or 3,3Ј-diaminobenzidine method. Alternatively, for the quantitation of the relative levels of tau protein, 125 I-labeled goat anti-mouse IgG was used as secondary antibody, and the blots were exposed to PhosphorImager plates. 10 g of total protein was loaded for the detection of tau on immunoblots using the antibodies T14/46, 25 g for the antibodies T1, PHF1, T3P, and PHF13, and 50 g for the antibodies AT8, AT270, PHF6, and 12E8.
Immunoprecipitation and GSK-3 Kinase Assays-To determine the kinase activities of overexpressed GSK-3␤, infected NT2N neurons were lysed in ice-cold lysis buffer (20 mM Pipes, pH 7.0, 10 mM NaCl, 0.5% Nonidet P-40, 0.05% ␤-mercaptoethanol, 5 mM EGTA, 50 mM NaF, 1 mM Na 3 VO 4 , and 1 M microcystin) supplemented with a mixture of protease inhibitors. Overexpressed GSK-3␤ was immunoprecipitated with a rabbit polyclonal anti-c-Myc antiserum. The immunoprecipitants were washed twice with lysis buffer and twice with GSK-3 kinase buffer (20 mM Hepes, pH 7.2, 10 mM MgCl 2 , 10 mM MnCl 2 , 1 mM dithiothreitol, 0.2 mM EGTA, and 5 M ATP). GSK-3 kinase assay was performed at 30°C for 15 min in 20 l of GSK-3 kinase buffer supplemented with 4 g of recombinant tau and 5 Ci of [␥-32 P]ATP for each reaction. The reaction was terminated by adding 5 ϫ sample buffer and boiling for 5 min. The entire reaction was then resolved on 10% SDS-PAGE, transferred to nitrocellulose membrane, and exposed to a PhosphorImager plate. GSK-3 kinase activity was represented by the level of 32 P incorporation into tau. The amount of immunoprecipitated GSK-3 present in each reaction was determined subsequently by immunoblot analysis with a monoclonal anti-GSK-3 antibody on the same nitrocellulose membrane and used to normalize the corresponding relative GSK-3 activity. To determine the effect of lithium on GSK-3 activity, the kinase reactions were performed in the presence of 0 -10 mM LiCl. To determine the reversibility of the lithium effect, the immunoprecipitated GSK-3 was incubated in GSK-3 kinase buffer with 10 mM LiCl for 20 min, washed three times with lithium-free GSK-3 kinase buffer, and then used to perform kinase assay. As controls, GSK-3 kinase reactions were either performed in the presence of lithium without wash after 20 min of lithium incubation, or without lithium incubation or wash.

Determination of Tau Binding to Microtubules and Alterations of Microtubule
Assembly-NT2N neurons were treated with 10 mM LiCl overnight before they were washed once with 37°C phosphate-buffered saline and lysed in 37°C RAB buffer (0.1 M Mes, 0.5 mM MgSO 4 , 1 mM EGTA, and 2 mM dithiothreitol, pH 6.8) supplemented with 0.1% Triton X-100, 20 M Taxol, 2 mM GTP, and a mixture of protease inhibitors. Cell lysates were homogenized with 20 strokes in a Dounce homogenizer and centrifuged for 20 min at 50,000 ϫ g, 25°C. The protein concentrations of the soluble fractions were adjusted to be the same, and the cytoskeletal fractions were resuspended in ice-cold RAB buffer of the same volume as the corresponding soluble fractions. To enrich tau, an equal fraction of the samples was boiled and clarified by centrifugation. An equal volume was resolved on 10% SDS-PAGE for immunoblot analysis with T14/46 for tau. For determining the distribution of ␣-tubulin subunits, an equal volume of the unboiled samples was used to perform immunoblot analysis with a mouse monoclonal antibody (MAb) to ␣-tubulin, a rat MAb to Tyr-tubulin, and a rabbit polyclonal anti-Glu-tubulin antiserum. The blots were incubated with 125 I-labeled secondary antibodies and exposed to PhosphorImager plates. Quantitation was performed with the ImageQuant software (Molecular Dynamics). The ratio of microtubule-bound tau (in the cytoskeletal fraction) to soluble tau was determined by comparing the T14/46 immunoreactivities. The ratios of polymerized to soluble ␣tubulin, Tyr-tubulin, or Glu-tubulin were determined and plotted in a similar manner.

Lithium Reduces Tau Phosphorylation in NT2N
Neurons in a Dose-dependent Manner-To examine the effects of lithium on tau phosphorylation in neuronal cells, NT2N neurons, which endogenously express tau, were treated with 0, 0.5, 1, 5, 10, or 25 mM LiCl. After 2 h of treatment, cells were lysed and an equal amount of total protein was resolved on SDS-PAGE for immunoblot analysis with a panel of anti-tau antibodies. As shown in Fig. 1, with increasing concentrations of lithium, T1 immunoreactivity increased and PHF1 immunoreactivity decreased, indicating a reduction of tau phosphorylation. This was accompanied by an increase in electrophoretic mobility of tau, detected by T14/46. The quantitative data on T1 and PHF1 immunoreactivities were normalized to the levels of T14/46 immunoreactivity and are plotted in Fig. 1B. These results demonstrate that lithium decreases the phosphorylation of tau in a dose-dependent manner in the NT2N neurons.
Lithium Reduces Tau Phosphorylation by Direct Inhibition of GSK-3 but Not through Phosphorylation-dependent Down-regulation of GSK-3-To test whether the effect of lithium on tau phosphorylation is a result of the inhibition of GSK-3, we first demonstrated that lithium directly inhibits GSK-3 in NT2N neurons. We overexpressed WT and a constitutively active mutant (S9A) form of GSK-3␤ in NT2N neurons using the SFV gene expression system. Both GSK-3␤/SFV constructs have a c-Myc tag at their COOH termini, and the overexpressed protein can be immunoprecipitated by a polyclonal anti-c-Myc antiserum. In Fig. 2A, overexpressed GSK-3␤ S9A was immunoprecipitated, and a GSK-3 kinase assay was performed in the presence of 0 -10 mM LiCl. The GSK-3 activities were assayed by monitoring 32 P incorporation into human recombinant tau. The amount of GSK-3 present in each reaction was determined by immunoblotting with a MAb to GSK-3 and used to normalize the relative GSK-3 activities. As shown in Fig. 2A, lithium directly inhibited the activity of GSK-3␤ S9A with an IC 50 of 3 mM. Similar results were obtained when GSK-3␤ WT was overexpressed in NT2N neurons (data not shown). Other monovalent cations (Na ϩ , K ϩ , Cs ϩ , and NH 4 ϩ ) did not have a similar effect (data not shown).
We also examined the in vivo effects of lithium on tau phos-phorylation in NT2N neurons that overexpressed GSK-3␤ WT or GSK-3␤ S9A. As shown in Fig. 2B, overexpression of both GSK-3 constructs increased tau phosphorylation because T1 (dephosphorylated) immunoreactivity decreased and PHF1 (phosphorylated) immunoreactivity increased compared with control (LacZ). When treated with LiCl, the increased tau phosphorylation induced by overexpression of both GSK-3␤ WT and GSK-3␤ S9A was abrogated (WTϩLi and S9AϩLi), as the changes in T1 and PHF1 immunoreactivities were reversed (Fig. 2B).
These experiments indicate that lithium reduces tau phosphorylation by direct inhibition of GSK-3. The serine 9 residue in GSK-3␤ is required for the phosphorylation-dependent inactivation, and mutating this residue to alanine in GSK-3␤ S9A abolishes the down-regulation of its activity by insulin and other growth factors (10, 39 -41). Because the effects of lithium on GSK-3 and tau phosphorylation are not affected by this mutation, the phosphorylation of serine 9 must not be involved in the inhibition of GSK-3 by lithium.
Lithium and GSK-3 Have Opposite Effects on the Same Phosphorylation Sites of Tau in NT2N Neurons-To examine the phosphorylation sites affected by lithium and GSK-3, we treated NT2N neurons with 10 mM LiCl for 2 h or infected the cells with GSK-3␤/SFV overnight. Control cells received no treatment. Cell lysates were resolved on 10% SDS-PAGE, and immunoblot was performed with a panel of site-specific phosphorylation-sensitive anti-tau antibodies. The results of a typical experiment are shown in Fig. 3A. The immunoreactivities of phosphorylated and nonphosphorylated tau were quantitated and normalized to the levels of T14/46 (total tau), as shown in Fig. 3B. With the overexpression of GSK-3, T1 (dephosphorylated residues 189 -207) immunoreactivity decreased, whereas PHF1 (phosphorylated serine 396/404), T3P (phosphorylated serine 396), PHF13 (phosphorylated serine 396), AT8 (phosphorylated serine 202 and threonine 205), AT270 (phosphorylated threonine 181), and PHF6 (phosphorylated threonine 231) immunoreactivities increased. Lithium treatment had the opposite effect on the immunoreactivities of the same antibodies (Fig. 3, A and B). However, phosphoryla- tion at the site recognized by 12E8 (phosphorylated serine 262) was not altered by GSK-3 or by lithium. These results demonstrate that lithium and GSK-3 affect the same sites of phosphorylation in tau. Thus it is likely that lithium exerts its effects on tau phosphorylation through the direct inhibition of GSK-3.
The Effect of Lithium on Tau Phosphorylation Is Not Blocked by the Inhibition of Phosphatases-The phosphorylation status of tau depends on a balance between kinase and phosphatase activities. A decrease in kinase activity or an increase in phosphatase activity, or both, could result in reduced tau phosphorylation. To test whether or not the effect of lithium on tau phosphorylation requires the activation of a known phosphatase, we treated NT2N neurons with lithium in the presence of 0 -50 nM calyculin A, a potent inhibitor of protein phosphatases 1 and 2A (42,43). By inhibiting these phosphatases, calyculin A induced a decrease in T1 immunoreactivity in a dose-dependent manner in cells not treated with lithium (Fig. 4A). Despite this inhibition of phosphatases by calyculin A, 10 mM LiCl strongly increased T1 immunoreactivity (Fig. 4A), suggesting that the effect of lithium on tau phosphorylation is not dependent on the activation of protein phosphatase 1 or 2A.
We also showed that the inhibition of phosphatases by caly-culin A and the overexpression of GSK-3 had additive effects in inducing hyperphosphorylation of tau in NT2N neurons. As shown in Fig. 4B, calyculin A induced a dose-dependent decrease in T1 immunoreactivity in cells infected with LacZ/SFV. A stronger decrease in T1 was observed when cells infected with GSK-3␤/SFV were treated with calyculin A.

The Effects of Lithium on GSK-3 and Tau Phosphorylation
Are Reversible-To study further the mechanism of the inhibition of GSK-3 by lithium, we determined whether the effects of lithium on tau phosphorylation and GSK-3 are reversible. NT2N neurons were treated with 10 mM LiCl for 2 h and subsequently washed and incubated with lithium-free medium for another 0.5 or 2 h. As controls, the cells were treated with lithium for 2, 2.5, or 4 h. Tau phosphorylation was detected by immunoblot analysis with T14/46 and T1. As shown earlier, lithium inhibits the phosphorylation of tau after 2-4 h of incubation as indicated by an increase in T1 immunoreactivity. This effect was reversed rapidly by the removal of lithium, since T1 immunoreactivity resumed to its control levels (Fig.  5A). We also demonstrated in an in vitro kinase assay that the inhibition of GSK-3 activity by lithium is reversible (Fig. 5B). When the GSK-3 kinase assay was performed in the presence of lithium after the immunoprecipitated GSK-3 was incubated with 10 mM LiCl for 20 min (LiClϩ, WashϪ), GSK-3 activity (represented by 32 P incorporation into tau) was strongly reduced compared with the control (no lithium incubation, LiClϪ, WashϪ). When cells were washed extensively after 20 min of lithium incubation, GSK-3 activity was recovered (LiClϩ, Washϩ) (Fig. 5B). These results indicate that lithium reduces tau phosphorylation in the NT2N neurons by direct and reversible inhibition of GSK-3 and are in agreement with a recent in vitro study showing that lithium reversibly inhibits GSK-3 (47).
Lithium Increases Tau Binding to Microtubules and Affects Microtubule Assembly-The phosphorylation status of tau affects its affinity for microtubules, and this also may alter the stability of microtubules (7,8). We therefore examined whether lithium treatment affects the microtubule binding of tau and microtubule assembly in NT2N neurons. After overnight treatment by 10 mM LiCl, NT2N neurons were lysed, and soluble (S) and pelletable cytoskeletal (P) fractions were obtained from the cell lysates. The tau protein in the cytoskeletal fraction (bound to microtubules) and in the soluble fraction (unbound to microtubules) was determined by immunoblotting analysis with T14/46 (Fig. 6A). Shown in Fig. 6B, the ratio of microtubulebound tau to soluble tau was increased significantly by lithium treatment. This promotion of tau binding to microtubules is likely a result of the lithium-induced decrease in tau phosphorylation.
We also examined the effects of lithium on microtubule assembly by immunoblotting with antibodies against ␣-tubulin, Tyr-tubulin, and Glu-tubulin. The assembly of microtubules by tubulin polymerization is a dynamic equilibrium (44). When ␣-tubulin is incorporated into microtubules, it undergoes posttranslational modifications such as detyrosination to generate the more stable Glu-microtubules (45,46). However, when Glu-microtubules are depolymerized, tyrosine residues are rapidly added back to generate Tyr-tubulin. Therefore, as shown in Fig. 6A (Control), in the normal postmitotic NT2N neurons, more Tyr-tubulins are found in the soluble fraction (S), whereas 90% of Glu-tubulins are present as Glu-microtubule polymers in the cytoskeletal fraction (P). The overall total ␣-tubulin subunits are also distributed with 70% as microtubule polymers (Fig. 6A, Control). After lithium treatment, the amounts of soluble ␣-tubulin and Tyr-tubulin decreased, but the ␣-tubulin and Tyr-microtubule polymers remained un- Lithium treatment affected all of these sites too, but its effect was the opposite of that of GSK-3. Neither GSK-3 nor lithium altered the phosphorylation on the 12E8 site (phosphorylated serine 262). changed or slightly increased (Fig. 6A, LiCl). This resulted in a shift of the equilibrium of microtubule polymerization favoring assembly, since the ratios of ␣-tubulin polymers to soluble ␣-tubulin subunits increased significantly (Fig. 6B). However, lithium treatment did not alter the distribution of Glu-tubulin and Glu-microtubules (Fig. 6, A and B). These effects of lithium on microtubule assembly may partly be attributed to the increased microtubule stability as a result of enhanced tau binding.

DISCUSSION
Lithium has been used widely for decades to treat bipolar (manic-depressive) disorder. Despite its remarkable efficacy, the molecular mechanisms underlying its therapeutic actions have not been elucidated fully. It has been proposed that the inhibition of GSK-3 by lithium could explain its mechanism of action (21), but this has not been demonstrated in nerve cells. In the current study, we demonstrated in neuronal NT2N cells that lithium treatment prevents the phosphorylation of tau through direct and reversible inhibition of GSK-3. This pro-vides another clue as to how lithium elicits its effects on the central nervous system.
One other recent study conducted in non-neuronal COS1 cells transiently transfected with GSK-3 and tau has shown that lithium treatment resulted in a decrease in tau immunoreactivity detected by the MAb AT8 (phosphorylated serine 202 and threonine 205) (47). However, this study did not examine other phosphorylation sites of tau, and the non-neuronal nature of COS1 cells limits its relevance to the neurons of the central nervous system. The NT2N neuronal culture system was used to conduct the current study. We demonstrated that apart from the AT8 site, lithium affects a series of other sites phosphorylated by GSK-3, including serine 202/threonine 205, serines 396/404, threonine 181, and threonine 231. The 12E8 site (phosphorylated serine 262, not a proline-directed serine site), on the other hand, is not regulated by GSK-3 and not affected by lithium.
We showed previously that insulin and insulin-like growth factor I have effects similar to those of lithium on tau phosphorylation and the binding of tau to microtubules in NT2N neurons (10). These effects are also mediated through the inhibi- LiCl for 2 h and subsequently washed with lithium-free medium and incubated for another 0.5 or 2 h. As controls, cells were treated with lithium without wash for 2, 2.5, or 4 h. Lithium-induced increase in T1 immunoreactivity was reversed rapidly after removal of lithium. Panel B, overexpressed GSK-3 was immunoprecipitated, incubated with or without 10 mM LiCl for 20 min (LiClϩ or Ϫ). GSK-3 kinase assay was performed in the presence of lithium (WashϪ) or after lithium was removed by the wash (Washϩ). Lithium inhibited GSK-3-mediated 32 P incorporation into tau, but the effect was reversed by the removal of lithium.
FIG. 6. Lithium enhances tau binding to microtubules and affects microtubule assembly. NT2N neurons were treated with or without 10 mM LiCl overnight. Cell lysates were separated into cytoskeletal fraction (P) and soluble fraction (S). Panel A, tau, ␣-tubulin (␣-tub), Tyr-tubulin (Tyr-tub), and Glu-tubulin (Glu-tub) present in each fraction were determined by immunoblot analysis. Panel B, the ratios of tau, ␣-tubulin, Tyr-tubulin, and Glu-tubulin in the cytoskeletal fraction versus those in the soluble fraction (P/S) were determined after quantitation and plotted. Lithium treatment significantly increased tau binding to microtubules and shifted the microtubule assembly equilibrium toward polymerization (n ϭ 4, *p Ͻ 0.01). The distribution of Glu-tubulin was not altered. tion of GSK-3, although by a different mechanism that involves signal transduction and phosphorylation-dependent inactivation of GSK-3. Therefore, lithium is not only an insulin mimic as a stimulator of glycogen synthesis in hepatocytes but also as a regulator of tau phosphorylation in neuronal cells.
Lithium also has been reported to inhibit the phosphorylation of other cytoskeletal proteins, e.g. in cultured chick sensory neurons, the newly synthesized mid sized neurofilament subunit (23); and in PC12 cells, chartins, another group of microtubule-associated proteins (48). Although the mechanism underlying these effects is currently not known, they may be mediated by the inhibition of GSK-3.
The effects of lithium on microtubule assembly are intriguing. In cultured chick sensory neurons, lithium blocks initial neurite outgrowth. But, if neurites are allowed to begin extending before exposure to lithium, further outgrowth ceases, and neurites appear frozen at the length achieved before treatment (24,48). Moreover, lithium treatment does not induce cell death, and the cytoskeletal alterations are completely reversible (24). Lithium has also been reported to induce rapid degradation of newly synthesized tubulin in neurons, resulting in a decrease in the fraction of tubulin present in the unassembled form. However the assembled microtubules are not affected by this degradation. The dynamic equilibrium of tubulin polymerization is actually shifted toward assembly, since the proportion of tubulin present in the assembled microtubule fraction is increased (49). Our results are consistent with these findings, and they suggest that the decreased tau phosphorylation and increased tau binding to microtubules may contribute to this stabilizing effect of lithium on assembled microtubules. Because neurite elongation is primarily a function of microtubule dynamics, this microtubule stabilizing effect, coupled with the increased degradation of nascent tubulin, may halt further neurite growth and in the meantime preserve the extant neurites.
One reason why the molecular mechanisms underlying the therapeutic effects of lithium have been difficult to characterize is that lithium could affect multiple cellular targets. It has been proposed that lithium targets key components of signal transduction pathways (50). One of the prevailing hypotheses is the inositol depletion hypothesis, which is based on the observation that lithium inhibits inositol monophosphatase (51,52). As a result of this inhibition, the cellular source of inositol could be depleted, and cells could therefore become unable to generate inositol 1,4,5-trisphosphate. Thus, the inositol 1,4,5-trisphosphate-dependent responses to extracellular stimulation would be blocked. However, this hypothesis cannot fully explain the effect of lithium on glycogen synthesis or on cell fate and development, because treatment with L690,330, a compound that is approximately 1,000-fold more potent than lithium in inhibiting inositol monophosphatase (53,54), does not parallel the dramatic teratogenic effects of lithium (21). Furthermore, the membrane-permeable analog of L690,330, termed L690,488, does not affect tau phosphorylation in NT2N neurons, 2 and thus inhibition of inositol monophosphatase does not appear to explain the effects of lithium on tau phosphorylation. The observations that lithium directly inhibits GSK-3 and decreases tau phosphorylation help to explain these effects and provide new insights into the pathogenesis and treatment of bipolar disease.
The hyperphosphorylation of tau and formation of paired helical filaments are believed to be critical events in the pathogenesis of Alzheimer's disease and a group of other neurodegenerative diseases (1,7,(55)(56)(57)(58). The extent and distribution of neurofibrillary lesions correlate reliably with the degree of dementia in Alzheimer's disease patients (59 -61). We have shown previously that overexpression of GSK-3 can induce an Alzheimer's disease-like phosphorylation of tau in neurons (10). It is likely that GSK-3 plays an important role in the normal and abnormal regulation of tau phosphorylation. By inhibiting GSK-3 and causing a decrease in tau phosphorylation, it is possible that lithium could be used as a therapeutic agent to block the hyperphosphorylation of tau and slow the progression of Alzheimer's disease. Future studies on the exact mechanism of this inhibition will facilitate the design of new GSK-3 inhibitors that could be used for the treatment of Alzheimer's disease and bipolar disease.