(Received for publication, April 15, 1997, and in revised form, May 9, 1997)
From the Department of Pharmacology and the
§ Department of Pathology and Laboratory Medicine,
University of Pennsylvania School of Medicine,
Philadelphia, Pennsylvania 19104
Hyperphosphorylated tau is the major component of paired helical filaments in neurofibrillary lesions associated with Alzheimer's disease. Hyperphosphorylation reduces the affinity of tau for microtubules and is thought to be a critical event in the pathogenesis of this disease. Recently, glycogen-synthase kinase-3 has been shown to phosphorylate tau in vitro and in non-neuronal cells transfected with tau. The activity of glycogen-synthase kinase-3 can be down-regulated in response to insulin or insulin-like growth factor-1 through the activation of the phosphatidylinositol 3-kinase pathway. We therefore hypothesize that insulin or insulin-like growth factor-1 may affect tau phosphorylation through the inhibition of glycogen-synthase kinase-3 in neurons. Using cultured human neuronal NT2N cells, we demonstrate that glycogen-synthase kinase-3 phosphorylates tau and reduces its affinity for microtubules and that insulin and insulin-like growth factor-1 stimulation reduces tau phosphorylation and promotes tau binding to microtubules. We further demonstrate that these effects of insulin and insulin-like growth factor-1 are mediated through the inhibition of glycogen-synthase kinase-3 via the phosphatidylinositol 3-kinase/protein kinase B signaling pathway.
Tau is a neuronal microtubule-associated protein found predominantly in axons (1). The function of tau is to promote tubulin polymerization and stabilize microtubules (2, 3). Tau, in its hyperphosphorylated form, is the major component of paired helical filaments (PHFs),1 the building block of neurofibrillary lesions in Alzheimer's disease (AD) brain (4-6). Hyperphosphorylated tau is also found in neurofibrillary lesions in a range of other central nervous system disorders (7). In AD patients, the extent and topographical distribution of the neurofibrillary lesions correlate reliably with the degree of dementia (8, 9).
Hyperphosphorylation impairs the microtubule binding function of tau. PHF-tau does not bind to microtubules unless it is dephosphorylated (10, 11). It has been hypothesized that the reduced binding ability of PHF-tau to microtubules, coupled with reduced levels of normal tau, destabilizes microtubules in AD. This results in the disruption of vital cellular processes, such as rapid axonal transport, and leads to the degeneration of affected neurons.
Several serine/threonine protein kinases have been shown to phosphorylate tau in vitro. They include mitogen-activated protein kinase (MAPK) (12, 13), glycogen-synthase kinase-3 (GSK-3) (14-16), cyclin-dependent kinase 5 (17, 18), cAMP-dependent protein kinase, Ca2+/calmodulin-dependent protein kinase II (19), and a recently cloned 110-kDa protein kinase (MARK) (20). However, it is not clear which of these kinases phosphorylate tau in neuronal cells and are authentic regulators of tau phosphorylation in vivo. One of these kinases, GSK-3, has a possible physiological role in regulating tau phosphorylation, because recent studies have demonstrated that it induces cellular tau phosphorylation in Chinese hamster ovary cells and COS cells transfected with tau (21, 22).
Two isoforms of GSK-3, GSK-3 (51 kDa) and GSK-3
(46 kDa), are
encoded by two different genes. They share 85% homology at the amino
acid level (23). The activities of both isoforms can be down-regulated
in response to insulin and growth factors, such as insulin-like growth
factor (IGF-1) (23-26). This down-regulation involves phosphorylation
of an N-terminal serine residue (serine 21 for GSK-3
and serine 9 for GSK-3
) (24-26). This phosphorylation is putatively mediated
by several protein kinases, e.g. mitogen-activated protein kinase-activated protein kinase 1 (MAPKAPK1 or p90rsk),
which lies on the MAPK cascade (24), and the 70-kDa S6 kinase (p70 S6K) (25) and protein kinase B (PKB or Akt) (26), which lie
downstream of the phosphatidylinositol 3-kinase (PI(3)K) pathway.
GSK-3 is abundant in the brain (27). Insulin, IGF-1, and their receptors are also found throughout the developing and mature vertebrate brain (28-30). But it is currently unclear whether GSK-3 activity is regulated by insulin and IGF-1 in neurons. Recent studies have shown that insulin and IGF-1 promote survival by signaling through the PI(3)K-PKB pathway in cultured cerebellar neurons (31). This led to our speculation that through the activation of PKB, insulin or IGF-1 may induce inhibition of GSK-3 and affect tau phosphorylation in neuronal cells.
To test this hypothesis, we conducted our study in human NT2N neurons, which are derived from a human teratocarcinoma cell line (NTera2/D1 or NT2) after treatment with retinoic acid (32, 33). These NT2N cells resemble late embryonic central nervous system neurons and express the fetal isoform of tau (33-35). Using this system, we demonstrate that insulin and IGF-1 reduce tau phosphorylation and promote tau binding to microtubules. This effect is mediated through the inhibition of GSK-3 via the PI(3)K-PKB signaling pathway.
Recombinant GSK-3 and PD98059 were purchased
from New England Biolabs; IGF-1 was from Promega; LY294002 was from
BIOMOL Research Laboratories; protein kinase A inhibitor was from
Upstate Biotechnology; [
-32P]ATP and
125I-labeled goat anti-mouse IgG were from NEN Life Science
Products; and the Semliki Forest Virus gene expression system was from
Life Technologies, Inc. Taxol was obtained from Dr. V. Narayanan of the
NCI, National Institutes of Health. Other chemicals were purchased from
Sigma.
NT2 cells were grown and maintained as described (33). Briefly, NT2 cells were treated with retinoic acid for 5 weeks before the neuron-like NT2N cells were separated from the parent NT2 cells and replated on 6-well plates previously coated with poly-D-lysine (10 µg/ml) and matrigel. These NT2N neurons were maintained in Dulbecco's modified Eagle's medium with high glucose (DMEM-HG), supplemented with 5% fetal bovine serum, penicillin/streptomycin, and mitotic inhibitors (1 µM cytosine arabinoside, 10 µM fluorodeoxyuridine, and 10 µM uridine). 2-4-week-old NT2N cells were used for experiments.
Overexpression of GSK-3Three GSK-3/SFV viral
constructs carrying C-terminal c-Myc tags (EQKLISEEDL) were made.
GSK-3
WT is the wild type GSK-3
; GSK-3
S9A is a constitutively
active form of GSK-3
, in which the serine 9 residue was mutated to
alanine; and GSK-3
KM is an inactive mutant engineered by mutating
lysines 85 and 86 to methionine and isoleucine. Another SFV construct
expressing
-galactosidase (LacZ) was used as a control. To
overexpress these constructs, NT2N cells were incubated for an hour
with viral stocks diluted in DMEM-HG (MOI = 10) and then
replenished with regular medium and incubated overnight. For
experiments that examined the effects of insulin on infected cells,
cells were incubated for 6 h after infection, serum-starved (in
DMEM-HG with 0.1% fetal bovine serum) for another hour, and then
treated with insulin (100 ng/ml) for 5 min.
To study the effects of insulin and IGF-1, NT2N cells were serum-starved in DMEM-HG with 0.1% fetal bovine serum for an hour and then treated with 100 ng/ml of insulin or 10 ng/ml of IGF-1 for 5 min. Alternatively, NT2N cells were pretreated with PD98059 (50 µM, 60 min), LY294002 (100 µM, 15 min), wortmannin (100 nM, 15 min), 8-Br-cAMP (2 mM, 15 min), or rapamycin (100 nM, 15 min) before insulin or IGF-1 treatment. Pervanadate was prepared by incubating one part of 500 mM H2O2 with five parts of 10 mM Na3VO4 at room temperature for 10 min (36). This stock was added to NT2N cells at 1:100 dilution for 2 h.
Immunoblot Analysis of Tau PhosphorylationNT2N cells were
lysed in ice-cold high salt RAB buffer (0.1 M MES, 0.5 mM MgSO4, 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 (37). After centrifugation for 20 min at 50,000 × g at 4 °C, the protein concentrations of the supernatants were determined by the bicinchoninic acid method (Pierce). An equal
amount of total protein was resolved on 10% SDS-polyacrylamide gel
electrophoresis for immunoblotting analysis with anti-tau antibodies.
The blots were developed by enhanced chemiluminescence (ECL) or
3,3-diaminobenzidine methods. Alternatively, for the quantitation of
the relative levels of tau protein, 125I-labeled goat
anti-mouse IgG was used as secondary antibody, and the blots were
exposed to PhosphorImager plates as described (37). 10 µg of total
protein was loaded for the detection of tau on immunoblots using the
monoclonal antibodies T14/46, 25 µg for the antibodies T1 and PHF1
and 50 µg for the antibodies AT8, AT270, and PHF6.
NT2N cells
were treated with 100 ng/ml of insulin or infected with GSK-3
S9A/SFV overnight before they were lysed in 37 °C RAB buffer
(essentially high salt RAB without NaCl) supplemented with 0.1% Triton
X-100, 20 µM taxol, 2 mM GTP, and a mixture
of protease inhibitors (38, 39). Cell lysates were Dounce homogenized 20 times and centrifuged for 20 min at 50,000 × g at
25 °C. The protein concentrations of the soluble fractions were
adjusted to be the same. The insoluble cytoskeleton fractions (pellet) were resuspended in ice-cold RAB buffer of the same volumes as the
corresponding soluble fractions. Equal volumes of both fractions (approximately 5 µg of total protein from each fraction) were resolved on 10% SDS-polyacrylamide gel electrophoresis for immunoblot analysis with a monoclonal anti-
-tubulin antibody. To enrich tau,
the samples were boiled for 5 min, incubated on ice for 5 min, and
clarified by centrifugation for 15 min at 4 °C. Equal volumes
(containing approximately 10 µg of total protein) were resolved
on 10% SDS-polyacrylamide gel electrophoresis for immunoblot analysis
with T14/46. 125I-Labeled goat anti-mouse IgG was used to
detect the immunoreactivities, and quantitation was performed
with the ImageQuant software (Molecular Dynamics). The ratios of
microtubule-bound tau (in cytoskeletal fraction)
versus soluble tau (in the soluble fraction) were
calculated according to T14/46 immunoreactivities.
Infected NT2N cells
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
Na3VO4, and 1 µM microcystin)
supplemented with a mixture of protease inhibitors. Overexpressed
GSK-3
containing a c-Myc tag was immunoprecipitated with a rabbit
polyclonal anti-c-Myc antiserum. The immunoprecipitates were washed
twice with lysis buffer and twice with GSK-3 kinase buffer (20 mM Hepes, pH 7.2, 10 mM MgCl2, 10 mM MnCl2, 1 mM dithiothreitol, 0.2 mM EGTA, and 5 µM ATP). GSK-3 kinase reaction
was performed at 30 °C for 15 min in 20 µl of GSK-3 kinase buffer
supplemented with 4 µg of recombinant tau and 5 µCi of
[
-32P]ATP and terminated by adding 5 × sample
buffer and boiling for 5 min. The entire reaction was then resolved on
10% SDS-polyacrylamide gel electrophoresis, transferred to
nitrocellulose membrane, and exposed to PhosphorImager plate. GSK-3
kinase activity was represented by the 32P incorporation
into tau and normalized by the amount of GSK-3 present in each reaction
determined by immunoblot analysis with a monoclonal anti-GSK-3
antibody. To determine endogenous GSK-3 activity, the monoclonal
anti-GSK-3 antibody was used for immunoprecipitation.
For PKB immunoprecipitation kinase assay, NT2N cells were harvested in
the same lysis buffer, and PKB was immunoprecipitated with a sheep
polyclonal anti-PKB antibody. The PKB kinase assay was performed as
described (31). Briefly, PKB kinase reaction was performed at 30 °C
for 15 min in 40 µl of PKB kinase buffer supplemented with 2 µg of
protein kinase A inhibitor, 10 µCi of [-32P]ATP, and
200 ng of recombinant GSK-3
. The amount of immunoprecipitated PKB
present in each reaction was detected by immunoblot analysis with the
anti-PKB antibody.
All the anti-tau antibodies used are mouse
monoclonal. T14 and T46 are phosphorylation-independent (40, 41); T1 is
nonphosphorylation-dependent (dephosphorylated residues
189-207) (1, 42), PHF1 (phosphoserine 396/404) (10, 43), PHF6
(phosphothreonine 231),2 PHF13
(phosphoserine 396),2 AT270 (phosphothreonine 181) (45,
46), AT8 (phosphoserine 202/phosphothreonine 205) (45, 46), and 12E8
(phosphoserine 262) (47) are phosphorylation-dependent. T1
was obtained from Dr. L. Binder; PHF1 was from Drs. P. Davis and S. G. Greenberg; and AT8 and AT270 were from Innogenetics. The rabbit
polyclonal anti-c-Myc was raised against synthetic peptide EQKLISEEDL.
The mouse monoclonal anti-GSK-3 was purchased from Transduction
Laboratories; the sheep polyclonal anti-PKB was from Upstate
Biotechnology; and the mouse monoclonal anti--tubulin was from
Amersham Corp.
To determine whether
insulin or IGF-1 stimulation alters tau phosphorylation, NT2N cells
were treated with or without insulin or IGF-1 for 5 min and lysed for
immunoblot analysis of tau phosphorylation with a panel of anti-tau
antibodies. T14/46, a mixture of two phosphorylation-independent
antibodies, detects total tau protein, whereas T1 detects
dephosphorylated tau at residues 189-207 (numbering according to the
largest isoform of central nervous system tau). With insulin and IGF-1
treatment, T1 immunoreactivity increased, indicating a reduction of tau
phosphorylation. This was confirmed by the decrease in AT8
immunoreactivity, which is specific for phosphorylated serine 202 and
threonine 205 (Fig. 1, A and B). The reduction of tau phosphorylation was induced by 100 ng/ml of
insulin and 10 ng/ml of IGF-1. At these concentrations, insulin and
IGF-1 bind with high affinity to their cognate receptors, because the
affinity for cross-reaction is 100-1000 times lower (60). Therefore it
is possible that both insulin and IGF-1 receptors mediate this
effect.
We then examined whether insulin or IGF-1 stimulation affects tau
binding to microtubules. Soluble and cytoskeletal fractions of cell
lysates were prepared from control and insulin-treated NT2N cells. The
tau protein and -tubulin in each fraction were detected by
immunoblot analysis (Fig. 1C). The ratio of
microtubule-bound tau (P) versus soluble tau
(S) was determined and plotted in Fig. 1D. This
ratio was significantly higher in cells treated with insulin than in
control cells (Fig. 1D), indicating an enhanced tau binding
to microtubules.
To determine whether GSK-3 phosphorylates tau in
neuronal cells, we overexpressed wild type (WT) and two mutant forms
(S9A and KM) of GSK-3 in NT2N cells using the SFV gene expression system. The S9A mutation abolishes
phosphorylation-dependent regulation at serine 9 and
results in a constitutively active enzyme. KM has a double mutation and
is enzymatically inactive. All three overexpressed GSK-3
carried a
c-Myc-tag at their C termini and were immunoprecipitated with
polyclonal anti-c-Myc antiserum. GSK-3 kinase assay was performed with
recombinant tau as substrate. Fig. 2A
demonstrates that GSK-3
S9A and WT phosphorylated tau in
vitro, whereas GSK-3
KM was inactive despite of a similar expression level. An irrelevant SFV construct expressing
-galactosidase (LacZ) was used as control.
To determine whether tau phosphorylation is altered by overexpression
of GSK-3 in vivo, we detected tau in infected NT2N cells
by immunoblotting with a panel of site-specific and
phosphorylation-sensitive anti-tau antibodies. Fig. 2B
demonstrates that in cells infected with the active GSK-3
/SFVs (S9A
and WT), T1 (dephosphorylated tau) levels decreased, whereas PHF1
(phosphoserine 396/404), AT270 (phosphothreonine 181), and PHF6
(phosphothreonine 231) levels increased, indicating elevations of
phosphorylation at these sites. The phosphorylation levels also
increased at sites recognized by AT8 (phosphoserine
202/phosphothreonine 205) and PHF13 (phosphoserine 396) (data not
shown). However, the immunoreactivity to 12E8 (phosphoserine 262) was
not affected by GSK-3 (Fig. 2B). Cells infected with LacZ/SFV were used as control in which tau phosphorylation levels did
not differ from noninfected cells (data not shown). The expression of
GSK-3
KM also did not alter tau phosphorylation (data not shown).
We also examined the effects of GSK-3 on tau binding to
microtubules. As shown in Fig. 2 (C and D),
significantly less tau was bound to microtubules in cells infected with
GSK-3
S9A/SFV, because the ratio of cytoskeletal tau
versus soluble tau was decreased. These data suggest that in
neuronal cells, GSK-3 phosphorylates tau at multiple sites and reduces
the affinity of tau for microtubules.
It has been shown previously that in non-neuronal cells,
insulin and IGF-1 down-regulate GSK-3 activity by inducing the
phosphorylation of a serine residue (serine 9 in GSK-3 and serine 21 in GSK-3
, 23-26). To determine whether insulin mediates its effects
on tau phosphorylation through the inhibition of GSK-3 in neurons, we first examined whether or not in NT2N cells insulin inhibits GSK-3
activity through the phosphorylation of serine 9. To do this, the cells
were infected with either GSK-3
S9A/SFV or GSK-3
WT/SFV and
subsequently treated with insulin. Overexpressed GSK-3
was immunoprecipitated, and the kinase activities were determined quantitatively using the in vitro kinase assay and
normalized to the amount of GSK-3 in each reaction. As shown in Fig.
3A, wild type GSK-3
activity
(WT) was inhibited by insulin, whereas GSK-3
S9A activity
(S9A) remained unchanged. The endogenous GSK-3 activity was
also assayed in noninfected NT2N cells by immunoprecipitating both
GSK-3
and GSK-3
with a monoclonal anti-GSK-3 antibody. The
normalized relative endogenous GSK-3 activity was also decreased by
insulin treatment (Fig. 3A, Endogenous). These
results indicate that the phosphorylation of serine 9 is involved in
the insulin-mediated inhibition of GSK-3
in neurons.
To further demonstrate that insulin regulates tau phosphorylation by
the inhibition of GSK-3, we infected NT2N neurons with GSK-3 S9A/SFV
or GSK-3
WT/SFV and subsequently treated the cells with insulin. The
total tau levels were not altered by infection or by insulin treatment
(T14/46, Fig. 3B). Similar to what was shown in
Fig. 2, in cells not treated with insulin, overexpression of GSK-3
S9A and GSK-3
WT resulted in a decrease in T1 and an increase in
PHF1. However, insulin treatment only reduced tau phosphorylation in
cells expressing GSK-3
WT but not in cells expressing GSK-3
S9A,
because an increase in T1 and a decrease in PHF1 can be detected only
in cells infected with GSK-3
WT/SFV (Fig. 3B). These
experiments demonstrate that tau phosphorylation cannot be regulated by
insulin unless GSK-3 activity can be inhibited through serine 9. This
strongly suggests that insulin regulates tau phosphorylation through
the inhibition of GSK-3.
Three different
protein kinases have been proposed to phosphorylate and inhibit GSK-3.
They are MAPKAPK1, which lies downstream of the MAP-kinase cascade, and
p70 S6K and PKB, which lie on the PI(3)K pathway. All three have been
shown to phosphorylate GSK-3 at serine 9 and GSK-3
at serine 21 (24-26).
To test which of these kinases and pathways mediate(s) the effects of
insulin and IGF-1 on tau phosphorylation and microtubule binding, we
utilized specific inhibitors of these pathways as tools. Wortmannin
(48), which inhibits PI(3)K, blocked the insulin and IGF-1-induced
dephosphorylation of tau (Fig. 4, A and
C). LY294002, which is another specific PI(3)K inhibitor
(49), produced the same results (Fig. 4, B and
C). PD98095 (50), which specifically inhibits the activation
of MAPKAPK1 by the MAPK cascade, had no effect; nor did the cAMP
analogue 8-Bromo-cAMP (51). LY294002 also abolished the insulin-induced
promotion of microtubule binding of tau (Fig. 1C). These
results indicate that insulin and IGF-1 exert their effects on tau by
signaling through the PI(3)K pathway.
Both p70 S6K and PKB are downstream of the PI(3)K pathway. There is evidence that p70 S6K activation can be promoted by both PI(3)K and PKB. To test whether p70 S6K plays a role in the regulation of tau phosphorylation by insulin and IGF-1, we inhibited p70 S6K with rapamycin (52), which blocks the phosphorylation and activation of p70 S6K. As shown in Fig. 4A, rapamycin did not alter the insulin-induced dephosphorylation of tau.
These results point to the PI(3)K-PKB pathway. To examine the
involvement of PKB, we treated NT2N cells with insulin and IGF-1, immunoprecipitated PKB, and tested PKB activity by an in
vitro kinase assay. As shown in Fig. 5A,
PKB activity, indicated by 32P incorporation into human
recombinant GSK-3, was greatly increased by insulin and IGF-1
treatment. We also treated NT2N cells with pervanadate, which was
reported to be a strong activator of PKB (26), and examined the effect
on tau phosphorylation. Fig. 5B shows that the
phosphorylation level of tau was greatly reduced by this treatment,
because an increase in T1 and a decrease in PHF1 immunoreactivities
were accompanied by an increase in the electrophoretic mobilities of
the tau protein bands as detected by T14/46 and T1. These data support
that the PI(3)K-PKB pathway mediates the effects of insulin and IGF-1
on tau phosphorylation.
Current hypotheses regarding the pathogenesis of AD include the
deposition of -amyloid as a result of aberrant amyloid precursor protein metabolism and formation of neurofibrillary lesions by aggregation of hyperphosphorylated tau protein. However, it seems that
amyloid deposits are not sufficient for dementia, because large numbers
of amyloid deposits are found in some cognitively normal individuals
(53). On the other hand, it has been confirmed that cognitive deficits
do not occur until dystrophic neurites containing PHFs and
neurofibrillary tangles have developed (8, 9, and 54). Thus
hyperphosphorylation of tau is believed to be a key event in the
pathogenesis of the neurofibrillary pathology and the dementia of
AD.
Several kinases have been shown to phosphorylate tau in vitro. But they do not necessarily do so in neuronal cells in vivo. For example, MAPK phosphorylates tau and generates a number of PHF-tau epitopes in vitro (12, 13). However, activating MAPK in cultures of primary neurons or transfected COS cells expressing tau does not increase the levels of phosphorylation for any PHF-tau epitopes (22). Another study also showed that stimulation of MAPK by v-raf transformation of fibroblasts fails to induce hyperphosphorylation of transfected tau (55). Therefore it is important to identify the kinases that are truly responsible for tau phosphorylation in neurons in vivo.
GSK-3 is one promising candidate. It phosphorylates tau in
vitro (14-16) and in non-neuronal cells transfected with tau (21, 22). Moreover, multiple phosphorylation sites of tau are reported to be
affected, including the sites recognized by
phosphorylation-dependent antibodies PHF1 (serines 396/404),
T3P (serine 396), AT8 (serine 202/threonine 205), AT270 (threonine
181), and AT180 (threonine 231) (21). Our results from the current
study confirmed the role of GSK-3 in tau phosphorylation in neuronal
cells. We also demonstrated that besides the previously reported
phosphorylation sites, overexpression of GSK-3 in NT2N cells also
increases phosphorylation at the sites recognized by antibodies T1
(dephosphorylated 189-207), PHF13 (serine 396), and PHF6 (threonine
231). But the phosphorylation level at the site recognized by 12E8
(serine 262) is not affected.
Our study also demonstrated that overexpression of GSK-3 reduces tau
binding to microtubules in NT2N cells. This agrees with a recent study
using transfected Chinese hamster ovary cells (56). Hyperphosphorylation of tau reduces its affinity for microtubules (10,
11), but it is currently unclear about the relative contributions made
by individual phosphorylated residues. A study using Chinese hamster
ovary cells transfected with tau has shown that phosphoserine residue
396, which lies right next to the microtubule-binding domain of tau,
makes a significant contribution toward the reduced microtubule binding
(10). However, the role of phosphoserine 262 (recognized by 12E8),
which lies within the microtubule-binding domain, is controversial. One
study suggests that the phosphorylation on this site alone eliminates
microtubule binding of tau (44), whereas another study shows that this
phosphorylation is not sufficient for abrogating such binding (47). Our
current study supports the latter point of view, because overexpression
of GSK-3 does not affect the phosphorylation of this site but still
causes reduced microtubule binding of tau.
In this study, we first report that tau phosphorylation and function
can be regulated through a signal transduction pathway by a hormone or
a growth factor. We showed that insulin and IGF-1 induce inhibition of
GSK-3, which results in tau dephosphorylation and increased microtubule
binding of tau. Summarized in Fig. 6 are the putative
signal transduction pathways through which insulin and IGF-1 might
exert their effects on tau. We demonstrated that the PI(3)K-PKB pathway
is involved in this action, but not the MAPK-MAPKAPK1 or the PI(3)K-p70
S6K pathways.
Insulin, IGFs and their receptors are found throughout the developing and mature vertebrate brain (28-30). Although the source of insulin in the brain remains a controversial issue, transcripts for IGFs are abundant in the central nervous system (29, 30). In addition, it has been shown that both insulin and IGF-1 receptor mRNAs are expressed in many different neuronal types and in ependymal cells of choroid plexus in the rat brain (28). Binding studies showed that these receptors are located on the cell bodies and dendrites of neurons and are enriched in the hippocampus and olfactory bulb of the rat brain (28). It is conceivable that the functions of insulin and IGF-1 in the nervous system are beyond the classic glucoregulatory effects observed in liver, muscle, and fat. Indeed, there is evidence that insulin and IGF-1 stimulate neuronal protein synthesis (57), increase neurite growth (58), induce tubulin mRNAs (59), and promote survival in neurons (31).
Insulin and IGF-1 receptors are structurally similar and belong to the
same family of receptor protein tyrosine kinases. There is evidence
that they have similar signaling modes: both phosphorylate insulin
receptor substrate-1 and activate PI(3)K, MAPK, and c-fos (60). In non-neuronal cells, the cellular actions of insulin are mainly
mediated via several signaling pathways involving changes in protein
phosphorylation (61). It is likely that at least some steps of these
pathways are shared among the classic target tissues and the nervous
system. We have shown both insulin and IGF-1 receptors are expressed in
NT2N cells, and upon stimulation by these ligands, tyrosine
phosphorylation of insulin receptor substrate-1 and PI(3)K is
induced.3 Our results presented in the
current study suggest that the inhibition of GSK-3 via the PI(3)K-PKB
pathway is also a common event. This leads to reduced phosphorylation
and activation of glycogen synthase in muscle, liver, and fat and
reduced phosphorylation of tau in neuronal cells. Another event that
occurs after insulin treatment in non-neuronal cells is the activation
of certain protein phosphatases, e.g. protein phosphatase 1. Activation of certain protein phosphatases may also affect tau
phosphorylation and microtubule stability in neuronal cells (35, 62).
But it is not known at this time whether insulin activates any
phosphatases in neurons. Nevertheless, our data suggest that the
dephosphorylation of tau induced by insulin is probably not mediated by
the activation of phosphatases but by the inhibition of GSK-3, because
the expression of GSK-3 S9A abolishes the effect.
The identification of a signal transduction pathway that regulates tau phosphorylation brings new insights to the study of the pathogenesis of AD. It raises the possibility that defects on this pathway may contribute to molecular pathology of the disease. Indeed, defects of glucose metabolism in AD brains have been well documented and are suggested to be the result of desensitized brain insulin receptors (63). This is also supported by the observation of increased insulin levels in cerebrospinal fluid of AD patients after an oral glucose tolerance test (64). If these observations indicate defects in the insulin receptor or its signaling pathways in AD, considering the effects of insulin on tau phosphorylation and function, long term malfunctioning may lead to the hyperphosphorylation of tau and impairment of tau function. Further studies need to be done to investigate whether in AD brains any abnormalities can be found in GSK-3, insulin, and IGF-1 receptors or their signaling pathways. This hypothesis should also be taken into consideration in future development of therapeutic strategies.
We thank Dr. John Q. Trojanowski for reading
the manuscript. We also thank Dr. M. Goedert for providing GSK-3 S9A
cDNA and Dr. P. Klein for providing GSK-3
WT and KM cDNAs.
Drs. L. Binder, P. Davis, and S. G. Greenberg are acknowledged for
providing some of the anti-tau antibodies.