Available via license: CC BY 4.0
Content may be subject to copyright.
THE
Joma
OF
BIOL~GICAL CHEMISTRY
0
1994 by
The
American Society for Biochemistry
and
Molecular Biology,
Inc.
Vol.
269, No.
20,
Issue
of
May
20,
pp.
14566-14574, 1994
Printed
in
U.S.A.
Glycogen Synthase Kinase-3P
Is
a Dual Specificity Kinase
Differentially Regulated by Tyrosine and Serinemhreonine
Phosphorylation*
(Received for publication, December
2,
1993, and
in
revised form, March
10,
1994)
Q.
May Wang, Carol
J.
Fiol,
Anna
A.
DePaoli-Roach, and Peter
J.
Roach
From
the Department
of
Biochemistry and Molecular Biology, Indiana University School
of
Medicine,
Indianapolis, Indiana
46202-5122
The enzyme glycogen synthase kinase-3
(GSK-3)
has
been implicated in the control of several metabolic en-
zymes and transcription factors in response to extracel-
lular signals. In the past, the enzyme has been consid-
ered to be a protein Ser/Thr kinase although it was
recently reported to contain
Tyr(P)
(Hughes,
K.,
Niko-
lakaki,
E.,
Plyte,
S.
E.,
Totty,
N.
E,
and Woodgett,
J.
Ft.
(1993)
EMBO
J.
12, 803408).
A
cDNA encoding rabbit
skeletal muscle
GSK-3P
was cloned and expressed in
Escherichia coli
as an active protein kinase, with appar-
ent
M,
46,000,
capable of phosphorylating several
known
GSK-3
substrates. Recombinant
GSK-30
autophospho-
rylated on Ser,
Thr,
and
Tyr
residues although the en-
zyme already contained
Tyr(P)
as judged by its recogni-
tion by anti-Tyr(P) antibodies. The net result of the
autophosphorylation was a =-fold reduction in enzyme
activity.
GSK-3a,
purified
from
rabbit muscle, also
un-
derwent autophosphorylation but
only
on Ser and Thr
residues. In this case, the autophosphorylation stabi-
lized the enzyme activity compared with the control
lacking ATP/Mg2‘.
Of
several phosphatases tested, the
A-phage phosphatase was the most effective in dephos-
phorylating at Ser and
Thr
residues but did not dephos-
phorylate at
Tyr
residues. The action of the A-phospha-
tase caused a reactivation of
GSK-30
to
-80%
of
the
starting activity. The protein tyrosine phosphatase
PTPlB
was able to dephosphorylate at
Tyr
residues
leading to a reduction in enzyme activity.
A
truncated
form of
GSK3P,
apparent
M,
40,000,
had a significantly
higher specific activity, was defective in autophospho-
rylation, and was not inactivated in the autophospho-
rylation reaction. We conclude that
GSK-3P
is a dual
specificity protein kinase in the same sense as the mito-
gen-activated protein kinase/ERK family
of
enzymes.
Phosphorylation at different residues differentially con-
trols enzyme activity, SerRhr phosphorylation causing
inactivation and
Tyr
phosphorylation resulting in in-
creased activity.
Protein SerlThrlTyr kinases, among the most prevalent
forms of regulatory proteins in cells (reviewed by Edelman
et
al.,
1987; Hunter 1991), have been classified on the basis
of
their specificity for the modification of Ser/Thr
or
Tyr
residues
in substrate proteins. With the availability of primary sequence
information
for
a
large number of protein kinases, new en-
zymes can often be typed according to signature sequences
characteristic of the two biochemical classes (Hanks
et
al.,
Grants
Dm7221
(to
P.
J.
R)
and
DK36569
(to
A.
D.
R).
The costs
of
*
This
work was supported in part
by
National Institutes of Health
publication
of
this article were defrayed in part
by
the payment
of
page
in
accordance
with
18
U.S.C. Section
1734
solely
to indicate this fact.
charges. This article must therefore be hereby marked
“aduertisement”
1988). The classification, however, was shown not
to
be abso-
lute with the discovery of the so-called ”dual specificity” en-
zymes that are capable of modifying Ser, Thr,
or
Tyr
residues
(Posada and Cooper, 1992; Lindberg
et al.,
1992). Of these en-
zymes, two of the most visible are the
MAP
kinase’
or
ERK
family (Wu
et
al.,
1991; Crews
et
al.,
1991) and the MEK
or
MAP
kinase kinase family (Gomez and Cohen, 1991; Crews
et
al.,
1992; Wu
et
al.,
1993; Zheng and Guan, 1993a).
MAP
kinase
autophosphorylates
at
both
Tyr
and Thr residues in the “TEY”
sequence (Wu
et
al.,
1991) located in the conserved protein
kinase subdomain VI11 as defined by Hanks
et
al.
(19881, al-
though no
Tyr
phosphorylation has been observed
so
far
in
exogenous substrates. Full activity of
MAP
kinase requires
both phosphorylations (Anderson
et
al.,
1990; Nishida and
Gotoh, 1993). One initial proposal was that activation of
MAP
kinase could be achieved by allosteric stimulation of the auto-
phosphorylation reaction via
an
activator protein (Seger
et
al.,
1991). However, it was also shown that
MAP
kinase could be
phosphorylated and activated by
a
separate protein kinase
called
MAP
kinase kinase
or
MEK (Gomez and Cohen, 1991;
Ahn
et
al.,
1991). MEK can modify both the
Thr
and the
Tyr
residues of
MAP
kinase (Gomez and Cohen, 1991; Nishida and
Gotoh, 1993; Zheng and Guan, 1993a) which makes it too
a
dual specificity protein kinase.
Glycogen synthase kinase
3
(GSK-3) was discovered and
named for
its
involvement in the control of glycogen metabo-
lism (Woodgett, 1991) and has been implicated in regulating
metabolic enzymes such
as
glycogen synthase (Cohen, 1986;
Roach, 1990) and ATP-citrate lyase (Hughes
et
al.,
1992).
GSK-3 also phosphorylates regulatory subunits of the cytosolic
and glycogen associated forms
of
the type I protein SerlThr
phosphatase (Hemmings
et
al.,
1982b; DePaoli-Roach, 1984;
Fiol
et
al.,
1988;
Dent
et
aZ.,
1989)
and the regulatory subunit
of
cyclic AMP-dependent protein kinase (Hemmings
et
al.,
1982a).
More recently, the
c-jun
(Boyle
et
al.,
1991; de Groot
et
al.,
1992;
Nikolakaki
et
al.,
1993), c-myb (Plyte
et
al.,
19921, c-myc (Plyte
et
al.,
19921, CREB2 and CREM (de Groot
et
al.,
1993) tran-
scription factors have been found to be substrates for GSK-3
and, at least in some cases, the phosphorylation may correlate
with altered function. Other recent work has shown that
GSK-3 phosphorylates the microtubule-associated tau proteins
and might induce
a
state similar to that found in Alzheimer’s
disease (Hanger
et al.
1992; Mandelkow
et
al.,
1992; Ishiguro
et
al.,
1993). Finally, GSK-3 has been shown
to
phosphorylate the
The abbreviations used
are:
MAP
kinase, mitogen-activated protein
kinase; GSK-3, glycogen synthase kinase-3; PAGE, polyacrylamide gel
electrophoresis; CREB,
cyclic
AMP-responsive element
binding
protein;
PVDF, polyvinylidene
difluoride.
CREM, cyclic AMP-responsive element modulator;
bp,
base
pair(s);
C.
J.
Fiol,
J.
S.
Williams,
Q.
M.
Wang,
P.
J.
Roach,
and A.
M.
Andrisani,
manuscript
submitted.
14566
This is an Open Access article under the CC BY license.
Glycogen Synthase Kinase
3
14567
largest subunit of eIF-2B, the factor involved in guanine nucle-
otide exchange on the protein synthesis initiation factor eIF-2
(Welsh and Proud, 1993). Sites phosphoylated by GSK-3 can be
divided into two classes.
For
some substrates, prior phospho-
rylation
of
the substrate,
to
form the motif -S-X-X-X-S(P)-, is a
strict requirement whereas in other substrates, no previous
phosphorylation is needed (Roach 1991; Wang
et al.,
1994). In
either case, many of the GSK-3 sites have Pro residues close to
the modified Ser
or
Thr (Plyte
et al.,
1992; Mandelkow
et
al.,
1992).
Multiple forms of GSK-3 have been identified. In mammals,
molecular cloning of rat brain cDNAs revealed the existence of
two closely related isoforms termed GSK-Sa and GSK-30
(Woodgett, 1990). The a-isoform is a 51-kDa protein sharing
95% identity with the 47-kDa p-isoform in the kinase domain.
In
Drosophila,
the
zeste-white3lshaggy
gene, which has critical
functions during embryogenesis, encodes multiple protein
products, via alternate splicing
of
the message (Siegfried
et al.,
1992; Ruel
et al.,
1993b). These proteins have
85%
identity
to
mammalian GSK-3 in their protein kinase domains (Woodgett
1991; Siegfried
et
al.,
1992) and GSK-3P, but not GSK-3a, is
able
to
restore wild-type phenotype in
zeste-white3 lshaggy
mu-
tants (Ruel
et al.,
1993a). In addition, the
shaggy
proteins re-
semble mammalian GSK-3 in such biochemical properties as
substrate specificity and regulation by post-translational modi-
fication (de Groot
et al.,
1992; Hughes
et al.,
1993). In the
budding yeast
Saccharomyces cerevisiae,
three genes are
known that encode proteins with significant similarity to
GSK-3. These are
MCKI
(Dailey
et al.,
1990; Neigeborn and
Mitchell, 1991; Shero and Hieter, 19911,
MDSl
(Puziss
et
al.,
1994) and
MRK13
which have 44,
58,
and 54% identity
to
the
kinase domain of GSK-3, respectively. The proteins encoded by
MCKl
and
MDSl
have been shown
to
phosphorylate a specific
GSK-3 peptide substrate (Puziss
et al.,
1994).
For
many years, little was known about the control of GSK-3.
Initial reports described the activation
of
the enzyme
by
insulin
(Yang
et al.,
1988; Villa-Moruzzi, 1989), but other data have
suggested the opposite. Ramakrishna and Benjamin (1988) de-
scribed insulin-induced inactivation of GSK-3 in adipocytes
while Welsh and Proud (1993) likewise observed inactivation
of
the enzyme in Chinese hamster ovary cells stimulated with in-
sulin. The mechanistic basis for such controls is still not un-
derstood, but recent experiments point
to
a possible role
for
phosphorylation. Goode
et al.
(1992) reported that
in vitro
pro-
tein kinase
C
was able
to
phosphorylate and inactivate GSK-3/3,
but not GSK-3a. In another study, it was found that GSK-3 iso-
lated
from
mammalian tissue
or
as a recombinant protein ex-
pressed in insect cells was modified at
Tyr
residues (Hughes
et
al.,
1993). The tyrosine was identified as Tyr-216
of
GSK-30,
which is equivalent
to
the tyrosine in the regulatory TEY se-
quence of
MAP
kinases.
As
in
MAP
kinase, this tyrosine phos-
phorylation appeared
to
enhance activity. In the present study,
we report the successful expression of active GSK-30 in
Escherichia coli
and demonstrate that GSK-3p is a dual speci-
ficity kinase in the same sense as
MAF'
kinase. Autophospho-
rylation occurred on Ser, Thr, and Tyr residues and correlated
with a net inactivation of the enzyme. Subsequent dephospho-
rylation at SerPThr residues restored activity whereas dephos-
phorylation at
Tyr
led
to
further inactivation. Therefore, GSK-3
is one of the few examples of an enzyme whose activity
is
regu-
lated in opposite senses by
Tyr
and Ser/Thr phosphorylation.
EXPERIMENTAL PROCEDURES
Molecular Cloning
of
GSK-3f3 cDNA-Based on known rabbit muscle
GSK-3 peptide sequences (Woodgett, 1990), degenerate oligonucleotide
T. A. Hardy, D. Wu, and
P.
J.
Roach, unpublished data.
primers corresponding to residues ELQIMR (sense) and EMNPNY (an-
tisense) were designed. Polymerase chain reaction with
Amplitaq
DNA
polymerase (Perkin-Elmer Cetus) was carried out in a programmable
heating block. Ten cycles of low stringency amplification consisting of
1
min of denaturation
at
94 "C, 4 min of annealing
at
37 "C, and
3
min of
polymerization
at
70 "C were first
run
using 10 pmol each of sense and
antisense primers and
a
rabbit skeletal muscle cDNA template which
was synthesized from rabbit skeletal muscle total RNA by
a
reverse
transcriptase reaction under conditions described previously (Graves et
al.,
1993). Subsequently, another
50
pmol of primers were added to the
reaction and 30 cycles of higher stringency reactions
(1
min of denatur-
ation
at
94
"C,
2 min of annealing at
50
"C,
3
min of polymerization at
70 "C) were performed. The amplified fragment, with
a
size of 580 bp,
was then subcloned into the pTZ19U vector for sequence analysis. After
its sequence was confirmed to correspond to GSK-3/3,
this
fragment was
labeled with [CX-~'P]ATP by random priming and used to screen an
unamplified random primed rabbit skeletal muscle hgtll cDNA library
constructed
as
described previously (Zhang et al.,
1989).
Prehybridiza-
tion of the filters (166,000 recombinants) was carried out at 65 "C for 4
h in
a
solution consisting of
10
x
Denhardt's
(1
x
Denhardt's solution:
0.02% (w/v) each of Ficoll, polyvinylpyrrolidone, and gelatin), 6
x
SSPE
(1
x
SSPE:
0.15
M
sodium chloride, 10 mM sodium phosphate, and
1
mM
EDTA, pH 7.4),
0.05%
sodium pyrophosphate, 0.1%
SDS,
and
0.1
mg/ml
of Torula RNA. Hybridization was
at
65 "C for
18
h in the same solution
except for addition of 2.0
x
lo6
counts/min/ml radiolabeled probe. The
filters were washed with 6
x
SSC
(1
x
SSC:
0.15
M
sodium chloride, 15
m
sodium citrate, pH 71,
0.05%
sodium pyrophosphate, and 0.1%
SDS
for
15
min
at
room temperature, and then twice
at
65
"C
with
1
x
SSC,
0.05%
sodium pyrophosphate, and 0.1%
SDS
for 30 min each. Five
positive clones were identified from screening 166,000 recombinant
phages and were taken to plaque purity (Sambrook et
al.,
1989). Nucle-
otide sequence analysis was performed by the dideoxy chain termina-
tion method (Sanger et
al.,
1977) using both universal and sequence
specific primers. One of these five clones was shown
to
contain a full-
length GSK-3P cDNA.
Expression
of
Recombinant GSK-3P in
E.
coli-The full-length cDNA
contained an internal EcoRI site within the coding region, and
so
the
two EcoRI-EcoRI subfragments were independently subcloned into the
pTZ19U plasmid. The 5'-fragment was digested with EcoRI and Eco57
I
to
generate a fragment lacking the first 45-bp sequence downstream of
the start codon.
Two
complementary oligonucleotides
(41
and 45 bp)
were synthesized and annealed to generate
a
double-stranded DNA
with an overhanging NdeI site
at
the 5'-end and an overhanging Eco57I
site at the 3'-end. This double-stranded DNA adaptor was ligated to the
Eco57I-EcoRI fragment
so
that the ATG start codon and the missing 45
bp of coding sequence were restored. The resulting fragment was li-
gated with PET-3c vector cut at the
NdeI
and the EcoRI sites to give an
intermediate construct.
The
3'-fragment, corresponding to the
COOH
terminus of GSK-3P, was excised from pTZl9U by restriction with
EcoRI, ligated into the intermediate construct at the EcoRI site, and the
orientation checked by restriction mapping. This ligated intermediate
construct had the full coding region of GSK-3p but now lacked the T7
transcription terminator which had been excised by cutting with EcoRI
and NdeI. Thus, the intermediate vector was cut at
a
unique vector
Aat
I1 site downstream of the stop codon, filled in, and then the entire
GSK-3p, coding sequence excised with NdeI. Unmodified PET-3c was
cut at the BamHI site, filled in, and then treated with NdeI. The GSK-3
coding fragment was then ligated into the new PET-3c vector via the
5'-NdeI site and
via
blunt ends at the 3'. The functional PET-GSK-3p
plasmid was
thus
formed. E. coli cells BL21DE3, transfected with
pET-GSK-3p, were grown at 37
"C
until the OD,,, was about 0.6 and
then induced with 0.1
lll~
isopropyl-P-thiogalactoside
for 6 h at 30 "C.
Purification
of
GSK-3"Purification of
GSK-3
activity from rabbit
skeletal muscle was carried out as previously reported (Fiol
et
al.,
1990).
For recombinant GSK-3P purification, E. coli cells expressing GSK-30
were collected from 1-liter cultures and resuspended in homogenization
buffer (buffer A) containing
10
mM
Tris,
pH 7.4,5% sucrose,
2
mM EDTA,
2 mM EGTA,
1
mM dithiothreitol, 2 mM benzamidine, 0.5 mM phenyl-
methylsulfonyl fluoride, 10 pg/ml leupeptin, and
10
&mI aprotinin.
The cell suspension was passed twice through
a
French press (900
poundslsquare inch), and the supernatant was collected by centrifuga-
tion
at
11,000
x
g
for
15
min. The supernatant was then batch-absorbed
for 2
h
onto
200
ml of DE52 (Whatman) equilibrated previously in buffer
A. The unbound material, which contained GSK-36 activity, was loaded
onto a phosphocellulose (P-11, Whatman) column (2.5
x
30 cm) which
had been washed with buffer A containing
50
mM NaCl. The GSK-3
kinase activity was eluted with a linear gradient of 50-500
mM
NaCl in
buffer A (total 360 ml). The pooled kinase activity was concentrated to
14568
Glycogen Synthase Kinase
3
less than
5
ml with an Amicon Centriprep
10,
diluted
1:lO
with buffer
A, and loaded onto
a
Cibacron Blue-Sepharose column
(1
x
20 cm)
equilibrated with
50
n"
NaCl in buffer A. A
50-750
mM NaCl gradient
in buffer A (total
50
ml) was used to elute the kinase activity. Fractions
(0.75
ml)
containing GSK-30 activity were pooled, concentrated to
1
ml
as
described above, diluted with 9 ml of buffer A, and finally chromato-
graphed on an S-Sepharose
Fast
Flow (Pharmacia LKB Biotechnology
Inc.) column
10.5
x
20 em). GSK-3 activity, which eluted around 250
n"
NaCl from the column, was dialyzed against
10
mM
Tris,
pH
7.4,
0.2%
mercaptoethanol,
50%
glycerol and stored at
-20
"C. The isolation of the
proteolytic fragment of GSK-3j3 was performed in the same way except
that several protease inhibitors (leupeptin, aprotinin, and phenylmeth-
ylsulfonyl fluoride) were not included in the buffer during the purifica-
tion.
Protein Kinase Assays-GSK-3 kinase activity was measured rou-
tinely by assay of phospho-CREB peptide phosphorylation. A typical
peptide phosphorylation with the kinase was performed
at
30 "C in
15
pl of reaction containing
50
PM
phospho-CREB, 100
p~
[y3'P]ATP
(500-
2000 countdmidpmol), 10 mM MgCl,,
1
mM dithiothreitol,
5%
glycerol,
0.1
mg/ml bovine serum albumin, 30 mM
Tris,
pH
7.4.
The 32P-labeled
peptide was recovered on Whatman p81 phosphocellulose paper,
washed four times with
75
mM phosphoric acid for a total of 20 min, and
counted by liquid scintillation spectrometry. Preparation of the phos-
pho-CREB peptide was
as
described.' Briefly, the 13-amino-acid pep-
tide, -KRREILSRRPSYR-, with sequence derived from CREB protein
was phosphorylated by the catalytic subunit of CAMP-dependent pro-
tein kinase
at
30 "C for
15
h in
a
reaction containing
50
mM
ws,
pH
7.4,
1.5
mM ATP,
10
n"
MgCI,, 1.2% 2-mercaptoethanol. The phosphory-
lated CREB peptide was purified by Sep-Pak C18 cartridge (Millipore)
and resuspended in water. Phosphorylation of the COOH-terminal ser-
ine by CAMP-dependent protein kinase makes the CREB petide a spe-
cific substrate for GSK-3.
Autophosphorylation of GSK-3 was accomplished by incubation
at
30 "C for the specified times with the buffer described above for the
peptide assay except that
25
p~
[Y-~~PIATP
(1000-5000
countdmid
pmol) was used. For the measurement of the phosphate incorporated
into GSK-3 through autophosphorylation, an aliquot of the reaction mix
was removed and added to SDS sample buffer (final composition: 62.5
mM
Tris,
pH
6.8,
2%
SDS,
10%
glycerol,
5%
2-mercaptoethanol, 0.25%
bromphenol blue). Phosphoproteins were then separated by SDS-PAGE
on 12% gels and detected by autoradiography. For quantitation, the
32P-labeled proteins were excised and the Cerenkov radiation counted.
For the measurement of GSK-3 kinase activity, another aliquot was
assayed for phospho-CREB peptide phosphorylation by adding to reac-
tion mix adjusted to give final phosphorylation conditions as described
in the preceding paragraph.
Protein Phosphatase Assay-Homogeneous recombinant bacterioph-
age A-protein phosphatase, Yersinia protein tyrosine phosphatase
Yop51, and rat brain protein tyrosine phosphatase 1B were kindly pro-
vided by Dr
J.
E. Dixon (University of Michigan). The catalytic subunit
of type-1 phosphatase was recombinant protein produced in
E.
coli4
The
type-2A catalytic subunit was isolated from rabbit muscle (DePaoli-
Roach,
1984).
Dephosphorylation of GSK-3 with protein phosphatases
was carried out
at
30 "C in buffer containing 30
n"
Tris,
pH
7.4,
0.1
mg/ml bovine serum albumin,
5%
glycerol, and
100
PM
MnCl, (except for
phosphatase 2A). For the dephosphorylation of autophosphorylated
GSK-3, protein phosphatases were added into the reaction mix contain-
ing GSK-30 autophosphorylated for 30 min as described above, and
incubated in the kinase reaction buffer plus
100
p~
MnCl, for the
specified times.
For
analysis
of
32P-labeled GSK-3 dephosphorylation,
SDS
sample buffer was added
at
the indicated time and the phospho-
proteins were separated by SDS-PAGE on
a
12% gel and detected by
autoradiography. For the evaluation of the effect of dephosphorylation
on GSK-3 kinase activity, the phosphatase reaction was terminated
at
the indicated times by the addition of
10
mM NaF (final concentration)
for
serine/threonine-phosphatase
and
100
J.IM
sodium orthovanadate
(final concentration) for protein-tyrosine phosphatase 1B.
An
aliquot of
the samples containing the dephosphorylated GSK-3 was removed and
assayed for its kinase activity using phospho-CREB as substrate. Con-
trol samples contained phosphatases and their corresponding inhibitors
throughout the incubation period.
Immunoblotting-Purified chicken polyclonal anti-GSK-3 peptide
antibodies and horseradish peroxidase-conjugated rabbit anti-chicken
IgG antibodies were kindly supplied by
Dr.
J.
C. Lawrence (Washington
University). The anti-GSK-3 antibodies were raised against a synthetic
peptide based on the sequence
85KKVLQDKRFKNRELQIMRKLD105
of
I.
Park and A. A. DePaoli-Roach, unpublished data.
GSK-3P which is identical in GSK-3a. Monoclonal anti-Tyr(P) antibody
(PY20) and peroxidase-conjugated polyclonal anti-mouse IgG antibod-
ies were purchased from ICN and Sigma, respectively. Protein samples
were resolved by SDS-PAGE and transferred to a PVDF membrane
(Immobilon-P, Millipore). The membranes were then incubated with
anti-GSK-3 or anti-Tyr(P) antibody PY20 and further probed with a
peroxidase-conjugated secondary antibody. Immunoreactive proteins
were revealed with the enhanced chemiluminescence detection system
(Amersham Corp.).
Other Materials and Methods-For phosphoamino acid analysis,
phosphorylated proteins were resolved by SDS-PAGE using 12% poly-
acrylamide gel and transferred
to
PVDF membranes. Pieces of mem-
brane containing 32P-labeled GSK-3 were excised and hydrolyzed in
5.7
N
HCl at 110 "C for 90 min. Phosphoamino acids were separated by
one-dimensional thin layer electrophoresis and detected by autoradiog-
raphy
as
described previously (Zioncheck
et
al.,
1986). Protein concen-
trations were determined by the Bradford
(1976)
protein assay using
bovine serum albumin
as
a
standard. [y-32P]ATP was from Dupont-New
England Nuclear. Chemicals and reagents for gel electrophoresis were
purchased form Bio-Rad. Restriction enzymes and other molecular bi-
ology reagents were obtained from Bethesda Research Laboratories.
RESULTS
Molecular Cloning, Expression, and Characterization
of
Rab-
bit Skeletal Muscle
GSK-3p"The GSK-3 used for most previ-
ous work in this laboratory was isolated from rabbit skeletal
muscle. This purification yields relatively small amounts of
purified protein due to its low abundance. To facilitate further
study of GSK-3, we sought to obtain
a
rabbit muscle cDNA
clone for GSK-3. Based on known peptide sequences derived
from rabbit skeletal muscle GSK-3 (Woodgett, 19901, we de-
signed degenerate sense and antisense oligonucleotides for
polymerase chain reaction amplification. A 580-bp fragment
was amplified from rabbit skeletal muscle cDNA and shown by
sequencing to encode a portion of GSK-3P. Using this fragment
as
a
probe to screen a rabbit skeletal muscle cDNA library, we
identified five positive clones, one of which was
a
full-length
GSK-3P cDNA and the others partial clones of GSK-3P. The
full-length cDNA contained an open reading frame of 1260
nucleotides (data not shown). This size is identical
to
that of the
rat brain GSK-3p cDNA and would encode
a
420-amino-acid
GSK-30 with
a
molecular mass of 46 kDa. At the nucleotide
level, the coding region of the rabbit skeletal muscle GSK-3P
shared 92% identity to
rat
brain GSK-3P. At the amino acid
level, these two GSK-3Ps differed in only 3 amino acids, all
close to the COOH terminus.
The GSK-BP coding region was inserted under the control of
the T7 410 promoter in the PET-3c vector of Studier
et
aE.
(1990)
as described under "Experimental Procedures."
An
isopropyl-
P-thiogalactoside-inducible
46-kDa protein in crude extracts of
E.
coli
cells transformed with the PET-GSK-3P expression vec-
tor
was
clearly seen in Coomassie Blue-stained gels. The pro-
tein was absent in extracts of bacteria carrying the control PET
plasmid without GSK-3 insert (data not shown). Similarly,
GSK-3 activity was detected in crude extracts of cells carrying
the PET-GSK-3P vector but not the control (data not shown).
The active GSK-3P was purified -180-fold as described un-
der "Experimental Procedures." Analysis
of
the GSK-3P by
SDS-PAGE indicated that
a
46-kDa polypeptide was the major
band on the gel, representing up
to
70%
of the protein in the
final preparation (Fig.
IA).
The apparent molecular weight
of
46,000 was close to that predicted from the sequence (Fig.
IA).
When an aliquot of the recombinant GSK-3P preparation was
analyzed by Western blotting with polyclonal anti-GSK-3 pep-
tide antibodies,
a
single immunoreactive protein of 46 kDa was
detected (Fig.
lB,
lane
3)
confirming that the iSOprOpYl-p-D-
thiogalactopyranoside-inducible
protein was recombinant rab-
bit skeletal muscle GSK-3P. We also analyzed
a
current prepa-
ration of GSK-3 purified from rabbit muscle. The antibodies
recognized a single dominant species with molecular mass 51
Glycogen Synthase Kinase
3
14569
A
B
kDa
kDa
94
-
67
-
45
-
..-
94
-
67
-
45
-
-8-
30
-
30
-
123
FIG.
1.
Expression
of GSK-Sf3
in
bacterial
cells.
Panel
A,
SDS-
PAGE of purified recombinant GSK-3p. Purified enzyme (400 ng) was
electrophoresed and the gel stained with Coomassie Blue.
Panel
B,
Western blot analysis of GSK-3.
An
aliquot
of
GSK-3a isolated from
rabbit skeletal muscle
(lane
1)
and GSK-3p (25 ng) expressed in
E.
coli
(lane 3)
were separated by gel electrophoresis, transfemed
to
a PVDF
membrane, and immunoblotted with chicken polyclonal anti-GSK-3 an-
tibodies as described under "Experimental Procedures." In
lane
2
a
mixture
of
GSK-3a and
p
was analyzed.
kDa (Fig.
lB,
lane
1
),
the predicted mass of
GSK-3a.
We con-
clude that the enzyme purified from rabbit muscle was pre-
dominantly
GSK-3a.
Recombinant
GSK-3P
was able
to
phosphorylate several pro-
teins previously identified as substrates for
GSK-3
(Plyte et al.,
1992).
Table I shows kinetic parameters for some of these sub-
strates. The CREB peptide, in keeping with earlier results:
was not detectably phosphorylated by
GSK-3P
unless previ-
ously phosphorylated with CAMP-dependent protein kinase
(Table
I).
Similarly, phosphorylation of recombinant rabbit
muscle glycogen synthase was totally dependent on prior phos-
phorylation by casein kinase I1 (data not shown), similar to the
recent results of Zhang et
al.
(1993)
with
GSK-3
isolated from
rabbit muscle.
We also examined the effects of various compounds on re-
combinant
GSK-3P
kinase activity using phospho-CREB pep-
tide as substrate. Among the compounds evaluated, heparin
(25
pg/ml) was found
to
cause stimulation of
GSK-3P
activity
to
160%
of the control (data not shown). The IC,, values for sev-
eral inhibitory compounds are shown in Table
11.
GSK-3p
pep-
tide phosphorylation activity was inhibited by NaC1, NaF, poly-
lysine, p-glycerol phosphate, and several divalent cations.
Several amino acid-specific reagents inhibited
GSK-3P
efi-
ciently. This result could implicate histidine, lysine, and argi-
nine residues as being important
for
GSK-3P
activity, but more
detailed work will be required to confirm this conclusion.
Autophosphorylation
of
Recombinant
GSK-3p
at
Ser,
Thr,
and
!&
Residues-Both rabbit skeletal muscle
GSK-3a
iso-
lated from the tissue and
GSK-3P
expressed in
E.
coli were able
to
catalyze an autophosphorylation reaction when incubated
with ATPMg2' (Fig.
2A
1.
Autophosphorylation caused a reduc-
tion in the electrophoretic mobility of
GSK-3a
as has been
reported previously (Hemmings et al.,
1981;
Tung and Reed,
1989),
whereas no effect on the mobility of the recombinant
GSK-3P
was observed (compare Figs.
1
and
2).
The stoichiom-
etry of the incorporation of radioactive phosphate into
GSK-3P
was approximately
1.2
mol of phosphatdmol kinase? Phos-
phoamino acid analysis of autophosphorylated
GSK-3a
and
GSK-3P
revealed the presence of Ser(P) and Thr(P) in both
Obviously, in
this
determination we measure only the additional
phosphate introduced over and above any phosphate already present in
the GSK-3.
TABLE
I
Substrate specificity
of
recombinant GSK-3P
for the phospho-CREB peptide using the P81
filter assay. Phosphoryla-
Phosphorylation of myelin basic protein by GSK-36 was analyzed as
tion of K-casein and phosvitin was carried out in a modified kinase
buffer with 0.5
m~
[y3*PJATP. Phosphorylated substrate was precipi-
tated onto filter paper washed with trichloroacetic acid as previously
described (Graves
et al.,
1993). Inhibitor-2 was incubated with GSK-3P
in the presence
of
0.125 mM ATP as previously described (Wang
et
al.,
1994). For quantitation, the 32P-labeled inhibitor-2 was run on a 12%
SDS-PAGE, excised from the gel, and the radioactivity quantitated by
Cerenkov counting.
Substrate
"In,
Km
nmollminlmg
Phospho-CREB peptide 2400 200
p?4
Myelin basic protein 80 59
w
K-Casein 106 114
p~
Phosvitin 140 5 mg/ml
Inhibitor-2 650 16
PM
TABLE
I1
Inhibitors
of
GSK-3p kinase activity
Different effectors of GSK-3p kinase were analyzed using the phos-
pho-CREB peptide as substrate as described under "Experimental Pro-
cedures." At least
six
different concentrations
of
each inhibitor were
examined. The IC,,
is
the concentration at which 50%
of
the control
protein kinase activity is inhibited.
Inhibitors
IC,
p-Glycerol phosphate 50 mM
Diethyl-pyrocarbonate 0.2
mM
Pyridoxal-5'-phosphate 0.6
m~
Phenylglyoxal 2 m~
Poly-lysine 0.4 mg/ml
NaCl
160
m~
NaF 35
mM
MnCl, 0.4
n"
CaCl, 4.5
mM
isoforms. Although quantitatively less,
a
significant amount
of
"(P)
was detected in
GSK-3P
(Fig.
2B).
This
Tyr
phospho-
rylation occurred on the same time scale as Ser/Thr phospho-
rylation since phosphoamino acid analysis at different times
indicated no great differences in the relative proportions of the
phosphoamino acids (data not shown). Replacement of Mg2'
with Mn2+ as divalent cation in the reaction buffer significantly
reduced the autophosphorylation rate of
GSK-3P
but did not
change the phosphoamino acid pattern (data not shown).
Also,
if the phospho-CREB peptide substrate was included in the
autophosphorylation reaction, the incorporation of phosphate
into
GSK-3P
was suppressed significantly (data not shown).
GSK-3a
was predominantly autophosphorylated on serine resi-
dues with a much smaller amount of ThdP) (Fig.
2B).
Anti-
Tyr(P)
antibodies were used to analyze for Tyr(P) in
GSK-3
preparations after SDS-PAGE and transfer
to
PVDF mem-
branes. Both
GSK-3a
and
GSK-3p
reacted with the anti-Tyr(P)
antibodies even before incubation with ATPMg2' (Fig.
20.
After
the autophosphorylation reaction, no great change in the
GSK-3a
signal was seen but there was a significant increase in
the signal with
GSK-3P,
consistent with the results of 32P-
labeling just described. This increase can be visually enhanced
if shorter exposures are made but the non-quantitative nature
of the chemiluminescence detection system makes
it
hard
to
judge exactly the relative amounts of
Tyr(P)
already present
versus what
is
added during the autophosphorylation reaction.
Effects
of
Autophosphorylation on
GSK-3
Activity-GSK-3
activity toward phospho-CREB peptide was measured as a
function of autophosphorylation (Fig.
3,
A
and
B).
There was a
time-dependent decrease in activity, to
a
level
of
20-35%
of the
starting value. The enzyme was stable over this period in the
absence of ATPMg2' (Fig.
3A),
and the amount of protein re-
14570
Glycogen Synthase Kinase
3
A
kDa
94
-
67
-
45
-
30
-
B
C
30
-
12 12 1234
FIG.
2.
Autophosphorylation
of
GSK-3.
Panel
A,
autophosphorylation of GSK-3a
(lane
1
)
and recombinant GSK-3p
(lane
2).
An
autoradio-
gram after SDS-PAGE is shown.
Panel
B,
phosphoamino acid analysis
of
autophosphorylated GSK-3a
(lane
1)
and GSK-3P
(lane
2).
The migration
of
standards
of
SedP)
(P-ser),
Thr(P)
(P-thr),
and TydP)
(P-tyr)
are indicated. A autoradiogram
of
the thin layer plate is shown.
Panel
C,
anti-TydP)
immunoblot. Equal amounts
of
rabbit muscle GSK-3a
(lanes
1
and
3)
and bacterially expressed GSK-3P
(lanes
2
and 4) proteins as judged by
Western blot using anti-GSK-3 antibodies were incubated in the kinase buffer with
(lanes
3
and
4)
or without
(lanes
1
and
2)
ATP/Mg2' at
30
"C
for
30 min. The samples were analyzed with monoclonal anti-wp) antibody as described.
mained constant as judged by Western blotting (Fig. 3C). In
parallel experiments, GSK-3a isolated from skeletal muscle
was also tested
for
effects
of
autophosphorylation on activity
(Fig.
4,
A and
B).
With this enzyme, the activity remained
constant in the presence
of
ATPMg2' (Fig.
4A
1.
However, there
was significant loss of protein kinase activity in the control
incubated without ATP/Mg2* (Fig.
4A).
In separate experi-
ments, decreased activity toward another substrate, phospha-
tase inhibitor-2, was also observed after autophosphorylation of
GSK-3P (data not shown).
Bacterially expressed GSK-3P was relatively stable during
and after purification. However, in one preparation of the re-
combinant GSK-3/3, several of the protease inhibitors (see "Ex-
perimental Procedures") were omitted from the buffers, and the
resulting preparation contained a predominant species of
40
kDa whose generation we attributed
to
partial proteolysis. This
proteolyzed fragment of GSK-3P was formed gradually during
the purification (data not shown). The fragment was enzymati-
cally active. Indeed, when the amounts of the truncated and the
intact forms
of
GSK-3P were normalized by their ability
to
phosphorylate phospho-CREB peptide (Fig.
5A
),6
SDS-PAGE
and immunoblotting with anti-GSK-3 antibodies revealed that
the intact GSK-3P sample contained much more immunoreac-
tive protein than the truncated form (Fig.
5B).
Thus,
the pro-
teolytic fragment had a significantly higher specific activity
than the intact form, by as much as 10-fold. The truncated
enzyme was also severely impaired in its ability
to
autophos-
phorylate (Fig.
5C),
and similar results were obtained if the
wild-type and truncated enzymes were incubated at normal-
ized concentrations rather than activities (not shown). Consist-
ent with these observations, incubation with ATPMg2' did not
correlate with any significant decrease in activity (Fig.
6).
In
preliminary studies, trypsin treatment of the autophospho-
rylated GSK-3P also resulted in increased protein kinase ac-
tivity (data not shown).
Opposing Effects
of
fir
and SerlThr Phosphorylation on
GSK-3P
Activity-It was important
to
establish that the inac-
tivation associated with the autophosphorylation
of
GSK-3P
was reversible and not in some way due
to
permanent inacti-
vation of the enzyme. Initial efforts
to
remove phosphorylated
Ser and Thr residues from autophosphorylated GSK-3P uti-
The reason that the reaction rate remains close
to
constant and does
not diminish over the course
of
the incubations in Fig.
3A
is that the
presence
of
the peptide substrate strongly suppresses the autophospho-
rylation.
lized several different phosphatases including rabbit muscle
type
1
and type 2A phosphatase catalytic subunits as well as
recombinant A-phage phosphatase
(zhuo
et al., 1993) which
acts on Tyr(P) as well as Ser(P)/Thr(P). Of these enzymes, type
2A was active toward GSK-3 but by far the most effective en-
zyme was the A-phosphatase which removed a significant
amount of the 32P label associated with GSK-3P (Fig. 7A, lane
2).
Phosphoamino acid analysis indicated that the A-phospha-
tase primarily dephosphorylated SedP) and ThdP), leaving
Tyr(P) as the dominant phosphoamino acid (Fig. 7B, lane
1
).
In
efforts to remove the Tyr(P), both the Yersinia Yop51 and the
mammalian protein
Tyr
phosphatases 1B were tested. The
Yersinia enzyme, under our conditions, was ineffective (data
not shown) but PTP 1B catalyzed dephosphorylation of GSK-3P
(Fig. 7A, lane
3).
Because of the relatively small proportion
of
Tyr(P), the dephosphorylation was only clearly visible
from
phosphoamino acid analysis (Fig. 7B, lane
3).
The effect of the
&phosphatase and PTP 1B on the protein kinase activity of
autophosphorylated GSK-3P was also examined.
As
shown in
Fig. 7C, treatment of GSK-3P that had been allowed
to
auto-
phosphorylate for 30 min with A-phosphatase resulted in a
significant restoration of the kinase activity, whereas tyrosine
phosphatase incubation caused a further loss of the kinase
activity as compared with the untreated control sample.
As
noted earlier, the GSK-3P contains Tyr(P) as judged by reac-
tivity with anti-phosphotyrosine antibodies. It could also con-
tain SedP) and/or ThdP), but we have no easy way
to
test this.
The possibility exists that the phosphate already contained in
the GSK-3P might affect its autophosphorylation characteris-
tics. GSK-3P that had first been treated with PTP 1B under the
conditions of Fig. 7 showed a qualitatively similar inactivation
by autophosphorylation as untreated enzyme (data not shown).
In other experiments, recombinant GSK-3P with an NH,-ter-
minal polyHis sequence was incubated with A-phosphatase and
purified via a Ni column prior
to
incubation with ATP and
Me.
A
similar inactivation
to
that recorded in Fig.
3A
was ob-
served.' We conclude that the presence of pre-existing phos-
phate in the GSK-3P does not alter the in vitro autophospho-
rylation, at least in any major way.
DISCUSSION
We report here that GSK-3P is a dual specificity protein
kinase that is differentially regulated by Ser/Thr and
Tyr
phos-
Q.
M.
Wang and
P.
J.
Roach, unpublished data.
Glycogen Synthase Kinase
3
14571
0
10
20
30
40
50 60 70
time
(min)
-
-
m
4
46kDa
2345678
w
0
m
o
0
a
0
46kDa
2345678
B
C
FIG.
3.
Regulation
of
recombinant
GSK-Sf3
kinase activity by
autophosphorylation.
Purified recombinant GSK-3P was subjected
to
autophosphorylation as described under “Experimental Procedures.”
Aliquots from the reaction were removed at the following times (rnin):
0
(lane
1
),
2
(lane
2),
6
(lane
3),
10
(lane
4),
20
(lane
5),
30
(lane
6),
45
(lane
7),
and
60
(lane
8).
The samples, containing
25
ng of GSK-3P,
were analyzed by autoradiography for phosphoproteins
(panel
B),
by
immunoblotting using anti-GSK-3 peptide antibodies
(panel
C),
or
were
analyzed for their ability
to
phosphorylate phospho-CREB peptide
(panel
A
).
Activity
is
expressed as a percentage of the first time point in
which the peptide phosphorylation assay was performed immediately
after the GSK-3 had been mixed with cold autophosphorylation buffer
(A,
with ATP/Mg2‘,
0,
without ATP/Mg2‘).
phorylation. Mammalian GSK-3 has thus far been considered
to
be
a
SerPrhr-specific protein kinase based on
its
amino acid
sequence and
its
phosphorylation of exogeneous substrates, but
our results suggest that GSK-3P
is
capable of catalyzing auto-
phosphorylation on
Tyr
as well as on SerPrhr. This reaction was
detected by
32P
incorporation
in vitro
with purified enzyme.
However, the protein already contained some Tyr(P) prior
to
incubation with ATP
as
judged by ita reaction with anti-Tyr(P)
antibody. We are unable to judge whether the enzyme also
contains Ser(P) and/or Thr(P). Since
E.
coli
is
unlikely
to
con-
tain protein
Tyr
kinase activity, we conclude that the
Tyr
phos-
phorylation was the result of autophosphorylation of GSK-3P
within the
E.
coli
cells. This
is
the
first
demonstration that
mammalian GSK-3
is
a dual specificity protein kinase, in the
same sense as
MAP
kinase (Wu
et al.,
1991; Posada and Cooper,
1992). It is interesting, though, that one of the yeast homologs
of GSK-3, MCK1, was shown
to
be associated with protein
Tyr
kinase activity although
it
was impossible at the time
to
say
definitively that MCKl was itself the enzyme responsible (Dai-
ley
et al.,
1990). Similar
to
the
MAP
kinase family, however,
GSK-3 has not been shown previously
to
modify
Tyr
residues in
exogenous substrates (Plyte
et al.,
1992), and in recent studies
where we have been especially alert
to
the possibility we ob-
0
20
40
60
time
(rnin)
4-
51
kDa
12345678
FIG.
4.
Effect
of
autophosphorylation
on
GSK-Sa
activity.
Au-
tophosphorylation time course of GSK-3a isolated form skeletal muscle.
The time points were the same as indicated in legend to Fig. 3.
An
autoradiogram after SDS-PAGE
is
shown
(panel
B).
Panel
A,
activity as
a function of phosphorylation. In a parallel experiment, the assays
of
muscle GSK-3a were performed exactly as described for recombinant
GSK-3p in the legend to Fig. 3
(A,
with ATP/Mg2‘;
0,
without ATP/
Me).
served no evidence for Tyr(P) in
CREB,
inhibitor-2,
or
myelin
basic protein after phosphorylation by GSK-3.7
In contrast to our results, Hughes
et al.
(1993) could not
detect autophosphorylation on
Tyr
by either GSK-3a
or
GSK-3P even though these authors had discovered that their
GSK-3 contained
Tyr(P).
Their enzymes, however, were either
purified from animal tissues
or
expressed in insect cells. The
most likely explanation for the apparent discrepancy lies in the
fact that our recombinant GSK-3P was produced in
E.
coli.
Presumably, the enzymes isolated from eukaryotic cells were
already fully modified on the relevant
Tyr
residue(s) whereas
enzyme expressed in bacteria was not. In fact, we partially
confirmed the results of Hughes
et
al.
(1993) since our GSK-3a,
enzyme purified from muscle, also did not autophosphorylate
on
Tyr
residues.
As
noted, GSK-3a behaved differently from GSK-3P in auto-
phosphorylation reactions
in vitro.
Besides the lack of
Tyr
phos-
phorylation, the autophosphorylation of GSK-3a was not inac-
tivating. If anything, the presence of ATP and Mg2‘ stabilized
the GSK-3a activity compared with a control lacking ATP, in
which the activity decreased with time of incubation. However,
whether this stabilization can be attributed
to
the occurrence of
autophosphorylation
or
simply
to
the presence ofATP and Mg2‘
cannot be distinguished. The GSK-3a did contain Tyr(P) as
judged by
its
interaction with anti-Tyr(P) antibodies. At this
time, we
do
not know whether these differences in the behavior
of GSK-3a and GSK-3P reflect specific properties of the differ-
ent isoforms
or
the fact that one was produced in
E.
coli
and the
other was isolated from mammalian cells. Efforts are underway
to
clone and express the rabbit skeletal muscle GSK-3a isoform
in
E.
coli
to
address this issue.
Within the protein kinase family, GSK-3’s nearest neighbors,
based on sequence alignments of the catalytic domain, include
the cdc2 and
MAP
kinase/ERK families (Hanks and Quinn,
1991). Recent biochemical observations are consistent with this
14572
Glycogen Synthase Kinase
3
A
B
kDa
94
-
67
-
45
-
0
30
-
12
time
(rnin)
C
kDa
94
-
67
-
45
-
30-
0
12
FIG.
5.
Activity
of
truncated
GSK-Sp.
Panel
A,
phospho-CREB
peptide phosphorylation by recombinant GSK-3P and
its
proteolytic
fragment. The amounts of the two preparations were normalized by
their peptide phosphorylation activity
(0,
GSK-3 intact form;
0,
pro-
teolytic fragment).
Panel
B,
immunblotting of GSK-3p intact form
(lane
1)
and the proteolytic fragment
(lane
2)
using anti-GSK-3 antibodies.
Panel
C, autophosphorylation of GSK-3P
(lane
1)
and
its
proteolytic
form
(lane
2).
The enzymes were analyzed by autoradiography after
SDS-PAGE. Equal amounts
(40
milliunits) of the intact and the trun-
cated GSK-SP, normalized on the basis of their peptide phosphorylation
activities, were analyzed by immunoblotting, or for incubation
with
100
p~
[-y-32P]ATP
in
the autophosphorylation reaction. One unit of GSK-3
activity is defined as the amount of enzyme
that
transfers
1
nmol of
phosphatdmin into the phospho-CREB peptide.
classification. Hughes
et
al.
(1993) were,
as
noted, the
first
to
detect Tyr(P) in GSK-3 and furthermore identified the
site
of
modification as Tyr-216 of GSK-3p which corresponds
to
the
regulatory "TEY" sequence of the
MAP
kinase/ERK family of
enzymes. Phosphorylation at this residue, which
is
conserved
in all
known
GSK-3-like enzymes, appears to be activating and,
although we have not yet mapped the site of
Tyr
autophospho-
rylation in GSK-3p,
it
is
possible that the same
site
is
involved.
Our observation that GSK-3p can be classified as
a
dual speci-
ficity enzyme makes another parallel with enzymes of the
MAP
kinase/ERK class. One
of
the distinguishing sequences for this
group of enzymes
is
the "Y-R-A-P-E" motif in the conserved
kinase sub-domain VIII. GSK-3 shares
a
slightly more ex-
tended similarity
to
MAP
kinase and STY/clk, another dual
specificity enzyme (Howell
et
al.,
1991; Wu
et
al.,
19911, in the
sequence R-X-Y-R-A-P-E (where X
is
aromatic). This region of
the molecule in GSK-3, cdc2, and
MAP
kinase/ERK
is
just
COOH-terminal of regulatory phosphorylations (Gould
et
al.,
01
0
10
20
30
40
time
(min)
intact and the truncated
GSK-3p
activity.
Equal amounts of the
FIG.
6.
Comparison
of
the effect of autophosphorylation
on
the
two forms of recombinant GSK-3P (based on activity toward phospho-
CREB peptide) were incubated with
100
p~
ATP for autophosphoryla-
tion, and aliquots of the reaction (containing
4
milliunits of kinase) were
removed
at
the indicated times and tested for phospho-CREB phospho-
rylation activities
(0,
truncated enzyme; A, intact enzyme). Activity is
expressed as described in the legend to Fig.
3.
1991; Wu
et
al.,
1991; Posada and Cooper, 1992). One final
feature that may be common to these three enzyme families has
to
do with substrate specificity. Although these enzymes almost
certainly phosphorylate distinct substrates
in vivo,
the se-
quences surrounding the sites phosphorylated often include the
motif SR-P, usually together with other determinants such as a
basic residue for cdc2 (Kemp and Pearson, 1990), another
NH,-
terminal
Pro
for
MAP
kinase (Clark-Lewis
et
al.,
1991; Davis,
1993) and
a
COOH-terminal phosphate group for GSK-3
(Roach, 1991).
A key question for GSK-3
is
how the enzyme
is
regulated
in
vivo.
Phosphorylation of
Tyr
would appear
to
correlate with
activation whereas modification of Serfl'hr residues, by auto-
phosphorylation, by protein kinase C (Goode
et
al.,
19921,
or
by
S6 kinase (Sutherland
et
al.
1993; Sutherland and Cohen,
1994) inactivates. Not all SerPrhr sites have been identified
although the fact that
a
truncated GSK-3p was both 10-fold
more active and impaired in autophosphorylation
is
consistent
with the loss of
a
phosphorylated inhibitory domain. The trun-
cation, by
-6
kDa as judged by SDS-PAGE, could be accommo-
dated
at
either end of the GSK-3p molecule without excising
conserved kinase sequences. We have identified one major
phosphopeptide from the full-length autophosphorylated GSK-
3p; this
is
at the
NH,
terminus and contained both serine and
threonine residues?
It
has also been suggested that
sites
for
protein kinase C, which acts only on the GSK-3p isoform, are
located
NH,
terminally (Goode
et
al.,
1992). Finally, phospho-
rylation of GSK-3 by S6 kinase apparently occurs at the NH,
terminus (Sutherland
et
al.
1993; Sutherland and Cohen,
1994). Although the information
is
not complete,
it
is
sugges-
tive that inactivating SerPrhr autophosphorylations occur at
the
NH,
terminus of the molecule. In surveying protein kinases
regulated by phosphorylation,
it
is
interesting that phospho-
rylation
is
only quite rarely inactivating. One prime example
is
of course the NH,-terminal phosphorylations of the
Tyf
and Thr
residues in ~34'~'~ (Norbury
et
al.,
1991; Krek and Nigg, 1991).
Another well
known
case
is
that of p6WSm and related enzymes
in which
a
COOH-terminal
Tyr
phosphorylation site
is
inacti-
vating (reviewed by Cooper and Howell, 1993).
Since both activating and inhibitory phosphorylations can
occur in GSK-3, the possibility exists for differential control by
protein phosphatases, depending on whether
Tyr
or
SerPrhr
residues are involved. Neither the type
1
nor type 2A phos-
phatases were particularly effective in dephosphorylating
at
SerPrhr. Perhaps some regulatory or targetting component was
lacking or else some other type of phosphatase might act on
GSK-3.
It
is
of interest that the
MAP
kinase phosphatases
appear
to
be enzymes with rather narrow substrate specifici-
ties compared with the protein Ser/Thr phosphatases (Zheng
Glycogen Synthase Kinase
3
14573
A
kDa
94
-
67
-
45
-
30
-
B
123
123
C
ISO-,
0-
time
(rnin)
FIG.
7.
Dephosphorylation
of
autophosphorylated
GSK-Sf3.
Panel A, treatment
of
autophosphorylated GSK-3p
(lune
1)
with A-phosphatase
(lane
2)
and protein tyrosine phosphatase
1B
(lune
3).
Recombinant GSK-3p was autophosphorylated at 30
"C
for
30 min and then was treated
with the phosphatases at 30
"C
for another
40
min.
An
autoradiogram after SDS-PAGE is shown.
Panel
B,
phosphoamino acid analysis
of
32P-labeled GSK-3P proteins treated with phosphatases.
Lane
1,
by A-phosphatase; lane
2,
no treatment;
lune
3,
by tyrosine phosphatase
1B.
Panel
C,
effect
of
dephosphorylation
of
GSK-3P on protein kinase activity. The autophosphorylated GSK-3P was treated with different phosphatases
for
40
min and then tested
for
peptide phosphorylation activities as described.
and Guan, 1993b; Sun
et
al.,
1993).
Another question, arising from the data presented here,
is
whether autophosphorylation of GSK-3 has any physiological
significance. Most protein kinases autophosphorylate and
changes in activity often ensue (Krebs, 1986). However,
a
physiological role for autophosphorylations as
a
regulatory
mechanism
is
largely unclear with the important exception of
the receptor-tyrosine kinase family (Fantl
et
al.,
1993).
As
men-
tioned in the Introduction, in initial studies of
MAP
kinase
control, one mechanism proposed was that an activator might
exist that could stimulate the autophosphorylation and hence
increase protein kinase activity (Seger
et
al.,
1991).
This
pro-
posal lost favor when
bona
fide
MAP
kinase kinases were found
(Gomez and Cohen, 1991;
Ahn
et
al.,
19911, but in principle
there
is
no reason why such mechanisms could not occur.
Hughes
et
al.
(1993) had inferred that
a
separate protein
Tyr
kinase must be involved in modification of Tyr-216 of GSK-3P
although, as noted, they were unaware of the possibility for
autophosphorylation at
Tyr
residues. We did test one obvious
candidate protein kinase,
MEK2
(Zheng and Guan, 19931, an
activator of
MAP
kinase, and found that
it
was unable to phos-
phorylate GSK-3P.8 Similar questions can be posed of the Serl
"hr
phosphorylations although an
inactivating
autophospho-
rylation would set up
a
different and rather interesting kind of
regulatory system.
A
protein kinase that constitutively turned
itself off by autophosphorylation would tend
to
dampen any
activation caused by dephosphorylation and would provide for
a
built-in feed-back inhibition.
An
important future goal will be
to
establish the roles of the different phosphorylations in de-
termining GSK-3 function.
Acknowledgments-We are indebted
to
John
C.
Lawrence Jr (Wash-
ington University) for supplying
us
with anti-GSK-3 antibodies,
to
Jack
E.
Dixon (University
of
Michigan)
for
Yersiniu, A-phage and
PTP
1B
protein phosphatases, and
to
Kun-Liang Guan (University
of
Michigan)
for
providing a sample
of
the MEK2 protein kinase.
REFERENCES
Ahn, N. G., Seger,
R.,
Bratlien,
R.
L.,
Diltz, C. D.,
'lbnks,
N.
K,
and Krebs, E. G.
Anderson, N. G., Maller, J.
L.,
'lbnks, N.
K,
and Sturgill,
T.
W.
(1990)
Nature
343,
Boyle, W. J., Smeal,
T.,
Defize,
L.
H.
K,
Angel, P., Woodgett, J.
R.,
Karin, M., and
Bradford, M. M.
(1976)
Anal.
Biochem.
72,248-254
Clark-Lewis,
I.,
Sanghera, J.
S.,
and Pelech,
S.
L.(1991)J.
Biol.
Chem.
266,15180-
(1991)
J.
Biol.
Chem.
266.42204227
651-653
Hunter,
T.
(1991)
Cell
64,573-584
15184
Q.
M.
Wang, K-L. Guan, and
P.
J. Roach, unpublished data.
Cohen, P.
(1986)
in
The
Enzymes
(Boyer, P. D., and Kreb, E. G., eds) Vol.
17A,
pp.
Cooper, J. A,, and Howell, B.
(1993)
Cell
73,1051-1054
Crews, C. M., Alessandrini, A. A,, and Erikson,
R.
L.
(1991)
Pm.
Natl.
Acad. Sei.
Crews, C. M., Alessandrini,
A.
A., and Erikson,
R.
L.
(1992)
Science
268,478-480
Dailey, D., Schieven, G.
L.,
Lim, M.-Y., Marquardt, H., Gilmore,
T.,
Thorner, J., and
Davis,
R.
J.
(1993)
J.
BioL.
Chem.
268,14553-14556
de Groot,
R.
P., Auwerx, J., Bourous, M., and Sassone-Corsi, P.
(1992)
Oncogene
7,
de Groot,
R.
P., den Hertog,
J.,
Vandenheede, J.
R.,
Goris, J., and Sassone-Corsi, P.
Dent, P., Campbell, D. G., Hubbard, M. J., and Cohen, P.
(1989)
FEBS
Lett.
248,
DePaoli-Roach, A. A.
(1984)
J.
Biol.
Chem.
269,12144-12152
Edelman, A. M., Blumenthal, D.
IC,
and Krebs, E. G.
(1987)Annu.
Rev.
Binchem.
Fantl, W., Johnson, D. E., and Williams, L.
T.
(1993) Annu.
Rea
Binchem.
62,
Fiol,
C.
J., Haseman, J.
H.,
Wang,
Y.,
Roach, P.
J.,
Roeske,
R.
W., Kowalczuk, M.,
Fiol,
C.
J., Wang, A., Roeske,
R.
W., and Roach, P. J.
(1990)
J.
BWZ.
Chem.
266,
Gomez, N., and Cohen, P.
(1991)
Nature
363, 170-173
Goode, N., Hughes,
IC,
Woodgett,
J.
R., and Parker, P. J.
(1992)
J.
Biol.
Chem.
287,
Gould,
K.
L.,
Moreno,
S.,
Owen, D. J., Sazer,
S.,
and Nurse, P.
(1991)
EMBO
J.
10,
Graves, P.
R.,
Hass, D. W., Hagedorn,
C.
H., DePaoli-Roach, A. A., and Roach, P.
J.
Guan,
IC-L.,
and Dixon, J.
E.
(1991)
J.
Biol.
Chem.
286,17026-17030
Hanger, D. P., Hughes,
K,
Woodgett, J.
R.,
Brion, J.-P., andAnderton, B. H.
(1992)
Hanks,
S.
K,
and Quinn, A. M.
(1991)
Methods Enzymol.
200.38-62
Hanks,
S.
IC,
Quinn, A. M., and Hunter,
T.
(1988)
Science
241,42-52
Hemmings, B. A,, Yellowlees, D., Kernohan, J. C., and Cohen, P
(1981)
Eur
J.
Hemmings, B. A,, Aitken,
A.,
Cohen,
P.,
Rymond, M., and Hofmann,
E
(1982a)Eur.
Hemmings, B. A,, Resink,
T.
J., and Cohen, P.
(1982b)
FEBS
Lett.
160,319-324
Howell, B.
W.,
Afar, D. E. H., Lew, J., Douville,
E.
M. J., Icely, P.
L.
E.,
Gray, D. A.,
Hughes,
IC,
Pulverer, B. J., Theocharous, P., and Woodgett, J.
R.
(1991)
Eur.
J.
Hughes,
K.
Ramakrishna.
S.,
Benjamin, W. B., and Woodgett,
J.
R.
(1992)
Bio-
461-497,
Academic Press, San Diego, CA
U.
S.
A.
88,8845-8849
Martin, G.
S.
(1990)
Mol. Cell. Biol.
10, 6244-6256
841-847
(1993)
EMBO
J.
12,3903-3911
61-12
56,567-613
453-481
and DePaoli-Roach, A. A.
(1988)
Arch. Biochem. Biophys.
267,797-802
6061-6065
16878-16882
32974309
(1993)
J.
Biol.
Chem.
268,6394-6401
Neurosic.
Lett.
147, 58-62
Biochem.
119,443451
J.
Biochem.
127,473481
and Bell,
J.
C.
(1991)
Mol. Cell. BioL
11,
566-572
Biochern.
203,305-311
;hem.
J..266,309-314
Hughes,
K,
Nikolakaki,
E.,
Plyte,
S.
E., 'lbtty, N. F., and Woodgett, J.
R.
(1993)
Hunter,
T.
(1991)
Methods Enzyml.
200,3437
Ishiguro,
K,
Shiratsuchi, A., Sato,
S.,
Omori, A., Arioka,
M.,
Kobayashi,
S.,
Uchida,
Kemp, B. E., and Pearson,
R.
B.
(1990)
Den& Biochem.
Sci.
16,342-346
Krebs, E. G.
(1986)
in
The
Enzymes
(Boyer, P. D., and Krebs, E. G., eds)
Vol17,
pp.
Krek, W., and Nigg, E. A.
(1991)
EMBO
J.
10,3331-3341
Lindberg,
R.
A,, Quinn, A. M., and Hunter,
T.
(1992)
"kends
Biochem.
Sci.
17,
Mandelkow, E."., Drewes,
G.,
Biernat, J., Gustke, N., Van Lint, J., Vandenheede,
Neigeborn, L., and Mitchell, A. P.
(1991)
Genes
&
Deu.
5,533-548
Nikolakaki, E., Coffer P. J., Hemelsoet,
R.,
Woodgett, J. R., and Defize, L.
H.
K
EMBO
J.
12,803-808
T.,
and Imahori,
IC
(1993)
FEBS
Lett.
326,167-172
3-20,
Academic Press, San Diego
114-119
J.
R.,
and Mandelkow, E.
(1992)
FEBS
Lett.
314,315-321
(1993)
Oncogene
8,833-840
14574
Glycogen Synthase Kinase
3
Nishida,
E.,
and Gotoh, Y.
(1993)
Dends
Biochem.
Sci.
18, 128-131
Norbury, C., Blow,
J.,
and Nurse,
P.
(1991)
EMBO
J.
10,33213329
Plyte,
S.
E.,
Hughes, K., Nikolakaki, E., Pulverer,
B.
J.,
and Woodgett,
J.
R.
(1992)
Posada,
J.,
and Cooper,
J.
A.
(1992)
Mol.
Biol.
Cell
3,
583492
Puziss,
J.
W., Hardy, T.
A,,
Johnson, R. B., Roach,
P.
J.,
and Hieter,
P.
(1994)
Mol.
Ramakrishna,
S.,
and Benjamin, W. B.
(1988)
J.
Biol.
Chem.
263, 12677-12681
Roach, P.
J.
(1986)
FASEB
J.
4,2961-2968
Roach,
P.
J.
(1991)
J.
Biol.
Chem.
266,14139-14142
Ruel, L., Bourouis, M., Heitzler, P., Pantesco, V., and Simpson,
P.
(1993a)
Nature
362, 557-560
Ruel,
L.,
Pantesco,
V,
Lutz,Y., Simpson, P., and Bourouis, M.
(1993b)
EMBO
J.
12,
1657-1669
Sambrook,
J.,
Fritsch, E. F., and Maniatis, T.
(1989)
Molecular Cloning: a Labo-
ratory Manual,
Cold Spring Harbor laboratory, Cold Spring Harbor,
NY
Sanger, F., Nicklen,
S.,
and Coulson,
A.
R.
(1977)
Proc. Natl. Acad.
Sci.
U.
S.
A.
74,
5463-5467
Seger, R., Ahn, N. G., Boulton, T. G., Yancopoulos, G. D., Panayatotos, N., Radz-
iejewska, E., Erickson,
L.,
Bratlien, R. L., Cobb, M.
H.,
and Krebs,
E.
G.
(1991)
Proc. Natl. Acad.
Sci.
U.
S.
A.
88,
6142-6146
Biochem.
Biophys.
Acta
1114, 147-162
Cell.
Biol.
14, 831-839
Shero,
J.
H., and Hieter,
P.
(1991)
Genes
&
Deu.
5, 549-560
Siegfried, E., Chou, T.-B., and Perrimon, N.
(1992)
Cell
71,1167-1179
Studier,
F.
W., Rosenberg,
A.
H., Dunn,
J. J.,
and DubendorIT,
J.
W.
(1990)
Methods
Sun,
H., Charles, C. H., Lau, L. F., and Tonks, N.
K.
(1993)
Cell
75,487-493
Sutherland, C. and Cohen,
P.
(1994)
FEBS Lett.
338,3742
Enzymol.
185, 60-89
Sutherland,
C.,
Leighton,
I.
A,,
and Cohen,
P.
(1993)
Biochem.
J.
296, 15-19
Tung, H. Y., and Reed, L.
J.
(1989)
J.
Biol.
Chem.
264,2985-2990
Villa-Moruzzi, E.
(1989)
FEBS Lett.
258, 208-210
Wang, Q. M., Park,
I.-K,
Fiol, C.
J.,
Roach,
P.
J.,
and DePaoli-Roach,
A.
A.
(1994)
Welsh,
G.
I.,
and Proud, C. G.
(1993)
Biochem.
J.
294,625629
Woodgett,
J.
R.
(1990)
EMBO
J.
9, 2431-2438
Woodgett,
J.
R.
(1991)
Pends Biochem. Sci.
16, 177-181
Wu, J., Rossomando,
A.
J.,
Her, J.-H., Vecchio, R.
D.,
Weber, M.
J.,
and Sturgill,
T.
W.
(1991)
Proc. Natl. Acad.
Sci.
U.
S.
A.
88,9508-9512
Wu,
J.,
Harrison,
J.
K., Vincent, L.
A,,
Haystead, C., Haystead, T.
A.
J.,
Michel, H.,
Hunt,
D.
E,
Lynch, K. R., and Sturgill,
T.
W.
(1993)
Proc. Natl. Acad.
Sci.
Yang, S.-D., Ha, L.-T., and Fung, T:J.
(1988)Biochem.
Biophys.
Res.
Commun.
151,
U.
S.
A.
.
90, 173-177
Zhang, W., Browner, M. F., Fletterick, R.
J.,
DePaoli-Roach,
A. A.,
and Roach,
P.
J.
61-69
Zhang, W., DePaoli-Roach,
A.
A,,
and Roach,
P.
J.
(1993)
Arch. Biochem Biophys.
(1989)
FASEB
J.
3,2532-2536
Zhang, Z.-Y., Clemens,
J.
C., Schubert, H.I., Stuckey,
J.
A.,
Fisher, M. W. F., Hume,
304,219-225
Zheng, C.-E, and Guan, K.-L.
(1993a)
J.
Biol.
Chem.
268, 11435-11439
D.
M., and Dixon,
J.
E.
(1992)
J.
Biol.
Chem.
267,23759-23766
Zheng, C.-F., and Guan, K.-L.
(1993b)
J.
Biol.
Chem.
268, 16116-16119
Zhuo,
S.,
Clemens,
J.
C., Hakes, D.
J.,
Barford D., and Dixon,
J.
E.
(1993)
J.
€501.
Zioncheck,
T.
F., Harrison, ML., and Geahlen, R. L.
(1986)
J.
Biol.
Chem.
261,
Biochemistry
33,
143-147
Chem.
268, 17754-17761
15637-15643