Báo cáo khoa học: Insights into the activation of brain serine racemase by the multi-PDZ domain glutamate receptor interacting protein, divalent cations and ATP ppt

Insights into the activation of brain serine racemase by the
multi-PDZ domain glutamate receptor interacting protein,
divalent cations and ATP
Florian Baumgart
1
, Jose
´
M. Manchen
˜
o
2
and Ignacio Rodrı
´
guez-Crespo
1
1 Departamento de Bioquı
´
mica y Biologı
´
a Molecular, Facultad de Ciencias Quı
´
micas, Universidad Complutense de Madrid, Spain
2 Grupo de Cristalografia Macromolecular y Biologı
´
a Estructural, Instituto Rocasolano, CSIC, Madrid, Spain
Serine racemase (SR) is a pyridoxal phosphate con-
taining enzyme that catal yses th e conversion of the
naturally occurring amino a cid l-serine into its race-
mic counterpart d-serine [1,2]. In brain tissues,
d-serine occupies the so-called ‘glycine site’ withi n
the NMDA receptor together with the neurotrans-

mitte r glutamate. Initially, Wolosker and coworkers
performed elegant studies that resulted in the mole-
cular cloning of the enzyme [3] and determined that
it was essentially expressed in astrocytes [4,5]. The
observation t hat certain neurones displayed signifi-
cant levels of d-serine [6] was confirmed when the
expression of SR was detected in neuronal cells as
well [7,8].
Keywords
calcium activation;
D-serine; GRIP; PDZ
domain; serine racemase
Correspondence
I. Rodrı
´
guez-Crespo, Departamento de
Bioquı
´
mica y Biologı
´
a Molecular, Facultad de
Ciencias Quı
´
micas, Universidad
Complutense, 28040 Madrid, Spain
Fax: +34 91 394 4159
Tel: +34 91 394 4137
E-mail:
(Received 7 June 2007, revised 6 July 2007,
accepted 11 July 2007)

doi:10.1111/j.1742-4658.2007.05986.x
Brain serine racemase contains pyridoxal phosphate as a prosthetic group
and is known to become activated by divalent cations such as Ca
2+
and
Mg
2+
, as well as by ATP and ADP. In vivo, brain serine racemase is also
activated by a multi-PSD-95 ⁄ discs large ⁄ ZO-1 (PDZ) domain glutamate
receptor interacting protein (GRIP) that is usually coupled to the GluR2 ⁄ 3
subunits of the a-amino-3-hydroxy-5-methylisoxazole-4-propionic acid
Ca
2+
channel. In the p resent study, we analysed the mechanisms b y which ser-
ine racemase becomes activated by GRIP, divalent cations and ATP. We
show that binding of PDZ6 of GRIP to serine racemase does not result in
increased d-serine production. However, full-length GRIP does augment
significantly enzymatic activity. We expressed various GRIP shorter con-
structs to map down the regions within GRIP that are necessary for serine
racemase activation. We observed that, whereas recombinant proteins con-
taining PDZ4-PDZ5-PDZ6 are unable to activate serine racemase, other
constructs containing PDZ4-PDZ5-PDZ6-GAP2-PDZ7 significantly aug-
ment its activity. Hence, activation of serine racemase by GRIP is not a
direct consequence of the translocation towards the calcium channel but
rather a likely conformational change induced by GRIP on serine race-
mase. On the other hand, the observed activation of serine racemase by
divalent cations has been assumed to be a side-effect associated with ATP
binding, which is known to form a complex with Mg
2+
ions. Because no

mammalian serine racemase has yet been crystallized, we used molecular
modelling based on yeast and bacterial homologs to demonstrate that the
binding sites for Ca
2+
, ATP and the PDZ6 domain of GRIP are spatially
separated and modulate the enzyme through distinct mechanisms.
Abbreviations
AMPA, a-amino-3-hydroxy-5-methylisoxazole-4-propionic acid; AMPpcp, phosphomethylphosphonic acid adenylate ester; GRIP, glutamate
receptor interacting protein; hSR, human serine racemase; PDB, protein databank; PDZ, PSD-95 ⁄ discs large ⁄ ZO-1; PLP, pyridoxal
phosphate; SR, serine racemase.
FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4561
Sequence comparison studies have shown that SR is
homologous to other type II fold-pyridoxal phosphate
enzymes. However, we observed that, unlike other
amino acid racemases, Ca
2+
,Mn
2+
,Mg
2+
and other
divalent cations were able to activate recombinant
brain SR through a process that involved the stabiliza-
tion of the enzyme [9]. The addition of chelating
agents such as EDTA or EGTA almost completely
abrogated the synthesis of d-serine. Further work dem-
onstrated that d-serine synthesis by a brain purified
SR was activated by ATP, ADP and GTP as well as
by divalent cations [10]. Likewise, when SR was
expressed recombinantly in human embryonic kidney

cells and purified, both divalent cations and ATP or
ADP activated the enzyme [11]. In addition, these two
reports showed that ATP was not hydrolysed during
turnover, hence suggesting a novel, allosteric role for
this nucleotide during d-serine synthesis.
By means of the yeast two-hybrid approach, three
proteins have been shown to interact with brain SR:
glutamate receptor interacting protein (GRIP) [1,12],
PICK1 [13] and Golga3 [14]. GRIP is a large protein
with seven PSD-95 ⁄ discs large ⁄ ZO-1 (PDZ) domains
that associates with the GluR2 ⁄ 3 subunits of the
a-amino-3-hydroxy-5-methylisoxazole-4-propionic acid
(AMPA) subtype of glutamate receptors using PDZ4
and PDZ5 [15]. PICK1, a protein that contains a sin-
gle PDZ domain, also binds to the GluR2 ⁄ 3 subunits
of the AMPA receptors as well as to protein kinase C
[13]. Both GRIP and PICK1 bind to the four carboxy-
terminal amino acids of SR (-Ser-Val-Ser-Val-COOH
in the human enzyme and -Thr-Val-Ser-Val-COOH in
the mouse enzyme) [12,13]. Interestingly, coexpression
of GRIP with SR caused a five-fold increase in the
d-serine released into the medium [1], whereas infec-
tion of astrocytes with a GRIP adenovirus resulted in
a two- to three-fold increase [12]. Finally, Golga3
binds to the first 66 amino acids of brain SR and,
remarkably, this binding also results in an increased
d-serine synthesis. However, activation of SR by
Golga3 takes place through a decreased ubiquitylation
of SR and diminished proteasomal degradation [14].
Because calcium activates SR [9,11], it is conceivable

that GRIP binding might target SR to the proximity
of the Ca
2+
-permeable AMPA channel. Alternatively,
due to the promiscuous interaction of GRIP with
numerous cellular proteins, GRIP binding to SR might
bring it to the close proximity of some other protein
responsible for the observed increase in d-serine syn-
thesis. Finally, GRIP might also activate brain SR
through a putative conformational change that might
result in an increased catalytic rate. With that in mind,
we analysed in detail the mechanisms by which the
hepta-PDZ domain protein GRIP activates brain SR.
We show that GRIP induces a conformational change
in SR responsible for the observed increase in d-serine
synthesis.
Results
Binding of PDZ6 of GRIP to brain SR
When SR is transfected in COS7 cells, high levels of
d-serine can be detected in the supernatant. Conse-
quently, we transfected brain SR in COS7 cells in the
presence and absence of transfected PDZ6 domain of
GRIP or full-length GRIP (Fig. 1A). In the three cases,
similar levels of transfected SR were obtained (Fig. 1A,
upper panel). Remarkably, although full-length GRIP
was able to increase significantly the activity of SR, its
PDZ6 domain was unable to increase the levels of
d-serine in the supernatant. To investigate whether
the PDZ6 domain of GRIP was efficiently transfected
in COS7, we performed an immunoblotting, which

resulted in a positive band at the expected molecular
mass (approximately 13 kDa; Fig. 1A, upper right-
hand panel). Next, to rule out the possibility that SR
and the PDZ6 domain of GRIP might be unable to
interact, we immunoprecipitated the transfected SR
from the COS7 cells and investigated whether the
PDZ6 domain of GRIP effectively associated with
the carboxy-terminal end of SR. As shown in Fig. 1A,
the PDZ6 domain of GRIP was found to be in associa-
tion with SR, an observation that confirms the proper
binding of both proteins.
Next, we decided to analyse the binding of the PDZ6
domain of GRIP to brain SR using recombinant pro-
teins. We expressed recombinant brain SR in Escheri-
chia coli using the pCWori vector as previously
described [9,16], as well as the isolated PDZ6 domain
of GRIP (residues 665–768). Both recombinant pro-
teins were purified to homogeneity (Fig. 1B). Because
the PDZ6 domain of GRIP is just a part of the entire
protein, we confirmed by circular dichroism that the
protein was properly folded (Fig. 1B, right panel). The
far-UV circular dichroism spectrum of the PDZ
domain of GRIP indicates an abundance in b-sheet
content, as confirmed from the crystal structure of this
domain [17].
Once we had confirmed that the PDZ6 domain of
GRIP was properly folded, we analysed the activity of
SR under different conditions in the presence and
absence of this binding protein (Fig. 1C). When SR is
purified in the presence of divalent cations, the

addition of extra Ca
2+
or Mg
2+
has a limited effect
on its catalytic activity. The chelating agent EDTA
Activation of brain SR by GRIP, calcium and ATP F. Baumgart et al.
4562 FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS
significantly diminished catalytical activity, whereas
ATP significantly increased the racemase activity
(Fig. 1C). When ATP plus Ca
2+
⁄ Mg
2+
was added to
the reaction mixture, the synthesis of d-serine reached
almost a three-fold increase over basal activity levels.
When recombinant PDZ6 domain of GRIP was added
to SR in a 2 : 1 molar ratio, no significant changes in
activity could be observed. This clearly indicates that
the binding of the PDZ6 domain by itself is not
responsible for the previously reported increase in SR
activity [1,12].
Binding of brain SR to recombinant constructs of
GRIP that included additional PDZ modules
GRIP is a very large protein (1112 amino acids) com-
posed of at least nine protein modules. The architecture
of this protein is depicted in Fig. 2A. Each PDZ
domain is comprised of approximately 100 amino acids
and they are conserved in terms of folding, although

each PDZ module is able to interact with a different
subset of target proteins [18,19]. GRIP contains two
clusters of three PDZ domains, since PDZ1, PDZ2 and
PDZ3 are consecutive in the sequence as well as PDZ4,
PDZ5 and PDZ6. On the other hand, PDZ7 is close to
the carboxy-terminal end of the amino acid sequence
[15]. The two clusters of PDZ domains, as well as the
second cluster and PDZ7, are separated by large
domains of unknown function referred to as GAP1 and
GAP2 (Fig. 2A). Binding of GRIP to SR occurs
through the direct interaction of the PDZ6 domain of
GRIP and the final four amino acids of SR [12].
We performed the recombinant expression in E. coli
and the purification of two larger fragments of GRIP
and incubated the purified proteins with SR to analyse
changes in activity. A hexa-His tag was introduced at
the N-terminal end and a FLAG tag was introduced at
the carboxy-terminal end. Expression and purification
conditions were optimized to obtain a homogeneous
band by SDS ⁄ PAGE. When the purified protein that
contained the cluster PDZ4-PDZ5-PDZ6 (residues 468–
768) was incubated with SR in a 2 : 1 molar ratio,
we failed to observe the expected increase in d-serine
A
B
C
Fig. 1. Binding of SR to the PDZ6 domain of GRIP. (A) COS7 cells
were transfected with a SR plasmid and then, 24 h post-transfec-
tion, they were trypsinized, split into three flasks and transfected
with a PDZ6, a full-length GRIP or an empty plasmid. The amount

of the released
D-serine into the medium was determined in the
three cases (left panel). The efficient expression of SR (upper left
panel) and PDZ6 domain of GRIP (upper right panel) in the trans-
fected COS7 cells was determined by immunodetection by wes-
tern blot. The association between the PDZ6 domain of GRIP and
SR was determined through the immunoprecipitation of SR and the
immunodetection of FLAG-tagged PDZ6 (bottom right panel).
(B) Coomassie Blue-stained SDS ⁄ PAGE gels of purified recombi-
nant SR (left) and the PDZ6 domain of GRIP (middle). The circular
dichroism spectrum of purified PDZ6 domain of GRIP is shown in
the right panel. (C)
D-Serine synthesis by recombinant SR in the
absence (black bars) and presence (grey bars) of a two-fold molar
excess of the PDZ6 domain of GRIP under different assay condi-
tions. Data are representative of three independent experiments.
F. Baumgart et al. Activation of brain SR by GRIP, calcium and ATP
FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4563
synthesis (Fig. 2B). Although a limited rise in activity
was observed when both ATP and Ca
2+
were present,
this increase is far from the expected five-fold [1] or two-
to three-fold [12] increase that was reported when both
proteins were transfected in mammalian cells. Remark-
ably, when we incubated the purified protein that
contained the PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end of
GRIP (residues 468–1112) with SR in a 2 : 1 molar
ratio, a significant increase in activity was obtained
when no additional ATP was added (Fig. 2C). This

increase was over two-fold when purified SR lacked
additional Ca
2+
or ATP, as well as when additional
Ca
2+
had been included in the reaction mixture.
Characterization of the activation of brain SR
by GRIP(468–1112)
We next analysed whether larger ratios of PDZ6
domain:SR might be able to increase the synthesis of
d-serine. We tested up to a ratio of 23 : 1 using the
recombinant PDZ6 domain of GRIP (residues 665–768)
and recombinant SR (Fig. 3A). Even at this large excess
of PDZ6, the increase in racemization activity was extre-
mely limited. However, when we tested larger ratios of
recombinant PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end of
GRIP (residues 468–1112), a significant increase in the
racemase activity was observed (Fig. 3B). Due to solu-
bility problems, we could not go beyond a 5.8 ratio of
GRIP fragment:SR but, at this point, we observed a
2.8-fold augmentation in racemase activity. Hence, this
result indicates that certain residues of GRIP that lie
outside the binding site for SR present within PDZ6 are
responsible for a complete activation of the enzyme.
Next, we maintained a 1 : 1 molar ratio of the recom-
binant protein PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end to
racemase and inspected the catalytic properties of the
latter at increasing concentrations of the substrate l-ser-
ine in the absence of additional ATP (Fig. 3C). We were

able to determine that binding of this fragment of GRIP
to SR changed both the K
m
and the V
max
of the reaction.
The K
m
for d-serine synthesis increased from 1.9 mm in
the absence of GRIP fragment (filled circles) to 3.4 in
its presence (empty circles), whereas the V
max
of the
reaction augmented 1.65-fold (from approximately
115 ± 8 nmolÆmg
)1
Æmin
)1
to 190 ± 7 nmolÆmg
)1
Æ
min
)1
; Fig. 3C, insert). Consequently, binding of PDZ4-
PDZ5-PDZ6-GAP2-PDZ7-end of GRIP induces a
conformational change in recombinant SR that is
responsible for its increased catalytic properties.
We next considered whether the binding of GRIP
to SR might result in a shift in the response towards
calcium. In the presence of ATP, a preparation of

recombinant SR that had been purified in the presence
D-Serine Synthesis (%)
50
100
150
200
250
300
D-Serine Synthesis (%)
100
200
300
400
+ ATP
mock
+ EDTA
+ Ca
2+
+ Ca
2+
/ATP
+ ATP
mock
+ EDTA
+ Ca
2+
+ Ca
2+
/ATP
C

N
PDZ1
GAP2
GAP1
PDZ7
PDZ4
PDZ5
PDZ6
PDZ3
PDZ2
468
665
768
1112
A
B
C
75
50
37
25
75
50
37
25
Fig. 2. D-Serine synthesis by purified recombinant SR in the pres-
ence of the PDZ4-PDZ5-PDZ6 and PDZ4-PDZ5-PDZ6-GAP2-PDZ7-
end modules of GRIP. (A) GRIP is formed by seven PDZ domains
separated by two large GAP domains of unknown function. The
residue numbering indicative of the beginning and end of each

domain is shown on the right.
D-Serine synthesis is measured
under various conditions in the absence (black bars) or presence
(grey bars) of GRIP constructs. The effect of (B) PDZ4-PDZ5-PDZ6
and (C) PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end is shown together with
a Coomassie Blue-stained SDS ⁄ PAGE (inserts). Data are represen-
tative of three independent experiments.
Activation of brain SR by GRIP, calcium and ATP F. Baumgart et al.
4564 FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS
of divalent cations, but afterwards dialysed, was incu-
bated with increasing levels of Ca
2+
in the absence
and presence of PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end
of GRIP (Fig. 3D). Values above 2 mm Ca
2+
could
not be reached due to protein precipitation in the
assay. In both cases, recombinant SR became activated
in the 1–100 lm range of Ca
2+
. The activation curve
was not displaced towards lower Ca
2+
concentrations,
a clear indication that GRIP binding does not result in
a higher sensitivity of SR towards this divalent cation.
ATP binding does not involve significant changes
either in the quaternary structure of SR or in its
kinetic properties

At this point, we considered the mechanism by which
the addition of ATP or ADP might augment the cata-
lytic properties of SR. It should be noted that the highly
homologous threonine dehydratase, an E. coli enzyme,
binds to and becomes activated by AMP, in a process
that results in the decrease of the K
m
for threonine
from 70 mm in the absence to 5 mm in the presence of
this cofactor [20]. At the same time, binding of AMP to
E. coli threonine dehydratase results in the association
of the protein monomers into tetramers [21]. Accord-
ingly, we inspected the catalytic properties of SR in the
presence and absence of added ATP. In the presence of
100 lm ATP, the enzyme becomes activated but, inter-
estingly, the V
max
for d-serine synthesis increases 2.2-
fold, whereas the K
m
of the reaction remains almost
unchanged (approximately 3.2 mm in the absence
against 3.0 mm in the presence of ATP) (Fig. 4A).
Remarkably, ATP binding is not involved in protein
oligomerization. We previously showed that brain SR is
found in solution in a dimer–tetramer equilibrium [9],
with the dimer eluting at approximately 12.8 mL in a
Superdex-200 column. Addition of 10 lm ATP plus
100 lm Mg
2+

to brain SR did not induce changes in the
elution profile (Fig. 4B). Although a slight increase in
the population of high molecular mass oligomers eluting
at approximately 10 mL was observed, the population
of dimers remained the most abundant in solution.
Molecular modelling of human SR (hSR) using
the crystal coordinates of homologous enzymes
from Schizosaccharomyces pombe and Thermus
thermophilus
Although no mammalian serine racemases have been
resolved to date, the structures of two homologous
racemases have been recently solved and their atomic
coordinates deposited in public databases. The three-
dimensional structure of SR from S. pombe has been
solved in the presence of Mg
2+
as well as with the
ATP analogue phosphomethylphosphonic acid adenyl-
ate ester (AMPpcp) [protein databank (PDB) codes:
molar ratio (GRIP/SR)
D-Ser synthesis (%)
50
100
150
200
250
molar ratio (GRIP/SR)
0 5 10 15 20 0 1 2 3 4 5 6
D-Ser synthesis (%)
50

100
150
200
250
[L-Ser] (mM)
0 1020304050
D-Ser synthesis (A
411 nm
)
0.00
0.01
0.02
0.03
0.04
0.05
0.06
1/[L-Ser]
-0.5 0.0 0.5 1.0 1.5
1/A
411 nm
0
20
40
60
80
additional Ca
2+
added
100 n
M

1 µ
M
10 µM
100 µ
M
1 m
M
D-Ser synthesis (A
411 nm
)
0.02
0.04
0.06
0.08
+ EDTA
A
C
D
B
Fig. 3. Characterization of SR activation by the PDZ4-PDZ5-PDZ6-
GAP2-PDZ7-end module of GRIP.
D-Serine synthesis by recombi-
nant SR in the presence of increasing molar ratios of (A)
the recombinant PDZ6 domain or (B) PDZ4-PDZ5-PDZ6-GAP2-
PDZ7-end module of GRIP. (C) The kinetic properties of
D-serine
racemization were determined in the absence (filled circles) or pres-
ence (empty circles) of a two-fold molar excess of PDZ4-PDZ5-
PDZ6-GAP2-PDZ7-end module of GRIP towards SR at increasing
substrate concentrations. The Lineweaver–Burk plot is depicted in

the insert. These assays were performed in the presence of 1 m
M
Ca
2+
and in the absence of ATP. (D) Increase in D-serine synthesis
at increasing Ca
2+
concentrations by recombinant SR in the
absence (filled circles) or presence (empty circles) of a two-fold
molar excess of the PDZ4-PDZ5-PDZ6-GAP2-PDZ7-end module of
GRIP. Data are representative of three independent experiments.
F. Baumgart et al. Activation of brain SR by GRIP, calcium and ATP
FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4565
1V71 and 1WTC]. Likewise, the atomic coordinates of
threonine deaminase from T. thermophilus have been
obtained by X-ray crystallography in the presence of
Ca
2+
(PDB code: 1VE5). With the aim of building a
full-atom 3D model of hSR, we carried out a tertiary
structure prediction approach by using three different
servers available on the Internet (see Experimental pro-
cedures). The results provided by the program were
fully consistent in that all of them rendered the same
hit with the highest score: the SR from the fission
yeast S. pombe (40% sequence identity with hSR). In
the case of 3d-jury, the threading predictions were
stopped because it readily considered 1V71 as a signifi-
cant hit. The other potential template provided by sp3
and 3d-shotgun with a similar but lower score is thre-

onine deaminase from T. thermophilus (TtTD; PDB
code: 1VE5) (41% sequence identity with hSR). Both
proteins, which belong to the tryptophan synthase
b-subunit-like pyridoxal-phosphate-dependent enzymes
superfamily, share the same polypeptide chain fold.
The 3D comparison of the Ca atoms of the two crystal
structures yielded a root mean square (rms) deviation
of 1.3 A
˚
for 289 Ca atoms. Thus, the template selected
for building a 3D model for hSR was the SR from
S. pombe (SpSR). Although the 3D models provided
by both 3d-shotgun and sp3 are indistinguishable
(rmsd ¼ 0.75 A
˚
for 318 Ca-atoms), it should be noted
that the last server provides a model for the first four
residues of hSR and also for the last 12 amino acid
residues, which were not modelled by 3d-shotgun.
Superimposition of the crystal structure of SpSR
and the herein proposed full-atom model for hSR
reveals that the pyridoxal phosphate (PLP) binding site
is conserved. The coenzyme is located deep between
the two subdomains. The lysine residue Lys56 in hSR
would be homologous to Lys57 in SpSR, Lys51 in
TtTD, or Lys62 in threonine deaminase from E. coli
[22], which are covalently bound to PLP through the
formation of a Schiff base. Conversely, the microenvi-
ronment of the phosphate moiety of PLP is essentially
conserved, mainly interacting with the tetraglycine

loop (Gly185 to Gly188), which is highly conserved in
PLP-dependent enzymes. Other residues interacting
with the PLP coenzyme are Phe55, which embraces the
pyridine ring, Ser313 and Asn154. The only significant
difference in the vicinity of PLP between hSR and
SpSR is precisely the presence of this last residue
Asn54 instead of Tyr152 in SpSR.
The crystal structure of both SpSR and TtTD
revealed a common cation binding site, which in the
first case is occupied by Mg
2+
and in the second by
Ca
2+
(Fig. 5A). In both cases, the metal is hexavalent-
ly coordinated. The cation binding site is formed by
two carboxylate-containing residues (Glu208, Asp214
in SpSR and Glu203, Asp209 in TtTD), a main chain
carbonyl oxygen (Gly212 in SpSR and Ala207 in
TtTD) and three well-ordered water molecules, which
in turn are hydrogen-bonded to main chain carbonyl
groups. Both the geometry and the distances perfectly
agree with the coordination of these metals [23]. The
Fig. 4. Binding of ATP to SR changes the
V
max
of the racemization reaction but not
the oligomeric state of the enzyme.
(A) Michaelis–Menten representation and
Lineweaver–Burk plot of the activation of

brain SR by ATP. Assays were performed in
the presence of 1 m
M Ca
2+
plus 1 mM
Mg
2+
. (B) Gel filtration elution profile of
recombinant SR (approximately 400 lg per
run) in the absence (upper trace) and pres-
ence (bottom trace) of 10 l
M ATP plus
100 l
M Mg
2+
. Data are representative of
four independent experiments.
Activation of brain SR by GRIP, calcium and ATP F. Baumgart et al.
4566 FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS
corresponding region of hSR is perfectly superimpos-
able with the above crystal structures, indicating that
the cation binding site is conserved, thus explaining
the Ca
2+
-binding ability of hSR. In this case, the
homologous amino acid residues in hSR putatively
involved in Ca
2+
coordination are: Glu210, Asp216
and Ala214.

Based on the crystal structure of SpSR (PDB code:
1V71), a putative model for the dimeric hSR can be
easily built (Fig. 5B). In this structure, two monomers
related by a crystallographic two-fold axis tightly asso-
ciate forming dimeric species. It is worth noting that
the presence of an extended C-terminal end in the
model for hSR herein proposed would not preclude
this association (Fig. 5B). The putative interface
between hSR monomers would be made up primarily
of hydrophobic residues essentially contributed by
regular structure elements of the polypeptide chain.
Estimation of the hydrophobic surface area with nac-
cess from the ligplot package reveals that approxi-
mately 76% of the contact area is hydrophobic, which
is typical of obligate complexes. Additionally, the crys-
tal structure of SpSR in the presence of AMPpcp has
also been deposited (PDB code: 1WTC), revealing a
putative ATP binding site. According to the present
model for hSR, the ATP analogue would be located
between hSR monomers in a shallow manner, mainly
interacting with polar residues (Fig. 5B). Remarkably,
comparison of the crystal structures of SpSR in the
presence and absence of AMPpcp shows two impor-
tant aspects. First, binding of the ATP analogue does
not induce large conformational changes in the protein
and, second, the ligand does not modify the dimeric
state of the protein, which is in agreement with the
results herein provided.
Finally, the proposed structure for the dimeric hSR
permits the visualization of a tentative model for the

hSR:PDZ6 complex (Fig. 5C). This model has been
constructed assuming that the C-terminal eight resi-
dues of hSR adopt an extended conformation similar
to that of the octapeptide of human liprin-a com-
plexed with the GRIP1 PDZ6 domain [17], with an
identical mode of interaction of PDZ6. Although this
model should only be considered tentatively, it per-
mits the identification of the regions of hSR that can
be directly affected by the binding of the PDZ
domain.
Discussion
Fluorescence studies from numerous groups have dem-
onstrated that astrocytes respond to neurotransmitters
not with action potentials, like most neurones, but
A
B
C
Fig. 5. (A) Calcium binding site of hSR. The polypeptide chains of
hSR (yellow) and SpSR (blue) are superimposed. Calcium is shown
as a green sphere; water molecules (wat) present in the crystal
structure of SpSR are shown as red spheres. Labels for the resi-
dues are for hSR. Carboxylate-containing residues and carbonyl oxy-
gens involved in cation binding are shown as sticks. The figure was
generated with
PYMOL [32]. Positioning of the Ca
2+
, ATP and PDZ6
domain binding sites in the model of hSR. (B) 3D Model of the
dimeric hSR. The relative position of the PLP (red sticks), Ca
2+

(magenta spheres) and AMPpcp (orange sticks) are shown. The
N- and C-terminal ends of hSR are indicated as N and C, respec-
tively. Magnesium ions complexed with the nucleotide present in
the crystal structure of SpSR (PDB code: 1WTC) are shown as
green spheres. (C) Putative 3D model of the PDZ6:HSR complex.
The figure was generated with
PYMOL.
F. Baumgart et al. Activation of brain SR by GRIP, calcium and ATP
FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4567
with propagating waves of intracellular calcium ions.
In response to these calcium oscillations, glial cells
release ‘gliotransmitters’ such as glutamate, ATP and
d-serine. Released glutamate and d-serine occupy
binding sites in the NMDA receptors of neurones
hence regulating neuronal function. In this context, we
analysed how brain SR becomes activated by three
independent factors: divalent cations, nucleotides such
as ATP and the multi-PDZ domain protein GRIP.
When an adenovirus that contained GRIP was used to
infect mice the cerebellar d-serine concentration was
augmented two-fold [12]. One likely explanation for
this observation is that the PDZ4-PDZ5 domains of
GRIP might interact with the AMPA channel, target-
ing brain SR bound to the PDZ6 domain of GRIP
towards its proximity. Under these circumstances,
brain SR might become activated due to the calcium
influx through the channel. Conversely, brain SR acti-
vation might occur through a conformational change
induced upon GRIP binding. Using purified recombi-
nant proteins, we have been able to establish that

GRIP binding to brain SR induces a conformational
change that increases the racemase activity by over
two-fold. Although the PDZ6 domain of GRIP is
involved in the interaction with the last four carboxy-
terminal residues of SR, this association is not enough
to induce SR activation. Our results indicate that addi-
tional PDZ modules are necessary to trigger the
observed activation. In this regard, it must be noted
that multi-PDZ domain proteins are known to display
‘communication’ between individual modules. For
example, neither the PDZ4, nor the PDZ5 domain of
GRIP bind to the GluR2 subunit of the AMPA chan-
nel independently, with the concerted action of both
being necessary for the association [15]. Analysis of
the amino acid sequence of GRIP reveals the presence
of two GAP domains that separate the two three-PDZ
domain tandems, PDZ1-PDZ2-PDZ3 from PDZ4-
PDZ5-PDZ6 and the latter from PDZ7. Activation of
brain SR requires the presence of GAP2 together with
PDZ7. Interestingly, in the absence of additional ATP,
the activation induced by the presence of the PDZ4-
PDZ5-PDZ6-GAP2-PDZ7-end of GRIP reached the
highest value. The conformational change induce by
GRIP binding on brain SR results in changes in both
the K
m
and the V
max
for d-serine synthesis and the
response curve for activation by calcium remained

unaltered.
Considering that the cellular concentration of ATP
(3–6 mm) is well above that needed for SR activation
[11], it is intriguing to elucidate what might be the
exact role of the nucleotide in vivo. Because amino acid
racemization in PLP-containing racemases is not an
ATP-driven reaction and ATP is not hydrolysed dur-
ing catalysis [10,11], the nucleotide could exert an allo-
steric role. Inspection of the 3D model of brain SR
reveals that ATP is positioned in the monomer–mono-
mer interface. However, crystals of the homologous
SpSR reveal that these enzymes establish identical
monomer–monomer interactions in the presence and
absence of the nucleotide. According to our results,
the absence of modulation of the oligomeric state
of the enzyme by the nucleotide is in clear contrast
with the behaviour displayed by the homologous
bacterial threonine deaminase [20,21]. Our data also
indicate that, in the absence of added ATP, brain SR
is more readily regulated by GRIP binding. It is then
conceivable that, in the microenvironments where SR
is present, cellular ATP levels might be low, hence per-
mitting a tight regulation through ATP binding.
Our finding that SR, GRIP and the GluR2 subunit
of the AMPA channel are able to form a ternary com-
plex appears to indicate that SR might be positioned
close to this calcium channel, hence modulating its
activity.
Experimental procedures
Chemicals and antibodies

Ultrapure l-serine was purchased from NovaBiochem
(Laufelfingen, Switzerland). Horseradish peroxidase was
obtained from Roche Molecular Biochemicals (Mannheim,
Germany). SR monoclonal and GRIP antibodies were
obtained from Transduction Laboratories. d-Serine,
d-amino acid oxidase from porcine kidney and FLAG anti-
body were purchased from Sigma (St Louis, MO, USA).
Buffers and common laboratory reagents were also
obtained from Sigma. Using a campus facility, we injected
pure recombinant brain SR and pure PDZ4-PDZ5-PDZ6-
GAP2-PDZ7-end GRIP fragment into rabbits. Injections
were performed every week over a 6-week period. The anti-
serum was then obtained and tested.
Cloning of SR and the GRIP constructs PDZ4-
PDZ5-PDZ6, PDZ6 and PDZ4-PDZ5-PDZ6-GAP2-
PDZ7-end into pCWori and recombinant protein
expression
The plasmid encoding the cDNA of mouse SR was a gener-
ous gift from S. H. Snyder (Johns Hopkins University,
Baltimore, MD, USA) and GRIP1 cDNA in pBK was kindly
given to us by R. Huganir (Johns Hopkins University).
We have previously described the recombinant expression
and purification of brain SR [9]. All the constructs were
cloned in the pCWori plasmid that possessed a hexa-His tag
Activation of brain SR by GRIP, calcium and ATP F. Baumgart et al.
4568 FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS
at the 5¢-end in frame when the NdeI site was used (N-termi-
nus of the recombinant protein) [9,16]. In addition, by PCR,
we introduced a SalI site at the 3¢-end of the amplified gene
followed by eight amino acids that encoded for the FLAG

epitope and a XbaI site (at the carboxy-terminal end of the
protein). Hence, all constructs were cloned between NdeI and
SalI to ensure that there would be a tag at each end of the
protein. In constructs with PDZ4 at the 5¢-end, we performed
a PCR with GRIP as template using the oligonucleotide 5¢-
GGACAGGTTGTT
CATATGGAAACACA-3¢, which intro
duced the NdeI site (underlined sequence) as forward primer.
The reverse oligonucleotide that annealed at the 3¢-end of the
gene and introduced the FLAG epitope was 5¢-GG
TCT
AGAGCTTATCGTC ATCGTCCTTGTAGTCGACTGTG
TTAGT-3¢. The SalI and XbaI sites are underlined. For
PDZ6 cloning (320 bp) we used the oligonucleotide 5¢-GAT
GAG
CATATGAGTTCCCGGGCG-3¢ as forward primer
(introducing a NdeI site) and 5¢-TGA
GTCGACGGGGAT
GGGCAGCTT-3¢ (introducing a SalI site) as reverse primer.
The amplification of the PDZ4-PDZ5-PDZ6 GRIP construct
was performed using the forward primer 5¢-GGACAGGT
TGTT
CATATGGAAACCACA-3¢ (introducing a novel
NdeI site) together with 5¢-TGA
GTCGACGGGGATGGG
CAGCTT-3¢ (introducing a novel SalI site).
The amplified DNA fragments were subcloned into a
pGEM-T vector (Promega, Madison, WI, USA) and con-
firmed by automated DNA sequencing. Then, they were
double digested with NdeI plus SalI and ligated into the

corresponding sites of pCWori. In all cases, the hexa-His
was at the N-terminal end of the protein and the FLAG
epitope at the carboxy-terminal end. BL21 (DE3)-compe-
tent cells (Novagen, Merck Chemicals Ltd, Nottingham,
UK) were routinely transformed with the respective pCW-
ori constructs and an overnight 10 mL culture was used to
inoculate a 2.8 L flask containing 0.75 L of 2 · yeast ⁄ tryp-
tone medium. Typically, 1.5 L of cell cultures were grown
in the presence of 100 mgÆL
)1
ampicillin at 37 °Ctoan
absorbance of 1.0 and induced adding 1 mm isopropyl thio-
b-d-galactoside. The E. coli cultures were then grown at
30 °C (to avoid degradation, especially of the PDZ con-
structs) at 220 r.p.m. for 16–20 h before the cells were har-
vested by centrifugation at 5140 g for 30 min in an F10
rotor (Sorvall, Norwalk, CT, USA). The cell pellets were
frozen in plastic bags as thin films and stored at )80 °C.
Immunoblot and immunoprecipitation
Regarding immunoblot, immunoprecipitation and confocal
analysis, we followed standard cellular biology protocols, as
reported in previous studies performed by our group [24,25].
Cloning of PDZ6 into pCDNA3
The DNA of FLAG-tagged PDZ6 was amplified from the
pCWori-PDZ6 construct, using 5¢-ATGCACCATCACC
A
GAATTCCCATATG-3¢ as forward primer and 5¢-CAT
GTTTGACAGCTTAT
TCTAGAG-3¢ as reverse primer con
taining EcoRI and XbaI restriction sites (underlined

sequences). The PCR product was double digested with
EcoRI plus XbaI and ligated into the corresponding sites of
pCDNA3. Thus, the PCR product of PDZ6 that we obtained
contained the sequence for a C-terminal FLAG tag for
immunoprecipitation and immunodetection. The PDZ6-
pCDNA3 plasmid was subsequently confirmed by automated
DNA sequencing.
Determination of D-serine concentration
A routine colorimetric assay with 200 lL of total sample
was used with l-serine as a substrate, coupling the appear-
ance of d-serine to commercial d-amino acid oxidase and
horseradish peroxidase, plus the peroxidase substrate
O-phenylenediamine. The d-serine that was produced dur-
ing the incubation period was degraded by d-amino acid
oxidase, which specifically targets d-amino acids generating
-keto acid, ammonia, and hydrogen peroxide. The hydro-
gen peroxide was quantified using horseradish peroxidase
and O-phenylenediamine, which turns yellow upon oxida-
tion. The activity of serine racemase was determined in
the presence of 20 mm Mops, pH 8.1, 10 lL of purified
enzyme (5 lg approximately), 10 mml-serine, 0.03 m m
dithiothreitol, 5 lm PLP, 50 lgÆmL
)1
O-phenylenediamine,
1nm FAD, 0.2 mgÆmL
)1
d-amino acid oxidase, and
0.01 mgÆmL
)1
horseradish peroxidase. The reactions

(200 lL, final volume) were incubated at 37 °C for 2 h
before measuring the absorbance (A
411 nm
) with a Beckman
DU-7 spectrophotometer (Beckman Coulter Inc., Fullerton,
CA, USA). The d-serine present in a given sample was
determined by correlating the absorbance (A
411 nm
) with
d-serine calibration curves. d-Serine measurements were
not influenced by the concentration of l-isomers present in
the sample. Because standard commercial l-serine prepara-
tions contain trace amounts of d-serine, it was necessary to
purchase ultrapure l-serine from NovaBiochem to achieve
a reasonable signal-to-noise ratio. Once optimized, this
three-enzyme assay was able to detect d-serine formed in
an unknown 100 lL sample in the linear range from 50 lm
to 1 mm (up to 0.1 lmol total of d-serine). All the measure-
ments were performed in triplicate.
To measure d-serine production by COS7 transfected with
SR, we used the second part of the colorimetric assay
described above. Transfected cells were incubated for
approximately 8 h with phenol-red-free Dulbecco’s modified
Eagle’s medium supplemented with 10 mml-serine. Col-
lected supernatants were boiled for 10 min and centrifuged
before the assay. To estimate d-serine levels, we typically
analyzed a 300 lL sample adding 10 nm FAD, 10 lgÆmL
)1
horseradish peroxidase, 50 lgÆmL
)1

O-phenylenediamine
and 100 lgÆmL
)1
d-amino acid oxidase in a final volume
of 400 lL. Samples were incubated for 5 h at 37 °C,
F. Baumgart et al. Activation of brain SR by GRIP, calcium and ATP
FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4569
centrifuged in a table-top microcentrifuge at 16 000 g and
subsequently measured at 411 nm with a spectrophotometer.
Circular dichroism measurements
CD spectra were recorded on a Jasco J-715 spectropolarime-
ter (Jasco Inc., Easton, MD, USA) using a 0.1 cm path
length cell at 25 °C. The temperature in the cuvette was reg-
ulated with a Neslab RT-111 circulating water bath (Neslab
Inc., Portsmouth, NH, USA). The buffer used was 50 mm
Tris, pH 7. A minimum of five spectra were accumulated
for each sample and the contribution of the buffer was
always subtracted. The resultant spectra were smoothed
using j715 noise reduction software provided with the CD
spectrophotometer.
Gel filtration analysis of recombinant SR in the
presence and absence of ATP
Generally, we followed a previously published protocol [9].
Aliquots of 200 lL of approximately 2 mgÆmL
)1
of recom-
binant racemase eluted from the Ni-nitrilotriacetic acid
affinity resin were injected into a GP 250 plus fast protein
liquid chromatography system equipped with two P-500
pumps and a Superdex HR200 column (Amersham Bio-

sciences, Piscataway, NJ, USA). Separation was performed
at 25 °C and protein detection was performed at 280 nm.
The flow rate was kept at 1 mLÆmin
)1
and the buffer
consisted of 50 mm Tris, pH 7.0, 50 mm NaCl, 1 mm
dithiothreitol in the presence or absence of 10 lm ATP plus
100 lm Mg
2+
. We were unable to use concentrations of
ATP above 10 lm due to the elevated absorbance signal
that we obtained.
Modelling hSR
The servers used in this work for prediction of the tertiary
structure of hSR were: sp3 (ui.
edu ⁄ hzhou ⁄ anonymous-fold-sp3.html) [26], 3d-shotgun
(INUB predictor; ) [27] and
3d-jury (BioInfoBank Meta server; />submit_wizard.pl) [28]. Currently, these servers are consid-
ered among the best performers for structure prediction as a
result of the CAFASP4 experiment (Critical Assessment of
Fully Automated Structure Prediction) [28]. Considering the
prediction results, a full-atom 3D model of hSR was built by
using the atom coordinates of SR from S. pombe (PDB code:
1V71) because it was provided as the best template for the
human homologue. Sequence alignments were analysed with
clustal w [29]. The quality of the final structure was
assessed with the verify3d program [30] and also with
procheck [31]. In the first case, the 3D profile score is high
(approximately 60), which is typical for correct structures
and, in the second case, the program indicated that the

stereochemistry of the model is correct, with 99.3% of amino
acid residues in the allowed regions of the Ramachandran
plot (data not shown). 3D Superposition of protein struc-
tures and other analyses were performed with software o
[23].
Acknowledgements
We would like to thank Dr Martı
´
nez del Pozo for
many useful comments and corrections of the manu-
script. We are also grateful to Dr Galve-Roperh for
numerous comments and suggestions. This work was
supported by grant BMC2006 05395 from the Spanish
DGICYT.
References
1 Boehning D & Snyder SH (2003) Novel neural modula-
tors. Annu Rev Neurosci 26, 105–131.
2 Mustafa AK, Kim PM & Snyder SH (2004) D-Serine as
a putative glial neurotransmitter. Neuron Glia Biol 1,
275–281.
3 Wolosker H, Blackshaw S & Snyder SH (1999) SR: a
glial enzyme synthesizing D-serine to regulate gluta-
mate-N-methyl-D-aspartate neurotransmission. Proc
Natl Acad Sci USA 96 , 13409–13414.
4 Schell MJ, Molliver ME & Snyder SH (1995) D-serine,
an endogenous synaptic modulator: localization to
astrocytes and glutamate-stimulated release. Proc Natl
Acad Sci USA 92, 3948–3952.
5 Mothet JP, Parent AT, Wolosker H, Brady RO Jr, Lin-
den DJ, Ferris CD, Rogawski MA & Snyder SH (2000)

D-serine is an endogenous ligand for the glycine site of
the N-methyl-D-aspartate receptor. Proc Natl Acad Sci
USA 97, 4926–4931.
6 Yasuda E, Ma N & Semba R (2001) Immunohistochemical
evidences for localization and production of D-serine in
some neurons in the rat brain. Neurosci Lett 299, 162–164.
7 Kartvelishvily E, Shleper M, Balan L, Dumin E &
Wolosker H (2006) Neuron-derived D-serine release
provides a novel means to activate N-methyl-D-aspar-
tate receptors. J Biol Chem 281, 14151–14162.
8 Williams SM, Diaz CM, Macnab LT, Sullivan RK &
Pow DV (2006) Immunocytochemical analysis of D-ser-
ine distribution in the mammalian brain reveals novel
anatomical compartmentalizations in glia and neurons.
Glia 53, 401–411.
9 Cook SP, Galve-Roperh I, Martinez del Pozo A &
Rodriguez-Crespo I (2002) Direct calcium binding
results in activation of brain SR. J Biol Chem 277,
27782–27792.
10 Neidle A & Dunlop DS (2002) Allosteric regulation of
mouse brain SR. Neurochem Res 27, 1719–1724.
Activation of brain SR by GRIP, calcium and ATP F. Baumgart et al.
4570 FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS
11 De Miranda J, Panizzutti R, Foltyn VN & Wolosker H
(2002) Cofactors of SR that physiologically stimulate
the synthesis of the N-methyl-D-aspartate (NMDA)
receptor coagonist D-serine. Proc Natl Acad Sci USA
99, 14542–14547.
12 Kim PM, Aizawa H, Kim PS, Huang AS, Wickramasin-
ghe SR, Kashani AH, Barrow RK, Huganir RL, Ghosh

A & Snyder SH (2005) SR: activation by glutamate neu-
rotransmission via glutamate receptor interacting pro-
tein and mediation of neuronal migration. Proc Natl
Acad Sci USA 102, 2105–2110.
13 Fujii K, Maeda K, Hikida T, Mustafa AK, Balkissoon R,
Xia J, Yamada T, Ozeki Y, Kawahara R, Okawa M et al.
(2006) SR binds to PICK1: potential relevance to schizo-
phrenia. Mol Psychiatry 11, 150–157.
14 Dumin E, Bendikov I, Foltyn VN, Misumi Y, Ikehara Y,
Kartvelishvily E & Wolosker H (2006) Modulation of
D-serine levels via ubiquitin-dependent proteasomal
degradation of SR. J Biol Chem 281, 20291–20302.
15 Dong H, O’Brien RJ, Fung ET, Lanahan AA, Worley
PF & Huganir RL (1997) GRIP: a synaptic PDZ
domain-containing protein that interacts with AMPA
receptors. Nature 386, 279–284.
16 Rodriguez-Crespo I, Gerber NC & Ortiz de Montellano
PR (1996) Endothelial nitric-oxide synthase. Expression
in Escherichia coli, spectroscopic characterization and
role of tetrahydrobiopterin in dimer formation. J Biol
Chem 271, 11462–11467.
17 Im Y-J, Park SH, Rho SH, Lee JH, Kang GB, Sheng M,
Kim E & Eom SH (2003) Crystal structure of GRIP1
PDZ6-peptide complex reveals the structural basis for
class II PDZ target recognition and PDZ domain-medi-
ated multimerization. J Biol Chem 278, 8501–8507.
18 Hung AY & Sheng M (2002) PDZ domains: structural
modules for protein complex assembly. J Biol Chem
277, 5699–5702.
19 Mothet JP, Pollegioni L, Ouanounou G, Martineau M,

Fossier P & Baux G (2005) Glutamate receptor activa-
tion triggers a calcium-dependent and SNARE protein-
dependent release of the gliotransmitter D-serine. Proc
Natl Acad Sci USA 102, 5606–5611.
20 Dunne CP, Gerlt JA, Rabinowitz KW & Wood WA
(1973) The mechanism of action of 5¢-adenylic acid-acti-
vated threonine dehydrase. IV. Characterization of
kinetic effect of adenosine monophosphate. J Biol Chem
248, 8189–8199.
21 Gerlt JA, Rabinowitz KW, Dunne CP & Wood WA
(1973) The mechanism of action of 5¢-adenylic
acid-activated threonine dehydrase. V. Relation
between ligand-induced allosteric activation and the
protomeroligomer interconversion. J Biol Chem 248,
8200–8206.
22 Gallagher DT, Gilliland GL, Xiao G, Zondlo J, Fisher
KE, Chinchilla D & Eisenstein E (1998) Structure and
control of pyridoxal phosphate dependent allosteric
threonine deaminase. Structure 6, 465–475.
23 Jones TA, Zou JY & Cowan SW (1991) Improved
methods for building protein models in electron density
maps and the location of errors in these models. Acta
Crystallogr A 47, 110–119.
24 Navarro-Lerida I, Corvi MM, Barrientos AA, Gavilanes
F, Berthiaume LG & Rodriguez-Crespo I (2004) Palmi-
toylation of inducible nitric-oxide synthase at Cys-3 is
required for proper intracellular traffic and nitric oxide
synthesis. J Biol Chem 279, 55682–55689.
25 Navarro-Lerida I, Portoles MT, Barrientos AA, Gavil-
anes F, Bosca L & Rodriguez-Crespo I (2004) Induction

of nitric oxide synthase-2 proceeds with the concomitant
downregulation of the endogenous caveolin levels.
J Cell Sci 117, 1687–1697.
26 Zhou H & Zhou Y (2005) Fold recognition by
combining sequence profiles derived from evolution
and from depth-dependent structural alignment of
fragments. Proteins: Struct Funct Bioinform 58,
321–328.
27 Fischer D (2003) 3D-SHOTGUN: a novel, coopera-
tive, fold-recognition meta-predictor. Proteins 51,
434–441.
28 Ginalski K, Eloffson A, Fischer D & Richlewski L
(2003) 3D-Jury: a simple approach to improve
protein structure predictions. Bioinformatics 19, 1015–
1018.
29 Thompson JD, Higgins DG & Gibson TJ (1994) CLUS-
TAL W: improving the sensitivity of progressive multi-
ple sequence alignment through sequence weighting,
position-specific gap penalties and weight matrix choice.
Nucleic Acids Res 22, 4673–4680.
30 Eisenberg D, Luthy R & Bowie JU (1997) VERIFY3D:
assessment of protein models with three dimensional
profiles. Methods Enzymol 277, 396–404.
31 Laskowski RA, MacArthur MW, Moss DS & Thornton
JM (1993) PROCHECK: a program to check the stereo-
chemical quality of protein structures. J Appl Crystal-
logr 26, 283–291.
32 DeLano WL (2002) The PyMOL molecular graphics
system. DeLano Scientific, Palo Alto, CA.
F. Baumgart et al. Activation of brain SR by GRIP, calcium and ATP

FEBS Journal 274 (2007) 4561–4571 ª 2007 The Authors Journal compilation ª 2007 FEBS 4571