Site-directed mutagenesis of predicted active site residues in glutamate carboxypeptidase II

Article abstract of DOI:10.1124/mol.55.1.179

Glutamate carboxypeptidase II (GCP II) catalyzes the extracellular hydrolysis of the neuromodulator N-acetyl-aspartylglutamate to N-acetyl- aspartate and glutamate. GCP II also hydrolyzes γ-glutamyl bonds in folylpolyglutamate. The predicted amino acid sequence of GCP II displays similarities to aminopeptidases from Streptomyces griseus and Vibrio proteolyticus, whose crystal structures have been determined. These aminopeptidases are cocatalytic zinc metallopeptidases belonging to the peptidase family M28, Specific zinc and substrate ligands have been proposed in GCP II based on the amino acid sequence alignment to these M28 family members. In the present study, site-directed mutagenesis has been used to test the assignment of these putative ligands in human GCP II. Substitutions to the five putative zinc ligands resulted in severely reduced enzyme activity, although mutant protein was expressed as demonstrated by immunoblot analysis. In addition, substitutions of amino acids near the putative zinc ligands have identified other specific residues important for enzyme structure and/or function. Substitutions to putative substrate ligands were less perturbing, and increases in K(m) were observed for substitutions that introduced a large charge perturbation (e.g., Lys to Glu). The results from substitutions at the proposed zinc and substrate ligands are consistent with the assignment of these residues and suggest that GCP II has a three- dimensional structure similar to other members of the peptidase family M28.

Full text of DOI:10.1124/mol.55.1.179

0026-895X/99/010179-07$3.00/0
Copyright © by The American Society for Pharmacology and Experimental Therapeutics
All rights of reproduction in any form reserved.
MOLECULAR PHARMACOLOGY, 55:179–185 (1999).
Site-Directed Mutagenesis of Predicted Active Site Residues in
Glutamate Carboxypeptidase II
HENRY S. SPENO, RUTH LUTHI-CARTER, WENDY L. MACIAS, STACEY L. VALENTINE, AMIT R. T. JOSHI, and
JOSEPH T. COYLE
Laboratory of Molecular and Developmental Neuroscience, Massachusetts General Hospital-East, Charlestown, Massachusetts
Received August 3, 1998; accepted October 19, 1998
This paper is available online at http://www.molpharm.org
ABSTRACT
Glutamate carboxypeptidase II (GCP II) catalyzes the extracel- II. Substitutions to the five putative zinc ligands resulted in
lular hydrolysis of the neuromodulator N-acetyl-aspartylgluta- severely reduced enzyme activity, although mutant protein was
mate to N-acetyl-aspartate and glutamate. GCP II also hydro- expressed as demonstrated by immunoblot analysis. In addi-
lyzes ␥-glutamyl bonds in folylpolyglutamate. The predicted tion, substitutions of amino acids near the putative zinc ligands
amino acid sequence of GCP II displays similarities to ami- have identified other specific residues important for enzyme
nopeptidases from Streptomyces griseus and Vibrio proteolyti- structure and/or function. Substitutions to putative substrate
cus, whose crystal structures have been determined. These ligands were less perturbing, and increases in Km were ob-
aminopeptidases are cocatalytic zinc metallopeptidases be- served for substitutions that introduced a large charge pertur-
longing to the peptidase family M28. Specific zinc and sub- bation (e.g., Lys to Glu). The results from substitutions at the
strate ligands have been proposed in GCP II based on the proposed zinc and substrate ligands are consistent with the
amino acid sequence alignment to these M28 family members. assignment of these residues and suggest that GCP II has a
In the present study, site-directed mutagenesis has been used three-dimensional structure similar to other members of the
to test the assignment of these putative ligands in human GCP peptidase family M28.
Glutamate carboxypeptidase II (GCP II) hydrolyzes the including nonmyelinating Schwann cells and the neuromus-
abundant neuropeptide N-acetyl-aspartylglutamate (NAAG) cular junction, small intestine, kidney, and prostate (Slusher
to N-acetyl-aspartate and glutamate in the extracellular et al., 1992; Berger et al., 1995a,b; Troyer et al., 1995a). At
space (Robinson et al., 1987; Stauch et al., 1989; Serval et al.,
1990; Slusher et al., 1990). NAAG inhibits glutamate’s acti-
vation of the N-methyl-D-aspartate ionotropic receptor
(Sekiguchi et al., 1989; Puttfarcken et al., 1993; Burlina et
al., 1994; Grunze et al., 1996) and is an agonist at the
metabotropic glutamate receptor mGluR3 (Wroblewska et
al., 1997). GCP II terminates the action of NAAG and gen-
erates extrasynaptic glutamate in the brain. Inhibition of
GCP II by 2-(phosphonomethyl)pentanedioic acid (PMPA),
has been shown to be effective in preventing neuronal cell
death in models of ischemia (Harukuni et al., 1997). GCP II
also hydrolyzes ␥-glutamyl linkages in pteroylpolyglutamate
(Pinto et al., 1996). Conversion of pteroylpolyglutamate to
pteroylmono-glutamate is necessary for intestinal absorption
of folic acid.
some of these peripheral sites, GCP II may act on substrates
other than NAAG. The name glutamate carboxypeptidase II
(EC 3.4.17.21) has been recently assigned; the enzyme has
previously been called N-acetyl-␣-linked acidic dipeptidase or
NAALADase and prostate-specific membrane antigen or
PSM. Molecular characterizations of rat, pig, and human
forms of GCP II have been described (Israeli et al., 1993;
Carter et al., 1996; Bzdega et al., 1997; Halsted et al., 1998;
Luthi-Carter et al., 1998).
GCP II is a class II membrane glycoprotein with an appar-
ent molecular mass of 94 to 100 kDa. Class II membrane
proteins have a short cytoplasmic amino terminus, a single
membrane-spanning domain, and a large extracellular do-
main (Fig. 1). A model for human GCP II has recently been
proposed by Rawlings and Barrett (1997) based on a region of
similarity to the peptidase family M28 comprised of cocata-
lytic metallopeptidases. The catalytic domain of GCP II (Fig.
1) has been identified by sequence similarities to low molec-
ular weight aminopeptidases from Streptomyces griseus and
Vibrio proteolyticus (also known as Aeromonas proteolytica).
The X-ray crystal structures have been determined for the
GCP II is also expressed in selected sites outside the brain
This work was supported by National Association for Research on Schizo-
phrenia and Affective Disorders Hilton Senior Investigator Award (to J.T.C.),
National Research Service Award fellowships from the National Institute of
Mental Health 5-T32-MH14275–21 (to B.K. Madras), and 1 F32 MH11895–01
(to H.S.S.).
ABBREVIATIONS: GCP II, glutamate carboxypeptidase II; NAAG, N-acetyl-aspartylglutamate; NEP, neutral endopeptidase; PMPA, 2-(phospho-
nomethyl)pentanedioic acid.
179

180
Speno et al.
Streptomyces and Vibrio aminopeptidases at 1.75-Å and sates were then incubated in the presence of the radiolabeled sub-
strate N-acetyl-L-aspartyl-L-[3,4-3H]glutamate (NEN Life Science
1.8-Å resolution, respectively (Chevrier et al., 1994; Green-
blatt et al., 1997). These peptidases have a binuclear Znϩϩ
Products, Boston, MA). Products were separated from intact sub-
strate using disposable anion exchange columns and L-[3,4-3H]glu-
center at the active site in which the two zinc ions share a
tamate measured by scintillation spectrometry as described in a
bridging carboxylate ligand. In addition, the structure of
study by Robinson et al. (1987). Enzymatic activity of GCP II in 2 ␮g
carboxypeptidase G2 has been determined at 2.5-Å resolution
of total protein from the cell lysate was assayed in a final volume of
and has a central topology quite similar to the Vibrio amino-
peptidase, even though no significant amino acid sequence
identity exists (Rowsell et al., 1997). Carboxypeptidase G2
hydrolyzes the ␣-linked glutamate from folic acid. The binu-
clear zinc-binding sites found in carboxypeptidase G2 and the
Vibrio aminopeptidase are highly similar in structure and
ligand environment (Table 1).
250 ␮l using 30 nM N-acetyl-L-aspartyl-L-[3,4-3H]glutamate in 50
mM Tris, pH 7.4, at 37°C, 0.1% Triton X-100, 150 ␮M potassium
phosphate, and 1 mM CoCl2 (Robinson et al., 1987). Total protein
was assayed using the bicinchoninic acid assay (Pierce, Rockford,
IL). To test for metal ion dependence of mutant enzyme activity,
activity was recovered by the addition of Coϩϩ or Znϩϩ in the
presence of EDTA (Robinson et al., 1987). Because all transfects
were expressed in the same cell system, it is assumed that differ-
ences in kinetic characteristics reflect primarily the transfected spe-
cies.
To test amino acid assignments, putative ligands for Znϩϩ
in GCP II have been targeted for amino acid substitution
based on the alignment of the amino acid sequence to the
Streptomyces and Vibrio aminopeptidases. The five putative
Znϩϩ ligands and their positions in GCP II are: His-377,
Asp-387, Glu-425, Asp-453, and His-553, where Asp-387 is
predicted to be the bridging ligand (Table 1). In contrast to
both structural and catalytic mono-zinc sites, carboxylate
oxygen coordination predominates in binuclear zinc sites
(Vallee and Auld, 1993).
The crystal structure of the Vibrio aminopeptidase shows
10 residues lining a hydrophobic specificity pocket implicated
in substrate binding (Chevrier et al., 1994). In the amino acid
sequence alignment, Rawlings and Barrett (1997) noted that
four of the residues that line the specificity pocket of the
Vibrio enzyme are substituted by positively charged residues
(i.e., Arg-463, Lys-499, Arg-536, and Lys-545), which may
account in part for substrate specificity in GCP II (Table 2).
These and some of the other 10 predicted substrate ligands,
a subset of which are next to putative zinc ligands, have been
targeted for amino acid substitution (Tables 1 and 2).
Kinetic Analysis. For those mutants that showed activity,
Michaelis-Menten (saturation) kinetics were performed. The Km and
Vmax parameters were determined by a nonlinear least-squares fit of
the initial velocity versus substrate concentration plot using Prism
software (GraphPad, San Diego, CA) where the error in these pa-
rameters is represented by the 95% confidence interval. At each
substrate concentration, three time points were taken either at 15,
30, and 45 min for wild type or up to 120, 180, and 240 min for
mutants within the linear range of the assay. Background counts
from the coelution of intact substrate were subtracted from each
point by including a sample without enzyme. The concentration of
the labeled substrate [3H]NAAG was kept constant at 30 nM,
whereas the concentration of unlabeled substrate was varied from
100 to 3000 nM. The initial velocity, v (in femtomoles per minute), at
a given substrate concentration was determined from the slope of the
linear plot of product versus time where the error in the initial
velocity is represented by the S.E.M. of the 95% confidence interval
from the linear regression of the slope. To compare changes in Vmax
between mutant and wild-type enzymes, the Vmax parameter has
been normalized by the ratio of wild-type to mutant protein from the
relative intensity of the signal measured from an immunoblot (see
below).
Materials and Methods
Inhibitor Analysis. Activity of selected mutants was measured
in the presence of varying concentrations of inhibitors in a volume of
125 ␮l using 2 ␮g of total protein, 1 mM CoCl2, and 10 nM
[3H]NAAG. PMPA was a gift from Dr. Barbara Slusher (Guilford
Pharmaceuticals, Baltimore, MD). Pteroylpenta-␥-glutamate was
purchased from B. Schircks Laboratories (Jona, Switzerland).
[3H]NAAG is shipped in phosphate buffer, however phosphate inhi-
bition of GCP II did not interfere with kinetic analysis at the con-
centration of [3H]NAAG used. To exchange the buffer for inhibitor
analysis and to purify N-acetyl-L-aspartyl-L-[3,4-3H]glutamate (NEN
Life Science Products), up to 250 ␮l of 24 ␮M [3H]NAAG was applied
to a 1-ml bed volume of anion exchange resin AG 1-X8 (Bio-Rad,
Hercules, CA), washed with 3 ml of 1 M formic acid, and eluted in 3
ml of 7.5 M formic acid. The sample was lyophilized and resuspended
in 3 ml of H2O and lyophilized two more times. The pellet was
resuspended in 200 mM Tris, pH 9.5, and if necessary, the pH
adjusted to 7.0 with NaOH.
Immunoblot Analysis. To normalize for the amount of enzyme
used between wild type and mutants in the kinetic assays, the
relative amount of GCP II protein was quantitated by immunoblot-
ting (Western blot analysis). Cellular lysates were separated by 7%
SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose
in 10 mM 3-cyclohexylamino-1-propane sulfonic acid, pH 11, 1%
methanol, and probed with the monoclonal antibody 7E11-C5 at 3
␮g/ml (a generous gift from Drs. Gerald Murphy and Alton Boynton)
(Israeli et al., 1994). The 7E11-C5 antibody recognizes the N termi-
nus of GCP II (Troyer et al., 1995b). Because no mutations were near
this epitope, amino acid substitutions were not expected to interfere
with binding of the antibody. Horseradish peroxide-conjugated sec-
Site-Directed Mutagenesis. The cDNA for human GCP II has
been subcloned into the mammalian expression vector, pcDNA3 (In-
vitrogen Corporation, Carlsbad, CA). Single-stranded template was
used in an in vitro mutagenesis reaction according to Kunkel et al.
(1991). When possible, restriction sites were designed into or out of
the mutation site to facilitate screening. Mutations were confirmed
by DNA sequencing.
Radioenzymatic Assay of Enzyme Activity. To test for alter-
ations in function of the encoded proteins, plasmids containing the
confirmed mutations were transiently transfected into the prostate
cancer cell line PC3 (American Type Culture Collection, Rockville,
MD), which does not express endogenous GCP II activity or protein
(Carter et al., 1996). Transfections were carried out following a
calcium phosphate method in 50 mM HEPES, pH 7.05 (Graham and
van der Eb, 1973). After incubation for 48 h, cells were harvested at
4°C and sonicated in the presence of 50 mM Tris, pH 7.4, 0.5% Triton
X-100 in high-pressure liquid chromatography grade H2O. Cell ly-
Fig. 1. Primary structure of GCP II. The individual predicted domains
are drawn to scale demonstrating a short cytoplasmic domain, a single
membrane spanning domain based on hydropathy profiles, and a large
extracellular domain with nine potential N-glycosylation sites each indi-
cated by a “T”. Amino acid positions are indicated below at the end of each
region. The catalytic domain is predicted by similarities to aminopepti-
dases from S. griseus and V. proteolyticus (Rawlings and Barrett, 1997).

Site-Directed Mutagenesis of Glutamate Carboxypeptidase II
181
ondary antibody (Jackson ImmunoResearch Laboratories, Inc., West region. In the Streptomyces and Vibrio aminopeptidases, the
Grove, PA) was used to visualize the complex in conjunction with the
chemiluminescent substrate SuperSignal ULTRA (Pierce). Data
were collected with a Fluor-S MultiImager (Bio-Rad) and analyzed
by Multi-Analyst software (Bio-Rad). To test that the immunoblot
was linear in the range of experimental samples, the intensities were
measured as a function of time. The amount of GCP II protein is
expressed per microgram of total protein loaded. For qualitative
analysis, immunoblots were visualized using film autoradiography.
Deglycosylation of Native Protein. To characterize the two
bands observed in the immunoblot of transiently expressed GCP II,
samples were deglycosylated under denaturing conditions with pep-
tide-N-glycosidase according to the manufacturer’s recommended
protocol (Oxford GlycoSciences, Inc., Bedford, MA) and probed with
7E11-C5 antibody in an immunoblot as described above.
corresponding residue was hydrogen bonded to the nearby
histidine zinc ligand. 2) Substitution of Pro-388 to Ala re-
sulted in a large increase in Km. It has been noted in binu-
clear zinc metallopeptidases that a cis-peptide bond occurs
near one of the zinc ligands (Table 1). 3) The E424Q mutant
resulted in a reduced Vmax value along with an increase in
Km (Fig. 3 and Table 4). A glutamate implicated in catalysis
is typically found next to one of the zinc ligands in both bi-
and mononuclear zinc metallopeptidases (Rowsell et al.,
1997). 4) The S454A mutant had a decrease in Vmax. 5) The
Y552F mutant had an increased Km and decreased Vmax. In
the Vibrio and Streptomyces aminopeptidases the corre-
sponding positions (Ser-454 and Tyr-552) are implicated as
comprising part of the substrate binding pocket indicating
proximity to the active site.
Results
Substitutions to Putative Substrate Ligands. In con-
trast to substitutions to putative zinc ligands, single substi-
tutions to putative substrate ligands were less disruptive to
enzyme activity so that kinetic parameters could be obtained
for most of the mutants. For the positively charged residues
implicated in determining substrate specificity, including
Lys-500, conservative substitutions that exchanged Lys for
Arg did not markedly alter the kinetic parameters (Table 4).
However, when the substitution introduced a large charge
perturbation (e.g., Lys to Glu), the Km for substrate of these
mutants increased greatly. This was seen for the R463I,
R536E, and K545E mutants (Table 4, an example of which is
shown in Fig. 3A). Immunoblot analysis indicates that the
expression of these mutants is like wild type except that the
upper band is faint for the R463I mutant (data not shown).
No activity was detected for the K500E mutant (after a 16-h
incubation), and reduced expression was observed in an im-
munoblot (Fig. 2A). Little change in the kinetic parameters
was seen for the K499E and S501A mutants, even though
they are directly next to Lys-500, which has no activity when
substituted to Glu. Very little activity was seen for the
N519D mutant after a 16-h incubation, so kinetic parameters
could not be obtained. Finally, the T538V mutant had a large
increase in the Km parameter (Table 4). Residues correspond-
ing to Asn-519 and Thr-538 have been implicated in contrib-
uting to the substrate binding pocket of the aminopeptidases.
Enzyme Activity in the Presence of Inhibitors. To
further analyze mutants that may affect substrate binding,
Substitutions to Putative Zinc Ligands. Based on
amino acid sequence alignment to other cocatalytic zinc pep-
tidases of the M28 family, the catalytic domain and poten-
tially important functional groups have been identified in
GCP II (Rawlings and Barrett, 1997). Site-directed mutagen-
esis has been used to test the assignment of specific amino
acid residues implicated in zinc and substrate binding. Single
substitutions to putative zinc ligands resulted in a profound
loss of enzyme activity (Table 3). The severe loss of activity in
these mutants is not due to the inability of transfectants to
express GCP II protein, as determined by immunoblot anal-
ysis (examples shown in Fig. 2A). However, reduced expres-
sion was observed for some of these mutants (Table 3). The
only mutant in which kinetic parameters were obtainable
was D387N, a substitution at the putative bridging ligand
that resulted in both an increased Km value and a decreased
Vmax when compared with wild type (Fig. 3 and Table 4). The
profile of activity versus metal ion concentration did not
change for the D387N and E424Q (see below) mutants when
compared with wild-type enzyme, suggesting that these sub-
stitutions did not disrupt metal binding (data not shown).
Substitutions to Residues Next to Putative Zinc Li-
gands. Substitutions to residues next to each of the five
putative zinc ligands were also targeted. These included Asp-
379, Pro-388, Glu-424, Ser-454, and Tyr-552. The following
results are shown in Tables 3 and 4. 1) Conservative substi-
tutions at Asp-379 resulted in no detectable enzyme activity
(after a 16-h incubation) and in a reduced expression of
protein, underscoring the importance of the structure in this [3H]NAAG hydrolysis was measured in the presence of in-
TABLE 1
Putative zinc ligands and nearby residues targeted for substitution in GCP II
Amino acids targeted for substitution were based on comparison with the following proteins of known structure: aminopeptidases from S. griseus and V. proteolyticus, and
carboxypeptidase G2, CPG2 (Greenblatt et al., 1997; Chevrier et al., 1994; Rowsell et al., 1997). Interactions of residues with Zn2ϩ, Hg2ϩ, H2PO4Ϫ, and hydroxamate that
were observed in the crystal structures are indicated under “Comments”. Based on the alignment, certain predictions can be made: Asp-379 is proposed to hydrogen bond
to the putative zinc ligand His-377. Asp-387 is the putative bridging ligand. Pro-388 is proposed to be involved in a cis-peptide bond with Asp-387. Glu-424 is predicted to
be involved in catalysis.
GCP II
Streptomyces
Vibrio
CPG2
Comments
His-377
Asp-379
Asp-387
Pro-388
Glu-424
Glu-425
Asp-453
Ser-454
Tyr-552
His-553
His-85
His-97
His-112
Asp-114
Asp-141
Asp-142
Glu-175
Glu-176
Glu-200
Pro-201
Tyr-384
His-385
Zn2ϩ/Hg2ϩ
Asp-87
Asp-97
Asn-98
Glu-131
Glu-132
Asp-160
Met-161
Tyr-246
His-247
Asp-99
H-bond
Asp-117
Asp-118
Glu-151
Glu-152
Asp-179
Met-180
Ile-255
Zn2ϩ/Hg2ϩ bridging
cis bond
H2PO4Ϫ/hydroxamate
Zn2ϩ
Zn2ϩ/Hg2ϩ
Ϫ
H2PO4
His-256
Zn2ϩ

182
Speno et al.
hibitors or the unlabeled alternative substrate, pteroylpenta- (Table 4). The Y552F mutant showed dramatic changes in
␥-glutamate (Pinto et al., 1996). The IC50 was determined for the IC50 for PMPA, phosphate, glutamate, and pteroylpenta-
pteroylpenta-␥-glutamate in the presence of [3H]NAAG for ␥-glutamate (Table 6), suggesting a role for this residue in
some of the mutants (Table 5). An 8- and 9-fold increase was binding these inhibitors and/or NAAG.
seen in the IC50 for the R536E and K545E mutants, respec-
Immunoblot Analysis. To determine whether mutant
tively (Table 5), consistent with their increased Km for NAAG protein was expressed, immunoblot analysis was performed.
Mutant protein was expressed even for those mutants that
showed no activity (Fig. 2A). Two bands observed in the
immunoblot suggest that two different species were present.
One explanation for the two bands is differential glycosyla-
tion. To test this, the lysate from cells transfected with wild-
type cDNA (i.e., expression of native enzyme) was deglycosy-
lated under denaturing conditions and run on a gel for
immunoblot analysis. One band was predominately seen at
TABLE 2
Putative substrate ligands targeted for substitution
Amino acids proposed to make up part of the specificity pocket in GCP II are based
on the corresponding residues observed in the crystal structure for the Vibrio
aminopeptidase (Chevrier et al., 1994). Assignment of residues in Streptomyces is
based on the superimposition and amino acid sequence alignment to the Vibrio
aminopeptidase (Greenblatt et al., 1997). The carbonyl oxygen of Arg-202 in Strep-
tomyces hydrogen bonds with the zinc bound H2PO4Ϫ. In contrast to putative zinc
ligands and adjacent residues that are indicated in Table 1, putative substrate
ligands are not conserved. See legend for Table 1 for description of Comments.
84 kDa, consistent with the predicted size of the GCP II
GCP II
Streptomyces
Vibrio
Comments
polypeptide (i.e., deglycosylated protein) (Fig. 2B). The hu-
man prostate cancer cell line LNCaP overexpresses GCP II
(Israeli et al., 1994). The LNCaP cell lysates expressed one
Arg-463
Lys-499
Lys-500
Ser-501
Asn-519
Arg-536
Thr-538
Lys-545
Asp-173
Glu-198
Ile-193
Cys-223
Asp-200
Arg-202
Gly-217
Phe-219
Gly-223
Tyr-225
Cys-227
Met-242
Phe-244
Phe-248
Hydroxamate
Ϫ
H2PO4
TABLE 3
Summary of substitutions to putative zinc ligands and nearby residues
Single substitutions are indicated along with the resultant enzyme activity using
[3H]NAAG as substrate; “no” indicates no detectable enzyme activity, even after
incubations of 16 h. Lower case “yes” indicates very little but detectable activity such
that reliable kinetic parameters could not be obtained. See Table 4 for kinetic
parameters for mutants indicated by “Yes”. Italics indicate slightly altered migration
observed in an immunoblot.
GCP II
Substitution/Activity
Comments
His-377 Gly no
Asp-379 Glu noa
Asp-387 Glu no
Pro-388 Ala Yes
Glu-424 Gln Yes
Glu-425 Gln noa
Ala noa
Asn noa
Asn Yes
Gln noa Zn2ϩ/Hg2ϩ
H-bond
Leu noa Zn2ϩ/Hg2ϩ bridging
cis bond
Asp yesa
Asp noa
PO24Ϫ/hydroxamate
Zn2ϩ
Asp-453 Glu yesa Asn noa
Leu noa Zn2ϩ/Hg2ϩ
Ser-454
Tyr-552 Phe Yes
His-553 Gly no
Ala Yes
PO24Ϫ
Ala no
Gln noa Zn2ϩ
a Indicates a low level of expression in an immunoblot, see Fig. 2A for examples.
See legend for Table 1 for description of comments.
Fig. 2. Immunoblot analysis. A, approximately 10 ␮g of total protein from
whole-cell lysates from representative mutants were loaded and com-
pared to native enzyme (WT). The region shown is from about 70 to 120
kDa. B, deglycosylation of native enzyme in whole-cell lysates compared
with the isotype expressed in LNCaP cells. Lane 1, LNCaP no treatment;
lane 2, LNCaP incubated without glycosidase; lane 3, LNCaP with gly-
cosidase; lane 4, transfected sample incubated without glycosidase; and
lane 5, transfected sample incubated with glycosidase.
Fig. 3. Michaelis-Menten kinetics. Examples are given of saturation
kinetics from whole cell lysates. Two micrograms of total protein was
assayed in the presence of a fixed amount of [3H]NAAG. Error bars for the
initial velocity represent the S.E.M. of the 95% confidence interval of the
slope of the product versus time curve. A, native enzyme (WT) and a
representative substitution to a residue proposed to bind substrate,
R463I. B, substitutions to the putative catalytic Glu-424, E424Q, and the
putative bridging ligand Asp-387, D387N.

Site-Directed Mutagenesis of Glutamate Carboxypeptidase II
183
7E11-C5 immunoreactive species that migrated at a position altered kinetic parameters (Table 4) and a 2-fold increase in
that is in between the two bands from the transfected sam- the IC50 for pteroylpenta-␥-glutamate (Table 5). In general,
ples. Upon deglycosylation this band migrated at the same anomalous band migration (e.g., D387L in Fig. 2A) in the
position as the deglycosylated transfected sample (Fig. 2B).
immunoblot, indicated by italics in Tables 3 and 4, correlated
The migration pattern of the two bands around 100 kDa in with no detectable activity, with the exception of N459D. For
the immunoblot gives an indication of the glycosylated state. those mutants for which kinetic parameters were obtained,
A band at 84 kDa would indicate a deglycosylated form. None the two bands seen in the immunoblot migrated with wild-
of the mutants showed a band around 84 kDa, suggesting type protein, suggesting that glycosylation had not drasti-
that all of the mutants are glycosylated (Fig. 2A). To test cally changed and that the affect of the mutation was the
whether glycosylation may influence enzyme activity, the result of the substitution and not due to a change in glyco-
putative glycosylation site Asn-459, which is near the puta- sylation.
tive Znϩϩ ligand Asp-453, was substituted to Asp. The two
bands observed in an immunoblot ran slightly faster than
wild-type protein, indicating an alteration in the glycosylated
Discussion
state (data not shown). The N459D mutant also had slightly
Site-directed mutagenesis has been performed in the hu-
man gene for GCP II to test assignments of specific amino
acids implicated in metal ligand binding, substrate binding,
and other structural and functional aspects of enzyme catal-
ysis. Assignments of these residues are based on the amino
acid sequence alignment of GCP II to low molecular weight
aminopeptidases whose crystal structures have been deter-
mined (Chevrier et al., 1994; Rawlings and Barrett, 1997;
Greenblatt et al., 1997). These aminopeptidases from S. gri-
seus and V. proteolyticus are cocatalytic zinc metallopepti-
dases that are members of the peptidase family M28. Zinc
ligands are conserved between these aminopeptidases and
GCP II (Table 1). However, putative substrate ligands are
not conserved (Table 2), consistent with a difference in sub-
strate specificity. All residues targeted for mutagenesis that
resulted in altered kinetic parameters, including putative
substrate ligands, are conserved among the human, rat, and
pig.
Substitutions to Putative Zinc Ligands. Single substi-
tutions have been designed at each of the five putative Znϩϩ
ligands (Table 3). A topological diagram, based on the Vibrio
aminopeptidase structure, is shown in Fig. 4 summarizing
the mutagenesis results and indicating the corresponding
ligands in GCP II. In general, substitutions to putative zinc
ligands are more disruptive to enzyme activity and stability
as detected by immunoblot analysis than are substitutions to
putative substrate ligands. It is not uncommon for apo-pro-
teins to be less stable than their holoenzyme. For example,
increased protease susceptibility was observed in the apo-
enzyme of a mutated zinc binding site of peptide deformylase
from Escherichia coli (Meinnel et al., 1995).
TABLE 4
Kinetic parameters obtained for mutants
The Michaelis-Menten parameters Km and Vmax for NAAG were determined by
saturation kinetics for mutants and compared with native enzyme wild type (WT),
see Fig. 3 for examples. Vmax has been adjusted by the relative intensities of protein
expression quantitated by an immunoblot by the ratio of WT to mutant. Italics
indicate slightly altered migration observed in an immunoblot.
Km
Vmax
nM
fmol/min
WT
330 (267–393)
101 (95–106)
Zn2ϩ ligands and nearby residues
D387N
P388A
1475 (670–2280)
Ͼ3000
24 (18–31)
ND
14 (13–16)
35 (22–37)
26 (19–33)
E424Q
650 (489–812)
S454A
Y552F
90 (61–119)
527 (196–934)
Substrate ligands
R463K
328 (188–468)
1580 (1410–1750)
193 (117–269)
471 (267–675)
ND
332 (224–440)
ND
266 (214–318)
104 (91–117)
90 (83–94)
118 (108–129)
142 (121–165)
ND
77 (70–83)
ND
84 (78–89)
ND
99 (90–108)
132 (129–135)
73 (65–80)
R463I†
K499E
K500R
K500Ea
S501A
N519D
R536K
R536E
Ͼ3000
T538V
1341 (1113–1569)
272 (250–294)
1166 (917–1415)
K545R
K545E
Glycosylation
N459D
387 (189–586)
50 (42–58)
a Indicates low level of expression in an immunoblot. ND, indicates not deter-
mined. The 95% confidence limits are presented in parentheses.
TABLE 5
IC50 values for pteroylpenta-␥-glutamate
The only GCP II mutant in which substantial activity was
observed is for the putative bridging ligand, D387N. The
altered kinetic parameters for the D387N mutant suggest
IC50 was determined for the alternative substrate, pteroylpenta-␥-glutamate for
selected mutants. IC50 value was determined by varying the concentration of inhib-
itor in the presence of 1 mM Co2ϩ and 10 nM [3H]NAAG where the 95% confidence
interval is shown in parentheses.
IC50
TABLE 6
nM
IC50 values for inhibition of the Y552F mutant
The potent inhibitor PMPA is a glutamate analog with a phosphate substitution at
WT
85 (56–127)
Zn vicinity
D387N
P388A
E424Q
S454A
Substrate
R463I
Ϫ
the amine position (Jackson et al., 1996). Phosphate (H2PO4 ) is a low potency
249 (76–818)
174 (56–544)
124 (76–204)
97 (68–139)
inhibitor. Pteroylpenta-␥-glutamate (PPG) is also hydrolyzed by GCP II (Pinto et al.,
1996). Glutamate (Glu) is a product of NAAG hydrolysis. ␤-NAAG is a nonhydrolyz-
able NAAG analog. Conditions are described in the legend for Table 5.
IC50
128 (95–173)
148 (110–200)
705 (368–1348)
86 (59–125)
Wild type
Y552F
K500R
R536E
T538V
K545E
N-Glycos
N459D
PMPA (nM)
0.9 (0.6–1.4)
0.5 (0.4–0.7)
85 (56–127)
77 (34–175)
1.2 (0.6–2.6)
75 (52–109)
9.0 (4.5–17.9)
19 (10–37)
15 (14–17)
2.3 (1.3–4.1)
Ϫ
H2PO4 (mM)
768 (606–973)
PPG (nM)
Glu (␮M)
␤-NAAG (␮M)
170 (94–305)

184
Speno et al.
that this important residue plays either an indirect or direct that Pro-388 plays an important structural role in the active
role in catalysis. Although there is as yet no direct evidence site of GCP II that most likely affects substrate binding
for the perturbation of metal coordination in mutants con- indirectly.
taining substitutions to putative metal ligands, the results
Substitution of Glu-424 to Gln resulted in a decrease in
are consistent with the observation that known zinc ligands Vmax suggesting a role for this carboxylate in catalysis. In the
are highly conserved in the sequence alignment of the pepti- Vibrio aminopeptidase, the corresponding glutamate (Glu-
dase family M28.
151) has been proposed to be a general base in the catalytic
Substitutions to Residues Next to Putative Zinc Li- mechanism and has been shown to be involved in binding to
gands. Amino acid residues next to metal ligands can affect the bridging H2O or phosphate or to a hydroxamate inhibitor
the local structure and/or function of the enzyme. In the (Table 1) (Chevrier et al., 1996). In the mononuclear car-
crystal structures of the Streptomyces and Vibrio aminopep- boxypeptidase A and thermolysin, a glutamate residue, im-
tidases, the corresponding Asp-379 is shown to hydrogen plicated in the catalytic mechanism, is in a HEXXH motif in
bond to His-377. There may be a requirement for the proper which the Glu is the putative catalytic residue and the two
positioning or electronic interaction of His-377 such that His are Znϩϩ ligands. Substitutions of Glu to Asp or Gln in
substitution of Asp-379 to Glu or Asn dramatically perturbs this motif in peptide deformylase and NEP resulted in de-
the local structure and/or function (Table 3). This triad fea- creased activity (Devault et al., 1988; Meinnel et al., 1995).
ture (e.g., Asp-His-Znϩϩ) is also found in other proteases
Both Ser-454 and Tyr-552 may comprise part of the sub-
such as the mononuclear zinc site of thermolysin in which strate-binding site because their corresponding residues in
both zinc-associated histidines are also coordinated by either the Vibrio aminopeptidase help define a specificity pocket. In
an Asn or Asp. It has been proposed that the active site of addition, these amino acids are directly next to putative zinc
neutral endopeptidase (NEP) has a similar structure to ther- ligands, and the corresponding Tyr-552 in the Streptomyces
molysin (Benchetrit et al., 1988). A substitution similar to aminopeptidase has been shown in the crystal structure to
the present case in NEP also decreased enzyme activity in hydrogen bond with phosphate that is bound between the two
which the corresponding Asp of the triad (i.e., Asp-650) was zinc ions. The reduced Vmax suggests that these residues, like
mutated to Glu, Asn, or Ala, (Le Moual et al., 1994).
Glu-424, are important for enzymatic activity. However, the
A nonproline cis-peptide bond that involves the backbone corresponding Ser-454 and Tyr-552 are not conserved (Table
of the bridging aspartic acid ligand is observed in the crystal 1). These substitutions may indirectly perturb the local struc-
structure of the Vibrio and Streptomyces aminopeptidases ture or reflect mechanistic differences unique to each partic-
and carboxypeptidase G2. As pointed out by Rawlings and ular peptidase. Interestingly, the large increase in the IC50
Barrett (1997), a proline, which is more likely to accommo- for PMPA or phosphate in the presence of NAAG may indi-
date a cis-peptide bond, is at this corresponding position in cate that the Tyr-552 hydroxyl is important for the binding of
GCP II next to the putative bridging zinc ligand Asp-387. The these inhibitors. PMPA is a glutamate analog with a phos-
large increase in Km (Ͼ3 ␮M) for substitution to Ala suggests phate substitution at the amine position (Jackson et al.,
1996). Tyr-552 may possibly bind to the phosphate function-
ality of PMPA because phosphate inhibition was also per-
turbed in the Y552F mutant.
Putative Substrate-Binding Determinants. Four posi-
tively charged residues in GCP II have been predicted by Raw-
lings and Barrett (1997) to be involved in the binding of the
negatively charged substrates. The corresponding residues in
the Vibrio aminopeptidase make up part of the hydrophobic
substrate specificity pocket. The present data support the as-
signment of these specificity determinants. Because many res-
idues are involved in substrate binding, it might be expected
that any one residue would contribute only partially to the
overall affinity for substrate. In this case, a single substitution
would not be expected to eliminate binding. The data support
this concept, because certain substitutions alter but do not
abolish activity. Only when the substitution was a large charge
perturbation, such as Lys to Glu, has a large change in Km been
observed. The only prediction not substantiated was for Lys-
499. However, this amino acid is not conserved in the rat or pig
sequence (Bzdega et al., 1997; Halsted et al., 1998; Luthi-Carter
et al., 1998). Interestingly, there is another Lys conserved in the
Fig. 4. Predicted topological diagram of the catalytic domain of GCP II.
rat and pig sequences directly next to Lys-499 at position 500
wherein substitution to Glu resulted in a loss of enzyme activ-
ity. Location of these putative substrate binding residues is
shown in Fig. 4. An exchange of Lys and Arg did not seem to
perturb substrate binding (Table 4), supporting the role of the
positively charged nature of these residues in binding. Altered
binding of pteroylpenta-␥-glutamate by some of these mutants
indicates that NAAG and folylpolyglutamate may share com-
The predicted catalytic domain shown is from residues 287 to 587. Re-
gions of secondary structure are shown when aligned to the Vibrio ami-
nopeptidase. Arrows indicate ␤-sheets and cylinders ␣-helices. A sum-
mary of substitutions that resulted in altered enzyme activity is
indicated: 1) The putative zinc ligands are circled with their correspond-
ing positions indicated. 2) Neighboring residues to the putative zinc
ligands are indicated in bold. 3) Positively charged residues thought to
bind substrate are indicated in bold and a plus sign. 4) Two other residues
implicated in substrate binding are Asn-519 and Thr-538. Asn-459 is one
of three putative glycosylation sites located in the catalytic domain.

Site-Directed Mutagenesis of Glutamate Carboxypeptidase II
185
(1998) Folylpoly-␥-glutamate carboxypeptidase from pig jejunum: Molecular char-
acterization and relation to glutamate carboxypeptidase II. J Biol Chem 273:
20417–20424.
Harukuni I, Bhardwaj A, Hurn PD, Slusher BS and Traystman RJ (1997) Novel
NAALADase inhibitor, PMPA, improves infarct volume in experimental stroke.
Soc Neurosci Abst 23:898.3.
mon binding determinants. Our data indicate that specific res-
idues, including positively charged ones, influence substrate
binding either directly or indirectly and are consistent with the
model proposed by Rawlings and Barrett (1997).
Identification of Important Structural and Functional
Elements through Evolutionary Conservation. Important
structural similarities are often identified by similarities be-
tween mammalian and bacterial proteins. Interesting examples
for which mutagenesis studies have been conducted include
both metabotropic and ionotropic glutamate receptors, such as
the N-methyl-D-aspartate receptor in which similarities to bac-
terial amino acid binding proteins have been observed (O’Hara
et al., 1993; Kuryatov et al., 1994; Stern-Bach et al., 1994; Paas
et al., 1996). As mentioned previously, structural similarities
have been identified between NEP, a type II membrane zinc
metallopeptidase, and the mononuclear zinc containing thermo-
lysin (Benchetrit et al., 1988). Mutagenesis analyses of NEP
support assignments made from the sequence alignment even
though little amino acid identity exists between the two se-
quences (Marie-Claire et al., 1997 and references therein).
These studies, like the current one, suggest that the three-
dimensional structure of the mammalian proteins are similar to
their bacterial counterparts and that structural and functional
elements can be predicted based on these comparisons. Lastly,
our data from the mutagenesis studies may provide useful
information to aid in the design of pharmacologic reagents.
Israeli RS, Powell CT, Corr JG, Fair WR and Heston WDW (1994) Expression of the
prostate-specific membrane antigen. Cancer Res 54:1807–1811.
Israeli RS, Powell CT, Fair WR and Heston WDW (1993) Molecular cloning of a
complementary DNA encoding a prostate-specific membrane antigen. Cancer Res
53:227–230.
Jackson PF, Cole DC, Slusher BS, Stetz SL, Ross LE, Donzanti BA and Trainor DA
(1996) Design, synthesis, and biological activity of a potent inhibitor of the neu-
ropeptidase N-acetylated ␣-linked acidic dipeptidase. J Med Chem 39:619–622.
Kunkel TA, Bebenek K and McClary J (1991) Efficient site-directed mutagenesis
using uracil-containing DNA. Methods Enzymol 204:125–139.
Kuryatov A, Laube B, Betz H and Kuhse J (1994) Mutational analysis of the
glycine-binding site of the NMDA receptor: Structural similarity with bacterial
amino acid-binding proteins. Neuron 12:1291–1300.
Le Moual H, Dion N, Roques BP, Crine P, and Boileau G (1994) Asp650 is crucial for
catalytic activity of neutral endopeptidase 24–11. Eur J Biochem 221:475–480.
Luthi-Carter R, Berger UV, Barczak AK, Enna M and Coyle JT (1998) Isolation and
expression of a rat brain cDNA encoding glutamate carboxypeptidase II. Proc Natl
Acad Sci USA 95:3215–3220.
Marie-Claire C, Ruffet E, Antonczak S, Beaumont A, O’Donohue M, Roques BP and
Fournie´-Zaluski M-C (1997) Evidence by site-directed mutagenesis that arginine
203 of thermolysin and arginine 717 of neprilysin (neutral endopeptidase) play
equivalent critical roles in substrate hydrolysis and inhibitor binding. Biochemis-
try 36:13938–13945.
Meinnel T, Lazennec C and Blanquet S (1995) Mapping of the active site zinc ligands
of peptide deformylase. J Mol Biol 254:175–183.
O’Hara PJ, Sheppard PO, Thøgersin H, Venezia D, Haldeman BA, McGrane V,
Houamed KM, Thomsen C, Gilbert TL and Mulvihill ER (1993) The ligand-binding
domain in metabotropic glutamate receptors is related to bacterial periplasmic
binding proteins. Neuron 11:41–52.
Paas Y, Eisenstein M, Medevielle F, Teichberg VI and Devillers-Thie´ry A (1996)
Identification of the amino acid subsets accounting for the ligand binding speci-
ficity of a glutamate receptor. Neuron 17:979–990.
Pinto JT, Suffoletto BP, Berzin TM, Qiao CH, Lin S, Tong WP, May F, Mukherjee B
and Heston WDW (1996) Prostate specific membrane (PSM) antigen: A novel
member of folate hydrolase in human prostatic carcinoma cells. Clin Cancer Res
2:1445–1451.
Puttfarcken PS, Montgomery D, Coyle JT and Werling LL (1993) N-acetyl-L-spartyl-
L-glutamate (NAAG) modulation of NMDA-stimulated [3H]norepinephrine release
from rat hippocampal slices. J Pharmacol Exp Ther 266:796–803.
Rawlings ND and Barrett AJ (1997) Structure of membrane glutamate carboxypep-
tidase. Biochim Biophys Acta 1339:247–252.
Robinson MB, Blakely RD, Couto R and Coyle JT (1987) Hydrolysis of the brain
dipeptide N-acetyl-L-aspartyl-L-glutamate: Identification and characterization of a
novel N-acetylated ␣-linked acidic dipeptidase activity from rat brain. J Biol Chem
262:14498–14506.
Rowsell S, Pauptit RA, Tucker AD, Melton RG, Blow DM and Brick P (1997) Crystal
structure of carboxypeptidase G2, a bacterial enzyme with applications in cancer
therapy. Structure 5:337–347.
Sekiguchi M, Okamoto K and Sakai Y (1989) Low-concentration N-acetylaspartyl-
glutamate suppresses the climbing fiber response of Purkinje cells in Guinea pig
cerebellar slices and the responses to excitatory amino acids of Xenopus laevis
oocytes injected with cerebellar mRNA. Brain Res 482:87–96.
Serval V, Barbeito L, Pittaluga A, Cheramy A, Lavielle S and Glowinski J (1990)
Competitive inhibition of N-acetylated-alpha-linked acidic dipeptidase activity by
N-acetyl-L-aspartyl-␤-linked-L-glutamate. J Neurochem 55:39–46.
Slusher BS, Robinson MB, Tsai G, Simmons M, Richards SS and Coyle JT (1990) Rat
brain N-acetylated alpha-linked acidic dipeptidase activity: Purification and im-
munologic characterization. J Biol Chem 265:21297–21301.
Slusher BS, Tsai G, Yoo G and Coyle JT (1992) Immunocytochemical localization of
the N-acetyl-aspartyl-glutamate (NAAG) hydrolyzing enzyme N-acetylated
␣-linked acidic dipeptidase (NAALADase). J Comp Neurol 315:217–229.
Stauch BL, Robinson MB, Forloni G, Tsai G and Coyle JT (1989) The effects of
N-acetylated alpha-linked acidic dipeptidase (NAALADase) inhibitors on
[3H]NAAG catabolism in vivo. Neurocsci Lett 100:295–300.
Stern-Bach Y, Bettler B, Hartley M, Sheppard PO, O’Hara PJ and Heinemann SF
(1994) Agonist selectivity of glutamate receptors is specified by two domains
structurally related to bacterial amino acid-binding proteins. Neuron 13:1345–
1357.
Acknowledgments
We thank Drs. Gerald Murphy and Alton Boynton for the mono-
clonal antibody 7E11-C5, Drs. Rudy Tanzi and Robert Moir for
assistance in quantitating immunoblots, Dr. Craig Martin for dis-
cussion on kinetic analysis, and Drs. Neil Rawlings and Alan Barrett
for review of the manuscript.
References
Benchetrit T, Bissery V, Mornon JP, Devault A, Crine P and Roques BP (1988)
Primary structure homologies between two zinc metallopeptidases, the neutral
endopeptidase 24.11 (“enkephalinase”) and thermolysin, through clustering anal-
ysis. Biochemistry 27:592–596.
Berger UV, Carter RE, McKee M and Coyle JT (1995a) N-acetylated alpha-linked
acidic dipeptidase is expressed by non-myelinating Schwann cells in the peripheral
nervous system. J Neurocytol 24:99–109.
Berger UV, Carter RE and Coyle JT (1995b) The immunocytochemical localization of
N-acetylaspartylglutamate, its hydrolyzing enzyme NAALADase, and the
NMDAR1 receptor at a vertebrate neuromuscular junction. Neuroscience 64:847–
850.
Burlina AP, Skaper SD, Mazza MR, Ferrari V, Leon AL and Burlina AB (1994)
N-acetylaspartylglutamate selectively inhibits neuronal responses to N-methyl-D-
aspartic acid in vitro. J Neurochem 63:1174–1177.
Bzdega T, Turi T, Wroblewska B, She D, Chung HS, Kim H and Neale JH (1997)
Molecular cloning of a peptidase against N-acetylaspartylglutamate from a rat
hippocampal cDNA library. J Neurochem 69:2270–2277.
Carter RE, Feldman AR and Coyle JT (1996) Prostrate-specific membrane antigen is
a hydrolase with substrate and pharmacologic characteristics of a neuropeptidase.
Proc Natl Acad Sci USA 93:749–753.
Chevrier B, D’Orchymont H, Schalk C, Tarnus C and Moras D (1996) The structure
of the Aeromonas proteolytica aminopeptidase complexed with a hydroxamate
inhibitor: Involvement in catalysis of Glu-151 and two zinc ions of the co-catalytic
unit. Eur J Biochem 237:393–398.
Chevrier B, Schalk C, D’Orchymont H, Rondeau J-M, Moras D and Tarnus C (1994)
Crystal structure of Aeromonas proteolytica aminopeptidase: A prototypical mem-
ber of the co-catalytic zinc enzyme family. Structure 2:283–291.
Troyer JK, Beckett ML and Wright GL Jr (1995a) Detection and characterization of
the prostate-specific membrane antigen (PSMA) in tissue extracts and body fluids.
Int J Cancer 62:552–558.
Troyer JK, Feng Q, Beckett ML and Wright GL Jr (1995b) Biochemical character-
ization of the 7E11–C5.3 epitope of the prostate-specific membrane antigen. Urol
Oncol 1:29–37.
Vallee BL and Auld DS (1993) Cocatalytic zinc motifs in enzyme catalysis. Proc Natl
Acad Sci USA 90:2715–2718.
Wroblewska B, Wroblewski JT, Pshenichkin S, Surin A, Sullivan SE and Neale JH
(1997) N-acetylaspartylglutamate selectively activates mGluR3 receptors in trans-
fected cells. J Neurochem 69:174–181.
Devault A, Nault C, Zollinger M, Fournie´-Zaluski M-C, Roques BP, Crine P and
Boileau G (1988) Expression of neutral endopeptidase (enkephalinase) in heterol-
ogous COS-1 cells: Characterization of the recombinant enzyme and evidence for a
glutamic acid residue at the active site. J Biol Chem 263:4033–4040.
Graham FL and van der Eb AJ (1973) A new technique for the assay of infectivity of
human adenovirus 5 DNA. Virology 52:456–467.
Greenblatt HM, Almog O, Maras B, Spungin-Bialik A, Barra D, Blumberg S and
Shoham G (1997) Streptomyces griseus aminopeptidase: X-ray crystallographic
structure at 1.75 Å resolution. J Mol Biol 265:620–636.
Grunze HCR, Rainnie DG, Hasselmo ME, Barkai E, Hearn EF, McCarley RW and
Greene RW (1996) NMDA-dependent modulation of CA1 local circuit inhibition.
J Neurosci 16:2034–2043.
Send reprint requests to: Dr. Joseph Coyle, McLean Hospital, 115 Mill St.,
Belmont, MA 02178. E-mail: jcoyle@warren.med.harvard.edu
Halsted CH, Ling E-H, Luthi-Carter R, Villanueva JA, Gardner JM and Coyle JT