Comparison of human glutamate carboxypeptidases II and III reveals their divergent substrate specificities

Article abstract of DOI:10.1111/febs.13761

Glutamate carboxypeptidase III (GCPIII) is best known as a homologue of glutamate carboxypeptidase II [GCPII; also known as prostate-specific membrane antigen (PSMA)], a protease involved in neurological disorders and overexpressed in a number of solid cancers. However, mouse GCPIII was recently shown to cleave β-citrylglutamate (BCG), suggesting that these two closely related enzymes have distinct functions. To develop a tool to dissect, evaluate and quantify the activities of human GCPII and GCPIII, we analysed the catalytic efficiencies of these enzymes towards three physiological substrates. We observed a high efficiency of BCG cleavage by GCPIII but not GCPII. We also identified a strong modulation of GCPIII enzymatic activity by divalent cations, while we did not observe this effect for GCPII. Additionally, we used X-ray crystallography and computational modelling (quantum and molecular mechanical calculations) to describe the mechanism of BCG binding to the active sites of GCPII and GCPIII, respectively. Finally, we took advantage of the substantial differences in the enzymatic efficiencies of GCPII and GCPIII towards their substrates, using enzymatic assays for specific detection of these proteins in human tissues. Our findings suggest that GCPIII may not act merely as a complementary enzyme to GCPII, and it more likely possesses a specific physiological function related to BCG metabolism in the human body. Database: The X-ray structure of GCPII Glu424Ala in complex with BCG has been deposited in the RCSB Protein Data Bank under accession code 5F09.

Full text of DOI:10.1111/febs.13761

Comparison of human glutamate carboxypeptidases II and
III reveals their divergent substrate specificities
1,2
1
1,2
Michal Navratil1,2, Jan Tykvart1,2, Jirı Schimer , Petr Pachl , Vaclav Navratil
ꢁꢀ
,
ꢀ
ꢀ
ꢀ
3
1
1,2
ꢀ
ꢀ
ꢀ1,2
ꢀ
ꢀꢁ
Tibor Andras Rokob , Klara Hlouchova , Lubomır Rulısek and Jan Konvalinka
1 Institute of Organic Chemistry and Biochemistry, Gilead Sciences and IOCB Research Centre, Academy of Sciences of the Czech
Republic, Prague, Czech Republic
2 Department of Biochemistry, Faculty of Natural Sciences, Charles University in Prague, Czech Republic
3 Institute of Organic Chemistry, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary
Keywords
Glutamate carboxypeptidase III (GCPIII) is best known as a homologue
of glutamate carboxypeptidase II [GCPII; also known as prostate-specific
arene-binding site; GCPIII; prostate-specific
membrane antigen; QM/MM calculations;
membrane antigen (PSMA)], a protease involved in neurological disorders
and overexpressed in a number of solid cancers. However, mouse GCPIII
b-citryl-L-glutamate
was recently shown to cleave b-citrylglutamate (BCG), suggesting that
Correspondence
these two closely related enzymes have distinct functions. To develop a tool
to dissect, evaluate and quantify the activities of human GCPII and
GCPIII, we analysed the catalytic efficiencies of these enzymes towards
three physiological substrates. We observed a high efficiency of BCG cleav-
age by GCPIII but not GCPII. We also identified a strong modulation of
GCPIII enzymatic activity by divalent cations, while we did not observe
this effect for GCPII. Additionally, we used X-ray crystallography and
computational modelling (quantum and molecular mechanical calculations)
to describe the mechanism of BCG binding to the active sites of GCPII
and GCPIII, respectively. Finally, we took advantage of the substantial dif-
ferences in the enzymatic efficiencies of GCPII and GCPIII towards their
substrates, using enzymatic assays for specific detection of these proteins in
human tissues. Our findings suggest that GCPIII may not act merely as a
complementary enzyme to GCPII, and it more likely possesses a specific
physiological function related to BCG metabolism in the human body.
J. Konvalinka, Institute of Organic
Chemistry and Biochemistry, Gilead
Sciences and IOCB Research Centre,
Academy of Sciences of the Czech
Republic, Flemingovo n. 2, 166 10 Prague,
Czech Republic
Fax: +420 220 183 578
Tel: +420 220 183 218
E-mail: konval@uochb.cas.cz
(Received 10 March 2016, revised 25 April
2016, accepted 18 May 2016)
doi:10.1111/febs.13761
Database
The X-ray structure of GCPII Glu424Ala in complex with BCG has been deposited in the
RCSB Protein Data Bank under accession code 5F09.
Abbreviations
3D, three-dimensional; AAS, atomic absorption spectroscopy; ABS, arene-binding site; ACN, acetonitrile; BCG, b-citryl-L-glutamic acid (b-
citrylglutamate, beta-citrylglutamate, beta-citryl-L-glutamic acid, beta-citryl-L-glutamate); BSA, bovine serum albumin; C12E8, octaethylene
glycol monododecyl ether; DBU, 1,8-Diazabicycloundec-7-ene; DIEA, N,N-Diisopropylethylamine; EtOAc, ethylacetate; FolGlun, folyl-n-c-L-
glutamic acid; G3PDH, glyceraldehyde 3-phosphate dehydrogenase; GCPII, glutamate carboxypeptidase II; GCPIII, glutamate
carboxypeptidase III; HPLC, high-performance liquid chromatography; HRMS, high-resolution mass spectrometry; MeOH, methanol; MD/
MM, molecular dynamical/molecular mechanical calculations; NAAG, N-acetyl-L-aspartyl-L-glutamic acid; NMR, nuclear magnetic resonance;
OPA, orthophtalaldehyde; ORF, open reading frame; Pd(C), palladium on activated charcoal; PSMA, prostate-specific membrane antigen;
QM/MM, quantum mechanical and molecular mechanical calculations; QM/MM/MD, quantum mechanical, molecular mechanical and
molecular dynamical calculations; qPCR, quantitative polymerase chain reaction; rhGCPII, recombinant human glutamate carboxypeptidase II;
SDS/PAGE, sodium dodecylsulfate polyacrylamide gel electrophoresis; TFA, trifluoroacetic acid; TLC, thin-layer chromatography.
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
1

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
calcium or manganese ions facilitates cleavage of
BCG. The researchers hypothesized that the ability of
GCPIII to cleave BCG in the presence of specific
metal ions may be caused by replacement of one of its
two active site zinc ions by these metals [13].
Introduction
Glutamate carboxypeptidase III (GCPIII) is a binu-
clear zinc metallopeptidase that belongs to the MER-
OPS M28B peptidase subfamily (ID M28.012) [1].
GCPIII shares 67% sequence identity and 81%
sequence similarity with glutamate carboxypeptidase II
(GCPII, EC 3.4.17.21), one of the most well-character-
ized members of the M28B subfamily [2]. Additionally,
GCPIII adopts an almost identical 3D structure and
possesses very similar enzymatic activity to GCPII [3].
It is also predicted to be a type II transmembrane pro-
tein with amino acids 9–31 forming the transmem-
brane part of the structure [4] and its protease domain
facing the extracellular milieu.
BCG was first identified as a physiologically relevant
molecule in the 1970s, when it was isolated from new-
born rat brains [14]. Additionally, BCG-hydrolysing
activity, as well as BCG itself, has been detected in rat
testes [15,16]. Although BCG has been known for sev-
eral decades, its precise physiological function remains
unclear. One hypothesis suggested that BCG may serve
as an endogenous low-molecular-weight chelator of
iron or other biogenic metals [17]. However, the recent
unambiguous assignment of BCG-hydrolysing activity
to GCPIII increased the potential for elucidation of
BCG’s physiological role. Furthermore, the identifica-
tion of GCPIII as a BCG hydrolase suggests that
GCPIII may not function merely as a compensatory
enzyme for GCPII, but may play a different, physio-
logically relevant role in humans.
The physiological functions of GCPII include degra-
dation of N-acetyl-L-aspartyl-L-glutamate (NAAG,
Fig. 1), the most abundant peptide neurotransmitter in
the brain [3,5], and processing of polyglutamylated
folates (FolGlun, Fig. 1), vitamin B9 precursors,
enabling the absorption of free folate in the small
intestine [6,7]. Furthermore, GCPII, also known as
prostate-specific membrane antigen (PSMA), has been
investigated as a promising prostate cancer marker
[8,9] and an interesting molecular address for specific
anticancer drug delivery [10,11]. Because both GCPII
and GCPIII have similar enzymatic activities and there
is no specific monoclonal antibody against GCPIII
[12], unravelling the distinct physiological functions of
GCPII and GCPIII has been challenging.
We have shown previously that the GCPII tissue
expression profile differs in human and animal models,
which might have implications for its specific roles
[18]. The findings concerning GCPIII/BCG metabolism
have been obtained using animal models, mostly mice
or rats. GCPIII tissue distribution in the human body
is unknown, and it is crucial to validate and properly
characterize this novel metal-dependent enzymatic
activity for human GCPIII. Therefore, we set out to
perform a careful comparative analysis of human
GCPII and GCPIII activities towards their physiologi-
cal substrates NAAG, BCG and polyglutamylated
folates. Additionally, we describe the binding of BCG
into the active sites of GCPII and GCPIII using X-ray
crystallography and high-level (QM/MM) molecular
modelling, respectively. Finally, employing the enzy-
mological data, we were able to specifically detect
GCPII and GCPIII protein levels in human tissues, a
feat that had been impossible to achieve so far.
Until recently, GCPIII was recognized mainly as a
potentially complementary enzyme to GCPII. How-
ever, in 2011, Collard et al. [13] identified a novel
physiological substrate for mouse GCPIII, b-citryl-L-
glutamic acid (BCG, Fig. 1), which is cleaved to citrate
and L-glutamate by GCPIII but not GCPII. Interest-
ingly, the study also showed that the efficiency of
NAAG vs. BCG cleavage by GCPIII is metal-
dependent. The presence of zinc or manganese ions
facilitates NAAG cleavage, and the presence of
Fig. 1. Chemical structures of investigated substrates. From left to right: N-acetyl-L-aspartyl-L-glutamic acid (NAAG), n-c-L-glutamylated folic
acid (FolGlun) and b-citryl-L-glutamic acid (BCG).
2
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
Specific molecular recognition of these two close
enzyme homologues will enable the use of GCPII as a
selective molecular address for the anticancer drug
delivery and to validate its potential for the diagnosis
and prognosis of prostate cancer.
to five orders of magnitude greater [Fig. 2A, kcat/KM
ꢀ1
Ca2+) = 1.33 9 107 sꢀ1ꢁMꢀ1]. Additionally, we found
that the catalytic efficiency of human GCPIII is highly
metal-dependent. The presence of calcium ions
(GCPII) = 1.7 9 102 sꢀ1ꢁM
vs. kcat/KM (GCPIII,
increased the BCG cleavage efficiency by two orders of
ꢀ1
magnitude [kcat/KM (GCPIII) = 1.8 9 105 sꢀ1ꢁM
vs.
kcat/KM (GCPIII, Ca2+) = 1.33 9 107 sꢀ1ꢁMꢀ1], while
Results
zinc ions facilitated NAAG and FolGlu1 cleavage [kcat
ꢀ1
/
BCG hydrolysis by GCPIII depends on divalent
metal cations
KM (GCPIII, NAAG) = 2.16 9 105 sꢀ1ꢁM
vs. kcat
/
KM (GCPIII, NAAG, Zn2+) = 2.27 9 106 sꢀ1ꢁMꢀ1].
Manganese ions seemed to increase GCPIII activity
towards all tested substrates. The activation effect of
these divalent cations is mainly driven by an increase in
kcat values, while the KM values are less affected
(Fig. 2B,C). Importantly, we also tested the ability of
these cations to activate GCPII, and we did not observe
any notable metal-dependent activity (data not shown).
These data suggest that in the presence of calcium
cations, GCPII and GCPIII can be specifically
detected based on their different enzymatic activities.
Using recombinant human GCPII and GCPIII [19],
we analysed hydrolysis of three physiological sub-
strates in the presence or absence of calcium, man-
ganese(II) and zinc cations. The chemical structures of
the tested compounds, NAAG, FolGlun and BCG are
shown in Fig. 1.
The enzymological data, summarized in Fig. 2, indi-
cate that BCG is selectively hydrolysed by human
GCPIII. Although it is also cleaved by GCPII, the cat-
alytic efficiency for the GCPIII-catalysed reaction is up
Fig. 2. Characterization of BCG-, NAAG-
and FolGlu1-hydrolysing activity of GCPIII in
the absence or presence of divalent cations
(Ca2+, Mn2+ or Zn2+) and comparison to
GCPII. The reaction buffer was 25 mM
bistrispropane, pH 7.5, supplemented with
the cations where specified (Ca2+ 2.5 mM,
Mn2+ 0.25 mM and Zn2+ 0.10 mM). Error
bars represent SD. KM and kcat values were
obtained by nonweighted hyperbolic fit into
a single saturation curve consisting of 6–11
data points (n = 1). The panels show
comparisons of (A) catalytic efficiencies, (B)
kcat values and (C) KM values of GCPIII and
GCPII towards the panel of substrates BCG,
NAAG and FolGlu1 in the presence or
absence of calcium, manganese and zinc
ions. Asterisks indicate that the KM value
was lower than the lowest substrate
concentration (see Materials and methods
for more information).
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
3

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
GCPII processes NAAG or FolGlu1 with a two order
of magnitude higher catalytic efficiency than GCPIII,
while GCPIII processes BCG with a five order of mag-
nitude higher catalytic efficiency than GCPII.
The GCPII arene-binding site facilitates
processing of polyglutamylated folates
The GCPII structure includes an exosite called the
arene-binding site (ABS), which comprises Arg463,
Arg511 and Trp541 and is involved in the binding of
the pteroate moiety during the cleavage of glutamy-
lated folates [7,20]. A similar feature has not been
described in the structure of GCPIII. We investigated
the catalytic efficiency of GCPIII towards polyglu-
tamylated folates (FolGlu1–6), which bind to the ABS
of GCPII, and compared it with previous data on
GCPII cleavage efficiency of FolGlu1–6 [7]. As shown
in Fig. 3, GCPIII processes FolGlu2–6 with much
lower catalytic efficiency than GCPII, while processing
of monoglutamylated folate (FolGlu1) is comparable
between the two enzymes (Fig. 3A). The difference
resides mainly in the higher KM values for GCPIII
(Fig. 3B,C). These results are in good agreement with
previously reported structural data showing that
GCPII is unable to fully utilize the ABS for binding of
the pteroate moiety of FolGlu1, while the FolGlu2–3
substrates are able to bind into this exosite [7].
Occupancy of active site zinc ions is lower for
GCPIII than GCPII
The previously reported X-ray structures of GCPIII
suggested that the active site Zn1 and Zn2 atoms have
only partial occupancies in the 0.80–0.95 and 0.45–
0.80 ranges, respectively [21]. To further corroborate
these findings, we applied an atomic absorption spec-
troscopy technique to GCPII and GCPIII to determine
the precise amount of zinc atoms per molecule (for
experimental details, see Materials and methods). We
determined that GCPII contains 1.40 zinc atoms per
molecule and GCPIII contains 1.13. These results fur-
ther support the partial occupancy of Zn2 within the
GCPIII structure.
BCG binding mode in the GCPII active site: X-ray
structure
Fig. 3. Comparison of the FolGlu1–6-hydrolysing activity of GCPIII
and GCPII: (A) catalytic efficiencies, (B) kcat values and (C) KM
values. The reaction buffer was 25 mM Tris/HCl, pH 7.5,
without any added cations. Data for cleavage of FolGlu1–6 by
GCPII were measured previously [7]. Error bars represent SD.
KM and kcat values were obtained by nonweighted hyperbolic fit
To elucidate the differences in BCG cleavage by
GCPII and GCPIII, we prepared a diffraction-quality
crystal of inactive GCPII Glu424Ala in complex with
BCG. We solved the structure by molecular replace-
into
a single saturation curve consisting of 6–10 data points
ꢂ
ment and refined the final model to 1.85 A resolution
(n = 1).
4
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
Table 1. Data collection and refinement statistics.
structure, the glutamate moiety shows weaker electron
density (see Fig. 4B) than in crystal structures of com-
plexes with better substrates (e.g. FolGlun or NAAG
[7,22]). Interestingly, the glutarate sensor adopts two
equally populated conformations, the ‘closed’ and
‘open’ form (see Fig. 4C). Unlike the glutamate moi-
ety, the citrate moiety of BCG forms only one identifi-
able interaction with the enzyme: a salt bridge between
one of the citrate carboxyl groups and the guanidine
group of Arg534. This lack of interaction in the S1
pocket may be responsible for the substantially lower
affinity of BCG towards GCPII, compared to that of
other substrates such as NAAG. Accordingly, the
citrate moiety of BCG had much less well-defined elec-
tron density than the glutamate moiety (see Fig. 4B),
suggesting that this portion of BCG is not present in
one particular conformation within the GCPII active
site and likely possesses some degree of rotational free-
dom. These findings suggest that the binding of BCG,
and in particular its citrate moiety, into the GCPII
active site might partially prevent the closing of the
glutarate sensor. Finally, it should be noted that we
refined BCG to 70% occupancy, which is yet another
indication that the binding affinity of BCG for GCPII
is poor.
Compound
BCG
5F09
PDB ID
Data collection statistics
Space group
I222
Cell parameters
ꢂ
a (A)
ꢂ
b (A)
100.899
130.916
159.118
90,90,90
1
ꢂ
c (A)
a, b, c (°)
No. of molecules in AU
ꢂ
Wavelength (A)
ꢂ
Resolution (A)
ꢂ
Highest resolution shell (A)
0.918
1.85
1.96–1.85
89 051 (13 967)
4.51 (4.43)
98.8 (97.1)
8.1 (94.1)
13.29 (1.96)
39.43
No. of unique refl.
Multiplicity
Completeness (%)
Rmeas (%)a
Average I/r (I)
2 b
ꢂ
Wilson B (A )
Refinement statistics
ꢂ
Resolution range (A)
ꢂ
Highest resolution shell (A)
30.00–1.85
1.90–1.85
84 147 (4444)
6105 (322)
0.155 (0.311)
0.181 (0.336)
0.014
No. of refl. in working set
No. of refl. in test set
R value (%)c
Rfree value (%)d
ꢂ
RMSD bond length (A)
RMSD angle (°)
1.690
No. of atoms in AU
6402
BCG binding mode in the GCPIII active site: QM/
MM calculations
No. of protein atoms in AU
No. of ligand atoms in AU
No. of ion atoms in AU
No. of water molecules in AU
5804
22
The technical difficulties of obtaining diffraction-qual-
ity crystals for GCPIII in complex with BCG led us to
approximate the structure of GCPIII in complex with
BCG by quantum mechanics/molecular mechanics
4
562
2
ꢂ
Mean B value (A )
21.91
Ramachandran plot statisticse
Res. in favoured regions (%)
Res. in allowed regions (%)
Outliers
97.2
2.7
(QM/MM). We computed
a QM/MM homology
model of GCPIII in complex with BCG (see Fig. 5A).
The comparison of the obtained GCPII and GCPIII
structures in complex with BCG suggests that the
overall binding mode of this substrate is very similar
in both enzymes (see Fig. 5B). However, in contrast to
the flexibility of the BCG citrate moiety in the struc-
ture of the GCPII-BCG complex, both citrate carboxy-
lates are coordinated – specifically by Ser444, Arg524
and Arg526 in the QM/MM model of GCPIII. Addi-
tionally, the orientation of the BCG hydroxyl group in
GCPIII enables its coordination by Gly508. In the
GCPII-BCG structure, the orientation of this hydroxyl
group is different, and prevents such an interaction.
Val382
a
The data in parentheses denote the highest resolution shell. Rmeas
is the redundancy-independent merging R factor [25]. bEstimated by
SFcheck [26,27]. cR value = ||Fo| ꢀ |Fc||/|Fo|, where Fo and Fc are the
d
observed and calculated structure factors, respectively. Rfree is
equivalent to R value but is calculated for 5% of the reflections cho-
sen at random and omitted from the refinement process. eAs calcu-
lated by MolProbity [28,29]. The AviTEV-tag and first 11 amino acids
from the extracellular part of GCPII are not visible in the structure.
(PDB ID 5F09; data collection and refinement statis-
tics are shown in Table 1). The binding mode of BCG
within GCPII is depicted in Fig. 4A. The glutamate
moiety of BCG binds into the S10 pocket of GCPII in
an identical manner as described for FolGlun or
NAAG substrates [7,22]. It forms hydrogen bonds
with the side chains of Arg210, Asn257 and Tyr552,
and also with Lys699 and Tyr700 forming the so-
called ‘glutarate sensor’ [23]. However, in this
Thermodymamics of Zn2+/Ca2+ binding in the
active site: QM/MM calculations
After obtaining plausible QM/MM models of GCPIII
in complex with NAAG and BCG, it is
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
5

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
Fig. 4. X-ray structure of GCPII Glu424Ala
in complex with b-citryl-L-glutamic acid
(BCG, PDB ID 5F09). Images are rendered
in cross-eye stereo representation. Active
site zinc ions are shown in violet, GCPII
protein in yellow and BCG in pale cyan. In
panel (A), Tyr700 is truncated to provide an
unobscured view. Distances are shown in
ꢂ
A. Images were created using PYMOL [24].
(A) Detail of the active site of GCPII with
bound BCG. (B) The electron density map
2Fo ꢀ Fc for BCG in the GCPII-BCG
complex (PDB ID 5F09), contoured at 1.0r.
(C) The structure of GCPII in complex with
BCG features the glutarate sensor in two
conformations, closed (yellow) and open
(brown). Glutarate sensor is depicted as
main chain only except for Lys699 and
Tyr700 (only stubs are shown for the open
conformation).
computationally straightforward to address the possi-
bility of replacing one Zn2+ in the active site with a
Ca2+ ion. This hypothesis has been put forward previ-
ously [13], and here, we tested it using advanced QM/
MM modelling, which is arguably more rigorous and
accurate than MD/MM simulations (or even less elab-
orate models [13]). We surmised that the QM/MM
equilibrium structures and associated energetics may
provide structural definition of potential calcium bind-
ing in the GCPII/III active site as well as a clue to
explaining the observed differences in catalytic effi-
ciency of GCPII and GCPIII in the absence or pres-
ence of Ca2+. We carried out 12 QM/MM simulations
varying the enzyme (GCPII and GCPIII), substrate
(NAAG and BCG) and active site metals (Zn1–Zn2,
Ca1–Zn2, Zn1–Ca2). Eight of the QM/MM equilib-
rium geometries (optimized structures, Zn1–Zn2,
Zn1–Ca2) for the QM part (~ 280 atoms comprising
the active site) are deposited in Data S1. In brief,
substitution of Ca2+ for Zn2+ in the Zn2 position
(which is in all cases favoured by approximately
10–15 kcalꢁmolꢀ1 compared to Ca1-Zn2 structures)
induces several structural perturbations to the active
site (see Fig. 6). In comparison with the X-ray struc-
ture of GCPII in complex with BCG (PDB ID 5F09),
the heterometallic cluster Zn1-Ca2 is rotated relative
to the homometallic Zn1-Zn2 cluster so that Ca2 can
be bound in a typical pentagonal bipyramidal manner,
whereby His367 and a partially ionized hydroxyl of
the citrate moiety are the axial ligands. An active site
water forms part of the pentagonal coordination base,
and residues Asp443 and Asp377 are correspondingly
shifted to accommodate the Ca2 atom. Finally, we
compared the energetics of calcium binding in the
6
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
Fig. 5. QM/MM model of GCPIII in
complex with BCG and its comparison to
the X-ray structure of GCPII Glu424Ala in
complex with b-citryl-L-glutamic acid (BCG,
PDB ID 5F09). Images are rendered in
cross-eye stereo representation. Numbering
of amino acid residues in GCPIII is lower by
10 in comparison to GCPII. Tyr700/Tyr690 is
truncated to provide an unobscured view.
ꢂ
Distances are shown in A. Zinc atoms are
shown as spheres. Images were created
using PYMOL [24]. (A) The QM/MM model of
GCPIII in complex with BCG shows that
GCPIII binds BCG in a way that allows more
interactions. Protein is coloured green, BCG
pale cyan, zinc atoms violet. (B) Overlay of
GCPII-BCG complex with the QM/MM
model of GCPIII-BCG complex. It shows
that BCG is bound in a very similar manner
to both GCPII and GCPIII. The only two
differences are the presence of Ser509 in
GCPIII instead of Asn519 in GCPII and how
the citrate moiety is bound. GCPII-BCG
complex is coloured yellow, and GCPIII-BCG
in green.
GCPII and GCPIII sites in the presence of substrate
(BCG vs. NAAG). Employing the QM (DFT+D3/
TPSSh/def2-TZVP)/MM values, the calculations pre-
dict rather contradictory effects. For NAAG as a sub-
strate, calculated affinity of calcium for the active site
was marginally greater for GCPIII (~ 0.5 kcalꢁmolꢀ1).
We found the opposite pattern for BCG; the calcium
substitution was ~ 10 kcalꢁmolꢀ1 ‘easier’ for the
GCPII-BCG complex than the GCPIII-BCG complex.
Therefore, the calculations suggest that the observed
differences in the catalytic efficiencies may not depend
on the stability constants of calcium ions that poten-
tially fill the partially occupied Zn2 sites. Further
investigation into the observed metal ion dependence
is beyond the scope of the current study.
detection of GCPIII in a representative sample of
human tissues, analogously to an earlier report [13]
(Fig. 7A). Both assays were performed in the presence
of calcium ions because it affords the highest differ-
ence in GCPII and GCPIII catalytic efficiencies for the
substrates employed (for quantitative information, see
Materials and methods). Simultaneously, we quantified
the amount of mRNA coding for GCPII and GCPIII
in a set of human tissues to confirm the functionality
of the proposed enzymatic detection assays.
The amounts of GCPII and GCPIII transcripts were
analysed in commercially available panels of human
tissue cDNA libraries (Human MTC Panel I and II)
by quantitative PCR (qPCR). Each transcript was
amplified by a specific assay set consisting of a primer
pair for amplification and a fluorescent probe for visu-
alization. The cDNA amplification products were
resolved by agarose gel electrophoresis. Besides the
occasional weak formation of primer dimers in the
GCPII assay set and the occasional weak formation of
higher molecular weight products in the GCPIII assay,
only the product of expected size was observed in all
reactions. No products were formed in the negative
controls without template DNA, indicating that there
GCPII and GCPIII expression levels can be
specifically determined by enzymatic assays and
correspond to mRNA levels
Based on the results obtained from the enzymological
characterization of human GCPII and GCPIII, we
employed the NAAG cleavage assay for specific detec-
tion of GCPII and the BCG cleavage assay for specific
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
7

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
were no false positives or negatives due to amplifica-
tion of other nontargeted sequences (data not shown).
We also tested the selectivity of each assay set by
amplification of plasmids with subcloned sequences of
full-length GCPII or GCPIII and of isolated genomic
DNA. Both assay sets amplified only the target of
interest, and not the homologous transcript, and they
did not tend to detect the genomic sequence. We tested
selectivity against the genomic sequence because
cDNA libraries are often contaminated with the gen-
ome. To this point, we also quantified the amount of
the genome in the cDNA libraries by another assay
set, which showed us that the contamination was very
low to undetectable.
Amplification of a dilution series of plasmids with
subcloned sequences of either GCPII or GCPIII
showed linear dependency of determined Cq values on
the logarithm of template concentration over the com-
plete range of concentrations used. It also revealed a
PCR efficiency of over 80% for both assay sets.
The absolute amounts of both transcripts detected
in tissue cDNA libraries are shown in Fig. 7B. Because
the cDNA libraries were normalized to several differ-
ent housekeeping genes (G3PDH, alpha-tubulin, phos-
pholipase A2, beta-actin) and are mostly pooled from
multiple individuals, the results between tissues should
be comparable and should represent tissue distribution
of both transcripts.
As shown in Fig. 7, the protein levels of GCPII and
GCPIII (Fig. 7A) determined by enzymatic assays cor-
relate with the mRNA levels determined by qPCR
(Fig. 7B). These data indicate that the enzymatic
assays can be used for specific discrimination of
GCPII and GCPIII expression levels in various human
tissues. We detected a greater than one order of mag-
nitude higher amount of GCPIII compared to GCPII
in the testes. On the other hand, GCPIII expression in
the brain, kidney and small intestine is one to two
orders of magnitude lower than that of GCPII.
Fig. 6. QM/MM model of GCPIII in complex with BCG and with Zn2
atom replaced by calcium (II) (GCPIII-BCG-Ca2) compared to the X-
ray structure of GCPII in complex with BCG (GCPII-BCG, PDB ID
5F09). Images are rendered in cross-eye stereo representation. BCG
is depicted in stick representation. Please note that there is also a
hydrogen atom of BCG hydroxyl group shown in light grey, which is
attracted by the ionized carboxyl group of the citrate moiety, so that
the hydroxyl becomes partially ionized to coordinate the Ca2 atom.
Metal atoms are shown as spheres. (A) A comparison of GCPIII-
BCG-Ca2 QM/MM model (blue) with GCPII-BCG (yellow). (B) The
same as in (A), but rotated along y axis by 180°. (C) Detailed view of
the coordination matrix of Ca2 in the GCPIII-BCG-Ca2 QM/MM
model. BCG is coloured pale cyan, protein blue, and Zn1-Ca2 cluster
Discussion
In this study, we demonstrated that human GCPIII
shares a similar metal-dependent enzymological profile
as previously described for its mouse orthologue [13].
We observed facilitation of BCG-hydrolysing activity
by Ca2+ ions and NAAG-hydrolysing activity by
Zn2+ ions. Additionally, we discovered that GCPIII
also processes polyglutamylated folates (FolGlun) and
that this activity is facilitated by Zn2+ ions in a similar
manner as NAAG hydrolysis.
ꢂ
grey. Coordination distances are shown in A. (D) The same as in (C),
The concentration of calcium cations used for the
enzymological studies, 2.5 mM, was chosen to
but rotated along y axis by 180°.
8
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
Fig. 7. GCPII/III protein and mRNA levels in
human tissues. Error bars represent SD
[n = 2 for (A), n = 3 for (B)]. For an easy
comparison of protein and mRNA levels,
tissues in (A) and (B) are juxtaposed. (A)
Concentration of GCPII/III proteins in
selected human tissues. The reaction buffer
was 20 mM Tris, 0.15 M NaCl, 2.5 mM
CaCl2, 0.1% (v/v) Tween 20, pH 7.4. (B)
Quantitative PCR (qPCR) determination of
mRNA coding for GCPII and GCPIII. The
‘No. of transcripts’ shows the amount of
transcripts determined in 1.0 lL of 10-fold
diluted tissue cDNA library (for more
experimental details see Materials and
methods).
approximate the common concentrations of these
cations in human plasma (despite its free, so-called
ionic, concentration is usually two times lower [30]).
Although free zinc concentration is very low in blood
plasma [31], 0.10 mM concentration was used in our
study, mainly because we wanted to characterize the
saturation kinetics and because GCPIII is half-satu-
We attempted to address the structural basis for the
differences in substrate specificities between these two
homologues. Inspection of the GCPII and GCPIII 3D
structures revealed that, although their active sites are
structurally almost identical, there are two notable dif-
ferences: substitution of Asn519 in GCPII with Ser509
in GCPIII and substitution of Trp541 in GCPII with
Lys531 in GCPIII. The Asn to Ser substitution enables
two conformations of the Zn2-coordinating amino acid
Asp443 and is thus likely responsible for the lower occu-
pancy of the Zn2 ion within the GCPIII active site [21],
while the Trp to Lys substitution likely disrupts the
ABS [7,20]. Disruption of the ABS in GCPIII can be
exploited for development of inhibitors that are highly
specific towards GCPII [34]. We corroborated these
results by showing that endogenous substrates such as
FolGlun, which utilize the ABS in GCPII [7], have
substantially lower affinity (higher KM value) towards
GCPIII compared to GCPII. These findings suggest
that even though GCPIII is also expressed at low levels
in small intestine, the processing of polyglutamylated
folates will likely be performed primarily by GCPII.
GCPII inhibitors are currently heavily investigated as
potential theranostic agents against prostate cancer [10]
and various neurological disorders. Therefore, it is
advisable to design compounds that would not hinder
rated by zinc at
a relatively low concentration
(approximately 6 lM, unpublished observation). Fur-
thermore, zinc in the CNS serves as an orthograde,
transcellular messenger [32], and it is also secreted by
pancreas at the amount 1–2 mgꢁdayꢀ1 [33] (see also
below). Because both metals substantially facilitated
the corresponding enzymatic activities of GCPIII,
GCPIII has the potential to act as either a BCG-
or NAAG/FolGlun-hydrolysing enzyme. Its activity
could likely be fine-tuned by the Ca2+ and Zn2+ con-
centrations in a particular tissue. Therefore, we sug-
gest that GCPIII may have different enzymatic
functions within the human body based on its tissue
localization, the accessibility of its substrates and the
concentration of divalent cations in that particular
tissue. For a proper estimation of the function of
GCPIII, its expression profile will need to be
correlated with the distribution of its endogenous sub-
strates, in particular BCG.
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
9

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
the BCG-hydrolysing activity of its homologue GCPIII,
although its physiological relevance is not quite clear
yet.
Finally, we utilized enzymological data for specific
detection of GCPII and GCPIII protein levels in sev-
eral human tissues. Because there are no commercially
available specific antibodies against human GCPIII,
this approach represents a method for detection and
discrimination between these two enzymes in human
tissues. The fact that we detected lower levels of
GCPIII than GCPII in brain and small intestine, the
tissues where cleavage of NAAG and polyglutamy-
lated folates, respectively, takes place, supports the
assumption that BCG is the main physiological sub-
strate of GCPIII in humans. The high level of GCPIII
expression in human testes, and the high BCG-
hydrolysing activity also detected in rat testes [16],
further strengthens this hypothesis. We believe that a
reliable determination of the GCPIII expression pro-
file on a protein level, combined with data on BCG
levels in human tissues, will help to dissect and eluci-
date the primary physiological function of GCPIII in
the human body.
Furthermore, we attempted to elucidate the mecha-
nism of calcium-dependent BCG cleavage by determin-
ing structures of GCPII and GCPIII in complex with
BCG. Unfortunately, due to experimental difficulties,
we were able to prepare only protein crystals of GCPII
in complex with BGC. Therefore, we built a QM/MM
model of the GCPIII-BCG complex. As a control for
our QM/MM approach, we also built a QM/MM
structural model of GCPII in complex with NAAG
and BCG (which are available as X-ray structures,
PDB IDs 3BXM [22] and 5F09, respectively). The
QM/MM models of GCPII in complex with NAAG
or BCG and X-ray structures of inactive GCPII
(Glu424Ala) in complex with these compounds are
almost identical, with root-mean-square deviation val-
ꢂ
ues of 0.069 and 0.189 A, respectively (except for the
citrate moiety, which adopts a different conformation
– the same conformation as in the GCPIII-BCG
model, data not shown).
Regarding mRNA of GCPII in human tissues, there
is one discrepancy to be addressed. Cunha et al. [35]
also determined the amount of GCPII (PSMA in the
medical literature) in Clontech tissue cDNA libraries.
Their primers, like ours, were designed to detect only
the PSMA splice variant and not the intracellular var-
iant of PSMA. The relative PSMA expression levels in
various tissues that they determined [35] are in good
agreement with our results, except for PSMA expres-
sion in prostate. We determined the ratio of expression
of PSMA in the prostate and liver (the tissue with sec-
ond highest PSMA expression) to be approximately 2,
compared to 12 in the previous report. Although the
tissue libraries were normalized to four different refer-
ence genes by the vendor, Cunha et al. renormalized
their results to only one of these genes, beta-actin.
Unfortunately, they do not describe the details of the
beta-actin assay or the amounts of beta-actin deter-
mined. Therefore, it is not clear whether the higher
specificity of PSMA expression in their study was
caused by lower levels of beta-actin in prostate, or
whether the different results can be attributed to dif-
ferent tissue library lots.
Our structural data show an overall similar binding
of BCG into the GCPII and GCPIII active sites,
which was expected because the major difference in
the catalytic efficiencies of these two enzymes comes
from substrate turnover number (kcat) rather than
substrate binding (KM) (see Fig. 2B,C). Considering
our observation that GCPII is not activated by diva-
lent cations, we investigated the hypothesis that the
mechanism behind the activation of GCPIII-BCG-
hydrolysing activity might presumably involve
replacement of the loosely coordinated Zn2 ion from
GCPIII’s active site with a different ion, such as
Ca2+. However, the QM/MM calculations did not
provide a clear indication that the replacement of the
Zn2 ion with Ca2+ would be thermodynamically
more favourable for BCG as a substrate rather than
NAAG and rather exclude this hypothesis. However,
it should be kept in mind that QM/MM calculations
do not provide any information about whether Ca2+
is present or absent in the partially occupied GCPIII
active site. In addition, our computed values do not
provide any indication whether BCG – presumably a
better chelator than NAAG – assists in the Ca2+
uptake into the active site. This issue might be, in
principle, assessed experimentally (X-ray crystallogra-
phy or X-ray absorption techniques) and comple-
mented by QM/MM/MD calculations of the
activation energies of individual complexes (to anal-
yse the kinetic origin of the observed metal ion
dependency) which is well beyond the scope of the
presented work.
In summary, we present a thorough enzymatic and
structural characterization of the behaviour of human
GCPII and GCPIII towards their physiological sub-
strates. We also identified metal-dependent enzymatic
activity of human GCPIII, which we did not observe
for human GCPII. Finally, we confirmed BCG to be
a specific substrate of human GCPIII. Utilizing these
findings, we mapped the expression patterns of GCPII
and GCPIII in a panel of human tissues. Compared
to GCPII expression, we detected higher levels of
10
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
for the rest of the protein. In the current version, the
program employs TURBOMOLE 6.6 suite [47] for the QM
part and AMBER 8.0 with the ff14SB force field for the
MM part [48]. COMQUM utilizes the standard hydrogen-
link approach, and the technical details of the QM/MM
computational protocol have been described elsewhere
[22].
GCPIII in testes and lower levels in brain, kidney and
prostate. Additionally, this study is also the first to
show mRNA levels of GCPIII in the human tissues.
Materials and methods
Crystallization and data collection
All quantum chemical calculations were performed at the
density functional theory (DFT) level. Geometry optimiza-
tions were carried out at the Perdew–Burke–Ernzerhof
(PBE) level [49]. The DFT/PBE calculations were expedited
by expanding the Coulomb integrals in an auxiliary basis
set: the resolution-of-identity (RI-J) approximation. The
def-SV(P) basis set was employed for all atoms. The single-
point energies were then calculated using the TPSSH
method [50] as implemented in TURBOMOLE 6.6. For these
calculations, the def2-TZVPD basis set was employed for
all atoms.
Inactive GCPII with the engineered mutation Glu424Ala
[22] and AviTagTM affinity tag (Avidity, Aurora, CO, USA)
[19] was crystallized using the hanging drop vapour diffusion
method as described earlier [36] with the following modifica-
tions: (a) 1.0 lL precipitant was added to a 1.0 lL drop of
protein (concentrated to 3.6 mgꢁmLꢀ1) in 20 mM bistrispro-
pane, 20 mM NaCl, pH 7.4, and sealed over a 0.50 mL reser-
voir of the same precipitant; (b) 0.1 lL of 2.1 mM BCG in
MilliQ water (Merck Millipore, Billerica, MA, USA) was
added to the crystallization drop after 11 weeks and again
after 13 weeks (resulting in a calculated ratio of pro-
tein : substrate equal to 19 lM : 0.21 mM); (c) the crystal
was frozen in liquid nitrogen after 15 weeks; and (d) precipi-
tant was made up of 33% (v/v) pentaerythritol propoxylate
PO/OH 5/4 (Hampton Research, Aliso Viejo, CA, USA),
1.5% (w/v) PEG 3350 (Merck KGaA, Darmstadt, Germany)
and 0.10 M Tris/HCl (Merck KGaA), pH 8.0, in distilled
water. In the drop, one single crystal appeared after several
weeks. Crystallization approach of adding BCG right at the
setup was unsuccessful. Diffraction data were collected at
100 K at the beamline BL1412 operated by the Joint Berlin
MX-Laboratory at the BESSY II electron storage ring (Ber-
lin-Adlershof, Germany) [37], using a Pilatus 6M detector
(Dectris, Baden, Switzerland). Scaling and merging was done
with the programs XDS [38] and its graphical user interface
XDSAPP [39].
The protein (GCPII) setup
Structural models used in QM/MM calculations were
ꢂ
based on the 1.71 A X-ray crystallographic structure of the
inactive Glu424Ala mutant of GCPII in complex with
ꢂ
NAAG (PDB ID 3BXM [22]) and the 1.37 A structure of
the GCPIII with L-glutamate (3FF3 [21]). BCG was posi-
tioned in the active site using the GCPII (Glu424Ala)/
NAAG and GCPIII/L-Glu structures as templates. The
protein was equilibrated using the standard approach [22]:
(a) minimizing the positions of all hydrogen atoms in the
initial structure, (b) adding a solvation sphere with a
ꢂ
radius of 50 A (~ 12 700 and 11 500 water molecules in
total for GCPII and III, respectively) and (c) running a
1 ns simulated annealing molecular dynamics followed by
the final minimization of the whole system (with all nonhy-
drogens atoms kept at their crystallographic positions
throughout). Both GCPII and GCPIII structures were neu-
tralized by addition of ions (two Na+ for GCPII and four
Clꢀ for GCPIII). We assumed the standard protonation
states at pH 7 for all amino acids. For the histidine resi-
dues, the protonation status was assigned based on a
detailed study of the hydrogen bond network around the
residue and the solvent accessibility. In the GCPII struc-
ture, histidines 82, 347, 377, 553 and 573 were assumed to
be protonated on the Nd atom; histidines 56, 112, 124,
295, 396, 475, 689 and 697 on the Ne atom; and histidines
345 and 618 on both nitrogens. In the GCPIII structure,
histidines 124, 367, 543, 637 and 721 were assumed to be
protonated on the Nd atom; histidines 102, 285, 335, 608,
615, 679 and 687 on the Ne atom; and histidine 72 on both
nitrogens.
Structure determination was performed by molecular
replacement using the program MOLREP from the CCP4 soft-
ware package [27,40,41]. The previously solved structure of
recombinant human GCPII (PDB ID 4NGP [11]) without
inhibitor and water molecules was used as the starting
model. Refinement calculations were performed with the
program REFMAC 5.7 [42,43], and the refinement protocol
was interspersed with manual corrections to the model
using WINCOOT 0.7 [44]. The final models, together with
experimental amplitudes, were deposited into the RCSB
Protein Data Bank under the accession number 5F09. A
summary of structural parameters is displayed in Table 1.
Images were created with the programs PYMOL [24] and FFT
and MAPMASK [45].
QM/MM calculations
The quantum system consisted of ~ 280 active site atoms
and is deposited in Data S1. The so-called system 2 (part
of the system that is relaxed in the QM/MM calculations
by MM method) comprised 48 amino acids in the vicinity
All QM/MM calculations were carried out with the COM-
QUM program [46], which combines the quantum mechan-
ical approach, applied to the region most relevant for the
studied process, and the classical mechanical approach
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
11

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
of the quantum system, and it has been selected on a per-
ꢂ
residue based cut-off distance of 2.5 A from the quantum
Scientific, Waltham, MA, USA) and by NMR at Bruker
Avance ITM 400 MHz (Bruker, Billerica, MA, USA).
system. The rest of the protein (system 3) was kept frozen
in all QM/MM simulations.
Synthesis of b-citrylglutamate
Overall scheme for the six-step synthesis of BCG is shown
below (Scheme 1). Detailed description of each step and its
resulting intermediate or final compound (designated 1 to 5
or 6, respectively) is given in the following text in the order
of the synthesis (Scheme 1).
Synthesis of b-citryl-L-glutamic acid (BCG)
General information
All chemicals were purchased from Sigma-Aldrich, unless
stated otherwise. b-citrylglutamate (BCG) was purified
using preparative scale HPLC Waters Delta 600 (Waters
Corporation, Milford, MA, USA) (flow rate 7 mLꢁminꢀ1),
with column Waters SunFire C18 OBD Prep Column,
5 lm, 19 9 150 mm (Waters Corporation). The purity of
BCG was tested on the analytical Jasco PU-1580 HPLC
instrument (Jasco, Easton, MD, USA) (flow rate
1 mLꢁminꢀ1, Rt is shown below) employing the Waters C18
Analytical Column, 5 lm, 250 9 5 mm (Waters Corpora-
tion). The final product was of at least 99% purity. Struc-
ture was further confirmed by high-resolution mass
spectrometry (HRMS) at LTQ Orbitrap XL (Thermo
1. 2-tert-butyl 1,3-dimethyl 2-hydroxypropane-1,2,3-
tricarboxylate
About 5.3 g of dimethyl citrate (24 mmol, prepared as
described in [51]) was suspended in 15 mL of isobutylene
and 0.5 mL of sulfuric acid was added while cooled to
ꢀ78 °C. The reaction mixture cleared over time. After 18 h
of stirring at room temperature, the reaction mixture was
cooled down to ꢀ78 °C and poured to ice bath saturated
with bicarbonate. The water phase was then washed three
times by ethylacetate (EtOAc) (50 mL) and the organic
O
O
HO
O
O
O
O
O
O
O
O
O
a
c
b
HO
HO
HO
HO
OH
O
O
O
O
O
O
O
O
O
HO
O
1
2
3
d
O
HO
O
O
O
O
O
O
O
OH
OH
O
O
O
O
O
O
e
f
HO
HO
HO
OH
O
HN
O
HN
O
O
HO
O
5
4
6
(BCG)
Scheme 1. The scheme for synthesis of b-citryl-L-glutamic acid (BCG). The synthesis proceeded via five intermediate compounds,
designated 1–5, whereby BCG is compound 6. (a) Isobutylene, H2SO4; (b) NaOH, MeOH/H2O; (c) DBU, benzyl bromide, ACN; (d) TFA; (e)
NH2Glu(OBzl)OBzl, O-(Benzotriazol-1-yl)-N,N,N0,N0-tetramethyluronium tetrafluoroborate, DIEA, dimethylformamide; (f) H2, Pd(C), MeOH.
12
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
phases were combined, dried and evaporated yielding 5.3 g
The organic phase was washed two times with 10%
KHSO4, two times with saturated NaHCO3 and once with
brine. About 600 mg of compound 5 (yield = 60%) was
obtained as an oily product after evaporation and column
chromatography on silica (EtOAc, TLC Rf = 0.25).
of pure product (yield = 80%).
2. 3-(tert-butoxycarbonyl)-3-hydroxypentanedioic acid
About 5.2 g of compound 1 (18.7 mmol) was dissolved in
20 mL of methanol and was cooled down to 0 °C. About
20 mL of 2 M NaOH was then added in portions during
5 min. The reaction was left to proceed for 6 h. The reac-
tion mixture was then diluted by 200 mL of EtOAc and
was extracted two times by 10% KHSO4 (2 9 100 mL).
The organic phase was then dried and evaporated to yield
3.5 g of product (yield = 75%).
6. 3-{[(1S)-1,3-dicarboxypropyl]carbamoyl}-3-hydroxy-
pentanedioic acid (ß-citrylglutamate, BCG)
About 50 mg of compound 5 was dissolved in 2 mL of
methanol (MeOH) and catalytical amount of palladium on
activated charcoal [Pd(C)] was added. The reaction mixture
was first purged with flow of nitrogen, then with flow of
hydrogen. The benzyl groups were then removed during
3 h using slightly elevated pressure of hydrogen (1.05 Atm).
TLC analysis showed that all the reactants disappeared and
the reaction mixture was filtered and evaporated. The pro-
duct was purified using preparative scale HPLC (isocratic
2.5% ACN, Rt = 6 min). About 10 mg was isolated upon
dry freezing (isolated yield = 42%). 1H NMR (401 MHz,
DMSO) d 12.31 (bs, 4H), 7.74 (d, J = 8.0 Hz, 1H), 5.73
(bs, 1H), 4.25 (td, J = 8.3, 4.8 Hz, 1H), 2.77 (d,
J = 15.6 Hz, 1H), 2.68 (d, J = 15.7 Hz, 1H), 2.66 (d,
J = 15.2 Hz, 1H), 2.58 (d, J = 15.2 Hz, 1H), 2.37–2.21 (m,
2H), 2.07–1.98 (m, 1H), 1.89–1.78 (m, 1H). 13C NMR
3. 1,3-dibenzyl 2-tert-butyl 2-hydroxypropane-1,2,3-
tricarboxylate
About 1 g of compound 2 (4.03 mmol, 1.0 eq) was sus-
pended in 20 mL of acetonitrile (ACN) along with 1.2 mL
of 1,8-Diazabicycloundec-7-ene (DBU) (8.86 mmol, 2.0 eq).
About 1.05 mL of benzyl bromide was added in one por-
tion and the reaction mixture was refluxed for 2 h. The
TLC analysis (EtOAc) showed no starting material and the
whole mixture was evaporated. The crude product was dis-
solved in 100 mL of EtOAc and was washed two times
with 10% KHSO4, two times with saturated NaHCO3 and
once with brine, and the organic phase was dried and evap-
orated to obtain the titled compound as white solid (1.2 g,
yield = 70%). The product was used in next step without
further purification (purity approximately 90%).
(101 MHz, DMSO)
d 174.01, 173.40, 172.94, 171.41,
171.26, 73.47, 51.12, 42.90, 42.45, 29.66, 26.78. Analytical
HPLC Rt = 4.0 min. HRMS (ESI-) m/z for C19H22O7N2
[MꢀH]ꢀ calculated 389.13433, found 389.13424.
4. 1,3-dibenzyl 2-hydroxypropane-1,2,3-tricarboxylate
Expression and purification of recombinant
human GCPII, GCPIII and GCPII inactive variant
About 1.1 g of compound 3 (2.57 mmol) was dissolved in
3 mL of trifluoroacetic acid (TFA). The reaction mixture
was stirred and sonicated alternately for 25 min and then
the TFA was removed by flow of nitrogen. The residue was
then dissolved in EtOAc and was washed three times with
water to get rid of the residual TFA. The organic phase
was then dried and evaporated. About 940 mg of com-
pound 4 was obtained (yield = 95%) and it was used in the
synthesis without further purification.
The expression and purification protocols for all used
human recombinant proteins in this study, Avi-tagged
extracellular portion of GCPII [19], GCPIII [12], the inac-
tive variant of GCPII [7] and tag-free extracellular portion
of GCPII [52], have been described earlier.
GCPII and GCPIII activity assay
Because GCPIII activity is metal-dependent, MilliQ (Merck
Millipore) was used to prepare all solutions. Reactions with
BCG, NAAG (Merck KGaA) and folyl-n-c-L-glutamic
acids (FolGlu1–6) (Schircks Laboratories, Jona, Switzer-
land) were performed in 25 mM Tris/HCl, pH 7.5, (Merck
KGaA) or 25 mM bistrispropane HCl, pH 7.5, (Merck
KGaA) in a total volume of 215 lL in a 96-well plate
immersed in a water bath (Grant Instruments, Shepreth,
UK) at 37.0 °C [or in a lower throughput setting in 1.5 mL
eppendorf tubes using a thermo-block Bioer MB120 (MB-
102; Bioer, Hangzhou, China)]. Concentrations of substrate
stock solutions were determined by amino acid analysis
[53]. All components were pipetted on ice, and reactions
5. 2-{(1S)-[1,3-bis(benzylcarboxy)prop-1-yl]-
carbamoyl}-1,3-bis(benzylcarboxy)-2-carboxy-2-
hydroxy-propan-2-amide
About 500 mg of compound 4 (1.35 mmol, 1.1 eq) was dis-
solved in 5 mL of dimethylformamide along with 564 mg
of 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hex-
auorophosphate (1.49 mmol, 1.1 eq) and 840 lL of N,N-
Diisopropylethylamine (DIEA) (4.65 mmol, 3.5 eq). The
mixture was left stirring for 15 min and then 486 mg of
H2N-Glu(OBn)-COOBn*Tos (1.49 mmol, 1.0 eq) was
added in one portion. After 2.5 h, all volatiles were evapo-
rated and the residue was dissolved in 50 mL of EtOAc.
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
13

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
were started by adding substrate following a 5- to 13-min
preincubation at 37.0 °C. After 20 min, the reactions were
terminated by inhibiting the enzyme and by changing the
pH to 5.8 by addition of 10.0 lL stopping solution com-
posed of 0.37 M phosphoric acid (Penta, Prague, Czech
Republic) and 22 lM 2-(phosphonomethyl)-pentanedioic
acid in MilliQ water (Merck Millipore) (when 25 mM bis-
trispropane buffer, pH 7.5, was used) or 3.48 lL of 0.41 M
phosphoric acid (Penta), 72 lM 2-(phosphonomethyl)-pen-
tanedioic acid and 7.2 lM 2-mercaptoethanol (Merck
KGaA) in MilliQ water (Merck Millipore) (when 25 mM
Tris/HCl buffer, pH 7.5, was used). Reaction conditions
were designed to yield 10–30% substrate turnover. The
method for quantifying FolGlu1–6 substrates and FolGlu0–5
products was the same as described elsewhere [7], while
hydrolysis of NAAG and BCG was assessed with a novel
assay (see below for details). Using GRAFIT, we fitted a sat-
uration curve to 6–11 data points for every substrate
(whereby each point represented a single or average of a
duplicate experiment). Because of low KM values, the assay
was sometimes restricted by the limit of quantification of
the employed analytic method (indicated by an asterisk in
Fig. 2 and below), meaning that the KM value was smaller
than the lowest experimental substrate concentration. In
Fig. 2, the lowest substrate concentrations were, from left
to right, 1.5, 0.20, 0.20, 0.20, 0.32*, 0.088*, 0.20, 0.20, 0.20,
0.10, 0.14*, 0.045, 0.031, 0.030, 0.034 lM.
Atomic absorption spectroscopy to determine Zn
occupancy in the GCPIII active site
Following purification, approximately 10 lM GCPII and
GCPIII was dialysed against 10 mM MOPS (Duchefa,
Biochemie, Haarlem, the Netherlands; pH adjusted by
KOH), pH 7.4, using Slide-A-LyzerTM MINI Dialysis
Devices (Thermo Scientific). The buffer was prepared from
MilliQ water (Merck Millipore) and subjected to Chelex
100 (Bio-Rad, Hercules, CA, USA) before use. Dialysis
devices and glass vials used for protein preparation were
washed twice with EDTA solution and thoroughly with
MilliQ (Merck Millipore) water. Samples were analysed on
an Analyst 800 spectrometer (Perkin-Elmer, Waltham, MA,
USA) with electrothermal atomization using an EDL Zn
lamp. Temperatures of pyrolysis and atomization were
700 °C and 1500 °C, respectively. Palladium was employed
as
a modifier. The limit of detection for zinc was
0.1 ngꢁmLꢀ1. For the analysis itself, samples were diluted
200-fold. The analysed volume was 10 mL. Measurements
were performed at least in triplicate. The protein concentra-
tion was quantified by amino acid analysis [53].
Quantification of GCPII/III at the protein and
mRNA levels
Enzymatic assay
The total protein concentrations of tissue lysates were esti-
mated by a modified Bradford method (Protein Assay;
Bio-Rad), using BSA (Thermo Scientific) as a standard
and 97 mM HEPES (Merck KGaA) and 0.15 M NaCl
(Penta) pH 7.5 (NaOH; Penta) as a buffer. For the activity
assay, tissue lysates were diluted 5.3- to 7300-fold overall.
Buffer used for the specific detection of both GCPII and
GCPIII was prepared from MilliQ water (Merck Milli-
pore), composed of 20 mM Tris (Merck KGaA), 0.15 M
NaCl (Penta), 2.5 mM CaCl2 (Merck KGaA) and 0.1% (v/
v) Tween 20 (USB, Cleveland, OH, USA), pH 7.4 (HCl,
Penta) and filtered through a 0.22 lm filter (Merck Milli-
pore). As a substrate, [3H]NAAG and [3H]BCG were used
at 5 nM final concentration (mixed with cold substrates in
a 19 : 1 molar ratio to yield a total substrate concentration
of 100 nM) in a previously described radiometric assay [5],
in which the column volume was 1.0 or 0.5 mL and elu-
tion was performed with 2.4 or 1.2 mL of 1 M HCOOH
(Penta), respectively. The usual time of incubation was
17 h at 37.0 °C.
Quantification of glutamate in the enzymatic
assay
To quantify glutamate as a product of NAAG and BCG
cleavage, a novel method was developed. It is based on a
highly concentrated solution of orthophthalaldehyde
(OPA), a commonly used reagent for amino acid derivati-
zation [53]. The new derivatization formulation was a
0.10 M OPA, 0.30 M 3-mercaptopropionic acid, 4% acetoni-
trile (v/v) and 0.94 M bistrispropane, pH 9.6 solution. A
11.0 lL portion of the solution was added to 99 lL of a
reaction containing the cleavage product glutamate. This
improvement made it possible to surpass the need to con-
centrate the sample by lyophilization and redissolving it [7].
Sample derivatization and analysis was performed on an
Agilent 1200 or 1260 instrument (Agilent Technologies,
Santa Clara, CA, USA) similarly as described previously
[7], except that the analysis method was slightly modified.
The new method consisted of 13.80 min at 0.7% B,
0.05 min transition to 80.0% B, 1.00 min at 80.0%
B, 0.05 min transition back to 0.70% B, 0.60 min at 0.7%
B, injection of 20.0 lL of 16-fold diluted 85% phosphoric
acid (w/w; Penta), pH 2.5, (NaOH; Penta) to dissolve pre-
cipitate of metal hydroxides, and 10.50 min at 0.7% B. The
automatic derivatization procedure and injection performed
by the instrument just prior to analysis took approximately
7 min.
This particular buffer confers the needed differences in
specificities of GCPII towards NAAG and GCPIII towards
BCG: When GCPII is quantified according to NAAG
hydrolysis, GCPIII must be approximately 60-fold more
concentrated to yield a comparable turnover. On the other
hand, when GCPIII is quantified according to BCG hydrol-
ysis, GCPII does not cleave BCG at all even if its concen-
tration is 180-fold higher (data not shown).
14
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
slope and intercept values of the curves for the common
sequence were equal and corresponded to the average
value between the two plasmids.
Quantitative PCR
Amounts of GCPII transcript (encoded by the gene
FOLH1) were quantified by an assay set of forward and
reverse primers (sequences 50-CGG CTT CCT CTT CGG
GTG-30 and 50-GAT GTT CTC AGC TTT CAA TTC
ATC C-30) and a fluorescent hydrolysis probe (sequence 50-
ATC CTC CAA TGA AGC TAC TAA CAT TAC TCC
AA-30). This set was designed to amplify nucleotides 366–
471 in the GCPII transcript (NM_004476) to yield an
amplified product of 106 bps, which covers regions of
exons 1 and 2 and corresponds to amino acids 35–70 in the
longest open reading frame (ORF). The forward primer is
complementary to the exon 1–exon 2 junction, and it was
designed to enable selective amplification of the wild-type
GCPII protein-coding splice variant (usually termed
PSMA). This assay set was also designed to amplify neither
genomic sequences nor the prostate specific membrane anti-
gen like protein transcript, which is encoded by the pseudo-
gene FOLH1B.
To assess the selectivity of qPCR amplification, both
pcDNA4 plasmids with subcloned protein-coding sequences
of either GCPII or GCPIII (106 copies per reaction) were
amplified in the same qPCR setup by both assay sets, and
possible products were inspected by agarose gel elec-
trophoresis. For the same reason, amplification of isolated
human genomic DNA was tested in the same way. The
amount of genomic DNA was quantified by a set of pri-
mers of sequences 50-GAG AAC CGT TTG AAT GAA
ACT GAG-30 and 50-TTG GAT GAA CAG GAA TAC
TTG GAA GA-30 and a fluorescent hydrolysis probe of
sequence 50-ACA GCC TCT GCA ATT CCA CGC CTA
T-30, which detects the intron–exon junction of the FOLH1
gene (gene coding for GCPII).
The amount of both GCPII and GCPIII transcripts was
then measured in 1.0 lL of 10-fold diluted tissue cDNA
libraries obtained from Clontech (Human MTC Panel I
and II; Takara Bio, Kyoto, Japan).
Amounts of GCPIII transcript were quantified by an
assay set of forward and reverse primers (sequences 50-TTT
GGA CTT CTG GGT TCC AC-30 and 50-TGC TTC TCT
CCT GGA GTA TTT TG-30) and a fluorescent hydrolysis
probe, Universal probe #30 (Roche, 04687639001, Basel,
Switzerland). This set was designed to amplify nucleotides
1291–1354 in the GCPIII transcript (NM_005467) which
covers regions in exons 11 and 12 and corresponds to
amino acids 416–437 in the longest ORF. The expected size
of the amplified product was 64 bases; the size of a possibly
amplified genomic DNA sequence was 5308 bases.
All qPCR reactions were carried out in triplicate in
FrameStar 480/96 multiwell plates (4titude, Dorking, UK)
sealed with adhesive foil (Roche) using a LightCycler 480 II
instrument (Roche) in a total volume of 10 lL. Each reac-
tion consisted of a LightCycler 480 Probe Master (Roche)
diluted according to the manufacturer’s instructions, forward
and reverse primer (1 lM final concentration each), fluores-
cent probe (final concentration of Roche universal probes
was 50 nM; final concentration of custom probes was
100 nM) and 1 lL of sample or template DNA (positive and
nontemplate controls as well as interplate calibrators were
included on each plate). Initial denaturation for 3 min at
95 °C was followed by 45 cycles of 10 s at 95 °C, 30 s at
66 °C and 30 s at 72 °C. The threshold cycle numbers (Cq)
were then determined from fluorescence intensities acquired
during the qPCR runs by second derivative maximum
method using LIGHTCYCLER 480 software (Roche). The pres-
ence and the size of PCR products was analysed in all qPCR
reactions by agarose gel electrophoresis.
As a standard for absolute quantification, serial 10-fold
dilutions covering concentrations from 107 to 102 copies
per reaction of pcDNA4 plasmid with subcloned protein-
coding sequence of either GCPII (longest ORF from
NM_004476 coding amino acids 1–750, i.e. full-length
GCPII) or GCPIII (longest ORF from NM_005467 cod-
ing amino acids 1–740, i.e. full-length GCPIII) were
amplified with both assay sets. The initial concentration of
plasmid DNA (purified by QIAprep Spin Miniprep Kit;
Qiagen, Venlo, Netherlands) prior to dilution was deter-
mined spectrophotometrically by absorption at 260 nm
(Nanodrop ND-1000; Thermo Scientific). To enable pre-
cise comparison between the determined amounts of both
transcripts, obtained calibration curves were further nor-
malized against each other by quantification of common
regions in both plasmids. The ampicillin resistance gene
region was quantified with a set of primers of sequences
50-GCA GAA GTG GTC CTG CAA CT-30 and 50-AGC
TTC CCG GCA ACA ATT A-30 and Universal probe
#58 (Roche). In this way, two calibration curves were
obtained for each plasmid, one for the amplification of
target transcript and one for the common sequence.
Finally, the slope and intercept values of both curves were
transformed for each plasmid so that the transformed
Acknowledgements
We thank Prof Barbara Slusher (School of Medicine,
John Hopkins University, MD, USA) for her kind gift
ꢁ
of the 2-PMPA inhibitor; Radko Soucek for amino
acid analysis; Dr Sebastian Zoll for data collection,
ꢀ
scaling and molecular replacement; Jana Starkova for
cultivation of S2 cells; Dr Vlasta Korunova for AAS;
ꢀ
ꢁ
ꢀ
and Dr Pavel Sacha for recloning the inactive GCPII
Glu424Ala into the vector pMT/Bip/AviTEV/
rhGCPII. The financial support of the Czech Science
Foundation (GA CR, projects P208-12-G016 and
14-31419S) and InterBioMed I project LO 1302 and
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
15

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
9 Queisser A, Hagedorn SA, Braun M, Vogel W,
LO 1304 (to PP) from the Ministry of Education of
the Czech Republic is gratefully acknowledged.
Duensing S & Perner S (2015) Comparison of different
prostatic markers in lymph node and distant metastases
of prostate cancer. Mod Pathol 28, 138–145.
10 Foss CA, Mease RC, Cho SY, Kim HJ & Pomper MG
(2012) GCPII imaging and cancer. Curr Med Chem 19,
1346–1359.
11 Tykvart J, Schimer J, Barinkova J, Pachl P, Postova-
Slavetinska L, Majer P, Konvalinka J & Sacha P (2014)
Rational design of urea-based glutamate
Author contributions
MN performed protein purification, crystallization and
kinetic studies; JT and PP performed X-ray structure
refinement and validation; JS synthesized BCG; KH
prepared protein samples for AAS; LR and TAR
worked out the QM/MM models; VN quantified
GCPII/III in human tissues; JK initiated and led the
project; MN, JT, LR, KH, VN and JK analysed the
data and wrote the manuscript.
carboxypeptidase II (GCPII) inhibitors as versatile
tools for specific drug targeting and delivery. Bioorg
Med Chem 22, 4099–4108.
12 Tykvart J, Navratil V, Sedlak F, Corey E, Colombatti
M, Fracasso G, Koukolik F, Barinka C, Sacha P &
Konvalinka J (2014) Comparative analysis of
monoclonal antibodies against prostate-specific
membrane antigen (PSMA). Prostate 74, 1674–1690.
13 Collard F, Vertommen D, Constantinescu S, Buts L &
Van Schaftingen E (2011) Molecular identification of
beta-citrylglutamate hydrolase as glutamate
carboxypeptidase 3. J Biol Chem 286, 38220–38230.
14 Miyake M, Kakimoto Y & Sorimachi M (1978)
Isolation and identification of beta-citryl-L-glutamic
acid from newborn rat-brain. Biochim Biophys Acta
544, 656–666.
References
1 Hlouchova K, Navratil V, Tykvart J, Sacha P &
Konvalinka J (2012) GCPII variants, paralogs and
orthologs. Curr Med Chem 19, 1316–1322.
2 Pangalos MN, Neefs JM, Somers M, Verhasselt P,
Bekkers M, van der Helm L, Fraiponts E, Ashton D &
Gordon RD (1999) Isolation and expression of novel
human glutamate carboxypeptidases with N-acetylated
alpha-linked acidic dipeptidase and dipeptidyl peptidase
IV activity. J Biol Chem 274, 8470–8483.
3 Hlouchova K, Barinka C, Klusak V, Sacha P, Mlcochova
P, Majer P, Rulisek L & Konvalinka J (2007) Biochemical
characterization of human glutamate carboxypeptidase
III. J Neurochem 101, 682–696.
15 Miyake M, Kume S & Kakimoto Y (1982) Correlation
of the level of beta-citryl-L-glutamic acid with
spermatogenesis in rat testes. Biochim Biophys Acta 719,
495–500.
4 Bzdega T, Crowe SL, Ramadan ER, Sciarretta KH,
Olszewski RT, Ojeifo OA, Rafalski VA, Wroblewska B
& Neale JH (2004) The cloning and characterization of
a second brain enzyme with NAAG peptidase activity.
J Neurochem 89, 627–635.
5 Robinson MB, Blakely RD, Couto R & Coyle JT
(1987) Hydrolysis of the brain dipeptide N-acetyl-L-
aspartyl-L-glutamate. Identification and characterization
of a novel N-acetylated alpha-linked acidic dipeptidase
activity from rat brain. J Biol Chem 262, 14498–14506.
6 Chandler CJ, Wang TTY & Halsted CH (1986)
Pteroylpolyglutamate hydrolase from human jejunal
brush-borders – purification and characterization. J Biol
Chem 261, 928–933.
7 Navratil M, Ptacek J, Sacha P, Starkova J, Lubkowski
J, Barinka C & Konvalinka J (2014) Structural and
biochemical characterization of the folyl-poly-gamma-L-
glutamate hydrolyzing activity of human glutamate
carboxypeptidase II. FEBS J 281, 3228–3242.
8 Huber F, Montani M, Sulser T, Jaggi R, Wild P, Moch H,
Gevensleben H, Schmid M, Wyder S & Kristiansen G
(2015) Comprehensive validation of published
16 Miyake M, Innami T & Kakimoto Y (1983) A beta-
citryl-L-glutamate-hydrolysing enzyme in rat testes.
Biochim Biophys Acta 760, 206–214.
17 Hamada-Kanazawa M, Kouda M, Odani A,
Matsuyama K, Kanazawa K, Hasegawa T, Narahara
M & Miyake M (2010) Beta-citryl-L-glutamate is an
endogenous iron chelator that occurs naturally in the
developing brain. Biol Pharm Bull 33, 729–737.
ꢀ
18 Rovenska M, Hlouchova K, Sacha P, Mlcochova P,
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
Horak V, Zamecnık J, Barinka C & Konvalinka J (2008)
Tissue expression and enzymologic characterization of
human prostate specific membrane antigen and its rat and
pig orthologs. Prostate 68, 171–182.
19 Tykvart J, Sacha P, Barinka C, Knedlik T,
Starkova J, Lubkowski J & Konvalinka J (2012)
Efficient and versatile one-step affinity purification
of in vivo biotinylated proteins: expression,
characterization and structure analysis of
recombinant human glutamate carboxypeptidase II.
Protein Expr Purif 82, 106–115.
20 Zhang AX, Murelli RP, Barinka C, Michel J, Cocleaza
A, Jorgensen WL, Lubkowski J & Spiegel DA (2010) A
remote arene-binding site on prostate specific
membrane antigen revealed by antibody-recruiting
small molecules. J Am Chem Soc 132, 12711–12716.
immunohistochemical prognostic biomarkers of prostate
cancer—what has gone wrong? A blueprint for the way
forward in biomarker studies. Br J Cancer 112, 140–148.
16
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
21 Hlouchova K, Barinka C, Konvalinka J & Lubkowski
J (2009) Structural insight into the evolutionary and
pharmacologic homology of glutamate
exosite-binding inhibitors selective for glutamate
carboxypeptidase II. J Med Chem 58, 4357–4363.
35 Cunha AC, Weigle B, Kiessling A, Bachmann M & Rieber
EP (2006) Tissue-specificity of prostate specific antigens:
comparative analysis of transcript levels in prostate and
non-prostatic tissues. Cancer Lett 236, 229–238.
36 Barinka C, Starkova J, Konvalinka J & Lubkowski J
(2007) A high-resolution structure of ligand-free human
glutamate carboxypeptidase II. Acta Crystallogr F 63,
150–153.
carboxypeptidases II and III. FEBS J 276, 4448–4462.
22 Klusak V, Barinka C, Plechanovova A, Mlcochova P,
Konvalinka J, Rulisek L & Lubkowski J (2009)
Reaction mechanism of glutamate carboxypeptidase II
revealed by mutagenesis, X-ray crystallography, and
computational methods. Biochemistry 48, 4126–4138.
23 Mesters JR, Barinka C, Li WX, Tsukamoto T, Majer
P, Slusher BS, Konvalinka J & Hilgenfeld R (2006)
Structure of glutamate carboxypeptidase II, a drug
target in neuronal damage and prostate cancer. EMBO
J 25, 1375–1384.
24 DeLano WL (2004) Use of PYMOL as a
communications tool for molecular science. Abstr Pap
Am Chem Soc 228, U313–U314.
25 Diederichs K & Karplus PA (1997) Improved R-factors
for diffraction data analysis in macromolecular
crystallography. Nat Struct Biol 4, 269–275.
€
37 Mueller U, Darowski N, Fuchs MR, Forster R,
€
Hellmig M, Paithankar KS, Puhringer S, Steffien M,
Zocher G & Weiss MS (2012) Facilities for
macromolecular crystallography at the Helmholtz-
Zentrum Berlin. J Synchrotron Radiat 19, 442–449.
38 Kabsch W (2010) Xds. Acta Crystallogr D 66, 125–132.
39 Krug M, Weiss MS, Heinemann U & Mueller U (2012)
XDSAPP: a graphical user interface for the convenient
processing of diffraction data using XDS. J Appl
Crystallogr 45, 568–572.
26 Vaguine AA, Richelle J & Wodak SJ (1999)
SFCHECK: a unified set of procedures for evaluating
the quality of macromolecular structure-factor data and
their agreement with the atomic model. Acta
Crystallogr D 55, 191–205.
27 Dodson EJ, Winn M & Ralph A (1997) Collaborative
computational project, number 4: providing programs for
protein crystallography. Methods Enzymol 277, 620–633.
28 Davis IW, Leaver-Fay A, Chen VB, Block JN, Kapral
GJ, Wang X, Murray LW, Arendall WB, Snoeyink J,
Richardson JS et al. (2007) MolProbity: all-atom
contacts and structure validation for proteins and
nucleic acids. Nucleic Acids Res 35, W375–W383.
29 Chen VB, Arendall WB III, Headd JJ, Keedy DA,
Immormino RM, Kapral GJ, Murray LW, Richardson
JS & Richardson DC (2010) MolProbity: all-atom
structure validation for macromolecular
40 Vagin A & Teplyakov A (2000) An approach to multi-
copy search in molecular replacement. Acta Crystallogr
D 56, 1622–1624.
41 Bailey S (1994) The Ccp4 suite – programs for
protein crystallography. Acta Crystallogr D 50, 760–
763.
42 Murshudov GN, Vagin AA & Dodson EJ (1997)
Refinement of macromolecular structures by the
maximum-likelihood method. Acta Crystallogr D 53,
240–255.
43 Murshudov GN, Skubak P, Lebedev AA, Pannu NS,
Steiner RA, Nicholls RA, Winn MD, Long F & Vagin
AA (2011) REFMAC5 for the refinement of
macromolecular crystal structures. Acta Crystallogr D
67, 355–367.
44 Emsley P, Lohkamp B, Scott WG & Cowtan K (2010)
Features and development of Coot. Acta Crystallogr D
66, 486–501.
45 Winn MD, Ballard CC, Cowtan KD, Dodson EJ,
Emsley P, Evans PR, Keegan RM, Krissinel EB, Leslie
AGW, McCoy A et al. (2011) Overview of the CCP4
suite and current developments. Acta Crystallogr D 67,
235–242.
crystallography. Acta Crystallogr D 66, 12–21.
30 Ryden SE, Kirkish LS & McCann DS (1976) Evaluation
of serum ionic calcium measurement in a general hospital
population. Am J Clin Pathol 66, 634–638.
31 Magneson GR, Puvathingal JM & Ray WJ (1987) The
concentrations of free Mg2 + and free Zn2 + in
equine blood plasma. J Biol Chem 262, 11140–11148.
32 Frederickson CJ, Suh SW, Silva D, Frederickson CJ &
Thompson RB (2000) Importance of zinc in the central
nervous system: the zinc-containing neuron. J Nutr 130,
1471s–1483s.
46 Ryde U & Olsson MHM (2001) Structure, strain, and
reorganization energy of blue copper models in the
protein. Int J Quantum Chem 81, 335–347.
€
€
€
47 Ahlrichs R, Bar M, Haser M, Horn H & Kolmel C
(1989) Electronic structure calculations on workstation
computers: the program system turbomole. Chem Phys
Lett 162, 165–169.
33 Kelleher SL, McCormick NH, Velasquez V & Lopez V
(2011) Zinc in specialized secretory tissues: roles in the
pancreas, prostate, and mammary gland. Adv Nutr 2,
101–111.
48 Maier JA, Martinez C, Kasavajhala K, Wickstrom L,
Hauser KE & Simmerling C (2015) ff14SB: improving
the accuracy of protein side chain and backbone
parameters from ff99SB. J Chem Theory Comput 11,
3696–3713.
ꢁ ꢁꢀ
ꢁ
ꢀ
34 Tykvart J, Schimer J, Jancarık A, Barinkova J, Navratil
ꢀ
ꢁ
V, Starkova J, Sramkova K, Konvalinka J, Majer P &
ꢀ
ꢀ
ꢀ
ꢁ
ꢀ
Sacha P (2015) Design of highly potent urea-based,
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies
17

ꢀ
M. Navratil et al.
Substrate specificity of GCPII/III
49 Perdew JP, Burke K & Ernzerhof M (1996) Generalized
gradient approximation made simple. Phys Rev Lett 77,
3865–3868.
inhibition and enzymological analysis of recombinant
human glutamate carboxypeptidase II. J Neurochem 80,
477–487.
50 Tao J, Perdew JP, Staroverov VN & Scuseria GE
(2003) Climbing the density functional ladder:
nonempirical meta-generalized gradient approximation
designed for molecules and solids. Phys Rev Lett 91,
146401.
53 Roth M (1971) Fluorescence Reaction for Amino
Acids. Anal Chem 43, 880–882.
Supporting information
51 Guo HY, Naser SA, Ghobrial G & Phanstiel O (2002)
Synthesis and biological evaluation of new citrate-based
siderophores as potential probes for the mechanism of
iron uptake in mycobacteria. J Med Chem 45, 2056–
2063.
Additional Supporting Information may be found
online in the supporting information tab for this arti-
cle:
Data S1. Supplementary information.
52 Barinka C, Rinnova M, Sacha P, Rojas C, Majer P,
Slusher BS & Konvalinka J (2002) Substrate specificity,
18
The FEBS Journal (2016) ª 2016 Federation of European Biochemical Societies