Structural and computational basis for potent inhibition of glutamate carboxypeptidase II by carbamate-based inhibitors

Article abstract of DOI:10.1016/j.bmc.2018.11.022

A series of carbamate-based inhibitors of glutamate carboxypeptidase II (GCPII) were designed and synthesized using ZJ-43, N-[[[(1S)-1-carboxy-3-methylbutyl]amino]carbonyl]-L-glutamic acid, as a molecular template in order to better understand the impact

Full text of DOI:10.1016/j.bmc.2018.11.022

Accepted Manuscript
Structural and computational basis for potent inhibition of glutamate carboxy-
peptidase II by carbamate-based inhibitors
Cyril Barinka, Zora Novakova, Niyada Hin, Daniel Bím, Dana V Ferraris,
Bridget Duvall, Gabriel Kabarriti, Reiji Tsukamoto, Milos Budesinsky, Lucia
Motlova, Camilo Rojas, Barbara S. Slusher, Tibor András Rokob, Lubomír
Rulí šek, Takashi Tsukamoto
PII:
S0968-0896(18)31611-0
https://doi.org/10.1016/j.bmc.2018.11.022
BMC 14626
DOI:
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
27 September 2018
7 November 2018
14 November 2018
Please cite this article as: Barinka, C., Novakova, Z., Hin, N., Bím, D., V Ferraris, D., Duvall, B., Kabarriti, G.,
Tsukamoto, R., Budesinsky, M., Motlova, L., Rojas, C., Slusher, B.S., András Rokob, T., Rulí šek, L., Tsukamoto,
T., Structural and computational basis for potent inhibition of glutamate carboxypeptidase II by carbamate-based
inhibitors, Bioorganic & Medicinal Chemistry (2018), doi: https://doi.org/10.1016/j.bmc.2018.11.022
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Structural and computational basis for potent inhibition of
glutamate carboxypeptidase II by carbamate-based inhibitors
Cyril Barinka,1* Zora Novakova,1 Niyada Hin,2 Daniel Bím,5 Dana V Ferraris,6 Bridget Duvall,2
Gabriel Kabarriti,2 Reiji Tsukamoto,2 Milos Budesinsky,5 Lucia Motlova,1 Camilo Rojas,2,4 Barbara S.
Slusher,2,3 Tibor András Rokob,7† Lubomír Rulíšek,5* Takashi Tsukamoto,2,3*
1Institute of Biotechnology CAS, BIOCEV, Prumyslova 595, 252 50 Vestec, Czech Republic
3
4
2Johns Hopkins Drug Discovery, Department of Neurology, and Department of Molecular and
Comparative Pathobiology, Johns Hopkins University, Baltimore, Maryland 21205, United States
5Institute of Organic Chemistry and Biochemistry, Czech Academy of Sciences, Flemingovo nám. 2,
Prague 6, 166 10, Czech Republic
6McDaniel College, 2 College Hill, Westminster, Maryland 21157, United States
7Institute of Organic Chemistry, Research Centre for Natural Sciences, Hungarian Academy of Sciences,
1117 Budapest, Magyar Tudósok körútja 2, Hungary
*Address correspondence to:
T. T.: Tel: +1-410-614-0982; e-mail: ttskuamoto@jhmi.edu
C. B.: Tel: +420-325-873-777; e-mail: cyril.barinka@ibt.cas.cz
L. R. Tel +420-220-183-263; e-mail: lubomir.rulisek@uochb.cas.cz
†The author is no longer affiliated with the listed institution.

ABSTRACT
A series of carbamate-based inhibitors of glutamate carboxypeptidase II (GCPII) were designed and
synthesized using ZJ-43, N-[[[(1S)-1-carboxy-3-methylbutyl]amino]carbonyl]-L-glutamic acid, as a
molecular template in order to better understand the impact of replacing one of the two nitrogen atoms in
the urea-based GCPII inhibitor with an oxygen atom. Compound 7 containing a C-terminal 2-
oxypentanedioic acid was more potent than compound 5 containing a C-terminal glutamic acid (2-
aminopentanedioic acid) despite GCPII’s preference for peptides containing an N-terminal glutamate as
substrates. Subsequent crystallographic analysis revealed that ZJ-43 and its two carbamate analogs 5 and
7 with the same (S,S)-stereochemical configuration adopt a nearly identical binding mode while (R,S)-
carbamate analog 8 containing a D-leucine forms a less extensive hydrogen bonding network. QM and
QM/MM calculations have identified no specific interactions in the GCPII active site that would
distinguish ZJ-43 from compounds 5 and 7 and attributed the higher potency of ZJ-43 and compound 7 to
the free energy changes associated with the transfer of the ligand from bulk solvent to the protein active
site as a result of the lower ligand strain energy and solvation/desolvation energy. Our findings underscore
a broader range of factors that need to be taken into account in predicting ligand-protein binding affinity.
These insights should be of particular importance in future efforts to design and develop GCPII inhibitors
for optimal inhibitory potency.
Abbreviations Used: GCPII – glutamate carboxypeptidase II; PSMA – prostate-specific membrane
antigen; SAR – structure-activity relationship; NAAG – N-acetyl-aspartyl-glutamate; ZBG – zinc-binding
group; QM/MM – quantum mechanics/molecular mechanics
Keywords: glutamate carboxypeptidase II; prostate-specific membrane antigen; crystal structure;
metallopeptidase

1. Introduction
Glutamate carboxypeptidase II (GCPII, EC 3.4.17.21) is a zinc metallopeptidase that cleaves the C-
terminal glutamate from N-acetylaspartylglutamate (NAAG) producing N-acetylaspartate and glutamate.
GCPII is also known as prostate-specific membrane antigen (PSMA), a marker expressed on prostate
cancer cells, and as folate hydrolase (FOLH1) for its ability to cleave C-terminal glutamate residues from
folate polyglutamate. Therefore, the therapeutic utility of GCPII inhibitors has been investigated in a wide
range of diseases, including neurological disorders,1-3 prostate cancer,4 and inflammatory bowel disease
(IBD).5 A number of potent GCPII inhibitors have been reported to date. Nearly all of these inhibitors
have a general structure consisting of a zinc-binding group (ZBG) linked to a carboxylic acid-containing
moiety designed to occupy the S1’ glutamate recognition site of the enzyme.2 Zinc-binding groups
exploited in the design of GCPII inhibitors include phosphonate, phosphinate, phosphoramidate, thiol, and
hydroxamate groups which were previously utilized in inhibitors of other metallopeptidases. Urea-based
GCPII inhibitors first reported in 2001 by Kozikowski et al.6 are distinct from these traditional GCPII
inhibitors in that the urea group has rarely been seen in the pharmacophore of other metalloprotease
inhibitors. Since the first report, a number of urea-based GCPII inhibitors have been developed and
pharmacologically characterized.7-13 Their highly potent GCPII inhibitory activity coupled with their high
degree of chemical compatibility with conjugation processes has prompted many researchers to conjugate
them to other functional molecules to target PSMA for diagnosis and treatment of prostate cancer.4 For
instance, [177Lu]PSMA-617 (LuPSMA),14 a β-particle emitting PSMA ligand containing the urea scaffold
(Fig. 1a), is currently being tested in patients with castration-resistant prostate cancer.
Fig. 1. Chemical structures of [177Lu]PSMA-617, compounds 1-3, and ZJ-43. (a)
[177Lu]PSMA-617 contains a component acting as a urea-based GCPII inhibitor that binds

to the active site. (b) Compounds 1-3 possess either a urea or carbamate moiety. (c)
Chemical structure of ZJ-43.
More recently, in an attempt to achieve more desirable tissue distribution for radiotherapy, Yang et al.
developed a new class of GCPII inhibitors (Fig. 1b) in which either one of the two nitrogen atoms in the
urea-based inhibitors was replaced by an oxygen atom, forming a carbamate moiety. 15 Comparison of
these inhibitors that differ in the orientation of the carbamate moiety have revealed that those containing a
C-terminal 2-oxypentanedioic acid are generally more potent than those containing a C-terminal glutamic
acid (2-aminopentanedioic acid). For instance, as shown in Fig. 1, a substantial loss of potency was seen
when the N-carbamoyl glutamic acid of compound 1 is replaced by an N-carboxy-glutamic acid as seen in
compound 2. In contrast, compound 3, in which a carbamate moiety was inserted in the opposite
orientation, was found to be substantially more potent than compound 2. This SAR trend is rather
unexpected considering that the endogenous GCPII substrates possess a C-terminal glutamic acid and are
anticipated to be better mimicked by carbamates containing a C-terminal glutamic acid.
These observations prompted us to conduct systematic SAR studies of carbamate-based GCPII inhibitors
and subsequent structural and computational analyses. This is partly driven by our successful application
of these combined experimental and theoretical approaches - the latter represented mostly by hybrid
quantum and molecular mechanics (QM/MM) calculations - in understanding the GCPII catalytic actions
and its inhibition. In particular, QM/MM assisted in revealing the GCPII reaction mechanism, the role of
specific mutations in GCPII on the substrate binding16, 17 and last but not least, the molecular basis of the
binding of urea-based inhibitors in the internal cavity of GCPII.18
The inhibitory potency of urea-based inhibitors has been shown to be substantially enhanced by expanding
towards the substrate entry channel away from the glutamate recognition site 2. These high molecular
weight inhibitors may not provide suitable templates for assessing the intrinsic effects of the carbamate
moiety replacing the urea group. Consequently, we have chosen ZJ-43 (Fig. 1c),8 one of the smallest urea-
based GCPII inhibitors, as a molecular template for our studies as we expected a robust response in
inhibitory potency with modifications to the urea group. Subsequently, co-crystal structures were obtained
for some of these compounds and further investigated by QM/MM simulations to gain insights into the
structural basis for GCPII inhibition by carbamate-based inhibitors.
2. Results and discussion
2.1. Chemistry
ZJ-43 was synthesized using a method similar to that previously reported with minor modification (Fig.
2).8 Specifically, ethyl acetate was used as a solvent in the catalytic hydrogenation process. This modified
method was successfully applied to urea- and carbamate-based GCPII inhibitors 4-10 (Table 1) by

replacing benzyl leucine and/or dibenzyl glutamate systematically with other benzyl esters, including both
enantiomers of dibenzyl 2-hydroxypentanedioate and benzyl 2-hydroxy-4-methylvalerate.
Fig. 2. Synthesis of ZJ-43. Reagents and conditions: (a) triphosgene, triethylamine, DCM,
-78 °C to rt, 16 h, 62% yield; (b) H2, 5% Pd/C, EtOAc, 30 psi, 12 h, quantitative yield.
2.2. Inhibition of GCPII by urea- and carbamate-based inhibitors
Inhibitory potencies of the urea- and carbamate-based compounds were measured using N-acetyl-L-
aspartyl-[3H]-L-glutamate as a substrate and purified human recombinant GCPII. The results are
summarized in Table 1. In our assay, ZJ-43 showed subnanomolar inhibitory activity against GCPII with
an IC50 value of 0.50 nM. Substitution of D-leucine for L-leucine of ZJ-43 resulted in 32-fold loss of
GCPII inhibitory potency as observed with compound 4. Carbamate derivative 5 in which the L-leucine of
ZJ-43 was replaced by (S)-2-hydroxy-4-methylvaleric acid (L-leucic acid) also showed a weaker potency
as compared to ZJ-43 despite possessing the stereochemical configuration corresponding to that of ZJ-43.
Further decline in potency was observed upon introduction of (R)-2-hydroxy-4-methylvaleric acid (D-
leucic acid) as seen in compound 6. This finding is consistent with the trend seen in the urea series
described earlier. Subsequently, we reversed the orientation of the carbamate moiety by replacing the
glutamate portion of ZJ-43 with (S)-2-hydroxypentanedioic acid (L-2-hydroxyglutaric acid). Despite the
loss of the C-terminal glutamate moiety recognized in all known endogenous GCPII substrates, carbamate
derivative 7 was found to be substantially more potent than compound 5 with an IC50 value of 0.65 nM
and nearly equipotent to ZJ-43. The superior potency of 7 over 5 indicates that the trend initially observed
by Yang et al. applies to a wider range of carbamate-based GCPII inhibitors regardless of the size and
nature of the components expanding towards the substrate entry channel away from the glutamate
recognition site. In order to further elaborate on these findings, all remaining three diastereomers 8-10
were also tested in the GCPII assay. Substitution of D-leucine for L-leucine of 7 resulted in 14-fold loss in
potency as seen in compound 8. As described earlier, a similar degree of decline in potency was observed
when the L-leucine of ZJ-43 was replaced with D-leucine. It should be noted that compound 8 is
substantially more potent than compound 6 which shares the same stereoconfiguration but has a carbamate
moiety in the opposite orientation. As apparent from compounds 9 and 10, (R)-2-hydroxypentanedioic
acid (D-2-hydroxyglutaric acid) is not tolerated by GCPII as a substitute for (S)-2-hydroxypentanedioic
acid. This is in a good agreement with the specificity shown by GCPII towards substrates containing a
terminal L-glutamate.19

Table 1: Inhibition of GCPII by urea- and carbamate-based compounds.
Stereo-
chemistry
Stereo-
chemistry
Cmpd
Structure
IC50 (nM) Cmpd
Structure
IC50 (nM)
(S,S) 0.50 ± 0.11
(S,S) 0.65 ± 0.11
ZJ-43
7
8
(R,S) 15.9 ± 9.9
(R,S)
(S,R)
(R,R)
8.9 ± 0.1
630 ± 8
4
5
6
(S,S) 16.9 ± 7.2
(R,S) 531 ± 111
9
8530 ± 20
10
IC50 values were calculated from the inhibition curves using non-linear analysis and are shown in the
format mean ± standard deviation.
Based on our results, ZJ-43 and its analogs can be classified into four well-established clusters by their
inhibitory potency (Fig. 3). The first cluster contains the two subnanomolar inhibitors, ZJ-43 and
compound 7. The three low nanomolar inhibitors 4, 5, and 8 constitute the second cluster. The third cluster
consists of the two submicromolar inhibitors 6 and 9. Compound 10, the enantiomer (both chiral centers
are reversed) of 5, represent the least potent GCPII inhibitors with an IC50 value of 8.5 μM. As expected,
the S,S-stereoisomers (ZJ-43, compounds 5 and 7) are preferred by GCPII although the enzyme displays
some degree of tolerance to the R-configuration on the leucine side. The superior potency of compound 7
containing a terminal (S)-2-oxyglutarate over compound 5 containing a terminal L-glutamate is consistent
with the prior findings.15 Given that the endogenous GCPII substrates possess a C-terminal glutamic acid,
this SAR trend is rather unexpected considering that the endogenous GCPII substrates possess a C-
terminal glutamic acid and are anticipated and may have an important implication for the design of new
GCPII inhibitors.

Fig. 3. Dose response curves of in vitro GCPII inhibition by ZJ-43 and compounds 4-10
determined by the radioenzymatic assay.
2.3. Co-crystal structures of GCPII in complex with urea- and carbamate-based inhibitors
In order to gain further insight into the structural basis for GCPII inhibition by carbamate-based inhibitors,
we have determined co-crystal structures of GCPII in complex with compounds ZJ-43, 5, 7 and 8 to 1.52,
1.61, 1.63, and 1.68 Å resolution limits, respectively (Table 2). For all complexes, the interpretable Fo-Fc
electron density peaks representing the active site-bound ligands were observed and individual compounds
were fitted into the density in the final stages of the refinement (Fig. 4A). Positioning of ZJ-43, 5, and 7
possessing S,S-configuration in the internal cavity of GCPII is virtually indistinguishable (Fig. 4B). The
P1’ glutarate moiety occupies the S1’ pocket with the “canonical” binding mode observed previously in
other urea-based GCPII inhibitors.20, 21 At the non-prime side, the most extensive contacts are observed
between GCPII and the P1 carboxylate function. First, it forms four salt bridges with guanidinium groups
of Arg534 (2.8 Å) and Arg536 (2.9 Å and 3.0 Å) of the arginine patch. Furthermore, it is engaged in the
water-mediated H-bonds with the side chains of Ser454 (2.8 Å) and Arg534 (2.9 Å) as well as the
carbonyl of Ser517 (2.8 Å). The interaction of the glutarate portion of compound 8 possessing R,S-
configuration with GCPII appears nearly identical to that of the above three compounds (Fig. 4C), owing
to the stereochemistry corresponding to that of L-glutamate at the P1’ site. The decline in the inhibitory
potency of compound 8 can be attributed to the less extensive interaction network observed between
GCPII and the P1 carboxylate of D-leucine residue (Fig. 4D). Here, the direct interactions involve the
guanidinium group of Arg534 (2.9 Å) and the side-chain amide of Asn519 (2.9 Å) and water-mediated H-
bonds with side chains of Arg534 (2.9 Å), Arg536 (3.0 Å), and the carbonyl of Ser517 (2.7 Å).

Table 2: Crystallographic Data Collection and Refinement Statistics.
Data collection statistics
Inhibitor
PDB code
ZJ-43
6FE5
5
6EZ9
7
6F5L
8
6ETY
Space group
I222
I222
I222
I222
Unit-cell parameters a,
b, c (Å)
101.7,
130.7, 159.8
101.5, 130.8,
158.6
101.1, 130.3,
158.6
101.7, 130.7,
158.2
Wavelength (Å)
0.98
0.98
0.98
0.98
30.00-1.52
(1.61-1.52)
161939
(25398)
5.09 (4.83)
99.4 (97.3)
15.89 (2.23)
0.048
30.00-1.61
(1.71-1.61)
30 – 1.63(1.73 -
1.63)
30.00-1.68
(1.78-1.68)
Resolution limits (Å)
No of unique refl.
132737 (20491) 128331 (20423) 118798 (18462)
Redundancy
Completeness (%)
I/σ(I)
5.12 (5.03)
97.6 (94.2)
9.77 (1.99)
5.22 (5.28)
98.7 (98.2)
16.13 (2.11)
5.13 (4.95)
98.9 (96.1)
14.72 (2.08)
Rmerge
0.086 (0.382)
0.059 (0.906)
0.062 (0.693)
(0.551)
Refinement
30.00 - 1.61
(1.66 - 1.61)
30.00-1.52
(1.56-1.52)
30.00 – 1.63
(1.67 -1.63)
30.00 – 1.68
(1.72 – 1.68)
Resolution limits (Å)
Total number of
reflections
Number of reflections
in working set
Number of reflections
in test set
128675 (8838)
124617 (8547)
4058 (291)
115124 (8463)
111551 (8214)
3573 (249)
156947
(11078)
124566 (9292)
120817 (9007)
3749 (285)
4964 (375)
15.9/17.9
(26.2/26.7)
15.5/17.4
(27.1/32.1)
16.3/18.1
(24.2/26.0)
15.6/17.8
(24.7/27.4)
R/Rfree (%)
Total number of non-H
atoms
Number of non-H
protein atoms
Number of inhibitor
molecules
6579
5788
1
6600
5846
1
6638
5824
1
6843
5974
1
Number of water
molecules
568
531
579
628
Average B-factor (Å2)
29.42
27.97
38.62
26.33
32.92
31.58
41.16
27.84
32.44
31.01
40.90
25.00
33.72
31.86
43.91
26.31
Protein
Water molecules
Inhibitor
&Ramachandran Plot
(%)
Most favored
Additionally allowed
97 %
2 %
97 %
2 %
97 %
2 %
97 %
2 %
1 % (Lys
655 A/B,
Ser 656
A/B, Val
382)
1 % (Lys 655
A/B, Ser 656
A/B, Gly 335,Val
382)
1 % (Lys 655
A/B, Ser 656
A/B, Val 382)
1 % (Ser 656
A/B, Gly335,
Val 382)
Disallowed
R.m.s. deviations:
bond lengths (Å)
bond angles (°)
0.015
1.68
0.015
1.64
0.018
0.016
1.79
0.13
1.75
0.13
chiral centers (Ǻ3)
0.12
0.12

missing residues
AA 44-54
AA 44-54
AA 44-54
AA 44-54
* Values in parentheses are for the highest resolution shells.
& Structures were analyzed using the MolProbity package
Fig. 4. Structural characterization of GCPII-inhibitor complexes. (a) Fo-Fc maps (green)
for studied inhibitors are contoured at 3.0 σ and modeled inhibitors are shown in stick
representation with atoms colored green (carbon), red (oxygen), and blue (nitrogen). The
active-site zinc ions are shown as orange spheres. (b) Superposition of SS diastereomers
ZJ-43 (pink), 5 (green), and 7 (light orange) in the internal pocket of GCPII. (c)

Superposition of 7 (light orange) and its R,S-counterpart 8 (cyan). (d) Comparison of the
interaction network of the P1 carboxylate function with GCPII residues. Inhibitors 7 and 8
are shown in stick representation, GCPII residues as lines and water molecules as red
spheres. H-bonds and salt bridges are shown as dashed lines, distances in Ångströms.
2.4. QM/MM calculations
Given that X-ray data at >1.1 Å resolution do not enable direct observation of hydrogen atoms, our
structural data cannot provide a mechanistic explanation for the observed difference in inhibitory potency
between the two carbamate-based inhibitors 5 and 7. To obtain refined binding modes of the individual
compounds and insight into the molecular origin of the inhibitory potency, QM/MM calculations of the
three inhibitors ZJ-43, 5, and 7 bound to GCPII were carried out. The details of the computational
protocols, including the protein preparation and equilibration are mentioned below (Materials and
Methods section). The QM regions (324 atoms, or 325 atoms in the case of urea) of the equilibrium
QM/MM structures are depicted in Fig. 5 (coordinates are deposited as supplementary data). As shown in
Fig. 5, our QM/MM calculations did not reveal any ‘visible’ differences in the binding of ZJ-43, 5, and 7.
The binding modes of all three inhibitors, (a) ZJ-43, (b) compound 5, and (c) compound 7 are essentially
identical, resulting in similar hydrogen bonding pattern. A small difference is observed for the interaction
of urea based inhibitor ZJ-43 with Gly 518 residue, forming two hydrogen bonds using both NH groups
of the central urea unit. Since the hydrogen bonding is equivalent for both compounds 5 and 7, such an
interaction cannot be responsible for the lower inhibitory potency of compound 5. Thus, the diverse
inhibitory potency, particularly the lower potency of compound 5 as compared to ZJ-43 and 7, does not
appear to be originated from any specific ligand-protein interactions.
Fig. 5. Details of hydrogen-bonding interactions between GCPII residues and (a) ZJ-43, (b)
compound 5, and (c) compound 7. H-bonds shorter than 3.5 Å are shown as broken lines.
2.5. Calculations of ligand strain energies and solvation/desolvation energies

Subsequently, we examined whether the experimental data can be attributed to the different strain energy
of each ligand, i.e. the alteration of ligand structure from its global energetic minimum in bulk solvent
(water) to the bound state in the enzyme pocket and to the solvation/desolvation penalty of the compound
(ligand).
To map the conformational space of each compound, the exhaustive DFT-D3/COSMO-RS//MD-LLMOD
conformational sampling22 was carried out, followed by the highly-accurate DLPNO-CCSD(T)
calculations. This combination enables us to obtain a quantitative insight into the energetics governing
GCPII affinity for individual compounds. The energy data are summarized in Table 3 whereas full
energetic data are deposited as supplementary data (Table S1). The results suggeset that the lower
inhibitory activity of 5 originates in its unfavorable strain energy. In comparison to ZJ-43 and compound
7, Gstrain of compound 5 is higher by 3.2 kcal·mol-1 and 1.4 kcal·mol-1, respectively.
To further validate the accuracy and completeness of the conformational sampling algorithm, the low–
lying minima were compared with the experimental NMR data. Solely for the purpose of the comparison,
the conformer’s free energies were also computed in DMSO (used as the solvent in NMR measurements).
All three obtained global minima (deposited as full 3-D structural data in the SI) are fully consistent with
the J-coupling constants provided by the NMR (Table S2). This gives a validation of the computational
strategy used. It can be also mentioned that the order of the energetically low-lying isomers changes
between water (target solvent) and DMSO (solvent used for NMR).
We estimated solvation/desolvation energy by the relative differences between the COSMO-RS solvation
energies in water and octanol. Octanol represents a “generic” protein environment and therefore, even in
the relative scale, we admittedly neglect any specific interactions with the protein (more precisely,
interactions that would be different for the three ligands). The validity of such an approach is, of course,
justified only for the highly similar ligands, such as those studied herein. As can be seen in Table 3, adding
the solvation/desolvation energies to the strain energies for the studied compounds, the quantitative
agreement between calculation and experimental binding constants is obtained. This suggests that the
differences in binding of the three ligands shall be attributed to the combination of the ligand strain
energies and their differences in the octanol/water solvation and not to any specific interaction originating
in the protein.
Table 3: The computed ligand strain energies and solvation/desolvation energies for ZJ-43, 5, and 7.
Gglobal-local
Glocal-enzyme
Gstrain
ΔGdesolv
ΔGbind
(referenced to ZJ-43)
3.1
5.3
2.7
1.1
2.1
3.2
4.2
7.4
6.0
0.0
-0.9
-1.8
4.2
6.5
4.2
ZJ-43
5
7

All values are reported in kcal mol-1. Gglobal-local is the difference in Gibbs free energy of a substrate
corresponding to reorganization from the equilibrium geometry in solvent to the local energy minimum
obtained by the geometry optimization of the enzyme-bound structure; Glocal-enzyme is the change in the
Gibbs free energy of a substrate corresponding to reorganization from the above local minimum to the
exact enzyme-bound geometry (provided by the QM/MM); Gstrain is the sum of the first two terms; Gdesolv
corresponds to the Gibbs free energy change due to transfer from water to 1-octanol (referenced to the
ZJ-43); and ΔGbind is the sum of Gstrain and Gdesolv. This can be related to the inhibitory affinity of each
substrate.
3. Conclusion
Since the first co-crystal structure was solved for GCPII in complex with an inhibitor in 2006,23 many
additional structures have been reported for various GCPII-ligand complexes. Structure-based drug design
has consequently played a key role in guiding the design of GCPII inhibitors.24 While these efforts have
been primarily focused on the analysis of ligand-protein interactions, our SAR findings coupled with
crystallographic analysis and comprehensive computational studies highlight a broader range of factors
that need to be taken into account in predicting binding affinity to GCPII. These insights should be of
particular importance in future efforts to design and develop GCPII inhibitors for optimal inhibitory
potency.
4. Materials and methods
4.1. Chemical Synthesis of urea- and carbamate-based GCPII inhibitors
4.1.1. General
All solvents were reagent grade or HPLC grade. Unless otherwise noted, all materials were obtained from
commercial suppliers and used without further purification. Melting points were obtained on a Mel-Temp
1
apparatus and are uncorrected. H NMR spectra were recorded at 400 MHz. Chemical shifts are reported
in parts per million relative to TMS. The HPLC solvent system consisted of distilled water and
acetonitrile, both containing 0.1% formic acid. Analytical HPLC was performed on an Agilent 1200 Series
HPLC system equipped with an Agilent G1315D DAD detector (detection at 220 nm) and an Agilent
6120 Quadrupole MS detector. Unless otherwise specified, the analytical HPLC conditions involve a
gradient of 5% acetonitrile/95% water for 0.5 min followed by an increase to 40% acetonitrile/60% water
over 1.75 min and continuation of 40% acetonitrile/60% water for 1 min with an Eclipse Plus C18 column
(2.1 mm × 50 mm, 3.5 μm) at a flow rate of 1.25 mL/min. All final compounds tested were confirmed to
be of >95% purity by the analytical HPLC methods described above unless otherwise noted.
4.1.2. Synthesis of (S)-2-(3-((S)-1-carboxy-3-methylbutyl)ureido)pentanedioic acid (ZJ-43).
4.1.2.1. (S)-Dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentanedioate. A
solution of (S)-dibenzyl 2-aminopentanedioate 1.38 g (4.22 mmol, 1.0 eq.) in anhydrous DCM (60 mL) at

-78 °C was added triphosgene (426 mg, 1.43 mmol,0.34 eq.) dropwise in DCM (20 mL) and then
triethylamine (640 mg, 1.5 eq.). Upon warming to rt, (S)-benzyl 2-amino-4-methylpentanoate (930 mg,
4.22 mmol, 1.0 eq.), and triethylamine (430 mg, 1.0 eq.) were added and the reaction was stirred
overnight. The organic phase was washed with water (2 x 50 mL), brine (50 mL), and dried over MgSO4,
concentrated. The reside purified by silica gel chromatography (30%-50% EtOAc/hexanes) to give 1.50 g
of the product as a white solid (62% yield); 1H NMR (CDCl3) δ 0.91 (dd, J = 6.3, 3.5 Hz, 6 H) 1.20 - 1.35
(m, 3 H) 1.41 - 1.73 (m, 5 H) 1.92 - 2.05 (m, 1 H) 2.21 (d, J = 8.1 Hz, 1 H) 2.35 - 2.57 (m, 2 H) 4.14 (q, J
= 7.1 Hz, 2 H) 4.45 - 4.65 (m, 2 H) 4.95 (d, J = 8.6 Hz, 1 H) 5.07 - 5.27 (m, 7 H) 7.29 - 7.41 (m, 15 H).
4.1.2.2. (S)-2-(3-((S)-1-Carboxy-3-methylbutyl)ureido)pentanedioic acid (ZJ-43). A suspension of (S)-
dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentanedioate (1.20 g, 2.08 mmol)
and 5% Pd/C (106 mg, 0.94 mmol) in EtOAc (50 mL) was hydrogenated at 30 psi for 12 hours, filtered
through a pad of celite, and then triturated with diethyl ether to give 633 mg of ZJ-43 as a white solid
(quantitative yield); mp 64-67 °C; 1H NMR (DMSO-d6)  0.87 (dd, J = 12.5, 6.7 Hz, 6 H) 1.33 - 1.55 (m,
2 H) 1.54 - 1.77 (m, 2 H) 1.81 - 2.03 (m, 1 H) 2.15 - 2.37 (m, 2 H) 3.99 - 4.25 (m, 2 H) 6.27 (dd, J = 10.1,
8.6 Hz, 2 H) 12.42 (brs, 3 H). Analytical HPLC; retention time 0.617 min, m/z 303.1 [M - H]-.
4.1.3. Synthesis of (S)-2-(3-((R)-1-carboxy-3-methylbutyl)ureido)pentanedioic acid (4)
4.1.3.1. (S)-Dibenzyl 2-(3-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentanedioate. (S)-
Dibenzyl 2-(3-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido) was prepared as was described for
(S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentane-dioate except (R)-benzyl
2-amino-4-methylpentanoate was used in place of (S)-benzyl 2-amino-4-methylpentanoate. The reside
purified by silica gel chromatography (1% MeOH/chloroform): white solid (37% yield); 1H NMR (CDCl3)
δ 0.90 (d, J = 2.1 Hz, 3H), 0.91 (d, J = 2.0 Hz, 6H), 1.46-1.53 (m, 1H), 1.57-1.69 (m, 2H), 1.98-2.04 (m,
1H), 2.18-2.23 (m, 1H), 2.39-2.49 (m, 2H), 4.51-4.57 (m, 2H), 5.04-5.26 (m, 8H), 7.31-7.38 (m, 15).
4.1.3.2. (S)-2-(3-((R)-1-Carboxy-3-methylbutyl)ureido)pentanedioic acid (4). Compound 4 was
prepared as was described for ZJ-43 except (S)-dibenzyl 2-(3-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-
2-yl)ureido)pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate and the material was hydrogenated at 1 atm for 60 minutes.
Filtration through celite afforded a white solid (80% yield); mp 50-55 °C; 1H NMR (D2O) δ 0.81 (d, J =
5.9 Hz, 3H), 0.85 (d, J = 5.9 Hz, 3H), 1.54 (t, J = 7.2 Hz, 2H), 1.58-1.64 (m, 1H), 1.86-1.93 (m, 1H), 2.09-
2.14 (m, 1H), 2.41 (t, J = 7.4 Hz, 2H), 4.14-4.22 (m 2H). Analytical HPLC; retention time 0.74 min, m/z
303.1 [M - H]-.
4.1.4. Synthesis of (S)-2-(((S)-1-carboxy-3-methylbutoxy)carbonylamino)pentanedioic acid (5)

4.1.4.1. (S)-Dibenzyl 2-(((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yloxy)carbonylamino)-
pentanedioate. (S)-Dibenzyl 2-(((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yloxy)carbonylamino)-
pentanedioate was prepared as was described for (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate except (S)-benzyl 2-hydroxy-4-methylpentanoate25 was used in
place of (S)-benzyl 2-amino-4-methylpentanoate. The residue purified by silica gel chromatography (25%-
40% EtOAc/hexanes), but used in the subsequent step without further purification or characterization
(62% yield).
4.1.4.2. (S)-2-(((S)-1-Carboxy-3-methylbutoxy)carbonylamino)pentanedioic acid (5). Compound 5
was prepared as was described for ZJ-43 except (S)-dibenzyl 2-(((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yloxy)carbonylamino)-pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-
(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded a white
solid (99% yield); mp 38-40 °C; 1H NMR (D2O) δ 0.77-0.81 (m, 6H), 1.49-1.55 (m, 1H), 1.62-1.69 (m,
2H), 1.81-1.87 (m, 1H), 2.07-2.13 (m, 1H), 2.40 (t, J = 7.4 Hz, 2H), 4.08-4.21 (m, 1H), 4.76-4.79 (m, 1H).
LCMS (20% acetonitrile/80% H2O for 0.25 minutes followed by an increase to 85% acetonitrile/15% H2O
over 1.75 minutes and continuation of 85% acetonitrile/15% H2O until 4 minutes): retention time 0.28
min, m/z 304.1[M - H]-.
4.1.5. Synthesis of (S)-2-(((R)-1-carboxy-3-methylbutoxy)carbonylamino)pentanedioic acid (6)
4.1.5.1. (S)-Dibenzyl 2-(((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yloxy)carbonylamino)-
pentanedioate. (S)-Dibenzyl 2-(((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-yloxy)carbonylamino)-
pentanedioate was prepared as was described for (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate except (R)-benzyl 2-hydroxy-4-methylpentanoate was used in place
of (S)-benzyl 2-amino-4-methylpentanoate. The residue purified by silica gel chromatography (20%
EtOAc/hexanes): white solid (40% yield);1H NMR (CDCl3) δ 0.92 (m, 6H), 1.60-1.66 (m, 1H), 1.72-1.78
(m, 2H), 2.22-2.33 (m, 2H), 2.37-2.50 (m, 2H), 4.43-4.46 (m, 1H), 5.01-5.05 (m, 1H), 5.10 (s, 2H), 5.17-
5.21 (m, 4H), 5.56 (d, J = 7.7 Hz, 1H), 7.31-7.38 (m, 15H).
4.1.5.2. (S)-2-(((R)-1-Carboxy-3-methylbutoxy)carbonylamino)pentanedioic acid (6). Compound 6
was prepared as was described for ZJ-43 except (S)-dibenzyl 2-(((R)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yloxy)carbonylamino)-pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-
(benzyloxy)-4-methyl-1-oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded a white
solid (99% yield); mp 59-60 °C; 1H NMR (DMSO-d6)  0.82-0.92 (m, 6H), 1.46-1.57 (m, 1H), 1.60-1.79
(m, 3H), 1.89-1.99 (m, 1H), 2.24-2.35 (m, 2H), 3.92-3.98 (m, 1H), 4.65 (dd, J = 10.2, 3.2 Hz, 0.2H, minor
rotomer, 4.74 (dd, J = 9.2, 4.6 Hz, 0.8H, major rotomer), 7.24 (d, J = 7.9 Hz, 0.2H, minor rotomer), 7.69
(d, J = 7.9 Hz, 0.8H, major rotomer), 12.96 (brs, 3H). Analytical HPLC; retention time 0.422 min, m/z

304.1[M - H]-. Analytical HPLC; retention time 0.60 min, m/z 304.1[M - H]-.
4.1.6. Synthesis of (S)-2-((S)-1-carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (7)
4.1.6.1. (S)-Dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate.
(S)-Dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate was prepared
in a similar manner to (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
yl)ureido)pentanedioate, but with minor modifications. (S)-dibenzyl 2-hydroxypentanedioate 26 was added
to a solution of triphosgene 112 mg, 0.4 mmol, 0.5 eq.) and anhydrous pyridine (66 mg, 0.84 mmol, 1.1
eq.) in anhydrous DCM (30 mL) at 0 °C. The solution was stirred at this temperature for 45 minutes, was
washed with water (10 mL), and then the DCM layer was dried over MgSO4, filtered, and concentrated.
Without further purification or characterization, to the crude chloroformate (300 mg, 0.76 mmol) in
anhydrous THF (20 mL) at 0 °C was added DIEA (250 mg, 1.9 mmol, 2.5 eq.) and then (S)-benzyl 2-
amino-4-methylpentanoate hydrochloride (200 mg, 0.76 mmol, 1.0 eq.). The solution was stirred at this
temperature for 2h and then concentrated. The residue was dissolved in EtOAc (30 mL) and washed with
10% KHSO4 (25 mL), then sat’d NaHCO3 (25 mL), dried over MgSO4, filtered, and concentrated. The
reside was purified by silica gel chromatography (25%-40% EtOAc/hexanes) to give 340 mg of the
product as a white solid (78% yield); 1H NMR (CDCl3) δ 0.90 (dd, J = 6.4, 2.2 Hz, 6 H) 1.49-1.56 (m,
1H), 1.60-1.69 (m, 2H), 2.12-2.20 (m, 1H), 2.21-2.30 (m, 1H), 2.40-2.52 (m, 2H), 4.38-4.44 (m, 1H),
5.07-5.25 (m, 8H), 7.33-7.79 (m, 15H).
4.1.6.2. (S)-2-((S)-1-Carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (7). Compound 7 was
prepared as was described for ZJ-43 except (S)-dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
ylcarbamoyloxy)pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded 7 as a white solid (99% yield); mp
1
37-42 °C; H NMR (DMSO-d6) δ 0.84 (d, J = 6.4 Hz, 3H), 0.88 (d, J = 6.4 Hz, 3H), 1.42-1.57 (m, 2H),
1.63-1.69 (m, 1H), 1.85-1.90 (m, 1H), 1.96-2.04 (m, 1H), 2.27-2.42 (m, 2H), 3.90-3.96 (m, 1H), 4.71-4.75
(m, 1H), 7.65 (d, J = 8.3 Hz, 1H), 12.57 (brs, 3H). LCMS (20% acetonitrile/80% H2O for 0.25 minutes
followed by an increase to 85% acetonitrile/15% H2O over 1.75 minutes and continuation of 85%
acetonitrile/15% H2O until 4 minutes): retention time 0.43 min, m/z 304.1[M - H]-.
4.1.7. Synthesis of (S)-2-((R)-1-carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (8).
4.1.7.1. (S)-Dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate.
(S)-Dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate was prepared
as was described for (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
yl)ureido)pentanedioate except (R)-benzyl 2-amino-4-methylpentanoate hydrochloride was used in place

of (S)-benzyl 2-amino-4-methylpentanoate hydrochloride and (S)-dibenzyl 2-hydroxypentanedioate was
used in place of (S)-dibenzyl 2-aminopentanedioate . The reside was purified by silica gel chromatography
(25%-40% EtOAc/hexanes): white solid (45% yield); 1H NMR (CDCl3) δ 0.92 (m, 6H), 1.27 (m, 1H),
1.53-1.66 (m, 2H), 2.16-2.28 (m, 2H), 2.42 -2.51 (m, 2H), 4.43 (m, 1H), 5.08-5.21 (m, 6H), 5.31 (s, 1H),
7.33-7.36 (m, 15H).
4.1.7.2. (S)-2-((R)-1-Carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (8). Compound 8 was
prepared as was described for ZJ-43 except (S)-dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
ylcarbamoyloxy)pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded a white solid (99% yield); mp 48-
50 °C;1H NMR (D2O) δ 0.83 (d, J = 6.3 Hz, 3H), 0.86 (d, J – 6.3 Hz, 3H), 1.53-1.67 (m, 3H), 2.01-2.17
(m, 2H), 2.41-2.51 (m, 2H), 4.10 (m, 0.8H, major rotomer), 4.22 (m, 0.2H, minor rotomer), 4.86-4.88 (m,
1H). LCMS (20% acetonitrile/80% H2O for 0.25 minutes followed by an increase to 85% acetonitrile/15%
H2O over 1.75 minutes and continuation of 85% acetonitrile/15% H2O until 4 minutes): retention time
0.44 min, m/z 304.1[M - H]-.
4.1.8. Synthesis of (R)-2-((S)-1-carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (9).
4.1.8.1. (R)-Dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate.
(R)-Dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate was prepared
as was described for (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
yl)ureido)pentanedioate except (R)-dibenzyl 2-hydroxypentanedioate26 was used in place of (S)-dibenzyl
2-aminopentanedioate. The reside purified by silica gel chromatography (25%-40% EtOAc/hexanes):
white solid (76% yield); 1H NMR (CDCl3) δ 0.90-0.93 (m, 6H), 1.51-1.68 (m, 3H), 2.15-2.26 (m, 2H),
2.40-2.51 (m, 2H), 4.40-4.41 (m, 1H), 5.06-5.21 (m, 6H), 5.31 (d, J = 8.2 Hz, 1H), 7.33-7.36 (m, 15H).
4.1.8.2. (R)-2-((S)-1-Carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (9). Compound 9 was
prepared as was described for ZJ-43 except (R)-dibenzyl 2-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
ylcarbamoyloxy)pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded a white solid (99% yield); mp 37-
40 °C; 1H NMR (D2O) δ 0.83 (d, J = 6.3 Hz, 3H), 0.86 (d, J = 6.3 Hz, 3H), 1.50-1.57 (m, 3H), 1.97-2.09
(m, 2H), 2.37-2.48 (m, 2H), 4.09-4.13 (m, 0.7H, major rotomer), 4.20-4.24 (m, 0.3H, minor rotomer),
4.80-4.87 (m, 1H). LCMS (20% acetonitrile/80% H2O for 0.25 minutes followed by an increase to 85%
acetonitrile/15% H2O over 1.75 minutes and continuation of 85% acetonitrile/15% H2O until 4 minutes):
retention time 0.43 min, m/z 304.1[M - H]-.
4.1.9. Synthesis of (R)-2-((R)-1-carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (10).

4.1.9.2. (R)-Dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate.
(R)-Dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-ylcarbamoyloxy)pentanedioate was prepared
as was described for (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
yl)ureido)pentanedioate except (R)-dibenzyl 2-hydroxypentanedioate was used in place of (S)-dibenzyl 2-
aminopentanedioate and (R)-benzyl 2-amino-4-methylpentanoate was used in place of (S)-benzyl 2-
amino-4-methylpentanoate. The residue purified by silica gel chromatography (25%-40%
EtOAc/hexanes), but used in the subsequent step without further purification (40% yield).
4.1.9.2. (R)-2-((R)-1-Carboxy-3-methylbutylcarbamoyloxy)pentanedioic acid (10). Compound 10, was
prepared as was described for ZJ-43 except (R)-dibenzyl 2-((R)-1-(benzyloxy)-4-methyl-1-oxopentan-2-
ylcarbamoyloxy)pentanedioate was used in place of (S)-dibenzyl 2-(3-((S)-1-(benzyloxy)-4-methyl-1-
oxopentan-2-yl)ureido)pentanedioate. Filtration through celite afforded a white solid (99% yield); mp 40-
43 °C; 1H NMR (D2O) δ 0.82 (d, J = 6.3 Hz, 3H), 0.86 (d, J = 6.3 Hz, 3H) 1.56-1.68 (m, 3H), 2.02-2.19
(m, 2H), 2.45-2.50 (m, 2H), 4.03-4.13 (m, 1H), 4.83-4.87 (m, 1H). LCMS (20% acetonitrile/80% H2O for
0.25 minutes followed by an increase to 85% acetonitrile/15% H2O over 1.75 minutes and continuation of
85% acetonitrile/15% H2O until 4 minutes): retention time 0.42 min, m/z 304.1[M - H]-.
4.2. GCPII Assay
4.2.1. GCPII expression and purification
Expression and purification of the extracellular domain of human GCPII (amino acids 44 – 750) were
carried out as described previously.27 N-terminally tagged GCPII was overexpressed in Schneider’s S2
cells and concentrated to 1/10 of the original volume using tangential flow filtration TFF (Millipore
Mosheim, France). Concentrated medium was dialyzed against 50 mM Tris-HCl, 150 mM NaCl, pH 8.0,
at 4°C overnight and affinity purified on the Streptactin Sepharose column (IBA, Germany). The N-
terminal Strep tag was cleaved by the TEV protease (1:50 molar ratio) at 4°C overnight and the final
purification step on Superdex 200 16/60 column (GE Healthcare Bio-Sciences, Uppsala, Sweden) in 20
mM Tris-HCl, 150 mM NaCl, pH 8.0, yielded GCPII with >98% purity. The protein was concentrated to
10 mg/mL and kept at -80°C until further use.
4.2.2, IC50 determination
Inhibition constants of studied inhibitors were determined using the radioenzymatic assay with 3H-NAAG
(radiolabeled at the terminal glutamate) as a substrate.28 Briefly, rhGCPII (30 ng/ml) was preincubated in
the presence of increasing concentrations of inhibitors in 20 mM Tris, 150 mM NaCl, pH 8.0, for 15 min
at 37°C in the total volume of 80 µl. The reaction was initiated by addition of 40 µl of 0.31 µM NAAG
(Sigma) and 15 nM 3H-NAAG (50 Ci/mmol, Perkin Elmer) mixture (120 µl total reaction volume). The
reaction was terminated after 20 mins by 120 µl of 200 mM potassium phosphate, 50 mM EDTA, 2 mM

β-mercaptoethanol, pH 7.4. The released glutamate was separated from the reaction mixture by ion-
exchange chromatography and quantified by liquid scintillation. Duplicate reactions were carried out for
each experimental point. The data were fitted using the GraphPad Prism software (GraphPad Software,
San Diego, CA, USA) and IC50 values were calculated from the inhibition curves of three independent
experiments using a non-linear analysis protocol.
4.3. Structural biology
4.3.1. Crystallization and data collection
Diffracting crystals of GCPII/inhibitor complexes were obtained using procedures described previously 20.
Briefly, GCPII (10 mg/ml) was mixed with a 20 mM stock solution of a given inhibitor in water
(neutralized with NaOH) at the 20:1 molar ratio and the GCPII/inhibitor solution was then combined with
an equivalent volume of the reservoir solution (33% pentaerythritol propoxylate, (Sigma), 2.0%
polyethylene glycol 3350 (Sigma), and 100 mM Tris–HCl, pH 8.0). Diffraction quality crystals grew in
hanging-drops at 293 K for 14 days and then were harvested and flash frozen in liquid nitrogen directly
from crystallization droplets. Diffraction intensities for each complex were collected at 100 K using
synchrotron radiation at MX P13 beamline (EMBL Hamburg at the PETRA III storage ring, DESY,
Hamburg, Germany; 0.98 Å) equipped with the Dectris PILATUS 3 6M (Dectris AG, Swizerland). The
complete dataset for each complex was collected from a single crystal and data were processed using XDS
29. The final data collection statistics are shown in Table 2.
4.3.2. Structure determination, refinement, and analysis
Difference Fourier methods were used to determine structures of GCPII/inhibitor complexes with the
3D7H structure used as a starting model. Calculations were performed using Refmac 5.5.30 and the
refinement protocol was interspersed with manual corrections to the model employing the program Coot
0.6.31 The PRODRG server was used to generate restrains library and coordinate files for individual
inhibitors.32 Inhibitors were fitted into the positive electron density map in the final stages of the
refinement. Approximately 1,000 of the randomly selected reflections were kept aside for cross-validation
(Rfree) during the refinement process. The quality of the final models was evaluated using the
MOLPROBITY software33 and relevant statistics are summarized in Table 2.
4.4. Computational methods
4.4.1. QM/MM calculations
The computational protocols and the general QM/MM setup that was previously successfully applied in
mechanistic and structural studies of GCPII16, 18, 34 has been employed in this work. This also concerns the
initial protein structure (further denoted St1) used in the modelling. The St1 is the fully equilibrated all-
atom QM/MM model originating in the 1.71 Å resolution X-ray structure of the E424A mutant with the
N-Ac-Asp-Glu ligand (PDB code 3BXM). This included a few missing residues in the 3BXM crystal

structure, added solvation sphere of ~12 700 water molecules (radius of 50 Å from the center of the
protein), and two Na+ ions to neutralize the overall charge of the protein. Apart from utilizing the previous
(non-trivial) protein preparation and equilibration,34 we believe that the adopted approach also warrants
unbiased structure of the ligand binding. The full St1 structure (in PDB format) is deposited as the
supporting information. The ligands were then positioned into the St1 by its superimposition with
particular ligand-protein crystal structure reported in the experimental part of this work. The quantum
system consisted of 324 atoms for 7 and 5 ligand structures, and of 325 atoms for urea-based ZJ-43,
whereas the part of the protein that was allowed to relax in the QM/MM calculations comprised 29
residues and 10 water molecules in the vicinity of the QM system (R= 2.0 Å) – the same for all three
studied ligand-protein complexes.
4.4.2. Calculations of ligand strain energies and solvation/desolvation energies
The Gibbs free energy of the equilibrium structure found by the DFT-D3/COSMO-RS//MD-LLMOD
methodology was calculated as: G = Eel + Gsolv + EZPVE + pV – RT lnQ, where i) Eel term was obtained
using the DLPNO-CCSD(T)/aug-cc-pVDZ method35 as implemented in ORCA 4.0 program;36 ii) EZPVE
pV – RTlnQ term was obtained from frequency calculations with the rigid rotor/harmonic oscillator
approximation (for p = 1 bar); and iii) the Gsolv was calculated using the conductor-like screening model
for realistic solvation (COSMO-RS) with the radii-based isosurface cavity and the COSMOtherm
+
parameter set BP_TZVPD_FINE_C30_1701.ctd as available in COSMOtherm17 37, 38. The COSMO-RS
calculations were carried out following the recommended protocol: RI-BP86-D3/def2-TZVPD/cosmo (ε =
∞) ≡ ECOSMO,∞ and RI-BP86-D3/def2-TZVPD/in vacuo ≡ Ein vacuo single-point calculations were used to
calculate Gsolv = (ECOSMO,∞ – Ein vacuo) + μCOSMO-RS(T), where μ(T) is the temperature dependent COSMO-
RS chemical potential, as defined in 37. From the practical reasons, we further decompose the strain energy
into two parts: i) from equilibrium conformation to the local minimum obtained by geometry optimization
of the enzyme-bound geometry (Eglobal-local), and ii) from local minimum to the geometry in enzyme (Elocal-
enzyme). The first can be calculated using all contributions to the Gibbs free energy, while the second
neglects the thermal contributions to the Gibbs free energy, originating from the frequency calculations
(the EZPVE + pV – RT lnQ term), as the enzyme-bound conformation is in constrained geometry invoked by
the enzyme environment. The geometry optimizations of the enzyme-bound structures were employed
using the RI-TPSS-D3/def2-SV(P)/cosmo(ε = 80) method 39-42 with the Turbomole 7.2.1 program
(http://www.turbomole.com). The level of theory is consistent with that of the QM/MM calculations. The
desolvation energy (Gdesolv) is estimated from the difference in Gsolv calculated for the enzyme-bound
geometry for water and for 1-octanol, mimicking the enzyme environment. This is referenced to the urea-
based inhibitor ZJ-43, for which the ΔGsolv was set to zero.
Acknowledgment:

We acknowledge Iva Jelinkova, Barbora Havlinova, and Petra Baranova for the excellent technical
assistance and DESY (Hamburg, Germany), a member of the Helmholtz Association HGF, for the use of
experimental facilities. Parts of this research were carried out at the PETRA III storage ring and we thank
Isabel Bento for assistance in using the P13 beamline. This work was in part supported by the CAS (RVO:
86652036), MEYS CR (LM2015043 CIISB), and projects CIISB4HEALTH
(CZ.02.1.01/0.0/0.0/16_013/0001776) and BIOCEV (CZ.1.05/1.1.00/02.0109) from the ERDF and by
NIH grants R01CA161056 and P30MH075673 awarded to BSS. LR acknowledges the support from Grant
Agency of the Czech Republic (grant 17-24155S) and computer time at IT4I infrastructure (MEYS grant
LM2015070, Large Infrastructures for Research, Experimental Development and Innovations:
IT4Innovations National Supercomputing Center).
Supplementary Data
Table S1 Primary energy data: molecular energies in gas-phase, E(el), solvation energies, G(solv) and
thermal corrections.
Table S2 1H and 13C NMR data of compounds ZJ-43, 5, and 7
Table S3 1H spectrra of compounds 4-10.
Coordinates.zip Equilibrium geometries of the QM systems and one all-atom PDB structure (denoted St1
above)

References
[1] J. Zhou, J.H. Neale, M.G. Pomper, et al., NAAG peptidase inhibitors and their potential for diagnosis
and therapy. Nat. Rev. Drug Discov. 2005;4:1015-1026.
[2] D.V. Ferraris, K. Shukla, T. Tsukamoto, Structure-activity relationships of glutamate carboxypeptidase II
(GCPII) inhibitors. Curr. Med. Chem. 2012;19:1282-1294.
[3] J.J. Vornov, K.R. Hollinger, P.F. Jackson, et al., Still NAAG'ing After All These Years: The Continuing
Pursuit of GCPII Inhibitors. Adv Pharmacol. 2016;76:215-255.
[4] M. Eiber, W.P. Fendler, S.P. Rowe, et al., Prostate-Specific Membrane Antigen Ligands for Imaging and
Therapy. J Nucl Med. 2017;58:67S-76S.
[5] R. Rais, W. Jiang, H. Zhai, et al., FOLH1/GCPII is elevated in IBD patients, and its inhibition ameliorates
murine IBD abnormalities. JCI Insight. 2016;1.
[6] A.P. Kozikowski, F. Nan, P. Conti, et al., Design of remarkably simple, yet potent urea-based inhibitors
of glutamate carboxypeptidase II (NAALADase). J. Med. Chem. 2001;44:298-301.
[7] A.P. Kozikowski, J. Zhang, F. Nan, et al., Synthesis of urea-based inhibitors as active site probes of
glutamate carboxypeptidase II: efficacy as analgesic agents. J. Med. Chem. 2004;47:1729-1738.
[8] R.T. Olszewski, N. Bukhari, J. Zhou, et al., NAAG peptidase inhibition reduces locomotor activity and
some stereotypes in the PCP model of schizophrenia via group II mGluR. J Neurochem. 2004;89:876-885.
[9] R.C. Mease, C.L. Dusich, C.A. Foss, et al., N-[N-[(S)-1,3-Dicarboxypropyl]carbamoyl]-4-
[18F]fluorobenzyl-L-cysteine, [18F]DCFBC: a new imaging probe for prostate cancer. Clin Cancer Res.
2008;14:3036-3043.
[10] R.T. Olszewski, M.M. Wegorzewska, A.C. Monteiro, et al., Phencyclidine and dizocilpine induced
behaviors reduced by N-acetylaspartylglutamate peptidase inhibition via metabotropic glutamate
receptors. Biol Psychiatry. 2008;63:86-91.
[11] T. Yamamoto, A. Kozikowski, J. Zhou, et al., Intracerebroventricular administration of N-
acetylaspartylglutamate (NAAG) peptidase inhibitors is analgesic in inflammatory pain. Molecular pain.
2008;4:31.
[12] A.X. Zhang, R.P. Murelli, C. Barinka, et al., A remote arene-binding site on prostate specific
membrane antigen revealed by antibody-recruiting small molecules. J Am Chem Soc. 2010;132:12711-
12716.
[13] J.F. Feng, K.C. Van, G.G. Gurkoff, et al., Post-injury administration of NAAG peptidase inhibitor
prodrug, PGI-02776, in experimental TBI. Brain Res. 2011;1395:62-73.
[14] C. Kratochwil, F.L. Giesel, M. Eder, et al., [(1)(7)(7)Lu]Lutetium-labelled PSMA ligand-induced
remission in a patient with metastatic prostate cancer. Eur J Nucl Med Mol Imaging. 2015;42:987-988.
[15] X. Yang, R.C. Mease, M. Pullambhatla, et al., [(18)F]Fluorobenzoyllysinepentanedioic Acid
Carbamates: New Scaffolds for Positron Emission Tomography (PET) Imaging of Prostate-Specific
Membrane Antigen (PSMA). J Med Chem. 2016;59:206-218.
[16] V. Klusak, C. Barinka, A. Plechanovova, et al., Reaction mechanism of glutamate carboxypeptidase II
revealed by mutagenesis, X-ray crystallography, and computational methods. Biochemistry.
2009;48:4126-4138.
[17] P. Mlcochova, A. Plechanovova, C. Barinka, et al., Mapping of the active site of glutamate
carboxypeptidase II by site-directed mutagenesis. FEBS J. 2007;274:4731-4741.
[18] A. Plechanovova, Y. Byun, G. Alquicer, et al., Novel substrate-based inhibitors of human glutamate
carboxypeptidase II with enhanced lipophilicity. J Med Chem. 2011;54:7535-7546.
[19] B.S. Slusher, C. Rojas, J.T. Coyle, Chapter 367 - Glutamate Carboxypeptidase II A2 - Rawlings, Neil D,
in: G. Salvesen (Ed.) Handbook of Proteolytic Enzymes, Academic Press2013, pp. 1620-1627.
[20] C. Barinka, Y. Byun, C.L. Dusich, et al., Interactions between human glutamate carboxypeptidase II
and urea-based inhibitors: structural characterization. J Med Chem. 2008;51:7737-7743.

[21] K. Kopka, M. Benesova, C. Barinka, et al., Glu-Ureido-Based Inhibitors of Prostate-Specific Membrane
Antigen: Lessons Learned During the Development of a Novel Class of Low-Molecular-Weight Theranostic
Radiotracers. J Nucl Med. 2017;58:17S-26S.
[22] O. Gutten, D. Bim, J. Rezac, et al., Macrocycle Conformational Sampling by DFT-D3/COSMO-RS
Methodology. J Chem Inf Model. 2018;58:48-60.
[23] J.R. Mesters, C. Barinka, W. Li, et al., Structure of glutamate carboxypeptidase II, a drug target in
neuronal damage and prostate cancer. EMBO J. 2006;22:1375-1384.
[24] J. Pavlicek, J. Ptacek, C. Barinka, Glutamate carboxypeptidase II: an overview of structural studies and
their importance for structure-based drug design and deciphering the reaction mechanism of the
enzyme. Curr Med Chem. 2012;19:1300-1309.
[25] S.B. Buck, J.K. Huff, R.H. Himes, et al., Total synthesis and anti-tubulin activity of epi-c3 analogues of
cryptophycin-24. J Med Chem. 2004;47:3697-3699.
[26] M. Wiesner, B. Neff, Method for the production of bicyclic aromatic amino acids and intermediate
products thereof, US 7,371,854, 2008.
[27] R.E. Conway, C. Rojas, J. Alt, et al., Prostate-specific membrane antigen (PSMA)-mediated laminin
proteolysis generates a pro-angiogenic peptide. Angiogenesis. 2016;19:487-500.
[28] K. Hlouchova, C. Barinka, J. Konvalinka, et al., Structural insight into the evolutionary and
pharmacologic homology of glutamate carboxypeptidases II and III. FEBS J. 2009;276:4448-4462.
[29] W. Kabsch, Xds. Acta Crystallogr D Biol Crystallogr. 2010;66:125-132.
[30] G.N. Murshudov, P. Skubak, A.A. Lebedev, et al., REFMAC5 for the refinement of macromolecular
crystal structures. Acta Crystallogr D Biol Crystallogr. 2011;67:355-367.
[31] P. Emsley, B. Lohkamp, W.G. Scott, et al., Features and development of Coot. Acta Crystallogr D Biol
Crystallogr. 2010;66:486-501.
[32] A.W. Schuttelkopf, D.M. van Aalten, PRODRG: a tool for high-throughput crystallography of protein-
ligand complexes. Acta Crystallogr D Biol Crystallogr. 2004;60:1355-1363.
[33] V.B. Chen, W.B. Arendall, 3rd, J.J. Headd, et al., MolProbity: all-atom structure validation for
macromolecular crystallography. Acta Crystallogr D Biol Crystallogr. 2010;66:12-21.
[34] M. Navratil, J. Tykvart, J. Schimer, et al., Comparison of human glutamate carboxypeptidases II and III
reveals their divergent substrate specificities. FEBS J. 2016;283:2528-2545.
[35] M. Sparta, F. Neese, Chemical applications carried out by local pair natural orbital based coupled-
cluster methods. Chem Soc Rev. 2014;43:5032-5041.
[36] F. Neese, The ORCA program system. WIREs Comput Mol Sci. 2012;2:73-78.
[37] A. Klamt, Conductor-like Screening Model for Real Solvents: A New Approach to the Quantitative
Calculation of Solvation Phenomena. J. Phys. Chem. 1995;99:2224-2235.
[38] A. Klamt, V. Jonas, T. Bürger, et al., Refinement and Parametrization of COSMO-RS. J. Phys. Chem.
1998;102:5074-5085.
[39] K. Eichkorn, O. Treutler, H. Öhm, et al., Auxiliary basis sets to approximate Coulomb potentials.
Chem. Phys. Lett. 1995;240:283-290.
[40] S. Grimme, J. Antony, S. Ehrlich, et al., A consistent and accurate ab initio parametrization of density
functional dispersion correction (DFT-D) for the 94 elements H-Pu. J Chem Phys. 2010;132:154104.
[41] A. Klamt, G. Schüürmann, COSMO: a new approach to dielectric screening in solvents with explicit
expressions for the screening energy and its gradient. J. Chem. Soc. Perkin Trans. 2. 1993:799-805.
[42] J. Tao, J.P. Perdew, V.N. Staroverov, et al., Climbing the density functional ladder: nonempirical meta-
generalized gradient approximation designed for molecules and solids. Phys Rev Lett. 2003;91:146401.