Virtual screening and computational optimization for the discovery of covalent prolyl oligopeptidase inhibitors with activity in human cells

Article abstract of DOI:10.1021/jm3002839

Our docking program, Fitted, implemented in our computational platform, Forecaster, has been modified to carry out automated virtual screening of covalent inhibitors. With this modified version of the program, virtual screening and further docking-based optimization of a selected hit led to the identification of potential covalent reversible inhibitors of prolyl oligopeptidase activity. After visual inspection, a virtual hit molecule together with four analogues were selected for synthesis and made in one-five chemical steps. Biological evaluations on recombinant POP and FAPα enzymes, cell extracts, and living cells demonstrated high potency and selectivity for POP over FAPα and DPPIV. Three compounds even exhibited high nanomolar inhibitory activities in intact living human cells and acceptable metabolic stability. This small set of molecules also demonstrated that covalent binding and/or geometrical constraints to the ligand/protein complex may lead to an increase in bioactivity.

Full text of DOI:10.1021/jm3002839

Article
pubs.acs.org/jmc
Virtual Screening and Computational Optimization for the Discovery
of Covalent Prolyl Oligopeptidase Inhibitors with Activity in Human
Cells
Stephane De Cesco,† Sebastien Deslandes,† Eric Therrien,† David Levan,† Mickael Cueto,† Ralf Schmidt,‡
́
́
̈
Louis-David Cantin,‡ Anthony Mittermaier,† Lucienne Juillerat-Jeanneret,§ and Nicolas Moitessier*,†
†Department of Chemistry, McGill University, 801 Sherbrooke Street West, Montrea
́
l, Queb
́
ec H3A 0B8, Canada
‡AstraZeneca R&D Montrea
́
l, 7171 Frederick-Banting, St. Laurent, Quebec H4S 1Z9, Canada
́
́
́
§University Institute of Pathology, CHUV-UNIL, Rue du Bugnon 25, CH-1011, Lausanne, Switzerland
S
* Supporting Information
ABSTRACT: Our docking program, Fitted, implemented in
our computational platform, Forecaster, has been modified to
carry out automated virtual screening of covalent inhibitors.
With this modified version of the program, virtual screening and
further docking-based optimization of a selected hit led to the
identification of potential covalent reversible inhibitors of prolyl
oligopeptidase activity. After visual inspection, a virtual hit
molecule together with four analogues were selected for
synthesis and made in one−five chemical steps. Biological evaluations on recombinant POP and FAPα enzymes, cell extracts,
and living cells demonstrated high potency and selectivity for POP over FAPα and DPPIV. Three compounds even exhibited
high nanomolar inhibitory activities in intact living human cells and acceptable metabolic stability. This small set of molecules
also demonstrated that covalent binding and/or geometrical constraints to the ligand/protein complex may lead to an increase in
bioactivity.
INTRODUCTION
■
Prolyl oligopeptidase (POP/PREP) is a highly conserved and
widely distributed postproline endopeptidase. This enzyme,
which can accommodate proline and alanine residues in its
catalytic site, has been found to cleave neuropeptides in vitro.1,2
POP was identified as a drug target soon after its discovery. For
example, microorganism-derived POPs have been associated
with the infectious potential of pathogens.3 More importantly,
it has been reported that POP is involved in many functions of
the central nervous system and that abnormal POP activity,
especially in the brain,4 was associated with a number of
diseases.5−7 POP had first been implicated in memory
impairment. More recently, the potential of POP inhibition
in the treatment of neurodegenerative (e.g., Alzheimer’s
disease, Parkinson’s disease) and psychiatric disorders (e.g.,
bipolar disorder8) has been revealed.5,6,9 Although the function
of POP remains unclear, POP inhibitors were tested in animal
models and the role of POP in β-amyloid formation10 and brain
Figure 1. Selected known POP inhibitors.
damage11 was demonstrated.
As a result, POP inhibitors, either reversible covalent (e.g.,
depositions.10,14,18,19 The latter observation offers promise of
the relevance of this mechanism in Alzheimer’s disease.
However, despite these encouraging results, POP inhibitors
have yet to enter advanced clinical trials (e.g., proof-of-concept
Cbz-Gly-Prolinal, Cbz-Pro-Prolinal,12 JTP-481913,14) or non-
covalent (e.g., SUAM-122115 and S-1709216), have already
been prepared and evaluated in various experimental models
with promising results (Figure 1).9,17 Indeed, researchers have
been able to measure beneficial effects, and recovery in some
cases, on memory and learning abilities in animal models of
brain disorders, along with prevention of amyloid-like
Received: February 20, 2012
© XXXX American Chemical Society
A
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
studies). As discussed in a perspective article, the lack of data
on the penetration of these reported inhibitors into the brain is
a significant limitation for further development of clinically
active POP inhibitors.9 In fact, the first report on brain
penetration was only disclosed last year by Jalkanen and co-
workers, who demonstrated that JTP-4819, a compound which
had been evaluated in phase I clinical trials,20 was poorly
distributed in the brain. A close analogue, KYP-2047 (Figure
1), exhibited more effective inhibition of intracellular POP.21
We became interested in developing novel scaffolds targeting
POP that would have physiochemical properties suitable for
further investigation and development. Our initial contributions
to this field led to the design and synthesis of a first-generation
constrained prolyl oligopeptidase inhibitor (1, Figure 1)
exhibiting high nanomolar inhibitory potency in cell extract
and low micromolar activity in intact living cells.22 This first
series was designed by structural modification of a known
inhibitor. This ligand-based design was guided by docking of
the proposed structures using our docking program, Fitted.
Herein, we report the discovery of novel POP inhibitors using a
very different strategy combining virtual screening and in silico
optimization. We then describe their synthesis, biological
evaluations, and metabolic stability.
Docking Covalent Inhibitors with Fitted. In its current
version, Fitted first identifies the presence of aldehydes,
ketones, boronates, and nitriles as functional groups while the
protein residue capable of covalent binding is specified ahead of
time (in the form of a keyword). The program then selects
either the covalent or noncovalent mode by measuring the
distance between the reactive carbon or boron and the catalytic
residue. If this distance is below a threshold dependent on the
reactive group (e.g., the covalent ideal bond distance as defined
in the GAFF force field34 plus 1 Å), the covalent bond is made,
and additional bonds are either cleaved (O−H in Serine) or
formed (O−H in hemiacetal, His-H if the inhibitor is a
boronate) (Figure 2). All these steps are fully automated and
Virtual Screening and Design. Docking Covalent POP
Inhibitors. Literature reports support the formation of a
covalent bond between the catalytic serine of the active site
and a reactive functional group such as an aldehyde,
hydroxyacetyl, or nitrile, present on the inhibitor.23 The initial
hypothesis, originally based on kinetic studies, was confirmed
by X-ray crystallography of POP in complex with Cbz-Pro-
Prolinal (PDB code: 1QFS24).
To fulfill this structural requirement, a fully automated virtual
screening would therefore require the use of a program that is
able to dock and model the formation of the covalent bond in
the active site of the enzyme. Over the past decade, significant
advances have brought the docking methods to the front line of
drug design and discovery.25 In fact, the number of reported
docking methods and their successful applications to novel
protein−ligand identification is ever-increasing. However,
although some covalent drugs are reaching the market,26 little
effort has been devoted to the docking of such ligands. For
instance, existing docking programs including GOLD,27
AutoDock,28 and FlexX29 were modified to handle covalent
inhibitors. However, all these programs require the manual
definition of the ligand reacting functional groups, precluding
their use in the context of virtual screening. To address this
issue, our docking program, Fitted,30,31 has been modified to
automatically identify reactive functional groups appropriate for
covalent inhibition and form covalent bonds when the catalytic
serine and reactive groups are properly positioned. In 2009, we
first described the use of Fitted for the design of covalent POP
inhibitors.22 To the best of our knowledge, MacDOCK32 and
Fitted are the only two programs considering covalent binding
in a fully automated fashion. ICM was also modified to carry
out a virtual screen of a large database of ketones and aldehydes
on a cysteine protease but did not truly form a covalent bond.
Rather, the catalytic cysteine side chain was removed and the
protocol docked the stereomeric tetrahedral products which
were built prior to the actual docking. To model the covalent
binding, the protocol forced the tetrahedral adduct at the
original position of the cysteine residue.33 Interestingly, this
procedure was very successful in identifying several ubiquitin-
like poxvirus proteinase I7L inhibitors in a virtual screen.
Figure 2. Chemical processes modeled in Fitted. (a) Formation of
hemiacetal from aldehyde. (b) Formation of borate from boronate
ester. (c) Application to the docking of Cbz-Pro-prolinal: yellow,
docked; green, experimentally determined (X-ray crystal structure,
PDB code: 1QFS).
do not require additional user intervention. As soon as a
binding mode is proposed, it is scored. The current version of
the RankScore scoring function35 implemented in Fitted has
not been modified to account for covalent bond formation and
reactivity of the inhibitor, a major limitation of state-of-the-art
docking program. In practice, our current scoring function will
only evaluate the fit between the investigated inhibitors and
POP. We expect its accuracy to be sufficient to discriminate
active from inactive compounds. However, ranking based on
predicted relative potency for active compounds would not be
possible. Virtual screening with known inhibitors will validate
this assumption (see below). Initial validation of Fitted with
POP inhibitors and other covalent inhibitors demonstrated its
accuracy in generating accurate binding modes as illustrated in
B
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 2c. Protein flexibility as implemented in Fitted can also
be used in conjunction with the docking of covalent inhibitors.
Application to Virtual Screening. This version of our
docking program was next implemented in our platform
Forecaster36 and used to screen potential POP inhibitors. The
objective of this study was twofold. First, we aimed to discover
novel POP inhibitors through virtual screening, then we also
validated the covalent docking implementation in Fitted and its
ability to discriminate active from inactive inhibitors. The
designed workflow of actions is shown in Figure 3. First, a
the workflow shown in Figure 3, a crystal structure of POP
bound to an inhibitor was used (PDB code: 2XDW). It was first
converted from the PDB format into a format useable by our
platform (e.g., Forecaster adds hydrogen atoms, atomic partial
charges, atom types, checks for errors, orients water molecules
and polar residue side chains) and then prepared for docking.
The compounds were next docked into POP by Fitted in the
rigid protein mode. Although this may look like a complex
procedure, all these steps are simply selecting potential
inhibitors on one side and preparing the protein on the other
side and then using these structures for docking, all these steps
being fully automated in the platform.36 The docked
compounds were finally rank-ordered by scores, revealing that
the nine known inhibitors were ranked on the very top of the
list (within the top 1.5%, see Supporting Information),
indicating that Fitted was indeed able to discriminate active
covalent POP inhibitors from potentially inactive molecules.
The rank-ordered screened library was next analyzed in the
hope to extract information that could be transposed into an
active molecule. Analysis of the high scoring compounds
allowed the identification of compound 2 as illustrated in the
top panel of Figure 4. Visual inspection of the docked pose of 2
Figure 3. Workflow of actions used to identify potential covalent POP
inhibitors.
library of drug-like molecules was collected from the ZINC
database37,38 upon a substructure search for compounds
containing either an aldehyde or a nitrile group. Descriptors
such as the presence of a set of functional groups or molecular
weight were next added to all the molecules and were used to
filter out undesired molecules. These descriptors available in
Forecaster are described in detail in a recent report and will not
be discussed further herein.36 For this purpose, our program,
Reduce, was used to “Filter” (i.e., discard) any molecules
featuring undesired functional groups (e.g., more than one
potential covalent group or isocyanate) or violating the
Lipinski’s rules.36 This process yielded a total of 2798
molecules. To evaluate the ability to identify POP inhibitors,
nine known inhibitors (given as Supporting Information) were
included in this library by merging the files from the filtered
library with a separate file containing these nine known POP
inhibitors. This extended library was then prepared for docking
and screened using the Fitted program. On the right branch of
Figure 4. Top panel: potential hit discovered by virtual screening (2),
designed potential inhibitors 3 and 4. Bottom panel: compound 4
docked (yellow sticks) superposed with cocrystallized synthetic
unnatural dipeptide (green sticks) and POP. (PDB code: 2XDW).
guided further structural modifications in order to make
improved interactions with the enzyme. Optimization of 2
was conducted with respect to the pharmacophore we
previously proposed which includes the need for two hydrogen
bonds with Trp595 and Arg643 for optimal binding.7 Docking
followed by visual inspection of the poses validated the
structural modifications. Key modifications were (1) the
inversion of the amide bond, (2) rigidification of the ethyl
group of 2 by introduction of a pyrrolidine ring, (3)
rigidification of the scaffold by forming a five-membered ring
on the left side, and (4) introduction of a nitrile in place of the
aldehyde of compound 2 for covalent bond formation.
C
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Compound 3 is the result of modifications 1, 2, and 4, while
compound 4 is the result of all of the four modifications. The
bottom panel of Figure 4 shows the docking pose obtained for
compound 4 superposed on a cocrystallized ligand with POP
(PDB code: 2XDW).
Scheme 2
Synthesis. To evaluate the impact of the rigidification (3
into 4) on the biological activity and confirm the presence of a
covalent bond, 3, 4, and three analogues were selected for
synthesis. The synthesis of the acyclic analogue 3 is depicted in
Scheme 1. Phthalic anhydride 5 was first reacted with
a
Scheme 1
a
(a) Benzylamine, toluene, reflux; (b) NaBH3CN, TFA, MeOH, 45%
(over 2 steps); (c) maleic anhydride, toluene, reflux, 86%; (d) HCl
conc reflux, 78%; (e) Piv-Cl, Et3N, pyrrolidine, 63%; (f) BOP, Et3N,
DMF, 2-(S)-cyano-pyrrolidine, TFA, 72%.
intact living cells (Tables 1−2), expressing POP and DPPIV
activities. A fluorigenic POP specific substrate (Cbz-Gly-Pro-
AMC) was used to measure the POP activity while DPPIV
activity was measured using Gly-Pro-AMC.22 DPPIV is a prolyl-
aminodipeptidase which cleaves dipeptides with free N-
terminus and a proline residue at position 2 on the N-terminus
side. This activity, which is targeted for the treatment of
diabetes type II, should not be modulated by POP inhibitors.9
As shown in Table 1, four of the five tested compounds
a
(a) Benzylamine, acetone, 60%; (b) BOP, Et3N, H-Pro-NH2, MeCN,
70%; (c) TFAA, Et3N, THF, 77%; (d) pyrrolidine, 85%; or piperidine,
55%.
benzylamine to yield the corresponding benzoic acid 6 in
reasonable yield. The carboxylic acid 6 was next coupled with L-
prolinamide in presence of benzotriazole-1-yl-oxy-tris-(dime-
thylamino)-phosphonium hexafluorophosphate (BOP) and
further converted to the nitrile derivative 3 upon dehydration
of the terminal amide with trifluoroacetic anhydride. Although
coupling 2-(S)-cyano-pyrrolidine with 6 would have saved a
chemical step, it was found to be low yielding. In parallel, the
potential noncovalent inhibitor 8, an analogue of 3, was
obtained in a single step by condensation of commercially
available N-benzylphthalimide and pyrrolidine. This compound
stems from two of the above-mentioned modifications
(inversion of the amide bond of compound 2 and rigidification
of its ethyl group by introduction of a pyrrolidine ring). The
six-membered ring analogue 9 was also synthesized in a similar
manner using piperidine as a nucleophile.
Table 1. Inhibition of POP and DPPIV Activity in Cell
a
Extracts
b
c
compd
LN18, LN229, LNZ308
HCEC
3
+++
+++
++
+++
+++
++
8
9
4
+++
+++
+++
+++
14
a
++, >50% inhibition at 100 μM; +++, >90% inhibition at 20 μM.
LN18, LN229, LNZ308: human glioblastoma cells. HCEC, human
b
c
The synthesis of the bicyclic series started with a reductive
amination between furfural 10 and benzylamine producing the
amine 11 (Scheme 2). This was followed by a tandem
condensation with maleic anhydride/Diels−Alder cycloaddition
to afford the carboxylic acid 12. This intermediate underwent
dehydration/rearomatization reaction to yield 13 upon acidic
treatment. The synthesis was completed by conversion of the
carboxylic acid 13 into the potential POP inhibitor 14 by
coupling with pyrrolidine. Following a similar approach, the
designed bicyclic inhibitor 4 was obtained through the coupling
of 13 with 2-(S)-cyano-pyrrolidine.
brain-derived endothelial cells.
exhibited significant POP inhibitory potency at a concentration
of 20 μM in cell extracts. In addition, none inhibited the DPPIV
activity at concentration as high as 100 μM. In addition, the
measured potency is consistent throughout the cell lines. We
have previously found that substituting the five-membered ring
with a six-membered ring was detrimental to the inhibitor’s
activity (Figure 5).22 This contrasted with the finding from
Racys and co-workers who reported a very active inhibitor built
around a piperidine ring.39 In the present work, the substitution
of the pyrrolidine ring of 8 by a piperidine ring (9) led to a
decrease of inhibitory activity.
RESULTS AND DISCUSSION
■
Inhibitory Potency. The inhibitory potency of these five
compounds was assessed first in cell extracts of human
glioblastoma cells and brain-derived endothelial cells, then in
With these promising results in hand, these five active
compounds were further assessed for their inhibitory activity in
intact living cells (Tables 2 and 3 and Figure 6). This second
D
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
conditions. However, we have previously identified Cbz-
protected pseudopeptides (e.g., 17, Figure 5) that exhibit
better potency, showing that there is still room for improve-
ment.
Comparing IC50 values between cell extracts and intact cells
(Table 3) demonstrated an increased efficacy for the extracted
enzyme for compounds 3 and 4, but not 14, compared to intact
cells. This information suggests that cellular uptake may be a
limiting factor for compounds 3 and 4, but not 14.
Selectivity. In addition to POP, the only enzyme with
comparable prolyl-specific endoproteolytic enzymatic proper-
ties is FAPα.40 The cells used, either human glioblastoma cells
or human brain-derived endothelial cells, do not express FAPα
(data not shown, Juillerat et al, manuscript in preparation), thus
for these cells, either intact cells or cell extracts, selectivity is not
an issue. To evaluate the selectivity of the inhibitors, inhibition
of purified recombinant human POP/PREP and FAPα has
been assessed and KM determinations have been performed
(Tables 4 and 5).
Figure 5. Ring Size and Inhibitory Potency against POP.22
a
Table 2. Inhibition of POP Activity in Living Intact Cells.
b
c
compd
LN18, LN229, LNZ308
HCEC
3
+++
++
+++
++
8
9
−
−
+++
4
+++
The experiments with purified recombinant human POP
showed that the three selected inhibitors, compounds 3, 4, and
14, are significantly better inhibitors of rhPOP than of rhFAPα
(Table 4). This also demonstrated that KM values for the
rhPOP and for the enzyme activity in cell extracts (Table 5) are
very similar. rhFAPα also displayed a much less favorable KM
for the selected substrate than rhPOP or the cell extracts. Thus,
compounds 3, 4, and 14 proved to be very poor inhibitors for
FAPα activity compared to POP. Therefore, we can conclude
that the target enzyme in human brain-derived cells is POP/
PREP and that the inhibitors 3, 4, and 14 are selective for POP
versus FAPα.
Structure−Activity Relationship. The 2 orders of
magnitude difference in activity between compounds 3 (IC50
= 65 nM, HCEC) and 8 (IC50 = 20−100 μM) in cell extracts
might indicate the formation of a covalent bond as intended.
Overall, as shown in Figure 7, the geometrical constraint
brought by the lactam ring (8 into 14) or the putative covalent
bond made with the nitrile (8 into 3) were highly beneficial to
the inhibitory activity against POP activity. Combining the two
14
+++
+++
a
−, No inhibition at 100 μM; ++, >50% inhibition at 100 μM; +++,
>90% inhibition at 20 μM. LN18, LN229, LNZ308: human
glioblastoma cells. HCEC, human brain-derived endothelial cells.
b
c
set of experiments assessed the intracellular inhibition of POP
activity by these small molecules. The inhibition of DPPIV-like
activity by the compounds was not determined in intact cells
because no inhibition had been found in cell extracts. These
experiments demonstrated that not only are these compounds
inhibiting the POP activity, but they are also able to penetrate
the cells. To further evaluate their inhibitory potency, dose
response efficacy of the three more potent inhibitors,
compounds 3, 4, and 14, were compared (Figure 6) and IC50
values determined (Table 3) in intact cells and in cell extracts
We were also very pleased to see that three of these compounds
remain in the nanomolar range while going from cell extracts to
living cells while our previous series illustrated by compound 1
were active in the low micromolar range under the same
Figure 6. Dose−response inhibition of POP activity in cell extracts (top) and intact human endothelial (HCEC) and glioblastoma (LN18, LN229,
LNZ308) cells (bottom) by compounds 3, 4, and 14.
E
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Table 3. Inhibition of POP Activity (IC50, [nM]) in Cell Extracts and Living Intact Cells
cell extracts
intact living cells
a
a
a
b
a
a
a
b
compd
LN18
LN229
LNZ308
HCEC
LN18
LN229
LNZ308
HCEC
122
nd
68
500
68
200
73
700
65
nd
2000
500
450
610
2500
290
200
640
2000
290
500
680
3
250
300
580
4
81
45
45
45
14
620
500
350
620
a
b
LN18, LN229, LNZ308: human glioblastoma cells. HCEC, human brain-derived endothelial cells.
Table 4. Inhibition of Recombinant Human FAP-α and
POP/PREP by Compounds 3, 4, and 14
rhFAP-α
% @ 100 μM
rhPOP/PREP
compd
IC50 (μM)
% @ 2 μM
IC50 (nM)
3
34
28
11
>100
>100
>100
90
93
63
<50
<50
4
14
= 150
structural changes (14 into 4 or 3 into 4) revealed that the
additional lactam ring was not necessary for optimal binding as
4 is only slightly, but not significantly, more potent than 3.
When noncovalent binding occurs, introduction of a geo-
metrical constraint in the form of a ring (8 into 14) reduces the
entropy penalty of binding but often requires slight adjustments
of the binding mode and/or of the protein environment to
keep the same favorable interactions. However, when covalent
binding occurs, a second constraint in the form of a covalent
bond is added, precluding any adjustment of the binding mode
through translation or rotation of the inhibitor. It may also
affect protein flexibility which could result in an entropic
penalty.41 As a result, rigidification of 3 into 4 resulted in no
significant gain in inhibitory activity. One can question whether
the postulated covalent bond between the nitrile group and the
protein catalytic serine actually occurs. Although additional
work is needed to confirm this hypothesis, the large binding
affinity increase observed when the nitrile is added onto 8 (IC50
= ca. 10000 nM) to form 3 (IC50 ca. 100 nM) strongly suggests
the formation of a covalent bond as intended. Recently, crystal
structures of nitrile-containing inhibitors including 18 have
shown that indeed a covalent bond is formed with the catalytic
serine of POP (Figure 8).42 As with our compound 4, this
compound 18 features a cyanopyrrolidine ring.
Figure 7. Structure−activity relationships in cell extracts.
Figure 8. Cocrystallized covalent POP inhibitor.
Table 6. In Vitro Metabolic Stability in Human and Rat
Liver Microsomes
HLM
RLM
in vitro Clint
(μL/min/mg)
in vitro
t1/2 (min)
in vitro Clint
(μL/min/mg))
in vitro
compd
t1/2 (min)
a
a
1
10
44
13
25
89
14
134
31
nt
nt
Metabolic Stability. Prior to testing these compounds in
vivo, we thought to assess their metabolic stability in human
liver microsomes (HLM) (Table 6 and Figure 9). As the
following step would be to test these inhibitors in animals,
metabolism was also evaluated on rat liver microsomes (RLM).
The derivatives 8 and 9 were first assessed and revealed major
oxidation of the pyrrolidine ring on the right-hand side of these
inhibitors and minor N-debenzylation or aromatic oxidation on
the left-hand side of the compounds.
3
37
39
38
35
17
4
4
111
55
8
83
9
16
311
62
14
98
23
a
Not tested.
A close look at the metabolic stability revealed that compared
to the first generation constrained prolyl oligopeptidase
inhibitor 1, constrained compounds 4 and 14 maintained
similar metabolic stability whereas the corresponding less
restricted analogues 3, 8, and 9 exhibited increased metabolic
liabilities. We believe that the geometrical constrains of
compounds 1, 4, and 14 affects the recognition by metabolic
enzymes (i.e., cytochrome P450 enzymes), while the more
flexible compounds 3, 8, and 9 may adopt more favorable
conformations for recognition by cytochrome P450 enzymes.
Table 5. KM [mM] Values of Recombinant Human POP/PREP and FAPα for the Substrate
compd
LN18
LN229
0.0067
LNZ308
0.0081
HCEC
0.0070
rhPOP/PREP
0.0085
rhFAP-α
Z-Gly-Pro-AMC
0.0085
0.103
F
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
EXPERIMENTAL SECTION
■
Virtual Screening. The database of small molecules was
downloaded from the ZINC Web site (http://zinc.docking.org) and
further processed using a series of programs implemented in our
platform Forecaster.36 First, the library was subjected to filtering using
Reduce (formal charge: [−1,0], molecular weight [200,550], aldehyde
[0,1], should not include the following functional group: ester, primary
amine, acyl chloride, ketone, hydroxamate, isocyanide, azide). The
resulting compounds were next prepared for docking and docked to a
protein structure prepared from the PDB file using Prepare (PDB
code: 2XDW) and prepared for docking using Process. The docking
was carried out with Ser554 identified as a catalytic residue susceptible
to forming a covalent bond with reactive groups.
Synthesis. All commercially available reagents were used without
further purification unless otherwise stated. FTIR spectra were
recorded using a Perkin-Elmer Spectrum One FT-IR. 1H and 13C
NMR spectra were recorded on Varian Mercury 400 MHz, 300 MHz,
or Unity 500 spectrometers. Chemical shifts are reported in ppm using
the residual of deuterated solvents as an internal standard. Thin layer
chromatography visualization was performed by UV or by develop-
ment using KMnO4, H2SO4/MeOH, or cerium/molybdate solutions.
Chromatography was performed on silica gel 60 (230−40 mesh). Low
resolution mass spectrometry was performed by ESI using a
Thermoquest Finnigan LCQ Duo. High resolution mass spectrometry
was performed by EI peak matching (70 eV) on a Kratos MS25 RFA
double focusing mass spectrometer or by ESI on a Ion Spec 7.0 T
FTMS at McGill University. Prior to biological testing, reversed phase
HPLC was used to verify the purity of compounds on an Agilent 1100
series instrument equipped with VWD-detector, C18 reverse column
(Agilent, Zorbax Eclipse XDB-C18 150 mm × 4.6 mm, 5 μm), UV
detection at 254 nm. All tested compounds were at least 95% pure.
(S)-N-Benzyl-2-(2-cyanopyrrolidine-1-carbonyl)benzamide
(3). To a stirred solution of the carboxylic acid compound 643 (500
mg, 1.96 mmol), in MeCN (50 mL) under argon atmosphere, were
added sequentially BOP (1.04 g, 2.3 mmol), L-prolinamide (0.224 g,
1.96 mmol), and Et3N (601 μL, 4.4 mmol). The reaction mixture was
allowed to react for 15 h at room temperature. The reaction mixture
was quenched with a 10% HCl solution, extracted in EtOAc, and
washed successively with saturated NaHCO3, and brine. The organic
phase was dried over anhydrous Na2SO4, filtered, and concentrated in
vacuo to afford (S)-1-(2-(benzylcarbamoyl)benzoyl)pyrrolidine-2-
carboxamide. Without further purification. the amide was dissolved
in DCM (60 mL) and Et3N (863 μL, 6.2 mmol) was added. The
resulting solution was cooled down to 0 °C, and then TFAA (564 μL,
4 mmol) was added dropwise. After stirring for another 2 h, a
saturated solution of NaHCO3 was added and the organic phase was
dried over Na2SO4, and concentrated in vacuo to afford, after flash
chromatography (EtOAc/hexanes, 1:1), compound 3 (250 mg, 55%,
white solid); Rf = 0.10 (EtOAc/hexanes, 1:1). IR (film) υ max 3311,
Figure 9. Metabolic stability in human and rat liver microsomes.
To pinpoint the site of metabolism, metabolite identification
experiments were performed for all analogues (Figure 8). The
analysis revealed that, with exception of compound 3, multiple
oxidations of the pyrrolidine ring are mainly responsible for the
metabolic clearance, even more pronounced for analogue 9
containing a piperidine moiety. The main metabolite pathway
for analogue 3 led to a debenzylated metabolite.
CONCLUSION
■
The discovery of novel inhibitors of POP activity selective over
DPPIV activity was made possible through the combination of
virtual screening (hit identification) using Fitted, a docking
program considering covalent bond formation, virtual lead
optimization, synthesis, and biological evaluation. The medic-
inal chemist/docking-guided lead optimization ensures that
both predicted binding affinity and synthetic feasibility were
optimal. Thus a selection of five molecules was synthesized in
one (8 and 9), three (3), or five (4 and 13) chemical steps.
This study represents one of the rare virtual screening studies
applied to the discovery of covalent inhibitors. Interestingly,
with focused synthetic efforts, we were able to develop three
inhibitors that are 10 times more active intracellularly than was
our previous hit molecule 1, which required significantly more
synthesis efforts.22 Further investigation using recombinant
POP enzyme confirmed that POP is the target of these
inhibitors and that they are highly selective for POP over FAPα.
While the introduction of one geometrical constraint (either a
lactam ring or a reactive group) was found to increase the
inhibitory potency, the combination of the two modifications
did not produce the expected synergistic increase. In addition,
metabolism studies revealed an acceptable microsomal stability
for the bicyclic analogues at this stage, a promising indication
for further in vivo evaluation.
2925, 1633, 1404, 1303, 698 cm−1; [α]D20 = −5.4 (c 1.25, CHCl3). H
1
NMR (400 MHz, CDCl3, 3:1 mixture of rotamers) δ 1.91−2.30 (m,
4H), 3.19−3.33 (m, 1.3H), 3.55−3.69 (m, 0.7H), 4.42 (m, 1H), 4.55
(d, 0.75H, J = 6.6 Hz), 4.69 (m, 1H), 4.73 (d, 0.25H, J = 6.6 Hz), 6.69
(bs, 0.25H), 6.85 (bs, 0.75H), 7.27−7.74 (m, 9H). 13C NMR (75
MHz, MeOD) δ 172.59, 169.57, 140.07, 137.72, 134.53, 132.53,
130.98, 129.73 (2C), 128.85 (2C), 128.77, 128.40, 128.23, 119.75,
49.96, 47.65, 44.67, 31.65, 26.07. HRMS (ESI+) calcd for
[C20H19N3O2 + H]+, 334.15500; found, 334.15454;
N-Benzyl-2-(pyrrolidine-1-carbonyl)benzamide (8). To a
solution of benzylphthalimide (503 mg, 2.13 mmol) in THF (15
mL) was added pyrrolidine (0.20 mL, 2.41 mmol). The reaction vessel
was left to react for 24 h under argon atmosphere. The mixture was
evaporated to dryness in vacuo, and the final product was obtained
1
after recrystallization from methanol (551 mg, 85%). The H NMR
(400 MHz, DMSO-d6) was identical to the previously reported data.44
1H NMR (400 MHz, DMSO-d6): δ 1.70−1.83 (m, 4 H), 3.06 (t, 2 H, J
= 6.7 Hz), 3.36 (t, 2 H, J = 6.7 Hz), 4.41 (d, 2 H, J = 6.0 Hz), 7.23−
7.33 (m, 6 H), 7.44−7.53 (m, 2 H), 7.65 (d, 1 H, J = 7.4 Hz), 8.89 (t,
G
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
1 H, J = 5.8 Hz). HRMS (ESI+) calcd for [C19H20N2O2 + Na]+,
331.14170; found, 331.14244.
mixture was stirred at 0 °C for 1 h, followed by the addition of a
solution of pyrrolidine (0.07 mL, 0.889 mmol) and Et3N (0.12 mL,
0.889 mmol) in DCM (1.0 mL). After stirring for a further 2 h at 0 °C,
the mixture was allowed to warm to room temperature and stirring was
continued overnight. The reaction mixture was quenched with a 10%
HCl solution (15 mL), extracted with DCM, dried over Na2SO4, and
evaporated under reduced pressure to afford, after flash chromatog-
raphy (EtOAc/hexanes 4:1 to 1:0, Rf = 0.10), compound 14 (163 mg,
63%, white solid). IR (film) υ max 2968, 2925, 2875,1685, 1629, 1437
N-Benzyl-2-(piperidine-1-carbonyl)benzamide (9). To a sol-
ution of benzylphthalimide (0.462 g, 1.95 mmol) in THF (20 mL) was
added piperidine (0.20 mL, 2.02 mmol). The reaction vessel was set
aside for 92 h under argon atmosphere. The mixture was evaporated to
dryness in vacuo, and the final product (0.346 g, 55%) was obtained by
flash column chromatography (1:1 hexanes/ethyl acetate, Rf = 0.13).
1H NMR (400 MHz, DMSO-d6) δ 1.20−1.43 (m, 6H), 2.95 (bs, 2H),
3.31−3.49 (m, 2H), 4.33 (d, 2H, J = 6.0 Hz), 7.14−7.18 (m, 2 H),
7.25 (s, 2H), 7.26 (s, 2H), 7.36−7.45 (m, 2H), 7.58 (dd, 1H, J = 7.6,
1.3 Hz), 8.81 (t, 1H, J = 6.0 Hz). 13C NMR (75 MHz, CDCl3) δ
169.83, 167.31, 138.28, 135.04, 132.82, 130.78, 129.38, 128.97, 128.60
(2C), 127.91 (2C), 127.35, 125.85, 48.45, 43.94, 42.62, 26.10, 25.37,
24.31. HRMS (ESI+) calcd for [C20H22N2O2 + Na]+, 345.15735;
found, 345.15823. IR (film) υ max 3305, 2938, 2856, 2350,1600, 1596,
1490 cm−1.
1
cm−1. H NMR (300 MHz, CDCl3) δ 1.91−1.96 (m, 4H), 3.06 (s,
1H), 3.24 (s, 1H), 3.65 (s, 1H), 3.80 (s, 1H), 4.21 (s, 2H), 4.61 (s,
1H), 4.84 (s, 1H), 7.23−7.36 (m, 7H), 7.50 (t, 1H, J = 6.0 Hz). 13C
NMR (75 MHz, CDCl3) δ 167.05, 166.89, 141.79, 136.95, 136.00,
135.46, 131.82, 128.92 (2C) 128.40 (2C), 127.86, 126.20, 123.39,
49.34, 48.27, 46.46, 45.71, 26.02, 24.63. HRMS (ESI+) calcd for
[C20H20N2O2 + H]+, 321.15975; found. 321.15956.
(S)-1-(2-Benzyl-3-oxoisoindoline-4-carbonyl)pyrrolidine-2-
carbonitrile (4). To a stirred solution of compound 5 (50 mg, 0.186
mmol) in DMF (1.0 mL) were added sequentially BOP (99 mg, 0.223
mmol), (S)-pyrrolidine-2-carbonitrile TFA salt (39 mg, 0.186 mmol),
and Et3N (0.06 mL, 0.408 mmol). The mixture was stirred for 15 h at
room temperature and quenched with a 10% HCl solution (10 mL),
extracted with EtOAc, dried over Na2SO4, and concentrated in vacuo
to afford, after flash chromatography (EtOAc/hexanes, 7:3, Rf = 0.22),
compound 4 (59 mg, 92%, white solid). 1H NMR (500 MHz, acetone-
d6) δ 2.01−2.44 (m, 4H), 3.29 (bs, 1.5H), 3.73 (m, 0.5H), 4.42 (s,
2H), 4.78 (m, 2H), 4.96 (m, 1H), 7.28−7.45 (m, 6H), 7.62−7.70 (m,
2H). 13C NMR (75 MHz, CDCl3) δ 167.72, 166.80, 141.86, 136.56,
133.03, 132.00, 128.96 (2C), 128.30 (2C), 127.92, 126.32, 124.47,
124.27, 118.77, 49.48, 48.96, 48.09, 46.45, 32.17 (0.3C), 30.41 (0.7C),
25.12 (0.7C), 23.50 (0.3C). IR (film) υ max 3014, 1681, 1636, 1413
cm−1. HRMS (ESI+) calcd for [C21H19N3O2 + H]+, 346.15500; found,
346.15468.
Inhibitory Potency. The human glioblastoma-derived cell lines
LN18, LN229, and LNZ308 were a kind gift of A. C. Diserens,
Neurosurgery Department, Lausanne, Switzerland, the immortalized
human brain-derived HCEC cells were kindly provided by D.
Stanimirovic, Ottawa, Canada. Cells were grown in DMEM culture
medium containing 4.5 g/L glucose, 10% fetal calf serum (FCS)
(HCEC cells), or 5% FCS (LN18, LN229, and LNZ308 cells), and
antibiotics (all from Gibco, Basel, Switzerland). One to two days
before evaluation, cells were seeded in 48-well plates (Costar, Corning,
NY) in complete medium in order to reach confluence on the day of
experiment. On the day of experiment, the culture medium was
removed, and either 200 μL of phosphate-buffered saline (PBS, pH
7.2−7.4) were added in half of the wells or 200 μL of PBS containing
0.1% Triton X-100 (Fluka, Buchs, Switzerland) were added in the
other half of the wells for the evaluation of the inhibition of enzyme
activities in intact cells or cell extracts, respectively. Experiments were
performed in duplicate wells. The synthetic molecules were dissolved
at 10 mg/mL in methanol and then diluted 1:10 in H2O, and 1 or 5 μL
of the water solution were added to duplicate PBS and PBS-Triton
wells, followed after 5−10 min at room temperature by either 1 μL of
Gly-Pro-AMC (DPPIV activity) or Z-Gly-Pro-AMC (POP activity)
substrates (1 mg/mL DMSO, both from Bachem, Bubendorf,
Switzerland), final concentration 10 μM. Increase in fluorescence at
λex/λem = 360/460 nm was recorded for 30 min at 37 °C in a
thermostatted multiwell fluorescence reader (Cytofluor, PerSeptive
BioSystems, Switzerland). For the determination of IC50, cells in PBS
or in PBS-Triton X-100 were exposed to decreasing concentrations of
the inhibitors, and then determination of residual activity was
measured and plotted against inhibitor concentration. IC50 values
were determined graphically.
N-Benzyl-1-(furan-2-yl)methanamine (11). A mixture of furan-
2-carboxaldehyde (1.3 mL, 15.6 mmol) and benzylamine (1.79 mL,
16.39 mmol) in benzene (31 mL) was heated at reflux with a Dean−
Stark for 4 h. The solvent was evaporated to dryness, and the crude
imine was dissolved in MeOH (15 mL) and then cooled to 0 °C.
Sodium cyanoborohydride (1.47 g, 23.4 mmol) and trifluoroacetic acid
(1.26 mL, 16.4 mmol) were added successively to the mixture. After
30 min, the mixture was allowed to warm to room temperature, and
stirring was continued for a further 2 h. The mixture was evaporated
under reduced pressure and the residue dissolved in EtOAc (20 mL)
and washed with 1 M NaOH (10 mL) and saturated NaCl solution
(10 mL). The organic phase was dried over Na2SO4 and concentrated
in vacuo. The residue was purified by flash chromatography (EtOAc/
hexanes, 1:4, Rf = 0.15) to afford 11 (1.32 g, 45%, light-yellow oil). IR
(film) υ max 3063, 3028, 2831, 1602, 1496, 1454, 1360, 1335, 1146,
1
1008, 730, 697 cm−1. H NMR (300 MHz, CDCl3) δ 1.73 (s, 1H),
3.80 (s, 4H), 6.20 (s, 1H), 6.33 (s, 1H), 7.23−7.38 (m, 6H). 13C NMR
(75 MHz, CDCl3) δ 153.82, 141.83, 139.90, 128.42, 128.27, 127.03,
110.11, 107.05, 52.80, 45.37. HRMS (ESI+) calcd for [C12H13NO +
H]+, 188.10699; found, 188.10692.
2-Benzyl-1-oxo-1,2,3,6,7,7a-hexahydro-3a,6-epoxyisoin-
dole-7-carboxylic Acid (12). To a stirred solution of 11 (1.30 g,
6.94 mmol) in toluene (100 mL), maleic anhydride (0.68 g, 6.94
mmol) was added, and the resulting mixture was heated to reflux for 5
h. The mixture was cooled to room temperature and further stirred
overnight. The precipitate was collected by filtration, washed with
Et2O (25 mL), and dried under vacuum. The product 12 (1.71 g,
86%) was isolated as a white solid. IR (film) υ max 3030, 1734, 1658,
1
1477, 1359, 1206 cm−1; mp (CHCl3) 170−174 °C. H NMR (300
MHz, CDCl3) δ 2.84 (d, 1H, J = 9.0 Hz), 2.98 (d, 1H, J = 9.0 Hz),
3.65 (d, 1H, J = 12.0 Hz), 3.83 (d, 1H, J = 12.0 Hz), 4.42 (d, 1H, J =
15.0 Hz), 4.63 (d, 1H, J = 15.0 Hz), 5.24 (s, 1H), 6.41 (s, 2H), 7.28−
7.32 (m, 5H), 10.77 (bs, 1H). 13C NMR (75 MHz, CDCl3) δ 173.58,
172.50, 137.08, 135.28, 134.99, 128.90, 127.93, 127.80, 88.68, 82.17,
51.01, 48.51, 46.97, 45.50. HRMS (ESI+) calcd for [C16H15NO4 +
H]+, 286.10738; found, 286.10696.
2-Benzyl-3-oxoisoindoline-4-carboxylic Acid (13). A solution
of compound 12 (1.70 g, 5.96 mmol) was dissolved in concd HCl (30
mL) and refluxed for 3 h. The mixture was then cooled to room
temperature, stirred overnight, and concentrated in vacuo. MeOH (20
mL) was added, and the solid was collected by filtration affording
compound 13 (1.25 g, 78%, tan solid). IR (film) υ max 3046, 1869,
1
1708, 1598, 1579, 1496 cm−1; mp (CHCl3) 178−182 °C. H NMR
(300 MHz, CDCl3) δ 4.44 (s, 2H), 4.86 (s, 2H), 7.32−7.40 (m, 5H),
7.62 (d, 1H, J = 6.0 Hz), 7.68 (t, 1H, J = 6.0 Hz), 8.40 (d, 1H, J = 9.0
Hz). 13C NMR (75 MHz, CDCl3) δ 169.53, 165.41, 141.89, 135.07,
133.31, 132.35, 129.35, 129.22, 129.08, 128.37, 128.32, 126.99, 50.41,
47.24. HRMS (ESI+) calcd for [C16H14NO3 + Na]+, 268.09682;
found, 268.09644.
Recombinant human prolyl oligopeptidase/PREP (rhPOP/PREP,
0.5 mg/mL, catalogue no. 4308-SE) and fibroblast activation protein α
(rhFAPα, 0.5 mg/mL, catalogue no. 3715-SE) were obtained from
R&D Systems (Abingdon, UK). The stock solutions were diluted in
PBS containing 1 mg/mL bovine serum albumin (BSA, Sigma), a
protein concentration comparable to the protein concentration of cell
extracts. For comparison of inhibitory potencies, IC50 and KM in cell
2-Benzyl-7-(pyrrolidine-1-carbonyl)isoindolin-1-one (14). To
a 0 °C solution of compound 5 (216 mg, 0.808 mmol) and Et3N (0.12
mL, 0.889 mmol) in DCM (2.0 mL) was added dropwise a solution of
pivaloyl chloride (0.10 mL, 0.808 mmol) in DCM (1.0 mL). The
H
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
extracts with rhPOP/PREP and rhFAPα, cells were grown in 6 cm
Petri dishes (Falcon, BD, Erembodegem, Belgium) and extracted in 3
mL of PBS-Triton (resulting in solutions of ∼0.5 mg protein/mL). For
the evaluation of the inhibition of the recombinant human enzymes,
the experimental design was similar to the experimental design of cell
extracts. For KM determinations, increasing substrate concentrations
were added to the cell extracts in PBS-Triton or to the recombinant
enzymes in PBS-BSA, and the KM values were determined graphically
using a double-reciprocal Michaelis−Menten plot.
Metabolic Stability Assay (Clint). The test substances at 1 μM
concentration were mixed with 0.1 M KH2PO4 buffer pH 7.4 and liver
microsomes (BD-Gentest) at 0.5618 mg/mL in a 96-deep well plate
and incubated at 37 °C. NADPH was added to start the reaction. After
0, 10, 20, and 30 min, an aliquot of the incubation mixture was
transferred to a 384-well plate containing an equal amount of
acetonitrile to stop the reaction. The 384-well plate was centrifuged for
30 min at 9000g and under cooling (4 °C). The supernatants were
then analyzed on an Agilent 1100 LC-MSD system. The robot used
for this assay is a TECAN Genesis (8-tip pipettes). All steps in the
assay, i.e., pipetting, incubation, etc., were performed by the robot
except for the centrifugation step.
REFERENCES
■
(1) Szeltner, Z.; Rea, D.; Juhasz, T.; Renner, V.; Fulop, V.; Polgar, L.
Concerted Structural Changes in the Peptidase and the Propeller
Domains of Prolyl Oligopeptidase are Required for Substrate Binding.
J. Mol. Biol. 2004, 340, 627−637.
(2) Polgar, L. Unusual secondary specificity of prolyl oligopeptidase
and the different reactivities of its two forms toward charged
substrates. Biochemistry 1992, 31, 7729−7735.
(3) Bastos, I. M. D.; Grellier, P.; Martins, N. F.; Cadavid-Restrepo,
́
G.; De Souza-Ault, M. R.; Augustyns, K.; Teixeira, A. R. L.; Schrevel,
J.; Maigret, B.; Da Silveira, J. F.; Santana, J. M. Molecular, functional
and structural properties of the prolyl oligopeptidase of Trypanosoma
cruzi (POP Tc80), which is required for parasite entry into
mammalian cells. Biochem. J. 2005, 388, 29−38.
(4) Irazusta, J.; Larrinaga, G.; Gonz·lez-Maeso, J.; Gil, J.; Meana, J. J.;
Casis, L. Distribution of prolyl endopeptidase activities in rat and
human brain. Neurochem. Int. 2002, 40, 337−345.
(5) Szeltner, Z.; Polgar, L. Structure, function and biological
relevance of prolyl oligopeptidase. Curr. Protein Pept. Sci. 2008, 9,
96−107.
(6) Mannisto, P. T.; Vanalainen, J.; Jalkanen, A.; Garcia-Horsman, J.
̈
̈
̈
Metabolite Identification. Sample preparation: Compounds were
incubated at 50 μM with 1 mg/mL of human liver microsomes (BD-
Gentest) in the presence of 1 mM NADPH at 37 °C for 30 min. After
precipitation with acetonitrile and centrifugation, the supernatant
diluted with water (1:1) was analyzed by LC/MS.
A. Prolyl oligopeptidase: a potential target for the treatment of
cognitive disorders. Drug News Persp. 2007, 20, 293−305.
(7) Brandt, I.; Scharpe, S.; Lambeir, A. M. Suggested functions for
prolyl oligopeptidase: a puzzling paradox. Clin. Chim. Acta 2007, 377,
50−61.
LC/MS conditions and data analysis: An Acquity UPLC/Q-TOF
system (Waters) with UV detector was employed for data acquisition.
Gradients made up from 10 mM ammonium formate buffer and
MeCN at a flow of 0.5 mL/min were applied on a HSS T3 column
(1.8 μm, 2.1 mm × 100 mm) connected to the mass spectrometer
operating in positive electrospray ionization.
(8) Williams, D. H.; O’Brien, D. P.; Sandercock, A. M.; Stephens, E.
Order changes within receptor systems upon ligand binding: receptor
tightening/oligomerisation and the interpretation of binding param-
eters. J. Mol. Biol. 2004, 340, 373.
(9) Lawandi, J.; Gerber-Lemaire, S.; Juillerat-Jeanneret, L.;
Moitessier, N. Inhibitors of prolyl oligopeptidases for the therapy of
human diseases: defining diseases and inhibitors. J. Med. Chem. 2010,
53, 3423−3438.
(10) Shinoda, M.; Toide, K.; Ohsawa, I.; Kohsaka, S. Specific
inhibitor for prolyl endopeptidase suppresses the generation of
amyloid ? protein in NG108−15 cells. Biochem. Biophys. Res. Commun.
1997, 235, 641−645.
(11) Laitinen, K. S. M.; Van Groen, T.; Tanila, H.; Venalainen, J.;
Mannisto, P. T.; Alafuzoff, I. Brain prolyl oligopeptidase activity is
associated with neuronal damage rather than beta-amyloid accumu-
lation. NeuroReport 2001, 12, 3309−3312.
Q-TOF Premier MSE method was used in full scan mode, and
fragmentation data were obtained by collision energy ramping.
MetaboLynx (Waters) was used to analyze the Q-TOF MSE data.
ASSOCIATED CONTENT
■
S
* Supporting Information
Structure of the 10 active molecules added to the library. This
material is available free of charge via the Internet at http://
pubs.acs.org.
(12) Cunningham, D. F.; O’Connor, B. Identification and initial
characterisation of a N-benzyloxycarbonyl-prolyl-prolinal (Z-Pro-
prolinal)-insensitive 7-(N-benzyloxycarbonyl-glycyl-prolyl-amido)-4-
methylcoumarin (Z-Gly-Bro-NH-Mec)-hydrolysing peptidase in bo-
vine serum. Eur. J. Biochem. 1997, 244, 900−903.
(13) Kuhns, L. R.; Loder, R. T.; Farley, F. A.; Hensinger, R. N.
Nuchal cord changes in children with os odontoideum: evidence for
associated trauma. J. Pediatr. Orthop. 1998, 18, 815−819.
(14) Toide, K.; Shinoda, M.; Iwamoto, Y.; Fujiwara, T.; Okamiya, K.;
Uemura, A. A novel propyl endopeptidase inhibitor, JTP-4819, with
potential for treating Alzheimer’s disease. Behav. Brain Res. 1997, 83,
147−151.
AUTHOR INFORMATION
■
Corresponding Author
*Phone: 514-398-8543. Fax: 514-398-3797. E-mail: nicolas.
moitessier@mcgill.ca.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
(15) Saito, M.; Hashimoto, M.; Kawaguchi, N.; Shibata, H.; Fukami,
H.; Tanaka, T.; Higuchi, N. Synthesis and inhibitory activity of acyl-
peptidyl-pyrrolidine derivatives toward post-proline cleaving enzyme; a
study of subsite specificity. J. Enzyme Inhib. 1991, 5, 51.
(16) Barelli, H.; Petit, A.; Hirsch, E.; Wilk, S.; De Nanteuil, G.;
Morain, P.; Checler, F. S-17092−1, a highly potent, specific and cell
permeant inhibitor of human proline endopeptidase. Biochem. Biophys.
Res. Commun. 1999, 257, 657−661.
We thank AstraZeneca, NSERC, and CIHR for funding as well
as the Faculty of Science (McGill University) for the Fessenden
Professorship in Innovation to NM.
ABBREVIATIONS USED
■
AMC, aminocoumarin; BOP, benzotriazol-1-yl-oxy-tris-(dime-
thylamino)-phosphoniumhexafluorophosphate; DCM, di-
chloromethane; DMF, dimethylformamide; DMSO, dimethyl-
sulfoxide; DPP, dipeptidyl peptidase; HLM, human liver
microsomes; PBS, phosphate buffered saline; Piv, pivaloyl;
POP, prolyl oligopeptidase; RLM, rat liver microsomes; TFA,
trifluoroacetic acid; TFAA, trifluoroacetic anhydride; THF,
tetrahydrofuran
́ ́
(17) Lopez, A.; Tarrago, T.; Giralt, E. Low molecular weight
inhibitors of prolyl oligopeptidase: a review of compounds patented
from 2003 to 2010. Exp. Opin. Ther. Pat. 2011, 21, 1023−1044.
(18) Miyazaki, A.; Toide, K.; Sasaki, Y.; Ichitani, Y.; Iwasaki, T. Effect
of a prolyl endopeptidase inhibitor, JTP-4819, on radial maze
performance in hippocampal-lesioned rats. Pharmacol., Biochem.
Behav. 1998, 59, 361−368.
I
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
(19) Kato, A.; Fukunari, A.; Sakai, Y.; Nakajima, T. Prevention of
amyloid-like deposition by a selective prolyl endopeptidase inhibitor,
Y-29794, in senescence-accelerated mouse. J. Pharmacol. Exp. Ther.
1997, 283, 328−335.
(39) Racys, D. T.; Rea, D.; Fulop, V.; Wills, M. Inhibition of prolyl
̈
̈
oligopeptidase with a synthetic unnatural dipeptide. Bioorg. Med. Chem.
2010, 18, 4775−4782.
(40) Juillerat-Jeanneret, L. Prolyl-specific peptidases and their
inhibitors in biological processes. Curr. Chem. Biol. 2008, 2, 97−109.
(41) Kichik, N.; Tarrago, T.; Claasen, B.; Gairi, M.; Millet, O.; Giralt,
E. 15N Relaxation NMR Studies of Prolyl Oligopeptidase, an 80 kDa
Enzyme, Reveal a Pre-existing Equilibrium between Different
Conformational States. ChemBioChem 2011, 12, 2737−2739.
(20) Umemura, K.; Kondo, K.; Ikeda, Y.; Kobayashi, T.; Urata, Y.;
Nakashima, M. Pharmacokinetics and safety of JTP-4819, a novel
specific orally active prolyl endopeptidase inhibitor, in healthy male
volunteers. Br. J. Clin. Pharmacol. 1997, 43, 613−618.
(21) Jalkanen, A. J.; Hakkarainen, J. J.; Lehtonen, M.; Venalainen, T.;
̈
̈
(42) Kaszuba, K.; Rog
T.; Szeltner, Z.; St. Pierre, J. F.; García-Horsman, A.; Mannisto, P. T.;
́ ́
, T.; Danne, R.; Canning, P.; Fulop, V.; Juhasz,
̈
̈
Kaariainen, T. M.; Jarho, E.; Suhonen, M.; Forsberg, M. M. Brain
̈
̈
̈
̈
̈
Pharmacokinetics of Two Prolyl Oligopeptidase Inhibitors, JTP-4819
and KYP-2047, in the Rat. Basic Clin. Pharmacol. Toxicol. 2011, 109,
443−451.
Karttunen, M.; Hokkanen, J.; Bunker, A. Molecular dynamics,
crystallography and mutagenesis studies on the substrate gating
mechanism of prolyl oligopeptidase. Biochimie 2012, 94, 1398−1411.
(43) Perron, Y. G.; Dewitt, N. Y.; Lee, C., Phthalamido penicillins,
U.S. Patent 3,035,047 1962.
(22) Lawandi, J.; Toumieux, S.; Seyer, V.; Campbell, P.; Thielges, S.;
Juillerat-Jeanneret, L.; Moitessier, N. Constrained peptidomimetics
reveal detailed geometric requirements of covalent prolyl oligopepti-
dase inhibitors. J. Med. Chem. 2009, 52, 6672−6684.
(44) McCormick, J. E.; McElhinney, R. S.; McMurry, T. B. H.;
O’Brien, J. E. N,N′-Transphthaloylation in a Monoprotected Diamine.
Synthesis 2006, 983−988.
(23) Kahyaoglu, A.; Haghjoo, K.; Kraicsovits, F.; Jordan, F.; Polgar, L.
Benzyloxycarbonylprolylprolinal, a transition-state analogue for prolyl
oligopeptidase, forms a tetrahedral adduct with catalytic serine, not a
reactive cysteine. Biochem. J. 1997, 322, 839−843.
(24) Fulop, V.; Bocskey, Z.; Polgar, L. Prolyl oligopeptidase: an
̈
̈
̈
unusual beta-propeller domain regulates proteolysis. Cell 1998, 94,
161.
(25) Moitessier, N.; Englebienne, P.; Lee, D.; Lawandi, J.; Corbeil, C.
R. Towards the development of universal, fast and highly accurate
docking/scoring methods: A long way to go. Br. J. Pharmacol. 2008,
153.
(26) Singh, J.; Petter, R. C.; Baillie, T. A.; Whitty, A. The resurgence
of covalent drugs. Nature Rev. Drug. Discovery 2011, 10, 307−317.
(27) GOLD 3.0; CCDC: Cambridge, UK, 2005.
(28) AutoDock 4.0; The Scripps Research Institute: La Jolla, CA,
2006.
(29) FlexX 3.1.0; BioSolveIT: Sankt Augustin, Germany, 2008.
(30) Corbeil, C. R.; Englebienne, P.; Moitessier, N. Docking ligands
into flexible and solvated macromolecules. 1. Development and
validation of Fitted 1.0. J. Chem. Inf. Model. 2007, 47, 435−449.
(31) Corbeil, C. R.; Englebienne, P.; Yannopoulos, C. G.; Chan, L.;
Das, S. K.; Bilimoria, D.; L’Heureux, L.; Moitessier, N. Docking ligands
into flexible and solvated macromolecules. 2. Development and
application of Fitted 1.5 to the virtual screening of potential HCV
polymerase inhibitors. J. Chem. Inf. Model. 2008, 48, 902−909.
(32) Fradera, X.; Kaur, J.; Mestres, J. Unsupervised guided docking of
covalently bound ligands. J. Comput.-Aided Mol. Des. 2004, 18, 635−
650.
(33) Katritch, V.; Byrd, C. M.; Tseitin, V.; Dai, D.; Raush, E.; Totrov,
M.; Abagyan, R.; Jordan, R.; Hruby, D. E. Discovery of small molecule
inhibitors of ubiquitin-like poxvirus proteinase I7L using homology
modeling and covalent docking approaches. J. Comput.-Aided Mol. Des.
2007, 21, 549−558.
(34) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D.
A. Development and testing of a general Amber force field. J. Comput.
Chem. 2004, 25, 1157−1174.
(35) Englebienne, P.; Moitessier, N. Docking ligands into flexible and
solvated macromolecules. 5. Force-field-based prediction of binding
affinities of ligands to proteins. J. Chem. Inf. Model. 2009, 49, 2564−
2571.
(36) Therrien, E.; Englebienne, P.; Arrowsmith, A. G.; Mendoza-
Sanchez, R.; Corbeil, C. R.; Weill, N.; Campagna-Slater, V.; Moitessier,
N. Integrating Medicinal Chemistry, Organic/Combinatorial Chem-
istry, and Computational Chemistry for the Discovery of Selective
Estrogen Receptor Modulators with Forecaster, a Novel Platform for
Drug Discovery. J. Chem. Inf. Model. 2012, 52, 210−224.
(37) Irwin, J. J.; Shoichet, B. K. ZINCA Free Database of
Commercially Available Compounds for Virtual Screening. J. Chem.
Inf. Model. 2005, 45, 177−182.
(38) A Free Database for Virtual Screening ZINC; http://blaster.
docking.org/zinc/ (Accessed Jan 03, 2012).
J
dx.doi.org/10.1021/jm3002839 | J. Med. Chem. XXXX, XXX, XXX−XXX