Fragment Screening of Soluble Epoxide Hydrolase for Lead Generation - Structure-Based Hit Evaluation and Chemistry Exploration

Article abstract of DOI:10.1002/cmdc.201500575

Soluble epoxide hydrolase (sEH) is involved in the regulation of many biological processes by metabolizing the key bioactive lipid mediator, epoxyeicosatrienoic acids. For the development of sEH inhibitors with improved physicochemical properties, we performed both a fragment screening and a high-throughput screening aiming at an integrated hit evaluation and lead generation. Followed by a joint dose-response analysis to confirm the hits, the identified actives were then effectively triaged by a structure-based hit-classification approach to three prioritized series. Two distinct scaffolds were identified as tractable starting points for potential lead chemistry work. The oxoindoline series bind at the right-hand side of the active-site pocket with hydrogen bonds to the protein. The 2-phenylbenzimidazole-4-sulfonamide series bind at the central channel with significant induced fit, which has not been previously reported. On the basis of the encouraging initial results, we envision that a new lead series with improved properties could be generated if a vector is found that could merge the cyclohexyl functionality of the oxoindoline series with the trifluoromethyl moiety of the 2-phenylbenzimidazole-4-sulfonamide series.

Full text of DOI:10.1002/cmdc.201500575

DOI: 10.1002/cmdc.201500575
Full Papers
Fragment Screening of Soluble Epoxide Hydrolase for
Lead Generation—Structure-Based Hit Evaluation and
Chemistry Exploration
Yafeng Xue,*[a] Thomas Olsson,[b] Carina A Johansson,[a] Linda Öster,[a] Hans-Georg Beisel,[b]
Mattias Rohman,[a] David Karis,[b] and Stefan Bäckstrçm[a]
Soluble epoxide hydrolase (sEH) is involved in the regulation
of many biological processes by metabolizing the key bioactive
lipid mediator, epoxyeicosatrienoic acids. For the development
of sEH inhibitors with improved physicochemical properties,
we performed both a fragment screening and a high-through-
put screening aiming at an integrated hit evaluation and lead
generation. Followed by a joint dose–response analysis to con-
firm the hits, the identified actives were then effectively tri-
aged by a structure-based hit-classification approach to three
prioritized series. Two distinct scaffolds were identified as tract-
able starting points for potential lead chemistry work. The ox-
oindoline series bind at the right-hand side of the active-site
pocket with hydrogen bonds to the protein. The 2-phenylben-
zimidazole-4-sulfonamide series bind at the central channel
with significant induced fit, which has not been previously re-
ported. On the basis of the encouraging initial results, we envi-
sion that a new lead series with improved properties could be
generated if a vector is found that could merge the cyclohexyl
functionality of the oxoindoline series with the trifluoromethyl
moiety of the 2-phenylbenzimidazole-4-sulfonamide series.
Introduction
The oxidation products of arachidonic acid have been estab-
lished as bioactive lipid mediators and are regulators of many
biological processes. These products are formed mainly by the
action of three classes of enzymes: cyclooxygenases, lipoxyge-
nases, and cytochromes. The cytochromes convert arachidonic
acid into epoxyeicosatrienoic acids (EETs). EETs are considered
to have several important biological functions, as they have
been implicated in diverse medical conditions, including hy-
pertension,[1] inflammation,[2] pain,[3] and the metabolic syn-
drome.[4] Given that EETs are mainly metabolized by the
enzyme soluble epoxide hydrolase (sEH, EC 3.3.2.10), inhibition
of this enzyme presents a possibility to address several patho-
logical conditions.[5,6]
compounds for this indication is 1-(1-acetylpiperidin-4-yl)-3-
adamantan-1-ylurea (AR9281, Figure 1[5,15]), which was devel-
oped by Arete Therapeutics. AR9281 was advanced into pha-
se II clinical studies but did not progress further probably be-
cause of a lack of efficacy.[16]
Many pharmaceutical companies have been involved in de-
veloping inhibitors of sEH, initially targeting hypertension.[7,8]
The target was selected based on studies indicating that EETs
were important in regulating vascular function.[9,10] Vasodilation
in several organs, as a result of exposure to EETs, has been re-
ported by a number of groups.[11–14] One of the most studied
[a] Dr. Y. Xue, Dr. C. A. Johansson, Dr. L. Öster, Dr. M. Rohman,
Dr. S. Bäckstrçm
Figure 1. Representative sEH inhibitors[17–21] with their IC50 values for human
sEH.
Department Discovery Sciences, AstraZeneca R&D Gothenburg
Pepparedsleden 1, 431 83 Mçlndal (Sweden)
yafeng.xue@astrazeneca.com
[b] Dr. T. Olsson, Dr. H.-G. Beisel, Dr. D. Karis
Department Medicinal Chemistry, CVMD iMED
AstraZeneca R&D Gothenburg
The structure and catalytic mechanism of sEH have been de-
scribed in several reviews.[6,22] The X-ray crystal structures of
sEH in complex with inhibitors show a very large binding
pocket with a narrow channel in which the actual reaction
occurs.[23] The catalytic site consists of two tyrosine residues
Pepparedsleden 1, 431 83 Mçlndal (Sweden)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201500575.
ChemMedChem 2016, 11, 497 – 508
497
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
(Tyr383 and Tyr466) that activate the epoxide for ring opening
and an aspartate residue (Asp335) for concomitant nucleophilic
attack. The ester formed is then quickly hydrolyzed to the cor-
responding dihydroxyeicosatrienoic acid (DHET). Essentially all
published inhibitors of sEH bind to these two tyrosine residues
and the aspartate residue. A key element for this binding is
usually a carbonyl group of a urea or an amide, which forms
hydrogen bond(s) (H-bond) to the tyrosine residue(s), and the
NH group acts as a H-bond donor to the aspartate residue. Var-
ious amides and ureas have been described as sEH inhibitors
(Figure 1).[17–21]
lead generation of sEH inhibitors. Herein, we present our re-
sults from a recent effort of a combined fragment screening
and high-throughput screening (HTS) hit-finding approach and
subsequent structure-based hit evaluation.
Results and Discussion
Early assay validation and combined fragment screening
and high-throughput screening
In recent years, we have gradually adopted a holistic hit find-
ing approach for selected feasible targets that combines frag-
ment screening and HTS with joint chemistry efforts. The bene-
fit of this strategy has been supported by others.[30] In the case
of sEH, we had planned to run an integrated lead generation
by combining fragment screening and HTS. The plan was to
run a fragment screening in parallel with a HTS, which would
then be followed by joint hit evaluation and chemistry expan-
sion.
The major pharmacophores of these inhibitors are urea- and
amide-based motifs (Figure 1) that bear polar interactions to
the active-site residues. Other parts of the inhibitors are mostly
hydrophobic and mainly have a space-filling function (i.e., van
der Waals interactions), which overall renders most known sEH
inhibitors as having poor physical properties. The secondary
and tertiary pharmacophores include polar groups that interact
with sEH outside the central channel or are in contact with the
solvent.[6]
First, ligand-observed NMR spectroscopy was explored and
established as the primary assay format for fragment screening
(Figure 2). A total of 4056 compounds (including 2861 from
a targeted selection and 1195 from a general selection) were
tested in the NMR spectroscopy binding assay, which resulted
Fragment-based lead generation has been established as
a mature technology for developing drug-like lead series.[24]
For sEH, structure-based virtual screening (VS) was initially
used for fragment work that was focused on lead generation
and further medicinal chemistry expansion.[25,26] In the work of
Tanaka et al.,[25] in demonstrating the great power of VS with
exceptionally high hit rates, the authors noted that the hits
from such work were mainly urea and amide derivatives con-
taining diverse highly lipophilic substituents. In the case of
Xing et al.,[26] the authors showed that VS could be a powerful
tool in combination with combinatorial library design for effi-
cient lead generation. Recently, various computational ap-
proaches have been employed in silico VS for fragment screen-
ing to find sEH inhibitors. In particular, a number of recent
papers have illustrated how sophisticated computational re-
gimes can be applied to sEH to enable efficient in silico frag-
ment screening, for which a number of novel scaffolds have
been identified that are distinct from the general amide and
urea derivatives.[27,28] Notably, in the in silico VS performed by
Moser et al.,[28] the enforced filters included one to satisfy H-
bonding interactions to the active-site triad as an essential
pharmacophore. Furthermore, in most of the structure-based
VS procedures, the active-site pocket of sEH was treated or im-
plied as conformationally rigid. On the other hand, experimen-
tal screening by biochemical and biophysical assays should
allow sampling of a complete conformational space in addition
to identifying inhibitors that bind outside the central channel.
A recent example is fragment screening by X-ray crystallogra-
phy with cocktail soaking, by which a secondary amine-based
scaffold was found as a promising lead for further optimiza-
tion.[29]
Figure 2. 1H NMR signals, CPMG spectra. Signals within yellow boxes belong
to fragment hit binding to sEH. 1D NMR binding assay: A competition ex-
periment by using a known inhibitor of sEH [1-(1-adamantyl)-3-(1-methylsul-
fonyl-4-piperidyl)urea,[15] PDB ID: 5ALZ[32]] as the reference compound. A 1D
NMR ligand-observed CPMG experiment[33] was used for detection. The
assay was performed in three steps: 1) addition of a fragment mixture 2) ad-
dition of sEH protein to detect binding, and 3) addition of the reference
compound to assure specific binding to the active site.
in hit rates of 17 and 27% for the targeted and general libra-
ries, respectively. The overall hit rate was 20%. Next, we deter-
mined the initial binding affinity (Kd) by a NMR spectroscopy
assay with reporter displacement[31] (Figures S2 and S3, Sup-
porting Information). The Kd values for 300 compounds were
determined to fall within the range of 0.018 to 1.5 mm (for de-
tails, see the Supporting Information). Owing to the large
In addition to the consideration of novel intellectual proper-
ty space, there is a clear need to identify distinct scaffolds with
additional polar interactions (i.e., other than interactions with
the active-site residues) to enable the development of sEH in-
hibitors with improved physicochemical properties. Therefore,
we set out to find hit series with non-urea-based scaffolds for
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
498
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
volume of the active-site pocket of sEH, some of the fragment
hits might not bind to the same region as the reporter mole-
cule, which makes relevant Kd values difficult to obtain for
those hits. Nonetheless, binding affinity data were obtained for
the majority of the compounds tested.
100000 compounds with the HTS assay[34] with a hit rate of ap-
proximately 30%. The primary hits from both the fragment
screening and the HTS were then evaluated jointly by dose–re-
sponse analysis (Figure 3b) and subsequent structure-based
classification for chemical expansion (see below).
In parallel, a HTS campaign was started by using a biochemi-
cal assay on an Agilent RapidFire high-throughput mass spec-
trometry system (RapidFire-MS, Agilent Technologies). Prior to
the full execution of the HTS, we performed an extensive assay
validation of the RapidFire-MS assay, including a rigorous cross
validation between the fragment screening (NMR binding) and
the HTS assays (Figure 3a). First, 10000 compounds were
tested by using the HTS assay; this was followed by dose–re-
sponse analysis of the identified actives. Interestingly, an ex-
ceptionally good recovery was obtained despite the high hit
rate in the spot test. Furthermore, a subset (36 compounds) of
the HTS actives was then analyzed by NMR spectroscopy,
which resulted in a remarkable 92% recovery, from which 33
out of 36 compounds were confirmed as specific binders; this
demonstrated the robustness of the HTS assay (Figure 3a). In
a similar effort, a set of fragment hits (300 compounds) origi-
nating from the NMR spectroscopy screen were analyzed by
using the HTS assay and over 66% (200 out of 300 com-
pounds) showed inhibition within the concentration range
used (up to 250 mm) (Figure 3a). Finally, IC50 values from the
HTS assay and Kd values determined by NMR spectroscopy
were compared and they displayed excellent correlation (Fig-
ure S4). Relative to RapidFire-MS, the NMR spectroscopy (re-
porter displacement) assay for Kd determination is significantly
more resource demanding and time consuming owing to its
low throughput. On the basis of this result, we modified our
screening cascade so that the primary hits from both fragment
screening (NMR spectroscopy) and HTS were fed directly into
the HTS assay for dose–response analysis, instead of using
NMR spectroscopy for Kd determination (Figure 3b).
It is remarkable to note that such a high proportion of the
fragment hits (identified as binders) showed clear inhibition in
the biochemical assay (RapidFire-MS), as reflected by the excel-
lent correlation between the two assays (Figure S4). Our result
is consistent with the observation that NMR spectroscopy is
the most robust technique for fragment screening.[35] This is an
important example showing that rigorous early cross validation
between biochemical and biophysical assays can guide the im-
plementation of a more efficient screening cascade for hit find-
ing in lead generation.
Structure-based hit classification for integrated lead
generation
For confirmed actives, compounds were selected for crystalliza-
tion trials. The selection was made initially to map the binding
pocket, but it was subsequently shifted to a structure-based
classification of the hit series. The selection was done from an
affinity ranked hit list by taking into account the chemical
tractability and supposed novelty. Clustering methods based
on molecular graphs were not employed, as fragment-sized
molecules, in contrast to HTS-sized compounds, less frequently
share an actual 3D binding mode within such a 2D cluster.
The clustering was simply done by a 3D overlay of the experi-
mental crystal structures of sEH in complex with selected frag-
ments. From the initial set of crystal structures, analogues were
selected for testing in an iterative approach.
Compared to a traditional HTS, this process placed much
higher weight on the use of the structure information to guide
the hit evaluation. Interesting hits judged solely on their chem-
ical structures were not pursued further unless the X-ray struc-
ture of its ligand complex could be obtained. Such structure-
based hit series were then prioritized for further expansion by
With the progression of such combined fragment screening
and HTS, the observed high hit rate and the represented di-
verse scaffolds led us to decide not to screen the whole com-
pound collection by HTS. In total, we screened approximately
Figure 3. Schematic diagram for a) the cross validation of assays and b) the adapted screening cascade used for hit finding and hit expansion. 7CR assay
refers to 7-concentration dose–response analysis.
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
499
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 4. List of compounds from the prioritized series including some tool compounds. The IC50 [mm] and cLogP values are indicated. The hits are also la-
beled as primary hit (P), analogue (A), tool compound (T), or designed compound (D). Series are classified primarily based on their 3D binding modes as de-
termined by X-ray crystallography (refer to main text for details). Series 1 is represented by compound 2 (with active analogues 3–5 and synthesized 8),
series 2 is represented by compounds 9 and 10, and series 3 is represented by compounds 11–13 (active analogues 15 and synthesized 16). Compounds 1,
6, and 7 were used for comparison for series 1.
consideration of their chemical tractability and structure novel-
ty (see Figure 4 and the discussion below).
Overall shape of the active-site pocket and key interactions
of soluble epoxide hydrolase inhibitors
The active-site pocket of sEH displays a dumbbell shape, with
three active-site residues, Asp335, Tyr383, and Tyr466, located
in the central channel “holding” the midpoint of the
handle.[23,31] On the basis of the ligand–complex structures
available from the PDB, all existing sEH inhibitors bind by pen-
etrating the central channel with their urea-like motifs engag-
ed in H-bonding interactions with the active-site residues
(Figure 5). However, the flanking groups on either side of the
inhibitors rarely engage in significant polar interactions, and
they spread widely to cover the whole volume in the pocket.
For the sake of convenience, we will use the picture in
Figure 5 as the standard viewing orientation. In this view, the
Phe267 pocket as referred to in the literature[23] is part of the
right-hand side (RHS) pocket. Likewise, the Trp336 pocket that
is adjacent to the left of the central channel would be part of
the left-hand side pocket (LHS).
Figure 5. Electron density and binding mode for 4-[4-(1-adamantylcarba-
moylamino)cyclohexoxy]benzoic acid[15] (PDB ID: 5AM3[32]). Note that 5am3
will be used to refer to the bound ligand here and the same scheme will be
used for all cited PDB structures throughout the paper (i.e., small letters of
the PDB ID to refer to the respective bound ligand). The density is shown as
the calculated 2FoÀFc map contoured at 1s. Note that the urea motif is H-
bonded to the active-site residues Asp335, Tyr383, and Tyr466. The end car-
boxylate is H-bonded to Ser415. This view of the active-site pocket will be
used as the default orientation throughout the paper. The Trp336 and
Met339 residues are indicated in the LHS, whereas Phe267, Met419, Ser415,
His524, and Trp525 are located in the RHS.
Identifying important key hot spots of fragment binding
Our crystallographic effort started with the aim to completely
map the active-site pocket with experimental structural data to
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
500
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
gain deeper knowledge of the ligand interactions to guide ra-
tional structure-based design discussions. Therefore, a large
number of literature compounds and diverse series of con-
firmed fragment hits were fed into our crystallographic pro-
cess. In total, more than 50 structures of ligand complexes of
sEH were obtained, which provided a comprehensive experi-
mental mapping of the active-site pocket in terms of protein–
ligand interactions.[26] Superposition of the bound ligands in
these structures resulted in the same dumbbell shape as that
in the active-site pocket (Figure 6).
Figure 6. Overall shape of the active-site pocket with an overlay of sEH in-
hibitors or binders. The active-site residues are depicted as ball-and-stick
models. a) Cartoon models of sEH showing the overall fold and the active-
site pocket in red box; b) close-up view of the active-site pocket with bound
ligands overlaid from 52 structures.[32]
Figure 7. Superposition of ligands that bind to different regions of the
pocket: The active-site residues are depicted as ball-and-stick models to in-
dicate the central active-site channel. a) Compounds that bind through the
central channel; b) fragment hits that bind at the LHS; c) fragment hits that
bind at the RHS. There are two conserved H-bonds to the protein, as indicat-
ed by the dashed lines (one to the side chain of Asp496 and the other to
the backbone NH of Phe497).
Through this analysis, it became clear that there are certain
preferred binding regions with some consensus hot spots, in
addition to the central channel (Figure 7a). In comparison, the
LHS seems to be mostly occupied by larger hydrophobic
groups (Figure 7b).
On the other hand, fragments bound to the RHS mostly
align well in space with their aromatic ring structures in a stack-
ing position to the imidazole ring of the His524 residue
(Figure 5). Such p stacking to His524 has been observed previ-
ously, although no polar interactions were associated with the
stacking.[21] Furthermore, two polar interactions to the protein
seem to be important (Figure 7c). The majority of bound frag-
ments on this side seem to have both or at least one of these
interactions.
Identifying induced-fit binding
For the fragment hits that bind to the central channel, all
showed interactions to the active-site residues (Figure 7a). We
also determined the ligand–complex structure with compound
13 at 2.57 Š resolution (Figure S11 f). If the binding scaffold is
a ring structure (e.g., see compound 13), certain movements
of the protein residues need to take place to accommodate
the ligand. Significantly larger induced fit was found for com-
pound 13 (Figures 4 and 8). In this case, the side chain of
Tyr383 (OH) has to move approximately 2 Š to allow the imida-
zole ring of the ligand to enter the central channel. There are
no clear polar interactions to the sulfonamide group, and the
Figure 8. The unique binding mode and electron density for compound 13
(in cyan), 5am3 (in pink) is overlaid for comparison. The electron density is
shown as blue mesh for bound 13 (2FoÀFc map contoured at 1s). The
nearby Phe381 and Gln384 residues have to move (e.g., up to 5 Š for the
phenyl group of Phe381, which would otherwise collide with the cyclopentyl
group) to accommodate the bulky structure caused by the 7-sulfonamide
kink. In fact, significant portions of the secondary structure including part of
the upstream helix (in pink) and a loop region (residues 375–379) became
disordered in the crystal. The hydroxyl group of Tyr383 is H-bonded to the
sulfonamide nitrogen. The imidazole[4,5b]pyridine ring is in a stacking posi-
tion to the side chain of Trp336 and is also close to Met339. a) Side view;
b) front view (908 apart from a).
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
501
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
cyclopentyl group seems to open a new hydrophobic pocket
near Phe381 (Figure 8). Similar induced-fit features were also
observed for compound 16, for which the complex structure
was solved at 2.35 Š resolution. Interestingly, the binding
mode of compound 15 (which has 2-phenylbenzimidazole-5-
sulfonamide as the core, Figure S10b) did not show any in-
duced-fit features. It is evident that such induced-fit binding
depends on the specific substitution pattern of the benzimida-
zole ring. This feature of significant induced fit near the central
channel has not been reported previously,[26] and it can be
a new direction for sEH inhibitor design. Another region that
showed induced-fit binding of ligands is the 410s loop (see
below). Notably, compound 13 was identified from the HTS, al-
though hits with the same scaffold were also found in the
NMR spectroscopy screening (from the targeted library, see
below). The observation regarding protein flexibility upon
binding of fragments is in line with a recently reported exam-
ple on mitochondrial branched-chain aminotransferase,[36] for
which it was shown that combined fragment screening and
HTS enabled identification of the hit series with novel induced-
fit interactions.
Oxoindoline was identified as a novel hit series with polar
interactions to the right-hand side
Figure 9. Overlay and comparison of series 1 and literature compounds.[21, 25]
a) Overlay of 2 (pink) with 3ant (green) and 3i1y (blue); b) overlay of 2
(purple), 3 (pink), 4 (grey), 5 (light green), and 3ant (green); c) overlay of 2
(pink), 6 (beige), 7 (brown), 8 (yellow), and 3ant (green); d) overlay of 8
(yellow) and 6 (light green) showing the polar interaction between 8 and
the 410s loop, which moves towards the bound ligand. 5am3 is shown as
gray models. The 410s loop collapses upon binding of 8, which induces
large movement around Val416 (ꢀ4 Š for the Ca as indicated by the red
arrow). 3ant, 3i1y are literature inhibitors, see Figure S1.
On the basis of their binding modes and overlay with literature
sEH inhibitors, we selected three major series (i.e., prioritized
series) for further hit expansion and lead generation (Figure 4).
From the primary hits, series 1 is represented by compound 2
(with active analogues 3–5 and synthesized 8 from the hit ex-
pansion), series 2 is represented by compounds 9 and 10, and
series 3 is represented by compounds 11–13 (active analogue
15 and synthesized 16). Compounds 1, 6, and 7 were used for
comparison for series 1. Below, we briefly describe the chemis-
try exploration of the respective series.
pounds were selected and 84 were tested. More than 80%
were found to be active. Among the active analogues, com-
plex structures with 3, 4, and 5 were determined (Figure 9b).
As seen for other more potent inhibitors (e.g., 3i1y, Figure S1),
the groups occupying the Phe267 pocket seem to prefer an
out-of-plane/perpendicular orientation relative to the oxoindo-
line system.
For series 1, it is most striking to see that compounds 1 and
2, given their same core scaffolds, bind in completely different
regions (ꢀ15 Š apart) in the pocket: 1 binds at the LHS (Fig-
ure S11a), whereas 2 binds at the RHS by p stacking to His524
and is H-bonded to Asp496 and Phe497, which are part of the
identified hot spots (Figure 7c; see also Figure S11b). For com-
pound 1, the 3-benzyl group seems to be too big to fit into
the RHS pocket (hence preventing its binding to the RHS), and
this indicates that the 3D binding mode of the ligand is sensi-
tive to the specific substitution scheme, even with the same
core scaffold. For compound 2, its pyrazole moiety seems
mainly to fill up the Phe267 pocket[21] and to interact with the
central channel indirectly through a bridging water molecule
to Asp335. The Phe267 pocket is occupied by other known
sEH inhibitors[5,21,25,26,37] (Figure 9a; see also Figure S1). We rea-
soned that the significant polar interactions observed for 2 can
be used for further chemistry efforts to test the idea that new
leads can be generated without interactions to the central
channel.
For our design hypothesis, we also compared series 1 with
other fragment hits. Compound 6 displays potency similar to
that of 2, although 6 has only one H-bond to the protein (Fig-
ure 9c). Notably, the nitrogen atom of the pyridine ring inter-
acts with the 410s loop through a bridging water molecule.
Compound 7 has a phenol moiety as the core scaffold and is
H-bonded, through its hydroxyl group, to the protein (Phe497)
(Figure 9c). Most interesting is the observation that the car-
boxylate group from 7 occupies a part of the RHS pocket that
is similar to that occupied by the pyridine ring in 6. We consid-
er this region as a potential direction from which we can fur-
ther explore by adding relevant bulky substituents to form
polar interactions. Compound 8 was synthesized to test such
a design hypothesis, and its complex structure with sEH was
also determined (Figure 9d). The cyclohexyl group is virtually
superimposed on top of that of 5 but with a 1808 flip; this can
likely be attributed to the large size of the tetrahydropyran
We then searched our in-house compound collection to
identify structural analogues to 2 and similar structural motifs
and tested them in our RapidFire-MS assay. A total of 107 com-
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
502
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
ring, which is in a chair configuration with the ring oxygen
atom interacting with Ser407 and Leu417 through a bridging
water molecule (Figure 9d; see also Figure S10c). Interestingly,
the 410s loop seems to have “collapsed” towards 8 with
Val416 flipped from an outward position to an inward position.
This illustrates that the RHS pocket possesses a certain level of
flexibility (i.e., is able to constitute induced-fit binding to some
ligands).[32] The binding mode of 8 to a large extent confirmed
our design hypothesis, which provided good basis for further
chemistry exploration.
without strong polar interactions. Given that this did not offer
any clear route for chemistry expansion, we decided not to
pursue this series further.
2-Phenylbenzimidazole-4-sulfonamide was identified as
a hit series with novel interactions to the central channel
For series 3, we wanted to explore the hits that penetrate the
central channel but that have distinct features. Compound 12
is interesting, as it interacts with the central channel and is
also H-bonded to Asp496 through its phenolic hydroxyl group
(Figure 10b). Unlike previous sEH inhibitors, 12 has a secondary
amine group instead of a urea-based motif. Furthermore, com-
pared to series 1, the substituted phenyl moiety approximately
overlaps the benzene ring of the oxoindoline system, and the
tert-butyl group superposes the 5-substituent of the oxoindo-
line ring well (Figure 10b,c). We speculate that it might be
possible to combine these two features to a new lead series
that makes significant polar interactions in the RHS pocket and
to the central channel (Figure 10c). Compound 11 has a urea-
based motif, that is, a piperidine-1-carboxamide moiety, as pre-
viously reported[25,38] (e.g., 3ant and 3koo, see Figure S1). How-
ever, the large substituent in 11, that is, the dibenzothiophene-
sulfone moiety, occupies the Phe267 pocket mainly with a hy-
drophobic and space-filling function.
For series 2, we included 9 and 10 as representatives. Both
bind at the RHS pocket with p stacking to His524 but with dif-
ferent core scaffolds (Figure 10a). They are coplanar to each
other in the superposition and approximately overlap the end
groups of the literature inhibitors 1zd2, 1zd5, 3koo, and 4jnc
Inhibitors with larger hydrophobic groups occupying the
RHS pocket include an amide-based pyrazole[40] and a piperi-
dine derivative[41] (3otq and 4hai, see Figure S1), which are
both oriented perpendicular to the dibenzothiophenesulfone
moiety in 11 (Figure 10c). It would be desirable to combine
this large hydrophobic functionality with the features of
series 1. Next, we turned to the most interesting hit from
series 3. Compound 13 has an imidazo[4,5-b]pyridine group
that binds to the central channel with significant induced fit
(Figure 10d). Its potency is nearly 10-fold better (IC50 =19 nm)
than that of 2 from series 1 (IC50 =160 nm). Compound 15 was
tested as an active analogue; it has a benzimidazole core scaf-
fold. Changing from a 4-substituent to a 5-substituent elimi-
nated the need for a large rearrangement around the LHS
pocket (Figure 10d). Furthermore, owing to its smaller end
group (unsubstituted phenyl group at the 2-position of the
benzimidazole ring), compound 15 shows much weaker bind-
ing (10-fold difference in IC50 value; Figure 4) than 13 and dis-
plays dual binding modes similar to 14 (Figure S10), which in-
dicates that the aliphatic tail on the 5-sulfonamide unit can fit
in either side of the pocket with no clear discrimination. We
envisage that structures represented by 13 can be used as
a starting point for new sEH inhibitors. As an initial attempt,
16 was designed to simplify the ring structure and to test
a more feasible synthesis. Compounds 16 and 13 only differ by
one atom (benzene versus pyridine ring) and show similar po-
tency (IC50 =17 and 19 nm). To validate our design, we also ob-
tained the complex structure for 16. The only difference rela-
tive to 13 is the orientation of the cyclopentyl group, which is
rotated in 16 by approximately 808 around the sulfonamide SÀ
N bond (Figure 10d), likely as a result of differences in the elec-
tronic properties.
Figure 10. a) Overlay of series 2 with literature compounds: 9 (deep green),
10 (brown), 1zd2 (yellow), 1zd5 (blue), 3koo (cyan), and 4jnc (green)[23,38,39]
Binding modes for series 3: b) overlay of 12 (beige), 11 (blue), and 14
(purple); c) overlay of 11 (blue), 8 (yellow), 3otq (orange), and 4hai (light
.
green);[40,41] d) overlay of 15 (blue), 13 (cyan), 16 (orange), and 5am3 (white).
1zd2, 1zd5, 3koo, 4jnc, 3otq, and 4hai are literature inhibitors, see Figure S1.
(Figure 10a; see also Figure S1). Interestingly, the isoxazole
part of 10 seems to mainly fill up the space of the RHS pocket
with its ring nitrogen atom in contact with the solvent mole-
cules. Relative to the potency of 5 (series 1), that of 9 is signifi-
cantly lower (ꢀ50-fold based on the IC50 value); this decrease
in potency is likely due to weakened H-bonds to the Asp496
residue (distance changes from 2.7 to 3.2 Š) and the Phe497
residue (from 2.8 to 3.1 Š) in addition to the much smaller
group at the Phe267 pocket. In summary, we found that the
main binding mode for series 2 involves a planar structure
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
503
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
It was reported in 2012[42] that in animal models the inhibi-
tors of sEH elevated endogenous EET levels and promoted pri-
mary tumor growth and metastasis. Owing to the potential lia-
bility of such a target-related tumorigenesis effect, the sEH
project at AstraZeneca was terminated and our lead genera-
tion effort was stopped prematurely. No further work was con-
ducted.
pounds tested in the respective screens (i.e., total ꢀ4000 vs.
ꢀ100000, which gives 4:100, Table 1). However, there is
a trend that series 2 seems to be significantly over-represented
from the fragment screening (e.g., hit ratio 60:100 between
the fragment screening and HTS for the benzothiazinones)
than other scaffolds. This could be due to the fact that the
core scaffolds of series 2 (MW: ꢀ170 Da) are more enriched in
the fragment libraries than the HTS collection. In contrast,
series 3 (the 2-phenylbenzimidazole-4-sulfonamide scaffold)
was also found among the HTS hits but not from the generic
fragment library. This is consistent with the fact that the HTS
compounds have a size distribution towards larger MWs. Nota-
bly, the most potent 2-phenylbenzimidazole-4-sulfonamide
compound from HTS (compound 13 in Figure 4) has a MW of
approximately 410 Da.
After the closing phase of this project, 2-phenylbenzimida-
zole and related derivatives were reported as fragment hits for
sEH on the basis of elaborate in silico approaches.[27,43] Whereas
the best 2-phenylbenzimidazole compound from those in silico
screenings has a potency in the low micromolar range, com-
pound 16 as a representative of our 2-phenylbenzimidazole-4-
sulfonamides has significantly higher potency (IC50 =17 nm).
More importantly, the induced-fit features associated with the
2-phenylbenzimidazole-4-sulfonamide scaffold as revealed by
high-resolution crystal structures present distinct novel interac-
tions with sEH.
It should also be pointed out that the NMR spectroscopy
hits are much less potent than the hits from the HTS (e.g., the
best 2-phenylbenzimidazole hit from NMR spectroscopy has an
IC50 value of 185 mm, which is ꢀ10000-fold less potent than
compound 13). Without the HTS hits, it would probably have
limited our chance to obtain the structure information for this
scaffold. As we had concluded from earlier analysis, the success
rate to obtain complex structures seems to be directly correlat-
ed to their affinity in the case of sEH, for example, only 35%
success rate if the affinity is weak with an IC50 value higher
than 100 mm.[32] Given the fact that we were relying on the
structure information to guide the hit evaluation, this scaffold
might not be prioritized if we had not gotten the complex
structure with 13 (which originated from HTS).
Retrospective analysis of hits from the fragment screening
and the high-throughput screening
Given that both the fragment screening and the HTS gave
high hit rates, we wanted to understand how these two
screening campaigns contributed to the hit series identified.
Therefore, we performed a retrospective analysis of the priori-
tized series by looking at their origins, NMR spectroscopy
screening versus HTS as the primary screening, as well as tar-
geted versus generic libraries in the case of NMR spectroscopy
screening.
For the NMR spectroscopy screen, the 2-phenylbenzimida-
zole scaffold in series 3 was enriched significantly by the tar-
geted selection (Figure 4 and Table 1). This series showed
novel interactions to the central channel and afforded a poten-
tial new scaffold for sEH inhibitors. However, this scaffold was
not found among the hits from the generic fragment library,
whereas three such hits were identified from the targeted li-
brary. One clear difference between the generic and the target-
ed libraries lies in their mean molecular weights (MWs, 159 vs.
210 Da); this may be a reasonable explanation for the above
observation. Indeed, it was found that compounds bearing the
2-phenylbenzimidazole group (MW: 194 Da for the core scaf-
fold) simply did not exist in the generic library.
Conclusions
Through a combination of fragment screening and high-
throughput screening, that is, a parallel primary screening fol-
lowed by integrated hit-confirmation and structure-based hit
evaluation, we identified two scaffolds, oxoindoline and 2-phe-
nylbenzimidazole-4-sulfonamide, as new chemical series for
potential soluble epoxide hydrolase inhibitors. With our initial
chemistry effort, we demonstrated that the oxoindoline series
can be further expanded by retaining the polar interactions to
the right-hand side (H-bonds to Asp496 and Phe496), and spe-
cific interactions can be introduced to target the 410s loop
region by induced fit. In addition, we showed that the 2-phe-
nylbenzimidazole-4-sulfonamide series seem to afford synthetic
feasibility for further work. On the basis of the encouraging ini-
tial results, we envision that a new lead series with improved
Looking at the hits from NMR spectroscopy and HTS, the
scaffolds in series 1 and 3 showed comparable hit rates for
both screens (roughly proportional to the number of com-
Table 1. Origin of the key scaffolds versus number of hits in the prioritized series.
Series
Scaffold
Representative
compound[a]
NMR hits
Total hits
generic library
targeted library
NMR
HTS
>80
5
20
>40
1
2
oxoindoline
benzothiazinone
phenylisoxazole
2
9
10
13
4
2
1
0
2
1
2
3
6
3
3
3
3
2-phenylbenzimidazole
[a] See Figure 4 for structures of the primary hits.
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
504
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
ed, which resulted in a set of approximately 3000 compounds. This
set was sorted by the heteroatom fraction, which was followed by
a thorough visual inspection. The selection of compounds for a ge-
neric NMR spectroscopy screening library resulted in a set of 1202
compounds with a mean MW of 159 Da. Some basic statistics are
given in the Supporting Information. A total of 1195 of these com-
pounds were included in the primary fragment screen by NMR
spectroscopy.
properties (especially physicochemical properties) could be
generated if we combine features from 5-cyclohexylspiro[indo-
line-3,4’-tetrahydropyran]-2-one (8) and N-cyclopentyl-2-[4-(tri-
fluoromethyl)phenyl]-1H-benzimidazole-4-sulfonamide
(16).
The main challenge would be to find a vector that can merge
the cyclohexyl functionality of 8 with the trifluoromethyl
moiety of 16 (red circle in Figure 11). With the rich structure in-
formation, we envisage that functional groups such as sulfona-
mides or large aliphatic ring systems could be tested as poten-
tial candidates for such a purpose.
Targeted screening library: The targeted fragment screening li-
brary was built in the following way. Out of the AZ corporate com-
pound collection, all compounds with no more than 4 rotatable
bonds (SciTegic def.), no more than 18 HA, calculated logD7.4 (Sci-
Tegic def.) lower than or equal to 4, at least 1 ring system, and at
least 1 heteroatom were selected, which resulted in approximately
134000 compounds for the targeted virtual screening set I. A
second set, the targeted virtual screening set II, comprised all
16000 compounds in the AZ general fragment library. Both sets
were converted separately into conformational ensembles of 3D
structures for virtual screening. First, for all unique compounds
protonation states were set and tautomers were enumerated as
appropriate.[44] All undefined stereochemistry, except on pyramidal
nitrogen atoms, was then enumerated, while keeping all defined
stereochemical features.[45] Initial 3D models, as well as conforma-
tional ensembles were generated by using Omega 2.4.3[46] with
these parameters: buidff=mmff94s_NoEstat, enumNitrogen=true,
enumRing=true, ewindow=10.0, maxconfs=500, rms=0.5,
searchff=mmff94s NoEstat, and an in-house developed torsion
angle library.[47] The virtual screening of these compounds was per-
formed as a rigid docking by using FRED 2.2.5.[48] Each of the two
targeted virtual screening libraries was docked separately by using
a total of 20 boxes, derived from 7 fragmentations differing in size
and covered space of a reference compound [PDB ID: 3I28, re-re-
fined].
Figure 11. Overlay of compounds 8 (yellow) and 16 (green) binding to sEH.
The red circle indicates the overlap of the cyclohexyl group of 8 and with
the trifluoromethyl moiety of 16.
For each run separately, all hits were ranked by consensus scoring
with Chemgauss3, OEChemscore, and PLP, this way discarding the
worst poses by agreement of multiple different scoring functions,
rather than simply selecting the “top n” solutions for any single-
scoring function. The targeted screening sets I and II resulted in
4636 and 3630 compound hits, respectively. Of these, 55 and 85%
were, respectively, kept by random selection, which resulted in
a pooled set of 5821 virtual screening hits on a unique compound
level.
The present work illustrates the importance of early cross
validation of the different assays, as this guided us to adapt
the screening cascade. The cross validation was then followed
by an integrated hit evaluation by using a structure-based ap-
proach, as we initially planned. Such an integrated strategy en-
abled an effective hit-finding process that resulted in
a number of hit series with distinct interactions to soluble ep-
oxide hydrolase.
The targeted screening library was then finalized by several rounds
of thorough visual inspection and selection, which resulted in a tar-
geted NMR screening library of approximately 3000 compounds.
The initial set of 5821 compounds showed a mean MW of 210 Da.
Some basic statistics are given in the Supporting Information. After
several rounds of careful visual inspection, a subset of 2861 com-
pounds was included in the primary fragment screen by NMR spec-
troscopy. For further details of the method and a discussion, refer
to the Supporting Information.
Experimental Section
General comments: 11,12-Epoxy-(5Z,8Z,14Z)-eicosatrieonic acid
(11,12-EET) and 11,12-dihydro-(5Z,8Z,14Z)-eicosatrieonic acid (11,12-
DHET) were obtained from Caman Chemicals. Human soluble ep-
oxide hydrolase (sEH) was prepared in house. All other chemicals
were from Sigma, unless otherwise stated. Complete details of the
protein constructs and recombinant production are given in the
Supporting Information.
Molecular modeling: Proteins were prepared by using Schrçdinger
software,[49] 3D models were generated by using CORINA,[45] and
flexible dockings were performed with GOLD[50] by writing out all
poses for subsequent post-processing. For full details, refer to the
Supporting Information.
Library selection and molecular modeling
Fragment screening by NMR spectroscopy
Generic screening library: The AstraZeneca (AZ) general fragment
library currently comprises about 16000 compounds. Out of these,
compounds with no more than 12 heavy atoms (HA) were select-
The NMR spectroscopy experiments were conducted with a Bruker
600 or 800 MHz spectrometer equipped with a cryogenically
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
505
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
cooled TXI-probe head. A SampleRailTM sample delivery system
(Bruker BioSpin) connected the spectrometer with a Tecan sample
preparation robot. Each protein or ligand addition was done just
prior to the NMR spectroscopy experiment. To avoid dilution with
robot system liquid (water), an in-house constructed piston was
used for mixing the sample in the NMR tube. Ligand binding was
mode with the following transitions: 11,12-EET 321.3/303.2 and
11,12-DHET 339.2/135.
The ratio (R) of substrate (S) to product (P) was calculated by using
Equation (1). The percentage effect of a compound’s ability to in-
hibit the enzyme was calculated by using Equation (2). In the spot
test (single concentration at 20 mm), a tested compound that
showed at least 30% inhibition was regarded as a hit.
1
detected by using 1D H CPMG NMR spectroscopy experiments ac-
quired at 293 K with 256 scans and a repetition delay of 3 s. The
CPMG spin echo part of the sequence consisted of 96 pulses with
1808 flip angle, interspaced by 2 ms, to give a spin lock time of
196 ms. A dual selective pulse (shape sinc1, 3 ms) was used in the
excitation-sculpting scheme to suppress both the residual water
and DMSO signals simultaneously.[51] The NMR spectroscopy data
was processed with 2 Hz line broadening and base-line correction.
The resonance integrals were extracted by using TopSpin (Bruker).
The dissociation constants obtained from the mutual titration ex-
periments were determined by using the Levenberg–Marquardt
least-squares optimization[52] in Matlab (Mathworks Inc., imple-
mented by Kristofer Modig, unpublished results). The uncertainties
in the fitted parameters (Kd,1, Imin,1, Imax,1, Kd,2, Imin,2, and Imax,2) were es-
timated by scaling the experimental error to give a reduced c2 =1.
P
R ¼
ð1Þ
ð2Þ
ðP þ SÞ
ðIÀDMSOÞ
% Effect ¼ 100 Â
ðcontrolÀDMSOÞ
in which DMSO is the ratio for the well containing only DMSO and
control is the ratio for the well with a strong inhibitor (I).
Fragments were screened from cocktails containing six compounds
(prepared in a randomized way) each at a concentration of 50 mm.
The concentration of reference compound [1-(1-adamantyl)-3-(1-
methylsulfonyl-4-piperidyl)urea,[15] Figure 2] was 50 mm. Analysis of
the fragment screening was done by overlaying the spectra of the
fragment mixture, the fragment mixture in the presence of the
protein, and the fragment mixture in the presence of the protein
and the strong inhibitor in TopSpin (Bruker). Fragment hits were
identified by comparison with prerecorded reference spectra of all
fragments. Fragments showing at least 40% decrease in signal in-
tensity and a substantial regain of signal intensity upon addition of
the reference compound were regarded as hits. The buffer was
50 mm deuterated Tris pH 7.5, 50 mm NaCl, 10% D2O. The protein
stock concentration was 100 mm.
Crystallization and structure determination
For detailed experimental procedures, refer to Ref. [32]. The struc-
ture with compound 16 was obtained by co-crystallization. Apo-
protein (18 mgmLÀ1) was mixed with the compound to a concen-
tration of 5 mm and was incubated at room temperature for 3 h.
Thereafter, crystals were setup in the same manner as for the apo-
protein and were flash frozen without any addition of cryoprotec-
tion.
Structures of the ligand complexes with compounds 1–15 were
deposited in the PDB with IDs: 5ALO, 5ALY, 5ALN, 5AM0, 5ALD,
5AKH, 5AIA, 5AM4, 5AIC, 5ALE, 5AKE, 5AI9, 5ALS, 5AKZ, and
5AKF.[32] Details and statistics of data collection and refinement for
sEH in complex with 16 are recorded in Table S1. The atomic coor-
dinates and the structure factors of this sEH complex were deposit-
ed in the protein data bank (www.pdb.org) with PDB ID: 5FP0.
NMR spectroscopy determination of the Kd values of the fragments
was performed by using a 1D NMR reporter displacement assay.
The reporter signal intensity recovery was monitored using
a CPMG experiment. The Kd values were calculated from equations
derived for a three-component competitive equilibrium model in-
volving the reporter, the test compound, and protein relating the
signal recovery to the concentration of the competing fragment.
For more details, refer to the Supporting Information.
PYMOL[53] was used for structure comparison and preparing figures.
Note that capital letters are used for the PDB IDs and the same IDs
in small letters are used to refer to the ligand bound in that struc-
ture.
HTS screening
Synthetic chemistry
Plate set up: For the HTS screen, the compound (40 nL) was
added to a 384 Greiner polystyrene plate by using an acoustic dis-
penser (Ecco, Labcyte). The enzyme (20 mL) in 50 mm Tris, pH 7.4,
was added to the plate (Mulitdrop); the mixture was pre-incubated
for 15 min and then 11,12-EET (20 mL) (Multidrop) was added, and
the reaction was allowed to incubate for 30 min. The reaction was
stopped by adding water (40 mL) containing a strong inhibitor
(Multidrop). Final concentration was 5 nm enzyme, 1 mm 11,12-EET,
and 20 mm inhibitor.
Chemicals and solvents from commercially available sources were
purchased and used without further purification. 1H NMR spectra
were recorded with a Bruker Biospin GmbH 400, 500, or 600 MHz
spectrometer. Chemical shifts are reported in parts-per-million (d)
relative to [D6]DMSO at d=2.50 ppm, CDCl3 at d=7.26 ppm, or
CD3OD at d=3.31 ppm as an internal standard. All tested com-
1
pounds were purified to >95% purity, as determined by H NMR
spectroscopy and HRMS or HPLC–UV. Heating of reactions with mi-
crowaves was performed with a Biotage Initiator 2.45 GHz, 400 W
microwave for the times indicated. Compounds in Figure 4 (except
3, 4, 8, and 16) were obtained from the AstraZeneca corporate
compound collection. Experimental and spectroscopic details for
noncommercially available compounds 3, 4, 8, and 16 are de-
scribed in the Supporting Information. In general, optimizations of
the reaction procedures were not performed and yields are given
for isolated materials.
RapidFire-MS: The plates was loaded into a RapidFire 300 (Agilent
Technologies) coupled to an API 4000 mass spectrometer (SCIEX).
Buffer A: 0.2% formic acid in water and buffer B: 0.2% formic acid
in methanol. Samples were loaded onto a C18 column and nonvo-
latile components was removed by washing the column with buf-
fer A. Elution of 11,12 EET and 11,12-DHET was done by using
buffer. Detection was performed by using the positive ESI MRM
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
506
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[19] S. Anandan, H. K. Webb, D. Chen, Y. Wang, B. R. Aavula, S. Cases, Y.
Cheng, Z. N. Do, U. Mehra, V. Tran, J. Vincelette, J. Waszczuk, K. White,
K. R. Wong, L. Zhang, P. D. Jones, B. D. Hammock, D. V. Patel, R. Whit-
comb, D. E. MacIntyre, J. Sabry, R. Gless, Bioorg. Med. Chem. Lett. 2011,
21, 983–988.
[20] H. C. Shen, F. Ding, S. Wang, S. Xu, H. Chen, X. Tong, V. Tong, K. Mitra, S.
Kumar, X. Zhang, Y. Chen, G. Zhou, L. Pai, M. Alonso-Galicia, X. Chen, B.
Zhang, J. R. Tata, J. P. Berger, S. L. Colletti, Bioorg. Med. Chem. Lett. 2009,
19, 3398–3404.
[21] A. B. Eldrup, F. Soleymanzadeh, S. J. Taylor, I. Muegge, N. A. Farrow, D.
Joseph, K. McKellop, C. C. Man, A. Kukulka, S. De Lombaert, J. Med.
Chem. 2009, 52, 5880–5895.
[22] R. H. Ingraham, R. D. Gless, H. Y. Lo, Curr. Med. Chem. 2011, 18, 587–
603.
Author contributions
S.B. performed the experiments on protein production; H.-G.B. per-
formed the computational studies for fragment library design and
hit evaluation; C.A.J. performed the NMR spectroscopy experi-
ments; L.Ö. performed the crystallization experiments; M.R. per-
formed the HTS and dose–response analyses; D.K. participated in
the medicinal chemistry design and hit evaluation; T.O. led medici-
nal chemistry and the hit evaluation and wrote the Introduction;
Y.X. performed the structural analyses and comparison and wrote
and edited the manuscript; H.-G.B., T.O, and Y.X. participated in the
structure-based design; S.B., H.-G.B., C.A.J., L.Ö., M.R., and T.O.
wrote the relevant experimental parts; and H.-G.B., C.A. J., L.Ö.,
and T.O. provided critical reviews.
[23] G. A. Gomez, C. Morisseau, B. D. Hammock, D. W. Christianson, Protein
Sci. 2006, 15, 58–64.
[24] M. Baker, Nat. Rev. Drug Discovery 2013, 12, 5–7.
[25] D. Tanaka, Y. Tsuda, T. Shiyama, T. Nishimura, N. Chiyo, Y. Tominaga, N.
Sawada, T. Mimoto, N. Kusunose, J. Med. Chem. 2011, 54, 851–857.
[26] L. Xing, J. J. McDonald, S. A. Kolodziej, R. G. Kurumbail, J. M. Williams,
C. J. Warren, J. M. O’Neal, J. E. Skepner, S. L. Roberds, J. Med. Chem.
2011, 54, 1211–1222.
[27] J. Achenbach, F. Klingler, R. Blçcher, D. Moser, A. Häfner, C. B. Rçdl, S.
Kretschmer, B. Krüger, F. Lçhr, H. Stark, B. Hofmann, D. Steinhilber, E.
Proschak, ACS Med. Chem. Lett. 2013, 4, 1169–1172.
[28] D. Moser, J. Achenbach, F. Klingler, B. Estella, S. Hahn, E. Proschak,
Bioorg. Med. Chem. Lett. 2012, 22, 6762–6765.
[29] Y. Amano, E. Tanabe, T. Yamaguchi, Bioorg. Med. Chem. 2015, 23, 2310–
2317.
Acknowledgements
The authors thank Margareta Herslçf for her efforts in the early
stages of the project; Erik Olandersson, Kjell Johansson, and
Johan Cassel for their work in synthetic chemistry; Anna Öhrn for
her contribution in the NMR spectroscopy screening; Gisela Brän-
dØn and Gerhard Fischer for their contributions in the early struc-
ture work; and Victoria B. K. Ullah for her assistance in retrieving
some of the experimental information.
[30] M. Whittaker, R. J. Law, O. Ichihara, T. Hesterkamp, D. Hallett, Drug Dis-
covery Today Technol. 2010, 7, e163–e171.
[31] C. Dalvit, M. Fasolini, M. Flocco, S. Knapp, P. Pevarello, M. Veronesi, J.
Med. Chem. 2002, 45, 2610–2614.
[32] L. Öster, S. Tapani, Y. Xue, H. Käck, Drug Discovery Today 2015, 20, 1104–
1111.
Keywords: drug discovery
·
high-throughput screening
·
inhibitors
hydrolase
·
ligand complex structures soluble epoxide
·
[33] S. Meiboom, D. Gill, Rev. Sci. Instrum. 1958, 29, 688–691.
[34] J. W. M. Nissink, S. Schmitt, S. Blackburn, S. Peters, J. Biomol. Screening
2014, 19, 369–378.
[1] M. A. Hye Khan, T. S. Pavlov, S. V. Christain, J. Neckar, A. Staruschenko,
K. M. Gauthier, J. H. Capdevila, J. R. Falck, W. B. Campbell, J. D. Imig, Clin.
Sci. 2014, 127, 463–474.
[35] R. E. Hubbard, J. B. Murray, Methods Enzymol. 2011, 493, 509–531.
[36] S. M. Bertrand, N. Ancellin, B. Beaufils, R. P. Bingham, J. A. Borthwick, A.
Boullay, E. Boursier, P. S. Carter, C. Chung, I. Churcher, N. Dodic, M. Fou-
chet, C. Fournier, P. L. Francis, L. A. Gummer, K. Herry, A. Hobbs, C. I.
Hobbs, P. Homes, C. Jamieson, E. Nicodeme, S. D. Pickett, I. H. Reid, G. L.
Simpson, L. A. Sloan, S. E. Smith, D. O. Somers, C. Spitzfaden, C. J. Suck-
ling, K. Valko, Y. Washio, R. J. Young, J. Med. Chem. 2015, 58, 7140–
7163.
[37] G. A. Gomez, C. Morisseau, B. D. Hammock, D. W. Christianson, Biochem-
istry 2004, 43, 4716–4723.
[38] A. B. Eldrup, F. Soleymanzadeh, N. A. Farrow, A. Kukulka, S. De Lombaert,
Bioorg. Med. Chem. Lett. 2010, 20, 571–575.
[39] R. K. Thalji, J. J. McAtee, S. Belyanskaya, M. Brandt, G. D. Brown, M. H.
Costell, Y. Ding, J. W. Dodson, S. H. Eisennagel, R. E. Fries, J. W. Gross,
M. R. Harpel, D. A. Holt, D. I. Israel, L. J. Jolivette, D. Krosky, H. Li, Q. Lu,
T. Mandichak, T. Roethke, C. G. Schnackenberg, B. Schwartz, L. M. Shew-
chuk, W. Xie, D. J. Behm, S. A. Douglas, A. L. Shaw, J. P. Marino, Jr.,
Bioorg. Med. Chem. Lett. 2013, 23, 3584–3588.
[2] W. B. Campbell, Trends Pharmacol. Sci. 2000, 21, 125–127.
[3] S. D. Kodani, B. D. Hammock, Drug Metab. Dispos. 2015, 43, 788–802.
[4] K. Sodhi, K. Inoue, K. H. Gotlinger, M. Canestraro, L. Vanella, D. H. Kim,
V. L. Manthati, S. R. Koduru, J. R. Falck, M. L. Schwartzman, N. G. Abra-
ham, J. Pharmacol. Exp. Ther. 2009, 331, 906–916.
[5] C. Morisseau, B. D. Hammock, Annu. Rev. Pharmacol. Toxicol. 2013, 53,
37–58.
[6] H. C. Shen, B. D. Hammock, J. Med. Chem. 2012, 55, 1789–1808.
[7] W. K. Lin, J. R. Falck, P. Y. Wong, Biochem. Biophys. Res. Commun. 1990,
167, 977–981.
[8] J. D. Imig, X. Zhao, C. Z. Zaharis, J. J. Olearczyk, D. M. Pollock, J. W.
Newman, I. Kim, T. Watanabe, B. D. Hammock, Hypertension 2005, 46,
975–981.
[9] K. G. Maier, R. J. Roman, Curr. Opin. Nephrol. Hypertens. 2001, 10, 81–87.
[10] J. D. Imig, B. D. Hammock, Nat. Rev. Drug Discovery 2009, 8, 794–805.
[11] R. J. Roman, Physiol. Rev. 2002, 82, 131–185.
[12] W. B. Campbell, D. Gebremedhin, P. F. Pratt, D. R. Harder, Circ. Res. 1996,
78, 415–423.
[40] H. Y. Lo, C. C. Man, R. W. Fleck, N. A. Farrow, R. H. Ingraham, A. Kukulka,
J. R. Proudfoot, R. Betageri, T. Kirrane, U. Patel, R. Sharma, M. A. Hoer-
mann, A. Kabcenell, S. D. Lombaert, Bioorg. Med. Chem. Lett. 2010, 20,
6379–6383.
[41] S. Pecic, S. Pakhomova, M. E. Newcomer, C. Morisseau, B. D. Hammock,
Z. Zhu, A. Rinderspacher, S. Deng, Bioorg. Med. Chem. Lett. 2013, 23,
417–421.
[42] D. Panigrahy, M. L. Edin, C. R. Lee, S. Huang, D. R. Bielenberg, C. E. But-
terfield, C. M. Barnes, A. Mammoto, T. Mammoto, A. Luria, O. Benny,
D. M. Chaponis, A. C. Dudley, E. R. Greene, J. A. Vergilio, G. Pietramag-
giori, S. S. Scherer-Pietramaggiori, S. M. Short, M. Seth, F. B. Lih, K. B.
Tomer, J. Yang, R. A. Schwendener, B. D. Hammock, J. R. Falck, V. L. Man-
thati, D. E. Ingber, A. Kaipainen, P. A. D’Amore, M. W. Kieran, D. C. Zeldin,
J. Clin. Invest. 2012, 122, 178–191.
[13] K. G. Proctor, J. R. Falck, J. Capdevila, Circ. Res. 1987, 60, 50–59.
[14] D. Gebremedhin, Y. H. Ma, J. R. Falck, R. J. Roman, M. VanRollins, D. R.
Harder, Am. J. Physiol. 1992, 263, H519–H525.
[15] Y. X. Wang, L. N. Zhang, G. M. Rubanyi, (Arete Therapeutics, Inc., San
Francisco, USA), PCT Int. Pat. Appl. PCT/US2009/053863
(WO 2010025043 A1), 2010.
[16] “Evaluation of Soluble Epoxide Hydrolase (s-EH) Inhibitor in Patients with
Mild to Moderate Hypertension and Impaired Glucose Tolerance”, clinical
trial #NCT00847899 (Sponsor: Arete Therapeutics, Inc.); https://clinical-
trials.gov/ct2/show/NCT00847899.
[17] C. Morisseau, M. H. Goodrow, D. Dowdy, J. Zheng, J. F. Greene, J. R. San-
born, B. D. Hammock, Proc. Natl. Acad. Sci. USA 1999, 96, 8849–8854.
[18] I. H. Kim, F. R. Heirtzler, C. Morisseau, K. Nishi, H. J. Tsai, B. D. Hammock,
J. Med. Chem. 2005, 48, 3621–3629.
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
507
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[43] D. Moser, J. M. Wisniewska, S. Hahn, J. Achenbach, E. Buscató, F. Klingler,
B. Hofmann, D. Steinhilber, E. Proschak, ACS Med. Chem. Lett. 2012, 3,
155–158.
[44] Permute (version 0.8-4), proprietary software, AstraZeneca.
[45] Molecular Networks GmbH, Erlangen, Germany; www.molecular-net
works.com.
[46] Omega version 2.4.3, OpenEye Scientific Software, Santa Fe, USA;
www.eyesopen.com/omega; P. C. D. Hawkins, A. G. Skillman, G. L.
Warren, B. A. Ellingson, M. T. Stahl, J. Chem. Inf. Model. 2010, 50, 572–
584.
[49] Schrçdinger Release 2014-1: Maestro, Version 9.7, Schrçdinger LLC, New
York, NY, 2014.
[50] G. Jones, P. Willett, R. C. Glen, A. R. Leach, R. Taylor, J. Mol. Biol. 1997,
267, 727–748.
[51] T. L. Hwang, A. J. Shaka, J. Magn. Reson. Ser. A 1995, 112, 275–279.
[52] J. Pujol, Geophysics 2007, 72, W1–W16.
[53] D. Warren, The PyMOL Molecular Graphics System, Version 1.3, Schrç-
dinger LLC.
[47] J. Sadowski, J. Bostrom, J. Chem. Inf. Model. 2006, 46, 2305–2309.
[48] FRED version 2.2.5, OpenEye Scientific Software, Santa Fe, NM (USA);
www.eyesopen.com/oedocking.
Received: December 11, 2015
Published online on February 4, 2016
ChemMedChem 2016, 11, 497 – 508
www.chemmedchem.org
508
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim