Characterization of crystal water molecules in a high-affinity inhibitor and hematopoietic prostaglandin D synthase complex by interaction energy studies

Article abstract of DOI:10.1016/j.bmc.2018.08.014

Hematopoietic prostaglandin D synthase (H-PGDS) is one of the two enzymes that catalyze prostaglandin D₂ synthesis and a potential therapeutic target of allergic and inflammatory responses. To reveal key molecular interactions between a high-affinity ligand and H-PGDS, we designed and synthesized a potent new inhibitor (K_D: 0.14 nM), determined the crystal structure in complex with human H-PGDS, and quantitatively analyzed the ligand–protein interactions by the fragment molecular orbital calculation method. In the cavity, 10 water molecules were identified, and the interaction energy calculation indicated their stable binding to the surface amino acids in the cavity. Among them, 6 water molecules locating from the deep inner cavity to the peripheral part of the cavity contributed directly to the ligand binding by forming hydrogen bonding interactions. Arg12, Gly13, Gln36, Asp96, Trp104, Lys112 and an essential co-factor glutathione also had strong interactions with the ligand. A strong repulsive interaction between Leu199 and the ligand was canceled out by forming a hydrogen bonding network with the adjacent conserved water molecule. Our quantitative studies including crystal water molecules explained that compounds with an elongated backbone structure to fit from the deep inner cavity to the peripheral part of the cavity would have strong affinity to human H-PGDS.

Full text of DOI:10.1016/j.bmc.2018.08.014

Accepted Manuscript
Characterization of crystal water molecules in a high-affinity inhibitor and hem-
atopoietic prostaglandin D synthase complex by interaction energy studies
Daisuke Takaya, Koji Inaka, Akifumi Omura, Kenji Takenuki, Masashi
Kawanishi, Yukako Yabuki, Yukari Nakagawa, Keiko Tsuganezawa, Naoko
Ogawa, Chiduru Watanabe, Teruki Honma, Kosuke Aritake, Yoshihiro Urade,
Mikako Shirouzu, Akiko Tanaka
PII:
S0968-0896(18)31129-5
https://doi.org/10.1016/j.bmc.2018.08.014
BMC 14501
DOI:
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
10 July 2018
7 August 2018
10 August 2018
Please cite this article as: Takaya, D., Inaka, K., Omura, A., Takenuki, K., Kawanishi, M., Yabuki, Y., Nakagawa,
Y., Tsuganezawa, K., Ogawa, N., Watanabe, C., Honma, T., Aritake, K., Urade, Y., Shirouzu, M., Tanaka, A.,
Characterization of crystal water molecules in a high-affinity inhibitor and hematopoietic prostaglandin D synthase
complex by interaction energy studies, Bioorganic & Medicinal Chemistry (2018), doi: https://doi.org/10.1016/
j.bmc.2018.08.014
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Characterization of crystal water molecules in a high-affinity inhibitor and
hematopoietic prostaglandin D synthase complex by interaction energy studies
Daisuke Takaya*,1, Koji Inaka*,2, Akifumi Omura*,3, Kenji Takenuki3, Masashi Kawanishi3,
Yukako Yabuki1, Yukari Nakagawa1, Keiko Tsuganezawa1, Naoko Ogawa1, Chiduru Watanabe1,
Teruki Honma1, Kosuke Aritake4,5, Yoshihiro Urade4, Mikako Shirouzu1, and Akiko Tanaka1**
1RIKEN Center for Biosystems Dynamics Research, 1-7-22 Suehiro, Tsurumi, Yokohama
230-0045, Japan, 2MARUWA Foods and Biosciences, Inc. 170-1 Tsutsui-cho, Yamatokoriyama,
Nara 639-1123, Japan, 3Asahi Kasei Pharma Corporation, 632-1 Mifuku, Izunokuni, Shizuoka
4
410-2321, Japan, International Institute for Integrative Sleep Medicine (WPI-IIIS), University
of Tsukuba, Tsukuba, Ibaraki 305-8575, Japan, 5Daiichi University of Pharmacy, 22-1
Tamagawa, Minami, Fukuoka 815-8511, Japan.
*These authors contributed equally to this work.
**Correspondence should be sent to: Akiko Tanaka, Ph.D., RIKEN Center for Biosystems
Dynamics Research, 1-7-22 Suehiro-cho, Tsurumi, Yokohama 230-0045, Japan.
E-mail: aktanaka@riken.jp, Phone: +81-45-503-9451, Fax: +81-45-503-9450
1

Hematopoietic prostaglandin D synthase (H-PGDS) is one of the two enzymes that catalyze
prostaglandin D2 synthesis and a potential therapeutic target of allergic and inflammatory
responses. To reveal key molecular interactions between a high-affinity ligand and H-PGDS, we
designed and synthesized a potent new inhibitor (KD: 0.14 nM), determined the crystal
structure in complex with human H-PGDS, and quantitatively analyzed the ligand-protein
interactions by the fragment molecular orbital calculation method. In the cavity, 10 water
molecules were identified, and the interaction energy calculation indicated their stable binding
to the surface amino acids in the cavity. Among them, 6 water molecules locating from the deep
inner cavity to the peripheral part of the cavity contributed directly to the ligand binding by
forming hydrogen bonding interactions. Arg12, Gly13, Gln36, Asp96, Trp104, Lys112 and an
essential co-factor glutathione also had strong interactions with the ligand. A strong repulsive
interaction between Leu199 and the ligand was canceled out by forming a hydrogen bonding
network with the adjacent conserved water molecule. Our quantitative studies including crystal
water molecules explained that compounds with an elongated backbone structure to fit from the
deep inner cavity to the peripheral part of the cavity would have strong affinity to human
H-PGDS.
Key words: Hematopoietic prostaglandin D synthase; crystal water molecule; interaction
energy; fragment molecular orbital method; crystal structure analysis; drug design
2

1 Introduction
Prostaglandin (PG) D2, produced from arachidonic acid, is important in the pathogenesis of
inflammatory diseases1, 2 and regulation of physiological sleep.3, 4 Hematopoietic prostaglandin
D synthase (H-PGDS) is one of the two enzymes that catalyze the isomerization of PGH2 to
PGD2. Inhibition and knockout studies of H-PGDS have indicated its involvement in allergic
and inflammatory responses, and thus human H-PGDS is a potential therapeutic target.1, 2, 5
In 2003, Inoue et al. reported the crystal structures of human H-PGDS containing Mg2+ and
glutathione (GSH), an essential co-factor for the enzyme.6 The asymmetric unit of the crystal
lattice contained two homodimers of H-PGDS. The monomer of H-PGDS contains a large
catalytic cavity sandwiched between the N-terminal (residues 2-71) and C-terminal (residues
82-199) domains.6 The catalytic cavity of H-PGDS is subdivided into three regions: the inner
cavity surrounded by Arg14, Met99, Tyr152, and Asp96, the central cavity surrounded by
Trp104, and the thioate anion of GSH, and the peripheral solvent-exposed part of the pocket.7, 8
In 2004, the first crystal structure of human H-PGDS in complex with an inhibitor was
reported.9 Since then, many inhibitors and candidate drugs against H-PGDS have been
developed and reported by pharmaceutical companies.1, 8, 10
In high-affinity inhibitor design, the decision of whether to engage or displace conserved
water molecules is an important subject. Thus, the conserved water molecule near Leu199 in the
catalytic cavity of human H-PGDS has been drawing attention.8, 10-12 Thorarensen’s group at
3

Pfizer reported that the complex structures of H-PGDS with nanomolar potency inhibitors all
showed the hydrogen bonding between the conserved water molecule and ligands. They tried to
displace the water molecule by their new ligands to enhance the binding affinity due to the
entropic gain; however, they concluded that the maximal affinity for H-PGDS required ligands
to form a hydrogen bonding with the water molecule.13 Therefore, we aimed to identify all the
crystal water molecules binding stably in the H-PGDS catalytic cavity and quantify their
contributions to ligand binding by the fragment molecular orbital (FMO) calculation method.14
This information will be valuable for future drug-design of high-affinity H-PGDS inhibitor
molecules.
The interactions important in the ligand-protein binding are hydrophobic and electrostatic
interactions, and hydrogen bonding. To understand these interactions the quantum mechanics
(QM) calculation method is one of powerful approaches, because the method takes into
consideration the effects of donating and withdrawing electrons and π-π interactions. The FMO
method has been developed to apply the QM calculations to large biomolecules.14-20 The method
is achieved by dividing a large molecule into small pieces called fragments, such as amino acid
residues, water molecules, ions, and ligands. The FMO method evaluates the electronic states of
each fragment pair, and indicates their interaction magnitude in individual fragment levels with
the Inter-Fragment Interaction Energy (IFIE).21 Further, the Pair Interaction Energy
4

Decomposition Analysis (PIEDA) indicates the energy components of IFIE: electrostatic
interaction (ES), dispersion interaction (DI), charge transfer with higher-order mixed terms
energies (CT+mix), and exchange-repulsion (EX).20, 22, 23 The ES component mainly reflects the
hydrogen bonding and Coulomb interaction energies, and the DI component reflects the CH-π
and π-π interaction energies. Thus ES and DI components are important for protein-ligand
interactions.
In this paper, we designed and synthesized a high-affinity ligand for human H-PGDS, and
then their complex crystal structure was determined. Based on the structure, FMO calculations
were performed to quantify the ligand-protein interactions in fragment levels to reveal all the
stably bound water molecules in the cavity and their contributions to the ligand binding.
2 Materials and methods
2.1 Synthesis of F092
2.1.1 General
First, 4-(2-oxopyrrolidin-1-yl)benzoic acid (1) was converted to the N-Boc aniline (2) via a
Curtis rearrangement reaction under conventional conditions, and then the deprotection reaction
of the Boc group was accomplished using HCl in 1,4-dioxane to produce
1-(4-aminophenyl)pyrrolidin-2-one (3). Subsequently, the amidation reaction of 3 and
5

2-(pyridin-2-yl)pyrimidine-5-carboxylic acid was performed by the addition of
propylphosphonic anhydride (T3P®), to provide F092 (4) in good yield.
Progress of all reactions was monitored on Merck pre-coated silica gel plates using ethyl
acetate (EtOAc)/hexane as a solvent system. Spots were visualized by irradiation with
ultraviolet light (254 nm). Column chromatography was performed using Yamazen silica gel 60
(230 − 400 mesh). Proton (1H) NMR spectra were recorded on a Bruker Avance 400 instrument
using tetramethylsilane as an internal standard. Chemical shifts are given in parts per million
1
(ppm) (δ relative to the residual solvent peak for H). The following abbreviations are used:
singlet (s), doublet (d), triplet (t), and multiplet (m). ESI mass spectrometry was performed by
UPLC/MS (Waters).
2.1.2 tert-butyl (4-(2-oxopyrrolidin-1-yl)phenyl)carbamate (2).
Diphenyl phosphorazidate (DPPA) (0.251 ml, 1.2 eq.) was added dropwise to a solution of
4-(2-oxopyrrolidin-1-yl)benzoic acid (1) (201 mg, 1 eq.), tert-BuOH (0.93 ml, 10 eq.) and
trimethylamine (Et3N) (0.272 ml, 2.0 eq.) in toluene (4.8 ml) at room temperature (rt). The
resulting pale yellow solution was stirred at 120 ºC for 16.5 h under N2 atmosphere. The
resulting brown mixture was cooled to rt and saturated NaHCO3(aq) (5 ml) was added to the
mixture. The mixture was extracted twice with EtOAc (10 ml), and the combined organic
extracts were washed with brine (5 ml), dried over MgSO4, and concentrated under reduced
6

pressure. Purification of the crude product by flash chromatography on silica eluted with
hexian-EtOAc (50:50, 30:70 and 20:80) gave 2 (183 mg, 68%) as a white solid; RF (33:67
hexane-EtOAc) 0.45; 1H NMR (400 MHz, DMSO-d6) δ 7.51 (dd, J = 2.4, 2.0 Hz, 2H, Ph), 7.42
(d, J = 4.4 Hz, 2H, Ph), 3.78 (t, J = 7.2 Hz, 2H, NCH2), 2.45 (t, J = 7.2 Hz, 2H, CH2CO), 2.04
+
(tt, J = 7.2, 7.2 Hz, 2H, CH2), 1.47 (s, 9H, tert-Bu); ESI-MS m/z calculated for C15H20N2O3 H+
[M + H]+; 277.2. Found: 277.3.
2.1.3 1-(4-aminophenyl)pyrrolidin-2-one (3).
To a solution of 2 (183 mg, 1.0 eq.) in 1,4-dioxane (0.67 ml) was added 4 M HCl in
1,4-dioxane (3.3 ml) at rt. The resulting solution was stirred at rt for 3 h to precipitate. Obtained
precipitate was filtered, washed three times with 1,4-dioxane (5 ml), and dried under reduced
1
pressure at 40 ºC for 18 h to give 3 (116 mg, 83%) as a white solid; H NMR (400 MHz,
DMSO-d6) δ 9.99 (brs, 2H, NH2), 7.73 (d, J = 8.8 Hz, 2H, Ph), 7.33 (d, J = 8.8 Hz, 2H, Ph),
3.83 (t, J = 7.2 Hz, 2H, NCH2), 2.50 (t, J = 7.2 Hz, 2H, CH2CO), 2.07 (tt, J = 7.2, 7.2 Hz, 2H,
CH2); ESI-MS m/z calculated for C10H12N2O+ H+ [M + H]+; 177.1. Found: 177.3.
2.1.4 N-(4-(2-oxopyrrolidin-1-yl)phenyl)-2-(pyridin-2-yl)pyrimidine-5-carboxamide (4).
A solution of propylphosphonic anhydride in EtOAc (50 wt.%, 6.45 ml, 2 eq.) was added
to a solution of 3 (1.07 g, 1.1 eq.), 2-(pyridin-2-yl)pyrimidine-5-carboxylic acid (920 mg, 1.0
eq.) and Et3N (5.10 ml, 8.0 eq.) in CH2Cl2 (45.7 ml) at rt. The resulting solution was stirred at rt
7

for 1 h to cause precipitation, which was collected by filtration and washed with EtOAc (2 x 10
ml). The obtained solid was dried under reduced pressure at 40 ºC for 18 h to give 4 (840 mg,
51%) as a white solid. Saturated NaHCO3(aq) was added to the filtrate, then the mixture was
stirred at rt for 30 min to give a precipitate, which was collected by filtration and dried to give 4
1
(597 mg, 36%) as a white solid; H NMR (400 MHz, DMSO-d6) δ 10.67 (s, 1H, NH), 9.43 (s,
2H, NCH), 8.80 (d, J = 3.6 Hz, 1H, NCH), 8.47 (d, J = 8.0 Hz, 1H, NCH), 8.03 (t, J = 8.0 Hz,
1H, CH), 7.78 (d, J = 9.2 Hz, 2H, Ph), 7.69 (d, J = 9.2 Hz, 2H, Ph), 7.59 (dd, J = 3.6, 8.0 Hz,
1H, CH), 3.85 (t, J = 7.2 Hz, 2H, NCH2), 2.07 (tt, J = 7.2, 7.2 Hz, 2H, CH2); ESI-MS m/z
+
calculated for C20H18N5O2 H+ [M + H]+: 360.1. Found: 360.3.
2.2 Purification of Histag H-PGDS
Human Histag H-PGDS was expressed and purified as described previously.11 Briefly, the
DNA fragment of the full-length H-PGDS gene (NCBI accession no.: NP 055300) fused with a
6x-Histidine tag at the N-terminus was cloned into the pET28a vector (Novagen, Madison, WI,
USA) and expressed in E. coli one shot BL21 (DE3) (Thermo Fisher). The cells were grown in
LB medium at 37 ºC, induced with 1.0 mM IPTG, and then cultured further for 4 hr at 37 ºC.
The cells were collected and disrupted by sonication in 50 mM phosphate-buffer (pH 6.6),
containing 2 mM MgCl2, 0.4 mg/ml lysozyme, and 6 μg/ml DNase/RNase. After removal of the
cell debris by centrifugation, the supernatant was filtered through a 0.45 μm PVDF filter and
8

applied to a GSH-Sepharose 4B column. After washing the column with 50 mM
phosphate-buffer (pH 6.6) containing 2 mM MgCl2, the protein was eluted with 50 mM
Tris-HCl, pH 9.0 containing 10 mM GSH. The eluted H-PGDS fractions were washed three
times by PBS(-) and concentrated using an Amicon Ultra filtration device. Protein
concentrations were determined with a Bradford protein assay kit (Bio-Rad).
2.3 Evaluation of H-PGDS inhibitors by Surface Plasmon Resonance (SPR)
SPR spectroscopy measurement was performed at 25 ºC on a Biacore T-200 instrument (GE
Healthcare Life Science). An anti-histidine antibody (His Capture kit, GE Healthcare Life Science )
was immobilized at approximate 14,500 RU onto the sensor chip CM-5 (GE Healthcare Life
Science) using the Amine Coupling Kit (GE Healthcare Life Science) according to the protocol
supplied by the manufacturer. Then, Histag H-PGDS was injected and captured on the chip at
approximate 900 RU using PBS(-) as the running buffer. Running buffer was replaced with an assay
buffer containing 20 mM HEPES (pH7.4), 150 mM NaCl, 2 mM MgCl2, 2 mM GSH, 1% DMSO,
and 0.005% Tween20. Test compounds were 2-fold serially diluted in the assay buffer and tested for
binding to the captured Histag H-PGDS at a flow rate of 30 l/min. The raw data (n=3) were
processed by BIA evaluation Software (GE Healthcare Life Science). TFC-007, prepared according
to JP2007-51121A, was purchased from HanChem Co., Ltd. (Daejeon, Korea) and used as a
high-affinity control compound.
9

2.4 Additional purification and crystallization of human H-PGDS
Preparation of the human H-PGDS protein for crystallization was performed as reported.24
Additional purification of the protein was performed as before to obtain high quality crystals, by
chromatography on a SuperQ-5PW column (TOHSO). The eluted protein was concentrated to 5
mg/ml with an Amicon Ultra filter (Merck) equipped with YM-10 membrane. The sample thus
obtained was washed with 20 mM Tris-HCl (pH 8.0) buffer.
Crystallization of the complex with F092 was performed by the counter diffusion method
to avoid forming cluster crystals, using a Crystal Tube Kit (MB2004-CRT200, Confocal Science
Inc., Tokyo, Japan).25 An 8 l aliquot of the protein solution, containing 2.5 mg/ml H-PGDS
protein, 50 mM Tris-HCl (pH 8.5), 17.5% (w/v) PEG6000, 1 mM DTT, 1 mM GSH, 1 mM
MgCl2, and 1% dioxane was loaded into a glass capillary equipped with a 6 mm length gel-tube.
The assembled capillary was placed into 1 ml of the reservoir solution, containing 50 mM
Tris-HCl (pH 8.5), 35% (w/v) PEG6000, 2 mM DTT, 2 mM GSH, 2 mM MgCl2, and 2%
dioxane for several days. Single crystals of H-PGDS in complex with F092 were successfully
obtained.
2.5 Data collection and structure determination
Diffraction data of the H-PGDS crystals in the complex with F092 were collected at 100 K
on the i04 beam line at Diamond Light Source (Didcot, UK) using a camera system of
10

PILATUS6M. The crystals grown in the capillaries were extracted into the reservoir solution,
soaked in a cryo-protectant solution containing 15% glycerol, and flash-cooled in a stream of
nitrogen gas at 100 K prior to data collection. The diffraction images were integrated and scaled
by iMOSFLM26 and Aimless,27 from the CCP4 suite.28
The complex structure was determined by the molecular replacement method with
MOLREP29 by using the native structure as the search model (PDB ID : 1IYH). The model
re-building and the refinement of the structure were performed with Coot30 and REFMAC5.31 In
this report, electron density peaks, with the threshold cutoff of 1.0 and 3.0 in the 2Fo-Fc
Fourier and the Fo-Fc omit-map, respectively, were picked up as candidates of water molecules.
Then the candidates with appropriate hydrogen bond interactions with surrounding O and N
atoms were selected and assigned as water molecules. Statistics of data collection and structure
refinement are summarized in Table 2. The coordinates and structure factors have been
deposited in the PDB, with the PDB ID: 5YWX.
2.7 QM calculations
The coordination sets of H-PGDS homodimers (residues 2-199 in chains A and B)
deposited to the PDB as the F092 complex (5YWX) and the apo form (1IYH) were employed
for the FMO calculations. Crystal water molecules within 4.5 Å from any receptor atoms were
+
retained. N- and C-termini of the protein are in the form of -NH3 and -COO-, respectively.
11

Hydrogen atoms were added and concurrently optimized, using the Amber10: EHT force field
implemented in MOE [Molecular Operating Environment (MOE), 2016.08; Chemical
Computing Group Inc., QC, Canada, 2016]. In the structure optimization process based on
molecular mechanics, all heavy atoms were fixed at the X-ray structure coordinates. We then
performed all FMO calculations14, 21, 32 at the second-order Møller-Plesset perturbation theory
(MP2) level with a 6-31G* basis set, using the ABINIT-MP program [MIZUHO/BioStation
Viewer, version 3.0; Mizuho Information and Research Institute, Inc., Tokyo, Japan, 2013]. In
our FMO calculations, each single amino acid unit containing side chain, C, and N-terminal
side amide group, was assigned to a fragment. This fragmentation procedure is a default
behavior of ABINIT-MP for accurate energy calculations. In addition, GSH, F092, and the Mg2+
ion were treated as single fragments. IFIE was used to reveal their energy contributions, and
PIEDA22, 23 was further used to analyze the each energy component of IFIE: electrostatic (ES),
exchange-repulsion (EX), charge transfer with higher order mixed terms energies (CT+mix),
and dispersion interaction (DI).
~
~
~
~
~
ES
IJ
EX
IJ
CTmix
IJ
DI.
IJ
EIJ  E  E  E
 E .
Scheme 1
~
In this equation, EIJ is the IFIE between I-th and J-th fragments and each item of addition is
the energy component. Attractive and repulsive interactions are showed negative and positive
12

values for IFIE, respectively. Interactions of fragments in the chain A were used for the IFIE and
PIEDA studies.
3 Results and Discussion
3.1 Design and synthesis of F092
Carron et al.10 and Trujillo et al.13 pointed out that the strong affinity for H-PGDS required
the inhibitors to form a hydrogen bond with the conserved water molecule near Leu199. Thus,
we aimed to design and synthesize a new compound that shares the typical substructures of
H-PGDS inhibitors, a phenyl group and a heterocyclic group, and also interacts with the
conserved water molecule. Referring to the substructures reported by Sanofi-Aventis,
Astra-Zeneca, and Pfizer,13 F092 was empirically designed to have an elongated backbone
structure to fit the deep catalytic cavity (Fig. 1). We expected that the back bone structure of the
pyrimidine next to the pyridine would be important for F092 to obtain strong interactions with
the conserved water molecule by forming two hydrogen bondings between the O atom of the
water and the two nitrogen atoms of the rings.
13

The synthesis scheme of F092 is shown in Fig. 1. The Curtius rearrangement of
commercially available 4-(2-oxopyrrolidin-1-yl)benzoic acid (1), followed by HCl-mediated
deprotection of the N-Boc, afforded 1-(4-aminophenyl)pyrrolidin-2-one (3). The aniline, 3, was
coupled with 2-(pyridin-2-yl)pyrimidine-5-carboxylic acid in the presence of propylphosphonic
anhydride (T3P®) to produce F092 (4) in good yield.
As we expected, F092 had a high-affinity for human H-PGDS. The dissociation constant
(KD) of F092 from the purified human Histag H-PGDS determined by SPR was 0.14 nM +
0.010 nM (Table 1). It was comparable to that of the known clinical candidate compound,
TFC-0071 (KD = 0.37 + 0.017 nM). We further confirmed that F092 strongly inhibited the
enzyme activity of the purified H-PGDS (Aritake et al. unpublished results). Therefore, we
studied the H-PGDS crystal structure in complex with F092 to analyze the interactions between
ligand, protein, and water molecules.
14

Table 1 List of compounds discussed in this paper.
Compound
Structure
KD (nM)*
O
N
O
N
TFC-007
F092
0.37 ± 0.017
0.14 ± 0.010
O
N
N
N
O
N
N
N
N
O
O
N
*KD values were determined from the results of triplicate SPR measurements.
3.2 Structure analysis of H-PGDS in complex with F092
The human H-PGDS crystals in complex with F092 diffracted to 1.74 Å (PDB ID: 5YWX,
data statistics are shown in Table 2). F092 was located in the catalytic cavity and surrounded by
Arg12, Arg14, and Thr159 in the inner cavity, Trp104 and GSH in the central cavity, and Trp39
and Ala105 in the peripheral region (Fig. 2). F092 interacted hydrophobically with Arg14 and
Met99 in the inner cavity and Phe9 in the peripheral region. The pyrimidine ring of F092
formed typical π-π interactions with the indole side-chain of Trp104, because the pyrimidine
ring was parallel to the indole ring (Fig. 2).
15

Fig. 2 Interactions between F092 and H-PGDS.
Views from two different directions are drawn schematically. Stick models show the refined model of
F092 and GSH in the F092 complex (PDB ID: 5YWX, chain A). Blue meshes indicate the difference
electron density of an Fo-Fc omit map of the bound F092 calculated at 3.0. Five blue spheres show
the conserved water molecules (WaterA-WaterE) in the catalytic cavity. Their residue numbers shown
in the PDB data (5YWX) are as follows: WaterA: HOHA442, WaterB: HOHA515, WaterC: HOHA464,
WaterD: HOHA518, WaterE: HOHA482. Other five yellow spheres show the specific water molecules
in the chain A of the F092 complex (HOHA408, HOHA421, HOHA462, HOHA503 and HOHA578).
Table 2. Data processing and refinement statistics.
H-PGDS-F092 complex
Data collection
Space Group
P1
Cell constants (Å)
a=47.0, b=48.2, c=91.0
=99.0, =92.3,=90.0
No. of molecules in ASU
No. of crystals
4
1
Resolution Range (Å)
No. of independent reflections
Completeness (%)
Rmerge ( I, %)
40.0- 1.74
79,469
97.1 (95.6)
0.088 (0.523)
0.42
Average Mosaicity
I/(I) (%)
5.9 (1.9)
Refinement
Non-hydrogen atoms
Protein
6552
16

Compound
108
583
Solvent (H2O)
Rwork value, Rfree value
0.203, 0.253
0.011
RMSD bond length (Å2)
RMSD bond angles (ᵒ)
Mean B (protein) (Å2)
1.47
Main Chain
Side Chain
22.1
21.6
Mean B (solvent) (Å2)
23.5
Mean B (compound) (Å2)
22.4
Chiral center
0.081
0.008
0.246
5YWX
Planar groups
VDW repulsions
PDB ID
In the catalytic cavity of chain A of the F092 complex, total 10 water molecules were
located within 6.0 Å from their interacting atoms of F092. Among them, five water molecules in
the inner cavity were observed in all four chains contained in the crystal unit of the F092
complex. Thus, the five stable water molecules were named WaterA~WaterE. They are shown as
blue spheres in Fig. 2. In addition to the stable water molecules, five specific water molecules,
HOHA462, HOHA578, HOHA408, HOHA503, and HOHA421 were observed in the structure
of chain A (shown as yellow spheres in Fig. 2). In the F092 complex structure, the averaged
B-factor of the 5 stable water molecules (WaterA ~ WaterE) was 9.0 Å2. This value was much
lower than that of the C atoms (20.7 Å 2) of the protein molecule. Contrary, the other water
molecules observed in the central and peripheral regions (yellow spheres) had much higher
17

B-factors. Their averaged B-factor was 33.5 Å 2. This difference suggested that the five water
molecules in the inner cavity of H-PGDS were fixed tightly to the protein surface.
Overlay of the crystal water molecules and interacting amino acids in the apo form (PDB
ID: 1IYH) and the complex forms of H-PGDS (PDB IDs: 5YWX and 3KXO) are shown in Fig.
3. All chains of the reported apo form structure in common contained the five stable water
molecules (shown as magenta spheres in Fig. 3). They were located at the same coordinates
corresponding to the five stable water molecules (WaterA ~ WaterE) in the F092 complex (shown
as green spheres in Fig. 3). Thus, these water molecules (WaterA ~ WaterE) are considerable as
conserved water molecules. The conserved water molecules were also observed at the same
positions in the crystals of the KXO complex10 (shown as green spheres in Fig. 3). Fig. 3 also
shows that WaterA corresponds to “the primary water molecule near Leu199”, and WaterB
corresponds to “the second water molecule” reported before.10, 13
18

Fig. 3 Overlay of the crystal water molecules and interacting amino acids in the apo form and the
complex forms of H-PGDS.
Views from two different directions are drawn schematically. Green sticks show the refined model
of F092 surrounded by amino acid residues of the catalytic cavity. Lines and spheres colored with
green, magenta and cyan show superposed structures of the F092 complex (5YWX, chain A), the
apo form (1IYH, chain A) and the KXO complex (3KXO, chain A), respectively. Black dashed lines
show hydrogen bonding interactions between these water molecules and the amino acid residues of
the F092 complex.
F092 formed four hydrogen bonding networks containing WaterA-WaterB-WaterC,
HOHA462-HOHA578, HOHA408-HOHA503, and HOHA421 (Figs. 2 and 3). In detail, WaterA
interacted with the carbonyl O atom of Leu199 and the O atom of Thr159, and WaterB formed a
hydrogen bonding network with WaterA. WaterC interacted with the N atom of Arg12.
HOHA462 interacted with the N1 atom of Arg14, and HOHA578 interacted with the Sγ atom
of the GSH molecule. HOHA408 interacted with Nζ atom of Lys 112, and HOHA503 interacted
directly with the carbonyl O21 atom of F092. HOHA421 interacted with the carboxyl O atom of
Ala105 and the carbonyl O27 atom of F092. WaterA formed two hydrogen bonding interactions
with the pyridine and pyrimidine N atoms of F092, at the distance of 2.88 Å and 2.92Å,
respectively. The strong affinity of F092 for H-PGDS (Table 1) may be due to these direct
hydrogen bonding interactions.
19

3.3 Analysis of interaction energies between F092, crystal waters and H-PGDS
3.3.1 Interactions between F092 and H-PGDS
The FMO calculations indicated that the lowest IFIE between F092 and H-PGDS
fragments was observed with the attractive interaction between F092 and cofactor GSH (-53.03
kcal/mol). Since GSH consists of three amino acid units (Gly-Cys-Asp), it is plausible that the
absolute value of the IFIE for F092 with GSH was larger than that with other fragments
consisted of a single amino acid unit. In addition the sulfur atom of Cys of GSH formed a
hydrogen bonding interaction with the ligand. The remarkably low IFIE indicated the key role
of GSH in ligand-binding.
The IFIE values of the interactions between F092 and amino acid residues that located
within 6.0 Å from the any atom of F092 are shown in Fig. 4. The IFIE values of interactions
formed with Arg12, Gly13, Gln36, Asp96, Trp104, Lys112, Thr159, and Leu199 had large
absolute values, as compared to those with the others. These interactions were important in
F092 binding. Here we were aware that the IFIE value of the interaction between Leu199 and
F092 was positive and especially high (10.09 kcal/mol). Thus, we wondered why F092 had a
high affinity for H-PGDS, despite the strong repulsion of Leu199 calculated.
20

12.5
7.5
2.5
-2.5
-7.5
-12.5
Fig. 4 IFIE values of the interactions between amino acid residues and ligands in the F092
complex.
The interactions between F092 and amino acid residues located within 6.0 Å from the any
atom of the ligand were selected and their IFIE values were calculated.
To clarify which types of interactions were involved in the interactions, the PIEDA
calculations were performed for the F092 complex. Figure 5 shows the distribution of the
PIEDA of the interactions shown in Fig. 4. The ES values (shown in cyan bars) of the
interactions of Asp96 and Lys112 with the ligand were both remarkable. The value of Asp96
was -9.73 kcal/mol, and that of Lys112 was -12.50 kcal/mol. Thus, interactions of these residues
with F092 are dominated by the electrostatic binding. In contrast, highly repulsive ES values
were observed with Arg14 (6.12 kcal/mol) and Leu199 (12.16 kcal/mol). These indicated that
Arg14 and Leu199 are involved in electrostatically repulsive interactions for the ligand binding.
21

15.0
10.0
5.0
ES
EX
CT+mix
DI
0.0
-5.0
-10.0
-15.0
Fig. 5 Distribution of the PIEDA calculated with the F092 complex structure.
The interactions between amino acid residues and F092 shown in Fig. 4 were analyzed by the
PIEDA method. The electrostatic interaction (ES), exchange-repulsion (EX), charge transfer
with higher-order mixed terms energies (CT+mix), and dispersion interaction (DI) are
represented by cyan, orange, green, and purple, respectively.
The interactions of F092 with Arg14, Met99, and Trp104 had large negative DI values
(shown in purple bars), which meant that their dispersion interactions contributed to the F092
binding. These residues were hydrophobically interacted as described in Section 3.2. The
PIEDA analysis quantitatively indicated the remarkable aromatic interactions that contributed to
the ligand binding to H-PGDS. Summary of these interaction energy studies are represented in
Fig. 6.
22

Fig. 6 Visualization of interaction energies of F092 with amino acid residues, water
molecules, and GSH in the H-PGDS complex.
Figures in the top row show the summary of IFIE analysis for F092 fragment (shown in
yellow). The fragments with attractive and repulsive interactions are represented by red
and blue, respectively. Figures in the bottom row show the results of the PIEDA methods
for F092 fragment (shown in yellow), the main components of the stabilizing interactions
of fragments are represented by the following color scheme: ES, red and blue; EX, white
and pink; CT+mix, light blue and white; DI, green and white. CH-p and p-p interactions
are indicated by purple and orange dot lines, respectively, where they were analyzed using
the CHPI program.33
3.3.2 Significant interactions between water molecules and H-PGDS
The IFIE values between each water molecule and all amino acid residues in the apo form
(1IYH, chain A) and the F092 complex form (5YWX, chain A) were calculated by the FMO
method. Amino acid residues significantly interacting with the each crystal water molecule are
listed in Table 3. Significant interactions discussed in this paper were set to those with an
23

absolute IFIE values greater than or equal to 3.0 kcal/mol, which was an empirically threshold
in FMO study reported before.33 The lists of amino acids significantly interacting with the
conserved water molecules were almost identical between the two coordinates except Arg14,
whose side chain was rotated in the direction of WaterA and WaterE only in the F092 structure
(Fig. 3).
Table 3. Significant interactions between the water molecules and amino acids
in the catalytic cavities.
interacting amino acid*
water
1IYH
5YWX
Leu199, Thr159, Arg14
Arg14, Arg12
WaterA
WaterB
Leu199
Arg14, Arg12, Leu199
WaterC
Arg12, Cys156, Arg194
Arg12, Cys156, Arg194
Arg14, Ile17
WaterD
Arg14, Ile17, Ile18
WaterE
Ser64, Tyr152, Asp96
Ser64, Tyr152, Asp96, Arg14
Lys112, Ala105
Glu106, Lys107
Arg14, Trp104, Asp96
Lys112
HOHA408
HOHA421
HOHA462
HOHA503
HOHA578
-
-
-
-
-
Arg14
* Amino acids interacting to the water molecules with |IFIE| > 3 kcal/mol.
Among them, water-amino acid interaction pairs with the strongest energies were selected
and listed with their IFIE values (Tables 4 and 5). In our fragmentation of FMO method, single
amino acid fragment contains its side chain, C, and N-terminal side amide group, so that the
FMO fragment unit is slightly different from the standard amino acid residue unit. The point
should be noted when a carbonyl group of main chain closely related to the inter-fragment
24

interactions. For example, Arg14 fragment includes the carbonyl group of the peptide bond
between Gly13 and Arg14, and Glu106 fragment includes that between Ala105 and Glu106.
These strong interactions were almost indicated as hydrogen bindings in Fig. 3.
Table 4. IFIE values of the strongest water-amino acid interactions
with each water molecule in the catalytic cavity of the apo form.
IFIE (kcal/mol)
aa-water
-21.60
water-amino acid interaction*
water
amino acid* (aa)
Leu199
Arg14
WaterA
WaterB
WaterC
WaterD
WaterE
-8.70
Arg12
-18.13
Arg14
-8.27
Ser64
-8.62
*Amino acid residue contributing to the strongest interaction with each water molecule.
Table 5. Related IFIE values of the strongest water-amino acid interactions with each water
molecule in the catalytic cavity of the F092 complex form.
water-amino acid interaction
IFIE (kcal/mol)
water
WaterA
amino acid* (aa)
Leu199
Arg14
aa-water water-F092 aa-F092 aa-water-F092**
-16.72
-9.68
-11.54
-6.75
-0.09
-0.65
0.76
10.09
2.07
-18.16
-14.35
-24.29
-7.24
WaterB
WaterC
Arg12
-16.30
-8.66
-7.91
2.07
WaterD
Arg14
WaterE
Ser64
-8.06
0.30
-7.01
HOHA408
HOHA421
HOHA462
HOHA503
HOHA578
Lys112
Glu106
Arg14
-14.37
-11.15
-12.00
-4.94
-4.67
-10.57
-4.42
-9.11
0.68
-12.50
2.60
-31.54
-19.13
-14.35
-26.55
-4.29
2.07
Lys112
Arg14
-12.50
2.07
-7.04
* Amino acid residue contributing to the strongest interaction with each water molecule.
**The IFIE of aa-water-F092 shows summation of the IFIEs between aa-water, water-F092, and
aa-F092.
25

The calculated IFIEs shown in the columns of “IFIE aa-water” confirmed that all of the
conserved water molecules in both structures maintained their strong interactions with the
surface amino acids of the H-PGDS catalytic cavity (< -8.06 kcal/mol), which was much lower
than the reference threshold33 of significant interactions. Especially, WaterA, and WaterC showed
remarkable stabilities.
Table 4 indicates that WaterA is highly stable due to the strong interaction with Leu199
(-21.60 kcal/mol). The negative charge of carboxyl moiety of Leu199 at the C-terminus of
H-PGDS might cause the strong interaction. And WaterA is adjacent to the ligand binding site,
thus it must be a key molecule to design high-affinity ligands. The calculated IFIE between
Leu199 and WaterA explained the results of Thorarensen’s group,13 which is “high affinity for
H-PGDS would require ligands to form a hydrogen bond with WaterA than to displace the water
molecule”.
As shown in the “IFIE aa-water” column of Table 5, specific five water molecules in the
central and peripheral part of the catalytic cavity of the F092 complex also stably bound to the
interacting amino acid residues (< -4.94 kcal/mol). FMO calculation indicated that HOHA408
and HOHA421 formed strong hydrogen bondings with Lys112 and Glu106 that existed in the
peripheral solvent-exposed part of the C-terminal domain, and HOHA462 formed hydrogen
bonding with Arg14 in the central cavity. The existence of the 5 additional stable water
26

molecules (yellow spheres in Fig. 2) in the central and peripheral part of the catalytic cavity and
their formation of stable hydrogen bonding networks (section 3.3.3) explained our empirical
results that compounds with an elongated backbone structure to fit from the deep inner cavity to
the peripheral part of the cavity would have strong affinity to H-PGDS.
3.3.3 Hydrogen bonding networks of the water molecules in the F092 complex
The interaction energies between the 10 water molecules and F092 were calculated and
shown in the “IFIE water-F092” column of Table 5. WaterA, WaterB, HOHA421 and HOHA503
had large attractive interactions with F092 (-11.54, -6.75, -10.57 and -9.11 kcal/mol,
respectively). HOHA408 and HOHA462 also had significant interactions with F092. It
indicated that these 6 crystal water molecules including ones locating in the central cavity and
the peripheral part of the cavity, directly contributed to the stable ligand binding.
Addition to these direct interactions between water molecules and F092, interactions of
hydrogen bonding networks including water molecules and F092 should be considered. The
summation of IFIE values for the network interactions are indicated in the “IFIE aa-water-F092”
column of Table 5, which are the sums of the three types of IFIE values, IFIEs between amino
acid and water, water and F092, and amino acid and F092.
Table 5 indicates that Leu199 formed a strongly attractive interaction with WaterA (-16.72
kcal/mol), and a strong repulsive interaction with F092 (10.09 kcal/mol) as pointed in the
27

section 3.3.1. But in the F092 complex, a hydrogen bonding network of Leu199-WaterA-F092
was formed, and its IFIE value (“IFIE aa-water-F092”) was -18.16 kcal/mol. These data
indicated that the stable WaterA in H-PGDS formed a hydrogen bonding network and balanced
the repulsive interaction between Leu199 and F092. This result reaffirmed that the stable water
molecule, WaterA is important for the robust ligand binding.
WaterB also worked for constructing a stable hydrogen bonding network of
Arg14-WaterB-F092 (-14.35 kcal/mol) in the F092 complex and overcoming the effects of
repulsion between Arg14 and the ligand (2.07 kcal/mol). Despite WaterC scarcely had an
interaction with the ligand (-0.09 kcal/mol for F092), the Arg12-WaterC-F092 network was
stable due to the two strongly attractive interactions between Arg12-WaterC and Arg12-F092
(Table 5).
Specific five water molecules in the F092 complex also formed bridged hydrogen bonding
networks with the interacting amino acid(s) and F092. IFIE values shown in the “aa-water-F092”
column of Table
5 indicate that Lys112-HOHA408-F092, Lys112-HOHA503-F092,
Glu106-HOHA421-F092, Arg14-HOHA462-F092, and Arg14-HOHA578-F092 networks
largely contribute F092 binding (-31.54, -26.55, -19.13, -14.35, and -4.29 kcal/mol,
respectively).
28

To clarify the complexity of HOHA462 and HOHA578 contribution to the F092 binding,
we performed IFIE analyses and PIEDA calculations of their interactions (Table 6). HOHA462
interacted with Arg14 (-12.00 kcal/mol), HOHA578 (-7.09 kcal/mol), and F092 (-4.42 kcal/mol).
HOHA578 interacted with GSH (-21.01 kcal/mol) and Arg14 (-7.04 kcal/mol). These
interaction energies were more than twice the reference threshold of IFIE except that between
HOHA578 and F092 (0.68 kcal/mol). From PIEDA analysis, the strong interaction among the
water molecules, Arg14, and GSH were mainly contributed by ES components, indicating that
they formed hydrogen bonding interactions. Consequently, in addition to the effect of the WaterB,
the hydrogen bonding network of F092-GSH-HOHA578-HOHA462-Arg14 that functioned as
one group having significant interaction energies also contributed to cancel out the repulsion
force between Arg14 and F092.
Table 6. Interactions with HOHA462 and HOHA578 in the F092 complex.
PIEDA (kcal/mol)
interacting DIST*
IFIE
water
fragment
Arg14
(Å)
(kcal/mol)
ES
EX
CT+mix
DI
1.84
2.10
2.12
2.10
2.21
2.26
2.70
-12.00
-7.09
-4.42
-7.09
-21.01
-7.04
0.68
-17.00
-7.45
-3.78
-7.45
-27.49
-7.33
1.94
10.85
2.90
4.02
2.90
16.06
3.58
0.16
-3.19
-1.35
-2.16
-1.35
-5.56
-1.48
-0.65
-2.66
-1.19
-2.51
-1.19
-4.01
-1.81
-0.76
HOHA462 HOHA578
F092
HOHA462
GSH
HOHA578
Arg14
F092
*“DIST” means minimum distance (Å) between the two fragments. Hydrogen atoms were
included in the distance calculation.
29