H. Mori et al. / Bioorg. Med. Chem. Lett. 24 (2014) 3732–3735
3733
Figure 2. (A) Chemical structure of one of hSR inhibitors identified by our virtual
screening protocol. (B) Stereoview of binding mode of 1 with hSR determined by our
docking procedure. Hydrogen-bonding interactions are indicated by red dashed
lines.
Figure 1. Multi-filter virtual screening.
comparable to that of malonate (1.31 mM) in our assay system
(Fig. 3). To our knowledge, this is a first report of hSR inhibitor that
possesses potent inhibitory activity comparable to malonate and a
quite different structure from malonate analogs.
applying this first filter, we found 9890 chemical compounds as
‘first hits’.
These ‘first hits’ were subsequently screened by high-through-
put protein structure-based virtual molecular docking. In this
molecular docking, we used the X-ray crystal structures of hSR in
complex with malonate (PDB: 3L6B and 3L6R).14 In addition, we
prepared ligand-free form of hSR by homology modeling method,
in which the X-ray crystal structure of ligand-free rat SR (rSR,
PDB: 3HMK)14 was used as a template. This modeling provided
two structurally different ligand-free active sites of hSR, as the X-
ray structure of ligand-free rSR was deposited as a dimer. As a
result, a total of four active sites of hSR (two malonate-bound
and two ligand-free forms) were used in this protein structure-
based virtual screening. Docking was done with HTVS mode in
GLIDE version 5.5 of Schrödinger Suite 2009. In order to determine
the criteria value of GLIDE-HTVS-Score for this screening, we first
docked five already-known hSR inhibitors,13,18 of which inhibition
constants (Ki) were reported (Supplementary Fig. 2). The resulting
scores for these five inhibitors ranged from ꢀ10.44 to ꢀ4.92. We
found that these GLIDE-HTVS-Scores were well correlated with
RTln(Ki) for these five inhibitors (Supplementary Fig. 3), indicating
that GLIDE-HTVS-Score is useful for extracting candidate com-
pounds for hSR inhibitors. We set the criteria of GLIDE-HTVS-Score
as ꢀ6.5, which is somewhat lower than that of the query molecule
used in the first filter (ꢀ6.18). This was intended to search candi-
date compounds that may possess more potent inhibitory activi-
ties than the query molecule. In this second screening, we
extracted 128 compounds as ‘second hits’.
In order to reduce the number of virtual hits to purchase, the
above second hits were finally clustered by similarity analysis based
on 2D structural fingerprint.19,20 This clustering was done with Can-
vas 1.2 of Schrödinger Suite 2009. As a result, we selected 19 com-
pounds as the cluster representative, of which 18 compounds
were commercially available and subjected to biological evaluation.
The identified 18 compounds were assayed with recombinant
hSR with C2DC6D mutations14 in vitro as described below. We
could finally identify four small molecules with inhibitory activity
against hSR from eighteen virtual hits, corresponding to a ‘hit rate’
of 22%. This high hit rate suggested that our multi-filter virtual
screening protocol was quite useful for exploring novel hSR
inhibitors. The chemical structure for one of these four compounds
(1) is given in Figure 2A.21 The IC50 value of 1 (1.21 mM) was
The interaction mode between 1 and hSR was examined by using
a combination method of the molecular-docking calculation and the
molecular mechanics Poisson–Boltzmann surface area (MM-PBSA)
free energy analysis,22 as detailed in the Supplementary data. In
order to test our procedure, we first applied it to the malonate–
hSR complex. The resulting model, that is the top-ranked pose, could
reproduce all interactions between malonate and hSR observed in
the crystal structure (Supplementary Fig. 4). The positional and con-
formational RMSDs relative to the X-ray pose were 0.58 and 0.20 Å,
respectively. These results suggest that our procedure is appropriate
for producing reliable interaction models of 1 and hSR. The resulting
interaction model (the top-ranked pose) of 1 with hSR obtained by
our procedure is shown in Figure 2B. The compound 1 was found
to bind to the ligand-free form of hSR and form a total of four
hydrogen bonds with N86, K114, and D238 of hSR. The 2,3,4-triflu-
orobenzene of 1 was located in the sub-pocket formed by I236,
P284, V317, S321, and S322 of hSR to make hydrophobic interac-
tions. The 4-chlorobenzene of 1 was found to make van der Waals
interactions with PLP-K56, S83, S84, G85, N86, H87, P153, G239,
and S242 of hSR, which were residues consisting of malonate-
binding site in the crystal structure of malonate–hSR complex.
We tried to design some derivatives of 1 that were expected to be
more potent for hSR. As shown in the left side of Figure 4, the chlo-
rine atom of the 4-chlorobenzene of 1 was quite close to the hydro-
xyl group of S242 of hSR. This motivated us to consider the
substitution of the chlorine atom by fluorine atom with stronger
hydrogen-bond accepting ability, aiming at the formation of the
additional hydrogen bond with S242. We could also consider the
substitution of the chlorine atom by larger halogen atom such as
bromine to increase the van der Waals or hydrophobic interactions
with P153 of hSR. On the other hand, the surrounding of the 2,3,4-
trifluorobenzene of 1 was indicated in the right side of Figure 4.
The fluorine atoms of 2,3,4-trifluorobenzene of 1 seemed to be less
important in the interaction of 1 with hSR, because they were
exposed to solvents. Further, we could see that the hydrophobic
side-chain of I236 of hSR was located in the neighborhood of
the fluorine atom at position 4. Therefore, we considered the
replacement of the 2,3,4-trifluorobenzene of 1 to a simpler mono-
Mori, Hisashi
