R. Goswami et al. / Bioorg. Med. Chem. Lett. 25 (2015) 616–620
617
identification of matriptase inhibitors. Screening of our in-house
library of benzamidine analogs (MW <300) in a biochemical assay
resulted in the identification of weak hits (1 and 2) exhibiting a
afford amidine 20. The Boc-protected amino group (if present) was
deprotected using ethanolic-HCl to give the desired compounds.
Based on the fragment linking approach, compounds 4 and 5
were synthesized (Fig. 1a). Compound 4 (R isomer) showed poor
matriptase inhibition but 5 (S isomer) exhibited 11 and 3-fold
improved inhibition compared to 2 and 3, respectively, with a Ki
matriptase Ki17 of ꢁ80
lM. Fragments 1 and 2 showed potential
for further development because of their low molecular weight,
presence of a single amidine group and moderate matriptase bind-
ing affinity. We decided to retain one amidine functionality in fur-
ther designs because several co-crystal structures indicated that an
amidine group is an important binding feature for making strong
salt-bridge interactions with Asp189 in the S1 pocket.9–11,14–16
Both 1 and 2 were modeled in the matriptase catalytic domain
(cd) which predicted the phenyl sulfonamide (1) and piperidyl
ethyl amine (2) projecting towards S2/S4 and S10 regions, respec-
tively, with the amidine moiety making salt-bridge interaction
with Asp189 in S1 pocket, as observed in the crystallographic bind-
ing mode of benzamidine (Fig. 1a).16 Apart from these, the models
suggested opportunities for further optimization including binding
affinity improvement through incorporation of more hydrophobic
groups on the pendant benzene in 1 because this moiety is
expected to occupy space in the vicinity of Phe99 and Trp215
(Fig. 1a). Therefore, an analog (3) was synthesized replacing the
terminal benzene with naphthyl. Fragment 3 exhibited 3-fold
improvement in matriptase inhibition compared to 1. We further
wanted to explore if fragments 1 and 3 can be linked at the ethoxy
b-carbon (@, Fig. 1a) into a single compound having groups orient-
ing towards three distinct regions of the matriptase cd, viz., S1, S2/
S4 and S10 sites, then this might create more (specific?) interac-
tions and improve binding affinity. This fragment-linking approach
required SAR studies to map these three separate regions since this
approach has not been explored in earlier publications.
of 7.4 lM. This modification demonstrated the importance of ste-
reo-chemistry for binding and also validated our model-based
hypothesis for linking fragments 1 and 3. Although compound 5
exhibited better matriptase inhibition than 2 and 3, there was
room for further improving its inhibitory activity based on the
missing polar contacts involving the P2/P4 and P10 ends as shown
in Figure 1a. This idea was also supported by earlier publica-
tions9,14,15 that highlighted the importance of polar interactions
in the S2–S4 and S10 regions for binding affinity improvement.
Compound 5 was docked in the matriptase catalytic domain,
which suggested a binding mode resembling to 2 and 3 (Fig. 1b)
with a possibility of introducing polar contacts with S2/S4 or S10
residues. This compound 5 model also suggested that moving the
amidine group from para to meta position might shift the sulfon-
amide and carbonyl groups closer to the backbones of Trp215,
Gly216 and Gly219 and gain additional polar interactions
(Fig. 2a). This modification was nevertheless anticipated to com-
promise the critical catalytic Ser195 contact which the ether oxy-
gen of 5 is expected to engage (Fig. 1a) but at the same time
result in additional polar interactions with S2/S4 or S10 potentially
leading to improved binding and target inhibition. With this
hypothesis, compound 6 (Table 1) was synthesized and it showed
significant improvement of 25-fold in matriptase inhibition. A high
resolution (1.9 Å) crystal structure of 6 in complex with matriptase
showed that P10 ethyl amine interacts with His57 through a water
molecule and the P1 amidine binding in the S1 pocket result in a
number of polar contacts. Most importantly, the ether oxygen does
not make any contact with Ser195 (but interacts with Gln192
instead), and the sulfonamide and carbonyl interact with the back-
bones of Trp215, Gly216 and Gly219 residues.
The synthesis of O-(3-carbamimidoylphenyl)-L-serine amides is
outlined in Scheme 1. Methyl -serinate was sulfonylated using
L
arylsulfonyl chloride with NMM (N-methylmorpholine) in THF
resulting in sulfonamide 13, which was further hydrolyzed using
LiOH to obtain acid 14. Commercially available amines were cou-
pled with acid 14 using 1-ethyl-3-(3-dimethylaminopropyl)carbo-
diimide (EDCI) to obtain 15. The free hydroxyl group of 15 was
mesylated with methanesulfonyl chloride to give 16, which was
coupled with 3-hydroxybenzonitrile and resulted in 3-cyanophen-
oxy analog 17. Treatment of nitrile group of 17 with aq NH2OH14 in
EtOH converted it to N-hydroxyamidine (18), followed by acetyla-
tion to form N-acetoxyamidine (19) which was further reduced to
The significantly improved matriptase inhibition observed with
6 and its experimentally determined binding mode with matrip-
tase motivated us to further study the SAR for different P10 and
P4 groups, which are described in Table 1. Initially, compounds 7
and 8 with different P10 groups were synthesized to understand
the role of water-mediated interaction with His57 observed for
Figure 1. (a) Predicted binding modes of 1 (pink) and 2 (gray) compared to the crystallographic binding mode of benzamidine (green) in matriptase cd (PDB ID: 1EAX). Water
mediated interactions involving benzamidine not shown for clarity. indicates linking region for hybrid compounds 4 and 5; (b) overlay of the docked models of 1 (pink), 2
(gray) and 5 (yellow). Benzamidine, 1, 2, 3, 4 and 5 showed Ki of 400 lM (approx.), 78.8 lM, 80.7 lM, 25.4 lM, 98.6 lM and 7.4 lM, respectively, for matriptase.
Goswami, Rajeev
