A new class of salicylic acid derivatives for inhibiting YopH of Yersinia pestis

Article abstract of DOI:10.1016/j.bmc.2014.10.042

Previously, we identified a class of salicylic acid derivatives that display inhibitory activity against the protein tyrosine phosphatase YopH from Yersinia pestis. Because docking study suggested that the large phenyl ring attaching to the salicylic acid

Full text of DOI:10.1016/j.bmc.2014.10.042

Bioorganic & Medicinal Chemistry 22 (2014) 6781–6788
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
A new class of salicylic acid derivatives for inhibiting YopH
of Yersinia pestis
Mahesh P. Paudyal a, Li Wu c, Zhong-Yin Zhang c, Christopher D. Spilling a, Chung F. Wong a,b,
⇑
a Department of Chemistry and Biochemistry, University of Missouri–Saint Louis, One University Boulevard, St. Louis, MO 63121, United States
b Center for Nanoscience, University of Missouri–Saint Louis, One University Boulevard, St. Louis, MO 63121, United States
c Department of Biochemistry and Molecular Biology, and Chemical Genomics Core Facility, Indiana University School of Medicine, 635 Barnhill Drive, Indianapolis, IN 46202,
United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Previously, we identified a class of salicylic acid derivatives that display inhibitory activity against the
protein tyrosine phosphatase YopH from Yersinia pestis. Because docking study suggested that the large
phenyl ring attaching to the salicylic acid core might be exposed to the solvent and might not contribute
significantly to binding, we have developed a new class of compounds that no longer contain this phenyl
ring. We first devised a synthetic scheme for the compounds and then developed an automated compu-
tational screening model surrounding this synthetic scheme to help select a small number of compounds
for synthesis and experimental testing. Based on this computational screening model and the analysis of
the structure–activity relationship of our previous class of compounds, we have synthesized eight com-
pounds and found five that yield micromolar activity. When applying in a larger scale, the synthetic
scheme and the computational screening model developed here should help to identify even more potent
inhibitors in the future.
Received 17 September 2014
Revised 23 October 2014
Accepted 30 October 2014
Available online 5 November 2014
Keywords:
Anti-plague agents
Biodefense
YopH inhibitors
Molecular docking
Virtual screening
Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction
protein kinase LCK and inhibit T-cell activation.11 Together with
several Yersinia outer proteins, researchers also found that YopH
Yersinia pestis is the pathogen responsible for the bubonic pla-
gue that killed millions of people in the 14th century. Since then,
smaller scale epidemics have re-emerged several times, including
the Great Plague of London from 1665 to 1666 and the Third
Pandemic occurring in China in 1855. Although infection by this
bacterium is currently rare in the United States, it still occurs in
California, Utah, Arizona, Nevada, and New Mexico, with ten to
twenty cases per year and about one in seven infected died from
the disease. Worldwide, there are about 2000 cases per year and
about 10% of the infected did not survive.1 However, the bigger
present risk lies in its use as a biological weapon.2 In aerosol form,
the bacterium can be made more deadly and contagious.1,3
Pathogenic Yersinia injects cytotoxic virulent proteins into
mammalian cells to disturb their normal functions. One of these
proteins, YopH, is a potent tyrosine phosphatase,4 which dephos-
phorylates protein substrates in epithelial cells and macro-
phages.5–8 The focal adhesion kinase and the docking protein
p130CAS of mammalian cells were also identified as targets of
YopH.9,10 More recently, YopH was found to dephosphorylate the
disturbs the cytoskeleton of host cells, therefore disrupting
phagocytosis.12,13
The significance of YopH as a drug target stems from evidences
that its interference could inactivate the pathogenic effects of Yer-
sinia pestis. For example, altering the YopH gene to a nonfunctional
one removed the bacterium’s pathogenicity.14–16 Mutating the
essential catalytic cysteine residue to alanine also demolished pro-
tein tyrosine phosphatase activity and abolished the pathogenic
effects of the bacterium.5,6
Researchers have already identified some inhibitors of YopH.
These include p-nitrocatechol sulfate (pNCS).17 The crystal struc-
ture of the complex of pNCS with YopH was also determined,
building the foundation for structure-based design of YopH inhib-
itors.17 The inhibition constant of pNCS to YopH reached only
about 25 l
M17 but more potent inhibitors have also been found.
For example, Phan et al identified a hexapeptide mimic, Ac-
DADE-F2Pmp-L-NH2, of the protein’s natural substrate that
showed nM inhibition.18 They also determined the crystal struc-
ture of this hexapeptide mimic with YopH, providing another use-
ful piece of information for rational drug design. Aurintricarboxylic
acid has also been identified as a potent inhibitor of YopH (5 nM)
and it displays 6–120 fold selectivity in favor of YopH over a panel
⇑
Corresponding author. Tel.: +1 314 516 5318.
E-mail address: wongch@umsl.edu (C.F. Wong).
http://dx.doi.org/10.1016/j.bmc.2014.10.042
0968-0896/Ó 2014 Elsevier Ltd. All rights reserved.

6782
M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
O
tyrosine phosphate
binding site
N3
NHR
peripheral
n
O
binding site
R
NH
O
HO
O
2a, n = 2
HO
HOOC
O
NHR
2b, n = 4
HO
X
Y
N
HN
O
n
HO2C
click chemistry
n
N
N
Z
HOOC
1
3a, n = 2
3b, n = 4
linker n = 1, 2, 3and 4
heterocycle
(orients linker)
Figure 1. Linking two fragments by click chemistry to form 1st generation of
X, Y, Z =O, N, C, S, CH, NH
salicylic inhibitors of YopH.
Figure 4. Generic structure of the 2nd generation of YopH inhibitors with salicylic
acid core.
O
O
O
R
NH2
NaN3
Br
Br
N3
Cl
NHR
high yield and purity under extremely mild conditions.23,24 More
importantly, the cycloaddition reaction can be conducted in aque-
ous solution in the absence of deleterious reagents, allowing direct
screening. The azide-containing building blocks (2) have been syn-
thesized in a one-pot procedure, in which alkyl or aryl amines are
reacted with the acyl chloride linkers (4) in N,N-dimethylformam-
ide (DMF), followed by SN2 reaction with sodium azide to generate
the corresponding azides (2) (Fig. 2).
NHR
n
n
50-90%
n
4a, n = 2
4b, n = 4
5a, n = 2
5b, n = 4
(two steps)
2a, n = 2
2b, n = 4
Figure 2. Synthesis of azides.
of six mammalian protein tyrosine phosphatases.19 In addition,
Tautz et al screened the DIVERSetTM library (ChemBridge, Inc.)
of drug-like compounds, and identified furanyl salicylate com-
pounds as potent inhibitors of YopH.20 Some of these compounds
also showed good cell penetration. Furthermore, by using molecu-
lar docking and 3D-QSAR, Hu and Stebbins rationalized the binding
of derivatives of ketocarboxylic acids and squaric acid to YopH,
providing 3D-QSAR models that help predict the activities of sim-
ilar compounds to YopH.21 In our own work, we developed a new
We constructed an initial library of 80 compounds by using 20
amines and four hydrocarbon linkers with length varying between
1 and 4. (The results for 40 were published previously.22) 33 of
these compounds produced IC50 better than 20
lM, with the best
compounds in the low M range. On the other hand, based on flex-
l
ible ligand-flexible protein molecular docking, we noticed that the
bulky phenyl group attaching to the benzofuran moiety tended to
stick out into the surrounding solvent and thus might not contrib-
ute significantly to binding (Fig. 3).22 As it is inefficient to carry a
big group that increases molecular weight without significant
improvement to binding, we have developed the second series of
compounds shown in Fig. 4 that no longer carried this big group.
This Letter describes the synthetic scheme for this class of com-
pounds, the use of computational tools to help select basic building
blocks, and the biochemical screening results showing some of
these compounds to be active.
class of salicylic compounds presenting low l
M activity.22 In spite
of these encouraging findings, none of these compounds have yet
reached significant preclinical development. As many other factors
such as ADME/toxicology could diminish the potential of any of
these compounds from advancing to the clinical stage, the contin-
uing search of additional chemical scaffolds is still essential.
In this Letter, we describe the development of a second genera-
tion of our salicylic compounds. We first discuss some key features
of our first generation of compounds and explain our motivation
for developing a second generation of compounds.
Our first generation of salicylic compounds (3) uses two func-
tional groups, connected by a hydrocarbon linker, to target the
phosphotyrosine pocket and an adjacent binding pocket simulta-
neously to improve binding affinity and selectivity (Fig. 1). We cou-
ple the two functional groups shown in Figure 1 using click
chemistry. One fragment contains a benzofuran salicylic acid core
(1) to engage the active site. The other includes an amine (2) hold-
ing an R group intended to target a peripheral pocket. The two frag-
ments are connected by alkyl linkers with 1 to 4 methylene units.
Using click chemistry to synthesize these bidentate ligands offers
an expedient way to connect the two components together with
2. Methods
2.1. Synthetic scheme
The second-generation compounds (Fig. 4) maintain the proven
salicylic acid group, but add some five membered aromatic hetero-
cycles, which possess oxygen, sulfur and nitrogen atoms in differ-
ent position. The design incorporates ease of synthesis and has the
following attributes (Fig. 5). (i) Many salicylic acid derivatives (10)
are commercially available as starting materials. (ii) The Boc
Figure 3. One docking structure of a salicylic compound to YopH showing the exposure of the phenyl ring.

M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
6783
R
NH2
HO
R
NH
O
7
HO
O
HO
X
Y
HN
O
OH
HO2C
n
OH
O
X
Y
BocHN
HO2C
Z
n
HO2C
X
8
9
Z
10
6
Figure 5. Building blocks for the synthesis of 2nd generation inhibitors.
HO
R
O
O
O
N
O
O
1. NCS
2. Et3N
O
R
CO2H
OH
N
O
NH2OH HCl
EtOH, Py
TFA, TFAA
acetone
CO2Me
R
O
O
O
MeO2C
R = H
(10%)
R = CHO
11a, R =CHO (40%)
O
13
10a, R = CHO
R = C(=O)Me
12a, R = H (90%)
11b, R = C(=O)Me (71%)
11c, R = I (80%)
10b, R = C(=O)Me
10c, R = I
12b, R = Me (91%)
1. Dabco, PhMe, 80 o
CO2Me
C
R = Me
(13%)
85% over
2 steps
R = I
then Microwave170 oC
2. Boc2O
HO
N
O
O
N
OMe
Et3N
O
O
O
Cl
1.
MeO2C
O
O
HO
O
O
O
O
O
2. LiOH, THF, H2O
(68%2 steps)
16
N
Boc
15
14
Figure 6. Synthetic scheme of salicylic acid-heterocycle biaryls.
O
O
O
OR'
18
OR'
O
O
O
O
OR'
W
(RO)2B
O
O
O
B
X
PdCl2(PPh3)2
O
O
OR'
I
X
O
O
X
W = I, Br
O
O
O
X
PdCl2(PPh3)2, CuI
Et3N, THF
O
O
O
O
O
PdCl2(PPh3)2, CuI
Et3N, THF
B
B
11c
20
19
17
Figure 7. Synthetic scheme of salicylic acid-heterocycle biaryls.
O
S
O
O
CO2H
O
O
O
O
O
O
HO
COOH
O
22
O
O
O
CO2H
S
23
21
24
Figure 8. Biaryl acids prepared to date by cross-coupling.
General Scheme
R
NH
HO
O
TFA, CH2Cl2
1. TFA, CH2Cl2
NHBoc
NH2 EDC, HOBt, DMF
NHBoc
H
n
O
HN
2. EDC, HOBt, DMF
Y
X
Y
N
HO2C
R
O
R
Z
O
n
O
7
HO
Z
6
25a-d
O
HO
X
O
8a-d, n = 1-4
26
Specific Example
O
O
Boc
27
NH2
HO
N
H
N
S
O
O
TFA, CH2Cl2
rt, 1 h, 75%
N
N
Boc
EDC.HCl, HOBt, DMF
N
H
N
H
NH2
N
H
rt, 46 h, 72%
29
S
A0
S
28
O
OH
O
O
N
O
O
O
16
N
TFA/H2O(9:1)
rt, 24 h, 57%
O
N
O
N
H
H
S
EDC.HCl, HOBt, DMF
rt, 46 h, 40%
N
O
30
OH
OH
Figure 9. Synthesis of the 2nd generation inhibitors.

6784
M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
protected salicylic acid via a palladium catalyzed coupling (Suzuki)
reaction.
The salicylic acids (10a and 10b) (Fig. 6) were protected26 and
O
N
N
N
H2N
S
H2N
converted into the corresponding oximes (12a and 12b). Using a
recently published procedure27 the oxime (12b) was converted
into the pyrrole (14) (Fig. 6). Chlorination of the aldoxime (12a)
using N-chlorosuccimide (NCS),28 followed by in situ formation
of the nitrile oxide and addition to methyl propionate gave the
isoxazole (13). Alternatively, the isoxazole regioisomer (16) was
formed by the addition of the known nitrile oxide to the alkyne
(15). The alkyne (15) could be made in 2 steps from 5-iodo salicylic
acid (10c).29–31 Although structures (13) and (14) are interesting,
only the isoxazole (16) was formed in sufficiently high yield for
further application.
S
H2N
A0
A2
A10
Figure 10. Three amines chosen for making second generation salicylic
compounds.
protected a, b, c and d aminoacids (8) are commercially available
and ready for coupling. (iii) Many amines (7) are available to
explore the peripheral binding sites. (iv) Amide bond formation
is reliable25 and can be performed using parallel synthesis meth-
ods. The heterocyclic rings can either be built onto a protected sal-
icylic acid derivative or more efficiently, introduced intact onto a
NH+
N
N
N
S
S
S
NH2
NH2
A8
NH2
A9
NH2
NH2
A5
NH2
A2
NH2 NH2
A3 A4
H2N
A0
H2N
A1
A7
A6
Figure 11. 10 Amines used to construct a virtual library of 180 compounds.
O
O
O
O
O
H2N
O
H2N
H2N
H2N
O
O
H2N
NH
N
O
O
S
NH
O
O
N
N
H2N
N
O
S
N
O
O
O
O
HO
O
O
O
O
HO
O
-O
S1
HO
O
O-
OH
O-
OH
O-
S4
OH
-O
-O
O-
OH
O-
OH
O-
OH
Lib
S2
S3
S5
S9
S7
S8
S6
Figure 12. 9 Salicylic acid cores used to construct a virtual library containing 180 compounds.
Table 1
Summary of docking results for 180 virtual compounds
Lib1
Lib2
Lib3
Lib4
Lib5
Lib6
Lib7
Lib8
Lib9
Ave
0a
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
ꢀ7.68
ꢀ8.15
ꢀ7.99
ꢀ7.13
ꢀ7.63
ꢀ7.04
ꢀ7.34
ꢀ7.48
ꢀ7.66
ꢀ7.28
ꢀ7.13
ꢀ7.14
ꢀ7.42
ꢀ6.95
ꢀ7.08
ꢀ6.67
ꢀ7.73
ꢀ6.49
ꢀ7.02
ꢀ6.21
ꢀ7.261
ꢀ7.48
ꢀ7.43
ꢀ7.86
ꢀ7.31
ꢀ7.35
ꢀ7.21
ꢀ6.76
ꢀ7.69
ꢀ7.23
ꢀ6.6
ꢀ6.92
ꢀ7.07
ꢀ7.85
ꢀ6.99
ꢀ6.8
ꢀ7.57
ꢀ7.67
ꢀ7.08
ꢀ7.03
ꢀ7.06
ꢀ7.29
ꢀ8.07
ꢀ6.77
ꢀ7.21
ꢀ5.86
ꢀ6.7
ꢀ7.44
ꢀ6.38
ꢀ8.09
ꢀ7.2
ꢀ7.03
ꢀ7.66
ꢀ7.91
ꢀ7.45
ꢀ7.39
ꢀ7.37
ꢀ6.69
ꢀ6.88
ꢀ7.84
ꢀ6.87
ꢀ6.73
ꢀ6.91
ꢀ7.18
ꢀ7.07
ꢀ7.68
ꢀ6.93
ꢀ6.37
ꢀ6.78
ꢀ7.53
ꢀ6.25
ꢀ7.126
ꢀ7.17
ꢀ7.92
ꢀ6.96
ꢀ6.95
ꢀ7.22
ꢀ6.84
ꢀ7.39
ꢀ6.92
ꢀ7.41
ꢀ6.46
ꢀ7.29
ꢀ6.92
ꢀ6.96
ꢀ6.65
ꢀ6.71
ꢀ7.13
ꢀ7.07
ꢀ6.79
ꢀ6.78
ꢀ6.47
ꢀ7.0005
ꢀ7.51
ꢀ7.09
ꢀ7.18
ꢀ7.19
ꢀ7.31
ꢀ7.67
ꢀ7.63
ꢀ7.54
ꢀ7.56
ꢀ7.36
ꢀ6.96
ꢀ6.45
ꢀ6.94
ꢀ6.93
ꢀ7
ꢀ6.82
ꢀ7
ꢀ7.29
ꢀ7.37
ꢀ7.57
ꢀ7.18
ꢀ7.26
ꢀ7.17
ꢀ7.31
ꢀ7.14
ꢀ7.46
ꢀ6.78
ꢀ6.89
ꢀ6.88
ꢀ7.14
ꢀ6.81
ꢀ6.99
ꢀ6.84
ꢀ6.93
ꢀ6.73
ꢀ6.91
ꢀ6.29
ꢀ7.25
ꢀ7.39
ꢀ7.37
ꢀ7.27
ꢀ7.89
ꢀ7.69
ꢀ7.3
ꢀ7.2
ꢀ7.14
ꢀ7.11
ꢀ6.97
ꢀ7.81
ꢀ6.83
ꢀ6.23
ꢀ7.33
ꢀ7.47
ꢀ6.91
ꢀ6.71
ꢀ6.94
ꢀ6.81
ꢀ6.45
ꢀ6.83
ꢀ6.3
ꢀ6.74
ꢀ6.9
ꢀ6.3
ꢀ7.13
ꢀ6.87
ꢀ7.45
ꢀ6.63
ꢀ6.52
ꢀ6.67
ꢀ6.88
ꢀ7.02
ꢀ6.27
ꢀ6.23
ꢀ6.97
ꢀ7.25
ꢀ6.907
ꢀ6.87
ꢀ6.96
ꢀ7.23
ꢀ7.87
ꢀ6.83
ꢀ7
ꢀ6.6
ꢀ6.71
ꢀ7.04
ꢀ6.77
ꢀ7.16
ꢀ6.82
ꢀ6.92
ꢀ7.05
ꢀ6.61
ꢀ5.98
ꢀ7.029
ꢀ6.58
ꢀ6.87
ꢀ6.47
ꢀ6.72
ꢀ6.55
ꢀ6.94
ꢀ6.99
ꢀ6.45
ꢀ6.05
ꢀ6.8965
ꢀ6.77
ꢀ6.67
ꢀ6.78
ꢀ6.98
ꢀ6.07
ꢀ6.76
ꢀ7.56
ꢀ6.99
ꢀ7
7
8
9
Ave
ꢀ6.04
ꢀ7.1545
ꢀ6.9735
ꢀ7.0795
The largest averaged binding affinity for each salicylic acid core or for each amine are shown in bold (also italized for the longer linker.)
a
0–9 Labels the amines A0 to A9 in Figure 11, the first set for the shorter linker and the second set for the longer linker. Lib1–Lib9 corresponded to salicylic acid cores S1 to
S9 in Figure 12.

M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
6785
In a more direct and higher yielding approach, the biaryl moiety
was formed via a palladium catalyzed coupling of an aryl halide
and an aryl boronic acid. Several useful aryl boronic acid and
halides are commercially available. The protected iodo salicyclic
acid (11c)32 underwent smooth coupling with several aryl boronic
acids. However, the iodide (11c) could also be transformed into the
corresponding boronate (17)33,34 for coupling with aryl halide part-
ners (18) (Fig. 7). Using the chemistry described above, the biaryl
acids (21–24) shown in Figure 8 were prepared.
5. Performed molecular docking by distributing the docking
runs to many computer processors. In docking with Auto-
dock, we ran 50 runs for each compound using the MMFF94
force field42–48 for the ligand and the AMBER force field49
for the protein and employing the Lamarckian genetic algo-
rithm. Each run used a population size of 150 and was
stopped when 15000 generations were completed. The best
binding affinity was then reported.
Figure 9 shows the general scheme for making the desired YopH
inhibitors by coupling amines (7) to the linker (8) via amide bond
formation using a diimide reagent (e.g. DCC or EDC). The Boc group
was removed under standard conditions (TFA), and the resulting
amine was again coupled via amide bond formation to a salicylic
biaryl acid (26). A final treatment with TFA removed the acetonide
protecting group(s) to give the desired products (6).
This automated computational pipeline has made it easier to
create and screen large libraries using many chemical building
blocks. This Letter describes an initial test on a small library of
180 compounds. Figure 11 shows the 10 amine building blocks
and Figure 12 gives the 9 salicylic acid cores used to construct
the first small virtual combinatorial chemical library. As two
hydrocarbon linkers containing two or four carbons were used,
the library consisted of 10 ꢁ 9 ꢁ 2 = 180 compounds. Note that
the computational fragments were not necessarily the same as
the chemical building blocks actually used in the synthetic scheme.
We chose the fragments only for convenience in computationally
constructing the library.
Further experimental details and analytical data on compounds
that have been made can be found in Supplementary Materials.
2.2. Computer-assisted selection of building blocks for initial
synthesis
To improve the odds of coming up with useful drug candidates
without synthesizing many compounds, we have introduced com-
putational filters. Two filters were used in this initial study. One fil-
ter was based on analyzing structure–activity relationship of the
first-generation compounds; the other involved computationally
constructing all possible compounds based on a chosen set of
building blocks and performing molecular docking to identify good
binders.
Based on structure–activity relationship (SAR) analysis on the
first-generation of compounds, we selected amines A0 and A10
in Figure 10 because they gave active compounds for the first gen-
eration of salicylic compounds regardless of the choice of the
length of the linker.
Table 2
Measured IC50 for nine compounds synthesized
Structure
IC50 value (lM)
O
OH
OH
S
S
O
O
O
O
31
11.1 0.8
N
N
S
N
H
N
H
O
N
H
N
H
30
32
23 0.4
N
S
OH
O
OH
O
O
N
S
N
N
H
2.3. Automating the construction of virtual libraries
surrounding the synthetic scheme and the evaluation of
binding affinity by molecular docking
H
39.5 1.9
O
OH
OH
With the anticipation of the need to analyze a large number of
compounds using the above synthetic scheme to obtain sufficiently
good drug candidates, we automated the process of generating
large virtual combinatorial libraries based on the synthetic scheme
and performing molecular docking on the compounds in the
libraries to select those that were most worthwhile for experimen-
tal synthesis and biological screening. This computational filter has
helped to reduce the number of compounds to be made and tested
biologically, and thus can save time and resources in coming up
with useful drug candidates.
The computational pipeline started with a set of chosen basic
chemical building blocks, formed all possible combination of com-
pounds that could be made by the synthetic scheme already
described, and evaluated the resulting compounds for possible
activity by molecular docking and binding affinity estimation.
The pipeline included the following steps:
S
O
O
S
N
N
H
N
H
33
44.2 14.9
O
OH
OH
O
S
O
O
N
N
H
O
N
H
34
63
1
N
OH
OH
O
O
OH
OH
O
O
35
36
37
38
N
S
>20
>50
N
H
N
H
O
OH
OH
S
O
O
S
N
N
H
N
H
1. Specified all the chemical building blocks in the form of
SMILES strings.35–37
O
OH
OH
O
O
2. Constructed all possible combinations of the SMILES strings
of the building blocks to form the SMILES strings of all pos-
sible members of a virtual combinatorial library.
3. Used Open Babel38,39 to generate three-dimensional struc-
tures of the compounds from the SMILES strings.
4. Used prepare_ligand4.py of Autodock40,41 to set up the com-
pounds in the appropriate format for molecular docking.
>200
>200
S
N
H
O
N
N
H
O
OH
OH
O
O
S
N
H
N
H

6786
M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
2.4. Biochemical assay
synthesis, as this approach could tolerate larger uncertainty of
the model calculations. To this end, we averaged the results for
each amine over all the nine salicylic acid building blocks using
two different hydrocarbon linkers to determine which amines
most likely led to good binders. Likewise, we averaged the results
for all ten amines for each salicylic acid building block to decide
which blocks most likely yielded active compounds. The results
are summarized in Table 1.
We tested the effect of a compound on inhibiting YopH by using
an assay utilizing p-nitrophenyl phosphate (pNPP) as a substrate.
The YopH-catalyzed hydrolysis of pNPP in the presence of varying
concentration of a compound was assayed at 30 °C in a 200
tion system in a 96-well plate. Each reaction contained 2
l
l reac-
ll of the
appropriate amount of the compound dissolved in DMSO to make
up the appropriate desired final concentration. The solution also
contained 198 ll assay buffer (50 mM 3,3-dimethylglutarate,
1 mM EDTA, 1 mM DTT, pH 7.0 with an ionic strength of
0.15 mMM adjusted by addition of NaCl) containing 2 mM pNPP
and 10 nM YopH. The protein tyrosine phosphatase-catalyzed reac-
The results suggested that the salicylic acid core 1, 9, 6, and 8
(shown in Fig. 12) gave the best binding affinity on average. For
the amines, the analysis ranked 2, 8, 1, and 6 shown in Figure 11
to be the best four for the shorter linker, and 2, 4, 6, and 8 for
the longer linker. Therefore, we used salicylic acid core 1, 9, and
8 in the synthesis. We chose the fourth best rather than the third
best salicylic acid core because the fourth best (core 8) had a sim-
ilar chemical structure as the second best (core 9) so that we could
explore the effects of placing the sulfur differently in the five-
member ring. It turned out that the synthesis of the best computa-
tionally inspired salicylic acid core (core 1 in Fig. 12) gave low yield
and solubility; we therefore replaced it by the somewhat similar
core, core 7 instead. For amines, because amine 2 gave the best
binding affinity, we used it as a building block along with the
two suggested by the structure–activity analysis described above.
In addition, because amine 9 in Figure 11 gave the worst averaged
binding affinity for both linkers, we used it to make one compound
to serve as a negative control.
We synthesized nine compounds using these subsets of salicylic
acid cores and amines to see whether we could indeed come up
with active compounds and whether the negative control was
indeed inactive. Table 2 summarizes the results. The negative con-
trol indeed showed no activity. On the other hand, five of the
remaining eight compounds were active. Thus, the successful rate
of classifying the compounds into actives and inactives was 6/
9 = 67% using our design principles based on the analysis of struc-
ture–activity relation of the first-generation of compounds and on
molecular docking of 180 compounds to YopH. Hence, the two
principles we used here should help to prioritize compounds to
make to come up with even better actives in the future.
tion was started by addition of the enzyme. 2 ll of DMSO was used
as a control. The concentration of pNPP was fixed to the KM value
for YopH. The YopH-catalyzed hydrolysis of pNPP was measured
by monitoring the absorbance at 405 nM of the product p-nitro-
phenol continuously, with a SpectraMAX 340 microplate spectro-
photometer (Molecular Devices). The initial rate was obtained by
calculating the slope of the product versus the time curve. We
determined the IC50 value by plotting the relative protein tyrosine
phosphatase activity toward pNPP versus inhibitor concentration
and fitting to the following equation using Kaleidagraph.
Vi=V0 ¼ IC50=ðIC50 þ ½IꢂÞ
ð1Þ
where Vi was the reaction velocity when the inhibitor concentration
was [I], V0 was the reaction velocity with no inhibitor.
3. Results and discussions
3.1. Virtual screening of the library containing 180 compounds
Because of uncertainty in docking simulation and only one pro-
tein structure (PDB code: 1PA9) was used in the docking, we did
not attempt to use individual computed binding affinity of each
compound from the docking study to decide which compounds
should be made. Rather, we used the results collectively to derive
rules to guide the selection of promising building blocks for
S1:A1*
S4:A6
S7:A1
S2:A2
S3:A2
S6:A2
S9:A6
S5:A2
S8:A5
Figure 13. Docking structures with most favorable binding affinity for each of the 9 salicylic acid core. ⁄S1:A1 indicates a docking structure obtained by using salicylic acid
core S1 in Fig. 12 and amine A1 in Fig. 11. All structures were obtained by using the shorter linker.

M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
6787
Although we used an approximate model of virtual screening
that had not accounted for the conformational flexibility of the
protein, it is still useful to analyze some of the docking structures
to get some preliminary insights into how these compounds might
bind to the protein. Figure 13 shows the docking structures of the
compounds that gave the most favorable binding affinity for each
salicylic acid core. All these structures contained the shorter rather
than the longer linker. Consistent with our early docking study of
the first generation of compounds using a more realistic flexible-
protein model,22 the salicylic acid core bound to the phosphotyro-
sine-binding pocket whereas the other ends of the compounds
could bind to one of several possible secondary pockets. Even
though the salicylic acid core bound to the same pocket in all these
structures, their binding mode could be somewhat different in ori-
entation. Amine 2 (Fig. 11) were found in several of these struc-
tures and it was found to bind to different secondary pockets
depending on the salicylic acid cores to which it was attached.
On the other hand, amines such as 1 and 6 appeared to favor one
specific secondary pocket. Because of these diverse modes of pro-
tein recognition, simple comparative analysis of the chemical
structures of these compounds alone might not yield reliable
insights into structure–activity relationships. Quantitative docking
models, such as the one employed here, could add valuable guid-
ance to selecting compounds to make that are more likely to be
biologically active.
of compound evaluation preceded by computational selection fol-
lowed by actual synthesis and biological testing. Figure 4 also
shows other salicylic acid cores that we can make.
Acknowledgements
This work was supported in part by a University of Missouri
Research Board Award and NIH RO1 CA152194. The University of
Missouri Bioinformatics Consortium has provided useful computa-
tional resources. The funding providers played no role in study
design, in the collection, analysis and interpretation of data, in
the writing of the report, and in the decision to submit this article
for publication.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2014.10.042.
References
1. Plague: Overview. Fact sheet revised February 2005. WHO. http://
www.who.int/mediacentre/factsheets/fs267/en/.
2. McGovern, T. W.; Christopher, G. W.; Eitzen, E. M. Arch. Dermatol. 1999, 135,
311.
3. Inglesby, T. V.; Dennis, D. T.; Henderson, D. A.; Bartlett, J. G.; Ascher, M. S.;
Eitzen, E.; Fine, A. D.; Friedlander, A. M.; Hauer, J.; Koerner, J. F.; Layton, M.;
McDade, J.; Osterholm, M. T.; O’Toole, T.; Parker, G.; Perl, M.; Trish, M.; Russell,
P. K.; Schoch-Spana, M.; Tonat, K. JAMA 2000, 283, 2281.
4. Guan, K. L.; Dixon, J. E. Science 1990, 249, 553.
4. Conclusions and future directions
5. Bliska, J. B.; Guan, K. L.; Dixon, J. E.; Falkow, S. Proc. Natl. Acad. Sci. USA 1991, 88,
1187.
6. Bliska, J. B.; Clemens, J. C.; Dixon, J. E.; Falkow, S. J. Exp. Med. 1992, 176, 1625.
7. Hartland, E. L.; Green, S. P.; Phillips, W. A.; Robinsbrowne, R. M. Infect. Immun.
1994, 62, 4445.
8. Andersson, K.; Carballeira, N.; Magnusson, K. E.; Persson, C.; Stendahl, O.;
Wolfwatz, H.; Fallman, M. Mol. Microbiol. 1996, 20, 1057.
9. Black, D. S.; Bliska, J. B. EMBO J. 1997, 16, 2730.
10. Persson, C.; Carballeira, N.; Wolfwatz, H.; Fallman, M. EMBO J. 1997, 16, 2307.
11. Alonso, A.; Bottini, N.; Bruckner, S.; Rahmouni, S.; Williams, S.; Schoenberger, S.
P.; Mustelin, T. J. Biol. Chem. 2004, 279, 4922.
12. Chimini, G.; Chavrier, P. Nat. Cell Biol. 2000, 2, E191.
13. May, R. C.; Machesky, L. M. J. Cell Sci. 2001, 114, 1061.
14. Bolin, I.; Wolf-Watz, H. Mol. Microbiol. 1988, 2, 237.
15. Michiels, T.; Cornelis, G. R. Microb. Pathogen 1988, 5, 449.
16. Straley, S. C.; Bowmer, W. S. Infect. Immun. 1986, 51, 445.
17. Sun, J. P.; Wu, L.; Fedorov, A. A.; Almo, S. C.; Zhang, Z. Y. J. Biol. Chem. 2003, 278,
33392.
As our first-generation of salicylic acid-based YopH inhibitors
carry a large phenyl group that does not appear to contribute sig-
nificantly to binding affinity, we have developed a second genera-
tion of inhibitors that no longer carry this group. The synthetic
scheme will allow us to consider many more inhibitors using many
building blocks that are commercially available. Because the num-
ber of possible compounds that can be made is large, we have also
developed an automated computational pipeline to construct large
virtual combinatorial compound libraries surrounding our syn-
thetic schemes, and to estimate the binding affinity of the com-
pounds in the libraries by molecular docking to help select a
more manageable subset of compounds for experimental synthesis
and biological screening.
As proof of concept, we first selected building blocks based on
analyzing structure–activity relationships of the first-generation
of compounds and on using our computational pipeline to con-
struct an initial small library containing one hundred and eighty
compounds. Based on the insights obtained, we synthesized eight
compounds and one negative control. Five of the eight compounds
turned out to be active and the negative control indeed did not
show activity, giving a successful rate of 5/9 = 67%.
18. Phan, J.; Lee, K.; Cherry, S.; Tropea, J. E.; Burke, T. R.; Waugh, D. S. Biochemistry
(Mosc). 2003, 42, 13113.
19. Liang, F. B.; Huang, Z. H.; Lee, S. Y.; Liang, J.; Ivanov, M. I.; Alonso, A.; Bliska, J.
B.; Lawrence, D. S.; Mustelin, T.; Zhang, Z. Y. J. Biol. Chem. 2003, 278, 41734.
20. Tautz, L.; Bruckner, S.; Sareth, S.; Alonso, A.; Bogetz, J.; Bottini, N.; Pellecchia,
M.; Mustelin, T. J. Biol. Chem. 2005, 280, 9400.
21. Hu, X.; Stebbins, C. E. Bioorg. Med. Chem. 2005, 13, 1101.
22. Huang, Z.; He, Y.; Zhang, X.; Gunawan, A.; Wu, L.; Zhang, Z. Y.; Wong, C. F.
Chem. Biol. Drug Des. 2010, 76, 85.
23. Kolb, H. C.; Finn, M. G.; Sharpless, K. B. Angew. Chem., Int. Ed. Engl. 2001, 40,
2004.
Enlarging the virtual library containing many additional basic
building blocks will help to select other promising compounds to
synthesize and test biochemically. By searching the ZINC data-
base50 for amines with molecular weight smaller than 300 and
with fewer than 7 rotatable bonds, we have identified about
5000 compounds that are commercially available. Using these
compounds to construct larger virtual libraries of compounds that
can be made by our synthetic scheme will help to identify more
compounds that are worthwhile to make and evaluate for biologi-
cal activity experimentally to search for even better drug
candidates.
In addition to the choice of amines, our computational analysis
of one hundred and eighty virtual compounds has already sug-
gested some salicylic acid cores that are worthwhile to consider.
For example, salicylic acid core 6 in Figure 12 yielded the third best
averaged binding affinity and we shall include it in our next round
24. Kolb, H. C.; Sharpless, K. B. Drug Discovery Today 2003, 8, 1128.
25. Montalbetti, C. A. G. N.; Falque, V. Tetrahedron 2005, 61, 10827.
26. Bajwa, N.; Jennings, M. P. J. Org. Chem. 2006, 71, 3646.
27. Ngwerume, S.; Camp, J. E. J. Org. Chem. 2010, 75, 6271.
28. Lilienkampf, A.; Pieroni, M.; Wan, B.; Wang, Y.; Franzblau, S. G.; Kozikowski, A.
P. J. Med. Chem. 2010, 53, 678.
29. Huynh, T. H. V.; Mantel, M. L. H.; Mikkelsen, K.; Lindhardt, A. T.; Nielsen, N. C.;
Otzen, D.; Skrydstrup, T. Org. Lett. 2009, 11, 999.
30. Vazquez, J.; Tautz, L.; Ryan, J. J.; Vuori, K.; Mustelin, T.; Pellecchia, M. J. Med.
Chem. 2007, 50, 2137.
31. Xie, Y.; Liu, Y.; Gong, G.; Rinderspacher, A.; Deng, S. X.; Smith, D. H.; Toebben,
´
U.; Tzilianos, E.; Branden, L.; Vidovic, D.; Chung, C.; Schürer, S.; Tautz, L.;
Landry, D. W. Bioorg. Med. Chem. Lett. 2008, 18, 2840.
32. Casimiro-Garcia, A.; Schultz, A. G. Tetrahedron Lett. 2006, 47, 2739.
33. Murata, M.; Oyama, T.; Watanabe, S.; Masuda, Y. J. Org. Chem. 2000, 65, 164.
34. Murata, M.; Watanabe, S.; Masuda, Y. J. Org. Chem. 1997, 62, 6458.
35. E. Anderson, G. D. Veith, D. Weininger, SMILES:
A line notation and
computerized interpreter for chemical structures, U.S. EPA, Environmental
Research Laboratory-Duluth, Duluth, MN 55804, 1987.
36. Weininger, D. J. Chem Inf. Model. 1988, 28, 31.

6788
M.P. Paudyal et al. / Bioorg. Med. Chem. 22 (2014) 6781–6788
37. Weininger, D.; Weininger, A.; Weininger, J. L. J. Chem Inf. Modell. 1989, 29, 97.
38. O’Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.;
Hutchison, G. R. 2011, 3, 33.
44. Halgren, T. A.; Nachbar, R. B. J. Comput. Chem. 1996, 17, 587.
45. Halgren, T. A. J. Comput. Chem. 1996, 17, 616.
46. Halgren, T. A. J. Comput. Chem. 1996, 17, 553.
39. The Open Babel Package, 2011.
47. Halgren, T. A. J. Comput. Chem. 1996, 17, 520.
40. Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; Belew, R. K.;
Olson, A. J. 1998, 19, 1639.
41. Goodsell, D. S.; Morris, G. M.; Olson, A. J. J. Mol. Recognit. 1996, 9, 1.
42. Halgren, T. A. J. Comput. Chem. 1999, 20, 720.
48. Halgren, T. A. J. Comput. Chem. 1996, 17, 490.
49. Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M.;
Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. J. Comput. Chem. 2005, 26,
1668.
43. Halgren, T. A. J. Comput. Chem. 1999, 20, 730.
50. Irwin, J. J.; Shoichet, B. K. J. Chem. Inf. Model. 2005, 45, 177.