Inhibitors of type III secretion in Yersinia: Design, synthesis and multivariate QSAR of 2-arylsulfonylamino-benzanilides

Article abstract of DOI:10.1016/j.bmc.2007.07.047

Compound 1, 2-(benzo[1,2,5]thiadiazole-4-sulfonylamino)-5-chloro-N-(3,4-dichloro-phe nyl)-benzamide, was identified as a putative type III secretion inhibitor in Yersinia, and the compound thus has a potential to be used to prevent or treat bacterial infe

Full text of DOI:10.1016/j.bmc.2007.07.047

Bioorganic & Medicinal Chemistry 15 (2007) 6994–7011
Inhibitors of type III secretion in Yersinia: Design, synthesis
and multivariate QSAR of 2-arylsulfonylamino-benzanilides
Anna M. Kauppi,a C. David Andersson,a Henrik A. Norberg,b Charlotta Sundin,b
Anna Linussona and Mikael Elofssona,*
a
˚
˚
Department of Chemistry, Umea University, SE-90187 Umea, Sweden
˚
Innate Pharmaceuticals AB, Umestan Fo¨retagspark, SE-90347 Umea, Sweden
b
Received 19 February 2007; revised 18 July 2007; accepted 25 July 2007
Available online 22 August 2007
Abstract—Compound 1, 2-(benzo[1,2,5]thiadiazole-4-sulfonylamino)-5-chloro-N-(3,4-dichloro-phenyl)-benzamide, was identified as
a putative type III secretion inhibitor in Yersinia, and the compound thus has a potential to be used to prevent or treat bacterial
infections. A set of seven analogues was synthesized and evaluated in a type III secretion dependent reporter-gene assay with viable
bacterial to give basic SAR. A second set of 19 compounds was obtained by statistical molecular design in the building block and
product space and by subsequent synthesis. Evaluation in the reporter-gene assay showed that the compounds ranged from non-
active to compounds more potent than 1. Based on the data multivariate QSAR models were established and the final Hi-PLS model
showed good correlation between experimentally determined % inhibition and the calculated % inhibition of the reporter-gene
signal.
Ó 2007 Elsevier Ltd. All rights reserved.
1. Introduction
like conditions will likely identify completely new sets
of molecules, as recently shown for Vibrio cholerae
where a small molecule inhibitor of the transcriptional
activator ToxT, virstatin, prevented both toxin and pili
expression, protecting infant mice from colonization.10
The importance and potential of virulence mechanisms
as drug targets is further underscored by a recent and
extensive study of Salmonella that indicates that a strat-
egy focusing on metabolic pathways is unlikely to be
successful.11 Virulence blocking agents can also be em-
ployed in a chemical biology approach to learn and
understand more about protein function and the mech-
anisms underlying the complexity of bacterial
virulence.12
Although mankind has developed a large arsenal of
antibacterial agents, lethality in infectious diseases still
is the 2nd leading cause of death worldwide.1 Another
worrisome fact is the constantly increasing number of
microbial pathogens that develop or acquire resistance
against antibiotics currently in use.2,3 To combat resis-
tant bacteria, there is an obvious need for effective strat-
egies such as development of new antimicrobial drugs
against not yet exploited targets.4–6 Antibacterial drugs
in use today act on a limited number of targets and a
successful continuation of the antibiotic era will most
likely rely on drugs with other modes of action. It is
now clear that various pathogenic bacteria use related
virulence systems and it has been shown that some com-
ponents of certain virulence systems are conserved be-
The type III secretion (T3S) system is a virulence mech-
anism utilized by several Gram-negative pathogens such
as Yersinia spp., Salmonella spp., Shigella spp., Pseudo-
monas aeruginosa, enteropathogenic Escherichia coli,
enterohaemorrhagic E. coli, and Chlamydia spp.13 The
clinical relevance of these pathogens suggests that T3S
systems are potential targets for novel anti-infective
drugs.14,15 The plasmid encoded Ysc (Yersinia secretion)
T3S system in Yersinia16 is extensively studied and rela-
tively well understood and thus serves as a good model
system.17 Three of the eleven known species of Yersinia,
Y. pestis, Y. enterocolitica, and Y. pseudotuberculosis are
tween different species. These findings offer
a
possibility to develop novel antibacterial agents that tar-
get virulence,4,5,7,8 that is, the bacteria’s ability to cause
disease.9 Targeting growth and virulence under in vivo
Keywords: Type III secretion; Yersinia; Virulence; Inhibitors; Statisti-
cal molecular design: multivariate QSAR.
*
Corresponding author. Tel.: +46 90 786 9328; fax: +46 90 138885;
e-mail: mikael.elofsson@chem.umu.se
0968-0896/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2007.07.047

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
6995
pathogenic to mammals of which Y. pseudotuberculosis
and Y. enterocolitica are enteropathogens mainly caus-
ing gastroenteritis, whilst Y. pestis is the agent responsi-
ble for bubonic plague.17,18 The T3S system allows
direct transfer of the effector proteins into the host cell
cytosol, creating an environment supporting bacterial
survival and proliferation by targeting specific host pro-
teins. The Ysc T3S system is a complex machinery com-
posed of specific Yop chaperones (Sycs) that deliver the
effector proteins, Yersinia outer proteins (Yops) to the
secretion channel through which the Yops are delivered
into the host cell. The Ysc T3S secretion apparatus is
essential for the bacteria to evade the host immune de-
fense and compounds targeting this mechanism will
attenuate bacterial virulence without affecting their
growth. We hypothesize that compounds acting on the
T3S will allow the host to clear the infection while devel-
opment of resistance will be slow due to the high energy
cost for the bacteria to mutate virulence genes. More-
over, microbial resistance against antibiotics is often ob-
tained through horizontal gene transfer from the
resident micro-flora and during antibiotic treatment of
patients, a selection of already resistant micro-flora will
occur.19 Compounds that target the T3S system will not
affect the host micro-flora since it lacks T3S systems and
hence development of resistance by target mutations
cannot occur in the micro-flora.20 Moreover, nine of
the Ysc proteins have counterparts in almost all T3S sys-
tems and it has been demonstrated that some compo-
nents of the secretion systems are interchangeable
among different species,21 which provide evidence for
evolutionary conservation. Given that the T3S systems
are conserved among the Gram-negative bacteria, it is
likely that compounds targeting T3S machinery in Yer-
sinia might also affect the T3S system in other species.
Up to date only a limited number of studies regarding
virulence inhibitors and T3S as a potential target have
been presented.22–27
block space are that the characterization of physiochem-
ical properties will be made on a smaller set of com-
pounds and the number of reactants will be reduced in
a straightforward way.28 This procedure also makes it
possible to use the principal properties of building
blocks as variables in a hierarchical partial least square
regression to latent structures (Hi-PLS) analysis.34 This
regression technique makes the evaluation process more
direct; indicating not only what molecular properties
should be varied but also in what part of the molecule
appropriate changes can be introduced.
Herein we present a SMD based approach for evalua-
tion of the putative T3S inhibitor 1 (Fig. 1) that was
identified as a singleton in a screening campaign in via-
ble bacteria.22 Synthesis and evaluation of a focused li-
brary of analogues of 1 allowed establishment of a
multivariate QSAR model of inhibitors of T3S in
Y. pseudotuberculosis.
2. Results and discussion
2.1. Establishment of SAR
Compound 1 (Fig. 1) has earlier been identified as a sin-
gle hit within its class in a screening campaign for T3S
inhibitors. Further experiments confirmed that this com-
pound selectively targeted the T3S without affecting bac-
terial growth although the exact bacterial target is
unknown.22 A set of seven analogues with small varia-
tions in the structure was designed to examine this sin-
gleton’s potential as a T3S inhibitor (Table 1, 2–8).
The 2-arylsulfonylamino-benzanilides were synthesized
via a three-step synthesis from arylsulfonylchlorides,
anilines, and 2-nitrobenzoic acids (Scheme 1, Section
2.5 and Section 4) and evaluated in a previously
described reporter-gene assay.22,24 The chlorine in the
5-position of the benzoic acid is important since substi-
tution to hydrogen (Table 1, 5 and 6) resulted in non-ac-
tive compounds. However, shifting the chlorine from
5- to 4-position did not affect activity (Table 1, cf. 1
and 2) nor did the change to a 4, 5-difluoro derivative
(Table 1, 8). Substitution of both chlorines to hydrogens
at the aniline part (Table 1, 4) resulted in an almost non-
active compound, while if the chlorine in 3-position was
kept (Table 1, 7), the activity was not affected to any
large extent. The thiadiazole group on the arylsulfonyl-
group could be removed without any detrimental effect
on the activity (Table 1, 3). In conclusion, the small vari-
ations made in the initial compound set affected the bio-
logical activity where the aniline and sulfonylamino
groups appear to tolerate modifications while electron-
withdrawing substituents seemed to be beneficial on
the central benzoic acid. This basic knowledge obtained
with compounds 1–8 was considered in the SMD when
selecting building blocks for the next generation of a de-
signed library.
A general method for the identification of biologically
active compounds is to screen large compound collec-
tions in whole-cell phenotypic assays or assays based
on purified proteins, that is, high-throughput screening
(HTS). Interesting compounds with desired biological
activity are selected and must then be further developed
by iterative design, synthesis, and biological evaluation.
A systematic approach to investigate a singleton or a
compound class is to characterize a large number of vir-
tually possible compounds by theoretical molecular
descriptors and subsequent selection of compounds for
synthesis based on chemical diversity, for example by
using statistical molecular design (SMD).28,29 It is possi-
ble to select a smaller subset of compounds by using
SMD techniques without losing information contained
in the full library. One major advantage of SMD is that
the resulting information can be used for establishment
of multivariate quantitative structure–activity relation-
ship (QSAR) models which later on can be used for
interpretation of important molecular properties and
prediction of biological activity for new compounds
belonging to the same class.29–33 SMD can be based
either on the building blocks or on computer-generated
products. The major advantages of design in building
2.2. Design by selection in principal property space
The SAR study of the eight analogues of 1 (2–8, Table 1)
formed the basis for the design of the second set of com-

6996
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
NH
NO
2
a
b
2
COOH
i
S
N
ii
N
Cl
R
O
R
R
O
O
O
i
R
S
Cl
S
ii
NH
O
NH
O
O
S
N
N
iii
Cl
R
H
H
O
iii
R
Cl
1
Figure 1. (a) Compound 1 identified as a single hit in its class and (b) a general representation of 2-arylsulfonylamino-benzanilides, constructed of
anilines, 2-nitrobenzoic acids and arylsulfonylchlorides.
pounds. SMD was performed in a two-step process
where a first selection was carried out in the building
block space followed by a second design in the product
space based on principal properties from a principal
component analysis (PCA) of molecular descriptors
(Fig. 2).
was focused to include compounds in the vicinity to
compounds known to be active according to the initial
SAR study and to complement it with more diverse rep-
resentatives. The total selection resulted in 19 derivatives
of 2-arylsulfonylamino-benzanilides subjected for syn-
thesis and biological evaluation (Table 1, 9–27).
2.3. Selection of building blocks
2.5. Chemistry
ChemFinderACX2002Prod35 was used to search for
commercially available arylsulfonylchlorides, anilines,
and 2-nitrobenzoic acids. Databases were generated
and 1D and 2D descriptors were calculated in molecular
operating environment (MOE) 36 and the data was com-
pressed with PCA.37–39 Building blocks were selected by
visual inspection of mainly the two first principal com-
ponents, primarily describing size and lipophilicity/
polarity (Supplementary material). The selection was
made using the knowledge from the initial SAR study
and also taking synthetic feasibility and diversity into
consideration. For example, 2-nitrobenzoic acids with
substituents in position three and six were excluded in
order to avoid potential steric hindrance during the acy-
lations and sulfonylations. To increase the probability to
obtain a sufficient number of active compounds within a
limited set, only 2-nitrobenzoic acids with electron-with-
drawing substituents in position four and/or five were
selected (Table 1, cf. 1–8). Regarding the arylsulfonyl-
chlorides, and anilines a more diverse selection of build-
ing blocks were made (Supplementary material) since
they appeared to tolerate modifications (see above).
The selection process resulted in 12 arylsulfonyl chlo-
rides, 17 anilines, and three 2-nitrobenzoic acids (Sup-
plementary material).
The 2-arylsulfonylamino-benzanilides (Table 1, 2–27)
were synthesized in three steps from arylsulfonylchlo-
rides, anilines, and 2-nitrobenzoic acids (Scheme 1).
The 2-nitrobenzanilide was formed by reacting the
2-nitrobenzoic acid with aniline in toluene with PCl3 un-
der microwave-assisted heating at 150 °C for 15 min or
by reacting the 2-nitrobenzoic acid with aniline using
HATU (O-(7-azabenzotriazol-l-yl)-N,N,N0,N0-tetram-
ethyluronium hexafluoro-phosphate) and DIEA (diiso-
propylethylamine) in DMF (dimethylformamide)
overnight at room temperature.40 Generally it was ob-
served that for anilines with ionizable or basic atoms,
the reaction was performed in DMF with HATU (Table
1, 19–22). The nitro group was successfully reduced to
the corresponding amine with H2PO2Na41 or hydrazine
and FeCl3.42 Generally the hydrazine/FeCl3 method was
used when reducing 2-nitrobenzanilides halogenated in
position four on the benzoic acid and also later on due
to the observation that the corresponding amine was ob-
tained in higher purity. Reacting the resulting 2-amino-
benzanilide with an arylsulfonylchloride in pyridine with
DMAP (N,N-dimethylaminopyridine) under micro-
wave-assisted heating at 100 °C for 20 min gave the final
compound. After each step the products were purified
with flash chromatography generally with CH2Cl2 as
eluent. All intermediates were analyzed with LC–MS
and brought forward without further characterization.
The target compounds were analyzed and characterized
with 1H NMR, 19F NMR if relevant, and LC–MS (Sec-
tion 4 and Supplementary material). The total yields
were 5–74% over three steps. In total, a focused library
composed of totally 27 compounds was generated
(Table 1).
2.4. Library selection
The selected building blocks were combined to all possi-
ble virtual compounds resulting in 612 compounds for
which 1D and 2D molecular descriptors were calculated.
The data was compressed with PCA and descriptors
with no contribution to the model were excluded result-
ing in a model with five significant principal components
describing 89% of the variation (see Supplementary
material for complete list of descriptors). The final com-
pounds to synthesize were chosen by visual inspection of
the score plots showing the first, second and third
principal component (R2 = 0.74), mainly describing size,
polarity, and lipophilicity, respectively. The selection
2.6. Biological evaluation
All compounds were tested for their ability to inhibit the
T3S in Y. pseudotuberculosis. We have previously de-
scribed a screening system based on viable Y. pseudotu-
berculosis for the identification of potential T3S

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
6997
Table 1. Effect on luciferase light emission for strain YPIII-pIB29 (yopE-luxAB) in the presence of compounds 1–27 at different concentrations
No.
Compounda
% Inhibition of light emissionb
Concentration in lM
100
62
50
69
20
10
S
N
N
N
O
S
Cl
Cl
O
Cl
Cl
NH
O
O
1
2
3
4
5
8
2
3
5
68
8
2
8
48 11
NH
NH
Cl
S
N
O
S
O
NH
68
70
71
73
61
70
35
58
9
7
Cl
O
S
Cl
O
Cl
NH
O
NH
O
Cl
O
S
S
N
N
O
NH
4
20
1
9
1
1
2
ꢀ2
4
NH
Cl
S
N
N
N
N
O
S
Cl
O
Cl
Cl
NH
O
5
6
13
18
3
6
11
12
1
3
10
5
2
2
5
4
1
NH
NH
S
N
O
S
O
NH
O
O
ꢀ1
S
N
O
S
Cl
Cl
O
NH
7
55
80
2
0
47
4
4
13
2
2
3
2
1
NH
NH
Cl
O
S
N
N
O
S
Cl
NH
O
8
66
—
79
86
39
—
43
50
9
F
F
O
O
Cl
Cl
S
F
F
NH
O
c
—
9
—
NH
Cl
OH
O
O
Cl
S
NH
O
10
11
88
87
6
5
3
4
4
2
13
2
4
NH
Cl
Cl
OH
O
F
O
Cl
S
F
NH
O
15
NH
Cl
Cl
S
N
O
F
O
N
S
F
NH
O
12
63
3
50
6
17
2
9
5
NH
Cl
(continued on next page)

6998
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
Table 1 (continued)
No.
Compounda
% Inhibition of light emissionb
Concentration in lM
100
50
—
20
10
—
O
O
Cl
S
F
NH
O
c
—
13
—
Cl
NH
Cl
Cl
O
Cl
O
S
F
NH
O
14
15
72
51
2
6
70
3
33
8
4
11
3
0
O
F
NH
Cl
O
S
O
O
NH
O
32 12
9
5
NH
Cl
O
O
Cl
Cl
S
O
NH
O
16
72
3
72
1
67
2
43
3
NH
Cl
O
O
S
NH
O
O
17
18
15
69
5
2
10
74
4
1
7
2
1
4
3
2
NH
O
O
Cl
O
S
O
NH
O
O
66
37
Cl
NH
Cl
O
Cl
O
S
NH
O
O
O
N
19
20
21
1
7
8
8
3
2
8
7
7
7
1
1
4
2
4
1
2
0
3
0
0
1
2
NH
NH
Cl
O
S
O
NH
N
NO
2
Cl
O
S
O
NH
10
NH
N
Cl
O
S
O
NH
O
22
23
52
75
1
2
59
71
1
3
51
51
3
1
18
22
2
3
Cl
NH
N
Cl
O
Cl
O
O
S
F
NH
NH
F
F
F

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
6999
Table 1 (continued)
No.
Compounda
% Inhibition of light emissionb
Concentration in lM
100
49
50
23
20
10
O
S
F
O
O N
2
F
NH
O
24
25
26
2
2
5
5
3
2
2
85
9
1
5
3
ꢀ2
1
4
6
NH
F
F
OH
Cl
O
Cl
O
Cl
S
Cl
NH
O
92
92
69
57
NH
F
F
O
Cl
O
S
Cl
NH
O
2
1
NH
F
F
O
Cl
O
O N
2
S
Cl
NH
O
27
90
1
79
3
33
3
9
3
NH
F
F
a The compounds were prepared as described in Section 4.
b Mean values and standard deviations (calculated with the Gauss approximation formula) are from triplicates, and experiments were reproduced at
least twice.
c Compounds showed insufficient solubility.
bacteria and the plates were incubated for 1 h at room
temperature followed by 2 h at 37 °C to allow full induc-
tion of the T3S system.43,46 The luciferase activity was
measured within 30 min after the 2-h incubation and
inhibition was calculated as % inhibition of the resulting
light signal compared to control. Several compounds
showed inhibition of the light emission and some of
them proved to be more potent or equal to compound
1 (Table 1). We have previously shown that compound
1 (Fig. 1) does not affect growth in Y. pseudotuberculo-
sis.22 In order to verify that this finding is general, a
small set of the most potent compounds (Table 2) was
Scheme 1. Reagents and conditions: (a) PCl3, toluene, 150 °C, 15 min,
tested for toxicity in a bacterial growth inhibition assay
at different concentrations (10, 20, 50, and 100 lM).
Wild type YPIII-pIB102 (yopE-luxAB) was grown in
growth medium supplemented with Ca2+ in the presence
of compounds (Table 2) and the absorbance at 600 nm
was measured during six hours. In the presence of differ-
ent compounds, the bacteria grew equally well at all
concentrations compared to control with the growth
curves at 50 lM shown in Figure 3 indicating that the
compounds do not exert any general toxicity.
mwi or HATU, DMF, diisopropylethylamine, rt, o.n.; (b) H2PO2Na,
Pd/C, THF/water or N2H4, FeCl3, C (s), MeOH; (c) DMAP, pyridine,
100 °C, 10 min, mwi.
inhibitors from chemical libraries in a high-throughput
mode.22,24 The compounds (Table 1) were screened for
the ability to inhibit light emission originating from
the luciferase activity of the LuxAB protein produced
from a yopE-luxAB transcriptional fusion.43,44 The lucif-
erase activity is directly correlated to the amount of pro-
duced and secreted YopE protein.45 The evaluation was
carried out with the non-virulent YPIII-pIB29 (yopE-
LuxAB) strain of Y. pseudotuberculosis lacking the effec-
tor protein YopH. Cultures were grown in growth med-
ium lacking Ca2+, conditions allowing for induction of
Yop secretion by activation of T3S system when cultures
are shifted from 26 to 37 °C. Compounds were added to
the medium in 96-well plates followed by medium with
With the reporter-gene screening strategy, there is a risk
ending up with compounds inhibiting bacterial growth
and also that the compounds might influence luciferase
instead of the secretion machinery. To reinforce our pre-
viously described screening assay for identification of
potential T3S inhibitors in viable Y. pseudotuberculosis,
an additional assay measuring the enzymatic activity of
secreted YopH was added. For analysis of the enzymatic

7000
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
Figure 2. Overview over the design process, from the characterized building blocks to the final compound selections.
activity of YopH, a virulent wild type YPIII-pIB102
(yopE-luxAB) strain was used. The effector protein
YopH is a protein tyrosine phosphatase that primarily
targets focal adhesion proteins and kinases and its func-
tion is essential for virulence.47 The N-terminal domain
of YopH binds directly to tyrosine phosphorylated pro-
teins and the catalytic activity of dephosporylation can
be measured in vitro. Ninety-six-well plates containing
bacteria and compounds at different concentrations
were grown at 37 °C for 2 h prior to the measurement
of the enzymatic activity originating from secreted
YopH. Inhibition of the reporter gene signal, and con-
comitant reduction of YopH activity at compound con-
centrations with no effect on bacterial growth strongly
support that the compounds selectively target the T3S
machinery. A representative set of compounds, includ-
ing one inactive, (Table 2) was examined for the poten-
tial to inhibit YopH and it was found that the
compounds inhibited both read-out signals to roughly
the same extent (cf. Tables 1 and 2). Thus this set of
inhibitors is selective for T3S system in Y. pseudotuber-
culosis and has no effect on bacterial growth.
100 lM, Table 1) and the experiments were reproduced
at least twice. The multivariate QSAR model of the
compounds’ T3S inhibitory effect was established by
using a multi-Y Hi-PLS with the % inhibition at 20
and 50 lM as the response (Fig. 4). Due to insufficient
solubility for some compounds, it was not feasible to
use the response at 100 lM, and at concentrations below
20 lM several compounds showed low inhibitory effect.
The low solubility might also be the underlying reason
for compounds not reaching full inhibition despite a
clear dose–response at 10, 20, and 50 lM. In total, 10
out of 27 compounds showed an inhibitory effect
exceeding 40% at 20 lM concentration (Table 1).
PCA was performed for each class of building blocks, that
is, eight arylsulfonylchlorides, 10 anilines, and three 2-
nitrobenzoic acids that were present in the designed final
compounds (Fig. 5). The three building block sets were
investigated separately and structural descriptors repre-
senting molecular properties such as charge, size, polarity,
and lipophilicity were used (Table 3). A two-component
PCA model based on 56 descriptors with R2X = 0.84
was obtained for the sulfonylchlorides where the first
principal component described size and the second de-
scribe lipophilicity/polarity (Fig. 6). For the anilines, a
two-component PCA model with R2X = 0.74 based on
51 descriptors was achieved, where the first principal com-
ponent described size and the second described polarity/
lipophilicity (Fig. 6). Note that even though the second
principal component described the same property
2.7. Multivariate QSAR for T3S inhibition
In total, 27 2-arylsulfonylamino-benzanilides were
successfully synthesized and biologically evaluated
(Table 1). The inhibitory effect of the compounds was
determined as averages from triplicates in a lumines-
cence assay at four concentrations (10, 20, 50, and

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7001
Table 2. Inhibition of the enzymatic activity of YopH for strain YPIII-pIB102 (yopE-luxAB) in the presence of selected compounds at different
concentrations
No.
Compound
% Inhibition of YopH activitya
Concentration in lM
100
63
50
66
20
10
19
S
N
N
O
S
Cl
O
Cl
NH
O
1
2
1
62
2
8
NH
Cl
O
O
S
Cl
Cl
O
NH
O
16
17
18
34
34
59
3
4
2
35
25
64
2
2
2
60
19
60
2
3
3
35
14
33
7
5
7
NH
Cl
O
O
S
NH
O
O
NH
O
O
Cl
O
S
O
NH
O
Cl
NH
O
Cl
Cl
O
O
S
F
NH
O
23
25
27
70
81
81
1
2
2
63
80
74
2
1
1
54
82
67
2
1
2
27 10
NH
O
F
F
F
OH
O
S
Cl
O
Cl
Cl
NH
57
26
3
6
NH
Cl
F
F
O
S
Cl
O
O N
2
Cl
NH
O
NH
F
F
a Mean values and standard deviations (calculated with the Gauss approximation formula) are from triplicates, and experiments were reproduced at
least twice.
(lipophilicity/polarity) for both the anilines and sulfonyl-
chlorides, it is rotated. In the score plots for sulfonylchlo-
rides and anilines, three distinct groupings could be
observed. Building blocks were grouped into small hydro-
phobic, chlorinated, and hydrogen bond accepting build-
ing blocks as indicated in score plots (Fig. 6a and c). For
the three 2-nitrobenzoic acids, two indicator variables
describing the substitution pattern of the halogens were
added besides the one-component PCA model based on
39 descriptors with R2X = 0.99.
PLS regression.48 The main advantage with this ap-
proach is that it facilitates the interpretation of the
structural influence of the different parts of the mole-
cules on the biological activity. In addition, the PCA of-
fers the possibility to reduce the noise level in X before
the regression modeling and hence decreases the risk
of chance correlations. The regression resulted in a
two-component PLS model with an R2Y of 0.78 for
describing the % inhibition at 50 lM (Q2 = 0.38) and
an R2Y of 0.72 for the response at 20 lM (Q2 = 0.29)
(Table 4; Model 1). The multivariate QSAR model
was established with 22 compounds, from these, 16 were
derived from the second set of compounds and six
including the hit 1 from the primary SAR investigation.
Compound 26 (Table 1) was excluded from the model-
ing process due to the observed reversed dose–response
activity that could not be explained, that is, showing
good activity at 10 lM and no activity at higher concen-
trations. Moreover, 5 and 6 (Table 1) were excluded
These five new PCA score vectors describing the princi-
pal properties in the three building block sets and the
two terms describing the substitution pattern on the
benzoic acids were used as X variables in a multi-Y
Hi-PLS with the % inhibition at 20 and 50 lM as the re-
sponse (Fig. 4). In Hi-PLS the data matrix X is divided
into sub-blocks, here the characterization of the building
block sets, to each of which PCA is applied prior to the

7002
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
The normal probability plots of the residuals and corre-
lation plots of the experimentally determined and by the
model calculated % inhibition indicated a non-linear
behavior of the model (data not shown). Logit transfor-
mation of the % inhibition data could not compensate
for the non-linear relationship (data not shown). Hence,
to deal with non-linearity, square and interaction terms
were added to the linear terms followed by exclusion of
non-significant terms. This resulted in a two-component
Hi-PLS model described by the seven linear terms and
12 interaction terms with R2Y = 0.93 and Q2 = 0.71 for
the response at 50 lM and R2Y = 0.91 and Q2 = 0.64
for the response at 20 lM (Table 4, Model 2). The inter-
pretation of the linear main terms did not differ between
these two models but the model was significantly im-
proved according to explained variance and cross-vali-
dation with the addition of the square and interaction
terms (Table 4, Models 1 and 2). This final Hi-PLS mod-
el (Table 5, Model 2) gave a good linear correlation be-
tween the experimentally determined and by the model
calculated % inhibition with a root mean square of error
of estimation (RMSEE) of 8.54 for 50 lM (Fig. 7) and
an RMSEE of 8.91 for 20 lM.
Figure 3. Growth curves for wt YPIII-pIB102 in the presence of
different compounds as noted (Tables 1 and 2) at 50 lM and control
supplemented with 1% DMSO.
In all modeling it is important to validate the established
models and here a combination of cross-validation and
permutation was used for validation of the presented
multivariate QSAR models. The permutation test with
repetitive randomization of the order of the values in
the Y matrix showed that the multivariate QSAR model
NH
NH
NH
NH
NO
Cl
2
2
2
2
2
COOH
C1
Cl
F
Cl
Cl
F
F
F
A1
A2
A3
NH
A4
NH
NH
NO
2
NH
2
2
2
2
COOH
C2
Cl
F
N
O
O
O
N
NO
A5
A6
A7
A8
2
NH
NH
2
2
COOH
C3
Cl
F
A9
A10
OH
Cl
S
N
SO Cl
N
SO Cl Cl
2
SO Cl
2
2
SO Cl
2
O
Cl
Cl
B1
B2
B3
B4
B8
Figure 4. Schematic representation over the process for establishment
of the multi-Y Hi-PLS MQSAR model.
SO Cl
2
SO Cl O N
SO Cl
2
SO Cl
2
2
2
F
from the modeling process in view of the fact that these
compounds contained 2-nitrobenzoic acids that were
not included in the design of the second set of com-
pounds. Compounds showing insufficient solubility (Ta-
ble 1, 9 and 13) were also excluded from the modeling
process due to missing values.
B5
B6
B7
Figure 5. Arylsulfonylchlorides, anilines, and 2-nitrobenzoic acids that
were present in the designed product collection and that were included
in the Hi-PLS modeling process.

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7003
Table 3. Molecular descriptors used in the establishment of multi-Y PLS MQSAR model. Descriptors 1–45 represent 2D descriptors and 46–64 3D
descriptors all calculated with MMFF94x force field and partial charges calculated according to Gasteiger–Marsili method
No.
Abbrev.
Explanation
No.
Abbrev.
Explanation
1
Diameter
VDistEq
VDistMa
weinerPath
weinerPol
chi0v
Size
Size/shape
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
Q_VSA_FPOS
Q_VSA_FNEG
Q_VSA_FPPOS
Q_VSA_FPNEG
Q_VSA_FHYD
Q_VSA_FPOL
SlogP
Electronic property/polarity
Electronic property/polarity
Electronic property/polarity
Electronic property/polarity
Electronic property/polarity
Electronic property/polarity
Lipophilicity
2
3
Size/shape
Size/shape
4
5
Size/shape
Size/shape
6
7
chi0v_C
chi1v
Weight
Size/shape
Size/shape
8
SMR
TPSA
density
Size
Lipophilicity
9
Size
Size/shape
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
chi0
chi1
Size
Lipophilicity
Size/shape
Size/shape
vdw_area
vdw_vol
logP(o/w)
ASA
VAdjEq
VAdjMa
zagreb
Lipophilicity
Lipophilicity
Size/shape
Size/shape
Lipophilicity
Size
balabanJ
Kier1
Kier2
Size/shape
Size/shape
dens
pmi
Shape
Size/Polarity
Lipophilicity
Size/shape
Size/shape
vol
VSA
Kier3
KierA1
Size/shape
Size/shape
ASA+
ASAꢀ
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity
KierA2
KierA3
Size/shape
Size/shape
ASA_H
ASA_P
KierFlex
Q_PC+
Polarity
Lipophilicity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
Charge/polarity
DASA
CASA+
Q_PCꢀ
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity/electronic property
Lipophilicity
Q_RPC+
Q_RPCꢀ
Q_VSA_POS
Q_VSA_NEG
Q_VSA_PPOS
Q_VSA_PNEG
Q_VSA_HYD
Q_VSA_POL
CASAꢀ
DCASA
F ASA+
F ASAꢀ
FC ASA+
FCASAꢀ
FASA_H
FASA_P
Lipophilicity
Figure 6. (a) Score plot for anilines A1–A10 (Fig. 5) with corresponding (b) loading plot. (c) Scoreplot for arylsulfonylchlorides B1–B8 (Fig. 5) with
corresponding (d) loading plot. Explanation of the used molecular descriptors and their numbering in the loading plots (b) and (d) are given in Table 3.

7004
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
Table 4. Statistical data of the two-component multi-Y Hi-PLS models for T3S inhibitors with the % inhibition at 20 and 50 lM as the responses
based on the principal properties of the individual building blocks
Compound concentration (lM)
Model 1a
Model 2b
R2
Q2c
Q2d
R2
Q2c
Q2d
50
20
0.78
0.72
0.38
0.29
0.43
0.34
0.93
0.91
0.71
0.64
0.73
0.63
a Model 1: seven linear terms included.
b Model 2: seven linear terms and 12 interaction terms included.
c Cross-validation using eight rounds.
d Cross-validation with leave-one-out method.
Table 5. Explanation of the Hi-PLS model derived and interpretation of the most important factors for anilines and sulfonylchlorides
Building block set
Score vector
Main property
High value
Low value
Anilines
Anilines
t1_A
t2_A
t1_B
t2_B
Size/shape
Large
Small
Polar
Lipophilicity/polarity
Size/shape
Lipophilicity/polarity
Lipophilic
Large
Polar
Sulfonylchlorides
Sulfonylchlorides
Small
Lipophilic
The abbreviation of score vectors refers to the coefficients in Figure 8 which can be correlated to the score and loading plots in Figure 6.
non-linear terms were confounded with each other and
hence the interpretation was mainly focused on the lin-
ear terms.
Interpretation of the values of the regression coefficients
showed that the variations made in the aniline and sul-
fonylchloride parts had the largest effect on the inhibi-
tion of T3S (Fig. 8). In particular, the size and
lipophilicity of building blocks in the sulfonylchloride
set are important for good activity as indicated with
high absolute t1 and t2 values (Fig. 8 and Table 5).
Traced back to the original descriptors, t1 was mainly
described by size dependent descriptors such as molecu-
lar weight, surface area, and density (Fig. 6c and d). In
the second component, t2, the sulfonylchlorides are
mainly separated based on descriptors like logP describ-
ing lipophilicity (Fig. 6c and d). A subsequent inspection
of the score plot (Fig. 6c) revealed that the building
blocks located in the lower right quadrant were the ones
that gave the highest activity, that is, sulfonylchlorides
having two chlorine substituents on the phenyl ring
(Fig. 5).
Figure 7. Experimentally determined % inhibition versus calculated %
inhibition predicted by the two-component multi-Y Hi-PLS model
(Model 2, Table 4) at 50 lM concentration of different compounds.
The structure–activity relationship for the aniline part
was dominated by the hydrophobicity of the building
blocks as indicated by the high positive value of the
regression coefficient of t2 (Fig. 8 and Table 5). Building
blocks A are separated by size dependent descriptors
such as diameter, volume and surface area in the first
component, t1, and in the second component, t2, by
lipophilicity descriptors like logP and SlogP (Fig. 6a
and b). The score plot (Fig. 6a) reveals that it again is
the halogenated building blocks that were responsible
for the highest inhibition effect on the T3S (Fig. 5).
The variation introduced on the benzoic acid part of
the molecule did not seem to influence the inhibitory ef-
fect of the product, as implied by the small regression
coefficient values for the variables describing this part
(Fig. 8).
holds R2 and Q2 values significantly higher than the R2/
Q2 values observed for the randomized models. The R2
values of the permuted models had on average a value
of 0.46 for non-correlated Ys and a negative Q2 for both
responses. The rather high R2 value of the randomized
models seen here has also been observed for others per-
forming QSAR modeling49,50 where it was suggested
that the Q2 value is more appropriate when it comes
to estimating the validity of the models.49,50 In addition,
the use of eight cross-validation rounds and leave-one
out method gave similar satisfying results (Table 4, cf.
c and d) and no outliers were observed according to
DModX. Moreover, a good correlation between the
model at 50 and 20 lM was observed when comparing
the regression coefficients to each other with
a
R2 = 0.96, that is, the model has the same interpretation
for the different responses. The correlation matrix of the
terms included in the model showed that some of the
The modeling results show that the inhibitory effect of
this class of molecules is strongly driven by the hydro-

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7005
Figure 8. Coefficient plot for the two-component multi-Y Hi-PLS model at 50 lM with seven main terms and 12 interaction terms. A corresponds to
the anilines, B to sulfonylchlorides, and C to the 2-nitrobenzoic acids.
phobicity of the molecules and that a high degree of
halogenations, for example, chlorine substitution, is
preferred.
the sulfonylamide in compound 1 has a calculated
51
pKa of 6.2 and thus a large fraction of the molecules
might be deprotonated and thus charged at physiologi-
cal pH. The pH of the BHI medium used in the assays
is 7.0–7.4. Although substitution with halogens contrib-
utes to lipophilicity, it will also affect charge and thus
solubility. Furthermore, it cannot be excluded that the
molecules actually bind to the target in the charged
form.
3. Conclusions
Compound 1 was initially identified in a phenotype-
based reporter-gene assay and although additional stud-
ies indicated T3S specificity, the target protein is
unknown.22 Screening and evaluation using assays
based on living bacteria have many significant advanta-
ges but also drawbacks. By employing a living bacte-
rium as assay system, compound screening and
evaluation are carried out with the agent that causes
the disease and inhibitors with different modes of action
can be identified. Hurdles including membrane penetra-
tion, efflux, and bacterial metabolism are directly ad-
dressed. However, establishment of QSARs becomes a
challenge since structural changes might affect not only
target interactions. In this study, we evaluated a com-
pound set based on 1 in order to establish multivariate
QSAR models to explore the potential for compound
1 as a starting point for development of T3S inhibitors
as virulence blocking drugs. In total, 27 compounds
were prepared and evaluated and the final Hi-PLS mod-
el showed good correlation between experimentally
determined % inhibition and the calculated % inhibition
of the reporter-gene signal at 50 lM (Fig. 7). However,
it should not be ignored that the structural differences
between the compounds are relatively small and the 2-
phenylsulfonylaminobenzanilide scaffold was conserved
in all molecules. The model indicates several factors that
have a positive effect on activity. Hydrophobicity is
important and substitution patterns including chlorines
are beneficial. The fact that the compounds in general
are lipophilic with logP, 3.94–6.23 indicate limitations
regarding the possibility to obtain potent compounds
with sufficient solubility. However electron-withdrawing
substitutents like chlorine, in particular on the central
aromatic ring, also influence the electronic distribution
and thus the acidity of the sulfonylamide. At pH 7.0,
Despite the inherent challenges with data obtained from
assays using live bacteria, a good model was obtained
indicating that this strategy is powerful in development
of virulence inhibitors targeting T3S.
4. Experimental
4.1. Characterization and selection of building blocks and
synthetic targets
ChemFinderACX2002Prod35 was searched for commer-
cially available arylsulfonylchlorides, anilines, and
2-nitrobenzoic acids. The searches generated several
thousands of anilines, approximately five hundred
arylsulfonylchlorides and about hundred 2-nitrobenzoic
acids. Compounds with additives and a molecular
weight >200 were removed resulting in 512 anilines,
129 arylsulfonylchlorides, and 25 2-nitrobenzoic acids.
All molecules were energy minimized with the imple-
mented MMFF94 force field in MOE software36 and
characterized with 1D and 2D molecular descriptors.
PCA was used to compress the characterized building
block sets into their respective main ‘principal’ struc-
tures, that is, to describe the information in the struc-
tural descriptors using a smaller number of new
uncorrelated variables called principal components.
These principal components consist of scores (T) that
describe the main variation of the data and loadings
(P), that express the relationship between scores and
the original characterized building blocks.32,37–39 The

7006
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
principal component analysis was performed in the Sim-
caTM 10.5 and 11.052 and Evince Beta software.53
with no contribution were excluded (see Supplementary
material for complete list). Two indicator variables were
added to the score vectors to describe the halogens’ sub-
stitution pattern (4- and 5-substitution) on the central
fragment, the 2-nitrobenzoic acids.
The molecular descriptors for the characterized com-
pounds used in the PCA were selected based on their
contribution to the model and descriptors with negative
Q2 values were excluded. The number of significant prin-
cipal components used was selected based on their
eigenvalues with a cut-off value >2. Building blocks were
selected by cherry picking from the first and second
component score plots. The first component described
size and the second mainly hydrophobicity for their
respective building blocks. The selection of building
blocks was made considering the information gained
from the primary SAR investigation but also with the
purpose to make a diverse selection. From the initial
building block libraries, a selection was made resulting
in 12 arylsulfonylchlorides, 17 anilines, and three 2-
nitrobenzoic acids (Supporting material). These were
combined to 612 possible virtual compounds that were
energy minimized in the MMFF94 force field in
MOE36 and characterized with 1D and 2D molecular
descriptors. The data were compressed with PCA and
variables with negative Q2 values were excluded result-
ing in a five-component model with R2 value of 0.89
(see Supplementary material for complete list of vari-
ables). Compounds to synthesize were chosen by visual
inspection of the score plots in the first, second, and
third component mainly described by size, polarity,
and lipophilicity, respectively. In the cherry picking of
compound for synthesis, the selection was focused to in-
clude compounds in the vicinity of compounds known
to be active from the SAR study but also taking diver-
sity into account. The final selection resulted in 19 com-
pounds for synthesis.
In the multi-Y Hi-PLS, the extracted principal proper-
ties (scores, T) from the PCAs of the individual building
blocks and the two indicator variables were used as the
X matrix and % inhibition of the reporter gene signal as
the response (Y). In the Hi-PLS modeling all linear
terms were used, and square and interaction terms were
added to handle non-linear behavior of the model. The
use of a compression method prior to the PLS analysis,
that is, Hi-PLS, led to substantially fewer, and hence
more manageable number of square and interaction-
terms than for standard PLS.
The number of interaction and square terms was reduced
based on their coefficient values, with an absolute cut-off
value <0.075. The number of significant components in
Hi-PLS was decided by cross-validation using two inde-
pendent cross-validation rounds (eight classes and leave-
one out method). The Q2 was calculated according to:
PRESS
Q2 ¼ 1 ꢀ
;
SS
where PRESS is the predicted error sum of squares
when all objects have been left out once and SS is the to-
tal sum of squares of Y corrected for the mean.
4.3. Model and design validation
The quality of Hi-PLS models was analyzed through
their Q2 values, calculated by internal validation using
eight cross-validation rounds and the leave-one-out
method.52 To study the significance of the models, their
RMSEE values and the observations distance to the
model in X space (DModX) were examined. Moreover,
permutation tests were performed to investigate the sig-
nificance of the Hi-PLS models. In these tests, the order
of the responses, Y, was randomly permutated 50 times
and the procedure was repeated 35 times allowing for
calculation of mean values for the R2 and Q2 values
intercepting zero.50 The plot of the correlation coeffi-
cient between the original and permuted Y versus the
cumulative R2 and Q2 gives a regression line where the
intercept (R2 and Q2 when the correlation coefficient is
zero) is an estimate of the significance of the model.50,52
4.2. Characterization of building blocks and multivariate
QSAR
The building blocks were created in the MOE software36
and energy minimized with the implemented MMFF94
force field and partial charges were calculated with the
Gasteiger–Marsali method.54 The individual building
blocks were characterized with actively chosen 1D, 2D,
and 3D molecular descriptors known to describe impor-
tant features like size, lipophilicity, flexibility, shape, and
electronic properties.36
The multivariate regression method Hi-PLS was used to
relate the data matrix (X, descriptors) to the activity ma-
trix (Y, biological response). PLS regression is a com-
monly used method for QSAR analysis that maximizes
the covariance between the data matrix and the re-
sponse.55,56 The PCA were performed in the SimcaTM
10.5 and 11.0 and the PLS modeling were performed
using the Simca software 10.5 and 11.0.52
The quality of the SMD was evaluated in Modde52 by
analysis of the correlation matrix for the coefficients in
the final Hi-PLS model.
4.4. Model interpretation
Interpretation of the Hi-PLS model, i.e. what molecular
properties that influence the biological activity, was
made through two levels. First it was based on the
PLS model and secondly through the PCA score vec-
tors, T, with its corresponding loading vectors, P, that
originate from the compression of the original structural
descriptors. The influence of the variables’ importance
PCA was used to compress the characterized building
block sets into a few number of orthogonal principal
components or scores (T) describing the main variation
of the data.37–39 The molecular descriptors for the char-
acterized compounds used in the PCA were selected
based on their contribution to the model and descriptors

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7007
for good activity was based on the regression coefficient
values, where variables with large coefficient values
(both positive and negative) contribute much to the
activity whereas small do not. Here, this means that
PCA score vectors with large positive coefficient values
were positively correlated to the biological response,
while those with large negative coefficient values had a
negative effect on the % inhibition. To understand which
molecular properties that were linked to the PCA score
vectors, T, their corresponding loading vectors, P, were
investigated in more detail, which is discussed in more
detail in Section 2.
(N-morpholino) ethanesulfonic acid pH 5.0, 1.6 mM
DL-1,4-dithiothreitol). The plate was incubated for
15 min on a rotary shaker at 37 °C and the reaction
was stopped by the addition of 20 ll aqueous 1 M
NaOH. The inhibition of YopH secretion was quanti-
fied by absorbance measurement at 405 nm in a mi-
cro-plate reader (TECAN GENios, Gain 150). The
experiment was reproduced twice in independent
experiments.
4.8. Growth inhibition experiments
The experimental procedures were carried out essen-
tially as described before.22,24 Growth inhibition was
measured by growing bacteria at 37 °C in the presence
of different compound concentrations in 96-well plates
containing 100 ll bacterial culture medium diluted to
an OD600 = 0.05–0.1 in BHI medium with 2.5 mM
CaCl2. The experiment was carried out either in a
Molecular Device Spectramax 340 plate reader with
continuous shaking at 37 °C and a periodic reading of
absorption at 600 nm or by reading of absorbance at
595 nm in a TECAN GENios plate reader once an hour
and shaking carried out in a rotary shaker at 37 °C.
Growth rates were followed for up to 24 h.
4.5. Biological evaluation, strains and growth conditions
The biological evaluation was performed with Y. pseu-
dotuberculosis serotype III (YPIII) strains, pIB102
(yopE-luxAB) and pIB29 (yopE-luxAB). The different
strains were grown at room temperature on LB plates
as described elsewhere.22,44
4.6. Reporter-gene assay
The experimental procedures were carried out essen-
tially as described before.22,24 The Y. pseudotuberculosis
strain YPIII-pIB29 (yopE-luxAB) was grown overnight
in brain heart infusion (BHI) medium containing
5 mM EGTA and 20 mM MgCl2 for calcium depletion
at ambient temperature. The optical density at 600 nm
was adjusted to 0.15–0.25 in BHI. In parallel, the com-
pounds solved in DMSO to be tested were dispensed
in four different concentrations (100, 50, 20, and
10 lM final concentration) to the plates with a pipetting
robot (Hydra) in triplicates. All compounds were dis-
solved in DMSO and the final DMSO concentration
in the biological assays was kept at 1% or below. To
each well 50 ll of diluted bacteria was added with a
dispenser instrument (Multi Drop, Thermo Lab Sys-
tems). The plates were incubated at ambient tempera-
ture on a rotary shaker for 1 h followed by incubation
at 37 °C for 2 h. Within 30 min after incubation, 50 ll
of fresh decanal solution (10 ll/100 ml water) was
added to the wells and the luciferase activity was mea-
sured immediately in a micro-plate reader (TECAN
GENios, Gain 150; Integration time 20 ms). Experi-
ments were reproduced in at least three independent
experiments.
4.9. Chemistry
1H NMR spectra were recorded on a Bruker DRX-400
in CDCl3 [residual cloroform-d1 (dH 7.26 ppm) as inter-
nal standard], ((CD3)2CO) [residual acetone-d6 (dH
2.05 ppm) as internal standard] and ((CD3)2SO) [resid-
ual dimethylsulfoxide-d5 (dH 2.50 ppm) as internal stan-
dard]. 19F NMR spectra were recorded on a Bruker
DRX-400 in ((CD3)2CO) with 0.1% CFCl3 (dH 0 ppm)
as internal standard. Mass-spectra were recorded by
detecting positive (ES+) and negative (ESꢀ) molecular
ions with an electro spray Waters Micromass ZG 2000
instrument using an XTerraÒ MS C18 5 lm 4.6 50 mm
*
column and an H2O/acetonitrile eluent system. The
same LC-system was also used for purification with a
*
preparative XTerraÒ Prep MS C18 5 lm 19 50 mm col-
umn and an H2O/acetonirile eluent system.
˚
Methanol and pyridine were dried over 3 A molecular
sieves. All laboratory vessels were dried in oven
(100 °C) before coming in contact with hydrolysis sensi-
tive chemicals. Organic solutions were dried over
Na2SO4 before being concentrated. Amberlite IR-120
was washed with water and methanol. Microwave
heated reactions were carried out in EmrysTM process
vials (2–5 ml) using a SmithCreatorTM microwave instru-
ment from Biotage. TLC was performed on Silica gel 60
F254 (Merck) with detection by UV-light and staining
with Indanetroin hydrate 95 parts 0.2% in butanol with
5 parts 10% acetic acid for detection of amines.
4.7. YopH phosphatase assay
The Y. pseudotuberculosis strain YPIII-pIB102 (yopE-
luxAB) was grown overnight in brain heart infusion
(BHI) medium with 25 lg/ml chloramphenicol (Clm)
on a rotary shaker at 26 °C. The optical density at
600 nm was adjusted to 0.08 in calcium depleted BHI
(5 mM EGTA, 20 mM MgCl2) and 100 ll of the di-
luted bacteria was added to white 96-well plate (Fluo-
roMunc, Nunc) and different concentrations of the
substances (100, 50, 20, and 10 lM) were added. The
plate was incubated on a rotary shaker at 37 °C. After
2 h, 10 ll from every well was transferred to a new
transparent plate containing 90 ll aqueous substrate
mixture (25 mM p-nitrophenyl-phosphate, 40 mM 2-
4.9.1. Typical procedures for the formation of 2-nitro-
benzanilides. Method A: 2-Nitrobenzoic acid (1 equiv,
0.250 g) and aniline (1.2 equiv) were suspended in tolu-
ene (4 ml) and stirred for 5 min followed by addition of
PCl3 (0.7 equiv) and the reaction was carried out in a
microwave instrument at 150 °C for 15 min. The

7008
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
resulting solution was diluted with methanol (25 ml) and
toluene (25 ml). Excess aniline was removed from the
solution by treatment with Amberlite IR-120 (40 ml)
for 40 min followed by filtration through a layer of sil-
ica. To remove excess 2-nitrobenzoic acid, the solution
was washed with a 20% NaHCO3 solution once and
H2O twice. The organic phase was dried over MgSO4
(s), filtrated, and concentrated. The products were puri-
fied by flash chromatography typically with CH2Cl2 as
eluent.
4.9.3.1. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-5-
chloro-N-(3,4-dichloro-phenyl)-benzamide (1). H NMR
1
((CD3)CO) d 10.55 (br s, 1H), 9.74 (br s, 1H), 8.28–
8.33 (m, 2H), 8.07 (d, 1H, J = 2.0 Hz), 7.86 (dd, 1H,
J = 8.8, 7.2 Hz), 7.77 (d, 1H, J = 8.8 Hz), 7.64–7.70
(m, 3H). LC–MS [MꢀH]ꢀ calcd m/z 506.69, found:
506.98.
4.9.3.2. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-4-
chloro-N-(3,4-dichloro-phenyl)-benzamide (2). The sub-
stance was synthesized according to procedure B, D,
1
Method B: 2-Nitrobenzoic acid (1 equiv, 0.200 g) and
HATU (O-(7-azabenzotriazol-l-yl)-1,1,3,3-tetramethyl
uronium hexafluorophosphate) (1 equiv) were dissolved
in DMF (10 ml) followed by diisopropylethylamine and
the resulting solution was stirred for 15 min at rt. The
aniline (1 equiv) was added and stirring was continued
overnight at rt. Solution was transferred to a separating
funnel and diethyl ether (20 ml) was added. The organic
phase was washed with H2O and NaHCO3 (10%, aq) to
remove DMF and unreacted 2-nitrobenzoic acid. The
organic phase was dried over MgSO4 (s), filtrated, and
concentrated. The product was purified by flash chro-
matography, typically with heptane/EtOAc as eluent.
and E. H NMR ((CD3)2CO) d 10.88 (br s, 1H), 9.71
(br s, 1H), 8.41 (d, 1H, J = 7.2 Hz), 8.32 (d, 1H,
J = 9.2 Hz), 8.10 (br s, 1H), 7.89 (dd, 1H, J = 8.4 and
7.2 Hz), 7.79 (d, 1H, J = 2.0 Hz), 7.69–7.74 (m, 2H),
7.65 (d, 1H, J = 8.8 Hz), 7.08–7.17 (m, 1H). LSMS
[MꢀH]ꢀ calcd m/z 511.93, found 511.63.
4.9.3.3. 2-Benzenesulfonylamino-5-chloro-N-(3,4-di-
chloro-phenyl)-benzamide (3). The substance was synthe-
sized according to procedure B, C, and E. H NMR
1
((CD3)2CO) d 10.33 (s, 1H), 10.10 (br s, 1H), 8.06 (d,
1H, J = 2.4 Hz), 7.87 (d, 1H, J = 2.4 Hz), 7.80–7.82
(m, 2H), 7.67–7.73 (m, 2H), 7.57–7.62 (m, 3H), 7.47–
7.51 (m, 2H). LC–MS [MꢀH]ꢀ calcd m/z 453.97, found:
453.66.
4.9.2. Typical procedures for reduction of 2-nitrobenza-
nilides. Method C: 2-Nitrobenzanilide (1 equiv) was dis-
solved in THF (20 ml) and Pd/C (0.100 g ) was added
followed by H2PO2Na (8 equiv) dissolved in H2O
(10 ml). The reaction mixture was refluxed, and the
reduction was run to completion within 20 min. Pd/C
was filtered off and the product was extracted with
CH2Cl2. The organic phase was dried over MgSO4 (s), fil-
trated, and concentrated. The product was purified by
flash chromatography, typically with CH2Cl2 as eluent.
4.9.3.4. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-5-
chloro-N-phenyl-benzamide (4). The substance was syn-
thesized according to procedure A, C, and E. 1H
NMR ((CD3)2CO) d 10.76 (s, 1H), 9.49 (s, 1H), 8.33
(dd, 1H, J = 7.2, 0.8 Hz), 8.29 (d, 1H, J = 8.8 Hz), 7.85
(dd, 1H, J = 8.8, 7.2 Hz), 7.74–7.79 (m, 3H), 7.67–7.69
(m, 1H), 7.45–7.7.51 (m, 3H), 7.24–7.28 (m, 1H). LC–
MS [MꢀH]ꢀ calcd m/z 444.01, found 443.74.
Method D: 2-Nitrobenzanilide (1 equiv), FeCl3 (0.3
equiv), active carbon (0.015 g per 0.100 g starting mate-
rial) was suspended in MeOH (10 ml). The solution was
heated to 65 °C and hydrazine monohydrate (21 equiv)
was added dropwise. The reaction mixture was refluxed
3–6 h. Activated carbon was filtered off and solvent was
removed by rotary evaporation. The product was dis-
solved in EtOAc (10 ml) and washed with H2O. The or-
ganic phase was dried over MgSO4 (s), filtrated, and
concentrated. The product was purified by flash chro-
matography, typically with CH2Cl2 as eluent.
4.9.3.5. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-
N-(3,4-dichloro-phenyl)-benzamide (5). The substance
was synthesized according to procedure A, C, and E.
1H NMR (CDCl3) d 10.64 (s, 1H), 9.56 (s, 1H), 8.29
(d, 1H, J = 6.8), 8.24 (d, 1H, J = 9.2), 8.06 (d, 1H,
J = 2.4), 7.81 (dd, 1H, J = 8.78, 1.74), 7.74 (d, 1H,
J = 8.23), 7.66–7.70 (m, 1H), 7.59–7.66 (m, 2H), 7.44
(t, 1H, J = 7.96), 7.08 (t, 1H, J = 7.31). LC–MS [MꢀH]ꢀ
calcd 476.96, found 477.02.
4.9.3.6. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-
N-(4-chloro-phenyl)-benzamide (6). The substance was
synthesized according to procedure A, C, and E. 1H
NMR (CDCl3) d 10.69 (s, 1H), 9.47 (s, 1H), 8.29 (d,
1H, J = 7.13), 8.24 (d, 1H, J = 8.78), 7.91 (s, 1H), 7.88
(dd, 1H, J = 8.87, 1.83), 7.74 (d, 1H, J = 8.14), 7.65–
7.56 (m, 2H), 7.47–7.40 (m, 2H), 7.25 (d, 1H,
J = 7.77), 7.07 (t, 1H, J = 7.59). LC–MS [MꢀH]ꢀ calcd
m/z 443.01, found 443.00.
4.9.3. Typical procedure for the formation of 2-arylsulfo-
nylamino-benzanilides. Method E: The 2-aminobenzani-
lide (1 equiv) and sulfonyl chloride (1.3 equiv) were
dissolved in pyridine (4 ml) followed by 4-dimethyl-
aminopyridine (1 equiv) and the reaction was per-
formed in a microwave instrument at 100 °C for
20 min. Reaction mixture was transferred to a vessel
using toluene (25 ml) and MeOH (25 ml) and Amber-
lite IR-120 (40 ml) was added to remove pyridine,
DMAP and unreacted 2-aminobenzanilide. The solu-
tion was stirred for 40 min, filtered and the organic
phase was dried over MgSO4 (s), filtrated, and concen-
trated. The product was purified by flash chromatog-
raphy resulting in products with generally >90%
purity if not otherwise stated.
4.9.3.7. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-5-
chloro-N-(3-chloro-phenyl)-benzamide (7). The substance
was synthesized according to procedure A, C, and E. 1H
NMR (CDCl3) d 10.90 (s, 1H), 9.57 (s, 1H), 8.37 (d, 1H,
J = 7.0 Hz), 8.28 (d, 1H, J = 9.0 Hz), 7.90–7.83 (m, 2H),
7.76 (s, 1H), 7.67 (d, 1H, J = 8.3 Hz), 7.61 (d, 1H,
J = 8.2 Hz) 7.45 (t, 1H, J = 8.1), 7.26 (d, 1H,

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7009
J = 8.0 Hz) 7.12 (dd, 1H, J = 8.5, 1.8 Hz). LC–MS
[MꢀH]ꢀ calcd m/z 476.97, found 477.02.
(s, 1H), 7.82 (d, 1H, J = 2.4 Hz), 7.61–7.69 (m, 5H),
7.34 (t, 1H, J = 8.8 Hz). 19F NMR ((CD3)2CO) d
120.7 (s, 1F). LC–MS [MꢀH]ꢀ calcd m/z 504.92,
found 504.64.
4.9.3.8. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-
N-(3,4-dichloro-phenyl)-4,5-difluoro-benzamide (8). The
substance was synthesized according to procedure A,
C, and E. 1H NMR ((CD3)2SO) d 10.62 (br s, 1H),
10.41 (br s, 1H), 8.34 (d, 1H, J = 8.3 Hz), 8.25 (d, 1H,
J = 6.8 Hz), 7.75–7.85 (m, 3H), 7.64 (d, 1H,
J = 8.8 Hz), 7.52–7.60 (m, 1H), 7.46 (dd, 1H, J = 11.0
and 8.8 Hz). 19F NMR ((CD3)2SO) d 141.4 (d, 1F,
J = 4.3 Hz), ꢀ130.5 (d, 1F, J = 18.7 Hz). LC–MS
[MꢀH]ꢀ calcd m/z 513.95, found 513.60. 85% pure
according to 19F NMR.
4.9.3.14. 2-(4-Acetyl-benzenesulfonylamino)-4-chloro-
N-(3-chloro-4-fluoro-phenyl)-benzamide (14). The sub-
stance was synthesized according to procedure A, D,
1
and E. H NMR ((CD3)2CO) d 11.03 (br s, 1H), 10.66
(s, 1H), 7.93–8.01 (m, 5H), 7.84 (d, 1H, J = 8.4 Hz),
7.57–7.63 (m, 2H), 7.22 (t, 1H, J = 9.2 Hz), 6.98 (d,
1H, J = 8.0 Hz), 2.54 (s, 3H). 19F NMR ((CD3)2CO) d
121.8 (s, 1F). LC–MS [MꢀH]ꢀ calcd m/z 479.00, found
478.82.
4.9.3.9. 4-Chloro-2-(3,4-dichloro-benzenesulfonylami-
no)-N-(4-fluoro-phenyl)-benzamide (9). The substance
was synthesized according to procedure A, D, and E.
1H NMR ((CD3)2CO) d 6.68 (d, 1H, J = 2.0 Hz), 6.63
(d, 1H, J = 8.4 Hz), 6.46 (dd, 1H, J = 8.4, 2.0 Hz), 6.39
(dd, 1H, J = 8.8, 4.8 Hz), 6.30 (d, 1H, J = 8.4 Hz), 6.11
(d, 1H, J = 2.0 Hz), 5.80 (t, 2H, J = 8.8 Hz), 5.54 (dd,
1H, J = 8.4, 2.0 Hz). 19F NMR ((CD3)2SO) d 119.7 (s,
1F). LC–MS [MꢀH]ꢀ calcd m/z 470.95, found 470.89.
4.9.3.15. 4-Chloro-2-(3-fluoro-benzenesulfonylamino)-
N-(4-methoxy-phenyl)-benzamide (15). The substance
was synthesized according to procedure A, D, and E.
1H NMR ((CD3)2CO) d 11.19 (br s, 1H) 9.62 (br s,
1H), 7.81–7.88 (m, 1H), 7.63–7.70 (m, 2H), 7.53–7.58
(m, 4H), 7.35–7.42 (m, 1H), 7.24 (dd, 1H, J = 8.4,
2.0 Hz), 6.95 (d, 2H, J = 9.2 Hz), 3.80 (s, 3H). 19F
NMR (CDCl3) d 111.6 (s, 1F). LC–MS [MꢀH]ꢀ calcd
m/z 433.04, found 432.84.
4.9.3.10. 5-Chloro-2-(3,5-dichloro-2-hydroxy-benze-
nesulfonylamino)-N-(4-fluoro-phenyl)-benzamide (10).
The substance was synthesized according to procedure
4.9.3.16. 4-Chloro-2-(3,4-dichloro-benzenesulfonylami-
no)-N-(4-methoxy-phenyl)-benzamide (16). The sub-
stance was synthesized according to procedure A, D,
1
1
A, D, and E. H NMR ((CD3)2CO) d 10.79 (br s, 1H),
and E. H NMR ((CD3)2CO) d 11.08 (br s, 1H), 9.62
9.84 (br s, 1H), 7.98 (s, 1H), 7.84 (d, 1H, J = 2.4 Hz),
7.74–7.77 (m, 2H), 7.69 (t, 1H, J = 2.4 Hz), 7.66 (s,
1H), 7.62 (d, 1H, J = 2.4 Hz), 7.52 (dd, 1H, J = 8.9,
6.5 Hz), 7.13–7.18 (m, 1H). 19F NMR ((CD3)2CO) d
118.1 (s, 1F). LC–MS [MꢀH]ꢀ calcd m/z 486.95, found
486.67.
(br s, 1H), 7.92 (d, 1H, J = 2.0 Hz), 7.85 (d, 1H,
J = 8.5 Hz), 7.63–7.74 (m, 3H), 7.57 (d, 2H,
J = 9.0 Hz), 7.28 (dd, 1H, J = 8.5, 2.1 Hz), 6.95 (d, 2H,
J = 9.1 Hz), 3.81 (s, 3H). LC–MS [MꢀH]ꢀ calcd m/z
482.97, found 482.76.
4.9.3.17. 2-Benzenesulfonylamino-N-benzo[1,3]dioxol-
5-yl-4-chloro-benzamide (17). The substance was synthe-
sized according to procedure A, D, and E. H NMR
((CD3)2CO) d 11.06 (s, 1H), 9.68 (s, 1H), 7.84–7.87 (m,
3H), 7.73 (d, 1H, J = 2.0 Hz), 7.61–7.66 (m, 1H), 7.52–
7.56 (m, 2H), 7.35–7.36 (m, 1H) 7.23 (dd, 1H, J = 8.4,
2.0 Hz) 7.09–7.12 (m, 1H), 6.89 (d, 1H, J = 8.4 Hz),
6.06 (s, 2H). LC–MS [MꢀH]ꢀ calcd m/z 429.03, found:
428.72.
4.9.3.11. 4-Chloro-2-(3,5-dichloro-2-hydroxy-benzene-
sulfonylamino)-N-(3,4-difluoro-phenyl)-benzamide (11).
The substance was synthesized according to procedure
1
1
A, D, and E. H NMR ((CD3)2CO) d 11.28 (br s, 1H),
10.02 (br s, 1H), 7.97–8.04 (m, 1H), 7.95 (d, 1H,
J = 8.8 Hz), 7.50–7.57 (m, 2H), 7.35–7.43 (m, 2H),
7.25–7.34 (m, 1H), 6.95 (dd, 1H, J = 8.4, 2.0 Hz). 19F
NMR ((CD3)2CO)
d 145.0 (d, 1F, J = 22.8 Hz),
ꢀ137.5 (d, 1F, J = 20.8 Hz). LC–MS [MꢀH]ꢀ calcd m/z
504.94, found 504.74.
4.9.3.18. N-Benzo[1,3]dioxol-5-yl-4-chloro-2-(3,4-di-
chloro-benzenesulfonylamino)-benzamide (18). The sub-
stance was synthesized according to procedure A, D,
4.9.3.12. 2-(Benzo[1,2,5]thiadiazole-4-sulfonylamino)-
4-chloro-N-(3,4-difluoro-phenyl)-benzamide (12). The
substance was synthesized according to procedure A,
1
and E. H NMR ((CD3)2CO) d 10.89 (br s, 1H), 9.65
(br s, 1H), 7.92 (d, 1H, J = 2.0 Hz), 7.82 (d, 1H,
J = 8.4 Hz), 7.65–7.72 (m, 3H), 7.27–7.31 (m, 2H), 7.06
(dd, 1H, J = 8.4, 2.0 Hz), 6.84 (d, 1H, J = 8.4 Hz), 6.03
(s, 2H). LC–MS [MꢀH]ꢀ calcd m/z 496.95, found
496.70.
1
D, and E. H NMR ((CD3)2CO) d 10.93 (s, 1H), 9.85
(br s, 1H), 8.42 (d, 1H, J = 6.8 Hz), 8.31 (d, 1H,
J = 8.8 Hz), 7.84–7.94 (m, 2H), 7.67 (s, 1H), 7.63 (d,
1H, J = 8.4 Hz) 7.47–7.56 (m, 1H), 7.33–7.44 (m, 1H),
7.04–7.14 (m, 1H). 19F NMR ((CD3)2CO) d 143.3 (br
s, 1F), ꢀ137.0 (d, 1F, J = 22.8 Hz). LC–MS [MꢀH]ꢀ
calcd m/z 478.91, found 478.82.
4.9.3.19.
2-Benzenesulfonylamino-5-chloro-N-(4-di-
methylamino-phenyl)-benzamide (19). The substance
was synthesized according to procedure B, C, and E.
1H NMR ((CD3)2CO) d 10.89 (br s, 1H), 9.62 (br s,
1H), 7.85 (s, 1H), 7.81 (d, 2H, J = 7.6 Hz), 7.74 (d,
1H, J = 8.8 Hz), 7.51–7.64 (m, 4H), 7.45 (t, 2H,
J = 8.0 Hz), 7.07 (br s, 2H), 3.02 (s, 6H). LC–MS
[MꢀH]ꢀ calcd m/z 428.08, found 427.76.
4.9.3.13. 5-Chloro-N-(3-chloro-4-fluoro-phenyl)-2-(3,4-
dichloro-benzenesulfonylamino)-benzamide (13). The
substance was synthesized according to procedure A,
1
D, and E. H NMR ((CD3)2CO) d 10.16 (br s, 1H),
9.81 (br s, 1H), 7.94 (dd, 1H, J = 6.8, 2.4 Hz), 7.87

7010
A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
4.9.3.20. 5-Chloro-N-(4-dimethylamino-phenyl)-2-(3-
nitro-benzenesulfonylamino)-benzamide (20). The sub-
stance was synthesized according to procedure B, C,
4.9.3.26. N-(3,4-Dichloro-phenyl)-4,5-difluoro-2-(tolu-
ene-4-sulfonylamino)-benzamide (26). The substance was
synthesized according to procedure A, C, and E. 1H
NMR ((CD3)2CO) d 10.27 (br s, 1H), 9.69 (br s, 1H),
7.97 (d, 1H, J = 2.2 Hz), 7.81–7.89 (m, 1H), 7.55–7.65
(m, 5H), 7.25 (d, 2H, J = 8.4 Hz), 2.28 (s, 3H). 19F
NMR (CDCl3) d 131.5 (d, 1F, J = 24.0 Hz), ꢀ144.0 (d,
1F, J = 23.2 Hz). LC–MS [MꢀH]ꢀ calcd m/z 469.00,
found 468.76.
1
and E. H NMR ((CD3)2CO) d 10.93 (br s, 1H), 9.51
(br s, 1H), 8.45–8.48 (m, 1H), 8.36–8.44 (m, 1H), 8.12–
8.20 (m, 1H,), 7.70–7.83 (m, 3H), 7.57–7.64 (m, 1H),
7.48 (d, 2H, J = 9.2 Hz), 6.78 (d, 2H, J = 9.2 Hz), 2.98
(s, 6H). LC–MS [MꢀH]ꢀ calcd m/z 473.07, found
1
472.73. 75% pure according to H NMR.
4.9.3.21. 2-Benzenesulfonylamino-5-chloro-N-quinolin-
7-yl-benzamide (21). The substance was synthesized
according to procedure B, C, and E. 1H NMR
((CD3)2SO) d 10.71 (br s, 1H), 10.50 (br s, 1H), 8.82–
8.89 (m, 1H), 8.40–8.47 (m, 2H), 8.03 (d, 1H,
J = 9.1 Hz), 7.85–7.94 (m, 2H), 7.75 (d, 2H,
J = 7.6 Hz), 7.52–7.62 (m, 3H), 7.45–7.52 (m, 2H), 7.38
(d, 1H, J = 8.8 Hz). LC–MS [MꢀH]ꢀ calcd m/z
436.05, found 435.74.
4.9.3.27. N-(3,4-Dichloro-phenyl)-4,5-difluoro-2-(3-nitro-
benzenesulfonylamino)-benzamide (27). The substance
was synthesized according to procedure A, C, and E.
1H NMR ((CD3)2CO) d 10.37 (br s, 1H), 9.69 (br s,
1H) 8.49 (s, 1H), 8.39 (d, 1H, J = 8.0 Hz), 8.18 (d, 1H,
J = 8.0 Hz), 7.93 (s, 1H), 7.87 (dd, 1H, J = 10.9,
8.4 Hz), 7.79 (t, 1H, J = 8.1 Hz), 7.64 (dd, 1H,
J = 11.6, 7.2 Hz), 7.56 (s, 2H). 19F NMR (CDCl3) d
131.6 (d, 1F, J = 22.0 Hz), ꢀ142.4 (d, 1F, J = 24.0 Hz).
LC–MS [MꢀH]ꢀ calcd m/z 499.97, found 499.58.
4.9.3.22. 5-Chloro-2-(3,4-dichloro-benzenesulfonylami-
no)-N-quinolin-7-yl-benzamide (22). The substance was
synthesized according to procedure B, C, and E. 1H
NMR ((CD3)2SO) d 10.19–10.96 (m, 2H), 8.83 (dd,
1H, J = 4.0, 1.6 Hz), 8.33–8.43 (m, 2H), 8.00 (d, 1H,
J = 8.8 Hz), 7.91 (d, 1H, J = 1.9 Hz), 7.80–7.87 (m,
2H), 7.68 (d, 1H, J = 7.9 Hz), 7.51–7.63 (m, 3H), 7.34
(d, 1H, J = 8.8 Hz). LC–MS [MꢀH]ꢀ calcd m/z
503.97, found 503.65.
Acknowledgments
We thank the Swedish National Research Council, the
˚
Foundation for Technology Transfer in Umea, the Carl
Trygger foundation and Umea Biotech Incubator for
˚
support.
4.9.3.23. 4,5-Difluoro-2-(3-fluoro-benzenesulfonylami-
no)-N-(4-fluoro-phenyl)-benzamide (23). The substance
was synthesized according to procedure A, C, and
E. 1H NMR ((CD3)2CO) d 10.72 (br s, 1H), 9.66
(br s, 1H), 7.89 (t, 1H, J = 9.8 Hz), 7.87–7.93 (m,
1H), 7.66–7.69 (m, 2H), 7.59–7.65 (m, 2H), 7.50–7.57
(m, 2H), 7.34–7.40 (m, 1H), 7.14–7.19 (m, 2H). 19F
NMR (CDCl3) d 111.6 (s, 1F), ꢀ119.3 (s, 1F),
ꢀ131.6 (d, 1F, J = 23.6 Hz), ꢀ143.7 (d, 1F, J =
23.2 Hz). LC–MS [MꢀH]ꢀ calcd m/z 423.04, found
422.77.
Supplementary data
1
HPLC chromatograms and H and 19F NMR spectra
for compound 1–27. Structures of building blocks used
in the design. Score and loading plots used for building
block selection and descriptor set, score and loading
plots used for selection of synthetic targets. Supplemen-
tary data associated with this article can be found, in the
online version, at doi:10.1016/j.bmc.2007.07.047.
References and notes
4.9.3.24. N-(3,4-Difluoro-phenyl)-4,5-difluoro-2-(3-nitro-
benzenesulfonylamino)-benzamide (24). The substance
was synthesized according to procedure A, C, and E.
1H NMR ((CD3)2CO) d 10.40 (br s, 1H), 9.69 (br s,
1H), 8.48 (d, 1H, J = 2.0 Hz), 8.38 (ddd, 1H, J = 8.0,
2.0, 0.8 Hz), 8.16–8.18 (m, 1H), 7.70–7.87 (m, 3H),
7.58–7.67 (m, 1H) 7.29–7.36 (m, 2H). 19F NMR
(CDCl3) d 131.4 (d, 1F, J = 22.8 Hz), ꢀ138.6 (d, 1F,
J = 22.0 Hz), ꢀ142.4 (d, 1F, J = 22.4 Hz), ꢀ144.6 (d,
1F, J = 22.4 Hz). LC–MS [MꢀH]ꢀ calcd m/z 468.03,
found 467.74.
1. Projan, S. J. Curr. Opin. Microbiol. 2003, 6, 427–430.
2. Poole, K. Curr. Opin. Microbiol. 2001, 4, 500–508.
3. Levy, S. B. Adv. Drug Deliv. Rev. 2005, 57, 1446–1450.
4. Goldschmidt, R. M.; Macielag, M. J.; Hlasta, D. J.;
Barrett, J. F. Curr. Pharma. Des. 1997, 3, 125–142.
5. Lee, Y. M.; Almqvist, F.; Hultgren, S. J. Curr. Opin.
Pharmacol. 2003, 3, 513–519.
6. Walsh, C. Nat. Rev. Microbiol. 2003, 1, 65–70.
7. Alksne, L. E.; Projan, S. J. Curr. Opin. Biotech. 2000, 11,
625–636.
8. Alekshun, M. N.; Levy, S. B. Drug Disc. Tod. 2004, 1,
483–489.
9. Salyers, A. A.; Whitt, D. D. Bacterial Pathogenesis A
Molecular Approach, 2 ed.; ASM Press: Washington, DC,
USA, 2002.
10. Hung, D. T.; Shakhnovich, E. A.; Pierson, E.; Mekalanos,
J. J. Science 2005, 310, 670–674.
4.9.3.25. 2-(3,5-Dichloro-2-hydroxy-benzenesulfonyl-
amino)-N-(3,4-dichloro-phenyl)-4,5-difluoro-benzamide
(25). The substance was synthesized according to proce-
1
dure A, C, and E. H NMR ((CD3)2CO)d 10.36 (br s,
1H), 9.91 (br s, 1H), 8.06 (d, 1H, J = 2.8 Hz), 7.85–
7.93 (m, 1H), 7.73 (d, 1H, J = 2.6 Hz), 7.52–7.66 (m,
4H). 19F NMR (CDCl3) d 126.2 (d, 1F, J = 22.4 Hz),
ꢀ139.3 (d, 1F, J = 22.4 Hz). LC–MS [MꢀH]ꢀ calcd
m/z 538.90, found 538.67.
11. Becker, D.; Selbach, M.; Rollenhagen, C.; Ballmaier, M.;
Meyer, T. F.; Mann, M.; Bumann, D. Nature 2006, 440,
303–307.
12. Ward, G. E.; Carey, K. L.; Westwood, N. J. Cell
Microbiol. 2002, 4, 471–482.

A. M. Kauppi et al. / Bioorg. Med. Chem. 15 (2007) 6994–7011
7011
13. Hueck, C. J. Microbiol. Mol. Biol. Rev. 1998, 62, 379–433.
14. Muller, S.; Feldman, M. F.; Cornelis, G. R. Expert Opin.
Ther. Targets 2001, 5, 327–339.
15. Gauthier, A.; Finlay, B. B. Asm News 2002, 68, 383–387.
16. Cornelis, G. R.; Boland, A.; Boyd, A. P.; Geuijen, C.;
Iriarte, M.; Neyt, C.; Sory, M. P.; Stainier, I. Microbiol.
Mol. Biol. Rev. 1998, 62, 1315–1352.
34. Eriksson, L.; Johansson, E.; Lindgren, F.; Sjo¨strom, M.;
Wold, S. J. Comput. Aided Mol. Des. 2002, 16, 711–726.
35. CambridgeSoft Cambridge, USA, 2002.
36. MOE; 2004.3 ed. Montreal, Quebec, Canada.
37. Wold, H. Estimation of Principal Components and Related
Models by Iterative Least Squares; Academic Press: New
York, 1966.
17. Wren, B. W. Nat. Rev. Microbiol. 2003, 1, 55–64.
18. Perry, R. D.; Fetherston, J. D. Clin. Microbiol. Rev. 1997,
10, 35–66.
19. Normark, S.; Nilsson, C.; Normark, B. H. Science 2005,
307, 1211–1212.
20. Henriques Normark, B.; Normark, S. J. Int. Med. 2002,
252, 91–106.
21. Rosqvist, R.; Hakansson, S.; Forsberg, A.; Wolf-Watz, H.
EMBO J. 1995, 14, 4187–4195.
38. Wold, S.; Esbensen, K.; Geladi, P. Chemom. Intell. Lab.
Syst. 1987, 2, 37–52.
39. Jackson, J. E. A User’s Guide to Principal Components
Analysis; Wiley: New York, 1991.
40. Montalbetti, C. A. G. N.; Falque, V. Tetrahedron 2005,
61, 10827–10852.
41. Entwistle, I. D.; Jackson, A. E.; Johnstone, R. A. W.;
Telford, R. P. J. Chem. Soc., Perkin Trans. 1 1977, 443–
444.
22. Kauppi, A. M.; Nordfelth, R.; Uvell, H.; Wolf-Watz, H.;
Elofsson, M. Chem. Biol. 2003, 10, 241–249.
42. Gowda, S.; Kempe Gowda, B. K.; Channe Gowda, D.
Synth. Comm. 2003, 33, 281–289.
23. Gauthier, A.; Robertson, M. L.; Lowden, M.; Ibarra, J.
A.; Puente, J. L.; Finlay, B. B. Antimicrob. Agents
Chemother. 2005, 49, 4101–4109.
43. Olsson, O.; Koncz, C.; Szalay, A. A. Mol. Gen. Genet.
1988, 215, 1–9.
44. Forsberg, A.; Rosqvist, R. Inf. Agents Dis.-Rev. Iss.
Comm. 1993, 2, 275–278.
45. Pettersson, J.; Nordfelth, R.; Dubinina, E.; Bergman, T.;
Gustafsson, M.; Magnusson, K. E.; Wolf-Watz, H.
Science 1996, 273, 1231–1233.
46. Forsberg, A.; Wolf-Watz, H. Mol. Microbiol. 1988, 2,
121–133.
24. Nordfelth, R.; Kauppi, A. M.; Norberg, H. A.; Wolf-Watz,
H.; Elofsson, M. Infect. Immun. 2005, 73, 3104–3114.
25. Kauppi, A. M.; Nordfelth, R.; Hagglund, U.; Wolf-Watz,
H.; Elofsson, M. Adv. Exp. Med. Biol. 2003, 529, 97–100.
˚
26. Muschiol, S.; Bailey, L.; Gylfe, A.; Sundin, C.; Hultenby,
K.; Bergstro¨m, S.; Elofsson, M.; Wolf-Watz, H.; Nor-
mark, S.; Henriques-Normark, B. Proc. Natl. Acad. Sci.
U.S.A. 2006, 103, 14566–14571.
47. Viboud, G. I.; Bliska, J. B. Ann. Rev. Microbiol. 2005, 59,
69–89.
´
27. Wolf, K.; Betts, H. J.; Chellas-Gery, B.; Hower, S.;
48. Wold, S.; Trygg, J.; Berglund, A.; Antti, H. Chemom.
Intell. Lab. Syst. 2001, 58, 131–150.
49. Eriksson, L.; Verboom, H. H.; Peijnenburg, W. J. G. M.
J. Chemom. 1996, 10, 483–492.
50. Lindgren, F.; Hansen, B.; Karcher, W.; Sjostrom, M.;
Eriksson, L. J. Chemom. 1996, 10, 521–532.
51. http://www.chemaxon.com/demosite/marvin/index.html.
52. Umetrics; SIMCA-P+ 10.0, 10.5, 11.0 ed. Box 7960, S-907
Linton, C. N.; Fields, K. A. Mol. Microbiol. 2006, 61,
1543–1555.
28. Linusson, A.; Gottfries, J.; Lindgren, F.; Wold, S. J. Med.
Chem. 2000, 43, 1320–1328.
¨
29. Linusson, A.; Gottfries, J.; Olsson, T.; Ornskov, E.;
Folestad, S.; Norden, B.; Wold, S. J. Med. Chem. 2001,
44, 3424–3439.
˚
19 Umea, Sweden.
30. Hansch, C. Acc. Chem. Res. 1969, 2, 232–239.
31. Wold, S.; Dunn, W. J., III J. Chem. Inform. Model. 1983,
23, 6–13.
32. Eriksson, L.; Johansson, E. Chemom. Intell. Lab. Syst.
1996, 34, 1–19.
33. Larsson, A.; Johansson, S.; Pinkner, J. S.; Hultgren, S. J.;
Almqvist, F.; Kihlberg, J.; Linusson, A. J. Med. Chem.
2005, 48, 935–945.
˚
53. UmBio UmBio AB, Umea, Sweden.
54. Gasteiger, J.; Marsili, M. Tetrahedron 1980, 36, 3219–
3228.
55. Dunn, W. J., III; Wold, S.; Edlund, U.; Hellberg, S.;
Gasteiger, J. Quant. Struct.-Act. Relat. 1984, 3, 131–
137.
56. Ho¨skuldsson, A. J. Chemom. 1988, 2, 211–228.