Design, synthesis, and multivariate quantitative structure-activity relationship of salicylanilides-potent inhibitors of type III secretion in Yersinia

Article abstract of DOI:10.1021/jm070741b

Analogues to the salicylanilide N-(4-Chlorophenyl)-2-acetoxy-3,5- diiodobenzamide, 1a, an inhibitor of type III secretion (T3S) in Yersinia, were selected, synthesized, and biologically evaluated in three cycles. First, a set of analogues with variations in the salicylic acid ring moiety was synthesized to probe possible structural variation. A basic structure-activity relationship was established and then used to cherry-pick compounds from a principal component analysis score plot of salicylanilides to generate a second set. A third set with increased likelihood of biological activity was designed using D-optimal onion design. A quantitative structure-activity relationship model using hierarchical partial least-square regression to latent structures (Hi-PLS) was computed using PLS score vectors of building blocks correlated to the % inhibition of T3S as a response. A PLS discriminant analysis (PLS-DA) model was derived using the same descriptor set as that for the Hi-PLS model. Both models were validated with an external test set.

Full text of DOI:10.1021/jm070741b

J. Med. Chem. 2007, 50, 6177-6188
6177
Design, Synthesis, and Multivariate Quantitative Structure-Activity Relationship of
SalicylanilidessPotent Inhibitors of Type III Secretion in Yersinia
Markus K. Dahlgren, Anna M. Kauppi, Ing-Marie Olsson,† Anna Linusson, and Mikael Elofsson*
Department of Chemistry, Umeå UniVersity, SE-901 87 Umeå, Sweden
ReceiVed June 24, 2007
Analogues to the salicylanilide N-(4-Chlorophenyl)-2-acetoxy-3,5-diiodobenzamide, 1a, an inhibitor of type
III secretion (T3S) in Yersinia, were selected, synthesized, and biologically evaluated in three cycles. First,
a set of analogues with variations in the salicylic acid ring moiety was synthesized to probe possible structural
variation. A basic structure-activity relationship was established and then used to cherry-pick compounds
from a principal component analysis score plot of salicylanilides to generate a second set. A third set with
increased likelihood of biological activity was designed using D-optimal onion design. A quantitative
structure-activity relationship model using hierarchical partial least-square regression to latent structures
(Hi-PLS) was computed using PLS score vectors of building blocks correlated to the % inhibition of T3S
as a response. A PLS discriminant analysis (PLS-DA) model was derived using the same descriptor set as
that for the Hi-PLS model. Both models were validated with an external test set.
1. Introduction
Shigella spp., and enteropathogenic Escherichia coli, which all
require T3S for virulence.11 Y. pestis, the agent of Black Death,
Y. pseudotuberculosis, which causes adenitis and septicaemia,
and Y. enterocolitica, which inflicts a broad range of gas-
trointestinal syndromes, are all members of the Yersinia genus
and are pathogenic to man.12,13 Research has revealed the major
molecular events during Yersinia infection,11,14-16 and Yersinia
thus serves as an excellent model to study T3S and identify
inhibitors16,17 of the plasmid-encoded Ysc (Yersinia secretion)
T3S system. In contact with a eukaryotic cell, a number of Yop
Yersinia outer proteins (Yops) and specific Yop chaperones are
expressed. The chaperones protect the Yops on their way to
the Ysc apparatus, which is a membrane-associated complex
of some 20 proteins that span the inner and outer membrane of
the bacterium. At the Ysc apparatus, a number of accessory
proteins assist the translocation of the Yops to the cytosolic
lumen of the eukaryotic cell via a pore in the membrane that is
formed by certain Yersinia proteins. The effector proteins
(YopE, YopH, YpkA, YopJ, YopM, and YopT) all target
specific functions of the eukaryotic cell, including phagocytosis
and inflammatory responses.14
Excessive use of antibiotics is generally accepted to be the
main reason for increased antibiotic resistance among bacteria.1,2
The spread of resistance is hard to control since it proliferates
through gene transfer.3 Although more restrictive use of
antibiotics has a buffering effect, additional approaches to
microbe control are needed.4 Most research so far has been
focused on Gram-positive bacteria due to the fact that the most
serious cases of resistance have, thus far, occurred in Gram-
positives.5 It is, most likely, only a matter of time before the
resistant Gram-negative strains become an equally serious threat.
A century ago, tuberculosis, pneumonia, and diarrhea were
feared as they were lethal diseases. We are steadily approaching
that same situation today. Stomach disorders caused by Yersinia
spp., Salmonella spp., and Shigella spp. cause 18% of the
bereavements among children under 5 years in developing
countries, according to the world health organization (WHOa)
World Health Report 2005. Today, plague is a controlled
disease, but infections and deaths are still reported, and resistant
strains have been found.4 Additionally, one cannot overlook the
risk of multiresistant Y. pestis being used as a bioterrorism
weapon.6,7 Multiresistant bacteria might be the biggest threat
to global health according to WHO. Thus, it is of utmost
importance to develop new efficient antibacterial agents attack-
ing bacterial functions, not yet targeted by human efforts in
both Gram-positive and Gram-negative bacteria.8-10
The translocation of the Yops is essential to evade the host’s
immune defense, and it is of interest to design and synthesize
molecules that can inhibit the secretion or its regulation. This
could result in an antibacterial response without killing the
bacteria, which hopefully will hinder bacteria from developing
resistance but will also have an effect on bacteria that already
have developed resistance against conventional antibiotics. Since
the T3S systems are highly conserved among Gram-negative
bacteria, it is likely that compounds targeting the T3S machinery
in Yersinia also may affect the T3S machinery in other species,
as has been shown for salicylidenacylhydrazides that target T3S
in both Yersinia and Chlamydia.18-22 Up to date, a number of
studies regarding virulence inhibitors and T3S as a potential
target have been presented. Most studies involve salicyliden-
acylhydrazides,19-23 but additional structures have been re-
ported.19,24,25 In addition to these inhibitors that target the actual
T3S machinery or its regulation, screening-based approaches
have also been used to identify inhibitors of specific effector
proteins that are delivered into host cells by T3S.26,27 The
strategy to target virulence is attractive,8,28,29 and a recent study
The type III secretion (T3S) system of Yersinia is a virulence
system which is shared by many different bacteria, including
Salmonella spp., Pseudomonas aeruginosa, Chlamydia spp.,
* To whom correspondence should be addressed. E-mail: mikael.
elofsson@chem.umu.se. Telephone: +46-90-7869328. Fax: +46-90-
138885.
† Current address: Umetrics AB, Stortorget 21 SE-2113, Malmo¨, Sweden.
a Abbreviations: PLS, partial least-square regression to latent structures;
Hi-PLS, hierarchical PLS; T3S, type III secretion; PCA, principal
component analysis; DOOD, D-optimal onion design; WHO, world health
organization; Yops, Yersinia outer proteins; Ysc, Yersinia secretion; SAR,
structure-activity relationship; QSAR, quantitative structure-activity
relationship; BB, building block; SMD, statistical molecular design; PLS-
DA, PLS discriminant analysis; ETEC, Enterotoxigenic Escherichia coli;
TCS, two-component system; YPIII, Yersinia pseudotuberculosis serotype
III; BHI, brain heart infusion.
10.1021/jm070741b CCC: $37.00 © 2007 American Chemical Society
Published on Web 11/02/2007

6178 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
indicates the strong capacity of bacteria to overcome the effect
of inhibitors that target metabolic functions.30
A general method used today within the pharmaceutical
industry and academia to find starting points for drug develop-
ment is to screen large collections of compounds in various
assays.31 It is possible to identify small-molecule virulence
inhibitors in the absence of target structural information by using
high-throughput whole-cell phenotypic assays.32 These large-
scale screens are generally termed high-throughput screening.
Hits from the screening are further evaluated to verify that the
identified response is not due to side effects (false positives).
A basic structure-activity relationship (SAR) is usually estab-
lished to ensure that the structure can be varied and still retain
biological activity. To find more biologically active compounds,
various computational methods are frequently utilized.33 If a
protein structure is known, molecular docking of virtual libraries
into the binding site of the target protein can be done. Another
useful method is statistical molecular design34-36 (SMD), which
results in a well-balanced data set suitable for quantitative
structure-activity relationship (QSAR) modeling.35,37-40 One
way of performing a SMD is to characterize building blocks
(BBs) (i.e., the reactants) using molecular descriptors and then
make a subselection based on chemical diversity. The selection
yields a reduced BB library that is representative (containing
BBs covering most of the variation) of the full library. This
process is repeated for all of the different BB sets, and an
enumeration of a virtual library of products is made out of the
selected BBs. A new diverse selection is made from the product
space, lowering the number of compounds to be synthesized.
Thus, it is possible to select and synthesize a smaller subset of
compounds without losing much of the information contained
in the full library. One advantage of making a design in the BB
space rather than in the product space is that the interpretation
and use of the QSAR will be more direct since the model will
give information about desired properties of BBs. The QSAR
model, in turn, can be used to predict biological activity for
new compounds belonging to the same compound class.41
Models are established correlating an X block of descriptors to
a Y matrix of responses often using partial least-squares
regression to latent structures (PLS).42,43 Multivariate QSAR
models based on PLS also have great value as they allow
interpretation of molecular properties important for biological
activity. Often, molecular properties (calculated molecular
descriptors) are strongly correlated and, hence, cannot serve as
design variables in themselves in a multivariate SMD. Instead,
variables can be compressed by a multivariate analysis method,
typically principal component analysis (PCA),44 which will
result in a low-dimensional hyperplane. The orthogonal principal
components constituting the hyperplane, each corresponding to
a principal property (e.g., hydrophobicity and size), can be used
as design variables in a multivariate SMD. If a design is
performed on the basis of BB sets, it is possible to use the
principal properties of BBs as variables in a hierarchical PLS
(Hi-PLS) analysis.45 This regression technique gives information
about which part of the molecule can be modified to give the
desired changes in the responses. Another advantage of this
strategy is that it is possible to map regions where the synthetic
feasibility is limited, that is, finding regions where building
blocks have poor reactivity.
Figure 1. Compound 1a was identified as a single hit within its class.
between biological response and chemical structure. The predic-
tive ability of the models has been evaluated using predictions
of three compounds followed by synthesis and biological testing.
2. Results and Discussion
2.1. Establishment of SAR. Six analogues (1b, 2a, 2b, 3a,
3b, and 4; Table 1) to 1a were synthesized from readily available
starting materials in order to investigate what kind of alterations
could be performed on the salicylic ring moiety without losing
biological activity entirely. Acetylated salicylanilides were
synthesized in two steps. First, the amide was formed from an
aniline and a salicylic acid, and in a second step, the product
was acetylated using acetic anhydride (see section 2.3 and the
Experimental Section for details). The compounds were tested
in a T3S-linked reporter gene assay in Y. pseudotuberculosis,
essentially as described previously (see section 2.4 and the
Experimental Section for details).19 Biological data indicated
that the hydroxyl or acetylester group was important to retain
biological activity. The iodine at position 3 could be replaced
with a hydrogen to increase the activity, but only if the
compound was acetylated. If unacetylated, the replacement of
iodine at position 3 reduced the activity. If both iodines at the
third and fifth positions were replaced with hydrogens, the
activity was reduced for both the unacetylated and the acetylated
compounds. These initial results supported a further investigation
of the SAR as it was possible to modify the structure without
total loss of biological activity.
2.2. Library Design using Two Complementary SMDs.
The design strategy used was to first perform a SMD to identify
regions in chemical space where biologically active compounds
could be found and then, in a second cycle, do another SMD
around the biologically active compounds to be able to get a
more robust QSAR model later on. The first SMD used was a
cherry-picking selection from a PCA score plot. The aim was
to synthesize a small number of salicylanilides derived from
commercially available sets of aniline and salicylic acid
derivatives. These were then to be acetylated to yield a second
set of compounds. The preliminary knowledge gained from the
initial SAR study was used when choosing new BBs, which
were manually selected (see Experimental Section). Twenty-
five anilines and 22 salicylic acids (see Supporting Information)
were used to enumerate a virtual library, resulting in 550 virtual
unacetylated salicylanilides. The structural variation of the
virtual compounds was summarized with a PCA, and from the
resulting three component models (R2 ) 0.97, Q2 ) 0.96), 16
new unacetylated salicylanilides (5b-20b) were chosen by
visually cherry-picking mainly from the two first principal
components (Figure 2a and b). Most of the compounds were
chosen around compound 1b in order to increase the probability
that several biologically active compounds would be found.
Some of the compounds were chosen in such a way that they
should have some BBs in common in order to ensure that the
interpretation of results later on would be more straightforward.
All of the selected compounds were synthesized (5b-20b) and,
in a second step, acetylated to yield the acetylated salicylanilides
In this paper, we present our design strategy used to select
and synthesize analogues to the T3S inhibitor 1a (Figure 1),
identified earlier as a hit in a screening campaign,19 using three
cycles of iterative selection, synthesis, and biological evaluation.
Multivariate methods have been utilized to find correlations

Structure-ActiVity Relationship of Salicylanilides
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6179
Table 1. Percentile Inhibitions of Luciferase Light Emission for Strain YPIII-pIB29 (yopE-luxAB) in the Presence of Compounds 1-26 at Four
Different Concentrations
a For the compound ID’s, an a denotes R ) Ac, and a b denotes R ) H. b Means and standard deviations are calculated from triplicates, and experiments
were reproduced on at least two separate occasions. Blank fields correspond to lack of biological activity.
(5a-20a). All compounds were biologically evaluated in the
reporter gene assay, and of the acetylated compounds, 7 (5a-
7a, 14a, 16a-18a) out of the 16 showed a dose-dependent
response. In general, if an acetylated salicylanilide was active,
the unacetylated compound was active too.
The second SMD was used as a complement to the cherry-
picking selection to get a more robust data set for the QSAR
modeling. D-Optimal onion design (DOOD)46 is an extension
to D-optimal designs,47,48 and a predecessor of this technique
was here used to find additional compounds with complementary
properties to the ones selected through the cherry-picking SMD.
The major advantage of DOOD, compared to traditional
D-optimal designs, is that the design can be centered around a
compound of your own choice. Around this centerpoint, a
number of “onion” layers are chosen, and a D-optimal design
is performed within each layer. This approach allows chemical
knowledge and experience to play a part in the design process.
The design was performed on unacetylated compounds. BBs
were characterized with descriptors, and data was compressed
with PCA, resulting in three latent variables for both salicylic
acids and anilines, explaining 95% (R2 ) 0.95, Q2 ) 0.76) and
94% (R2 ) 0.94, Q2 ) 0.78), respectively, of the variation in
the matrices. The centerpoint in the design was determined
geometrically (see Experimental Section); essentially, a new
compound (22b) and, from two outer onion layers, five new
compounds (21b, 23b, 24b-26b) were selected by the DOOD.
Both unacetylated and acetylated versions of the compounds
were synthesized and tested (Table 1). Of the six acetylated
compounds from the DOOD, five showed biological activity.
The only inactive acetylated compound selected by the DOOD
(24a) was structurally very different from the other compounds.
The biologically active acetylated salicylanilides all had IC50

6180 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
Figure 2. (a) Score plot of 550 virtual salicylanilides. Compound 1b is marked with a filled triangle, and cherry-picked compounds are marked
with an open triangle. (b) Corresponding loading plot of descriptors used.
values below 20 µM, except compound 26a, which was a low-
potency inhibitor. The unacetylated compounds showed a similar
biological response as that of the acetylated salicylanilides,
except 24b, which showed an inhibitory effect.
Table 2. Additional Data for the Most Potent Inhibitors, Showing the
Luciferase Light Emission Inhibition at Four Different Concentrations
2.3. Synthesis. The acetylated salicylanilides were synthesized
in two steps. Anilines and salicylic acids were reacted under
microwave heating, using PCl3 in toluene, to form unacetylated
salicylanilides. In a second step, the compounds were acetylated
using acetic anhydride and phosphoric acid. The purified
1
compounds were analyzed and characterized with H NMR
spectroscopy and LC-MS (Experimental Section and Supporting
Information). The total yields were 2-60% over two steps. The
lower yields in the first step were mainly due to highly
deactivated anilines. In the second step, salicylanilides with
electron-deficient hydroxyl groups often gave much lower yields
than salicylanilides with higher electron density around the
hydroxyl group. Poor solubility in the solvent during reaction
might have contributed to the poor yields. It was possible to
identify trends in the PCA score plot where some regions had
products that could not be obtained in any quantifiable amount,
while others contained easily synthesized compounds (data not
shown). Over three cycles of selection, synthesis and biological
evaluation of a total library of 50 analogues to 1a were
1
synthesized (Table 1). Compound purities, estimated from H
NMR, are given in a table in the Supporting Information. In
addition, purities estimated from LC chromatograms are given
for key compounds.
2.4. Biological Evaluation. The biological evaluation was
performed using a previously published method.19,21 In total,
25 acetylated, and the corresponding nonacetylated, salicyla-
nilides were biologically evaluated. The inhibitory effect of the
compounds was determined as averages from triplicates in a
luminescence assay at four concentrations (100, 50, 20, and 10
µM; Table 1), and the experiments were reproduced on at least
two separate occasions. The more potent compounds were tested
at four additional concentrations (5, 2, 1, and 0.5 µM; Table
2). Of the 25 acetylated compounds tested, 15 showed a dose-
dependent response and thus could be used to establish a
multivariate QSAR model. Of these, three came from the SAR
study, seven from the cherry-picking selection, and five from
the DOOD. Some of the low-activity compounds did not show
a dose-dependent response, possibly due to poor solubility, and
therefore could not be used in the multivariate modeling later
on. The acetylated compounds were more polar than the
unacetylated, as indicated by TLC. The increase in lipophilicity
was most likely due to the possibility to form an intramolecular
a For the compound ID’s, an a denotes R ) Ac, and a b denotes R ) H.
b Means and standard deviations are calculated from triplicates, and
experiments were reproduced on at least two separate occasions.
hydrogen bond in the nonacetylated compounds. In general, if
an acetylated compound was biologically active, its unacetylated
counterpart was too, even though there were a few discrepancies
(i.e., 24a and 24b). Inhibition of protein secretion was inves-
tigated through Western analysis for one of the most potent
compounds (23a) and its unacetylated counterpart (23b), and it
was found that the level of light emission inhibition roughly

Structure-ActiVity Relationship of Salicylanilides
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6181
molecules to the response using PLS, but only after doing a
variable selection. The only way we could establish a model of
the BBs correlated to response was by using the methodology
outlined above.
Comparing compounds 14a with 26a and 5a with 15a, it
would seem that 3,5,6-trichlorosalicylic acid was a poor BB
for making acetylated salicylanilides as T3S inhibitors. Compar-
ing compound 6a with 21a, both containing the 4-trifluorom-
ethylsulfanylaniline building block, it became clear that there
were interactions between the aniline and the salicylic acid rings
which were also supported by the coefficients. The QSAR model
could explain complex correlations between structure and
activity that, otherwise, would be very difficult to understand
due to several nonlinearities, which makes it highly useful for
predictions. The biologically active compounds with a high
activity generally had an aniline ring moiety with a large-size,
highly oval-shaped, high dipole moment and large hydrophobic
and negatively charged surfaces. Last, the HOMO and LUMO
energies were generally low in the anilines. The salicylic ring
moiety in the active molecules with high activity generally had
a high dipole moment, high hardness, and high density. Large
negative charges, large negatively charged surfaces, and high
HOMO energies also characterized a salicylic acid, present in
potent inhibitors (e.g compounds 21a and 23a; Tables 1 and
2). Even though the model contained several nonlinear terms
that, in combination with the PLS score vectors highly correlated
nature, made a thorough analysis of the SAR complicated, it
could easily be used for predictions of large sets of molecules.
A shortcoming of the model was that it could not be used to
correctly predict inactive compounds that, in several cases, were
predicted as active. This was confirmed by the prediction of
the inactive compounds not used in the modeling. The underly-
ing reason is that active and inactive compounds have different
molecular properties and the latent variables used to describe
the active compounds do not have a relevance for description
of the inactive.
Figure 3. Schematic representation of the multivariate Hi-PLS QSAR
model. X1 and X2 are the BB data matrices with calculated descriptors,
while X3 and X4 are the extracted PLS score vectors. X5 is the square
terms of X3, and X6 is the square terms of X4. X7 is the interaction
term between X3 and X4.
matched inhibition of protein secretion, as seen for compound
1a (see Supporting Information).19
2.5. Multivariate QSAR for T3S Inhibition. Since a
compound could be inactive due to a host of reasons, like efflux
and the inability to pass through a membrane, only compounds
showing a dose response were used to make a multivariate
QSAR model. For the Hi-PLS multivariate QSAR model, the
inhibitory values at 20 and 10 µM were used as a multi-Y
response matrix since the data set had an even spread in
biological activity at these concentrations. All of the BBs that
were part of an active molecule were assembled in two new
tables, one for acetylated salicylic acids and one for anilines.
The BBs were characterized with a large number of descriptors,
and the X blocks were correlated to the multi-Y response matrix
using PLS, resulting in six PLS score vectors for the salicylic
2.6. PLS-DA Classification of Inactive and Active Acety-
lated Salicylanilides. The Hi-PLS multivariate QSAR model
was good for predicting and describing the biologically active
compounds but not useful for predicting inactive compounds.
It was therefore of interest to have a model that could be used
to classify active and inactive compounds. To be able to
discriminate active from inactive compounds, a PLS-discrimi-
nant analysis (PLS-DA)49 model was made on the acetylated
salicylanilides. Twelve out of the 25 compounds (1a-3a, 5a-
7a, 14a, 16a-17a, 21a-23a, and 25a) were classified as active
(at least 40% T3S inhibition at 50 µM). The same methodology
utilized in the Hi-PLS QSAR modeling was used to establish
the PLS-DA, using the % light emission inhibition at 50 µM as
the response. The correlation of the building blocks to the
multi-Y response matrix resulted in 11 score vectors for the
anilines (RX2 ) 1.0, RY2 ) 0.72, Q2 ) -0.27) and the acetylated
2
acids (RY2 ) 0.61, Q2 ) -0.09) and seven for the anilines (RY
) 0.72, Q2 ) 0.28). Following a number of steps (see
2
Experimental Section), a resulting two-component model (RX
) 0.34, RY2 ) 0.96, Q2 ) 0.82) was derived. The procedure is
shown in Figure 3. The model showed good correlation between
experimental and calculated biological response (Figure 4a and
b). A number of attempts were made to establish a multivariate
QSAR, using more conventional methods. An attempt using the
DOOD parameters for the building blocks was made and did
not yield a PLS model, as it only gave a negative value on Q2.
Local PCA models were made on anilines and salicylic acids,
and PCA score vectors were extracted and combined into a new
data table. However, using the PCA score vectors in a PLS gave
the same result as that from using the DOOD parameters. The
structural variation found in the PCA models simply did not
correlate to the biological response. If a majority of the
descriptors have nothing to do with a relevant problem, they
introduce noise in which the relevant variables drown. The
underlying problem is that it is difficult to decide beforehand
which variables are important for the activity. A model could
be computed only when correlating properties of the whole
salicylic acids (RX ) 1.0, RY ) 0.62, Q2 ) -0.28). The
2
2
2
2
resulting one-component PLS-DA model (RX ) 0.19, RY
)
0.75, Q2 ) 0.65) correctly discriminated between the two classes
of compounds (Figure 5). The model contained 15 linear terms,
1 square term, and 9 interaction terms (Figure 6). The first score
vector for the anilines and the first score vector for the salicylic
acids had large positive coefficients and shared a positive
interaction term. These terms were the most important, as
indicated by the size of their coefficients, and some of the
information contained in these terms was also found in the other
linear terms. Interpretation of the first linear term of the anilines
indicated that the aniline should have high density, rigidity, and

6182 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
Figure 4. Observed versus calculated plots for (a) 10 and (b) 20 µM compound concentrations, showing good correlations.
to the fact that the BBs used in the compound had been used in
more salicylanilides than the BBs used in 29a, while compound
27a did not have any BB present in any salicylanilide that the
models were based on.
2.8. Bacterial Growth. Earlier results showed that compound
1a did not affect growth at concentrations where T3S was
completely inhibited.19 It was previously found that the salicy-
lanilide 1b affected growth at much lower concentrations than
those of 1a.25 To verify that the light emission inhibition was
due to an effect on the T3S and not on growth, growth
experiments were performed on three acetylated compounds (6a,
23a, and 28a; Tables 1 and 3) and their unacetylated counter-
parts (6b, 23b, and 28b; Tables 1 and 3). Growth inhibition
was tested on Y. pseudotuberculosis, and experiments were
reproduced on at least two separate occasions (see Supporting
Information). Four of the compounds (23a, 23b, 28a, and 28b)
did not show any effect on bacterial growth at concentrations
at least 10 times higher than their estimated light emission IC50
values. Compound 6a had a slight effect on bacterial growth at
10 µM concentration but not nearly as serious of an effect as
that of compound 6b, which is in accordance with earlier results
for 1a and 1b.25 To see if there was any general bacterial growth
inhibition, an additional experiment was performed on Entero-
toxigenic Escherichia coli (ETEC), a Gram-negative bacterium
that lacks a T3S system. Growth of ETEC in the absence and
presence of the inhibitors was similar to the growth observed
in Y. pseudotuberculosis (data not shown). The fact that
unacetylated salicylanilides have been reported as ATP synthesis
inhibitors50 and that two (1b and 6b) out of four unacetylated
compounds tested for growth inhibition showed higher toxicity
than their acetylated counterparts led us to focus on the
acetylated compounds for the multivariate modeling, as the only
interest was on compounds acting on T3S (see discussion in
section 2.9.).
2.9. Possible Mode of Action. The bacterial growth experi-
ment results (see section 2.8.) indicated that the acetylated
compounds 1a and 6a might have had another mode of action
than that of the unacetylated compounds 1b and 6b, especially
considering the fact that unacetylated salicylanilides have been
reported to disrupt the proton motive force required for ATP
synthesis.50,51 A low pKa of the phenol along with electron-
withdrawing substituents in both ring systems was important
for salicylanilides to be effective proton motive force uncou-
plers.50 Looking at the structures of the compounds tested for
growth inhibition, the unacetylated compounds with two
electron-withdrawing iodines in the salicylic acid ring (1b and
6b) give rise to the most noticeable inhibition. This is probably
Figure 5. PLS-DA score plot of active (boxes) versus inactive
compounds (triangles).
dipole moment, negatively charged surfaces, and low HOMO
and LUMO energies. The first linear term of the salicylic acids
indicated that in order for the compounds to be active, they
should have a salicylic acid with low HOMO, high density,
dipole moment, and hardness, and large partial charges. The
other main terms had smaller coefficients, but their interaction
and square terms had large coefficients, showing that the model
contained several nonlinearities.
2.7. External Validation of Models. Three acetylated
salicylanilides (27a-29a; Table 3) were designed and synthe-
sized to validate the multivariate QSAR model and the PLS-
DA model. The predicted % inhibition of light emission for
the compounds 27a-29a is listed in Table 3. The three
compounds were mainly selected based on their predicted T3S
inhibitory effect, after 320 compounds had been predicted using
both of the multivariate models (see Supporting Information).
First, the compounds were predicted, using the PLS-DA model
to see if they would be predicted as active. The compounds
predicted as active were then predicted using the Hi-PLS
multivariate QSAR model. The three compounds chosen for
validating the models all had different substitution patterns and
different predicted biological activity. Compound 27a was
selected primarily due to the fact that none of the BBs in it had
been used in any previously synthesized acetylated salicylanilide
and that it was predicted as a potent inhibitor. The PLS-DA
model correctly predicted the three compounds to belong to the
class of active compounds. The Hi-PLS multivariate QSAR
model could correctly rank the three compounds. Compound
28a was most accurately predicted, and this is most likely due

Structure-ActiVity Relationship of Salicylanilides
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6183
Figure 6. PLS-DA coefficient plot with 15 linear, 1 square, and 9 interaction terms.
Table 3. Luciferase Light Emission Inhibitory Values and Predicted Inhibition for External Test Compounds
a For the compound ID’s, an a denotes R ) Ac, and a b denotes R ) H. b Means and standard deviations are calculated from triplicates, and experiments
were reproduced on at least two separate occasions.
due to the fact that the pKa should be lower in these compounds
and is probably the reason why growth inhibition is not observed
for the other compounds tested (23b and 28b). The acetyl group
of compounds 1a and 6a might be important to circumvent the
proton motive force uncoupling, but it is also possible that 1a
and 6a act as prodrugs and are metabolized into 1b and 6b,
which have the actual biological effect. It has earlier been
reported that unacetylated salicylanilides, with highly similar
substitution patterns to those of the compounds presented in
this paper, inhibit two-component systems (TCSs) of Gram-
positive bacteria.51 The compounds inhibit TCSs by targeting
the sensor kinase and causing aggregation through structural
alterations of the kinase.52 It was shown that the Gram-negative
bacteria E. coli was effected by the salicylanilides, but only
after treating the bacteria with a membrane-permeabilizing agent.
After this treatment, it was possible to observe levels of
antibacterial activity comparable to those manifested against
Gram-positive organisms. These results indicated that Gram-
negative bacteria contain the same TCS target as that of the
Gram-positive bacteria but that the compounds had problems
penetrating the outer membrane. There are a number of TCSs
in Y. pseudotuberculosis, but to the best of our knowledge, no
connection between a TCS and T3S has previously been
reported. Interestingly, a link between TCSs and T3S has been
established in Pseudomonas syringae and Salmonella.53,54
TCSs control virulence responses of a wide variety of
bacterial pathogens.55 Compounds targeting bacterial virulence
have a high chance of having an effect on resistant strains, and
there is a possibility that virulence inhibitors will be effective
for a long time before resistance becomes an issue. Research
on finding new antibacterial agents has traditionally focused
on inhibiting bacterial growth through various means (e.g.,
inhibiting metabolic pathways or disrupting cell wall formation),
a strategy not successful in delivering effective novel drugs
against diseases caused by resistant pathogens.56 This highlights
the importance of finding and developing inhibitors for novel
virulence targets.
3. Conclusions
Using three cycles of iterative selection, synthesis, and
biological testing, 50 analogues to 1a were synthesized and
biologically evaluated. First, five compounds (1b-4) were
selected and synthesized as the basis for a preliminary SAR
study. The SAR study was followed by a cherry-picking from
a PCA score plot that was used to select 16 new unacetylated
salicylanilides (5b-20b), which were synthesized and then
acetylated to yield another set of 16 compounds (5a-20a). In
order to ensure a sufficient number of active compounds, a
DOOD was utilized to design five new unacetylated salicyla-
nilides (21b-26b). The compounds from the DOOD were
synthesized and, in a second step, acetylated to yield the last
five compounds (21a-26a). After all compounds had been
biologically evaluated, a Hi-PLS multivariate QSAR model was
established, showing a good correlation between experimental
and predicted data for the biologically active acetylated com-
pounds. As a complement to the multivariate QSAR model, a

6184 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
where PRESS is the predicted error sum of squares when all objects
have been left out once and SS is the total sum of squares of Y
corrected for the mean.
PLS-DA model was made that could discriminate between active
and inactive acetylated compounds. To test the models predictive
capability, three new acetylated compounds (27a-29a) along
with their unacetylated counterparts (27b-29b) were synthe-
sized. Predictions proved to be accurate as the PLS-DA model
correctly could predict the compounds to be biologically active,
while the multivariate QSAR model could correctly rank the
three compounds.
4.4. PLS-DA. The same methodology, utilizing the exact same
pretreatment of BBs and the same descriptor set, was used for the
PLS-DA as that in the Hi-PLS MQSAR model. The criterion for
activity was set to 40% light emission inhibition at 50 µM. BBs
used in all compounds (1a-26b) were correlated to the response
matrix using PLS. After score vector extraction, forming of a new
table, and a new PLS regression, square and interaction terms were
calculated. Coefficients were removed using another set of rules
(see Supporting Information).
4.5. Biological Evaluation, Strains, and Growth Conditions.
The biological evaluation was performed with Yersinia pseudotu-
berculosis serotype III (YPIII) strain pIB29 (yopE-luxAB).19 The
bacteria were grown at room temperature on LB plates, as described
elsewhere.19,60
4.6. General Screening and Assay Conditions. The experi-
mental procedures were carried out essentially as described
before.19,21 The Yersinia pseudotuberculosis strain YPIII-pIB29
(yopE-luxAB) was grown over night in BHI (brain heart infusion)
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 compounds
dissolved in DMSO to be tested were dispensed in four different
concentrations (100, 50, 20 and 10 µM final concentration) to the
plates manually in triplicate. The final DMSO concentrations in
the biological assays were kept at 1% or below. To each well, 50
µL of diluted bacteria was added with a dispenser instrument (Multi
Drop, Thermo Lab Systems). The plates were incubated at ambient
temperature on a rotary shaker for 1 h, followed by incubation at
37 °C for 2 h. Within 30 min after incubation, 50 µL of fresh
decanal solution (10 µL/100 mL water) was added to the wells,
and the luciferase activity was measured immediately in the a
microplate reader (TECAN GENios, Gain 150; integration time of
20 ms). Experiments were run in triplicate and reproduced on at
least two separate occasions.
4.7. Growth Inhibition Experiments. The experimental pro-
cedures were carried out as described before.19,21 Growth inhibition
was measured by growing bacteria at 37 °C in the presence of
different compound concentrations in 96-well plates containing 100
µL of bacterial culture medium diluted to an OD600 of 0.1 in BHI
medium with 2.5 mM CaCl2. The experiments were carried out by
reading of the absorbance at 595 nm in a TECAN GENios plate
reader once an hour and by shaking carried out in a rotary shaker
at 37 °C. Growth rates were followed for up to 7 h.
4. Experimental Section
4.1. Cherry-Picking. ChemFinderACX2002Prod57 was used to
search for commercially available salicylic acids and anilines. BBs
were chosen manually based on availability, pricing, chemical
compatibility, and substitution pattern. Compounds with large
substituents at the ortho position to the carboxylic acid were filtered
out. The BBs were created in MOE58 and enumerated to produce
a library of 550 virtual products. 1D and 2D descriptors were
calculated using the same software (see Supporting Information).
The data set was imported to Simca 9.0,59 and data was centered,
unit-variance-scaled, and log-transformed. A PCA was done to
compress the data, followed by a variable selection. Data was again
centered, unit-variance-scaled, and log-transformed. Variables with
low contribution to R2 and Q2 were iteratively removed, and the
descriptor set was thereby reduced from 110 to 38. The number of
significant components in the model was determined by cross
validation using seven cross-validation rounds as implemented in
Simca 9.0.59
4.2. Complementary Design using DOOD. BBs were charac-
terized with 1D and 2D descriptors using MOE (see Supporting
Information). The BBs utilized to generate the virtual products for
the DOOD were the anilines and salicylic acids employed in the
synthesized compounds from the SAR and cherry-picking selec-
tions, plus a few additional BBs (see Supporting Information). The
descriptors were mainly describing different surfaces, electronic
properties, lipophilicity, and size. PCA was performed on both
salicylic acids and anilines to compress the variation in the data
set. The first three score values (t1, t2, and t3) for each BB group,
in combination with the descriptors SlogP and total polar surface
area for the entire product, comprised the candidate set variables
used for the DOOD selection. The three compounds with the highest
biological activity were set as vertices in an inner shell. The middle
of the vertices was set as the centerpoint in the design. The candidate
set was divided into layers so that the thickness of the two following
outer layers was equal to 10% of the thickness of the inner layer.
The earlier-synthesized compounds were set as inclusions in the
design, and new compounds were selected by the DOOD meth-
odology as applied in an earlier report by Olsson et al.46
4.3. Characterization of Building Blocks for Multivariate
QSAR Modeling. The BBs were created in MOE, and a rough
conformational search was performed by systematically rotating
all rotatable bonds 60°. All conformations were energy-minimized
with the implemented MMFF94 force field in MOE, and the
conformations with the lowest energy were saved. BBs were
characterized with a large number of 1D, 2D, and 3D descriptors
from MOE in addition to a number of descriptors from Spartan
(Linux/Unix 02 edition, AM1 calculations) (see Supporting Infor-
mation). A PLS regression was made of the data matrices describing
the anilines or salicylic acids to the multi-Y response matrix in
such a way that all BBs were listed together with their inhibition
values obtained for the concentrations of 10 and 20 µM. Thus, a
BB used in four active compounds would be listed with four
different inhibition values. The score vectors were extracted and
combined into a new data matrix. A new PLS regression was made
to correlate the X block to the multi-Y matrix. Interaction terms
and square terms were calculated, and using a set of rules (see
Supporting Information), coefficients were selectively removed.
The cross-validated predicted variation (Q2) was calculated using
seven cross-validation rounds according to
4.8. ETEC Growth Inhibition Experiments. Growth inhibition
was measured by growing ETEC bacteria at 37 °C in the presence
of different compound concentrations in 96-well plates containing
100 µL of bacterial culture medium diluted to an OD600 of 0.1 in
LB medium. The experiments were carried out by reading of the
absorbance at 595 nm in a TECAN GENios plate reader once an
hour and by shaking carried out in a rotary shaker at 37 °C. Growth
rates were followed for up to 24 h.
1
4.9. General Chemistry. H NMR spectra were recorded on a
Bruker DRX-400 in CDCl3, CD3OD, or (CD3)2SO. Mass spectra
were recorded by detecting negative (ES-) molecular ions with an
electrospray Waters Micromass ZG 2000 instrument using an
XTerra MS C18 5 µm 4.6 × 50 mm column and a H2O/acetonitrile/
formic acid eluent system. The same LC system was also used for
purification with a preparative XTerra Prep MS C18 5 µm 19 × 50
mm column and a H2O/acetonitrile eluent system. Microwave-
heated reactions were performed in Emrys process vials (5 mL)
using a SmithCreator microwave instrument from Biotage. TLC
was performed on Silica gel 60 F254 (Merck) with detection by UV
1
light. H NMR spectra were recorded using a 400 MHz Bruker
spectrometer.
4.10. Typical Procedure for the Formation of Salicylanilides.
Salicylic acid (0.7-1.0 mmol, 1 equiv) and aniline (0.8 equiv) were
suspended in toluene (5 mL) followed by addition of PCl3 (1 equiv).
The reaction was performed using a microwave instrument at
Q2 ) 1 - PRESS/SS

Structure-ActiVity Relationship of Salicylanilides
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6185
150 °C for 10 min. The reaction mixture was dissolved with EtOAc,
washed with water followed by NaHCO3 (aq., sat.), and extracted
with EtOAc. The organic phase was dried using anhydrous MgSO4
and filtered. After absorption onto silica gel, the crude products
were purified using flash chromatography with heptane/EtOAc as
the eluent.
7.24-7.08 (m, 3H), 2.43 (s, 3H). LCMS (m/z): [M - H-]- calcd
for [C15H10Br2FNO3], 427.89; found [-Ac+], 385.99.
N-(2-Fluorophenyl)-3,5-dibromo-2-hydroxybenzamide (7b).
1H NMR (DMSO): δ 10.91 (bs, 1H), 8.28 (d, J ) 2.3 Hz, 1H),
8.04 (d, J ) 2.2 Hz, 1H), 7.68-7.61 (m, 1H), 7.39-7.31 (m, 2H),
7.30-7.24 (m, 1H). LCMS (m/z): [M - H-]- calcd for [C13H8-
Br2FNO2], 385.88; found, 386.03.
4.11. Typical Procedure for Acetylation of Salicylanilides. To
a round-bottom flask (25 mL) containing the salicylanilide, excess
of acetic acid anhydride (15 mL) and a few droplets of phosphoric
acid (85%, aq.) were added. The reaction was performed at 70 °C
for 1 h. Addition of ice and water under cooling generally resulted
in a precipitate of the product. Flash chromatography (silica gel
using heptane/EtOAc systems as the eluent) or preparative LC-
MS was used if no pure precipitate of the product could be isolated.
1
N-(4-Heptylphenyl)-2-acetoxy-3,5-diiodobenzamide (8a). H
NMR (DMSO): δ 10.34 (s, 1H), 8.36 (d, J ) 2.0 Hz, 1H), 7.96
(d, J ) 2.0 Hz, 1H), 7.53 (d, J ) 8.5 Hz, 2H), 7.15 (d, J ) 8.5 Hz,
2H), 2.57-2.49 (m, 2H), 2.23 (s, 3H), 1.63-1.47 (m, 2H), 1.36-
1.17 (m, 8H), 0.92-0.79 (m, 3H). LCMS (m/z): [M - H-]- calcd
for [C22H25I2NO3], 603.98; found [-Ac+], 562.09.
N-(4-Heptylphenyl)-2-hydroxy-3,5-diiodobenzamide (8b). 1H
NMR (CDCl3): δ 12.90 (s, 1H), 8.18 (s, 1H), 7.84 (s, 1H), 7.77
(s, 1H), 7.44 (d, J ) 8.1 Hz, 2H), 7.23 (d, J ) 8.1 Hz, 2H), 2.65-
2.56 (m, 2H), 1.67-1.53 (m, 2H), 1.39-1.28 (m, 8H), 0.93-0.83
(m, 3H). LCMS (m/z): [M - H-]- calcd for [C20H23I2NO2], 561.97;
found, 562.09.
1
Compound purities, estimated from H NMR and LC UV traces,
are given in the Supporting Information.
N-(4-Chlorophenyl)-2-acetoxy-3,5-diiodobenzamide (1a). H
1
NMR (DMSO): δ 10.55 (s, 1H), 8.38 (d, J ) 2.0 Hz, 1H), 7.99
(d, J ) 2.0 Hz, 1H), 7.71-7.64 (m, 2H), 7.43-7.37 (m, 2H), 2.23
(s, 3H). LCMS (m/z): [M - H-]- calcd for [C15H10ClI2NO3],
539.84; found [-Ac+], 497.77.
N-(4-tert-Butylphenyl)-2-acetoxy-3,5-diiodobenzamide (9a). 1H
NMR (CDCl3): δ 8.25 (s, 1H), 8.01 (s, 1H), 7.70 (s, 1H), 7.47 (d,
J ) 8.3 Hz, 2H), 7.38 (d, J ) 8.4 Hz, 2H), 2.37 (s, 3H), 1.32 (s,
9H). LCMS (m/z): [M - H-]- calcd for [C19H19I2NO3], 561.94;
found [-Ac+], 520.04.
N-(4-tert-Butylphenyl)-2-hydroxy-3,5-diiodobenzamide (9b).
1H NMR (CDCl3): δ 12.91 (s, 1H), 8.19 (s, 1H), 7.84 (s, 1H),
7.77 (s, 1H), 7.51-7.39 (m, 4H), 1.33 (s, 9H). LCMS (m/z): [M
- H-]- calcd for [C17H17I2NO2], 519.93; found, 520.04.
N-(4-tert-Butylphenyl)-2-acetoxy-3,5,6-trichlorobenzamide (10a).
1H NMR (DMSO): δ 10.64 (s, 1H), 8.19 (s, 1H), 7.58-7.49 (m,
2H), 7.41-7.34 (m, 2H), 2.26 (s, 3H), 1.27 (s, 9H). LCMS (m/z):
[M - H-]- calcd for [C19H18Cl3NO3], 412.03; found [-Ac+],
370.10.
N-(4-Chlorophenyl)-2-hydroxy-3,5-diiodobenzamide (1b). 1H
NMR (DMSO): δ 13.02 (bs, 1H), 10.89 (bs, 1H), 8.35 (d, J ) 1.9
Hz, 1H), 8.21 (d, J ) 1.9 Hz, 1H), 7.73-7.68 (m, 2H), 7.49-7.43
(m, 2H). LCMS (m/z): [M - H-]- calcd for [C13H8ClI2NO2],
497.83; found, 497.92.
N-(4-Chlorophenyl)-2-acetoxy-5-iodobenzamide (2a). 1H NMR
(CDCl3): δ 8.13 (d, J ) 2.1 Hz, 1H), 7.96 (bs, 1H), 7.82 (dd, J )
2.2 Hz, 8.5 Hz, 1H), 7.53 (d, J ) 8.8 Hz, 1H), 7.38-7.30 (m, 2H),
6.92 (d, J ) 8.5 Hz, 1H), 2.32 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C15H11ClINO3], 413.94; found [-Ac+], 372.02.
N-(4-Chlorophenyl)-2-hydroxy-5-iodobenzamide (2b). 1H NMR
(DMSO): δ 11.70 (bs, 1H), 10.62 (bs, 1H), 8.15 (d, J ) 1.9 Hz,
1H), 7.77-7.66 (m, 3H), 7.46-7.40 (m, 2H), 6.81 (d, J ) 8.4 Hz,
1H). LCMS (m/z): [M - H-]- calcd for [C13H9ClINO2], 371.93;
found, 372.02.
N-(4-tert-Butylphenyl)-3,5,6-trichloro-2-hydroxybenzamide
1
(10b). H NMR (DMSO): δ 10.56 (bs, 1H), 10.49 (s, 1H), 7.84
(s, 1H), 7.60-7.56 (m, 2H), 7.39-7.34 (m, 2H), 1.27 (s, 9H).
LCMS (m/z): [M - H-]- calcd for [C17H16Cl3NO2], 370.02; found,
370.10.
N-(4-Chlorophenyl)-2-acetoxybenzamide (3a). 1H NMR
(CDCl3): δ 8.01 (s, 1H), 7.84 (d, J ) 7.6 Hz, 1H), 7.60-7.50 (m,
3H), 7.41-7.30 (m, 3H), 7.16 (d, J ) 8.1 Hz, 1H), 2.33 (s, 3H).
LCMS (m/z): [M - H-]- calcd for [C15H12ClNO3], 288.04; found
[-Ac+], 246.11.
N-(3,5-Dimethylphenyl)-2-acetoxy-3,5,6-trichlorobenzamide
1
(11a). H NMR (CDCl3): δ 7.64 (s, 1H), 7.34 (s, 1H), 7.19 (s,
2H), 6.84 (s, 1H), 2.33 (s, 6H), 2.30 (s, 3H). LCMS (m/z): [M -
H-]- calcd for [C17H14Cl3NO3], 384.00; found [-Ac+], 342.06.
N-(3,5-Dimethylphenyl)-3,5,6-trichloro-2-hydroxybenza-
mide (11b). 1H NMR (DMSO): δ 10.55 (bs, 1H), 10.42 (bs, 1H),
7.83 (s, 1H), 7.29 (s, 2H), 6.76 (s, 1H), 2.25 (s, 6H). LCMS
(m/z): [M - H-]- calcd for [C15H12Cl3NO2], 341.99; found,
342.06.
N-(4-Chlorophenyl)-2-hydroxybenzamide (3b). 1H NMR
(DMSO): δ 11.63 (bs, 1H), 10.49 (s, 1H), 7.95-7.90 (m, 1H),
7.80-7,73 (m, 2H), 7.46-7.40 (m, 3H), 7.01-6.93 (m, 2H). LCMS
(m/z): [M - H-]- calcd for [C13H10ClNO2], 246.03; found, 246.11.
N-(4-Chlorophenyl)-3,5-difluorobenzamide (4). 1H NMR
(DMSO): δ 10.48 (s, 1H), 7.82-7.77 (m, 2H), 7.71-7.64 (m, 2H),
7.58-7.59 (m, 1H), 7.46-7.41 (m, 2H). LCMS (m/z): [M - H-]-
calcd for [C13H8ClF2NO], 266.02; found, 266.10.
N-(3,5-Di-tert-butylphenyl)-2-acetoxy-5-bromobenzamide (12a).
1H NMR (CDCl3): δ 8.09-7.92 (m, 2H), 7.61 (dd, J ) 2.4 Hz,
8.6 Hz, 1H), 7.44 (s, 2H), 7.26-7.24 (m, 1H), 7.06 (d, J ) 8.6 Hz,
1H), 2.35 (s, 3H), 1.34 (s, 18H). LCMS (m/z): [M - H-]- calcd
for [C23H28BrNO3], 444.12; found [-Ac+], 402.07.
N-Phenyl-2-acetoxy-3,5-diiodobenzamide (5a). 1H NMR
(CDCl3): δ 8.26 (d, J ) 2.1 Hz, 1H), 8.02 (d, J ) 2.0 Hz, 1H),
7.73 (s, 1H), 7.56 (d, J ) 8.2 Hz, 2H), 7.40-7.34 (m, 2H), 7.19-
7.15 (m, 1H), 2.37 (s, 3H). LCMS (m/z): [M - H-]- calcd for
[C15H11I2NO3], 505.87; found [-Ac+], 463.97.
N-(3,5-Di-tert-butylphenyl)-5-bromo-2-hydroxybenzamide (12b).
1H NMR (CDCl3): δ 7.85 (s, 1H), 7.66 (d, J ) 2.3 Hz, 1H), 7.51
(dd, J ) 2.3 Hz, 8.9 Hz, 1H), 7.42 (d, J ) 1.6 Hz, 2H), 7.31-7.29
(m, 1H), 6.93 (d, J ) 8.9 Hz, 1H), 1.35 (s, 18H). LCMS (m/z):
[M - H-]- calcd for [C21H26BrNO2], 402.11; found, 402.22.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-acetoxy-3-isopro-
N-Phenyl-2-hydroxy-3,5-diiodobenzamide (5b). 1H NMR
(DMSO): δ 13.24 (bs, 1H), 10.77 (bs, 1H), 8.39 (d, J ) 1.9 Hz,
1H), 8.21 (d, J ) 1.8 Hz, 1H), 7.70-7.63 (m, 2H), 7.44-7.36 (m,
2H), 7.22-7.16 (m, 1H). LCMS (m/z): [M - H-]- calcd for
[C13H9I2NO2], 463.86; found, 463.89.
1
pylbenzamide (13a). H NMR (CDCl3): δ 8.42 (s, 1H), 8.37 (s,
N-(4-Trifluoromethylsulfanylphenyl)-2-acetoxy-3,5-diiodo-
1H), 7.64-7.58 (m, 1H), 7.50-7.45 (m, 1H), 7.33 (t, J ) 7.7 Hz,
1H), 6.88 (s, 1H), 3.87 (s, 3H), 3.10-3.01 (m, 1H), 2.34 (s, 3H),
2.32 (s, 3H), 1.25 (s, 3H), 1.23 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C20H22ClNO4], 374.05; found [-Ac+], 332.22.
1
benzamide (6a). H NMR (CDCl3): δ 8.29 (d, J ) 2.1 Hz, 1H),
8.00 (d, J ) 2.1 Hz, 1H), 7.88 (s, 1H), 7.70-7.61 (m, 4H), 2.38
(s, 3H). LCMS (m/z): [M - H-]- calcd for [C16H10F3I2NO3S],
605.83; found [-Ac+], 563.93.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-hydroxy-3-isopro-
1
N-(4-Trifluoromethylsulfanylphenyl)-2-hydroxy-3,5-diiodo-
benzamide (6b). 1H NMR (DMSO): δ 12.73 (bs, 1H), 10.99 (bs,
1H), 8.33 (d, J ) 1.8 Hz, 1H), 8.23 (d, J ) 1.6 Hz, 1H), 7.85 (d,
J ) 8.7 Hz, 2H), 7.75 (d, J ) 8.7 Hz, 2H). LCMS (m/z): [M -
H-]- calcd for [C14H8F3I2NO2S], 563.82; found, 563.74.
N-(2-Fluorophenyl)-2-acetoxy-3,5-dibromobenzamide (7a). 1H
NMR (CDCl3): δ 8.47-8.34 (m, 2H), 8.05 (s, 1H), 7.91 (s, 1H),
pylbenzamide (13b). H NMR (CDCl3): δ 8.55 (s, 1H), 8.29 (s,
1H), 7.42-7.33 (m, 2H), 6.94-6.85 (m, 2H), 3.92 (s, 3H), 3.48-
3.35 (m, 1H), 2.36 (s, 3H), 1.26 (s, 3H), 1.25 (s, 3H). LCMS
(m/z): [M - H-]- calcd for [C18H20ClNO3], 332.11; found,
332.29.
N-(4-Fluorophenyl)-2-acetoxy-3,5-diiodobenzamide (14a). 1H
NMR (DMSO): δ 10.48 (s, 1H), 8.38 (d, J ) 2.0 Hz, 1H), 7.99

6186 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
(d, J ) 2.1 Hz, 1H), 7.71-7.64 (m, 2H), 7.23-7.15 (m, 2H), 2.23
(s, 3H). LCMS (m/z): [M - H-]- calcd for [C15H10FI2NO3],
523.87; found [-Ac+], 481.96.
N-(4-Trifluoromethylsulfanylphenyl)-3,5,6-trichloro-2-hy-
droxybenzamide (21b). H NMR (DMSO): δ 10.97 (bs, 1H),
10.74 (s, 1H), 7.88 (s, 1H), 7.86-7.80 (m, 2H), 7.75-7.70 (m,
2H). LCMS (m/z): [M - H-]- calcd for [C14H7Cl3F3NO2S],
413.91; found, 413.98.
1
N-(4-Fluorophenyl)-2-hydroxy-3,5-diiodobenzamide (14b). 1H
NMR (MeOD): δ 8.27 (d, J ) 2.0 Hz, 1H), 8.19 (d, J ) 2.0 Hz,
1H), 7.68-7.63 (m, 2H), 7.15-7.07 (m, 2H). LCMS (m/z): [M -
H-]- calcd for [C13H8FI2NO2], 481.85; found, 481.96.
N-(5-Chloro-2-fluorophenyl)-2-acetoxy-3,5-diiodobenza-
1
mide (22a). H NMR (CDCl3): δ 8.50 (d, J ) 6.2 Hz, 1H), 8.31
N-Phenyl-2-acetoxy-3,5,6-trichlorobenzamide (15a). 1H NMR
(MeOD): δ 7.89 (s, 1H), 7.61 (d, J ) 7.9 Hz, 2H), 7.42-7.32 (m,
2H), 7.23-7.13 (m, 1H), 2.30 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C15H10Cl3NO3], 355.96; found [-Ac+], 314.07.
N-Phenyl-3,5,6-trichloro-2-hydroxybenzamide (15b). 1H NMR
(DMSO): δ 10.84-10.49 (m, 2H), 7.83 (s, 1H), 7.71-7.65 (m,
2H), 7.38-7.32 (m, 2H), 7.14-7.08 (m, 1H). LCMS (m/z): [M -
H-]- calcd for [C13H8Cl3NO2], 313.95; found, 314.05.
(s, 1H), 8.28 (s, 1H), 8.17 (s, 1H), 7.13-7.04 (m, 2H), 2.42 (s,
3H). LCMS (m/z): [M - H-]- calcd for [C15H9ClFI2NO3], 557.83;
found [-Ac+], 515.76.
N-(5-Chloro-2-fluorophenyl)-2-hydroxy-3,5-diiodobenza-
mide (22b). 1H NMR (MeOD): δ 8.31-8.18 (m, 2H), 7.89-7.80
(m, 1H), 7.31-7.17 (m, 2H). LCMS (m/z): [M - H-]- calcd for
[C13H7ClFI2NO2], 515.82; found, 515.76.
N-(4-Bromophenyl)-2-acetoxy-5-iodobenzamide (23a). 1H NMR
(DMSO): δ 10.53 (s, 1H), 7.99 (d, J ) 2.0 Hz, 1H), 7.95-7.89
(m, 1H), 7.65 (d, J ) 8.9 Hz, 2H), 7.53 (d, J ) 8.9 Hz, 2H), 7.09
(d, J ) 8.5 Hz, 1H), 2.18 (s, 3H). LCMS (m/z): [M - H-]- calcd
for [C15H11BrINO3], 457.89; found [-Ac+], 416.00.
1
N-(2-Fluorophenyl)-2-acetoxy-4-methylbenzamide (16a). H
NMR (CDCl3): δ 8.74 (s, 1H), 8.58-8.49 (m, 1H), 7.97 (d, J )
7.9 Hz, 1H), 7.22-7.04 (m, 4H), 7.01 (s, 1H), 2.42 (s, 3H), 2.41
(s, 3H). LCMS (m/z): [M - H-]- calcd for [C16H14FNO3], 286.09;
found [-Ac+], 244.21.
N-(4-Bromophenyl)-2-hydroxy-5-iodobenzamide (23b). 1H
NMR (DMSO): δ 11.70 (bs, 1H), 10.61 (bs, 1H), 8.14 (d, J ) 2.2
Hz, 1H), 7.73-7.65 (m, 3H), 7.58-7.53 (m, 2H), 6.81 (d, J ) 8.8
Hz, 1H). LCMS (m/z): [M - H-]- calcd for [C13H9BrINO2],
415.88; found, 416.00.
N-(2-Fluorophenyl)-2-hydroxy-4-methylbenzamide (16b). 1H
NMR (DMSO): δ 11.87 (bs, 1H), 10.67 (s, 1H), 8.23-8.16 (m,
1H), 7.91 (d, J ) 8.0 Hz, 1H), 7.35-7.28 (m, 1H), 7.26-7.14 (m,
2H), 6.86-6.78 (m, 2H), 2.30 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C14H12FNO2], 244.08; found, 244.16.
N-(4-Trifluoromethylsulfanylphenyl)-2-acetoxy-3,7-dibro-
monaphthamide (24a). 1H NMR (DMSO): δ 10.99 (s, 1H), 8.50
(d, J ) 2.0 Hz, 1H), 8.42 (s, 1H), 8.17 (d, J ) 9.0 Hz, 1H), 7.97
(dd, J ) 2.0 Hz, 9.0 Hz, 1H), 7.92-7.85 (m, 2H), 7.78-7.72 (m,
2H), 2.32 (s, 3H). LCMS (m/z): [M - H-]- calcd for [C20H12-
Br2F3NO3S], 559.88; found [-Ac+], 517.94.
N-(4-Trifluoromethylsulfanylphenyl)-3,7-dibromo-2-hydrox-
ynaphthamide (24b). 1H NMR (DMSO): δ 11.52 (bs, 1H), 11.17
(bs, 1H), 8.50 (s, 1H), 8.26 (d, J ) 2.0 Hz, 1H), 8.01 (d, J ) 9.1
Hz, 1H), 7.95-7.89 (m, 2H), 7.83 (dd, J ) 2.0 Hz, 9.1 Hz, 1H),
7.80-7.74 (m, 2H). LCMS (m/z): [M - H-]- calcd for [C18H10-
Br2F3NO2S], 517.87; found, 517.94.
N-(4-Bromophenyl)-2-acetoxy-3,5-diiodobenzamide (17a). 1H
NMR (CDCl3): δ 8.27 (s, 1H), 7.99 (s, 1H), 7.74 (s, 1H), 7.52-
7.41 (m, 4H), 2.36 (s, 3H). LCMS (m/z): [M - H-]- calcd for
[C15H10BrI2NO3], 583.79; found [-Ac+], 541.70.
N-(4-Bromophenyl)-2-hydroxy-3,5-diiodobenzamide (17b). 1H
NMR (DMSO): δ 12.98 (bs, 1H), 10.73 (bs, 1H), 8.36 (d, J ) 2.0
Hz, 1H), 8.23 (d, J ) 1.9 Hz, 1H), 7.68-7.63 (m, 2H), 7.62-7.56
(m, 2H). LCMS (m/z): [M - H-]- calcd for [C13H8BrI2NO2],
541.77; found, 541.85.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-acetoxy-3,5-di-
iodobenzamide (18a). 1H NMR (CDCl3): δ 8.45 (s, 1H), 8.36 (s,
1H), 8.27 (d, J ) 2.1 Hz, 1H), 8.12 (d, J ) 2.0 Hz, 1H), 6.90 (s,
1H), 3.89 (s, 3H), 2.37 (s, 3H), 2.34 (s, 3H). LCMS (m/z): [M -
H-]- calcd for [C17H14ClI2NO4], 583.86; found [-Ac+], 541.77.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-hydroxy-3,5-di-
iodobenzamide (18b). 1H NMR (DMSO): δ 13.18 (bs, 1H), 10.50
(bs, 1H), 8.39 (s, 1H), 8.23 (s, 1H), 7.53 (s, 1H), 7.19 (s, 1H), 3.82
(s, 3H), 2.27 (s, 3H). LCMS (m/z): [M - H-]- calcd for [C15H12-
ClI2NO3], 541.85; found, 542.00.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-acetoxy-5-methyl-
benzamide (19a). 1H NMR (CDCl3): δ 8.76 (s, 1H), 8.47 (s, 1H),
7.77 (d, J ) 2.0 Hz, 1H), 7.34-7.29 (m, 1H), 7.04 (d, J ) 8.2 Hz,
1H), 6.90 (s, 1H), 3.89 (s, 3H), 2.41 (s, 3H), 2.36 (s, 3H), 2.34 (s,
3H). LCMS (m/z): [M - H-]- calcd for [C18H18ClNO4], 346.08;
found [-Ac+], 304.20.
N-(4-Chloro-2-methoxy-5-methylphenyl)-2-hydroxy-5-meth-
ylbenzamide (19b). 1H NMR (DMSO): δ 11.45 (s, 1H), 10.79 (s,
1H), 8.39 (s, 1H), 7.81 (d, J ) 2.0 Hz, 1H), 7.26-7.20 (m, 1H),
7.14 (s, 1H), 6.20 (d, J ) 8.3 Hz, 1H), 3.88 (s, 3H), 2.28 (s, 3H),
2.27 (s, 3H). LCMS (m/z): [M - H-]- calcd for [C16H16ClNO3],
304.07; found, 304.15.
N-(4-tert-Butylphenyl)-2-acetoxy-5-fluorobenzamide (20a). 1H
NMR (CDCl3): δ 8.01 (s, 1H), 7.58 (dd, J ) 2.9 Hz, 5.5 Hz, 1H),
7.54-7.48 (m, 2H), 7.42-7.37 (m, 2H), 7.24-7.17 (m, 1H), 7.16-
7.11 (m, 1H), 2.34 (s, 3H), 1.32 (s, 9H). LCMS (m/z): [M - H-]-
calcd for [C19H20FNO3], 328.13; found [-Ac+], 286.25.
N-(4-tert-Butylphenyl)-5-fluoro-2-hydroxybenzamide (20b).
1H NMR (DMSO): δ 12.99 (bs, 1H), 7.63-7.53 (m, 3H), 7.33 (d,
J ) 8.6 Hz, 2H), 7.08-7.00 (m, 1H), 6.74-6.66 (m, 1H), 1.28 (s,
9H). LCMS (m/z): [M - H-]- calcd for [C17H18FNO2], 286.12;
found, 286.25.
N-(4-Trifluoromethylsulfanylphenyl)-2-acetoxy-3-chloroben-
1
zamide (25a). H NMR (CDCl3): δ 8.04 (s, 1H), 7.68-7.59 (m,
6H), 7.37-7.28 (m, 1H), 2.38 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C16H11ClF3NO3S], 388.00; found [-Ac+], 346.08.
N-(4-Trifluoromethylsulfanylphenyl)-3-chloro-2-hydroxyben-
zamide (25b). 1H NMR (DMSO): δ 12.25 (s, 1H), 10.84 (s, 1H),
7.96 (d, J ) 8.8 Hz, 1H), 7.89 (d, J ) 8.6 Hz, 2H), 7.75 (d, J )
8.6 Hz, 2H), 7.67 (d, J ) 8.8 Hz, 1H), 7.06-6.97 (m, 1H). LCMS
(m/z): [M - H-]- calcd for [C14H9ClF3NO2S], 345.99; found,
346.09.
N-(4-Fluorophenyl)-2-acetoxy-3,5,6-trichlorobenzamide (26a).
1H NMR (CDCl3): δ 7.66 (s, 1H), 7.56-7.48 (m, 2H), 7.43 (s,
1H), 7.12-7.03 (m, 2H), 2.30 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C15H9Cl3FNO3], 373.96; found [-Ac+], 332.06.
N-(4-Fluorophenyl)-3,5,6-trichloro-2-hydroxybenzamide (26b).
1H NMR (DMSO): δ 7.76-7.59 (m, 3H), 7.24-7.11 (m, 2H).
LCMS (m/z): [M - H-]- calcd for [C13H7Cl3FNO2], 331.94; found,
332.01.
N-(4-Chloro-3-methylphenyl)-2-acetoxy-5-chlorobenzamide
1
(27a). H NMR (CDCl3): δ 7.94 (s, 1H), 7.80 (d, J ) 2.5 Hz,
1H), 7.51 (s, 1H), 7.47 (dd, J ) 8.7 Hz, 6.1 Hz, 1H), 7.35-7.30
(m, 2H), 7.11 (d, J ) 8.7 Hz, 1H), 2.38 (s, 3H), 2.33 (s, 3H). LCMS
(m/z): [M - H-]- calcd for [C16H13Cl2NO3], 336.02; found
[-Ac+], 294.04.
N-(4-Chloro-3-methylphenyl)-5-chloro-2-hydroxybenza-
1
mide (27b). H NMR (DMSO): δ 11.75 (bs, 1H), 10.49 (s, 1H),
7.92 (d, J ) 2.7 Hz, 1H), 7.70 (d, J ) 2.2 Hz, 1H), 7.58 (dd, J )
2.4 Hz, 8.6 Hz, 1H), 7.46 (dd, J ) 2.7 Hz, 8.8 Hz, 1H), 7.40 (d, J
) 8.7 Hz, 1H), 7.01 (d, J ) 8.8 Hz, 1H), 2.34 (s, 3H). LCMS
(m/z): [M - H-]- calcd for [C14H11Cl2NO2], 294.01; found, 294.19.
N-(4-Trifluoromethylsulfanylphenyl)-2-acetoxy-5-iodobenza-
1
N-(4-Trifluoromethylsulfanylphenyl)-2-acetoxy-3,5,6-trichlo-
robenzamide (21a). 1H NMR (DMSO): δ 11.12 (s, 1H), 8.24 (s,
1H), 7.82-7.71 (m, 4H), 2.27 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C16H9Cl3F3NO3S], 455.92; found [-Ac+], 413.98.
mide (28a). H NMR (CDCl3): δ 8.31 (s, 1H), 8.04 (d, J ) 2.2
Hz, 1H), 7.77 (dd, J ) 8.5 Hz, 6.3 Hz, 1H), 7-59-7.64 (m, 4H),
6.89 (d, J ) 8.53 Hz, 1H), 2.31 (s, 3H). LCMS (m/z): [M - H-]-
calcd for [C16H11F3INO3S], 479.94; found [-Ac+], 438.11.

Structure-ActiVity Relationship of Salicylanilides
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6187
(18) Bailey, L.; Gylfe, A.; Sundin, C.; Muschiol, S.; Elofsson, M.;
Nordstrom, P.; Henriques-Normark, B.; Lugert, R.; Waldenstrom,
A.; Wolf-Watz, H.; Bergstrom, S. Small molecule inhibitors of type
III secretion in Yersinia block the Chlamydia pneumoniae infection
cycle. FEBS Lett. 2007, 581, 587-595.
(19) Kauppi, A. M.; Nordfelth, R.; Uvell, H.; Wolf-Watz, H.; Elofsson,
M. Targeting bacterial virulence: Inhibitors of type III secretion in
Yersinia. Chem. Biol. 2003, 10, 241-249.
(20) Muschiol, S.; Bailey, L.; Gylfe, A.; Sundin, C.; Hultenby, K.;
Bergstrom, S.; Elofsson, M.; Wolf-Watz, H.; Normark, S.; Henriques-
Normark, B. A small-molecule inhibitor of type III secretion inhibits
different stages of the infectious cycle of Chlamydia trachomatis.
Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 14566-14571.
(21) Nordfelth, R.; Kauppi, A. M.; Norberg, H. A.; Wolf-Watz, H.;
Elofsson, M. Small-molecule inhibitors specifically targeting type
III secretion. Infect. Immun. 2005, 73, 3104-3114.
(22) Wolf, K.; Betts, H. J.; Chellas-Gery, B.; Hower, S.; Linton, C. N.;
Fields, K. A. Treatment of Chlamydia trachomatis with a small
molecule inhibitor of the Yersinia type III secretion system disrupts
progression of the chlamydial developmental cycle. Mol. Microbiol.
2006, 61, 1543-1555.
N-(4-Trifluoromethylsulfanylphenyl)-2-hydroxy-5-iodobenza-
mide (28b). H NMR (MeOD): δ 8.28 (s, 1H), 7.84 (d, J ) 8.5
1
Hz, 2H), 7.75-7.64 (m, 3H), 6.80 (d, J ) 8.7 Hz, 1H). LCMS
(m/z): [M - H-]- calcd for [C14H9F3INO2S], 437.93; found,
437.96.
1
N-(4-Bromophenyl)-2-acetoxy-3-chlorobenzamide (29a). H
NMR (DMSO): δ 10.57 (s, 1H), 7.76 (dd, J ) 8.1 Hz, 6.6 Hz,
1H), 7.70-7.63 (m, 3H), 7.56-7.52 (m, 2H), 7.45 (t, J ) 7.9 Hz,
1H), 2.26 (s, 3H). LCMS (m/z): [M - H-]- calcd for [C15H11-
BrClNO3], 365.95; found [-Ac+], 324.07.
N-(4-Bromophenyl)-3-chloro-2-hydroxybenzamide (29b). 1H
NMR (MeOD): δ 7.88 (d, J ) 7.8 Hz, 1H), 7.64 (d, J ) 8.8 Hz,
2H), 7.56 (d, J ) 8.3 Hz, 1H), 7.52 (d, J ) 8.8 Hz, 2H), 6.96-
6.89 (m, 1H). LCMS (m/z): [M - H-]- calcd for [C13H9BrClNO2],
323.94; found, 324.00.
Acknowledgment. We thank the Swedish National Research
Council, the Swedish Governmental Agency for Innovation
Systems (VINNOVA), the Foundation for Technology Transfer
in Umeå, the Carl Trygger foundation, Innate Pharmaceuticals,
and Umeå Biotech Incubator for support.
(23) Bailey, L.; Gylfe, A.; Sundin, C.; Muschiol, S.; Elofsson, M.;
Nordstrom, P.; Henriques-Normark, B.; Lugert, R.; Waldenstrom,
A.; Wolf-Watz, H.; Bergstrom, S. Small molecule inhibitors of type
III secretion in Yersinia block the Chlamydia pneumoniae infection
cycle. FEBS Lett. 2007, 581, 587-595.
(24) Gauthier, A.; Robertson, M. L.; Lowden, M.; Ibarra, J. A.; Puente,
J. L.; Finlay, B. B. Transcriptional inhibitor of virulence factors in
enteropathogenic Escherichia coli. Antimicrob. Agents Chemother.
2005, 49, 4101-4109.
(25) Kauppi, A. M.; Nordfelth, R.; Hagglund, U.; Wolf-Watz, H.; Elofsson,
M. Salicylanilides are potent inhibitors of type III secretion in
Yersinia. AdV. Exp. Med. Biol. 2003, 529, 97-100.
Supporting Information Available: Information on BBs,
descriptors, and other variable selection criteria are given. A table
of compound purities, LC chromatograms for key compounds, and
1H NMR spectra of all compounds are also given. This material is
available free of charge via the Internet at http://pubs.acs.org.
(26) Lee, V. T.; Pukatzki, S.; Sato, H.; Kikawada, E.; Kazimirova, A.
A.; Huang, J.; Li, X. H.; Arm, J. P.; Frank, D. W.; Lory, S.
Pseudolipasin A is a specific inhibitor for phospholipase A(2) activity
of Pseudomonas aeruginosa cytotoxin ExoU. Infect. Immun. 2007,
75, 1089-1098.
(27) Tautz, L.; Bruckner, S.; Sareth, S.; Alonso, A.; Bogetz, J.; Bottini,
N.; Pellecchia, M.; Mustelin, T. Inhibition of Yersinia tyrosine
phosphatase by furanyl salicylate compounds. J. Biol. Chem. 2005,
280, 9400-9408.
References
(1) Levy, S. B. Antibiotic resistancesthe problem intensifies. AdV. Drug
DeliVery ReV. 2005, 57, 1446-1450.
(2) Poole, K. Multidrug resistance in Gram-negative bacteria. Curr. Opin.
Microbiol. 2001, 4, 500-508.
(3) Normark, S.; Nilsson, C.; Normark, B. H. MicrobiologysA pathogen
attacks while keeping up defense. Science 2005, 307, 1211-1212.
(4) Brundtland, G. H. OVercoming antimicrobial resistance. World
Health Report on Infectious Diseases; Geneva, Switzerland, 2000.
(5) Henry, C. M. Antibiotic resistance. Chem. Eng. News 2000, 78, 41-
58.
(28) Alksne, L. E.; Projan, S. J. Bacterial virulence as a target for
antimicrobial chemotherapy. Curr. Opin. Biotechnol. 2000, 11, 625-
636.
(6) Hawley, R. J.; Eitzen, E. M. Biological weaponssA primer for
microbiologists. Annu. ReV. Microbiol. 2001, 55, 235-253.
(7) 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, T. M.; Russell, P. K.; Schoch-Spana,
M.; Tonat, K. Plague as a biological weaponsMedical and public
health management. JAMA, J. Am. Med. Assoc. 2000, 283, 2281-
2290.
(8) Goldschmidt, R. M.; Macielag, M. J.; Hlasta, D. J.; Barrett, J. F.
Inhibition of virulence factors in bacteria. Curr. Pharm. Des. 1997,
3, 125-142.
(9) Lee, Y. M.; Almqvist, F.; Hultgren, S. J. Targeting virulence for
antimicrobial chemotherapy. Curr. Opin. Pharmacol. 2003, 3, 513-
519.
(29) Marra, A. Targeting virulence for antibacterial chemotherapy-
identifying and characterizing virulence factors for lead discovery.
Drugs R&D 2006, 7, 1-16.
(30) Becker, D.; Selbach, M.; Rollenhagen, C.; Ballmaier, M.; Meyer, T.
F.; Mann, M.; Bumann, D. Robust Salmonella metabolism limits
possibilities for new antimicrobials. Nature 2006, 440, 303-307.
(31) Burbaum, J. J.; Sigal, N. H. New technologies for high-throughput
screening. Curr. Opin. Chem. Biol. 1997, 1, 72-78.
(32) Hung, D. T.; Shakhnovich, E. A.; Pierson, E.; Mekalanos, J. J. Small-
molecule inhibitor of Vibrio cholerae virulence and intestinal
colonization. Science 2005, 310, 670-674.
(33) Jorgensen, W. L. The many roles of computation in drug discovery.
Science 2004, 303, 1813-1818.
(34) Linusson, A.; Gottfries, J.; Lindgren, F.; Wold, S. Statistical molecular
design of building blocks for combinatorial chemistry. J. Med. Chem.
2000, 43, 1320-1328.
(35) Linusson, A.; Gottfries, J.; Olsson, T.; O¨ rnskov, E.; Folestad, S.;
Norden, B.; Wold, S. Statistical molecular design, parallel synthesis,
and biological evaluation of a library of thrombin inhibitors. J. Med.
Chem. 2001, 44, 3424-3439.
(36) Martin, E. J.; Blaney, J. M.; Siani, M. A.; Spellmeyer, D. C.; Wong,
A. K.; Moos, W. H. Measuring diversitysExperimental-design of
combinatorial libraries for drug discovery. J. Med. Chem. 1995, 38,
1431-1436.
(37) Hansch, C. A quantitative approach to biochemical structure-activity
relationships. Acc. Chem. Res. 1969, 2, 232-239.
(10) Walsh, C. Where will new antibiotics come from? Nat. ReV.
Microbiol. 2003, 1, 65-70.
(11) Hueck, C. J. Type III protein secretion systems in bacterial pathogens
of animals and plants. Microbiol. Mol. Biol. ReV. 1998, 62, 379-
433.
(12) Perry, R. D.; Fetherston, J. D. Yersinia pestissEtiologic agent of
plague. Clin. Microbiol. ReV. 1997, 10, 35-66.
(13) Wren, B. W. The YersiniaesA model genus to study the rapid
evolution of bacterial pathogens. Nat. ReV. Microbiol. 2003, 1, 55-
64.
(14) Cornelis, G. R. Molecular and cell biology aspects of plague. Proc.
Natl. Acad. Sci. U.S.A. 2000, 97, 8778-8783.
(38) Wold, S.; Dunn, W. J., III. Multivariate quantitative structure-activity
relationships (QSAR): Conditions for their applicability. J. Chem.
Inf. Model. 1983, 23, 6-13.
(39) Eriksson, L.; Johansson, E. Multivariate design and modeling in
QSAR. Chemom. Intell. Lab. Syst. 1996, 34, 1-19.
(40) Larsson, A.; Johansson, S.; Pinkner, J. S.; Hultgren, S. J.; Almqvist,
F.; Kihlberg, J.; Linusson, A. Multivariate design, synthesis, and
biological evaluation of of peptide inhibitors of FimC/FimH protein-
protein interactions in uropathogenic Escherichia coli. J. Med. Chem.
2005, 48, 935-945.
(15) Cornelis, G. R.; WolfWatz, H. The Yersinia Yop virulon: A bacterial
system for subverting eukaryotic cells. Mol. Microbiol. 1997, 23,
861-867.
(16) Muller, S.; Feldman, M. F.; Cornelis, G. R. The type III secretion
system of Gram-negative bacteria: A potential therapeutic target?
Expert Opin. Ther. Targets 2001, 5, 327-339.
(17) Gauthier, A.; Finlay, B. B. Type III secretion system inhibitors are
potential antimicrobialssPresent in disease-causing Gram-negative
bacteria, components of this system might be good targets for novel
antimicrobial agents. ASM News 2002, 68, 383-387.

6188 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Dahlgren et al.
(41) Wold, S.; Josefson, M.; Gottfries, J.; Linusson, A. The utility of
multivariate design in PLS modeling. J. Chemom. 2004, 18, 156-
165.
(42) Wold, S.; Ruhe, A.; Wold, H.; Dunn, W. J. The collinearity problem
in linear regression. The partial least squares approach to generalized
inverses. SIAM J. Sci. Stat. Comput. 1984, 5, 735-743.
(43) Wold, S.; Sjostrom, M.; Eriksson, L. PLS-regression: a basic tool
of chemometrics. Chemom. Intell. Lab. Syst. 2001, 58, 109-130.
(44) Wold, S.; Esbensen, K.; Geladi, P. Principal component analysis.
Chemom. Intell. Lab. Syst. 1987, 2, 37-52.
(45) Eriksson, L.; Johansson, E.; Lindgren, F.; Sjo¨strom, M.; Wold, S.
Megavariate analysis of hierarchical QSAR data. J. Comput.-Aided
Mol. Des. 2002, 16, 711-726.
(46) Olsson, I. M.; Gottfries, J.; Wold, S. D-optimal onion designs in
statistical molecular design. Chemom. Intell. Lab. Syst. 2004, 73, 37-
46.
(47) deAguiar, P. F.; Bourguignon, B.; Khots, M. S.; Massart, D. L.;
PhanThanLuu, R. D-optimal designs. Chemom. Intell. Lab. Syst. 1995,
30, 199-210.
(48) Johnson, M. E.; Nachtseim, C. J. Some guidelines for constructing
exact D-optimal designs on convex spaces. Technometrics 1983, 25,
271-277.
(49) Sjo¨stro¨m, M.; Wold, S.; So¨derstro¨m, B. PLS discriminant plots.
Proceedings of PARC in Practice, Amsterdam, The Netherlands, June
19-21, 1985; Elsevier Science Publ. B. V.: North Holland, 1986;
pp 461-470.
(50) Terada, H.; Goto, S.; Yamamoto, K.; Takeuchi, I.; Hamada, Y.;
Miyake, K. Structural requirements of salicylanilides for uncoupling
activity in mitochondriasQuantitative-analysis of structure-
uncoupling relationships. Biochim. Biophys. Acta 1988, 936, 504-
512.
(51) Macielag, M. J.; Demers, J. P.; Fraga-Spano, S. A.; Hlasta, D. J.;
Johnson, S. G.; Kanojia, R. M.; Russel, R. K.; Sui, Z. H.; Weidner-
Wells, M. A.; Werblood, H.; Foleno, B. D.; Goldschmidt, R. M.;
Loeloff, M. J.; Webb, G. C.; Barrett, J. F. Substituted salicylanilides
as inhibitors of two-component regulatory systems in bacteria. J. Med.
Chem. 1998, 41, 2939-2945.
(52) Stephenson, K.; Yamaguchi, Y.; Hoch, J. A. The mechanism of action
of inhibitors of bacterial two-component signal transduction systems.
J. Biol. Chem. 2000, 275, 38900-38904.
(53) Kim, C.C.; Falkow, S. Delineation of upstream signaling events in
the salmonella pathogenicity island 2 transcriptional activation
pathway. J. Bacteriol. 2004, 186, 4694-4704.
(54) Xiao, Y. M.; Lan, L. F.; Yin, C. T.; Deng, X.; Baker, D.; Zhou, J.
M.; Tang, X. Y. Two-component sensor RhpS promotes induction
of Pseudomonas syringae type III secretion system by repressing
negative regulator RhpR. Mol. Plant-Microbe Interact. 2007, 20,
223-234.
(55) Barrett, J. F.; Isaacson, R. E. Annual Reports in Medicinal Chemistry;
Academic Press: San Diego, CA, 1995; pp 111-118.
(56) Payne, D. J.; Gwynn, M. N.; Holmes, D. J.; Pompliano, D. L. Drugs
for bad bugs: confronting the challenges of antibacterial discovery.
Nat. ReV. Drug DiscoVery 2007, 6, 29-40.
(57) ChemACX2002; CambridgeSoft: Cambridge, MA, 2002.
(58) MOE. Chemical Computing Group Inc.: Montreal, Quebec, Canada,
2004.
(59) Umetrics. SIMCA-P+ SIMCA-P+ 10.0, 10.5, 11.0; Box 7960, S-907
19 Umeå, Sweden.
(60) Forsberg, A.; Rosqvist, R. In-vivo expression of virulence genes of
Yersinia-pseudotuberculosis. Infect Agents Dis. 1993, 2, 275-278.
JM070741B