Statistical molecular design of a focused salicylidene acylhydrazide library and multivariate QSAR of inhibition of type III secretion in the Gram-negative bacterium Yersinia

Article abstract of DOI:10.1016/j.bmc.2010.02.022

A combined application of statistical molecular design (SMD), quantitative structure-activity relationship (QSAR) modeling and prediction of new active compounds was effectively used to develop salicylidene acylhydrazides as inhibitors of type III secretion (T3S) in the Gram-negative pathogen Yersinia pseudotuberculosis. SMD and subsequent synthesis furnished 50 salicylidene acylhydrazides in high purity. Based on data from biological evaluation in T3S linked assays 18 compounds were classified as active and 25 compounds showed a dose-dependent inhibition. The 25 compounds were used to compute two multivariate QSAR models and two multivariate discriminant analysis models were computed from both active and inactive compounds. Three of the models were used to predict 4416 virtual compounds in consensus and eight new compounds were selected as an external test set. Synthesis and biological evaluation of the test set in Y. pseudotuberculosis and the intracellular pathogen Chlamydia trachomatis validated the models. Y. pseudotuberculosis and C. trachomatis cell-based infection models showed that compounds identified in this study are selective and non-toxic inhibitors of T3S dependent virulence.

Full text of DOI:10.1016/j.bmc.2010.02.022

Bioorganic & Medicinal Chemistry 18 (2010) 2686–2703
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Statistical molecular design of a focused salicylidene acylhydrazide library
and multivariate QSAR of inhibition of type III secretion in the
Gram-negative bacterium Yersinia
*
Markus K. Dahlgren, Caroline E. Zetterström, Åsa Gylfe, Anna Linusson, Mikael Elofsson
Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden
a r t i c l e i n f o
a b s t r a c t
Article history:
A combined application of statistical molecular design (SMD), quantitative structure–activity relationship
(QSAR) modeling and prediction of new active compounds was effectively used to develop salicylidene
acylhydrazides as inhibitors of type III secretion (T3S) in the Gram-negative pathogen Yersinia pseudo-
tuberculosis. SMD and subsequent synthesis furnished 50 salicylidene acylhydrazides in high purity.
Based on data from biological evaluation in T3S linked assays 18 compounds were classified as active
and 25 compounds showed a dose-dependent inhibition. The 25 compounds were used to compute
two multivariate QSAR models and two multivariate discriminant analysis models were computed from
both active and inactive compounds. Three of the models were used to predict 4416 virtual compounds in
consensus and eight new compounds were selected as an external test set. Synthesis and biological eval-
uation of the test set in Y. pseudotuberculosis and the intracellular pathogen Chlamydia trachomatis vali-
dated the models. Y. pseudotuberculosis and C. trachomatis cell-based infection models showed that
compounds identified in this study are selective and non-toxic inhibitors of T3S dependent virulence.
Ó 2010 Elsevier Ltd. All rights reserved.
Received 4 November 2009
Revised 10 February 2010
Accepted 12 February 2010
Available online 18 February 2010
Keywords:
Statistical molecular design
QSAR
Type III secretion
Bacterial virulence
Antibacterial
Yersinia
Chlamydia
1. Introduction
Escherichia coli.4 The T3S machinery is a syringe-like organelle that
delivers toxins into the host cell cytosol and the bacteria thereby
Antibiotic resistance has evolved against the majority of all clin-
ically used antibiotics and it is clear that mankind is loosing ground
in the never-ending battle against bacterial infections. Current
antibiotics all target bacterial growth with low selectivity and
thereby affect both pathogenic bacteria and our benign endoge-
nous microflora. Alternative strategies are needed and recent find-
ings suggest that bacterial mechanisms required for infection,
virulence factors, can effectively be targeted.1–3
Type III secretion (T3S) is a virulence system found in a wide
range of clinically important Gram-negative pathogens including
Yersinia spp., Chlamydia spp., Shigella spp., Salmonella spp., Pseudo-
monas aeruginosa, and enteropathogenic and enterohemorrhagic
modulate the host responses and create a specific niche for coloni-
zation. In contrast to the toxins, which in most cases are unique for
each bacterial species, the T3S machinery is evolutionary con-
served and therefore constitutes a potential drug target.1–3,5
Screening-based approaches using phenotype cell-based assays
have resulted in synthetic and natural compounds that target
T3S in Yersinia spp., enteropathogenic E. coli, Salmonella typhimuri-
um, P. aeruginosa, and P. syringae.6–11 Continued efforts have led to
optimized compounds, tethered inhibitors, synthetic methods, and
structure–activity relationships.12–19 Recently it was reported that
plant phenolic acids induce expression of T3S system genes in the
plant pathogen Dickeya dadantii by moderation of a two-compo-
nent system20 and based on this finding p-coumaric acid was iden-
tified as a repressor of T3S system genes.21 Another recent
publication describes that the proton pump inhibitor omeprazole
blocks T3S in S. enterica.22 The salicylidene acylhydrazides (Fig. 1),
are promising since they target T3S in Yersinia pseudotuberculosis,
Abbreviations: BB, building block; T3S, type III secretion; T2S, type II secretion;
T4S, type IV secretion; PLS, partial least-square regression to latent ce:text>; Hi-PLS,
hierarchical PLS; PLS-DA, PLS discriminant analysis; ETEC, enterotoxigenic Esche-
richia coli; YPIII, Yersinia pseudotuberculosis serotype III; BHI, brain heart infusion;
DOOD, D-optimal onion design; PCA, principal component analysis; SAR, structure–
activity relationship; QSAR, quantitative structure–activity relationship; SMD,
statistical molecular design; MLR, multiple linear regression; SVM, support vector
machine; MM, molecular mechanics; Yop, Yersinia outer protein; FCS, fetal calf
serum; DMEM, Dulbecco’s modified Eagle’s medium; LB, Luria broth; OD, optical
density; RPMI, Roswell Park Memorial Institute; MIC, minimum inhibitory
concentration.
* Corresponding author. Tel.: +46 90 7869328; fax: +46 90 7867655.
E-mail address: mikael.elofsson@chem.umu.se (M. Elofsson).
Figure 1. The general structure of salicylidene acylhydrazides.
0968-0896/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2010.02.022

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2687
Shigella flexneri, enterohemorrhagic E. coli, and Salmonella spp. that
inflict a broad range of gastrointestinal infections and Chlamydia
spp. that cause sexually transmitted disease.13,23–27 In addition,
to the potential of these T3S inhibitors to be developed into
therapeutic agents they are important chemical probes to further
study the role of T3S in different systems. This is particularly
important for organisms like the obligate intracellular pathogen
Chlamydia spp. for which no tractable genetic system exists.23,28–32
The biological target of the salicylidene acylhydrazides is cur-
rently unknown and the biological evaluations have been per-
formed on whole bacterial cells. It cannot be excluded that the
compounds act on multiple targets or that the target(s) might vary
between different species. To further develop this compound class
it is necessary to gain a detailed understanding of what chemical
features that affect biological activity and how these can be
merged with, for example, pharmacokinetic requirements for
in vivo experiments. One strategy to achieve this is the computa-
tion of quantitative structure–activity relationship (QSAR)33–35
models based on compound sets obtained by statistical molecular
design (SMD)36,37 and biological data, typically from assays in vi-
tro. Despite its obvious advantages the combined application of
SMD and QSAR modeling have been described in relatively few
publications.16,17,37–41
The strength of SMD is that it results in a statistically balanced
compound set suitable for QSAR modeling and importantly SMD
can be performed without knowledge of the target. One way to
perform SMD is to mathematically describe building blocks (BBs),
that is, the reactants used to construct the compound class of inter-
est, with calculated or experimentally derived molecular descrip-
tors and then make a selection based on chemical properties. The
procedure is repeated for each BB set and an enumeration of the
selected BBs results in a virtual library. The products are then char-
acterized with descriptors and a new diverse selection is per-
formed in order to lower the number of synthetic targets and
still retain most of the information of the full data set. Alternatively
products can be selected by a statistically controlled combination
of the selected BBs without requirement for characterization of
the virtual products. Another advantage of making the selection
in the BB space rather than the product space is that the interpre-
tation of the QSAR will give information about the desired proper-
ties of BBs.
QSAR models are constructed from a set of biologically active
molecules, the training set, identified among compounds obtained
by, for example, SMD. The molecules are mathematically described
by calculated or experimentally derived molecular descriptors and a
regression technique, for example, multiple linear regression
(MLR),42 partial least-squares regression to latent structures
(PLS),43,44 or support vector machine (SVM),45 is used to correlate
the training set and its variables to the biological response. QSAR
models are generally only valid in the experimental domain covered
by the molecules in the trainingset since such models are interpolat-
ing. This means that the training set has to cover the molecular prop-
erties important for the biological response, that is, the training set
should represent a variety of chemical properties that might influ-
ence the biological response in order to get a spread in biological
activity. At the same time the training set should not contain too di-
verse compounds as it potentially could lead to many inactive com-
pounds and this can be prevented by applying basic SARs prior to
SMD.
different models could possibly all be useful, depending on the
task, as the models could potentially excel at predicting parts of
the prediction set. Recent publications highlight the usefulness of
consensus predictions using several QSAR models.46–48 The more
orthogonal models are relative to each other, the higher the predic-
tive power should be when they are used in consensus.48
In this paper, we use SMD to select diverse sets of salicylic alde-
hydes and hydrazides for synthesis of salicylidene acylhydrazides.
Of the 54 selected compounds 50 compounds could successfully be
synthesized and purified. All compounds were evaluated for T3S
inhibition in Y. pseudotuberculosis and several multivariate QSAR
models were computed. The predictive ability of the models was
evaluated and confirmed in consensus using an external set of
compounds. A panel of compounds was evaluated for virulence
inhibition in cell infection models.
2. Results and discussion
2.1. Statistical molecular design
The salicylidene acylhydrazides were assembled in one step
from salicylic aldehydes and hydrazides. Based on published data
it was clear that the aromatic rings of both the BBs tolerated sub-
stitution with different functional groups or exchange to heteroar-
omatic systems or fused aromatics without loss of biological
activity.13,24,29 After careful inspection of the compounds and their
corresponding biological activity, we could not deduce any clear
SAR in terms of substituent patterns. The salicylic aldehyde moiety
tolerated both polar and hydrophobic substituents while retaining
biological activity. This indicated that properties like pKa of the sal-
icylic phenol and partial charges of the carbon atoms in the sali-
cylic aldehyde aromatic ring might be of larger importance than
the hydrophobicity of the functional groups. There seemed to be
a highly non-linear relationship between the BBs constituting the
molecules and their biological response. Current data thus sug-
gested complex SARs and in view of that the target was unknown
it was of importance to select a statistically balanced set of com-
pounds, covering a wide range of properties, in order to get a com-
pound library with a spread in biological activity well suited for
QSAR modeling. From commercial vendors 48 salicylic aldehydes
(Fig. S1) and 92 hydrazides (Fig. S2) were selected (Section 4.1),
numerically described using calculated molecular descriptors and
subjected to SMD as schematically shown in Figure 2.
Low energy conformations of the 48 salicylic aldehydes were
numerically described with computed ab initio, molecular
mechanics (MM) and informatics descriptors resulting in total 70
descriptors (Table S1). In order to ensure that the MM descriptors
were not overrepresented and to reduce the impact of the hydro-
phobicity descriptors, a selection of MM descriptors were grouped
in two separate groups and summarized using PCA (Fig. 2a, Tables
S2–S4). The total data was compressed block wise using PCA,
resulting in a five component model (R2 = 0.85, Q2 = 0.53) for the
salicylic aldehyde BBs (Fig. 2a and Table S5). Conformational anal-
ysis of the 92 hydrazides was followed by a numerical description
of the structures through computation of a total of 97 ab initio,
MM, and informatics descriptors (Table S6). After removal of a
number of non-interpretable descriptors the remaining subset con-
sisting of 50 descriptors (Table S7) were summarized in one block
using PCA (Fig. 2b), resulting in a five component model (R2 = 0.88,
Q2 = 0.77). The PCA score vectors were used as design variables,
and for each data set a two-layer D-optimal onion design (DOOD)49
was performed. Seventeen salicylic aldehydes and 18 hydrazides
were selected by the designs. One salicylic aldehyde, 5-bromo-sal-
icylic aldehyde, which has been found in several salicylidene acy-
lhydrazides that inhibited T3S13,24 was added manually to make
It is common that several QSAR models are computed, for
example, when applying different regression methods, adding or
removing training set molecules, transforming molecular descrip-
tors or the biological response mathematically, and adding or
removing descriptors. Often it can be difficult to decide which
model to use for predictions, as statistics of the models may be
similar while the predictions of new compounds may differ. The

2688
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
a
PCA
PCA
MM charge
descriptors
MM hydro-
phobicity
X1
X2
descriptors
PCA
5 PCA
score
vectors
ab initio
descriptors
DOOD
X1 X2
X3
17 salicylic
aldehydes
b
Hydrazide
descriptors
PCA
5 PCA
score
vectors
DOOD
18 hydrazides
Figure 2. Schematic representation of the design process for (a) salicylic aldehydes and (b) hydrazides. For the salicylic aldehydes the MM descriptors describing charges and
hydrophobicity were grouped and summarized with PCA separately. The resulting score vectors (blocks X1 and X2) were combined with the ab initio descriptors and X3
(molecular weight and volume). PCA was used to summarize the variation of both BB sets and two libraries of BBs were designed using two separate two-layer DOODs.
the two BB sets of equal size. A final selection resulted in 54 virtual
products (Section 4.2).
pounds were tested for growth inhibition of Y. pseudotuberculosis
and E. coli at 50 and 100 lM and no or modest effects were ob-
served (Fig. S5).
2.2. Synthesis
2.4. Multivariate QSAR for T3S inhibition
The salicylidene acylhydrazides were synthesized in one step
from salicylic aldehydes and hydrazides by conventional or micro-
wave heating in absolute ethanol. The purified compounds were
analyzed and characterized with LC–MS and 1H NMR spectroscopy
using DMSO as solvent. A dynamic process gave rise to two rota-
mers for 34 of the salicylidene acylhydrazides. A total of 50 com-
pounds were successfully synthesized in 60–95% yield with a
purity of at least 95% (Tables 1 and S8). The three compounds con-
taining the BB butyric acid hydrazide could not be synthesized as
the BB proved to be unreactive under the reaction conditions used
(Fig. S3). 4-Methyl-[1,2,3]thiadiazole-5-carboxylic acid (2-hydro-
xy-3-methoxy-5-nitro-benzylidene)-hydrazide (Fig. S3) could not
be purified above 80% according to the integrated UV-trace from
LC and this compound was therefore omitted from further studies.
All our modeling efforts showed that there was a non-linear rela-
tionship between the molecular descriptions of the BBs constituting
the inhibitors and their biological response. Initially two models for
predictions, Hi-PLS-1 and Hi-PLS-DA-1, were computed based on
PLS score vectors by application of our previously published meth-
odology17 as schematically described in Figure 3 (model statistics
in Table 2). Hi-PLS-1 showed good correlation between experimen-
tal and calculated luciferase signal inhibition (Fig. 4) and Hi-PLS-
DA-1 showed good separation of active compounds and inactive
compounds along the direction of the PLS score vector (Fig. 5). To
getinterpretable models, groupeddescriptors were used to compute
two new models, Hi-PLS-2 and Hi-PLS-DA-2, as schematically
shown in Figure 6. The correlation between calculated and experi-
mental inhibition is good for Hi-PLS-2 (Fig. 7) while Hi-PLS-DA-2
was less efficient than Hi-PLS-DA-1 in separating actives from
non-actives (Fig. S6). All models except Hi-PLS-DA-2 had acceptable
and comparable statistics (R2Y above 0.65 and Q2 above 0.50, Table
2). Permutations of the response and subsequent model computa-
tions were performed and the result showed that the Hi-PLS-1 and
Hi-PLS-2 models were not a result of chance correlation (Figs. S7
and S8). The loading plots for Hi-PLS-1 (Fig. S9), and Hi-PLS-DA-1
(Fig. S10), and Hi-PLS-DA-2 (Fig. S11) are given in Supplementary
data.
2.3. Biological evaluation for QSAR modeling
The 50 salicylidene acylhydrazides were evaluated at eight con-
centrations (1.56–200 lM) in Y. pseudotuberculosis using a T3S
linked luciferase reporter-gene assay6,13 (Tables 1 and S9a, and
Fig. S24) and an enzymatic assay based on the secreted phospha-
tase, YopH.16 Compounds were classified as active if they showed
a dose-dependent effect with more than 40% inhibition at 50 lM
in the luciferase reporter-gene assay and a concomitant dose-
dependent reduction of the YopH activity (Fig. S4a). Based on these
criteria 18 of the 50 compounds were classified as active. Five com-
pounds (ME0163 and ME0190–ME0193) did not show an inhibi-
tory effect in the YopH assay (Fig. S4b) and possibly inhibit the
luciferase directly in a T3S independent mechanism. For ME0151
(Fig. S24) the inhibitory effect diminished in both the reporter-
The interpretation was based on the model coefficients of Hi-
PLS-2 as shown in Figure 8. The model is dominated by non-linear
terms, underlining the complex SAR of the compound class. The
calculated pKa of the phenol in position two of the salicylic alde-
hydes is the most important property in the model, participating
in seven interaction terms. Other important properties are the
electrostatic potential charges of the aromatic carbons in the sali-
cylic aldehyde ring that are involved in six interaction terms and
gene and YopH assay at concentrations above 100
lM and this is
most likely due to limited aqueous solubility. In addition, all com-

Table 1
Percentage inhibition of luciferase light emission for strain YPIII-pIB102 (yopE-luxAB) in the presence of compounds ME0150–ME0199 at 50 lM concentration
ID
Structure
Inhibitiona
ID
Structure
Inhibitiona
ID
Structure
Inhibitiona
HO
N
N
S
OH
OH
OH
OH
H
N
H
N
F
F
H
N
F
O
O
N
O
ME0150
60
1
ME0164
32
1
ME0178
22
6
N
N
F
O
O
O
O
Br
O
O2N
Cl
O
N
N
H
N
OH
H
N
OH
H
N
F
N
ME0151
ME0152
54
25
2
3
ME0165
ME0166
84
45
0
2
ME0179
ME0180
27
73
3
1
N
S
N
O
O
Br
O
OH
OH
O
OH
OH
H
N
H
N
H
N
O
F
N
H2N
N
O
N
N
S
N
O
O
Br
O
OH
O
OH
H
H
N
F
F
F
N
N
N
H
N
ME0153
36
3
ME0167
23
9
ME0181
9
3
O
O
O
N
Br
O
Br
Cl
Br
O2N
N
OH
H
N
OH
OH
H
N
H
N
N
N
N
ME0154
ME0155
27
32
6
2
ME0168
ME0169
87
86
0
1
ME0182
ME0183
18
23
8
4
O
O
O
O
Cl
Br
OH
OH
OH
OH
H
N
H
N
H
N
N
N
O
N
N
O
O
Cl
Br
O
O
OH
H
N
H
N
F
F
O
I
OH
N
H
N
O
N
F
O
N
ME0156
24 10
ME0170
33
3
ME0184
66
1
O
N
Br
O
O
O
I
F
Br
N
OH
H
N
OH
OH
H
N
H
I
N
O
N
N
ME0157
ME0158
30
26
5
8
ME0171
ME0172
39
10
3
3
ME0185
ME0186
35
39
7
2
O
O
O
O
NO2
OH
I
Cl
OH
H
N
H
N
O
N
OH
N
H
N
N
I
O
Cl
N
NO2
O
I
(continued on next page)

Table 1 (continued)
ID
Structure
Inhibitiona
ID
Structure
Inhibitiona
38 11
ID
Structure
Inhibitiona
O
O
O
OH
F
H
N
O
O
Cl
OH
OH
H
N
H
N
O
N
NO2
O
ME0159
43
3
ME0173
ME0187
30
1
O
N
N
Br
O
O
N
NO2
F
Br
OH
H
N
OH
OH
H
N
H
N
O
Cl
NO2
O
N
N
ME0160
ME0161
32
29
4
1
ME0174
ME0175
63
19
4
3
ME0188
ME0189
22
18
1
4
NO2
O
O
O
F
NO2
HO
Cl
OH
H
N
OH
H
N
NO2
N
OH
H
N
N
O
Cl
O
Cl
N
HO
OH
NO2
O
F
O
O
Cl
O
O
OH
H
N
OH
H
OH
N
H
N
N
S
N
Cl
ME0162
ME0163
ME0192
70
51
43
2
2
1
ME0176
ME0177
ME0195
27
61
23
0
2
2
ME0190
ME0191
ME0198
66
54
43
1
2
2
O
O
N
O
N
HO
OH
O
N
Cl
Cl
O
OH
F
OH
Cl
H
N
H
N
OH
H
N
H2N
N
O
N
O
NO2
O
HO
OH
O
N
Cl
HO
OH
O
OH
Br
H
N
H
N
O
OH
O
H
N
O
N
N
O
Cl
N
O
O
N
O2N
Cl
O
OH
O
H
N
OH
H
N
O
O
O
OH
O
H
N
O
N
S
N
ME0193
ME0194
50
36
3
1
ME0196
ME0197
16
3
8
ME0199
34
2
OH
OH
O
N
O
O
Br
O
OH
OH
H
N
H
N
O
OH
OH
H2N
N
N
9
O
NO2
O
Br
a
Means and standard deviations are calculated from triplicates, and experiments were reproduced on at least three separate occasions.

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2691
Salicylidene acylhydrazides
Salicylic
Aldehydes
Hydrazides
Characterization
with descriptors
Characterization
with descriptors
PLS
PLS
XA
XB
Y
Y
PLS score vector
extraction
X1
X2
Combination and
expansion of terms
PLS
Figure 4. Experimental versus calculated luciferase signal inhibition plot for Hi-
PLS-1 (50 lM compound concentration). The ID numbers in the plot are the last
two numbers of the compound IDs.
2
2
X1
X2
X1
X2
X1X2
Y
Figure 3. Schematic representation of calculation of Hi-PLS-1 and Hi-PLS-DA-1,
using hierarchical PLS score vectors. XA and XB are the BB data matrices with
calculated descriptors. X1 and X2 are the extracted PLS score vectors. Y is the
eight compounds were synthesized and biologically evaluated as
described in Sections 4.3 and 4.10 (Tables 3 and S9b). Five of the
selected compounds (ME0258, ME0259, and ME0261–ME0263)
had hydrazide BBs not used in any of the previously 50 synthesized
salicylidene acylhydrazides, while all of the salicylic aldehydes
were represented in the training set.
response (i.e.,
% light emission inhibition at 50 lM for Hi-PLS-1 and class
membership for Hi-PLS-DA-1).
Table 2
The compounds predicted as inactive were all inactive, and of the
five compounds predicted as active, three were active (Table 3). In
Hi-PLS-1, the three 5-bromo-salicylaldehyde containing com-
pounds (ME0259–ME0261) were predicted as having the highest
activity of the eight compounds. According to Hi-PLS-2 compound
ME0264 should be the most active. Both models predicted
ME0257 to have the lowest activity of the five active compounds.
ME0257 was correctly classified as active, but contrary to predic-
tions it had the highest activity of the five compounds predicted as
Statistics for the multivariate models Hi-PLS-1 and 2 and Hi-PLS-DA-1 and 2
Model
Score vector seta
R2X
R2Y
Q2
Hi-PLS-1
Hi-PLS-DA-1
Hi-PLS-2
SAL2 + HYD2
SAL1 + HYD1
—
—
0.19
0.12
0.29
0.21
0.67
0.67
0.69
0.43
0.51
0.55
0.53
0.35
Hi-PLS-DA-2
a
The two combined score vector sets used for the model in question. The score
vectors were derived from PLS models (see Section 4.4).
active and it was one of the four most potent compounds at 50
lM
in the total set of 58 salicylidene acylhydrazides. ME0260 was
correctly classified as active and ME0259, which was predicted as
active, was border line to being active, with a dose-dependent
inhibition. ME0261 and ME0264 were both incorrectly predicted
as active and the lack of activity could possibly be due to factors such
as efflux or the compounds having properties not covered by the
models. The models correctly classified inactive compounds but
could not correctly rank the active compounds. Importantly, the
models could handle predictions of compounds having BBs not
represented in the training set.
the shape of the salicylic aldehyde moiety which is participating in
five interaction terms. The substitution pattern on the salicylic
aldehyde ring is therefore of vital importance when designing
new salicylidene acylhydrazides as the functional groups on the
aromatic ring will affect the electrostatic potential charges, the
pKa of the phenol and the shape of the BB. As a consequence of
the non-linear SAR, the salicylic aldehyde BB should not be varied
independently of the hydrazide BB, since the biological response is
largely dependent on the interactions between the two BBs.
2.5. Evaluation of models using an external test set
2.6. Inhibition of Y. pseudotuberculosis infection
Combination of the 48 salicylic aldehydes and 92 hydrazides
gave 4416 virtual compounds of which 1916 were predicted as ac-
tive in Hi-PLS-1, 2208 in Hi-PLS-2, and 659 in Hi-PLS-DA-1. The
overlap of the three models was calculated to 327 compounds,
indicating a substantial difference between the predictions of the
models. From the 327 compounds predicted as active, five
(ME0257, ME0259–ME0261, and ME0264) were manually se-
lected. Additionally, three compounds predicted as inactive in all
models (ME0258, ME0262, and ME0263) were selected and the
The T3S linked luciferase reporter-gene assay and the enzy-
matic YopH assay, in combination with assays for general growth
inhibition clearly identifies putative virulence inhibitors but these
assays are however uncoupled from infection. Four active com-
pounds, ME0165, ME0166, ME0168 and ME0174 and one inactive,
ME0181, were therefore tested for virulence inhibition in an assay
where the mouse macrophage-like cell line J774 is infected with Y.
pseudotuberculosis (pIB102).29 A fully functional T3S system trans-
locates the effector proteins into the macrophage cytosol resulting

2692
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
Figure 5. PLS-DA Score plot (Hi-PLS-DA-1) of active (black boxes) and inactive (gray triangles) compounds. The plot shows a good separation of active and inactive
compounds along the direction of the score vector. The ID numbers in the plot are the last two numbers of the compound IDs.
Hydrazides
Salicylic aldehydes
Moment of
Inertia
Moment of
Inertia
X21
X22
X23
X24
X25
X11
Size
Descriptors
Atomic Partial
Charges
X12
Descriptors for
salicylic aldehyde
BB´s and
Surface
Descriptors
Size
Descriptors
X13
X14
X15
X16
PCA
PCA
hydrazide BB´s
Grouping
of
variables
Grouping
of
variables
Charge
Descriptors
Surface
Descriptors
Hydrophobicity
Descriptors
Charge
Descriptors
Variables
that did
not fit
Hydrophobicity
Descriptors
into any
group
PCA score vector
extraction and
combination with
ungrouped
PCA score vector
extraction and
combination with
ungrouped
X1U
X2U
variables
variables
XSAL
XHYD
Combination and
expansion of data
PLS
Y
2
2
XHYD
XSALXHYD
XSAL XHYD
XSAL
Figure 6. Schematic representation of calculation of Hi-PLS-2 and Hi-PLS-DA-2, using PCA score vectors from grouped variables. Descriptors describing similar properties
within the same BB set were grouped and data was summarized using PCA. The PCA score vectors from both BB sets were combined and expanded prior to PLS regression. Y is
the response (i.e., % light emission inhibition at 50 lM for Hi-PLS-2 and class membership for Hi-PLS-DA-2).
in cell death or reduced viability. The status of the J774 cells is
monitored using CalceinAM, that is, converted to a green fluores-
cent molecule in healthy cells. Wild-type Y. pseudotuberculosis re-
duced J774 viability and this effect was reversed by ME0168 that

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2693
not target toxin secretion in general (Fig. 9). The active compounds
ME0165 and ME0166 resulted in the same effects as ME0168
(Fig. S12). The fourth active compound ME0174 however failed
to inhibit virulence which was also true for the inactive compound
ME0181 (Fig. S12). Taken together these results show that our new
compounds not only were active when tested on bacteria alone but
also selective, non-toxic, and active in a complex cell-based infec-
tion model.
2.7. Inhibition of Chlamydia trachomatis infection
In order to investigate the specificity for T3S of the salicylidene
acylhydrazides, the external test set (ME0257–ME0264) was eval-
uated for the ability to inhibit growth of Chlamydia trachomatis. In
contrast to the extracellular pathogen Y. pseudotuberculosis, C. tra-
chomatis is an obligate intracellular pathogen that relies on T3S for
its growth. Chlamydia is therefore a suitable model organism to
evaluate if T3S inhibitors identified with Yersinia inhibit the evolu-
tionary conserved T3S system in other pathogens. Salicylidene acy-
lhydrazides do not inhibit entry of Chlamydia into host cells31 and
the T3S inhibition observed in Chlamydia models therefore requires
that the compounds penetrate the host cell membrane and enter
the compartment, inclusion, hosting the bacteria.23,28–30 The com-
pounds were added at five different concentrations to HeLa 229
cells one hour post infection with C. trachomatis. For each com-
pound the lowest concentration resulting in complete inhibition
of C. trachomatis, that is, the minimum inhibitory concentration
MIC (Table 3). Five of the eight tested salicylidene acylhydrazides
(ME0259–ME0261 and ME0263–ME0264) were able to com-
Figure 7. Experimental versus calculated luciferase signal inhibition plot for Hi-
PLS-2 (50 M compound concentration). The ID numbers in the plot are the last
two numbers of the compound IDs.
l
completely blocked the toxic effects at 50 lM (Fig. 9). Uninfected
control cells remained healthy at all compound concentrations
pletely inhibit growth at 50 lM or lower compound concentration.
Comparing these results with the biological data from the Y. pseu-
dotuberculosis assay, the overlap was determined to two active
(ME0259 and ME0260) and two inactive (ME0258 and ME0262)
compounds, which was in full agreement with the consensus pre-
dictions of the three multivariate models. The models could not
correctly predict the biological response of ME0261 and ME0264
in Yersinia. Interestingly the predictions for ME0261 and ME0264
are in accordance with the inhibition of Chlamydia growth. The
models cannot correctly predict the observed inability of
ME0257 to inhibit Chlamydia growth or the activity of ME0263.
These results show that the models hold information relevant for
indicating that ME0168 is non-toxic (Fig. 9). Infection by the Y.
pseudotuberculosis pIB604
DyopB T3S translocation mutant showed
that this strain is non-virulent and ME0168 had no effect under
these conditions (Fig. 9). To further conclude selectivity for T3S
dependent virulence mechanisms J774 cells were also infected
with enterotoxigenic E. coli (ETEC).50,51 Both Y. pseudotuberculosis
and ETEC are Gram-negative gastrointestinal pathogens but ETEC
lacks a T3S system and instead employs a type II secretion (T2S)
system for secretion of toxins.50 ETEC efficiently reduced macro-
phage health and this effect was not reversed by addition of
ME0168 indicating that the inhibitor is selective for T3S and do
Figure 8. Loading (w*c) column plot from Hi-PLS-2. The loadings in black were used for interpretation. The loadings in gray could not be removed as the variables were used
in non-linear terms important for the model.

2694
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
Table 3
Percentage inhibition of luciferase light emission for strain YPIII-pIB102 (yopE-luxAB) in the presence of external test compounds ME0257–ME0264 at 100, 50, 25, and 12.5 lM
concentrations
ID
Structure
100
Ma
50
l
25
l
12.5
Ma
Hi-PLS-1 Hi-PLS-2 Hi-PLS-DA-1
Chlamydia Cell viability Cell viability
l
Ma
Ma
l
MICb
(lM) 50
l
Ma (%)
25
l
Ma (%)
N
OH
H
N
ME0257
96
20
2
4
84
2
3
27
5
2
4
4
49
47
34
Border line active Inactive
67
4
72
1
N
O
Cl
OH
H
N
N
ME0258
ME0259
ME0260
26
38
44
10
11
22
4
13
15
12 25
Inactive
Active
Active
Inactive
79
60
65
3
1
2
83
77
73
2
1
3
O
OH
Br
H
N
N
58
71
5
4
6
2
3
4
9
2
73
64
48
46
50
Cl
O
O
OH
Br
H
N
N
25
O
OH
H
N
N
N
ME0261
ME0262
39
57
3
8
26
22
4
3
15
9
7
7
7
10 68
51
31
Active
50
74
33
3
1
75
58
1
2
O
O
Br
OH
H
N
O
15
1
20
Inactive
Inactive
Br
OH
H
N
N
ME0263
ME0264
24
6
17 15 20
6
7
18
4
3
7
26
53
34
62
Inactive
50
79
72
1
1
80
75
4
2
O
O2N
OH
H
N
F
20 15 28 10 19
Border line active 50
N
O
Compounds with at least 40% inhibition at 50
predicted class membership for Hi-PLS-DA-1 at 50
l
M compound concentration were defined as active. The predicted inhibition is given for Hi-PLS-1 and Hi-PLS-2 and the
M compound concentration. Minimum inhibitory concentration, MIC, of Chlamydia trachomatis growth was tested at 50,
M. Compound toxicity to the host cells used for Chlamydia growth, HeLa 229, was tested at 50 and 25 M. Compounds were defined as non-toxic when
HeLa229 cell viability was at least 70% relative to controls.
l
25, 12.5, 6.25, 3.13
l
l
a
Means and standard deviations are calculated from triplicates, and experiments were reproduced on at least three separate occasions.
Chlamydia MIC was tested in duplicates on at least three separate occasions.
b
T3S inhibition in Yersinia as well as growth inhibition of Chlamydia,
2.8. Mode of action
supporting a common mode of action. There are however some dif-
ferences, such as ME0263, that is, inhibiting Chlamydia growth but
is inactive in the Yersinia assays, or ME0257, that is, active in Yer-
sinia but unable to inhibit Chlamydia growth. These discrepancies
could possibly be related to membrane properties that differ be-
tween Yersinia, Chlamydia and its eukaryotic host cell.
To ensure that the observed inhibition of Chlamydia growth was
not a result of the salicylidene acylhydrazides being toxic to the in-
fected HeLa 229 cells, cell viability was measured (Table 3). Four of
the five external test set compounds, ME0260, ME0261, ME0263,
and ME0264, that completely inhibited Chlamydia growth were
not toxic to HeLa 229 cells (70% or higher cell viability compared
to the DMSO control at the lowest concentrations where Chlamydia
was inhibited). Only ME0259 was slightly toxic (60% cell viability
The first detailed investigation of the salicylidene acylhydrazides
as T3S inhibitors showed that the inhibition is instantaneous,
reversible, and that the compounds likely act directly on the secre-
tion machinery that span the inner and outer membranes of the bac-
terium.13 The T3S syringe-like apparatus with a needle and
membrane-spanning base is highly conserved while its regulation
and effector proteins often are unique for the different bacterial spe-
cies. The fact that the compounds inhibit T3S in Y. pseudotuberculo-
sis,13 Salmonella enterica,24,25 Shigella flexneri,26 enterohemorrhagic
E. coli,27 and Chlamydia spp.23,28–30 support a direct inhibition of
the T3S machinery. In contrast the intracellular pathogen Coxiella
burnetii that utilizes a type four secretion (T4S) system was unaf-
fected23 and in the present study we show that the toxicity from tox-
ins secreted by a T2S system in enterotoxigenic E. coli is unaffected at
at 50 lM compound concentration).

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2695
ME0168
140
120
100
80
0 µM
25 µM
50 µM
100 µM
60
40
20
0
wt
yopB mutant
Uninfected
ETEC
Figure 9. Infection of J774 macrophages with wild-type Y. pseudotuberculosis, a non-virulent yopB mutant strain of Y. pseudotuberculosis, and enterotoxigenic E. coli (ETEC) in
the presence of different concentrations of ME0168. Macrophage viability was quantified from the fluorescence signal originating from enzymatic conversion of the reagent
CalceinAM. Viability for uninfected cells corresponds to 100%. ME0168 is non-toxic and selectively rescues macrophages infected with wild-type Y. pseudotuberculosis. Error
bars are calculated with Gauss approximation formula.
concentrations that inhibit T3S. The T3S syringe-like machinery
shows remarkable similarity to the hook-basal body of the flagel-
lum.5,52 The flagellum is an organelle required for bacterial motility
and interestingly salicylidene acylhydrazides inhibit motility in Y.
pseudotuberculosis6 and S. enterica.24 This further supports the
hypothesis that the compounds target the conserved machinery di-
rectly. Further support comes from the finding that treatment of S.
flexneri with a salicylidene acylhydrazide resulted in increased num-
bers of T3S apparatuses without needles or with shorter needles.26
This indicates that the inhibitor blocks secretion of the needle pro-
tein component(s), the first level of protein export required for
assembly of a fully functional T3S machinery. In a recent study tran-
criptomics were used in a comprehensive analysis of the effects of
four salicylidene acylhydrazides on enterohemorrhagic E. coli.27
The results showed that the compounds inhibit expression of the
T3S system but it remains to be established if this effect is a result
from feed-back regulation originating from inhibition of the secre-
tion apparatus or if the compounds target a regulatory pathway.
An interesting feature of the salicylidene acylhydrazides is that
they can act as trivalent ligands in complexes with metal cations
and a number of complexes have been prepared and their structures
determined.53,54 Due to their chelation capacity the compounds
have been investigated as therapeutic agents against iron poison-
ing.55 Intracellular pathogens require iron for growth and addition
of Fe2+ and Fe3+ reversed the effect of salicylidene acylhydrazides
in C. trachomatis.30 The observed inhibition of growth of Chlamydia
spp. might thus be the result of iron deprivation rather than a
blocked T3S system. The fact that inhibition of T3S in Y. pseudotuber-
culosis could not be abrogated by addition of iron argues other-
wise.30 Transcriptomic analysis was recently carried out in E. coli
and showed that iron response genes were unaffected by the com-
pounds and the E. coli were therefore not suffering from iron deple-
tion.27 A plausible explanation is that Chlamydia spp. utilize T3S to
secure a supply of iron that enable intracellular growth.30
ture important global chemical features affecting, for example,
permeability and metabolic stability. The fact that QSAR models
could be computed and the models could be validated with an
external test set in two biologically different T3S dependent patho-
gens strongly support a common and selective mode of action on
the T3S apparatus directly or on its regulation.
3. Conclusions
Application of an SMD strategy resulted in 50 compounds and
after biological evaluation three QSAR models with acceptable sta-
tistics could be established. The models differed to a large extent
and predictions were therefore made in consensus. The models
could correctly classify six of eight compounds in an external test
set. To further establish the potential of this class of compounds
a selected set of inhibitors were evaluated in cell-based infection
models. It was found that compounds active in the first line assays
also displayed a selective and non-toxic virulence-blocking activ-
ity. In conclusion we illustrate that the combined use of SMD, QSAR
modeling, and validation with an external test set predicted in con-
sensus is an effective and general strategy for generation of infor-
mation important in the process of lead identification and
optimization.
4. Experimental section
4.1. Characterization of BBs
BBs were selected manually from Aldrich, Maybridge, ABCR, Ac-
ros, and Alfa Aesar based on availability, size and chemical compat-
ibility. For the hydrazide BBs only up to two connected ring
systems were allowed and all BBs with molecular weight lower
than 50 or higher than 300 were removed. In addition, BBs having
more than two atoms connecting the hydrazide part and the ring
system were removed. Finally BBs containing chiral centers were
removed. The final sets contained 48 salicylic aldehydes (Fig. S1)
and 92 hydrazides (Fig. S2). The BBs were created in MOE56 and
for the salicylic aldehydes a conformational analysis was per-
formed using OMEGA57 with default settings. The conformational
analysis for the hydrazide BBs was performed in MOE using sto-
chastic conformational search with default settings, except the
MM iteration limit that was changed from 200 to 800. The
Our QSAR models and a visual comparison of active and inactive
compounds reveal no clear and straightforward interpretation. It is
possible that the salicylidene acylhydrazides inhibit T3S as metal
complexes containing one or two compounds or that binding to
the target involves a metal ion. The structural features that the
QSAR models indicate as critical for activity can therefore describe
properties important for a binding mode that involves the protein
target and one or more metal ions. Since the compounds were eval-
uated in cell-based systems the QSAR models most likely also cap-

2696
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
conformations of both BB sets were imported to MOE and the
MMFF94X force field partial charges were calculated. The confor-
mations were energy minimized in MOE using the MMFF94X force
field with a root mean square gradient of 0.1. For each resulting
conformation, the AM1 energy was calculated in MOE and the
conformation with lowest energy for each BB was saved. The low
energy conformations were then further geometry optimized using
a Hartree–Fock calculation in Jaguar.58 In Jaguar a number of
descriptors (e.g., polarizability and moment of inertia) were calcu-
lated. The geometry optimized structures were imported back into
MOE where atomic partial charges were assigned for each BB using
AM1 calculations. A second set of descriptors was then calculated
in MOE and added to the descriptors calculated in Jaguar, resulting
in a total set of 70 descriptors for the salicylic aldehydes and 97
descriptors for the hydrazides. The total descriptor set for each
BB set is given in Supplementary data (Tables S1 and S6).
bottom, white) bacterial solution (100
addition of compound (1 L DMSO solution) to give the final com-
pound concentrations 1.56, 3.13, 6.25, 12.5, 25, 50, 100, and
200 M. The plates were incubated on a rotary shaker for 1 h at
room temperature followed by 2 h at 37 °C. After incubation,
10 L of the bacterial suspension was transferred to new 96-well
plates (NuncTM, flat bottom, transparent) containing YopH sub-
strate mixture (90 L, 25 mM p-nitro phenyl phosphate, 40 mM
2-(N-morpholino)ethanesulfonic acid, pH 5.0, and 1.6 mM
dithiothreitol in MQ water) in each well. The plates were incubated
at 37 °C for 15 min before NaOH (20 lL, 1 M) was added to each
well to quench the reaction. The absorbance was measured at
lL) was added, followed by
l
l
l
l
405 nm in a microplate reader (TECAN Infinite M200). For the
luciferase assay, decanal solution (50 lL, 10 lL decanal/100 mL
water) was added to the first set of plates and the light emission
was immediately detected with a microplate reader (TECAN
Infinite M200, Gain 150, integration time of 20 ms). All liquid
handling during screening were performed with a BiomekÒ NXP.
Experiments were run in triplicates and reproduced at least three
independent times.
4.2. Statistical molecular design
The design procedure is summarized in Figure 2, and explained
in detail here. For the salicylic aldehydes, descriptors describing
hydrophobicity (Table S2) and charges (Table S3) were grouped
separately and imported to Simca 12.0.59 Data was centered and
scaled to unit variance (as was the case for all PCA and PLS models)
and two local PCA models were generated to summarize the infor-
mation (Table S4). There was some overlap of descriptors in both
these PCA models since some descriptors described both polarity
and charges. The number of significant components in all PCA
models was determined by the eigenvalue of the last component
(components with an normalized eigenvalue ꢀ2.0 were considered
significant), cross-validation using seven cross-validation rounds
as implemented in Simca 12.0,59 and inspection of the loading
plots to see that the components held unique information. Com-
pounds with eigenvalues lower than 2.0 were accepted, as long
as the other two criteria were fulfilled. The resulting PCA score vec-
tors were combined with the ab initio descriptors, molecular
weight and volume (Table S5). A new PCA model was computed
to compress the data. The PCA score vectors were extracted and
used as design variables in a two-layer linear DOOD, as imple-
mented in MODDE 8.0.2.60 The hydrazide descriptors were
summarized using PCA, and the PCA score vectors were used in a
second DOOD, performed in the same manner as for salicylic alde-
hydes. We did not have any indication about which properties of
the hydrazides that could correlate with the biological activity,
and therefore we did not do a block wise PCA of the BBs. The final
products were selected by having both BB sets in separate columns
where the BBs were in randomized order. The two columns were
combined to form the first 18 virtual products. The hydrazide BB
column was shifted one step downwards so that the last BB in
the column became the first. The two BB columns were combined
to form the second set of 18 virtual products. A second downward
shift of the hydrazide BB column followed by combination with the
salicylic aldehyde BB column gave the final 18 virtual products.
Growth inhibition was measured according to published proto-
cols.6,13 Wild-type Y. pseudotuberculosis pIB102 and E. coli DH5
a
were grown overnight in Luria broth (LB) at room temperature
and 37 °C, respectively, followed by a dilution to an OD600 = 0.1
in LB medium containing 2.5 mM CaCl2. To the diluted bacteria
(100
lL) compound (1
lL DMSO stock solution) was added to give
a final compound concentration of 50 and 100
l
M. The plates were
incubated in 37 °C and the absorbance was measured at 595 nm in
a plate reader (SpectraMAX 340) for 23 h (Fig. S5).
4.4. Multivariate QSAR modeling
Only compounds showing a dose–response were used to estab-
lish QSAR models, since compounds could be inactive due to sev-
eral reasons, like poor solubility, efflux, and inability to pass
through the double bacterial cell membrane. Compounds showing
luciferase signal inhibition, but lacking inhibition of the control
YopH (ME0163, ME0190–ME0193), were excluded from all model-
ing. In most attempts to establish QSAR models, compound
ME0167 was predicted to have much higher inhibition than the ac-
tual experimental value. Therefore, ME0167 was excluded from all
modeling, as the poor inhibition could be due to other factors than
low luciferase signal inhibition, such as efflux.
For all modeling the inhibitory values at 50
centration was used as Y-response, since the dataset had the most
even spread in biological activity at this concentration. At 100
concentration some compounds started to precipitate and at
25 M concentration there were too few active compounds (i.e.,
lM compound con-
lM
l
more than 40% inhibition). For PLS discriminant analysis (PLS-
DA) classification, salicylidene acylhydrazides were classified as
active if they inhibited the luciferase light emission signal by at
least 40% at 50 lM compound concentration. Two sets of training
set compounds were used in the modeling, one for Hi-PLS-DA
(Obs_1) and one for Hi-PLS (Obs_2) models (Table 4). Obs_2
consisted of all compounds showing a dose–response and three
4.3. General assay conditions
The biological evaluation of the salicylidene acylhydrazides was
performed in Y. pseudotuberculosis according to previously pub-
lished methods6,13 using a T3S linked luciferase reporter-gene
assay (Tables 1 and 3, S9a, and S9b) and an enzymatic YopH assay
(data not shown). Y. pseudotuberculosis serotype III strain YPIII
pIB102 (yopE-luxAB) was grown overnight, at room temperature,
in brain heart infusion (BHI) containing ethylene glycol tetraacetic
acid (5 mM) and MgCl2 (20 mM) for calcium depletion. The optical
density of the bacteria measured at 600 nm (OD600) was adjusted
to 0.088–0.095 in BHI. To each well in a 96-well plate (NuncTM, flat
Table 4
PLS score vector sets used for Hi-PLS-1 and Hi-PLS-DA-1
Score vector seta
Training setb
No. of score vectorsc
R2Y
SAL1
HYD1
SAL2
HYD2
Obs_1
Obs_1
Obs_2
Obs_2
16
16
14
14
0.53
0.27
0.64
0.33
a
b
c
Set of score vectors computed from PLS models based on BBs.
Training set used to compute the PLS score vectors.
Number of score vectors computed.

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2697
additional inactive compounds (ME0172, ME0181, ME0197) that
all were calculated as inactive in the models, which was not the
case for all inactive compounds synthesized and tested
biologically.
generated.61 A positive intercept for Q2 would indicate that most
models would be significant no matter the order of the response
and thus the model would be insignificant and likely just a result
of chance correlation.
The salicylidene acylhydrazides with a measured biological re-
sponse were disassembled into their BBs and two columns, one
for each BB set, were formed. To each BB column the total
descriptor sets with ab initio, MM and informatics descriptors
were added (Tables S1 and S6). The descriptor matrices were im-
ported into Simca 12.0.59 PLS regression was made of the matri-
ces describing salicylic aldehydes or hydrazides to the luciferase
signal inhibition in such a way that all BBs were listed together
with their inhibition values from the corresponding molecules
the BBs were used in. Thus a BB used in three salicylidene acy-
lhydrazides would be listed with three different inhibition values.
PLS components were added until the cumulative value of R2X
reached 1.0. The PLS score vectors from each BB set were
extracted and combined, forming a new data matrix consisting
of all training set compounds and their score vector descriptor.
The PLS score vectors were expanded with square and interaction
terms. A variable selection, fully described in Supplementary data,
was performed to compute the final models (Table S10 for Hi-
PLS-1 and Table S11 for Hi-PLS-DA-1).
The methodology used to compute Hi-PLS-2 (using Obs_2) and
Hi-PLS-DA-2 (using Obs_1) is summarized in Figure 6 and de-
scribed in detail here. Descriptors describing similar properties
were grouped into six groups for the salicylic aldehydes and five
for the hydrazides (Table S12). Those descriptors that could not
be assigned to a group (e.g., pKa and HOMO) were kept as indepen-
dent descriptors (X1U and X2U). The grouped descriptors were
summarized using local PCA models, generating the score vectors
X11 ꢁ X16 for the salicylic aldehydes and X21 ꢁ X25 for the hydra-
zides (Table S13). The PCA score vectors from the local models and
the independent descriptors from both BB sets were combined
(Table S14) and expanded by square and interaction terms. Follow-
ing variable selection the final models could be computed (Table
S15 for Hi-PLS-2 and Table S16 for Hi-PLS-DA-2).
4.5. Evaluation of models using an external test set
All the combinations of BBs used in the SMD were enumerated.
Compounds with a DModX,62 that is, the residual standard devia-
tion, cutoff higher than 1.40 in the PLS models and 1.33 in the
PLS-DA model were excluded. The total set was predicted in the
three models and the compounds predicted as active (ꢀ40% inhibi-
tion of luciferase light emission for Hi-PLS-1 and Hi-PLS-2 and
class membership to the actives for Hi-PLS-DA-1) were recorded.
The overlap between the models was then calculated, disregarding
any compounds having a higher DModX than the cutoff in any of
the three models.
DModX is essentially a tool to detect model outliers, but can
also be used to limit the test set depending on the goal of the pre-
dictions. A compound is considered non-deviating if it has a smal-
ler value than the critical distance, D-crit,59 in DModX. D-crit is
calculated from the F-distribution of the last component of the
model in question. If the goal is to improve existing QSAR models
by inclusion of a few additional compounds, then a low DModX
cutoff and a wide range of biological activity might be good
choices. If the aim is to find one potent inhibitor and a large syn-
thetic effort can be put into identifying this compound, a predicted
response over 90% and a DModX two or three times the D-Crit
value might be wise choices. If the goal is to establish models that
theoretically can predict the entire virtual set of compounds based
on commercially available BBs, a new SMD could be performed
using some or all of the coefficients from the QSAR model. Since
QSAR models are interpolating, the SMD should in this case cover
the outer domain of the virtual product space, possibly excluding
extreme outliers as indicated by DModX.
4.6. Yersinia pseudotuberculosis and enterotoxigenic
Escherichia coli infection of J774A Cells
The fraction of sum of squares of Y that can be explained by one
PLS component (R2Y), was calculated as:
R2Y ¼ SSmod=SStot
The experimental procedures were essentially carried out as
described before29 with compounds ME0165–ME0167, ME0174,
and ME0181. J774A cells were seeded out into 96-well plates
(4 ꢂ 104 cells/well) in Dulbecco’s modified Eagle’s medium
where SSmod is the sum of square of the responses corrected for the
mean and SStot the total sum of squares of Y corrected for the mean.
The fraction of the total variation of Y that can be predicted by
one component (Q2), as estimated by cross-validation, was calcu-
lated using seven cross-validation rounds according to:
(DMEM) with 10% fetal calf serum (FCS) and gentamicin (3 lg/
mL), grown for 24 h at 37 °C in 5% CO2. Y. pseudotuberculosis
wild-type pIB102 and pIB604 yopB (translocation-deficient T3S
mutant) were grown in LB medium supplemented with kanamy-
cin (25 g/mL) at room temperature overnight, diluted 1/10 in
D
Q2 ¼ 1 ꢁ PRESS=SStot
l
DMEM and incubated on a rotary shaker at room temperature
1 h and at 37 °C for 2 h. Enterotoxigenic E. coli (ETEC) was grown
in LB medium overnight at 37 °C, diluted as above and grown at
37 °C for 3 h. The J774A cells were washed with phosphate buf-
fered saline (PBS, 100 lL) then fresh DMEM (50 lL) containing
the different compounds (ME0165–ME0167, ME0174, and
ME0181, added from DMSO stock solutions) and bacteria
(50 lL, OD600 = 0.002) were added to give 25, 50, and 100 lM fi-
nal compound concentrations. The final DMSO concentration was
kept below 0.5%. The plate was incubated for 14 h at 37 °C in 5%
where PRESS is the predicted error sum of squares when all objects
have been left out once.
Four models were computed with logit transformed response,
as implemented in Simca 12.0,59 in order to compensate for the
non-normalized distribution in activity of the training set and S-
shaped curvature of the experimental versus calculated luciferase
signal inhibition. The logit transformed models did not differ from
the non-transformed models in their predictions and interpretabil-
ity. The logit transformed models are given fully in Supplementary
data (p. S35 for an overview of the logit transformed models, Figs.
S13–S23, and Tables S17–S22).
All PLS models were evaluated with permutation tests, using
Simca 12.0,59 where the order of the biological response was ran-
domized using 300 permutations for each model (Figs. S7, S8,
and S21–S23). The cumulative R2 and Q2 gave a regression line
where the intercept is an estimate of the significance of the models
CO2. CalceinAM (Invitrogen, 20 lL of a 6 lM solution in PBS) was
added and the plate was incubated for 40 min at 37 °C in 5% CO2.
The fluorescence was measured in a microplate reader (TECAN
Infinite M200) at 485/535 nm (Fig. 9 and S12) and the results
were confirmed using fluorescent microscopy (data not shown).
As a rescue control gentamycin (5
infected with ETEC.
lg/mL) was added to the cells

2698
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
4.7. Determination of minimum inhibitory concentration (MIC)
for Chlamydia trachomatis
and washed with water. If no precipitate could be formed, the
products was purified using preparative LC–MS. Compound puri-
ties, calculated from LC–UV traces, are given in Supplementary
data (Table S8).
HeLa 229 cells cultured in 96-well microtiter plates were in-
fected with C. trachomatis serovar L2 at multiplicity of infection
0.3. Compounds or solvent (DMSO final concentration 1%) were di-
luted in RPMI 1640 medium with supplements (10% fetal calf ser-
4.10.1. 4-Methyl-[1,2,3]thiadiazole-5-carboxylic acid (6-bromo-
2-hydroxy-3-methoxy-benzylidene)-hydrazide (ME0150)
Two rotamers were obtained in a 1:0.73 ratio as a result of a
dynamic process. 1H NMR (major rotamer): d 12.59–12.36 (m,
1H), 9.90 (s, 1H), 8.88 (s, 1H), 7.16 (d, J = 8.7 Hz, 1H), 7.01 (d,
J = 8.7 Hz, 1H), 3.81 (s, 3H), 2.87 (s, 3H); 1H NMR (minor rot-
amer): d 12.78 (br s, 1H), 9.90 (s, 1H), 8.41 (s, 1H), 7.15 (d,
J = 8.7 Hz, 1H), 7.01 (d, J = 8.7 Hz, 1H), 3.84 (s, 3H), 2.92 (s, 3H);
LC–MS [MꢁH+]ꢁ calcd for [C12H11BrN4O3S]: m/z: 368.96; found:
369.03.
um, 2 mM L-glutamine, 8 lg/mL gentamicin) and added to the
infected cells one hour after infection. After 35 h incubation at
37 °C in a humidified atmosphere of 5% (v/v) CO2/air, cells were
fixed with methanol, immunostained (Pathfinder, Biorad) and
examined using fluorescent microscopy at 200ꢂ magnification.
All compounds were tested at 50, 25, 12.5, 6.25, and 3.13 lM and
tests were performed in duplicates at least three times. Minimum
inhibitory concentration was defined as the lowest tested concen-
tration where no intracellular growth was observed.
4.10.2. Isonicotinic acid (6-bromo-2-hydroxy-3-methoxy-
benzylidene)-hydrazide (ME0151)
4.8. Cell viability assay
1H NMR: d 12.68 (s, 1H), 12.60 (s, 1H), 9.00 (s, 1H), 8.85–8.80
(m, 2H), 7.89–7.84 (m, 2H), 7.16 (d, J = 8.7 Hz, 1H), 7.01 (d,
J = 8.7 Hz, 1H), 3.81 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for
[C14H12BrN3O3]: m/z: 348.00; found: 348.18.
HeLa 229 cells cultured in 96-well microtiter plates were incu-
bated with test compounds (50
final concentration 1%) diluted in non-colored DMEM with supple-
ments (0.5% fetal calf serum, 2 mM -glutamine, 8 g/mL gentami-
lM and 25 lM) or solvent (DMSO
L
l
cin) for totally 28 h at 37 °C in a humidified atmosphere of 5% (v/v)
CO2/air. Cell viability was evaluated using a resazurin (blue and
4.10.3. Cyano-acetic acid (6-bromo-2-hydroxy-3-methoxy-
benzylidene)-hydrazide (ME0152)
non-fluorescent, Sigma 40
l
M final concentration), that is, reduced
Two rotamers were obtained in a 1:0.45 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.31 (br s, 1H), 12.23
(s, 1H), 8.69 (s, 1H), 7.15 (d, J = 8.7 Hz, 1H), 6.99 (d, J = 8.7 Hz,
1H), 3.89 (s, 2H), 3.80 (s, 3H); 1H NMR (minor rotamer): d 12.23
(s, 1H), 10.39 (s, 1H), 8.47 (s, 1H), 7.14 (d, J = 8.7 Hz, 1H), 6.98 (d,
J = 8.7 Hz, 1H), 4.23 (s, 2H), 3.80 (s, 3H); LC–MS [MꢁH+]ꢁ calcd
for [C11H10BrN3O3]: m/z: 309.98; found: 310.06.
to resorufin (pink and highly fluorescent) in living cells.63,64 Fluo-
rescence was measured using a Tecan Safire plate reader with exci-
tation at 535 nm and emission at 595 nm. Resazurin was added
four hours prior to measurement and cell-free wells with medium
and compound were used to rule out chemical reduction of resazu-
rin by the compounds. Scoring was made relative to the DMSO
control. In addition, cell viability was scored microscopically. All
experiments were performed in triplicates and repeated at least
three times.
4.10.4. 2-Bromo-5-methoxy-benzoic acid (4-chloro-2-hydroxy-
benzylidene)-hydrazide (ME0153)
Two rotamers were obtained in a 1:0.45 ratio as a result of a
dynamic process. 1H NMR (major rotamer): d 12.10 (s, 1H),
11.31 (s, 1H), 8.47 (s, 1H), 7.64 (d, J = 8.2 Hz, 1H), 7.60 (d,
J = 8.8 Hz, 1H), 7.17 (d, J = 3.0 Hz, 1H), 7.01–6.95 (m, 3H), 3.80
(s, 3H); 1H NMR (minor rotamer): d 12.10 (s, 1H), 10.32 (s, 1H),
8.26 (s, 1H), 7.56 (d, J = 8.7 Hz, 1H), 7.28 (d, J = 9.0 Hz, 1H), 7.04
(d, J = 3.1 Hz, 1H), 7.02 (d, J = 3.2 Hz, 1H), 6.89–6.84 (m, 2H),
3.77 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for [C15H12BrClN2O3]: m/z:
380.96; found: 381.03.
4.9. General chemistry
Mass-spectra were recorded by detecting negative (ESꢁ) molec-
ular ions with an electro spray Waters Micromass ZG 2000 instru-
ment using an XTerraÒ MS C18
5
lM 4.6 ꢂ 50 mm column and an
H2O/acetonitrile/formic acid eluent system. The same LC-system
was also used for purification with a preparative XTerraÒ Prep
MS C18
5
l
M 19 ꢂ 50 mm column and an H2O/acetonitrile eluent
system. Reflux reactions were performed using a Radley carousel
12 reaction station. Microwave heated reactions were performed
in EmrysTM process vials (20 mL) using a SmithCreatorTM micro-
wave instrument from Biotage. 1H NMR spectra were recorded in
DMSO-d6 (residual DMSO-d5, dY = 2.50 ppm, as internal standard)
using a 400 MHz Bruker spectrometer.
4.10.5. 4-Bromo-benzoic acid (4-chloro-2-hydroxy-
benzylidene)-hydrazide (ME0154)
Two rotamers were obtained in a 1:0.11 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.17 (s, 1H), 11.53 (s,
1H), 8.63 (s, 1H), 7.89 (d, J = 8.5 Hz, 2H), 7.77 (d, J = 8.5 Hz, 2H),
7.64 (d, J = 8.1 Hz, 1H), 7.02–6.95 (m, 2H); 1H NMR (minor rot-
amer): d 12.17 (s, 1H), 11.83 (s, 1H), 8.32 (br s, 1H), 7.89 (d,
J = 8.5 Hz, 2H), 7.77 (d, J = 8.5 Hz, 2H), 7.70 (br s, 1H), 7.46 (br s,
1H), 6.91 (br s, 1H); LC–MS [MꢁH+]ꢁ calcd for [C14H10BrClN2O2]:
m/z: 350.95; found: 351.09.
4.10. Typical procedure for the formation of salicylidene
acylhydrazides
Salicylic aldehyde (0.7–1.4 mmol, 1 equiv) and hydrazide
(1.4 equiv) were suspended in absolute ethanol (10 mL) and re-
fluxed for 4 h, or heated using
a
microwave instrument at
4.10.6. Adamantan-1-yl-acetic acid (4-chloro-2-hydroxy-
benzylidene)-hydrazide (ME0155)
155 °C for 7 min. Remaining solid was filtered off and analyzed
using LC–MS. If the filtrand was impure, the filtrate and ethanol
(100 mL) was added and the mixture was heated under stirring
until all material was thoroughly dissolved. If the mixture could
not be dissolved completely by the addition of ethanol, DMSO
(2–5 mL) was added. If a heterogeneous mixture still remained,
a warm filtration was performed. From the resulting solution a
pure precipitate was generally formed by the addition of cold
water. The precipitate was filtered off using suction filtration
Two rotamers were obtained in a 1:0.39 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.52 (s, 2H), 8.33 (s,
1H), 7.55 (d, J = 8.1 Hz, 1H), 6.96 (s, 1H), 6.95–6.91 (m, 1H), 1.97
(s, 2H), 1.93 (br s, 3H), 1.70–1.53 (m, 12H); 1H NMR (minor rot-
amer): d 11.21 (s, 1H), 10.58 (br s, 1H), 8.19 (s, 1H), 7.64 (d,
J = 8.8 Hz, 1H), 6.96 (br s, 1H), 6.95–6.91 (m, 1H), 2.40 (s, 2H),
1.93 (br s, 3H), 1.70–1.53 (m, 12H); LC–MS [MꢁH+]ꢁ calcd for
[C19H23ClN2O2]: m/z: 345.13; found: 345.26.

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2699
4.10.7. 3,4-Dimethoxy-benzoic acid (2-hydroxy-3,5-diiodo-
4.10.14. 2-Oxo-2-[N0-(2,4,6-trihydroxy-benzylidene)-
benzylidene)-hydrazide (ME0156)
hydrazino]-acetamide (ME0163)
1H NMR: d 13.02 (s, 1H), 12.35 (s, 1H), 8.43 (s, 1H), 8.04 (s, 1H),
7.86 (d, J = 1.8 Hz, 1H), 7.61 (d, J = 9.0 Hz, 1H), 7.52 (s, 1H), 7.12 (d,
J = 8.5 Hz, 1H), 3.85 (s, 6H); LC–MS [MꢁH+]ꢁ calcd for
[C16H14I2N2O4]: m/z: 550.89; found: 550.99.
1H NMR: d 12.26 (s, 1H), 11.02 (s, 2H), 9.85 (s, 1H), 8.87 (s, 1H),
8.22 (s, 1H), 7.91 (s, 1H), 5.81 (s, 2H); LC–MS [MꢁH+]ꢁ calcd for
[C9H9N3O5]: m/z: 238.04; found: 238.17.
4.10.15. 3-Trifluoromethyl-benzoic acid (3-fluoro-2-hydroxy-
benzylidene)-hydrazide (ME0164)
4.10.8. (4-Fluoro-phenoxy)-acetic acid (2-hydroxy-3,5-diiodo-
benzylidene)-hydrazide (ME0157)
Two rotamers were obtained although the minor rotamer was
not assigned due to a ratio of less than 0.02:1. The amide and phe-
nol protons were exchange broadened. 1H NMR: d 8.68 (s, 1H), 8.28
(s, 1H), 8.25 (d, J = 8.0 Hz, 1H), 7.77 (d, J = 8.0 Hz, 1H), 7.81 (t,
J = 7.8 Hz, 1H), 7.43 (d, J = 7.8 Hz, 1H), 7.34–7.25 (m, 1H), 6.97–
6.89 (m, 1H); LC–MS [MꢁH+]ꢁ calcd for [C15H10F4N2O2]: m/z:
325.06; found: 325.19.
Two rotamers were obtained in a 1:0.22 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.66 (s, 1H), 12.23 (s,
1H), 8.37 (s, 1H), 8.05 (d, J = 1.9 Hz, 1H), 7.85 (d, J = 1.9 Hz, 1H),
7.20–7.13 (m, 2H), 7.06–7.00 (m, 2H), 4.73 (s, 2H); 1H NMR (minor
rotamer): d 11.82 (s, 1H), 10.68 (br s, 1H), 8.09 (s, 1H), 7.88 (br s,
1H), 8.05 (d, J = 1.9 Hz, 1H), 7.14–7.08 (m, 2H), 6.98–6.93 (m,
2H), 5.13 (s, 2H); LC–MS [MꢁH+]ꢁ calcd for [C15H11FI2N2O3]: m/z:
538.87; found: 538.98.
4.10.16. (4-Nitro-phenoxy)-acetic acid (3-fluoro-2-hydroxy-
benzylidene)-hydrazide (ME0165)
4.10.9. 3,5-Dichloro-benzoic acid (2-hydroxy-3,5-diiodo-
benzylidene)-hydrazide (ME0158)
Two rotamers were obtained in a 1:0.92 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.72 (s, 1H), 11.29 (br
s, 1H), 8.33 (s, 1H), 8.20 (d, J = 9.3 Hz, 2H), 7.57 (d, J = 7.9 Hz, 1H),
7.16 (d, J = 9.3 Hz, 2H), 7.31–7.24 (m, 1H), 6.94–6.83 (m, 1H),
5.34 (s, 2H); 1H NMR (minor rotamer): d 11.98 (br s, 1H), 10.22
(br s, 1H), 8.56 (s, 1H), 8.25 (d, J = 9.2 Hz, 2H), 7.39 (d, J = 7.9 Hz,
1H), 7.22 (d, J = 9.2 Hz, 2H), 7.31–7.24 (m, 1H), 6.94–6.83 (m,
1H), 4.91 (s, 2H); LC–MS [MꢁH+]ꢁ calcd for [C15H12FN3O5]: m/z:
332.07; found: 332.15.
1H NMR: d 12.71, (br s, 2H), 8.43 (s, 1H), 8.06 (d, J = 1.9 Hz, 1H),
7.96 (d, J = 1.8 Hz, 2H), 7.92 (br s, 2H); LC–MS [MꢁH+]ꢁ calcd for
[C14H8Cl2I2N2O2]: m/z: 558.80; found: 558.89.
4.10.10. 3,4,5-Trimethoxy-benzoic acid (2-hydroxy-3-methoxy-
benzylidene)-hydrazide (ME0159)
1H NMR: d 11.94 (s, 1H), 10.93 (s, 1H), 8.66 (s, 1H), 7.26 (s, 2H),
7.16 (d, J = 7.7 Hz, 1H), 7.03 (d, J = 7.6 Hz, 1H), 6.87 (t, J = 7.9 Hz,
1H), 3.86 (s, 6H), 3.81 (s, 3H), 3.73 (s, 3H); LC–MS [MꢁH+]ꢁ calcd
for [C18H20N2O6]: m/z: 359.12; found: 359.20.
4.10.17. 3-Methoxy-benzoic acid (3-fluoro-2-hydroxy-
benzylidene)-hydrazide (ME0166)
1H NMR: d 12.20 (s, 1H), 11.61 (s, 1H), 8.66 (s, 1H), 7.53 (d,
J = 7.7 Hz, 1H), 7.49 (d, J = 2.3 Hz, 1H), 7.46 (d, J = 7.8 Hz, 1H),
7.39 (d, J = 7.8 Hz, 1H), 7.19 (dd, J = 8.0 Hz, J = 1.6 Hz, 1H), 7.31–
7.26 (m, 1H), 6.96–6.87 (m, 1H), 3.84 (s, 1H); LC–MS [MꢁH+]ꢁ
calcd for [C15H13FN2O3]: m/z: 287.08; found: 287.22.
4.10.11. 2-Nitro-benzoic acid (2-hydroxy-3-methoxy-
benzylidene)-hydrazide (ME0160)
Two rotamers were obtained in a 1:0.70 ratio as a result of a
dynamic process. The phenol protons were exchange broadened.
1H NMR (major rotamer): d 9.28 (s, 1H), 8.31 (s, 1H), 8.21–8.17
(m, 1H), 7.91–7.85 (m, 1H), 7.82–7.72 (m, 1H), 7.69–7.65 (m,
1H), 6.94–6.90 (m, 1H), 6.77–6.73 (m, 1H), 6.72–6.66 (m, 1H),
3.76 (s, 3H); 1H NMR (minor rotamer): d 10.51 (s, 1H), 8.50 (s,
1H), 8.17–8.13 (m, 1H), 7.91–7.85 (m, 1H), 7.82–7.72 (m, 2H),
7.23–7.18 (m, 1H), 7.07–7.02 (m, 1H), 6.90–6.84 (m, 1H), 3.82
(s, 3H)); LC–MS [MꢁH+]ꢁ calcd for [C15H13N3O5]: m/z: 314.08;
found: 314.20.
4.10.18. 4-Methyl-[1,2,3]thiadiazole-5-carboxylic acid (3-
chloro-5-fluoro-2-hydroxy-benzylidene)-hydrazide (ME0167)
Two rotamers were obtained in a 1:0.10 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.39 (s, 1H), 10.57 (s,
1H), 8.46 (s, 1H), 7.92 (d, J = 2.4 Hz, 1H), 7.46 (dd, J = 8.7 Hz,
J = 2.4 Hz, 1H), 6.91 (d, J = 8.8 Hz, 1H), 2.97 (s, 3H); 1H NMR (minor
rotamer): d 12.39 (s, 1H), 10.97 (s, 1H), 8.55 (s, 1H), 7.84–7.80 (m,
1H), 7.46 (dd, J = 8.7 Hz, J = 2.4 Hz, 1H), 6.91 (d, J = 8.8 Hz, 1H), 2.84
(s, 3H); LC–MS [MꢁH+]ꢁ calcd for [C11H8ClFN4O2S]: m/z: 338.95;
found: 339.05.
4.10.12. 3-(4-Hydroxy-phenyl)-propionic acid (2,4,6-
trihydroxy-benzylidene)-hydrazide (ME0161)
Two rotamers were obtained in a 1:0.19 ratio as a result of a
dynamic process. 1H NMR (major rotamer): d 11.32 (s, 1H),
10.95 (s, 2H), 9.76 (s, 1H), 9.15 (s, 1H), 8.46 (s, 1H), 7.11–6.94
(m, 2H), 6.76–6.60 (m, 2H), 5.81 (s, 2H), 2.84–2.71 (m, 2H),
2.45–2.36 (m, 2H); 1H NMR (minor rotamer): d 11.11 (s, 1H),
10.52 (s, 2H), 9.76 (s, 1H), 9.15 (s, 1H), 8.36 (s, 1H), 7.11–6.94
(m, 2H), 6.76–6.60 (m, 2H), 5.82 (s, 2H), 2.84–2.71 (m, 2H),
2.69–2.63 (m, 2H); LC–MS [MꢁH+]ꢁ calcd for [C16H16N2O5]: m/z:
315.10; found: 315.18.
4.10.19. Isonicotinic acid (5-bromo-2-hydroxy-benzylidene)-
hydrazide (ME0168)
Two rotamers were obtained in a 1:0.10 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.35 (s, 1H), 11.12 (s,
1H), 8.88–8.70 (m, 2H), 8.65 (s, 1H), 7.92–7.75 (m, 3H), 7.44 (dd,
J = 8.7 Hz, J = 2.1 Hz, 1H), 6.91 (d, J = 8.7 Hz, 1H); 1H NMR (minor
rotamer): d 12.08 (br s, 1H), 10.31 (br s, 1H), 8.88–8.70 (m, 2H),
8.31 (s, 1H), 7.92–7.75 (m, 1H), 7.63 (br s, 2H), 7.35 (d, J = 7.9 Hz,
1H), 6.83 (d, J = 8.9 Hz, 1H); LC–MS [MꢁH+]ꢁ calcd for
[C13H10BrN3O2]: m/z: 317.99; found: 318.01.
4.10.13. 3-Chloro-benzo[b]thiophene-2-carboxylic acid (2,4,6-
trihydroxy-benzylidene)-hydrazide (ME0162)
Two rotamers were obtained in a 1:0.20 ratio as a result of a
dynamic process. 1H NMR (major rotamer): d 12.03 (2, 1H),
11.01 (s, 2H), 9.89 (s, 1H), 8.77 (s, 1H), 8.20–8.12 (m, 1H),
7.98–7.87 (m, 1H), 7.66–7.56 (m, 2H), 5.85 (s, 2H); 1H NMR (min-
or rotamer): d 12.12 (br s, 1H), 10.20 (s, 2H), 9.89 (s, 1H), 8.51 (s,
1H), 8.20–8.12 (m, 1H), 7.98–7.87 (m, 1H), 7.66–7.56 (m, 2H),
5.72 (s, 2H); LC–MS [MꢁH+]ꢁ calcd for [C16H11ClN2O4S]: m/z:
361.00; found: 361.08.
4.10.20. Cyano-acetic acid (5-bromo-2-hydroxy-benzylidene)-
hydrazide (ME0169)
Two rotamers were obtained in a 1:0.36 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.23 (s, 1H), 7.85 (d,
J = 2.5 Hz, 1H), 7.38 (dd, J = 8.7 Hz, J = 2.5 Hz, 1H), 6.91–6.83 (m,
1H), 4.24 (s, 2H); 1H NMR (minor rotamer): d 8.36 (s, 1H), 7.77

2700
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
(br s, 1H), 7.45–7.40 (m, 1H), 6.91–6.83 (m, 1H), 3.82 (s, 2H); LC–
(s, 3H); LC–MS [MꢁH+]ꢁ calcd for [C17H16Cl2N2O5]: m/z: 397.03;
MS [MꢁH+]ꢁ calcd for [C10H8BrN3O2]: m/z: 279.97; found: 280.10.
found: 397.10.
4.10.21. 2-Bromo-5-methoxy-benzoic acid (5-tert-butyl-2-
hydroxy-benzylidene)-hydrazide (ME0170)
4.10.28. 2-Nitro-benzoic acid (3,5-dichloro-2-hydroxy-
benzylidene)-hydrazide (ME0177)
Two rotamers were obtained in a 1:0.47 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.06 (s, 1H), 10.82 (s,
1H), 8.49 (s, 1H), 7.60 (d, J=8.8 Hz, 1H), 7.57–7.52 (m, 1H), 7.35 (dd,
J = 8.6 Hz, J = 2.4 Hz, 1H), 7.17 (d, J = 3.0 Hz, 1H), 7.02 (d, J = 3.1 Hz,
1H), 6.87 (d, J = 8.6 Hz, 1H), 3.81 (s, 3H), 1.27 (s, 9H); 1H NMR (min-
or rotamer): d 12.06 (s, 1H), 9.64 (s, 1H), 8.24 (s, 1H), 7.57–7.52 (m,
1H), 7.31 (d, J = 2.4 Hz, 1H), 7.23 (dd, J = 8.6 Hz, J = 2.4 Hz, 1H), 7.04
(d, J = 3.0 Hz, 1H), 6.98 (d, J = 8.8 Hz, J = 3.1 Hz, 1H), 6.72 (d,
J = 8.6 Hz, 1H), 3.77 (s, 3H), 1.17 (s, 9H); LC–MS [MꢁH+]ꢁ calcd
for [C19H21BrN2O3]: m/z: 403.06; found: 403.18.
Two rotamers were obtained in a 1:0.72 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.65 (br s, 1H), 12.35
(s, 1H), 8.43 (s, 1H), 8.18 (d, J = 8.0 Hz, 1H), 7.94–7.87 (m, 1H),
7.82 (d, J = 7.3 Hz, 2H), 7.72–7.68 (m, 1H), 7.22 (d, J = 7.2 Hz, 1H);
1H NMR (minor rotamer): d 12.13 (br s, 1H), 10.23 (br s, 1H),
8.26 (s, 1H), 8.22 (d, J = 8.1 Hz, 1H), 7.94–7.87 (m, 1H), 7.81–7.77
(m, 1H), 7.72–7.68 (m, 1H), 7.65 (d, J = 2.2 Hz, 1H), 7.53 (d,
J = 2.1 Hz, 1H); LC–MS [MꢁH+]ꢁ calcd for [C14H9Cl2N3O4]: m/z:
351.99; found: 352.15.
4.10.29. 3-(4-Hydroxy-phenyl)-propionic acid (2-hydroxy-4-
methoxy-benzylidene)-hydrazide (ME0178)
4.10.22. 4-Bromo-benzoic acid (5-tert-butyl-2-hydroxy-
benzylidene)-hydrazide (ME0171)
Two rotamers were obtained in a 1:0.40 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.49 (br s, 1H), 11.45
(s, 1H), 9.15 (s, 1H), 8.23 (s, 1H), 7.37 (d, J = 8.6 Hz, 1H), 7.06–
6.99 (m, 2H), 6.69–6.64 (m, 2H), 6.49–6.47 (m, 1H), 6.46 (d,
J = 2.3 Hz, 1H), 3.76 (s, 3H), 2.81–2.74 (m, 2H), 2.47–2.41 (m,
2H); 1H NMR (minor rotamer): d 11.11 (s, 1H), 10.31 (br s, 1H),
9.15 (s, 1H), 8.15 (s, 1H), 7.48 (d, J = 8.6 Hz, 1H), 7.06–6.99 (m,
2H), 6.69–6.64 (m, 2H), 6.50 (d, J = 2.4 Hz, 1H), 6.43 (d, J = 2.3 Hz,
1H), 3.73 (s, 3H), 2.81–2.74 (m, 4H); LC–MS [MꢁH+]ꢁ calcd for
[C17H18N2O4]: m/z: 313.12; found: 313.20.
Two rotamers were obtained in a 1:0.04 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.13 (s, 1H), 11.00 (s,
1H), 8.65 (s, 1H), 7.89 (d, J = 8.5 Hz, 2H), 7.77 (d, J = 8.5 Hz, 2H),
7.52 (d, J = 2.3 Hz, 1H), 7.35 (dd, J = 8.6 Hz, J = 2.4 Hz, 1H), 6.87 (d,
J = 8.6 Hz, 1H), 1.28 (s, 9H); 1H NMR (minor rotamer): d 11.83 (br
s, 1H), 9.81 (br s, 1H), 8.31 (s, 1H), 7.74–7.64 (m, 4H), 7.45 (br s,
1H), 7.26 (br s, 1H), 6.79 (br s, 1H), 1.21 (s, 9H); LC–MS [MꢁH+]ꢁ
calcd for [C18H19BrN2O2]: m/z: 373.05; found: 373.06.
4.10.23. Adamantan-1-yl-acetic acid (5-tert-butyl-2-hydroxy-
benzylidene)-hydrazide (ME0172)
4.10.30. 3-Chloro-benzo[b]thiophene-2-carboxylic acid (2-
hydroxy-4-methoxy-benzylidene)-hydrazide (ME0179)
Two rotamers were obtained in a 1:0.33 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.07 (s, 1H), 11.30 (s,
1H), 8.56 (s, 1H), 8.21–8.10 (m, 1H), 7.98–7.91 (m, 1H), 7.68–7.58
(m, 2H), 7.50 (d, J = 8.5 Hz, 1H), 6.57–6.45 (m, 2H), 3.79 (s, 3H); 1H
NMR (minor rotamer): d 12.07 (s, 1H), 10.14 (s, 1H), 8.34 (s, 1H),
8.21–8.10 (m, 1H), 7.98–7.91 (m, 1H), 7.68–7.58 (m, 2H), 7.55 (d,
J = 9.1 Hz, 1H), 6.57–6.45 (m, 1H), 6.40 (br s, 1H), 3.72 (s, 3H);
LC–MS [MꢁH+]ꢁ calcd for [C17H13ClN2O3S]: m/z: 359.02; found:
359.10.
Two rotamers were obtained in a 1:0.34 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.45 (br s, 1H), 11.00
(br s, 1H), 8.35 (s, 1H), 7.46 (d, J = 2.4 Hz, 1H), 7.31 (dd, J = 8.6 Hz,
J = 2.5 Hz, 1H), 6.89–6.77 (m, 1H), 1.97 (s, 2H), 1.93 (br s, 3H),
1.78–1.52 (m, 12H), 1.26 (s, 9H); 1H NMR (minor rotamer): d
11.19 (br s, 1H), 9.93 (br s, 1H), 8.19 (s, 1H), 7.65 (d, J = 2.4 Hz,
1H), 7.26 (dd, J = 8.6 Hz, J = 2.5 Hz, 1H), 6.89–6.77 (m, 1H), 2.41
(s, 2H), 1.93 (br s, 3H), 1.78–1.52 (m, 12H), 1.25 (s, 9H); LC–MS
[MꢁH+]ꢁ calcd for [C23H32N2O2]: m/z: 367.24; found: 367.28.
4.10.24. 3,4-Dimethoxy-benzoic acid (3-chloro-5-fluoro-2-
hydroxy-benzylidene)-hydrazide (ME0173)
4.10.31. 2-[N0-(2-Hydroxy-4-methoxy-benzylidene)-
hydrazino]-2-oxo-acetamide (ME0180)
1H NMR: d 12.28 (s, 2H), 8.57 (s, 1H), 7.61 (d, J = 8.3 Hz, 1H),
7.55–7.44 (m, 3H), 7.12 (d, J = 8.5 Hz, 1H), 3.85 (s, 6H); LC–MS
[MꢁH+]ꢁ calcd for [C16H14ClFN2O4]: m/z: 351.05; found: 351.10.
1H NMR: d 12.29 (s, 1H), 11.39 (s, 1H), 8.64 (s, 1H), 8.26 (s, 1H),
7.95 (s, 1H), 7.37 (d, J = 8.6 Hz, 1H), 6.52 (dd, J = 8.5 Hz, J = 2.4 Hz,
1H), 6.48 (d, J = 2.4 Hz, 1H), 3.77 (s, 3H); LC–MS [MꢁH+]ꢁ calcd
for [C10H11N3O4]: m/z: 236.06; found: 236.21.
4.10.25. 3-(4-Fluoro-phenyl)-propionic acid (3-chloro-5-fluoro-
2-hydroxy-benzylidene)-hydrazide (ME0174)
Two rotamers were obtained in a 1:0.52 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.52 (s, 1H), 7.62–7.40
(m, 2H), 7.26–6.91 (m, 4H), 4.72 (s, 2H); 1H NMR (minor rotamer):
d 8.27 (s, 1H), 7.62–7.40 (m, 2H), 7.26–6.91 (m, 4H), 5.14 (s, 2H);
LC–MS [MꢁH+]ꢁ calcd for [C16H13ClF2N2O2]: m/z: 339.03; found:
339.12.
4.10.32. 3-Trifluoromethyl-benzoic acid (3,5-di-tert-butyl-2-
hydroxy-benzylidene)-hydrazide (ME0181)
1H NMR: d 12.40 (s, 1H), 12.20 (s, 1H), 8.60 (s, 1H), 8.28 (s, 1H),
8.25 (d, J = 7.9 Hz, 1H), 8.01 (d, J = 7.8 Hz, 1H), 7.83 (t, J = 7.8 Hz,
1H), 7.33 (d, J = 2.2 Hz, 1H), 7.25 (d, J = 2.2 Hz, 1H), 1.42 (s, 9H),
1.29 (s, 9H); LC–MS [MꢁH+]ꢁ calcd for [C23H27F3N2O2]: m/z:
419.19; found: 419.25.
4.10.26. 3,5-Dichloro-benzoic acid (3-chloro-5-fluoro-2-
hydroxy-benzylidene)-hydrazide (ME0175)
4.10.33. (4-Nitro-phenoxy)-acetic acid (3,5-di-tert-butyl-2-
hydroxy-benzylidene)-hydrazide (ME0182)
1H NMR: d 12.44 (br s, 1H), 12.06 (br s, 1H), 8.58 (s, 1H), 7.97 (d,
J = 1.9 Hz, 2H), 7.92 (t, J = 1.9 Hz, 1H), 7.55–7.50 (m, 2H); LC–MS
[MꢁH+]ꢁ calcd for [C14H8Cl3FN2O2]: m/z: 358.95; found: 359.04.
Two rotamers were obtained in a 1:0.28 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.11 (br s, 1H), 11.97
(br s, 1H), 8.46 (s, 1H), 8.25 (d, J = 9.1 Hz, 2H), 7.30 (s, 1H), 7.23
(s, 1H), 7.22 (d, J = 9.2 Hz, 2H), 4.92 (s, 2H), 1.39 (s, 9H), 1.28 (s,
9H); 1H NMR (minor rotamer): d 12.11 (br s, 1H), 10.46 (s, 1H),
8.46 (s, 1H), 8.20 (d, J = 9.5 Hz, 2H), 8.19 (s, 1H), 7.30 (s, 1H), 7.18
(d, J = 9.2 Hz, 2H), 5.34 (s, 2H), 1.39 (s, 9H), 1.28 (s, 9H); LC–MS
[MꢁH+]ꢁ calcd for [C23H29N3O5]: m/z: 426.20; found: 426.25.
4.10.27. 3,4,5-Trimethoxy-benzoic acid (3,5-dichloro-2-
hydroxy-benzylidene)-hydrazide (ME0176)
1H NMR: d 12.45 (br s, 1H), 12.37 (br s, 1H), 8.57 (s, 1H), 7.65 (d,
J = 2.4 Hz, 1H), 7.62 (d, J = 2.4 Hz, 1H), 7.28 (s, 2H), 3.87 (s, 6H), 3.74

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2701
4.10.34. 3-Methoxy-benzoic acid (3,5-di-tert-butyl-2-hydroxy-
benzylidene)-hydrazide (ME0183)
8.65 (d, J = 2.9 Hz, 1H), 8.59 (d, J = 2.9 Hz, 1H), 8.30 (s, 1H), 2.44
(s, 2H), 1.94 (br s, 3H), 1.74–1.53 (m, 12H); LC–MS [MꢁH+]ꢁ calcd
for [C19H22N4O6]: m/z: 401.14; found: 401.19.
1H NMR: d 12.27 (s, 1H), 12.16 (s, 1H), 8.58 (s, 1H), 7.52 (d,
J = 7.7 Hz, 1H), 7.48 (t, J = 7.7 Hz, 1H), 7.47 (d, J = 7.7 Hz, 1H), 7.32
(d, J = 2.1 Hz, 1H), 7.22 (d, J = 2.2 Hz, 1H), 7.22–7.17 (m, 1H), 3.85
(s, 3H), 1.42 (s, 9H), 1.29 (s, 9H); LC–MS [MꢁH+]ꢁ calcd for
[C23H30N2O3]: m/z: 381.22; found: 381.29.
4.10.41. 3,4-Dimethoxy-benzoic acid (4-diethylamino-2-
hydroxy-benzylidene)-hydrazide (ME0190)
1H NMR: d 11.63 (s, 1H), 11.51 (s, 1H), 8.41 (s, 1H), 7.55 (dd,
J = 8.3 Hz, J = 1.9 Hz, 1H), 7.48 (d, J = 1.9 Hz), 1H), 7.18 (d,
J = 8.8 Hz, 1H), 7.08 (d, J = 8.5 Hz, 1H), 6.27 (dd, J = 8.7 Hz,
J = 2.3 Hz, 1H), 6.12 (d, J = 2.3 Hz, 1H), 3.83 (s, 3H), 3.83 (s, 3H),
3.36 (q, J = 7.0 Hz, 4H), 1.11 (t, J = 7.0 Hz, 6H); LC–MS [MꢁH+]ꢁ
calcd for [C20H25N3O4]: m/z: 370.17; found: 370.28.
4.10.35. 3-Trifluoromethyl-benzoic acid (2-hydroxy-3,5-
dimethoxy-benzylidene)-hydrazide (ME0184)
Two rotamers were obtained in a 1:0.09 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.18 (s, 1H), 10.03 (s,
1H), 8.70 (s, 1H), 8.28 (s, 1H), 8.24 (d, J = 2.8 Hz, 1H), 4.87 (s, 2H),
3.79 (s, 3H), 3.72 (s, 3H); 1H NMR (minor rotamer): d 11.94 (s,
1H), 10.03 (s, 1H), 8.90 (s, 1H), 8.42 (s, 1H), 8.16 (br s, 1H), 8.08
(d, J = 8.0 Hz, 1H), 7.92 (br s, 1H), 6.59 (br s, 1H), 6.55 (br s, 1H),
3.78 (s, 3H), 3.61 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for [C17H15F3N2O4]:
m/z: 367.09; found: 367.11.
4.10.42. (4-Fluoro-phenoxy)-acetic acid (4-diethylamino-2-
hydroxy-benzylidene)-hydrazide (ME0191)
Two rotamers were obtained in a 1:0.34 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.45 (br s, 1H), 11.20
(s, 1H), 8.33 (s, 1H), 7.20–7.07 (m, 3H), 7.06–6.99 (m, 2H), 6.28–
6.23 (m, 1H), 6.13–6.08 (m, 1H), 4.62 (s, 2H), 3.39–3.26 (m, 4H),
1.10 (t, J = 7.0 Hz, 6H); 1H NMR (minor rotamer): d 11.20 (s, 1H),
9.83 (br s, 1H), 8.10 (s, 1H), 7.36 (d, J = 8.8 Hz, 1H), 7.20–7.07 (m,
2H), 6.96–6.89 (m, 2H), 6.25–6.21 (m, 1H), 6.13–6.08 (m, 1H),
5.01 (s, 2H), 3.39–3.26 (m, 4H), 1.10 (t, J = 7.0 Hz, 6H); LC–MS
[MꢁH+]ꢁ calcd for [C19H22FN3O3]: m/z: 358.15; found: 358.23.
4.10.36. Isonicotinic acid (2-hydroxy-3-methoxy-5-nitro-
benzylidene)-hydrazide (ME0185)
Two rotamers were obtained in a 1:0.08 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.84–8.79 (m, 2H), 8.79
(s, 1H), 8.31–8.27 (m, 1H), 7.87–7.82 (m, 2H), 7.81–7.77 (m, 1H),
3.97 (s, 3H); 1H NMR (minor rotamer): d 8.77–8.68 (m, 2H), 8.42
(br s, 1H), 7.93 (br s, 1H), 7.72 (br s, 1H), 7.65 (br s, 2H), 3.94 (s,
3H); LC–MS [MꢁH+]ꢁ calcd for [C14H12N4O5]: m/z: 315.07; found:
315.15.
4.10.43. 3,5-Dichloro-benzoic acid (4-diethylamino-2-hydroxy-
benzylidene)-hydrazide (ME0192)
Two rotamers were obtained in a 1:0.03 ratio as a result of a dy-
namic process. Due to the small integrals of the minor rotamer, the
amide proton could not be identified. 1H NMR (major rotamer): d
11.92 (s, 1H), 11.25 (s, 1H), 8.42 (s, 1H), 7.93 (d, J = 1.8 Hz, 2H),
7.87–7.83 (m, 1H), 7.23 (d, J = 8.8 Hz, 1H), 6.27 (dd, J = 8.8 Hz,
J = 2.2 Hz, 1H), 6.12 (d, J = 2.1 Hz, 1H), 3.36 (q, J = 7.0 Hz, 4H),
1.10 (t, J = 7.0 Hz, 6H); 1H NMR (minor rotamer): d 9.81 (s, 1H),
8.14 (s, 1H), 7.78 (br s, 1H), 7.71 (br s, 2H), 7.15 (d, J = 8.8 Hz,
1H), 6.21 (d, J = 9.5 Hz, 1H), 6.06 (br s, 1H), 3.47–3.40 (m, 4H),
1.05 (t, J = 7.0 Hz, 6H); LC–MS [MꢁH+]ꢁ calcd for [C18H19Cl2N3O2]:
m/z: 378.08; found: 378.08.
4.10.37. Cyano-acetic acid (2-hydroxy-3-methoxy-5-nitro-
benzylidene)-hydrazide (ME0186)
Two rotamers were obtained in a 1:0.40 ratio as a result of a dy-
namic process. The phenol protons were exchange broadened. 1H
NMR (major rotamer): d 11.87 (s, 1H), 8.33 (s, 1H), 8.25–8.19 (m,
1H), 7.74 (d, J = 2.6 Hz, 1H), 4.26 (s, 2H), 3.96 (s, 3H); 1H NMR (min-
or rotamer): d 11.87 (s, 1H), 8.48 (s, 1H), 8.25–8.19 /m, 1H), 7.76 (d,
J = 2.6 Hz, 1H), 3.95 (s, 3H), 3.85 (s, 2H); LC–MS [MꢁH+]ꢁ calcd for
[C11H10N4O5]: m/z: 277.06; found: 277.18.
4.10.44. 3,4,5-Trimethoxy-benzoic acid (2,3,4-trihydroxy-
benzylidene)-hydrazide (ME0193)
4.10.38. 2-Bromo-5-methoxy-benzoic acid (2-hydroxy-3,5-
dinitro-benzylidene)-hydrazide (ME0187)
1H NMR: d 11.80 (s, 1H), 11.48 (s, 1H), 9.46 (s, 1H), 8.50 (s, 1H),
8.47 (s, 1H), 7.25 (s, 2H), 6.79 (d, J = 8.4 Hz, 1H), 6.40 (d, J = 8.4 Hz,
1H), 3.87 (s, 6H), 3.73 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for
[C17H18N2O7]: m/z: 361.10; found: 361.14.
Two rotamers were obtained in a 1:0.32 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.57 (br s, 1H), 12.37
(s, 1H), 8.81 (d, J = 2.9 Hz, 1H), 8.76 (d, J = 2.9 Hz, 1H), 8.66 (s,
1H), 7.62 (d, J = 8.9 Hz, 1H), 7.20 (d, J = 3.0 Hz, 1H), 7.08–7.02 (m,
1H), 3.81 (s, 3H); 1H NMR (minor rotamer): d 12.57 (br s, 1H),
12.37 (s, 1H), 8.65 (d, J = 2.9 Hz, 1H), 8.43 (s, 1H), 8.39 (d,
J = 2.9 Hz, 1H), 7.59 (d, J = 8.8 Hz, 1H), 7.08–7.02 (m, 1H), 7.01 (d,
J = 3.1 Hz, 1H), 3.79 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for
[C15H11BrN4O7]: m/z: 436.97; found: 436.97.
4.10.45. 2-Nitro-benzoic acid (2,3,4-trihydroxy-benzylidene)-
hydrazide (ME0194)
Two rotamers were obtained in a 1:0.80 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.14 (br s, 1H), 11.91
(br s, 1H), 11.08 (br s, 1H), 9.51 (br s, 1H), 8.30 (s, 1H), 8.14 (d,
J = 7.3 Hz, 1H), 7.90–7.85 (m, 1H), 7.79 (d, J = 7.3 Hz, 1H), 7.77–
7.73 (m, 1H), 6.83 (d, J = 8.5 Hz, 1H), 6.40 (d, J = 8.5 Hz, 1H); 1H
NMR (minor rotamer): d 9.51 (br s, 1H), 9.15 (br s, 1H), 8.50 (br
s, 1H), 8.37 (br s, 1H), 8.19 (dd, J = 8.2 Hz, J = 0.9 Hz, 1H), 8.13 (s,
1H), 7.90–7.85 (m, 1H), 7.79 (d, J = 7.3 Hz, 1H), 7.66 (dd,
J = 7.6 Hz, J = 1.3 Hz, 1H), 6.57 (d, J = 8.6 Hz, 1H), 6.27 (d,
J = 8.5 Hz, 1H); LC–MS [MꢁH+]ꢁ calcd for [C14H11N3O6]: m/z:
316.05; found: 316.12.
4.10.39. 4-Bromo-benzoic acid (2-hydroxy-3,5-dinitro-
benzylidene)-hydrazide (ME0188)
The amide and phenol protons were exchange broadened. 1H
NMR: d 8.80 (d, J = 2.7 Hz, 1H), 8.79 (s, 1H), 8.74 (d, J = 2.7 Hz,
1H), 7.91 (d, J = 8.4 Hz, 2H), 7.78 (d, J = 8.4 Hz, 2H); LC–MS
[MꢁH+]ꢁ calcd for [C14H9BrN4O6]: m/z: 406.96; found: 406.93.
4.10.40. Adamantan-1-yl-acetic acid (2-hydroxy-3,5-dinitro-
benzylidene)-hydrazide (ME0189)
4.10.46. 3-(4-Hydroxy-phenyl)-propionic acid (5-bromo-2-
hydroxy-3-methoxy-benzylidene)-hydrazide (ME0195)
Two rotamers were obtained in a 1:0.72 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.30 (s, 1H), 7.37–7.31
(m, 1H), 7.15–7.09 (m, 1H), 7.06–6.98 (m, 2H), 6.69–6.62 (m,
Two rotamers were obtained in a 1:0.18 ratio as a result of a dy-
namic process. The phenol protons were exchange broadened. 1H
NMR (major rotamer): d 12.02 (s, 1H), 8.76 (d, J = 2.8 Hz, 1H),
8.77 (d, J = 2.8 Hz, 1H), 8.53 (s, 1H), 2.03 (s, 2H), 1.94 (br s, 3H),
1.74–1.53 (m, 12H); 1H NMR (minor rotamer): d 11.43 (s, 1H),

2702
M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2H), 3.82 (s, 3H), 2.85–2.73 (m, 2H), 2.50–2.42 (m, 2H); 1H NMR
(minor rotamer): d 8.22 (s, 1H), 7.37–7.31 (m, 1H), 7.15–7.09 (m,
1H), 7.06–6.98 (m, 2H), 6.69–6.62 (m, 2H), 3.83 (s, 3H), 2.85–
2.73 (m, 4H); LC–MS [MꢁH+]ꢁ calcd for [C17H17BrN2O4]: m/z:
391.03; found: 391.18.
1.27 (s, 9H); LC–MS [MꢁH+]ꢁ calcd for [C19H22N2O2]: m/z:
309.16; found: 309.30.
4.10.53. 2-Chloro-benzoic acid (5-bromo-2-hydroxy-
benzylidene)-hydrazide (ME0259)
Two rotamers were obtained in a 1:0.49 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.46 (s, 1H), 7.81 (d,
J = 2.6 Hz, 1H), 7.63–7.41 (m, 5H), 6.90 (d, J = 8.7 Hz, 1H); 1H
NMR (minor rotamer): d 8.25 (s, 1H), 7.63–7.41 (m, 4H), 7.36–
7.30 (m, 2H), 6.77 (d, J = 8.6 Hz, 1H); LC–MS [MꢁH+]ꢁ calcd for
[C14H10BrClN2O2]: m/z: 350.95; found: 351.09.
4.10.47. (4-Nitro-phenoxy)-acetic acid (2-hydroxy-3,5-
dimethoxy-benzylidene)-hydrazide (ME0196)
Two rotamers were obtained in a 1:0.72 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.65 (br s, 1H), 8.87 (s,
1H), 8.34 (s, 1H), 8.20 (d, J = 9.3 Hz, 2H), 7.15 (d, J = 9.3 Hz, 2H),
6.84 (d, J = 2.8 Hz, 1H), 6.63 (d, J = 2.8 Hz, 1H), 5.35 (s, 2H), 3.81
(s, 3H), 3.72 (s, 3H); 1H NMR (minor rotamer): d 11.75 (br s, 1H),
9.91 (s, 1H), 8.55 (s, 1H), 8.25 (d, J = 9.3 Hz, 2H), 7.21 (d,
J = 9.3 Hz, 2H), 6.70 (d, J = 2.7 Hz, 1H), 6.65 (d, J = 2.8 Hz, 1H),
4.87 (s, 2H), 3.79 (s, 3H), 3.72 (s, 3H); LC–MS [MꢁH+]ꢁ calcd for
[C17H17N3O7]: m/z: 374.10; found: 374.18.
4.10.54. 3-Methoxy-benzoic acid (5-bromo-2-hydroxy-
benzylidene)-hydrazide (ME0260)
1H NMR: d 12.13 (s, 1H), 11.28 (s, 1H), 8.62 (s, 1H), 7.79 (d,
J = 2.5 Hz, 1H), 7.52 (d, J = 7.8 Hz, 1H), 7.49–7.41 (m, 3H), 7.21–
7.16 (m, 1H), 6.91 (d, J = 8.8 Hz, 1H), 3.84 (s, 3H); LC–MS [MꢁH+]ꢁ
calcd for [C15H13BrN2O3]: m/z: 347.00; found: 347.06.
4.10.48. 2-[N0-(5-Bromo-2-hydroxy-3-methoxy-benzylidene)-
hydrazino]-2-oxo-acetamide (ME0197)
Two rotamers were obtained in a 1:0.04 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.47 (1H), 10.59 (br s,
1H), 8.73 (s, 1H), 8.29 (s, 1H), 7.96 (s, 1H), 7.34 (d, J = 2.1 Hz, 1H),
7.17 (d, J = 2.1 Hz, 1H), 3.83 (s, 3H); 1H NMR (minor rotamer): d
11.87 (br s, 1H), 9.85 (br s, 1H), 8.73 (s, 1H), 8.25 (s, 1H), 8.04 (s,
1H), 7.27 (d, J = 2.2 Hz, 1H), 7.13 (d, J = 2.3 Hz, 1H), 3.83 (s, 3H);
LC–MS [MꢁH+]ꢁ calcd for [C10H10BrN3O4]: m/z: 313.98; found:
314.04.
4.10.55. Phenyl-acetic acid (5-bromo-2-hydroxy-benzylidene)-
hydrazide (ME0261)
Two rotamers were obtained in a 1:0.72 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.37 (s, 1H), 7.74 (d,
J = 2.6 Hz, 1H), 7.43–7.19 (m, 6H), 6.89–6.83 (m, 1H), 3.56 (s,
2H); 1H NMR (minor rotamer): d 8.21 (s, 1H), 7.80 (d, J = 2.6 Hz,
1H), 7.43–7.19 (m, 6H), 6.89–6.83 (m, 1H), 3.96 (s, 2H); LC–MS
[MꢁH+]ꢁ calcd for [C15H13BrN2O2]: m/z: 331.01; found: 331.18.
4.10.49. 3-Methoxy-benzoic acid (2-hydroxy-3,5-dimethoxy-
benzylidene)-hydrazide (ME0198)
4.10.56. Phenyl-acetic acid (5-bromo-2-hydroxy-3-methoxy-
benzylidene)-hydrazide (ME0262)
1H NMR: d 11.99 (br s, 1H), 10.43 (br s, 1H), 8.64 (s, 1H), 7.54–
7.40 (m, 3H), 7.15 (d, J = 8.2 Hz, 1H), 6.69 (d, J = 2.6 Hz, 1H), 6.64 (d,
J = 2.6 Hz, 1H), 3.83 (s, 3H), 3.80 (s, 3H), 3.74 (s, 3H); LC–MS
[MꢁH+]ꢁ calcd for [C17H18N2O5]: m/z: 329.11; found: 329.23.
Two rotamers were obtained in a 1:0.78 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 11.87 (s, 1H), 10.76 (s,
1H), 8.38 (s, 1H), 7.34 (d, J = 2.3 Hz, 1H), 7.34–7.19 (m, 5H), 7.14 (d,
J = 2.3 Hz, 1H), 3.82 (s, 3H), 3.55 (s, 2H); 1H NMR (minor rotamer):
d 11.41 (s, 1H), 9.63 (s, 1H), 8.25 (s, 1H), 7.43 (d, J = 2.4 Hz, 1H),
7.34–7.19 (m, 5H), 7.12 (d, J = 2.3 Hz, 1H), 3.95 (s, 2H), 3.84 (s,
3H); LC–MS [MꢁH+]ꢁ calcd for [C16H15BrN2O3]: m/z: 361.02;
found: 361.06.
4.10.50. 3-Chloro-benzo[b]thiophene-2-carboxylic acid (5-
bromo-2-hydroxy-3-methoxy-benzylidene)-hydrazide
(ME0199)
Two rotamers were obtained in a 1:0.31 ratio as a result of a dy-
namic process. The amide and phenol protons were exchange
broadened. 1H NMR (major rotamer): d 8.58 (s, 1H), 8.16–8.08
(m, 1H), 8.00–7.88 (m, 1H), 7.66–7.55 (m, 2H), 7.36 (br s, 1H),
7.15–7.10 (m, 1H), 3.84 (s, 3H); 1H NMR (minor rotamer): d 8.41
(s, 1H), 8.16–8.08 (m, 1H), 8.00–7.88 (m, 1H), 7.66–7.55 (m, 2H),
7.36 (br s, 1H), 7.15–7.10 (m, 1H), 3.84 (s, 3H); LC–MS [MꢁH+]ꢁ
calcd for [C17H12BrClN2O3S]: m/z: 436.93; found: 437.00.
4.10.57. 4-Methyl-benzoic acid (5-tert-butyl-2-hydroxy-
benzylidene)-hydrazide (ME0263)
1H NMR: d 12.01 (s, 1H), 11.12 (s, 1H), 8.63 (s, 1H), 7.85 (d,
J = 8.2 Hz, 2H), 7.49 (d, J = 2.5 Hz, 1H), 7.39–7.31 (m, 3H), 6.87 (d,
J = 8.6 Hz, 1H), 2.39 (s, 3H), 1.28 (s, 9H); LC–MS [MꢁH+]ꢁ calcd
for [C19H22N2O2]: m/z: 309.16; found: 309.30.
4.10.51. Isonicotinic acid (4-chloro-2-hydroxy-benzylidene)-
hydrazide (ME0257)
4.10.58. 4-Nitro-benzoic acid (3-fluoro-2-hydroxy-
benzylidene)-hydrazide (ME0264)
Two rotamers were obtained in a 1:0.09 ratio as a result of a dy-
namic process. 1H NMR (major rotamer): d 12.27 (s, 1H), 11.37 (s,
1H), 8.84–8.76 (m, 2H), 8.67 (s, 1H), 7.88–7.79 (m, 2H), 7.67 (d,
J = 8.1 Hz, 1H), 7.04–6.95 (m, 2H); 1H NMR (minor rotamer): d
12.03 (s, 1H), 10.47 (s, 1H), 8.75–8.69 (m, 2H), 8.33 (s, 1H), 7.65–
7.59 (m, 2H), 7.43–7.37 (m, 1H), 6.92–6.84 (m, 2H); LC–MS
[MꢁH+]ꢁ calcd for [C13H10ClN3O2]: m/z: 274.04; found: 274.18.
Two rotamers were obtained in a 1:0.09 ratio as a result of a dy-
namic process. For the minor rotamer, the phenol and amide pro-
tons were exchange broadened. 1H NMR (major rotamer): d 12.48
(s, 1H), 11.44 (s, 1H), 8.70 (s, 1H), 8.43–8.37 (m, 2H), 8.22–8.15 (m,
2H), 7.44 (d, J = 7.8 Hz, 1H), 7.34–7.26 (m, 1H), 6.97–6.89 (m, 1H);
1H NMR (minor rotamer): d 8.43–8.30 (m, 3H), 8.01–7.94 (m, 2H),
7.26–7.16 (m, 2H), 6.82 (br s, 1H); LC–MS [MꢁH+]ꢁ calcd for
[C14H10FN3O4]: m/z: 302.06; found: 302.15.
4.10.52. Phenyl-acetic acid (5-tert-butyl-2-hydroxy-
benzylidene)-hydrazide (ME0258)
Acknowledgements
Two rotamers were obtained in a 1:0.62 ratio as a result of a dy-
namic process. The phenol and amide protons were exchange
broadened. 1H NMR (major rotamer): d 8.41 (s, 1H), 7.48 (d,
J = 2.5 Hz, 1H), 7.37–7.19 (m, 6H), 6.86–6.79 (m, 1H), 3.56 (s, 2H),
1.25 (s, 9H); 1H NMR (minor rotamer): d 8.25 (s, 1H), 7.66 (d,
J = 2.5 Hz, 1H), 7.37–7.19 (m, 6H), 6.86–6.79 (m, 1H), 3.93 (s, 2H),
We thank the Swedish National Research Council, the Swedish
Governmental Agency for Innovation Systems (VINNOVA), the Carl
Trygger foundation, and Creative Antibiotics for support. The work
was carried out with affiliation to Umeå Centre for Microbial
Research.

M. K. Dahlgren et al. / Bioorg. Med. Chem. 18 (2010) 2686–2703
2703
28. Muschiol, S.; Bailey, L.; Gylfe, A.; Sundin, C.; Hultenby, K.; Bergstrom, S.;
Elofsson, M.; Wolf-Watz, H.; Normark, S.; Henriques-Normark, B. A. Proc. Natl.
Acad. Sci. U.S.A. 2006, 103, 14566.
29. Bailey, L.; Gylfe, A.; Sundin, C.; Muschiol, S.; Elofsson, M.; Nordstrom, P.;
Henriques-Normark, B.; Lugert, R.; Waldenstrom, A.; Wolf-Watz, H.;
Bergstrom, S. FEBS Lett. 2007, 581, 587.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2010.02.022.
30. Slepenkin, A.; Enquist, P. A.; Hagglund, U.; de la Maza, L. M.; Elofsson, M.;
Peterson, E. M. Infect. Immun. 2007, 75, 3478.
References and notes
31. Muschiol, S.; Normark, S.; Henriques-Normark, B.; Subtil, A. BMC Microbiol.
2009, 9, 75.
1. Marra, A. Drugs R. D. 2006, 7.
32. Prantner, D.; Nagarajan, U. M. Infect. Immun. 2009, 77, 76.
33. Hansch, C. A. Acc. Chem. Res. 1969, 2, 232.
2. Clatworthy, A. E.; Pierson, E.; Hung, D. T. Nat. Chem. Biol. 2007, 3, 541.
3. Escaich, S. Curr. Opin. Chem. Biol. 2008, 12, 400.
34. Wold, S.; Dunn, W. J., III J. Chem. Inf. Model. 1983, 23, 6.
35. Eriksson, L.; Johansson, E. Chemom. Intell. Lab. Syst. 1996, 34, 1.
36. Martin, E. J.; Blaney, J. M.; Siani, M. A.; Spellmeyer, D. C.; Wong, A. K.; Moos, W.
H. J. Med. Chem. 1995, 1431, 38.
4. Hueck, C. J. Microbiol. Mol. Biol. Rev. 1998, 62, 379.
5. Keyser, P.; Elofsson, M.; Rosell, S.; Wolf-Watz, H. J. Int. Med. 2008, 264, 17.
6. Kauppi, A. M.; Nordfelth, R.; Uvell, H.; Wolf-Watz, H.; Elofsson, M. Chem. Biol.
2003, 10, 241.
37. Linusson, A.; Wold, S.; Norden, B. Mol. Divers. 1999, 4, 103.
38. Holm, L.; Frech, K.; Dzhambazov, B.; Holmdahl, R.; Kihlberg, J.; Linusson, A. J.
Med. Chem. 2007, 50, 2049.
39. Larsson, A.; Johansson, S.; Pinkner, J. S.; Hultgren, S. J.; Almqvist, F.; Kihlberg, J.;
Linusson, A. J. Med. Chem. 2005, 48, 935.
40. Linusson, A.; Gottfries, J.; Olsson, T.; Örnskov, E.; Folestad, S.; Norden, B.; Wold,
S. J. Med. Chem. 2001, 44, 3424.
41. Nurbo, J.; Roos, A. K.; Muthas, D.; Wahlstrom, E.; Ericsson, D. J.; Lundstedt, T.;
Unge, T.; Karlen, A. J. Pept. Sci. 2007, 13, 822.
7. Gauthier, A.; Robertson, M. L.; Lowden, M.; Ibarra, J. A.; Puente, J. L.; Finlay, B. B.
Antimicrob. Agents Chemother. 2005, 49, 4101.
8. Linington, R. G.; Robertson, M.; Gauthier, A.; Finlay, B. B.; MacMillan, J. B.;
Molinski, T. F.; van Soest, R.; Andersen, R. J. J. Nat. Prod. 2006, 69, 173.
9. Felise, H. B.; Nguyen, H. V.; Pfuetzner, R. A.; Barry, K. C.; Jackson, S. R.; Blanc, M.
P.; Bronstein, P. A.; Kline, T.; Miller, S. I. Cell Host Microbe 2008, 4, 325.
10. Iwatsuki, M.; Uchida, R.; Yoshijima, H.; Ui, H.; Shiomi, K.; Matsumoto, A.;
Takahashi, Y.; Abe, A.; Tomoda, H.; Omura, S. J. Antibiot. 2008, 61, 222.
11. Pan, N. J.; Brady, M. J.; Leong, J. M.; Goguen, J. D. Antimicrob. Agents Chemother.
2009, 53, 385.
42. Davies, M. J. Am. Stat. Assoc. 1961, 56, 729.
43. Wold, S. R. A.; Wold, H.; Dunn, W. I. SIAM J. Sci. Stat. Comp. 1984, 5, 735.
44. Wold, S.; Sjostrom, M.; Eriksson, L. Chemom. Intell. Lab. Syst. 2001, 58.
45. Cortes, C.; Vapnik, V. Mach. Learn. 1995, 20, 273.
46. Matthews, E. J.; Kruhlak, N. L.; Benz, R. D.; Contrera, J. F.; Marchant, C. A.; Yang,
C. Toxicol. Mech. Methods 2008, 18, 189.
12. Kauppi, A. M.; Nordfelth, R.; Hagglund, U.; Wolf-Watz, H.; Elofsson, M. Adv. Exp.
Med. Biol. 2003, 529, 97.
13. Nordfelth, R.; Kauppi, A. M.; Norberg, H. A.; Wolf-Watz, H.; Elofsson, M. Infect.
Immun. 2005, 73, 3104.
14. Sun, J. S.; Han, X. W.; Yu, B. Synlett 2005, 437.
47. Zhu, H.; Tropsha, A.; Fourches, D.; Varnek, A.; Papa, E.; Gramatica, P.; Oberg, T.;
Dao, P.; Cherkasov, A.; Tetko, I. V. J. Chem. Inf. Model. 2008, 48, 766.
48. Muchmore, S. W.; Debe, D. A.; Metz, J. T.; Brown, S. P.; Martin, Y. C.; Hajduk, P. J.
J. Chem. Inf. Model. 2008, 48, 941.
15. Tsuchiya, S.; Sunazuka, T.; Hirose, T.; Mori, R.; Tanaka, T.; Iwatsuki, M.;
ÅŒmura, S. Org. Lett. 2006, 5577, 8.
16. Kauppi, A. M.; Andersson, C. D.; Norberg, H. A.; Sundin, C.; Linusson, A.;
Elofsson, M. Bioorg. Med. Chem. 2007, 15, 6994.
49. Olsson, I. M.; Gottfries, J.; Wold, S. Chemom. Intell. Lab. Syst. 2004, 73, 37.
50. Tauschek, M.; Gorrell, R. J.; Strugnell, R. A.; Robins-Browne, R. M. Proc. Natl.
Acad. Sci. U.S.A. 2002, 99, 7066.
17. Dahlgren, M. K.; Kauppi, A. M.; Olsson, I.-M.; Linusson, A.; Elofsson, M. J. Med.
Chem. 2007, 50, 6177.
18. Kline, T.; Felise, H. B.; Barry, K. C.; Jackson, S. R.; Nguyen, H. V.; Miller, S. I. J.
51. Kaper, J. B.; Nataro, J. P.; Mobley, H. L. T. Nat. Rev. Microbiol. 2004, 2, 123.
52. Cornelis, G. R. Nat. Rev. Microbiol. 2006, 4, 811.
53. Koh, L. L.; Kon, O. L.; Loh, K. W.; Long, Y. C.; Ranford, J. D.; Tan, A. L. C.; Tjan, Y. Y.
J. Inorg. Biochem. 1998, 72, 155.
54. Gudasi, K.; Patil, M.; Vadavi, R.; Shenoy, R.; Patil, S.; Nethaji, M. Transition Met.
Chem. 2006, 31, 580.
55. Zu, D.; Liu, R. C. H. Med. Res. Rev. 2002, 22, 26.
Med. Chem. 2008, 51, 7065.
19. Kline, T.; Barry, K. C.; Jackson, S. R.; Felise, H. B.; Nguyen, H. V.; Miller, S. I.
Bioorg. Med. Chem. Lett. 2009, 19, 1340.
20. Yang, S.; Peng, Q.; San Francisco, M.; Wang, Y.; Zeng, Q.; Yang, C.-H. PLoS ONE
2008, 3, e2973.
21. Li, Y.; Peng, Q.; Selimi, D.; Wang, Q.; Charkowski, A. O.; Chen, X.; Yang, C. H.
Appl. Environ. Microb. 2009, 75, 1223.
56. MOE. Molecular Operating Environment (MOE), Version 2005.06, Chemical
Computing Group, Inc.: Montreal, Quebec, Canada, 2005.
57. OMEGA. OpenEye Scientific Software, Inc., 2.3.2; 9 Bisbee, Suite D, Santa FE, New
Mexico, United States of America.
22. Puiac, S.; Negrea, A.; Richter-Dahlfors, A.; Plant, L.; Rhen, M. Antimicrob. Agents.
Chemother. 2009, 53, 2402.
23. Wolf, K.; Betts, H. J.; Chellas-Gery, B.; Hower, S.; Linton, C. N.; Fields, K. A. Mol.
Microbiol. 2006, 61, 1543.
58. Jaguar. Version 6.5, Schrödinger, LLC, New York, NY, 2006.
59. Umetrics. SIMCA-P+ 12.0; Box 7960, S-907 19 Umeå, Sweden.
60. MODDE. Umetrics, 8; Box 7960, S-907 19 Umeå, Sweden.
61. Lindgren, F.; Hansen, B.; Karcher, W.; Sjostrom, M.; Eriksson, L. J. Chemom.
1996, 10, 521.
24. Negrea, A.; Bjur, E.; Ygberg, S. E.; Elofsson, M.; Wolf-Watz, H.; Rhen, M.
Antimicrob. Agents Chemother. 2007, 51, 2867.
25. Hudson, D. L.; Layton, A. N.; Field, T. R.; Bowen, A. J.; Wolf-Watz, H.; Elofsson,
M.; Stevens, M. P.; Galyov, E. E. Antimicrob. Agents. Chemother. 2007, 51, 2631.
26. Veenendaal, A. K. J.; Sundin, C.; Blocker, A. J. J. Bacteriol. 2009, 191,
563.
62. Wold, S. Pattern Recognit. 1976, 8, 127.
63. O’Brien, J.; Wilson, I.; Orton, T.; Pognan, F. Eur. J. Biochem. 2000, 267, 5421.
64. Lindhagen-Persson, M.; Vestling, M.; Reixach, N.; Olofsson, A. Amyloid 2008, 15,
240.
27. Tree, J. J.; Wang, D.; McInally, C.; Mahajan, A.; Layton, A.; Houghton, I.;
Elofsson, M.; Stevens, M. P.; Gally, D. L.; Roe, A. J. Infect. Immun. 2009,
77, 4209.