Antimycobacterial Pyrroles: Synthesis, Anti-Mycobacterium tuberculosis Activity and QSAR Studies

Article abstract of DOI:10.1016/S0968-0896(00)00061-4

A number of known antifungal pyrrole derivatives and some newly synthesized compounds (5-33) were tested in vitro against Mycobacterium tuberculosis CIP 103471. The majority of tested compounds were efficient antimycobacterial agents showing MIC values ranging from 0.5 to 32 μg/mL. A 3-D-QSAR study has been performed on these pyrrole derivatives to correlate their chemical structures with their observed inhibiting activity against M. tuberculosis. Due to the absence of information on a putative receptor responsible for this activity, classical quantitative structure-activity relationships (QSAR) and comparative molecular field analysis (CoMFA) have been applied. A model able to well correlate the antimycobacterial activity with the chemical structures of pyrrole derivatives 5-33 has been developed which is potentially helpful in the design of novel and more potent antituberculosis agents. The combination of CoMFA with classical QSAR descriptors led to a better hybrid 3-D-QSAR model, that successfully explains the structure-activity relationships (r²=0.86) of the training set. A comparison between the QSAR, CoMFA and mixed QSAR-CoMFA models is also presented. The hybrid model is to be preffered, however, because of its lowest values of the average absolute error of prediction toward a limited external test set.

Full text of DOI:10.1016/S0968-0896(00)00061-4

Bioorganic & Medicinal Chemistry 8 (2000) 1423±1432
Antimycobacterial Pyrroles: Synthesis, Anti-Mycobacterium
tuberculosis Activity and QSAR Studies
Rino Ragno, a Garland R. Marshall, a,* Roberto Di Santo, b Roberta Costi, b
Silvio Massa, c Ra€aello Rompei d and Marino Artico b,*
aCenter for Molecular Design, Washington University, St. Louis, MO 63110, USA
b
Á
Dipartimento di Studi Farmaceutici, Universita degli Studi di Roma ``La Sapienza'', P. le A. Moro 5, 00185 Rome, Italy
c
Á
Dipartimento Farmaco Chimico Tecnologico, Universita degli Studi di Siena, Banchi di Sotto 55, 53100 Siena, Italy
Á Á
Cattedra di Microbiologia Applicata, Facolta di Scienze Matematiche, Fisiche e Naturali, Universita di Cagliari, via Porcel 4,
d
09124 Cagliari, Italy
Received 10 November 1999; accepted 17 February 2000
AbstractÐA number of known antifungal pyrrole derivatives and some newly synthesized compounds (5±33) were tested in vitro
against Mycobacterium tuberculosis CIP 103471. The majority of tested compounds were ecient antimycobacterial agents showing
MIC values ranging from 0.5 to 32 mg/mL. A 3-D-QSAR study has been performed on these pyrrole derivatives to correlate their
chemical structures with their observed inhibiting activity against M. tuberculosis. Due to the absence of information on a putative
receptor responsible for this activity, classical quantitative structure±activity relationships (QSAR) and comparative molecular ®eld
analysis (CoMFA) have been applied. A model able to well correlate the antimycobacterial activity with the chemical structures of
pyrrole derivatives 5±33 has been developed which is potentially helpful in the design of novel and more potent antituberculosis
agents. The combination of CoMFA with classical QSAR descriptors led to a better hybrid 3-D-QSAR model, that successfully
explains the structure±activity relationships (r2=0.86) of the training set. A comparison between the QSAR, CoMFA and mixed
QSAR±CoMFA models is also presented. The hybrid model is to be preferred, however, because of its lowest values of the average
absolute error of prediction toward a limited external test set. # 2000 Elsevier Science Ltd. All rights reserved.
Introduction
of new potent and selective antituberculosis agents, the
current therapeutic outlook is poor. Only a few deriva-
tives were found endowed with some antimycobacterial
The search for novel antibacterial agents active against
Mycobacterium tuberculosis and other atypical myco-
bacteria is urgent due to the lack of e€ectiveness of
known antituberculosis agents against opportunistic
pathogens as a consequence of rapidly emerging resis-
tance. Furthermore, immunocompromised patients
observed with AIDS, or after transplantation, are easily
infected by pathogenic fungi, protozoa and myco-
bacteria, leading rapidly to death. In particular, myco-
bacteria have recently increased their virulence and
about 30 million of people are predicted by the World
Health Organization (WHO) to catch tuberculosis in the
near future.1 5 Despite the e€orts of the pharmaceutical
companies engaged in the design, synthesis and assays
activity, including Upjohn oxazolidinone
1
(U-
100480),6,7 the nitroimidazo[2,1-b]oxazole 2,8,9 some
broad-spectrum esters of pyrazinoic acid 310 and 5-
chloropyrazinamide 411 (Chart 1). While the mechan-
isms of action of clinical antituberculosis agents have
been intensively investigated, the details of their inter-
action with their target enzymes are not known.12,13
Recently, pyrrole derivatives have emerged as che-
motherapeutic agents potentially useful for inhibiting
the activities of M. tuberculosis and other atypical
mycobacteria, including M. avium complex, an oppor-
tunistic pathogen that greatly contributes to the death
of AIDS patients. Porretta and co-workers14 have
reported the antimycobacterial activity of a pyrrole
derivative, and Di Santo et al.15 more recently ascribed
appreciable inhibiting action to pyrrolnitrin and some
related nitropyrroles 5±12 and 19±23.
*Corresponding authors. Garland R. Marshall: Tel.: +1-314-362-
1567; fax: +1-314-362-0234; garland@gpc.wust.l.edu; Marino Artico:
Tel./fax: +39-6-446-2731; artico@uniroma1.it.
0968-0896/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0968-0896(00)00061-4

1424
R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
Keeping in mind the above results, we tested against M.
tuberculosis a number of pyrrole derivatives (5±12, 19±24
and 31±33)15 18 and some newly synthesized compounds
(13±18 and 25±30) related to the above pyrroles (Table 1),
in which the phenyl moiety was replaced by a pyridine
ring typical of isoniazid (INH), which is nowadays the
most potent antituberculosis agent in clinical practice.
With the aim of establishing the role of the pyrrole ring
as a pharmacophoric group and, hence, its in¯uence on
the antimycobacterial activity, we have undertaken a
3-D-QSAR study on this set of pyrrole derivatives.
Since no information regarding their putative receptor is
available, classical quantitative structure±activity rela-
tionships (QSAR) and comparative molecular ®eld
analysis (CoMFA)19 21 were used to correlate the anti-
mycobacterial activity of compounds 5±33 (Table 1)
showing some level of inhibitory potency (Tables 2 and
3) against M. tuberculosis.
Chart 1.
Table 1. Pyrrole derivatives used for training set and test set
Compd
X
Y
R
R1
R2
R3
MICa (mg/mL)
Reference
5
6
7b
8
COOEt
NO2
NO2
NO2
NO2
H
H
H
H
H
H
Cl
Cl
H
H
Cl
Cl
H
H
H
H
H
C2H5
C2H5
H
H
H
H
Cl
Cl
H
C2H5
H
H
H
H
Cl
Cl
Cl
H
Cl
Cl
H
Cl
H
H
H
H
Cl
H
16
1
16
8
8
32
15,16
15,17
15,17
15,17
15,17
17
15,17
15,16
Ð
Ð
Ð
Ð
Ð
9
H
10
11b
12
13
14
15
16
17
18
19
20
21
22b
23
24b
25
26
27
28
29b
30
31
32
33b
NO2
NO2
H
OH
OCH3
OCH3
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
CH3
CH3
H
H
H
H
H
>250
16
16
>250
16
16
250
>250
16
8
16
125
8
250
0.50
4
16
4
COOEt
COOEt
CONHNH2
COOEt
COOEt
COOEt
CONHNH2
COOH
COOEt
COOEt
COOEt
CH2OH
CH(OH)C6H5
COOEt
COOEt
CONHNH2
COOEt
COOEt
CH2OH
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Cl
H
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
H
CH3
C2H5
H
H
Cl
OCH3
OCH3
H
Ð
15,16
15,16
15,16
15,16
15
18
Ð
Ð
Ð
Ð
Ð
Ð
18
18
18
F
F
H
H
H
Cl
Cl
H
Cl
CH3
NO2
CH3
Ð
Ð
Ð
CH2
CH2
CH2
SO2
SO2
SO2
Ð
Ð
Ð
Ð
Ð
Ð
Ð
H
H
H
F
OCH3
>250
4
16
Ð
Ð
Ð
Ð
1
1
aMinimum inhibiting concentration of test derivatives against Mycobacterium tuberculosis CIP 103471. Reference drugs: INH (MIC=0.2 mg/mL)
and SM (MIC=0.7 mg/mL). The assays were done in triplicate. The degree of variation was found always within 50%.
bUsed for test set.

R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
Table 2. Training set: QSAR parameters and recalculated pMICs
1425
Compd
Exp pMIC
HlogP
DipMom
LUMO
Vol
G_CDS
Fitted pMIC
Mixed
COMFA
QSAR
5
6
8
9
4.19
5.41
4.51
4.51
3.87
4.21
4.13
2.61
4.16
4.18
2.99
2.66
4.34
4.62
4.34
4.62
5.83
4.92
4.31
4.97
4.91
4.46
5.68
4.49
4.10
4.10
4.10
2.84
3.77
2.45
0.06
2.39
2.93
3.53
1.26
5.44
5.32
5.34
4.83
4.53
5.02
2.26
4.27
2.43
5.60
5.55
5.50
7.50
4.79
5.76
5.76
4.89
6.14
6.29
6.60
6.95
6.06
6.27
4.44
5.06
4.53
5.20
7.90
6.76
6.74
3.09
5.29
6.03
6.54
0.17
0.65
0.65
0.67
0.67
0.52
0.18
0.25
0.22
0.24
0.39
0.39
0.49
0.06
0.37
0.35
0.53
0.09
0.42
1.01
1.58
0.66
0.71
217.41
191.88
192.69
192.42
186.43
245.67
197.49
179.61
215.03
231.40
232.08
213.33
285.72
318.64
322.29
284.59
301.98
321.83
283.50
320.64
284.11
401.27
404.80
2.66
4.72
5.01
5.11
8.86
0.32
4.08
9.72
1.39
1.01
3.00
8.72
3.09
0.58
0.82
2.91
1.44
0.14
6.68
2.93
6.05
0.99
1.73
4.21
5.37
4.53
4.57
3.86
4.17
4.12
2.51
4.28
4.16
3.05
2.56
4.26
4.58
4.46
4.53
5.49
4.95
4.67
4.41
5.00
5.00
5.13
4.20
5.44
4.51
4.81
4.12
3.72
3.82
2.59
4.16
4.26
3.76
2.72
4.44
4.49
4.52
4.48
5.48
4.90
4.00
4.13
4.74
5.15
5.40
4.16
5.37
4.49
4.80
4.14
3.60
3.78
2.68
4.07
4.28
3.82
2.82
4.46
4.46
4.61
4.43
5.47
4.87
3.98
4.15
4.76
5.20
5.45
10
12
13
14
15
16
17
18
19
20
21
23
25
26
27
28
30
31
32
Table 3. Test set: actual and residual pMICs
CoMFA model
Mixed model
QSAR model
Compd
Experimental
Pred.
Residual
Pred.
Residual
Pred.
Residual
7
4.21
2.73
3.41
3.21
2.90
5.69
4.43
4.14
5.19
4.24
4.31
4.85
0.22
1.41
1.78
1.03
1.41
0.84
5.06
3.85
4.41
3.69
4.31
4.57
0.85
1.12
1.00
0.48
1.41
1.12
4.01
3.80
4.42
4.01
4.44
4.62
1.21
1.07
1.01
0.80
1.54
1.07
11
22
24
29
33
Chemistry
Molecular Modeling and QSAR Methods
Derivatives used for this study have been prepared
either according to procedures described in previous
works15 18 or by the pathway depicted in Scheme 1.
All the computational work was performed on Silicon
Graphics computers (Indigo2 R10000 195 MHz and
Octane 225 MHz). The measurements of biological
activity used to develop the QSAR and the CoMFA
were expressed as (pMIC)= log(MIC), where MIC is
the minimal inhibiting concentration expressed as molar
concentration (Tables 2 and 3). To the four derivatives
which have been found to be `inactive' (MIC >250 mg/
mL) an activity value equal to 50% of that of the com-
pound reported as the less active was arbitrarily
assigned. Although such data might contribute to a
disturbing source of noise, it is, nevertheless, true that
the exclusion of a non-negligible number of observa-
tions would imply serious loss of information.24,25
Condensation of pyridine-4-carboxaldehyde with tri-
ethylphosphonoacetate in the presence of anhydrous
K2CO3 furnished 3-(4-pyridinyl)-2-propenoic acid ethyl
ester,22 that was annulated with the proper toluene-4-
sulfonylalkyl isocyanide23 in the presence of sodium
hydride to give pyrroles 13 and 17. Potassium salt of 13
reacted with arylsulfonyl chlorides in the presence of 18-
crown-6 as a catalyst to obtain sulfones 28 and 29 and
lithium aluminum hydride reduction of the latter com-
pound furnished the related carbinol 30.
Alkylation of pyrroles 13 and 17 with formation of
derivatives 15, 16, 25 and 26 was performed by phase
transfer catalysis using tetrabutylammonium hydrogen
sulfate and the proper alkyl halide in a mixture of
dichloromethane and concentrated aqueous NaOH.
Finally, treatment of esters 13, 17 and 25 with hydrazine
hydrate in boiling ethanol gave carboxyhydrazides 14,
18 and 27, respectively.
For systems in which no information exists about the
binding site, it is a well established assumption to con-
sider that a similar class of molecules (congeneric series)
binds to the putative receptor site adopting a similar
geometry and orientation. Since structural overlap is the
simplest hypothesis, we decided to test its self-con-
sistency before deriving a more complicated model. For
this reason, no particular e€ort was made to explore the

1426
R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
Scheme 1.
conformational space of the tested molecules to seek
alternative binding modes. A goal of future work will be
to determine possible binding geometries that de®ne a
common pharmacophoric pattern and generate a 3-D
receptor model.
energetic cuto€ of 30 kcal/mol with no electrostatic
interactions at sterically bad contacts. For both the
CoMFA ®elds and the ®eld-®t alignment, the atomic
point charges were obtained from a semiempirical
calculation using the AM1 hamiltonian implemented in
MOPAC93. Regression analyses were done using the
SYBYL implementation of the PLS algorithm, initially
with leave-one-out (LOO) cross-validation to reduce the
possibility of obtaining chance correlations and eight
principal components (PCs).29
All the molecules were built using the SKETCH module
implemented in the program SYBYL26 starting from the
lowest energy conformation of the searched template
molecule (32). Before any alignment trial, the geometry
of each molecule was optimized by a simple minimiza-
tion to the nearest local minimum using the MMFF9427
force ®eld implemented in SYBYL.
The optimal number of PCs was then chosen on the
basis of the highest cross-validated q2 value, the smallest
standard error of prediction (SEP), and the minimum
number of components. To improve the signal-to-noise
ratio, the minimum sigma value was set to 2.0 kcal/mol.
The steric and electrostatic ®eld columns were weighted
according to the CoMFA-STD default scaling option,
where a ®eld is considered as a whole and every
CoMFA variable is a€ected by the overall ®eld mean
and standard deviation. Final PLS with no cross-vali-
dation was then carried out using the optimal number
of PCs.
CoMFA calculations were performed with the QSAR
module of SYBYL and the following parameters. The
grid in which the aligned molecules were embedded was
regularly spaced (2 A) with dimensions of 22Â22Â16 A,
the same region used to align the molecules by the ®eld-
®tting techniques. A further re®nement of the region
was done with the PLS region focusing28 option in the
QSAR-SYBYL module. Region focusing is an iterative
procedure that re®nes a model by increasing the weight
for those lattice points that are most pertinent to the
model. Steric and electrostatic interaction energies were
calculated using a carbon sp3 probe with a +1 charge,
a distance-dependent dielectric constant (1/r), and an
Along with the standard CoMFA ®elds, the in¯uences
of some classical QSAR parameters were also taken into
account. Several parameters were computed for each

R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
1427
molecule and added to the CoMFA analysis spread-
sheet: (i) the program HINT30 version 2.25 was used to
estimate the logarithm of the partition coecient
(HlogP);31 (ii) from MOPAC93 (1SCF calculation,
AM1 hamiltonian) the HOMO and LUMO energies;32
(iii) the ionization potential and the dipole moment; (iv)
from the AMSOL (1SCF calculation, AM1 hamilto-
nian) program the molecular volume (Vol) along with
the cavity dispersion energies were obtained (G-CDS,
G-P and G-P-CDS).32
CoMFA model derived. The cross-validation analysis
using only the CoMFA ®elds led to a model with an
optimal number of principal components (PCs) of 4, a
q2 value of 0.44 and a standard error of prediction
(SEP) of 0.78 log units. The corresponding non-cross-
validated analysis (4 PCs) resulted in a ®tted r2 value of
0.933 (Fig. 1), with a standard error of estimation (SEE)
of 0.229 (F-test value=62.729) (CoMFA MODEL).
The contributions of the steric and electrostatic ®elds
were, respectively, 50.5 and 49.5%. The model did not
show very good predictive statistical results (q2 only
0.44) but was still valuable, since it was shown that
CoMFA models showing a q2 higher than 0.3 have a
low probability of chance correlation.29 The low value
of q2 may be due to some de®ciencies in the molecular
alignment, but several attempts to improve the CoMFA
model varying the alignment did not lead to better
results; for instance, the more obvious atom by atom
alignment based on the common skeleton atoms led to a
CoMFA model with almost the same statistical char-
acteristics, but slightly worse than that obtained by the
®eld-®t alignment.
Results and Discussion
Antimycobacterial activity
Pyrrole derivatives 5±33 were assayed in vitro against
M. tuberculosis CIP 103471 and their activities are
reported in Table 1 together with those of INH and
streptomycin (SM) used as reference drugs. Most test
derivatives inhibited the growth of M. tuberculosis at
MIC values ranging from 0.5 to 250 mg/mL. In parti-
cular, compound 25 showed the highest potency (0.5 mg/
mL) comparable to reference drugs, while four deriva-
tives, namely 11, 14, 18 and 28, were devoid of any
antituberculosis activity. Imidazoles 32±33 and nitro-
pyrrole 6 showed MIC=1 mg/mL, similar to that of
streptomycin, while compounds 26, 28 and 30 were 4
times less potent. The remaining derivatives showed
interesting MIC values (8±32 mg/mL) with the exception
of the slightly active pyrroles 17, 22 and 24 (MIC=125±
250 mg/mL).
The combination of the previous CoMFA with classical
QSAR descriptors led to a better 3-D-QSAR model.
After several trials, the best relationship was obtained
by including the HINT calculated logP (HlogP), the
dipole moment (DipMom), the LUMO energy, the
molecular volume (Vol) and the cavity-dispersion-
solvent structure free energy (G_CDS). The optimal
number of PCs for this hybrid model was found to be 3,
for which a q2 value of 0.73 was obtained with a SEP of
0.45 pMIC units. With 3 PCs, a r2 value of 0.86 was
obtained (HYBRID MODEL), with a SEE of 0.32 units
(Fig. 2).
Molecular modeling and QSAR
The calculated QSAR parameters were correlated with
the in vitro measured inhibitory activities of the synthe-
sized compounds. The molecular data set was split in a
training set of 23 molecules and a test set of 6 molecules
(Tables 2 and 3). The splitting was based on two prin-
In contrast with classical QSARs, it is not possible to
write a regression equation for CoMFA 3-D-QSAR
models because of the large number of variables. On the
other hand, the CoMFA model has the advantage that
the 3-D correlated results can be graphically displayed.
cipal concepts: (1) an activity range (ÁpMICTrainingSet
=
2.61±5.83; ÁpMICTestSet=2.73±5.69) of the two sets as
close as possible and (2) the maximum chemical diver-
sity of the test set.
In developing a CoMFA model, the alignment of the
training set is the most important step. Because no
structural information existed on the nature of the
receptor and, hence, of the binding site, the ®eld-®t
procedure was used to align the training set. Compound
32 was used as template molecule on the basis of being
both one of the more sterically constrained as well as
one of the more active. The random search module
implemented in SYBYL was used to select the con-
formation of the template molecule. The MMFF94
force ®eld with the default settings was used. The lowest
energy conformation obtained was then used to build all
the other molecules of the training and test sets. A ®eld-
®t alignment with subsequent minimization was carried
out twice to obtain the ®nal alignment used in the ana-
lysis. As the compounds belong to a congeneric series,
the particular conformation chosen is not expected to
have a signi®cant e€ect on the statistical quality of the
Figure 1. CoMFA model: training set ®tting and test set predictions.

1428
R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
The interpretation of these maps is highly subjective; the
absence of lattice points does not mean that a given
pharmacophoric element has no in¯uence on the biolo-
gical activity, but only that almost all the compounds
included in the training set exert similar steric and/or
electrostatic in¯uence in that particular area, or, alter-
natively, that the lack of polyhedra in some regions
delineates a less explored space.
As can be seen in Figure 3, large green polyhedra run
around the aromatic moiety directly attached to the
pyrrole indicating a good correlation of the activity with
the presence of bulky substituents on the 2-aryl portion.
Yellow polyhedra situated around the space where the
ethyl group of the carboxyethyl substituent is present in
almost all the molecules indicate a decreasing activity
induced with the presence of some bulky group in that
position. However, due to the ¯exible nature of the ethyl
moiety and because the molecules after the ®eld ®tting
were minimized it is not possible to make any assump-
tions regarding this particular portion of the space. A
rather important blue spot is visible around the region
where the lone pair of the carboxyl oxygen is located.
Since the presence of a lone pair donating atom is
almost always present, we can assume that for some
level of activity a hydrogen bonding donor in that
region is required. Finally two large red polyhedra are
located where the sulfonyl moiety exists for some com-
pounds included in the training set. In this case we
assume that the conformations used represent the bind-
ing geometry, and the red polyhedra represent a region
of the putative binding site where a hydrogen bond
could occur.
Figure 2. Hybrid model: training set ®tting and test set predictions.
The CoMFA steric and electrostatic ®eld contour maps
give information about those regions where a modi®ca-
tion of the chemical structure can enhance, or decrease,
the biological activity. The surfaces are the representa-
tion of those lattice points, where di€erences in ®eld
values are correlated with di€erences in activity. In
Figure 3 are shown the maps for the CoMFA and the
mixed models, that are coincident, where the red con-
tours represent regions of decreased tolerance for posi-
tive charge (20% contribution), the blue contours
represent regions of decreased tolerance for negative
charge (80% contribution), the yellow contours repre-
sent regions of low steric tolerance (20% contribution)
and the green contours represent regions of high steric
tolerance (80% contribution).
Because of the high contribution to the ®nal QSAR
analysis, a PLS analysis was also carried out using only
the QSAR descriptors, giving a q2 of 0.70 and a stan-
dard error of prediction (SEP) value of 0.46 pMIC units
Figure 3. CoMFA electrostatic and steric STDEV*COEFF contours plot from the analysis CoMFA and Mixed. Negative charge not favorite areas
(contribution level 80%) are represented by the red polyhedra. Negative charge-favored areas (contribution level 20%) are represented by blue
polyhedra. Sterically favored areas (contribution level 80%) are represented by green polyhedra. Sterically unfavored areas (contribution level 20%)
are represented by yellow polyhedra. Along with the contours are displayed also two molecules, one active in cyan 32 and one less active in white 17.

R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
1429
using 3 PCs (Fig. 4). The non-cross-validated PLS
resulted in a conventional correlation coecient (r2)
value of 0.83 (QSAR MODEL) and a standard error of
estimation (SEE) of 0.36 log unit (F-test value=30.0).
Here is the QSAR regression equation obtained using
only the traditional QSAR parameters:
error of prediction (1.00) and SDEP (1.04). All the
models show moderate overprediction as can be seen
from Figures 1, 2 and 4, where the experimental biolo-
gical activities are plotted versus the predicted activities.
In almost all the cases, the models tend to predict a
greater value for this limited test set.
pMIC ˆ2:437 ‡ ꢀ0:245†ÃHlogP ‡ ꢀ0:313†
ÃDipMom ꢀ0:969†ÃLUMO ꢀ0:003†
ÃVol ‡ ꢀ0:097†ÃG CDS
Conclusions
We have described the synthesis, antimicrobial activity
and QSAR studies of pyrrole derivatives endowed with
anti-M. tuberculosis activity. A number of these com-
pounds showed good potency in in vitro assays and
pyrrole 25 resulted as active as SM and two and half
times less potent than INH used as reference drugs. This
compound will be considered as lead for further inves-
tigations on antimycobacterial pyrrole derivatives.
From the regression equation, it is possible to observe
that there is a strong correlation between the parameters
and the biological activity. Since all parameters used are
global descriptors of the molecule as a whole, it is not
possible to relate speci®c molecular structural di€er-
ences to the contribution of each variable and little
molecular design insight is obtained.
A variation of the CoMFA method with the inclusion
of some classical QSAR descriptors has led to the de®-
nition of a hybrid QSAR±CoMFA model that success-
fully explains the structure±activity relationships
(r2=0.86) of the training set. Considering the limited
range of pMIC (ÁpMIC=3) we did not expect to
obtain an excellent predictive QSAR model but only to
®nd out some relationship to be used in the design of
more potent compounds. When we applied an external
test set the HYBRID MODEL was capable of dis-
criminating between the more and less active compounds
showing reasonable error of prediction (SDEP=1.04).
Such a value of SDEP could also be obtained if one
used the average of the training set as a predictor but in
this case no trend is obtained. Moreover from the asso-
ciated graphical CoMFA contour maps it is possible to
derive some suggestions to modify our lead compound
and design new potential antimycobacterial agents. In
particular the large green polyhedra around the aro-
matic moiety in position 4 of pyrrole ring suggests the
introduction of a small lipophilic substituent that could
increase the antimycobacterial activity. In fact, con-
sidering the obtained classical QSAR equation, a higher
value of logP and G_CDS along with a limited
enhancement of the molecular volume could lead to
more potent derivatives. Moreover the CoMFA maps
suggest the introduction of more molecular diversity to
explore those regions that were not found relevant in
the 3-D-QSAR.
In general, a modi®cation leading to an increase of the
overall lipophilicity should lead to an increase of activ-
ity (high positive coecients for HlogP and G_CDS).
On the other hand, the overall bulk of the molecules
does not have to increase too much (small negative
coecient for the molecular volume) to decrease activ-
ity. In respect to the other two important parameters,
LUMO and DipMom, they seem to be in agreement
with what was observed for the previous CoMFA maps.
Large variations of the electrostatic nature of the mole-
cule may lead to large variations of the biological
activity (high coecients for both the dipole moment
and the LUMO energy of opposite sign).
A set of six compounds reserved from training was used
to test the ecacy of the obtained models. The models
show almost the same statistical coecients, although
the best overall is the mixed QSAR±CoMFA model
showing the lowest value of both the average absolute
Experimental
Melting points were determined on a Buchi 530 melting
point apparatus and are uncorrected. Infrared (IR)
spectra (Nujol mulls) were recorded on a Perkin±Elmer
1
297 instrument. H NMR spectra were recorded at 200
MHz on a Bruker AC 200 spectrometer. Chemical shifts
are reported in d (ppm) units relative to the internal
reference tetramethylsilane (TMS). All compounds were
routinely checked by TLC and 1H NMR. TLC was
performed by using Stratocrom SIF Fluka (silica gel
precoated plates with ¯uorescent indicator) or Strato-
crom ALF Fluka (aluminum oxide precoated plates
Figure 4. Hybrid model: training set ®tting and test set predictions.

1430
R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
with ¯uorescent indicator). Developed plates were
visualized by UV light. Chromatographic puri®cations
were performed on Merck aluminum oxide (70±230
mesh) and Merck silica gel (70±230 mesh). Solvents
were reagent grade and, when necessary, were puri®ed
and dried by standard methods. Concentration of solu-
tions after reactions and extractions involved the use of
a rotary evaporator operating at a reduced pressure of
approximately 20 Torr. Organic solutions were dried
over anhydrous sodium sulfate. Analytical results
agreed to within‹0.40% of the theoretical values.
Microanalyses were performed by Laboratories of
Dipartimento di Scienze Farmaceutiche, Universita di
Padova, Italy. Chemical and physical data of newly
synthesized compounds 13±18 and 25±30 are reported in
Table 4.
mL) and extracted with ethyl acetate (3Â100 mL). The
organic extracts were collected, washed with brine
(3Â200 mL) and dried. Removal of the solvent gave a
crude product which was chromatographed on alumi-
num oxide column (ethyl acetate as eluent) to give pure
17 (1.5 g, 36% yield); IR: n 2700 (NH) and 1680 (CO)
cm 1; 1H NMR (CDCl3): d 1.18 (m, 6H, CH3), 2.58 (q,
2H, CH2CH3), 4.16 (q, 2H, OCH2CH3), 7.27 (m, 2H,
pyridine C3-H and C5-H), 7.43 (d, 1H, J1,2=3.0 Hz,
pyrrole C2-H), 8.57 (m, 2H, pyridine C2-H and C6-H),
9.10 (bs, 1H, NH); anal. C14H16N2O2 (244.29) C, H, N.
This procedure was used for the synthesis of 13 starting
from toluene-4-sulfonylmethylisocyanide (TosMIC).
Spectroscopic data of 13 are reported below.
1
13: IR: n 3100 (NH) and 1710 (CO) cm 1; H NMR
(DMSO-d6): d 1.21 (t, 3H, CH3), 4.15 (q, 2H, CH2), 7.19
(m, 1H, pyrrole C5-H), 7.47±7.56 (m, 3H, pyridine C3-
H and C5-H and pyrrole C2-H), 8.47 (m, 2H, pyridine
C2-H and C6-H), 11.77 (bs, 1H, NH); anal. C12H12
N2O2 (216.24) C, H, N.
Syntheses
Speci®c examples presented below illustrate general
synthetic procedures.
Ethyl 3-(4-pyridinyl)-2-propenoate. A solution of pyri-
dine-4-carboxaldehyde (5.0 g, 0.047 mol) in absolute
ethanol (80 mL) was added into a well stirred mixture of
triethylphosphonoacetate (12.6 g, 0.056 mol) and anhy-
drous potassium carbonate (19.5 g, 0.141 mol). The
resulting suspension was stirred at 70 ꢀC for 2.5 h, then
the solvent was removed and the residue was treated
with water (300 mL) and extracted with chloroform
(3Â150 mL). The organic extracts were collected,
washed with brine (3Â200 mL) and dried. Evaporation
of the solvent gave pure ethyl 3-(4-pyridinyl)-2-pro-
penoate (8.33 g, 100% yield).22
Ethyl 1-(4-nitrobenzenesulfonyl)-4-(4-pyridinyl)pyrrole-3-
carboxylate (29). Potassium (490 mg, 0.013 mol) was
added into a well stirred solution of 13 (3.0 g, 0.014
mol) in anhydrous tetrahydrofuran (150 mL) and the
mixture was re¯uxed until the metal disappeared. The
cooled mixture was treated with 18-crown-6 (3.7 g,
0.014 mol), then a solution of 4-nitrobenzenesulfonyl
chloride (3.1 g, 0.014 mol) in anhydrous tetrahydro-
furan (250 mL) was dropped into the above mixture
over 30 min. After addition the solution was stirred at
room temperature for 2.5 h, then the solvent was
removed. The residue was treated with water (200 mL)
and extracted with ethyl acetate (3Â100 mL). The
organic extracts were collected, washed with brine and
dried. Removal of the solvent gave a residue which was
chromatographed on aluminum oxide column (ethyl
Ethyl 5-ethyl-4-(4-pyridinyl)pyrrole-3-carboxylate (17).
A solution of ethyl 3-(4-pyridinyl)-2-propenoate (3.0 g,
0.017 mol) and 1-(4-methylphenylsulfonyl)propyliso-
cyanide23 (3.8 g, 0.017 mol) in anhydrous dimethylsulf-
oxide:diethyl ether (30:60 mL) mixture was added by
dropping into a well-stirred suspension of sodium
hydride (0.068 mol, 2.7 g of 60% suspension in white
oil) in anhydrous diethyl ether (80 mL) under argon
stream. When addition stopped the mixture was stirred
at room temperature for 15 min, treated with water (200
acetate as eluent) to furnish 29 (2.19 g, 39% yield); IR: n
1700 (CO) cm
1
;
1H NMR (CDCl3): d 1.26 (t, 3H,
CH3), 4.23 (q, 2H, CH2), 7.26 (d, 1H, J2,5=2.8 Hz,
pyrrole C5-H), 7.33 (m, 2H, pyridine C3-H and C5-H),
7.88 (d, 1H, J2,5=2.8 Hz, pyrrole C2-H), 8.16 (m, 2H,
benzene C3-H and C5-H), 8.42 (m, 2H, benzene C2-H
Table 4. Chemical and physical data of derivatives 13±18 and 25±30
Compd
Formula
Mp (ꢀC)
Recrystallization solventa
Yield (%)
Reaction time (h)
Chromatographic systemb
13
14
15
16
17
18
25
26
27
28
29
30
C12H12N2O2
C10H10N4O
134±135
270±272
58±60
99±100
199±201
257±259
Oil
a
b
c
c
d
e
Ð
Ð
d
c
a
a
99
73
84
74
36
12
55
20
33
90
39
69
0.25
0.25
2.5
2.5
0.25
3
1.5
1.5
1
Ð
Ð
A
A
A
Ð
B
B
Ð
C
C
13H14N2O2
C14H16N2O2
14H16N2O2
C
C12H14N4O
C19H17C1N2O2
C
21H22N2O2
C17H15C1N4O
19H18N2O4S
C18H15N3O6S
17H16N2O3S
Oil
165±167
107±108
174±175
135±137
C
1
3
0.5
A
Ð
C
aa, benzene; b, ethanol; c, benzene/cyclohexane; d, toluene; e, isopropanol.
bA, aluminum oxide/ethyl acetate; B, silica gel/ethyl acetate; C, aluminum oxide/chloroform.

R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
1431
and C6-H), 8.58 (m, 2H, pyridine C2-H and C6-H);
anal. C18H15N3O6S (401.39) C, H, N, S.
2.5 Hz, pyrrole C5-H), 7.41±7.44 (m, 3H, pyridine C3-H
and C5-H and pyrrole C2-H), 8.53 (m, 2H, pyridine C2-
H and C6-H); anal. C14H16N2O2 (244.29) C, H, N.
Compound 28 was obtained with the same procedure;
IR: n 1720 (CO) cm 1; 1H NMR (CDCl3): d 1.26 (t, 3H,
CH2CH3), 2.45 (s, 3H, CH3), 4.23 (q, 2H, CH2), 7.23±
7.39 (m, 5H, pyrrole C5-H, pyridine C3-H and C5-H
and benzene C3-H and C5-H), 7.83±7.88 (m, 3H, pyr-
role C2-H and benzene C2-H and C6-H), 8.57 (m, 2H,
pyridine C2-H and C6-H); anal. C19H18N2O4S (370.42)
C, H, N, S.
1
26: IR: n 1700 (CO) cm 1; H NMR (CDCl3): d 1.00
and 1.19 (2t, 6H, CH3), 2.46 (q, 2H, CH2CH3), 4.12 (q,
2H, OCH2CH3), 5.11 (s, 2H, CH2), 7.08±7.38 (m, 8H,
pyridine C3-H and C5-H, pyrrole H and benzene H),
8.57 (m, 2H, pyridine C2-H and C6-H); anal. C21H22
N2O2 (334.42) C, H, N.
4-(4-Pyridinyl)pyrrole-3-carboxyhydrazide (14). A mix-
ture of 13 (500 mg, 2.3 mmol), hydrazine hydrate (3
mL) and ethanol (1.5 mL) was re¯uxed for 15 min.
After cooling the precipitate which formed was ®ltered
and recrystallized from ethanol to obtain pure 14 (340
mg, 73% yield); IR: n 3280 (NH and NH2), 1630 (CO)
cm 1; 1H NMR (DMSO-d6): d 4.28 (bs, 2H, NH2), 7.24
(s, 2H, pyrrole H), 7.50 (m, 2H, pyridine C3-H and
C5-H), 8.40 (m, 2H, pyridine C2-H and C6-H), 9.08 (bs,
1H, CONH), 11.40 (bs, 1H, NH); anal. C10H10N4O
(202.22) C, H, N.
1-(4-Methylbenzenesulfonyl)-4-(4-pyridinyl)pyrrole-3-
methanol (30). A solution of 28 (300 mg, 0.8 mmol) in
anhydrous tetrahydrofuran (6 mL) was added by drop-
ping into a well stirred suspension of lithium aluminum
hydride (70 mg, 1.8 mmol) in the same solvent (6 mL).
After addition the mixture was stirred at room tem-
perature for 30 min and then carefully treated with cru-
shed ice. The inorganic precipitate was removed and the
solution was concentrated and shaken with brine, dried
and evaporated to give 30 (180 mg, 69% yield); IR: n
3180 (OH) cm
1
;
1H NMR (CDCl3): d 2.41 (s, 3H,
CH3), 2.49 (bs, 1H, OH), 4.61 (s, 2H, CH2), 7.26±7.45
(m, 6H, pyrrole H, pyridine C3-H and C5-H and ben-
zene C3-H and C5-H), 7.80 (m, 2H, benzene C2-H and
C6-H), 8.52 (m, 2H, pyridine C2-H and C6-H); anal.
C17H16N2O3S (328.39) C, H, N, S.
This procedure was used for the synthesis of derivatives
18 and 27.
18: IR: n 3320, 3160 and 3080 (NH and NH2), 1635
1
(CO) cm 1; H NMR (DMSO-d6): d 1.11 (t, 3H, CH3),
2.52 (q, 2H, CH2), 4.25 (bs, 2H, NH2), 7.18 (m, 3H,
pyridine C3-H and C5-H, pyrrole H), 8.44 (m, 2H, pyr-
idine C2-H and C6-H), 8.92 (bs, 1H, CONH), 11.22 (bs,
1H, NH); anal. C12H14N4O (230.27) C, H, N.
Ethyl 1-(4-chlorobenzyl)-4-(4-pyridinyl)pyrrole-3-carb-
oxylate (25). A solution of 4-chlorobenzyl chloride (740
mg, 4.6 mmol) in dichloromethane (14 mL) was added
by dropping into a well stirred mixture of 13 (1.0 g,
4.6 mmol) and tetrabutylammonium hydrogen sulfate
(140 mg, 0.4 mmol) in a mixture of dichloromethane (14
mL) and 50% aqueous NaOH (5 mL) cooled at 0 ꢀC.
The resulting mixture was stirred at room temperature
for 15 h, then diluted with water (50 mL) and extracted
with chloroform (3Â25 mL). The organic extracts were
collected, washed with brine (3Â50 mL) and dried.
Removal of the solvent gave a residue which was chro-
matographed on silica gel column (ethyl acetate as elu-
ent) to give pure 25 (860 mg, 55% yield); IR: n 1700
(CO) cm 1; 1H NMR (CDCl3): d 1.26 (t, 3H, CH3), 4.22
(q, 2H, CH2CH3), 5.05 (s, 2H, CH2), 6.76 (d, 1H, J=2.2
Hz, pyrrole C5-H), 7.11±7.42 (m, 7H, pyrrole C2-H,
pyridine C3-H and C5-H and benzene H), 8.53 (m, 2H,
pyridine C2-H and C6-H); anal. C19H17ClN2O2 (340.81)
C, H, N, Cl.
1
27: IR: n 3180 (NH and NH2), 1630 (CO) cm 1; H
NMR (DMSO-d6): d 4.33 (bs, 2H, NH2), 5.14 (s, 2H,
CH2), 7.30±7.46 (m, 8H, pyridine C3-H and C5-H, pyr-
role H and benzene H), 8.41 (m, 2H, pyridine C2-H and
C6-H), 9.16 (bs, 1H, NH); anal. C17H15ClN4O (326.79)
C, H, N, Cl.
Microbiology
Compounds. Test derivatives and drug references were
dissolved in DMSO at the initial concentration of 10
mg/mL and stored in cold place until use.
Cytotoxicity. The cytotoxicity of above compounds was
tested on VERO cell monolayers (ICN-FLOW), grown
in Dulbecco's modi®ed MEM (GIBCO Lab. Inc.) with
2% fetal calf serum. Six-well culture plates were inocu-
lated with 9Â104 cells. After 24 h the compounds were
added and after ®ve further days the cells were detached
from wells, trypsinized and counted in a Neubauer
chamber. The minimal toxic dose (MTD50) was the
concentration of drugs that induced a reduction of 50%
of cell growth with respect to the control.
By this procedure were prepared compounds 15, 16 and
26; their spectroscopic data are reported below.
15: IR: n 1705 (CO) cm 1; 1H NMR (CDCl3): d 1.26 (t,
3H, CH2CH3), 3.71 (s, 3H, CH3), 4.22 (q, 2H, CH2),
6.74 (d, 1H, J2,5=2.5 Hz, pyrrole C5-H), 7.35 (d, 1H,
J2,5=2.5 Hz, pyrrole C2-H), 7.46 (m, 2H, pyridine C3-
H and C5-H), 8.53 (m, 2H, pyridine C2-H and C6-H);
anal. C13H14N2O2 (230.27) C, H, N.
Antimycobacterial activity. All compounds were pre-
liminarily assayed against two freshly isolated clinical
strains, M. fortuitum CA10 and M. tuberculosis B814,
according to di€usion method in agar.33 Growth
media were Mueller-Hinton (Difco) containing 10% of
OADC (oleic acid, albumin and dextrose complex) for
16: IR: n 1705 (CO) cm 1; 1H NMR (CDCl3): d 1.27 (t,
3H, OCH2CH3), 1.49 (t, 3H, CH2CH3), 3.97 (q, 2H,
CH2CH3), 4.22 (q, 2H, OCH2CH3), 6.78 (d, 1H, J2,5=

1432
R. Ragno et al. / Bioorg. Med. Chem. 8 (2000) 1423±1432
M. fortuitum and Middlebrook 7H9 agar (Difco) with
10% of ADC (albumin dextrose complex) for M.
tuberculosis. Substances were tested at the single dose of
100 mg/mL. Compounds that showed activity in the
preliminary test were assayed for inhibitory activity
against M. tuberculosis CIP 103471 in comparison with
streptomycin (SM) and isoniazid (INH) as reference
drugs. A broth microdilution assay was used for rapidly
growing strains.33 35 Minimum inhibitory concentra-
tions (MICs in mg/mL) for each compound were deter-
mined. The MICs for test derivatives and controls were
determined by the BACTEC 460 TB method.36,37
15. Di Santo, R.; Costi, R.; Artico, M.; Massa, S.; Lampis, G.;
Deidda, D.; Pompei, R. Bioorg. Med. Chem. Lett. 1998, 8,
2931.
16. Massa, S.; Di Santo, R.; Mai, A.; Botta, M.; Artico, M.;
Panico, S.; Simonetti, G. Farmaco. 1990, 45, 833.
17. Massa, S.; Di Santo, R.; Costi, R.; Mai, A.; Artico, M.;
Retico, A.; Apuzzo, G.; Artico, M.; Simonetti, G. Med. Chem.
Research 1993, 3, 192.
18. Massa, S.; Di Santo, R.; Artico, M.; Costi, R.; Di Filippo,
C.; Simonetti, G.; Retico, A.; Artico, M. Eur. Bull. Drug
Research 1992, 1, 12.
19. Cramer, R. D.III; Patterson, D.; Bunce, J. J. Am. Chem.
Soc. 1988, 110, 5959.
20. Cramer, R. D., III; DePriest, S.; Patterson, D.; Hecht, P.
In 3-D-QSAR in Drug Design; Kubinyi, H., Ed.; ESCOM:
Leiden, 1993; pp 443±485.
21. Cramer, R. D., III; Wold, S. B. US Patent 5,025,388, 18
June 1991.
22. Hallinan, E. A.; Hagen, T. J.; Husa, R. K.; Tsymbalov, S.;
Rao, S. N.; van Hoeck, J.-P.; Ra€erty, M. F.; Stapelfeld, A.;
Savage, M. A.; Reichman, M. J. Med. Chem. 1993, 36, 3293.
23. van Leusen, A. M.; Bouma, R. J.; Possel, O. Tetrahedron
Lett. 1975, 3487.
Acknowledgements
Thanks are due to Ministero della Sanita, Istituto
Superiore di Sanita, II Progetto Tubercolosi 1997 (Grant
No. 96/D/T48) and to Italian MURST for partial support.
References
24. Greco, G.; Novellino, E.; Fiorini, I.; Nacci, V.; Campiani,
G.; Ciani, S. M.; Garofalo, A.; Bernasconi, P.; Mennini, T. J.
Med. Chem. 1994, 37, 4100.
25. Ta®, A.; Anastassopoulous, J.; Theophanides, T.; Botta,
M.; Corelli, F.; Massa, S.; Artico, M.; Costi, R.; Di Santo, R.;
Ragno, R. J. Med. Chem. 1996, 39, 1227.
26. In Tripos Associates, Inc., 1699 S Hanley Rd, St. Louis,
MO 63144, USA.
27. Halgren, T. A. Journal of Computational Chemistry 1996,
17, 490.
28. Lindgren, F.; Geladi, P.; Rannar, S.; Wold, S. J. Chemo-
metrics 1994, 8, 349.
29. Clark, M.; Cramer, R. I. Quant. Struct. Act. Relat. 1993,
12, 137.
30. In Kellogg, K. E.; Abraham, D. J.; Richmond, VA 23298±
0540.
31. Kellog, G.; Semus, S.; Abraham, D. HINT: J. Comput.
Aided Mol. Design 1991, 5, 545.
32. Cramer, C. J. Acc. Chem. Res. 1993, 26, 599.
33. Hawkins, J. E.; Wallace, R. J.; Brown, B. A. In Manual of
Clinical Microbiology; Balows, W. J.; Hausler, K. L.; Her-
mann, K. L.; Isemberg, H. D.; Shadomy, H. J., Eds. ASM:
Washington, DC, 1991; pp 1138±1152.
34. National Committee for Clinical Laboratory Standards.
Antimycobacterial Susceptibility Testing. Proposed Standard
M24; NCCLS: Villanova, PA, 1990.
35. Brown, B. A.; Swenson, J. M.; Wallace, R. J., Jr. In Clin-
ical Microbiology Procedures Handbook American Society for
Microbiology: Washington, DC, 1992; Vol. I.
36. Helfets, L. Antimicrob. Agents Chemother. 1996, 40, 1759.
37. Inderlied, C. B.; Salenger, M. In Manual of Clinical
Microbiology, 5th ed.; Murray, P. R.; Baron, E. J.; Pfaller, M.
A.; Tenover, F. C.; Volken, R. H., Eds.; American Society of
Microbiology: Washington, DC; pp 1385±1404.
1. Baker, R. W.; Mitscher, L. A.; Arain, T. M.; Shawar, R.;
Stover, C. K. Ann. Rep. Med. Chem. 1996, 31, 161.
2. Duncan, K. Exp. Opin. Ther. Patents 1997, 7, 129.
3. Niccolai, D.; Tarsi, L.; Thomas, R. J. Chem. Commun. 1997,
2333.
4. Horsburgh, C. R., Jr. The New England Journal of Medicine
1991, 324, 1332.
5. Heifets, L. Antimicrob. Agents Chemother. 1996, 40, 1759.
6. Barbachyn, M. R.; Hutchinson, D. K.; Brickner, S. J.;
Cynamom, M. H.; Kilburn, J. O.; Klemens, S. P.; Glickman,
S. E.; Grega, K. C.; Hendges, S. K.; Toops, D. S.; Ford, C.
W.; Zurenke, G. E. J. Med. Chem. 1996, 39, 680.
7. Ashtekar, D. R.; Costa-Pereira, R.; Shrinivasan, T.; Iyyer,
R.; Vishvanathan, N.; Rittel, W. Diagn. Microbiol. Infect. Dis.
1991, 14, 465.
8. Nagarajan, K.; Shankar, R. G.; Rajappa, S.; Shenopy, S. J.;
Costa-Pereira, R. Eur. J. Med. Chem. 1989, 24, 631.
9. Ashtekar, D. R.; Costa-Pereira, R.; Nagarajan, K.; Vishva-
nathan, N.; Bhatt, A. D.; Rittel, W. Antimicrob. Agents Che-
mother. 1993, 37, 183.
10. Cynamon, M. H.; Gimi, R.; Gyenes, F.; Sharpe, C. A.;
Bergmann, K. E.; Han, H. J.; Gregor, L. B.; Rapolu, R.;
Luciano, G.; Welch, J. T. J. Med. Chem. 1995, 38, 3902.
11. Cynamon, M. H.; Speirs, R. J.; Welch, J. T. Antimicrob.
Agents Chemother. 1998, 42, 462.
12. Johnsson, K.; Schultz, P. G. J. Am. Chem. Soc. 1994, 116,
7425.
13. Dessen, A.; Quemard, A.; Blanchard, J. S.; Jacobs, W. R.,
Jr; Sacchettini, J. C. Science 1995, 24, 1638.
14. Deidda, D.; Lampis, G.; Fioravanti, R.; Biava, M.; Porretta,
G. C.; Zanetti, S.; Pompei, R. Antimicrob. Agents Chemother.
1998, 42, 3035.