Identification of antitubercular benzothiazinone compounds by ligand-based design

Article abstract of DOI:10.1021/jm3008882

1,3-Benzothiazin-4-ones (BTZs) are a novel class of TB drug candidates with potent activity against M. tuberculosis. An in silico ligand-based model based on structure-activity data from 170 BTZ compounds was used to design a new series. Compounds were tested against a panel of mycobacterial strains and were profiled for cytotoxicity, stability, and antiproliferative effects. Several of the compounds showed improved activity against MDR-TB while retaining low toxicity with higher microsomal, metabolic, and plasma stability.

Full text of DOI:10.1021/jm3008882

Brief Article
pubs.acs.org/jmc
Identification of Antitubercular Benzothiazinone Compounds by
Ligand-Based Design
Tomislav Karoli,† Bernd Becker,† Johannes Zuegg,† Ute Mollmann,‡ Soumya Ramu,† Johnny X. Huang,†
̈
and Matthew A. Cooper*,†
†Institute for Molecular Bioscience, University of Queensland, St. Lucia, Queensland 4072, Australia
‡Department of Molecular and Applied Microbiology, Leibniz Institute for Natural Product Research and Infection Biology, Hans
Knoell Institute, Beutenbergstrasse 11a, D-07745 Jena, Germany
S
* Supporting Information
ABSTRACT: 1,3-Benzothiazin-4-ones (BTZs) are a novel class of TB drug candidates with potent activity against M.
tuberculosis. An in silico ligand-based model based on structure−activity data from 170 BTZ compounds was used to design a
new series. Compounds were tested against a panel of mycobacterial strains and were profiled for cytotoxicity, stability, and
antiproliferative effects. Several of the compounds showed improved activity against MDR-TB while retaining low toxicity with
higher microsomal, metabolic, and plasma stability.
Rv3790), which with DprE2 (Rv3791) is critical for the
synthesis of cell-wall arabinans and catalyzes the epimerization
of decaprenylphosphoryl-β-D-ribose to decaprenylphosphoryl-
β-D-arabinose.5 It has been shown6 that BTZ compounds act as
suicide inhibitors after enzymatic activation of the nitro as the
nitroso and subsequent covalent reaction with Cys387 of
DprE1resulting in enzymatic inhibition.
A review of the available pharmacokinetics of these BTZ
compounds revealed that many had poor solubility and that the
bioavailability and ADME/T properties could be improved.
Therefore, the principle goal of this study was to design a
second generation series with increased solubility and to
explore the metabolic stability and toxicity profile of the new
compounds.
INTRODUCTION
■
Tuberculosis (TB), an infectious disease most commonly
caused by Mycobacterium tuberculosis (Mtb), claims ∼2 million
lives each year. According to the most recent WHO report, 4.7
million cases of TB were recorded in 2009, 250 000 of which
were estimated to involve multidrug resistant TB (MDR-TB).
Of these MDR-TB cases, around 7% (up to 33% in some
countries) have been shown to be extensively drug resistant
(XDR-TB).1 The increasing failure of the current therapies for
TB and the dire lack of drugs in the development pipeline
highlight the urgent need to develop new drugs.2
1,3-Benzothiazin-4-ones (BTZs, Figure 1) are a novel class of
TB drug candidates with potent activity against Mtb in vitro
METHODOLOGY AND CHEMISTRY
■
As indicated in Figure 1, the lead 2 can be broken down
conceptually into a BTZ core bearing three substituents (R1,
R2, and R3). The structure−activity relationship (SAR) around
groups R1 and R2 has been found to be very restrictive with
little modification tolerated, while considerable variation in the
structure of the R3 group is possible. In this study, the groups
at R1 and R2 positions were fixed to NO2 and CF3,
respectively, and only variations at position R3 were explored.
A BTZ specific ligand-based model was derived from 541
MIC values for 160 of the 170 compounds from the BTZ
series3 and 21 different bacterial organism/strains. The model
was built using Bayesian statistics by correlating 2D molecular
fingerprints of the compounds to their classification of being
active (MIC < 1 μM) or inactive (MIC > 1 μM). For a
compound predicted to be active, a predicted pMIC value was
derived by interpolation from its nearest structural neighbor
(most similar actives). Models were only derived for those
Figure 1. Lead benzothiazinone structures. R1 acts as site of covalent
addition to DprE1 after enzymatic activation, and an electron
withdrawing group at R2 is required. Variation is permitted in R3.
Both enantiomers (chiral center highlighted by ∗) are equipotent in
vitro.
and in vivo.3 Previous studies3,4 identified the racemic
benzothiazinone 1 (BTZ-10526038) and its chiral analogue 2
(BTZ-10526043, BTZ043) (Figure 1) as lead structures.
Compound 2 exhibits a minimum inhibitory concentration
(MIC) of 1 ng/mL against Mtb H37Rv, a factor of 20 below
that of the frontline TB drug isoniazid (INH). The target of 2 is
decaprenylphosphoryl-β-D-ribose 2′-epimerase (DprE1,
Received: June 29, 2012
Published: August 23, 2012
© 2012 American Chemical Society
7940
dx.doi.org/10.1021/jm3008882 | J. Med. Chem. 2012, 55, 7940−7944

Journal of Medicinal Chemistry
Brief Article
Figure 2. Amines selected by ligand-based design for exploration of the R3 position after addition to the BTZ core.
a
Scheme 1. Preparation of Spiroketal BTZ Derivatives
a
(i)EtOH, amines 3a−ae, 60 °C; (ii) diol, DIPEA, TMSOTf, 0 °C, DCM, then ketone, TMSOTf, DCM, rt; (iii) 10% Pd/C, Et3SiH or cyclohexene,
MeOH; (iv) 4, EtOH, 60 °C.
organism/strains having experimental MIC data for more than
20 compounds in the BTZ set, resulting in predictions for
seven strains (Supporting Information Table S1). The
predictability was then assessed by leave-one-out cross-
validation. In all cases a good level of prediction was achieved.
For example, for Mtb H37Rv, the model is able to correctly
identify as actives 90% of the total number of truly active
compounds, without including a single inactive compound in
the prediction. For this model in particular, 60% of active
compounds are assigned a predicted MIC within 10-fold of its
corresponding experimental MIC. Poor aqueous solubility had
been identified as a characteristic property of the BTZ series,
and the most highly active compounds exhibit a low solubility.
Therefore, predictive solubility models were included in the
design of new compounds (Pipeline Pilot, Accelrys).7
containing 7.5 million commercial compounds, selecting
structurally similar amines. After removal of molecules
containing reactive groups, 2797 commercially available amines
were selected. For each of the 2797 amines the corresponding
BTZ compound was generated using virtual combinatorial
chemistry. The resulting virtual compounds were filtered for
molecular weight (350−500 Da), rotatable bonds (<8), and
predicted solubility (log S > −6.5). Further filtering was carried
out using the predictive MIC model for Mtb H37Rv, removing
compounds that were either predicted to be inactive or outside
the applicability domain of the model. The resulting 333
compounds were divided into 20 clusters using the FCFP_4
2D fingerprints and nearest-neighbor clustering. Within each
cluster the three compounds with highest predicted solubility
(log S) were selected.
On the basis of these models, a target list of novel BTZ
analogues was generated to select a diverse set of R3
substituents with good predicted MIC and good predicted
solubility. R3 amines from the 170 BTZ compounds were used
as templates for virtual screening against the chemical catalogue
From these compounds 28 amines (3a−ab) were selected
based on availability and cost (Figure 2). A further 3 amines
(3ac−ae) outside the predictive models were also chosen. The
target compounds were prepared by nucleophilic addition of
amines to thiomethyl benzothiazinone 44,8 to give the BTZ
7941
dx.doi.org/10.1021/jm3008882 | J. Med. Chem. 2012, 55, 7940−7944

Journal of Medicinal Chemistry
Brief Article
a
Table 1. Antibacterial, Stability, and Toxicity Testing Data for Selected BTZ Compounds
MIC [μg/mL]
stability [%]
toxicity [μM]
MIC [μg/mL]
stability [%]
toxicity [μM]
b
M. smeg,
c
g
HUVEC,
g
b
M. smeg,
c
g
HUVEC,
g
buffer,
buffer,
e
plasma,
f
e
plasma,
f
M. fort. B,
d
K-562,
h
M. fort. B,
d
K-562,
h
compd
MDR-TB
microsome
HeLa
compd
MDR-TB
microsome
HeLa
i
2
0.008
0.00156
≤0.015
0.004
0.012
0.03
98
8
5q
0.25
101
ND
i
107/97
98/45
91
11
>50
20
37
41
17
21
48
18
30
39
>0.1
64/92
61/80
96
ND
i
≤0.125
0.004
ND
5a
5z
18
i
86/114
43/32
14
ND
92/114
6/7
30
>0.03
0.004
38
5g
5m
5n
5p
0.125
0.05
5ac
7
82
>50
>50
>50
16
108/93
28/37
78
0.0125
≤0.015
0.015
96/71
120/75
98
≤0.125
0.015
0.0125
0.03
95/115
13/2
98
0.0125
>0.1
51/56
81/29
82
22
36
i
0.25
ND
11b
11e
0.015
>50
>50
>50
>50
>50
43
i
0.025
≤0.125
0.001
0.00625
0.03
99/111
42/13
95
ND
0.00625
>0.03
0.004
90/88
i
i
ND
ND/ND
>50
>50
>50
99
90/104
6/8
0.0031
≤0.015
98/103
6/9
a
b
c
d
All compounds showed CC50 of >10 μM against HepG2 and HEK293 cell lines. M. smegmatis 700084. M. fortuitum B. M. tuberculosis 2745/09
e
f
g
h
i
MDR. Human/mouse. Liver microsomes human/mouse. GI50. CC50. Not determined.
analogues 5a−ae (Scheme 1). This process allows for a very
convenient point of convergence in the synthesis, as 4 can be
prepared on large scale and nucleophilic amines can then be
readily added into the R3 position. For analogues 5t, 5x, 5y,
and 5aa insufficient quantities of the desired products were
obtained for assay after several attempts at synthesis and the
compounds were not pursued further for this study.
assays using the 7H9. Active compounds were then tested for
mammalian cell cytotoxicity against HepG2 and HEK293 cells.
By use of this initial testing data, the active compounds were
subjected to further testing for MIC against M. smegmatis
SG987, M. vaccae 10670, M. fortuitum B, M. aurum SB66, and
MDR-TB 2745/09, stability in buffer and human and mouse
liver microsomes, antiproliferative effects, and cytotoxicity
(Tables 1 and S4).
The methodology used to design 5a−ab selects only amines
that are commercially available. To explore the region around
the spiroketal of 2, an additional group of compounds was
designed where chemical steps were required in preparation of
the amines (Scheme 1). While formation of the ketals using
standard Dean−Stark conditions was found to be tedious on
small scale, a one pot version of the TMSOTf catalyzed
addition of the bis-TMS ether of the diol9 proved to be very
rapid and yielded clean products. This procedure could be
applied to ketones already installed on the BTZ core (e.g., 6a,
6b), or on a N-protected aminoketone (10a−g) before
deprotection and condensation with 4. For several of the
spiroketal compounds, diastereomers were formed at the spiro
junction. In most cases the diastereomers were not separable
and the compounds were assayed as mixtures. However, for
two compounds these diastereomers were separable by
preparative HPLC and were assayed independently (11c/11d
and 11f/11g, respectively).
To compare the compounds in this study with the previous
results, we tested each compound in MIC assays for M.
smegmatis 700084 using two different media: Middlebrook 7H9
and Mueller−Hinton broth. Most of the compounds showed
only slightly different MIC values, within a log difference in the
two broths (Table S3). The only exception was 5ac, which
showed 3 orders of magnitude higher activity (lower MIC)
when assayed in Mueller Hinton broth. This compound also
exhibits very low aqueous solubility (around 60 nM), which
may be responsible for the different MIC values. In general all
the assays using the MHB as growth medium resulted in
slightly lower MIC values (higher activity) compared to the
DISCUSSION AND CONCLUSIONS
■
MIC testing revealed several compounds that were more active
or equivalent to 2 against M. smegmatis (5p, 5m, 5z, 5ac, 11b,
11e, and 7) and MDR-TB (5a, 5p, 5z, 5ac, 11b, and 11e). The
observed therapeutic indices (CC50/MIC) were excellent for
candidate antimicrobials (>10 000), notably for 5p (>50 000)
and 5ac (>12 500), for which the indices greatly exceeded that
of compound 2 (>6250). The differences were more
pronounced when comparing antiproliferative effects (GI50/
MIC of >50 000, >12 500, and 1190, respectively). The
compounds were generally stable to buffer and human and
mouse plasma and liver microsomes. However, reduced
stability was observed for 5g, 5n, 5m, 5p, 5z, and 11e. With
a view to future in vivo studies, it was interesting to note the
difference in stability in the presence of human or mouse
microsomes observed for some compounds (eg 2, 5m, 5n, and
7). Evidence that the newly designed compounds retain the
mode of action of compound 2 was shown in the lack of activity
against M. aurum SB66 as the orthologous enzyme possesses a
serine rather than a cysteine at residue 387, known to be critical
to the action of the BTZ class.3
Of the 24 compounds tested from the in silico design
approach, 13 were found to have activities against M. smegmatis
of better than 1 μg/mL, while only two compounds were found
to be inactive at our highest test concentration (2 μg/mL),
indicating about a 90% success rate for the model at predicting
active compounds (Table S2). The kinetic solubility of a
selected group of compounds was measured at pH 7.4 using the
7942
dx.doi.org/10.1021/jm3008882 | J. Med. Chem. 2012, 55, 7940−7944

Journal of Medicinal Chemistry
Brief Article
methods described by Alsenz,10 Hoelke,11 and Lipinski.12 For
two compounds there was a large variation observed between
the duplicate determinations and they could only be placed
qualitatively in the partially soluble band.13 As can be seen in
Table 2, the predicted solubility values were found to be within
a log unit of those measured experimentally. Four compounds
had equivalent or improved solubility relative to 2.
Each group or area in the plot contains highly active
compounds and low or inactive compounds, suggesting that
the target does not have a single preferred structure for the R3
position and that it should be possible to tune compound
pharmacokinetic properties by varying the R3 moiety. Indeed,
highly active compounds have been found with solubilities
ranging from 60 nM (5ac) to 12 μM (5m) and calculated log P
values ranging over 3 orders of magnitude.
The ligand based modeling method used has shown a
remarkable ability to distinguish actives from inactives for
antitubercular BTZs and has allowed the exploration of the
SAR within the broad structural space of the training set. The
ability to bias the compound set toward any desired calculated
physical property allows the refinement of the pharmacokinetic
properties of the products. The extensive toxicity and stability
data obtained allow us to select compound classes to progress
into more detailed SAR studies and subsequent animal models.
Table 2. Comparison of Predicted and Measured Solubility
measured result
compd
predicted log S
log S
μM
6.8
2
−4.73
−5.09
−5.31
−5.03
−4.83
−5.38
−5.85
−5.17
−4.94
−6.05
−6.94
−5.16
−5.79
−5.74
5e
5f
1.63
1.82
5h
5j
partially soluble
partially soluble
5k
5m
5q
5w
5z
5ac
−4.66
21.8
11.9
8.1
ASSOCIATED CONTENT
* Supporting Information
−4.93
−5.09
−4.98
−5.53
−7.23
■
S
10.6
2.97
0.060
Details of synthesis and characterization, modeling, and
biological assays (including the first round assays and the full
MIC and toxicity panel of the second round). This material is
available free of charge via the Internet at http://pubs.acs.org.
To compare the newly designed compound in this study and
the previous BTZ compound, a structure diversity analysis was
performed. By use of 2D fingerprints (ECFP_6) and Tanimoto
distances, a distance matrix between each of the original BTZs
and our new compounds was calculated. For visual
representation of the distances between each compound,
standard multidimensional scaling (MDS) has been applied
to reduce the multidimensional distance matrix to two
dimensions (distances within the two-dimensional plot
represent the structural distances in the matrix)
AUTHOR INFORMATION
Corresponding Author
*Phone: +61 7 3346 2044. Fax: +61 7 3346 2090. E-mail: m.
cooper@uq.edu.au.
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
As Figure 3 illustrates, the compounds designed and
synthesized in this study were within the same broad structure
We acknowledge the assistance of David Vidal and Jordi
Mestres of Chemotargets and Albert Hinnen, Vadim Makarov,
and Stuart Cole for helpful discussions. Funding was provided
by the National Health and Medical Research Council
(NHMRC) Australia Fellowship 511105.
ABBREVIATIONS USED
■
pMIC, −log10(MIC expressed as molarity)
REFERENCES
■
(1) WHO. Towards Universal Access To Diagnosis and Treatment
of Multidrug-Resistant and Extensively Drug-Resistant Tuberculosis by
2015. http://www.who.int/tb/publications/2011/mdr_report_2011/
en/index.html (accessed May 16, 2012).
Figure 3. Multidimensional scaling (MDS) plot of the structural
similarity/diversity of analogues of 2, using 2D chemical fingerprints
(ECFP_6) and Tanimoto distances to calculate the diversity between
compounds. Original BTZ compounds are shown as diamonds and the
newly designed compounds in this study as circles. The MIC activity is
indicated by the color of the symbols, showing compounds with low
MIC activity (MIC > 1 μg/mL) in dark red, medium MIC activity (1
< MIC < 0.01 μg/mL) in orange, and high MIC activity (MIC < 0.01
μg/mL) in green.
(2) (a) Ma, Z.; Lienhardt, C.; McIlleron, H.; Nunn, A. J.; Wang, X.
Global tuberculosis drug development pipeline: the need and the
reality. Lancet 2010, 375 (9731), 2100−2109. (b) Villemagne, B.;
́
Crauste, C.; Flipo, M.; Baulard, A. R.; Deprez, B.; Willand, N.
Tuberculosis: the drug development pipeline at a glance. Eur. J. Med.
Chem. 2012, 51, 1−16.
(3) Makarov, V.; Manina, G.; Mikusova, K.; Mollmann, U.; Ryabova,
O.; Saint-Joanis, B.; Dhar, N.; Pasca, M. R.; Buroni, S.; Lucarelli, A. P.;
Milano, A.; De Rossi, E.; Belanova, M.; Bobovska, A.; Dianiskova, P.;
Kordulakova, J.; Sala, C.; Fullam, E.; Schneider, P.; McKinney, J. D.;
Brodin, P.; Christophe, T.; Waddell, S.; Butcher, P.; Albrethsen, J.;
Rosenkrands, I.; Brosch, R.; Nandi, V.; Bharath, S.; Gaonkar, S.;
Shandil, R. K.; Balasubramanian, V.; Balganesh, T.; Tyagi, S.; Grosset,
J.; Riccardi, G.; Cole, S. T. Benzothiazinones kill Mycobacterium
tuberculosis by blocking arabinan synthesis. Science 2009, 324 (5928),
801−804.
space as the original BTZ compounds. While the ligand-based
in silico design is biased by the training set of BTZ compounds,
the selection of compounds for synthesis was targeted for
maximum structural diversity.
In the MDS plot the compounds fall into three structural
clusters: (i) linear amines, (ii) cyclic amines (piperidines and
piperazines), and (iii) the closely related piperidinespiroketals.
7943
dx.doi.org/10.1021/jm3008882 | J. Med. Chem. 2012, 55, 7940−7944

Journal of Medicinal Chemistry
Brief Article
(4) Makarov, A. V.; Cole, S. T.; Mollmann, U. Benzothiazinone
Derivatives and Their Use as Antibacterial Agents. U.S. Patent Appl.
US 2009/0239851 A1, 2009.
(5) (a) Mikusova, K.; Huang, H.; Yagi, T.; Holsters, M.; Vereecke,
D.; D’Haeze, W.; Scherman, M. S.; Brennan, P. J.; McNeil, M. R.;
Crick, D. C. Decaprenylphosphoryl arabinofuranose, the donor of the
D-arabinofuranosyl residues of mycobacterial arabinan, is formed via a
two-step epimerization of decaprenylphosphoryl ribose. J. Bacteriol.
2005, 187 (23), 8020−8025. (b) Wolucka, B. A. Biosynthesis of D-
arabinose in mycobacteriaa novel bacterial pathway with implica-
tions for antimycobacterial therapy. FEBS J. 2008, 275 (11), 2691−
2711.
(6) (a) de Jesus Lopes Ribeiro, A. L.; Degiacomi, G.; Ewann, F.;
Buroni, S.; Incandela, M. L.; Chiarelli, L. R.; Mori, G.; Kim, J.;
Contreras-Dominguez, M.; Park, Y.-S.; Han, S.-J.; Brodin, P.; Valentini,
G.; Rizzi, M.; Riccardi, G.; Pasca, M. R. Analogous mechanisms of
resistance to benzothiazinones and dinitrobenzamides in Mycobacte-
rium smegmatis. PLoS One 2011, 6 (11), e26675. (b) Trefzer, C.;
Rengifo-Gonzalez, M.; Hinner, M. J.; Schneider, P.; Makarov, V.; Cole,
S. T.; Johnsson, K. Benzothiazinones: prodrugs that covalently modify
the decaprenylphosphoryl-beta-D-ribose 2′-epimerase DprE1 of
Mycobacterium tuberculosis. J. Am. Chem. Soc. 2010, 132 (39),
13663−13665.
(7) Tetko, I. V.; Tanchuk, V. Y.; Kasheva, T. N.; Villa, A. E. P.
Estimation of aqueous solubility of chemical compounds using E-state
indices. J. Chem. Inf. Comput. Sci. 2001, 41 (6), 1488−1493.
(8) Makarov, V. A. Process for the Preparation of 2-Amino-
Substituted 1,3-Benzothiazine-4-ones. WO 2011/132070 A1, 2011.
(9) Tsunoda, T.; Suzuki, M.; Noyori, R. A facile procedure for
acetalization under aprotic conditions. Tetrahedron Lett. 1980, 21 (14),
1357−1358.
(10) Alsenz, J.; Kansy, M. High throughput solubility measurement in
drug discovery and development. Adv. Drug Delivery Rev. 2007, 59 (7),
546−567.
(11) Hoelke, B.; Gieringer, S.; Arlt, M.; Saal, C. Comparison of
nephelometric, UV-spectroscopic, and HPLC methods for high-
throughput determination of aqueous drug solubility in microtiter
plates. Anal. Chem. 2009, 81 (8), 3165−3172.
(12) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubility and
permeability in drug discovery and development settings. Adv. Drug
Delivery Rev. 1997, 23 (1−3), 3−25.
(13) Bevan, C. D.; Lloyd, R. S. A high-throughput screening method
for the determination of aqueous drug solubility using laser
nephelometry in microtiter plates. Anal. Chem. 2000, 72 (8), 1781−
1787.
7944
dx.doi.org/10.1021/jm3008882 | J. Med. Chem. 2012, 55, 7940−7944