Novel inhibitors of DNA gyrase: 3D structure based biased needle screening, hit validation by biophysical methods, and 3D guided optimization. A promising alternative to random screening

Article abstract of DOI:10.1021/jm000017s

Random screening provided no suitable lead structures in a search for novel inhibitors of the bacterial enzyme DNA gyrase. Therefore, an alternative approach had to be developed. Relying on the detailed 3D structural information of the targeted ATP bindin

Full text of DOI:10.1021/jm000017s

2664
J . Med. Chem. 2000, 43, 2664-2674
Novel In h ibitor s of DNA Gyr a se: 3D Str u ctu r e Ba sed Bia sed Need le Scr een in g,
Hit Va lid a tion by Biop h ysica l Meth od s, a n d 3D Gu id ed Op tim iza tion . A
P r om isin g Alter n a tive to Ra n d om Scr een in g
Hans-J oachim Boehm, Markus Boehringer,* Daniel Bur, Hans Gmuender, Walter Huber, Werner Klaus,
Dirk Kostrewa, Holger Kuehne, Thomas Luebbers, Nathalie Meunier-Keller, and Francis Mueller
Preclinical Research, Pharmaceuticals Division, F. Hoffmann-La Roche Ltd., CH-4070 Basel, Switzerland
Received J anuary 18, 2000
Random screening provided no suitable lead structures in a search for novel inhibitors of the
bacterial enzyme DNA gyrase. Therefore, an alternative approach had to be developed. Relying
on the detailed 3D structural information of the targeted ATP binding site, our approach
combines as key techniques (1) an in silico screening for potential low molecular weight
inhibitors, (2) a biased high throughput DNA gyrase screen, (3) validation of the screening
hits by biophysical methods, and (4) a 3D guided optimization process. When the in silico
screening was performed, the initial data set containing 350 000 compounds could be reduced
to 3000 molecules. Testing these 3000 selected compounds in the DNA gyrase assay provided
150 hits clustered in 14 classes. Seven classes could be validated as true, novel DNA gyrase
inhibitors that act by binding to the ATP binding site located on subunit B: phenols, 2-amino-
triazines, 4-amino-pyrimidines, 2-amino-pyrimidines, pyrrolopyrimidines, indazoles, and 2-hy-
droxymethyl-indoles. The 3D guided optimization provided highly potent DNA gyrase inhibitors,
e.g., the 3,4-disubstituted indazole 23 being a 10 times more potent DNA gyrase inhibitor than
novobiocin (3).
In tr od u ction
The identification of an appropriate lead structure is
a vital step for any pharmaceutical research project.
Usually, lead structures are identified by random
screening of hundreds of thousands of compounds.
However, in some cases no suitable leads are found by
this approach. Without lead structures research projects
usually have to be discontinued even if they aim to cover
urgent medical needs with outstanding market poten-
tials.
Seeking novel inhibitors of DNA gyrase (E.C. 5.99.1.3),
we have been facing a similar situation. DNA gyrase is
a well-established antibacterial target.1,2 It is an es-
sential, prokaryotic type II topoisomerase with no direct
mammalian counterpart. It is involved in the vital
processes of DNA replication, transcription, and recom-
bination. DNA gyrase catalyzes the ATP-dependent
introduction of negative supercoils into bacterial DNA
as well as the decatenation and unknotting of DNA. The
enzyme consists of two subunits, A and B, of molecular
mass 97 and 90 kDa, respectively, with the active
enzyme being an A2B2 complex. The A subunit of DNA
gyrase is involved in DNA breakage and reunion while
the B subunit catalyzes the hydrolysis of ATP. DNA
gyrase is inhibited by quinolones, coumarins, and cy-
clothialidines (Figure 1), which, however, all have their
own limitations. Quinolones, e.g., ciprofloxacin (1),
which inhibit the DNA breakage-reunion cycle by
binding to the subunit A and by blocking the gyrase-
DNA complex, are successfully used as broad spectrum
F igu r e 1. Known inhibitors of DNA gyrase: ciprofloxacin (1),
cyclothialidine Ro 09-1437/000 (2), and novobiocin (3).
antibiotics in the clinic. Unfortunately, resistance to
quinolones has already emerged,3,4 and the fear of
iuvenile arthropathy prevents their use in pediatrics.5
The two other known classes of DNA gyrase inhibitors,
cyclothialidines,6 e.g., Ro 09-1437/000 (2), and cou-
marins, e.g., novobiocin (3), bind to the ATP recognition
* Address correspondence to Markus Boehringer, F. Hoffmann-La
Roche Ltd., Pharmaceuticals Division, Preclinical Research, PRBC,
Bldg. 15/136, CH-4070 Basel, Switzerland. Tel: ++41 61 688 8507.
Fax: ++41 61 688 6459. E-mail: markus.boehringer@roche.com.
10.1021/jm000017s CCC: $19.00 © 2000 American Chemical Society
Published on Web 06/23/2000

Novel Inhibitors of DNA Gyrase
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14 2665
F igu r e 2. (a) Compound A cannot bind to the active site due
to sterical interference of an inappropriate side chain. (b)
Compound B lacking the inappropriate side chain can bind to
the active site. (c) Four needle type compounds.
site located in the subunit B.7,8 This target site may offer
advantages such as a lack of cross-resistance with
quinolones. Novobiocin (3) was clinically used against
Staphylococcus aureus, including multi-resistant S.
aureus (MRSA). However, it suffers from toxicity and a
rapidly developing resistance.9,10 As demonstrated by
the cyclothialidines, this type of resistance can be
overcome.11 The limitations of the cyclothialidines are
their insufficient in vivo activities due to a class specific
rapid and extensive glucuronidation of the essential
phenol moiety.12
To overcome the limitations of the known DNA gyrase
inhibitors, we have been searching for novel structures.
However, beside the known inhibitors, quinolones, cou-
marins, and cyclothialidines, random screening has not
provided other suitable hits. Therefore, we have been
developing a new rational approach to generate lead
structures by using the detailed 3D structural informa-
tion of the ATP binding site located on subunit B. This
approach combines as key techniques (1) an in silico
screen for potential low molecular weight inhibitors, (2)
a robust and sensitive biased high throughput DNA
gyrase screen based on the results of the in silico screen,
(3) validation of the screening hits relying mainly on
biophysical methods, and (4) a 3D guided optimization
process.
F igu r e 3. The common hydrogen bond network between three
different ligands and DNA gyrase, including critical water
molecules (Wat), determined by X-ray analysis. (a) The
phenolic moiety of cyclothialidine forms hydrogen bonds with
Asp73, Wat45, and Wat46. SAR of cyclothialidine derivatives
revealed that the hydroxyl group hydrogen bonding to Wat46
can be replaced, e.g., by MeO, with gain in binding affinity,
indicating that this H-bond is not essential. (b) Hydrogen
bonds formed by the carbamate moiety of novobiocin in the
DNA gyrase binding pocket. (c) Hydrogen bonds between the
adenine part of ADPNP and the DNA gyrase binding pocket.
availibility problems. (3) The needle and its derivatives
are most likely structures relatively easy and inexpen-
sive to be synthesized compared to complex natural
products.
In Silico Need le Scr een in g
Beside profound SAR for inhibition of DNA gyrase by
cyclothialidine derivatives,13,14 detailed three-dimen-
sional structural information was available about the
ATP binding site of DNA gyrase from the Gram negative
Escherichia coli as well as from the Gram positive S.
aureus. At that time we had access to the X-ray
structures of the N-terminal 43 kDa or 24 kDa frag-
ment, respectively, of subunit B of DNA gyrase from E.
coli complexed with the substrate analogue ADPNP8 as
well as complexed with the two structurally different
inhibitors cyclothialidine 2 and novobiocin (3).7 Al-
though these ligands bind to the same enzyme pocket,
they occupy common as well as very different sub-
regions. In the inner part of the pocket they all share a
common binding motive: each of them donates a
hydrogen bond to an aspartic acid side chain (Asp73)15
and accepts a hydrogen bond from a conserved water
molecule (Wat45) (Figure 3). We reasoned that a novel
inhibitor should have the ability to form these two key
hydrogen bonds. In addition, it should contain a lipo-
philic part satisfying the steric and lipophilic require-
ments of the enzyme pocket. This pharmacophore
hypothesis is schematically summarized in Figure 4.
A 3D database screening was performed to identify
molecules that could fulfill our pharmacophore hypoth-
esis. The programs LUDI16,17 and CATALYST18 were
employed to search the Available Chemicals Directory
(ACD) and a part of the Roche compound inventory
(RCI), i.e., a combined data set that contained about
Con cep t of Need le Scr een in g
There are definitely various reasons why a compound
could be inactive in an enzyme assay. A sterically or
electronically inappropriate side chain, for example,
might be the reason for converting an active core
structure into an inactive compound (Figure 2). As a
consequence of such considerations we became inter-
ested in screening for low molecular weight inhibitors,
i.e., for inhibitors that are reduced to the minimal
structural elements while still fitting perfectly into the
active site and fulfilling the essential binding require-
ments to exhibit a maximal biological effect. We call
these low molecular weight inhibitors (MW < 300)
needles since they should be able to penetrate into deep
and narrow channels and subpockets of active sites like
a fine, sharp needle probing the surface of an active site
(Figure 2).
The attractiveness of the needle screening lies in the
small and compact structure of the needle: (1) The
needle is devoid of any unnecessary structural elements
that, for example, might result in toxicity or metabolic
instability. (2) A later introduction of tailored side
chains during a following optimization process does not
per se lead to oversized molecules having, e.g., bio-

2666 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14
Boehm et al.
F igu r e 4. Initial binding hypothesis used for in silico screen-
ing by LUDI and Catalyst.
F igu r e 5. Typical needle structures obtained from the LUDI
search.
350 000 compounds. LUDI was used to dock small
needle type molecules into the binding site. The 3D
structures of needles were generated by the computer
program CORINA.19 Only those molecules were ac-
cepted as valid hits that in silico form both critical
hydrogen bonds. Typical needle structures obtained
from the LUDI search are shown in Figure 5. Not only
these depicted unsubstituted needles were selected but
also appropriately substituted analogues with a MW <
300. The LUDI search resulted in a list of about 200
molecules that were tested for their DNA gyrase inhibi-
tory activity. The CATALYST search required the
precise topological definition of essential ligand enzyme
interactions, i.e., H-bond donors/ acceptors, van der-
Waals interactions, and excluded volumes. The precise
positions and types of these key interactions were
defined in the cyclothialidine-DNA gyrase X-ray struc-
ture using our molecular modeling program MOLOC.20
An RCI subdatabase containing multiple conformations
of each molecular entity was searched. Molecules with
a molecular weight of >400 were excluded. Clustering
using Tanimoto as implemented in the Daylight sys-
tem21 and manual selection of representatives led to a
total of 400 potential DNA gyrase inhibitors. In addition,
several closely related analogues of the hits from the
two 3D database searches were included in the list of
compounds to be tested.
F igu r e 6. Selected hits of the biased needle screening and
their maximal noneffective concentration (MNEC) determined
in the supercoiling assay. Bold: proposed needle. Dashed
bonds: postulated H-bonds to Asp73 and water Wat45. In the
case of the hydroxymethyl-indole 4i it was assumed that the
hydroxymethyl group displaces Wat45.
the assay was set up in a way that allowed detecting
not only highly potent but also weak inhibitors, i.e., that
allowed testing compounds even in high concentrations.
Instead of a supercoiling assay22 usually used to test
DNA gyrase inhibitory activity and used to run the
above-mentioned random screening, a coupled, spectro-
photometric ATPase assay23 was employed containing
the enzymes DNA gyrase, pyruvate kinase, and lactate
dehydrogenase. Due to a higher tolerance of DMSO in
the ATPase assay (e2.5%; maximal DMSO concentra-
tion in the supercoiling assay: 0.3%) compounds could
be assayed in concentrations up to 0.5 mM. ATP
hydrolysis was determined monitoring the disappear-
ance of NADH. The assay was run in the 96 well plate
format.
Furthermore, kinase inhibitors known from literature
were docked into the binding site of DNA gyrase.
Promising structures, i.e., those that might be able to
fulfill the pharmacophore hypothesis, were included in
the DNA gyrase assay.
Relying on the results of the in silico screening, just
600 compounds were tested initially. Then, close ana-
logues of the first hits were assayed too. Overall, 3000
compounds were screened in the biased needle screen
providing 150 hits. Most of them clustered in 14 classes.
As expected, all hits were only weak inhibitors of DNA
gyrase. Their maximal noneffective concentration (MNEC)
was in the range of 5-64 µg/mL, i.e., 2 to 3 orders of
magnitude higher than the MNEC of novobiocin (3;
MNEC ) 0.25 µg/mL) or cyclothialidine (2; MNEC )
0.05 µg/mL). Representative hits are depicted in Figure
6.
Bia sed Need le Scr een in g for DNA Gyr a se
In h ibitor s
Only weak DNA gyrase inhibitors were expected to
be found among the hits from in silico needle screening,
since most of them have already been tested in the
random screening performed a few years ago leading
to the discovery of cyclothialidine only. Consequently,

Novel Inhibitors of DNA Gyrase
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14 2667
Ta ble 1. Fourteen Needle Classes of ATPase Assay Hits, Their
Validation by Supercoiling Assay (SC), Analytical
Ultracentrifugation (AUC), Surface Plasmon Resonance (SPR),
NMR Spectroscopy (NMR), and X-ray Analysis (XR), and Their
Selection for Further Optimization (SEL)a
Hit Va lid a tion
A strong emphasis was put on early validation of the
screening hits and elimination of false positives due to
the following considerations: (1) The hit list of any
screening contains most likely a number of false posi-
tives. False positives may arise because of various
reasons, e.g., unspecific binding or induction of unspe-
cific protein aggregation. (2) Our screening assay was
a coupled assay. Thus, a hit was not per se a DNA
gyrase inhibitor. (3) In our search for novel inhibitors
of DNA gyrase, we were focusing on compounds that
act by binding to the ATP binding pocket located on
subunit B. However, neither the used ATPase assay nor
the supercoiling assay distinguished between com-
pounds interfering specifically with the ATP binding site
and others.
Our hit validation protocol consisted of (1) a super-
coiling assay, (2) elaboration of an initial SAR with
respect to the 3D structure of the active site, (3)
analytical ultracentrifugation experiments, (4) surface
plasmon resonance binding studies, (5) heteronuclear
1H/15N correlation NMR spectroscopy, and (6) X-ray
analysis.
All ATPase hits, i.e., compounds with a MNEC e 64
µg/mL, were subjected to the supercoiling assay in order
to identify and eliminate those compounds from the hit
list that acted in the ATPase assay by interfering with
pyruvate kinase or lactate dehydrogenase. DNA gyrase
inhibition (i.e., MNEC e 32 µg/mL) was observed for
compounds of all classes except for pyrazoles, hydrox-
amic acids, and 2-aminothiazoles (Table 1). Using the
3D structure of the ATP binding site, preliminary SARs
were elaborated based on supercoiling data of the
initially screened compounds as well as of some newly
synthesized analogues. A flat SAR (i.e., no change in
biological activity despite large changes in chemical
structure), as, for example, observed in the case of
catechols, was interpreted as a first indication of
unspecific binding. The question of specific binding was
addressed in more detail by several biophysical binding
studies. In all these binding studies the 24 kDa N-
terminal fragment of DNA gyrase subunit
B of
S. aureus11 which comprised the ATP binding site
was used as a model of the whole subunit B and as
a model of the active DNA gyrase, i.e., the A2B2
tetramer.
Bin d in g Stu d ies by An a lytica l Ultr a cen tr ifu ga -
tion (AUC) an d Su r face P lasm on Reson an ce (SP R).
AUC and SPR were both used as high throughput
methods to identify compounds that bound to the 24
kDa fragment. Since the ATP binding pocket is located
on the 24 kDa fragment, only those compounds that
bound to this fragment could potentially bind to the ATP
pocket. In the case of AUC, binding was detected in a
sedimentation equilibrium experiment, i.e., the sample
was centrifuged at the velocity required for the sedi-
mentation-diffusion equilibrium of the 24 kDa frag-
ment. In a typical experiment, compounds were tested
at 100 µM in the presence of 10 µM protein. At
equilibrium, the concentration profile of the potential
ligand was determined by absorption measurement at
different points in the sample cell. Binding of the small
ligand to the heavier 24 kDa fragment results in a
characteristic concentration profile.
a
Dashed bonds: postulated H-bonds to Asp73 and water Wat45.
In the case of the hydroxymethyl-indoles it was assumed that the
hydroxymethyl group displaces Wat45. +: inhibition/binding
observed. (+): very weak inhibition/binding observed. -: no
inhibition/binding observed. nd: not determined. X: class selected
for further optimization.
In the case of surface plasmon resonance binding
experiments, the 24 kDa fragment of DNA gyrase B was
immobilized on a carboxymethyldextran-modified gold
surface in concentrations of 700-8000 pg/mm2. The
compounds to be tested were exposed to these sensors
in micromolar concentrations, i.e., 2.5-250 µM. Re-
sponses that were qualitatively similar to responses for
3 were interpreted as indication of binding. Under the
experimental conditions used for AUC or SPR, respec-

2668 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14
Boehm et al.
1
F igu r e 7. Shift differences of H/15N cross-peaks of backbone amide NHs in ppm observed by heteronuclear 1H/15N correlation
NMR spectroscopy upon addition of potential ligands.
tively, no information on the precise site of binding could
be derived. Overall, consistent results were obtained by
AUC and SPR (Table 1).
1
Sp in -Sp in J -Cou p led Heter on u clea r H/15N Cor -
r ela tion NMR Sp ectr oscop y (NMR) was used as a
highly sensitive and specific method to determine if
binding of a ligand to the 24 kDa fragment of DNA
gyrase occurred andsmore importantlysif binding oc-
curred at the ATP binding site. This method relies on
the fact that interactions of a small ligand with a protein
manifest themselves in changes of the positions of
resonance lines of the macromolecule.24,25 The effects
F igu r e 8. X-ray structure of 5 complexed to the 24 kDa
of binding were easily observed in the spectrum of the
15N isotopically labeled 24 kDa fragment. The elabo-
rated assignment of about 80% of all NMR signals26
allowed the sequence specific identification of residues
with a shifted N-H signal upon binding of the ligand
and subsequently the delineation of the binding epitope
within the structure of the 24 kDa fragment. In the case
of 5, the most pronounced shift differences clustered in
the three regions around Glu50, Ile78, and His135
(Figure 7). Inspection of the 3D structure7,8 revealed
that these amino acids cluster all exclusively around the
ATP binding cavity (Figure 8). Therefore, binding of 5
to the ATP recognition site was postulated. A similar
pattern of shift differences was observed for novobiocin
(3), known to bind into the ATP pocket, verifying the
binding hypothesis of 5. The final proof was achieved
by X-ray structure analysis of the 24 kDa fragment
complexed with 5.27 Similar patterns of shift differences
as for 3 and 5 were observed for many other compounds
tested, e.g., 4f, indicating binding to the ATP site (Table
1). Due to the different structures of the tested com-
pounds, the extent of the induced shifts was of course
not always as pronounced as in the case of 5, even for
compounds binding with a similar affinity. As expected,
no peak shifts were observed in the case of the quinolone
1. For some tested ATPase assay hits no binding or even
binding to other sites was detected. For example the
fragment27 illustrating shift differences of 1H/15N cross-peaks
of protein backbone NHs which are observed by heteronuclear
1H/15N correlation NMR spectroscopy upon addition of 5.
Depicted is the protein backbone of the 24 kDa fragment
(white), the indazole 5 (green), the water molecule Wat45 (red),
and all protons of the backbone amide NHs which were shifted
>0.16 ppm (white balls). The extent of the induced cross-peak
shifts is displayed proportionally by the sizes of the Hs, i.e.,
white balls. The shifted cross-peaks cluster exclusively around
the ATP binding pocket. The picture was prepared by MOLOC
and WebLab ViewerPro.
catechole 4b did obviously not bind to the ATP binding
site although it exhibited a more potent DNA gyrase
inhibitory activity in the supercoiling assay than 4f or
5 (MNEC ) 63 µg/mL). Similar to catecholes, no binding
to the ATP site could be observed for 2-aminothiazoles,
thiophenylamides, and carbamates despite their inhibi-
tory activity in the supercoiling assay.
X-r a y An a lysis was used to verify binding of a
compound to the ATP binding site and as a tool to guide
the following optimization of these ligands. Instead of
the 24 kDa fragment of DNA gyrase subunit B from S.
aureus used for AUC, SPR, and NMR experiments, a
loop deletion mutant of it was used which behaved more
favorable during crystallization experiments.27,28 This
mutant was obtained by deletion mutagenesis of resi-
dues 97-119 linking directly residues Thr96 and Ser120.

Novel Inhibitors of DNA Gyrase
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14 2669
Ta ble 2. Dissociation Constants KD of Needles Determined by
Heteronuclear 1H/15N Correlation NMR Spectroscopy
F igu r e 9.
fragment. The results are summarized in Table 2.
Clearly, the highest affinity was observed for 12a , the
lowest for 8a and 10a .
With respect to the X-ray data of the ATP binding
site,27 the relative magnitude of the KD values could be
rationalized by the following considerations. (1) The
indazole 12a fitted much better into the ATP binding
cavity whereas the smaller 7a -10a led to voids, giving
rise to less favorable van der Waals interactions. (2) The
water molecule Wat15 (Figure 3) was most likely
displaced by 12a (as observed in the X-ray structure of
5), resulting in a favorable entropy contribution. In the
case of the smaller compounds 7a -10a , Wat15 was very
likely not displaced (as observed in the X-ray structure
of 4d ). (3) It was assumed that Wat15 was forming a
favorable hydrogen bond with N3 of 7a or 9a , respec-
tively (as observed in the X-ray structure of 4d ). For
8a and 10a such H-bonds were not possible without
breaking the H-bond network to Wat45 and Asp73.
F igu r e 10. Different needle classes containing identical side
chains.
According to X-ray data the loop deletion mutant
comprised the ATP recognition site as in the unmodified
24 kDa fragment and as in the published DNA gyrase
X-ray structures7,8 which all describe complexes of E.
coli DNA gyrase fragments. Unspecific binding of a
ligand to a protein can lead to unspecific protein
aggregation and, consequently, to protein sedimentation
during crystallization experiments. Such unfavorable
compounds were identified during AUC experiments
and excluded from crystallization trials, raising its
success rate. Using this prescreening procedure and the
loop deletion mutant protein, numerous X-ray struc-
tures were solved for a range of very different ligand
structures. Already during the hit validation phase the
X-ray structures of the complexes with the indazole 5,
the 2-amino-triazine 4d , and the pyrrolopyrimidine 6
were determined.27
3D Gu id ed Op tim iza tion P r ocess
The preliminary SARs of the validated needles, the
detailed SARs for cyclothialidines13,14 and coumarins,32
andsmost importantlysthe 3D structural knowledge of
the active site derived from the X-ray structures of the
24 kDa fragment complexed with ADPNP, 2, 3, 5 or 4d
or 6,7,8,27 respectively, were employed to guide the
optimization of the validated hits. Since the principal
considerations for the optimization were identical for
all validated needle classes, we will focus in the follow-
ing on indazoles only.
As a first criteria we postulated that a DNA gyrase
inhibitor must not only form a hydrogen bond network
with Asp73 and the very firmly bound water molecule
Wat45, but it must also contain a side chain stacking
on the Glu50-Arg76 salt bridge (Figure 11). This hy-
pothesis allowed the explanation of the improved DNA
gyrase inhibitory activity of 4h compared to 12a (Table
3). With a second generation of indazoles it was at-
tempted to introduce additional structural elements that
were able to form H-bonds to Arg136.33 This second
postulate was realized by compounds such as 13. The
indazole 13 was obtained by conversion of 1434 with
4-methyl-7-thiocoumarin followed by treatment with
TFA (Scheme 1). In contrast to our expectation, not
more than just minor improvements of DNA gyrase
As a consequence of this hit validation process we
could focus in the subsequent hit exploration and
optimization on phenoles, 2-amino-triazines, 4-amino-
pyrimidines, 2-amino-pyrimidines, pyrrolopyrimidines,
indazoles, and 2-hydroxymethylindoles (Table 1).
Ra n k in g of Va lid a ted Need les
Although the hit validation process allowed reducing
the initial hit list from 14 structural classes to seven,
we felt a need to prioritize and rank the validated needle
classes. The classical way would be to synthesize and
characterize close analogues, i.e., compounds belonging
to different needle classes but containing identical side
chains, e.g., 7b29-31-12b (Figure 10). However, 12b
would hardly be comparable to 7b-11b since the
different substituent sites and exit vectors would result
in a different overall topology. Comparing just the
unsubstituted needles 7a -12a in the ATPase assay or
supercoiling assay was not possible, since they were all
inactive in these assays (MNEC > 250 µg/mL). However,
the high sensivity of heteronuclear 1H/15N correlation
NMR spectroscopy allowed determining the dissociation
constant KD of the unsubstituted needles to the 24 kDa

2670 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14
Boehm et al.
Sch em e 2. Synthesis of Indazoles 22, 23 and 27
F igu r e 11. Refined binding model: the additional sites of
expected, favorable interactions between a ligand and the ATP
binding pocket are marked in the X-ray structure of the 24
kDa fragment complexed with 5. With the exception of the
Asp73 side chain the protein is presented by its surface only.
Dashed lines: observed H-bonds between inhibitor, protein,
and conserved water Wat45. The picture was prepared by
MOLOC and WebLab ViewerPro.
Ta ble 3. DNA Gyrase Inhibition by Indazoles
a
Reagents and conditions: (a) N2H4‚H2O, HOCH2CH2OH, 2 h,
160 °C; (b) (1) TBDMSCl, imidazole, DMF, 12 h, rt; (2) (BOC)2O,
Et3N, DMAP, MeCN, 3 h, rt; (3) NBS, (PhCO2)2, CCl4, 4.5 h, 78
°C; (c) 4-methyl-7-thiocoumarin, Et3N, CH2Cl2, 4 h, rt; (d) (1) BnBr
or p-tBu-BnBr, KF, DMF, 4 h, rt; (2) TFA, CH2Cl2, 4 h, rt; (e)
3-mercapto-benzoic acid methyl ester (24), Et3N, CH2Cl2, 4 h, rt;
(f) (1) BnBr, KF, DMF, 2 h, rt; (2) LiOH, THF, MeOH, H2O, 5 h,
rt.
ing a second side chain. This series of 3,4-disubstituted
indazoles was synthesized according to the procedure
outlined in Scheme 2. A pronounced improvement of
inhibitory activity could be observed for 22. Optimizing
the DNA gyrase inhibitory activity of 22 led to 23
containing an additional tert-butyl group. It was about
10 times more active than 3 in the DNA gyrase assay
while still being structurally much less complex than
3. The final proof of our binding hypotheses was
achieved by the X-ray structure of 27 complexed to the
24 kDa fragment (Figure 12). In this structure, (1) the
indazole scaffold formed the postulated H-bond network
with Asp73 and Wat45, (2) the mercapto-benzoic acid
side chain stacked on the Glu50-Arg76 salt bridge and
was H-bonded to Arg136, and (3) the benzyloxy side
chain formed van der Waals interactions with the
lipophilic area Ile78-Pro79-Ile94.
1
MNECs are determined in the supercoiling assay, except for
12a .
Sch em e 1. Synthesis of 3-Substituted Indazole 13a
a
Reagents and conditions: (a) (1) 4-methyl-7-thiocoumarin,
Et3N, CH2Cl2, 2 h, rt; (2) TFA, CH2Cl2, 3 h, rt.
Con clu sion s
inhibition were observed within this series. To exploit
the potential of van der Waals interactions between a
ligand and a lipophilic area formed by Ile78, Pro79, and
Ile94, a third series of indazoles was prepared contain-
Our novel approach combining (1) an in silico needle
screen, (2) a biased DNA gyrase needle screen, (3) a hit
validation protocol, and (4) a 3D guided optimization
process has led to a series of potent, novel inhibitors of

Novel Inhibitors of DNA Gyrase
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14 2671
Su r fa ce P la sm on Reson a n ce Bin d in g Exp er im en ts:36,37
Surface plasmon resonance studies were performed on a
Biacore 2000 instrument. A CM5 sensor chip was used,
equipped with a carboxymethyldextran-modified gold surface.
The 24 kDa fragment of DNA gyrase B was immobilized in
three channels of the sensor chip using the fourth channel as
a reference. The 24 kDa fragment was immobilized according
to the procedure described by J ohnsson et al.38 The protein
solution needed for the immobilization procedure was freshly
prepared and contained 100 µg/mL 24 kDa fragment in 9 mM
acetate, 1 mM Hepes buffer at pH 4.9. In a typical immobiliza-
tion, different amounts of protein were bound on the surfaces
of the three flow channels, e.g., 700, 5524, and 7807 pg/mm2.
The amount of bound protein was controlled via different time
intervals of contact. Excess activated carboxyl functions on the
dextran layer were quenched after the immobilization by
contacting the sensor surface with ethanolamine. The sensor
surface was washed shortly (30 s) with an SDS solution (1%)
to dissolve protein aggregates on the surface formed during
the immobilization. The activity of the immobilized 24 kDa
fragment was controlled after the immobilization procedure
and between successive binding experiments by contacting the
surface with a solution of 3. The association and dissociation
rate constants (kd, ka) and the equilibrium sensor response (Req)
were taken as quality criteria. Typical values of kd and ka for
3 were 3-4 × 10-3 s-1 and 1-3 × 106 M-1 s-1. The equilibrium
sensor response depended on the amount of protein im-
mobilized and the concentration of 3 in solution. The binding
capacity of the immobilized protein was in the order of 0.1-
0.15 depending on the amount of protein immobilized.
F igu r e 12. X-ray structure of the 24 kDa fragment complexed
with 27. With the exception of the Asp73 and Arg136 side
chains, the protein is presented by its surface only. Dashed
lines: observed H-bonds between inhibitor, protein, and
conserved water Wat45. The picture was prepared by MOLOC
and WebLab ViewerPro.
DNA gyrase. Running the in silico screening prior to
the enzyme assay allowed us to focus early on the most
promising candidates in our compound library, increas-
ing the hit rate and being highly time- and cost-efficient.
Besides the classical ways of validating screening hits
by biological experiments and preliminary SARs, bio-
physical methods were applied to eliminate rapidly and
unambiguously false positives. The high sensitivity of
NMR spectroscopy was used to elaborate a ranking of
the validated needles with respect to their binding
affinities. Such a ranking may have a decisive influence
on the prioritization of tasks and on the selection of the
most promising needle for the following lead optimiza-
tion processes. The optimization process itselfsconverting
the screening hits into potent lead structuresswas
strongly guided by the detailed knowledge of the active
site’s 3D structure, enabling us to focus the synthetic
efforts early on the most promising side chains of the
inhibitors. The lead optimization is exemplified by the
indazole series resulting in 23, a 10 times more potent
DNA gyrase inhibitor than novobiocin (3). Overall, our
novel approach represents an attractive alternative to
random screening.
During a time interval of 5 min the compounds to be tested
were exposed to the sensor surface as 2.5, 25, or 250 µM
solutions, respectively, in a Hepes buffer (10 mM, pH 7.5, 150
mM NaCl, 3 mM EDTA, 0.005% polysorbate 20 (v/v), 1%
DMSO).
Two decisive criteria were used to identify the compounds
binding to the 24 kDa fragment: (1) a positive sensor response
after the contact interval; (2) a ranking of the sensor responses
detected for the different channels that is comparable to the
ranking observed for controls, i.e., for 3. This ranking was
interpreted as indication for protein specificity of the binding.
NMR Sp ectr oscop y for Hit Va lid a tion . A typical two-
1
dimensional H-15N correlated NMR sample contained 7 mM
of compound to be tested, 0.7 mM 24 kDa fragment of DNA
gyrase subunit B, 75 mM NH4CO3, 0.02% NaN3, 2.5% DMSO
in 90% H2O, 10% D2O, pH 8.6. The spectra were recorded at
600 MHz on a Bruker DMX 600 NMR instrument at a
temperature of 300 K. Spectral parameters were 512 complex
points in F2 at a width of 9615 Hz (proton dimension) and 64
complex points in F1 at a width of 1766 Hz (nitrogen dimen-
sion). The radio frequency carriers were set at the water
resonance line (4.75 ppm) for 1H and at 117.0 ppm for 15N.
Normally 32 scans were acquired for each FID resulting in a
total measurement time of 66 min. Data were transferred to
a SGI Unix workstation and processed using the NIH software
package NMRPipe.39 Spectral interpretation was performed
with the aid of the XEASY40 program. Herein, the 2D spectrum
of a protein-ligand solution was compared to a reference
spectrum of the apo-enzyme that contained 2.5% DMSO only,
but no ligand. The secondary chemical shifts were calculated
from the 1H and 15N shift coordinates of the cross-peaks:
Exp er im en ta l Section
DNA Gyr a se Assa ys. Assay conditions of the ATPase and
supercoiling assay were described previously.22,23 Both assays
led to at least similar, if not identical MNECs.
24 k Da F r a gm en ts. The 24 kDa N-terminal fragments of
the DNA gyrase subunit B of S. aureus and the corresponding
loop deletion mutant were obtained as described previously.11,28
The 24 kDa fragment used for AUC, SPR, and NMR consisted
of the residues 15-223, including the mutation A18V. Except
the deleted residues 97-119 the loop deletion mutant used
for X-ray analysis consisted of the whole N terminus ending
with residue 223.
{[∆δ(1H)]2 + [∆δ(15N)0.3]2}1/2
.
An a lytica l Ultr a cen tr ifu ga tion . Ultracentrifugation ex-
periments were performed at 20 °C on a Beckman Optima XL-
A. A run contained up to 21 samples of 100 µL volume at a
ligand concentration yielding a ligand absorbance of g0.1 OD
at a wavelength preferably >280 nm. Typical ligand concen-
trations were 100 µM. The concentration of the 24 kDa
fragment of DNA gyrase B was adapted to the ligand concen-
tration in order to result a 10-fold ligand excess. The buffer
contained 75 mM NH4CO3, pH 8.6, 0.02% NaN3, 2.5% DMSO.
A ligand free sample was centrifuged in each run as a control.
Data were analyzed according to DISCREEQ.35 Optimal veloc-
ity was 18 000 rpm.
NMR Sp ectr oscop y for Ra n k in g of Need les. For each
1
compound the two-dimensional H-15N correlated NMR spec-
trum was recorded at 10 different concentrations ranging from
0 to 22 mM and ligand/protein ratios from 0 to ∼50, respec-
tively. The samples were prepared by adding small amounts
of a 200 mM ligand stock solution in DMSO to the protein
solution containing the 15N labeled 24 kDa fragment at a
concentration of 0.45 mM in 75 mM ammonium phosphate
buffer, pH 8.6. To account for shifts induced by DMSO a
titration was performed adding just DMSO to the protein
solution. The spectra were recorded and processed as described
for the hit validation. For the determination of dissociation

2672 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14
Boehm et al.
constants (KD), nonlinear least-squares fits of shifts versus
ligand concentration were performed using the public domain
program xmgr (version 3.01) according to the equation δ )
(A0X/(X + A1)) + A2, with δ equal to the change in chemical
shift (1H or 15N), X ) concentration of ligand, A0 ) maximum
change in chemical shift, A1 ) KD, and A2 ) (chemical shift)
offset.
mmol) in acetonitrile (3 mL) was added dropwise. The solution
was allowed to warm to room temperature stirred for 3 h.
Evaporation of the solvent and chromatography (hexane/
EtOAc 4:1) provided 18 (468 mg, 73%) as an oil. 1H NMR
(DMSO-d6): δ 7.63 (1H, d, J ) 7.8), 7.41 (1H, t, J ) 7.8), 6.75
(1H, d, J ) 7.8), 2.59 (3H, s), 1.62 (9H, s), 1.01 (9H, s), 0.34
(6H, s). MS (EI+): m/z 362 (M+).
X-r a y. Experimental details for the crystallization experi-
ments and X-ray analysis were described by Kostrewa et al.27
and Dale et al.28
3-Br om om eth yl-4-(ter t-bu tyl-d im eth yl-sila n yloxy)-in -
d a zole-1-ca r boxylic Acid ter t-Bu tyl Ester (19). To a solu-
tion of 18 (440 mg, 1.21 mmol) in CCl4 (5 mL) heated to reflux
was added a mixture of N-bromosuccinimide (215 mg, 1.21
mmol) and benzoylperoxide (29 mL, 0.12 mmol) in small
portions. The mixture was refluxed for another 4.5 h, then
cooled to room temperature, and filtered. After evaporation of
the solvent and chromatography (hexane/EtOAc 15:1), 19 was
obtained as an oil (385 mg, 72%). 1H NMR (DMSO-d6): δ 7.71
(1H, d, J ) 8.0), 7.53 (1H, t, J ) 8.0), 6.87 (1H, d, J ) 8.0),
4.97 (2H, s), 1.68 (9H, s), 1.08 (9H, s), 0.42 (6H, s). MS (EI+):
m/z 440 (M+).
4-(ter t-Bu tyl-dim eth yl-silan yloxy)-3-(4-m eth ylcou m ar in -
7-ylsu lfa n ylm et h yl)-in d a zole-1-ca r b oxylic Acid ter t-
Bu tyl Ester (20). A solution of 19 (380 mg, 0.86 mmol) and
4-methyl-7-thiocoumarin (182 mg, 0.95 mmol) in CH2Cl2 (5
mL) containing Et3N (132 µL, 0.95 mmol) was stirred 4 h at
room temperature. H2O (10 mL) was added, and the mixture
was extracted with EtOAc. The organic extracts were washed
with brine, dried (Na2SO4), and concentrated. Chromatography
(hexane/EtOAc 1:1) provided 20 (428 mg, 90%) as a white solid.
1H NMR (DMSO-d6): δ 7.66 (1H, d, J ) 8.0), 7.64 (1H, d, J )
8.0), 7.54 (1H, d, J ) 1.5), 7.47 (1H, t, J ) 8.0), 7.31 (1H, dd,
J 1 ) 8.0, J 2 ) 1.5), 6.81 (1H, d, J ) 8.0), 6.33 (1H, s), 4.71
(2H, s), 2.40 (3H, s), 1.60 (9H, s), 0.97 (9H, s), 0.35 (6H, s). MS
(EI+): m/z 552 (M+).
Ch em istr y. Gen er a l In for m a tion . Compounds 1, 3, 4b,
4e, 7a , 8a , 9a , 10a , 12a , 4-methyl-7-thiocoumarin, and 3-mer-
capto-benzoic acid were commercially available. The syntheses
of the compounds 2,13,14 4a ,41 4f,42 4h ,43 5,43 7b,30,31 11a ,4414,34
and 1645,46 were described already. Unless indicated otherwise,
commercially available chemicals and solvents were used
without further purification. Silica gel 60 (0.04-0.063 mm;
Merck) was used for flash chromatography. 1H NMR spectros-
copy was performed on a Bruker 250 MHz or 400 MHz
machine using the indicated solvent with TMS as internal
standard. Chemical shifts δ are given in ppm relative to TMS,
coupling constants J in Hz.
7-(1H -In d a zole -3-ylm e t h ylsu lfa n yl)-4-m e t h ylcou m -
a r in (13). To a solution of 14 (90 mg, 0.29 mmol) and 4-methyl-
7-thiocoumarin (61 mg, 0.32 mmol) in CH2Cl2 (3 mL) was
added Et3N (44 µL, 0.32 mmol). After the mixture was stirred
for 2 h at room temperature, H2O (10 mL) was added and the
mixture was extracted with EtOAc. The organic layer was
washed with brine and dried (Na2SO4) before the solvent was
evaporated. The obtained oil was dissolved in CH2Cl2 (4 mL),
and TFA (0.3 mL) was added. After 3 h at room temperature
the reaction mixture was neutralized with saturated NaHCO3
solution and extracted with EtOAc. The organic layer was once
again washed with brine and dried (Na2SO4), and the solvent
was evaporated. The residue was precipitated from CHCl3 and
hexane, providing 13 (71 mg, 76%) as a white powder. 1H NMR
(DMSO-d6): δ 12.94 (1H, s), 7.89 (1H, d, J ) 7.8), 7.63 (1H, d,
J ) 7.8), 7.55-7.45 (2H, m), 7.39-7.30 (2H, m), 7.12 (1H, t, J
) 7.8), 6.31 (1H, s), 4.75 (2H, s), 2.38 (3H, s). MS (EI+): m/z
322 (M+). HRMS calcd for C18H14N2O2S 322.0776, found
322.0773.
4-Ben zyloxy-3-(4-m eth ylcou m ar in -7-ylsu lfan ylm eth yl)-
in d a zole-1-ca r boxylic Acid ter t-Bu tyl Ester (21). KF (42
mg, 0.72 mmol) was added to a suspension of 20 (200 mg, 0.36
mmol) and benzyl bromide (52 µL, 0.43 mmol) in DMF (10 mL).
After the mixture was stirred for 4 h at room temperature,
H2O (20 mL) was added and the mixture was extracted with
EtOAc. The combined organic layers were washed with brine
and dried (Na2SO4), and the solvent was evaporated. The
residue was precipitated from CHCl3 and hexane yielding 21
3-Meth yl-1H-in d a zol-4-ol (16). A solution of 2,6-dihy-
droxyacetophenone (15; 50 g, 331 mmol) in ethyleneglycol (750
mL) was added dropwise to a solution of hydrazine hydrate
(33 g, 659 mmol) in ethyleneglycol (250 mL). The reaction
mixture was stirred for 20 min at room temperature, then
heated to 160 °C for 2 h. The solution was allowed to cool to
room temperature, and it was poured into H2O (2L). The pH
was adjusted to pH 6 by adding AcOH (25 mL), and the
mixture was extracted with EtOAc (4 × 1L). The combined
organic layers were washed with 5% Na2SO3, dried (Na2SO4),
and concentrated. Chromatography (toluene/acetonitrile 7:3)
1
(153 mg, 80%) as a white solid. H NMR (DMSO-d6): δ 7.67-
7.46 (7H, m), 7.33-7.26 (3H, m), 6.99 (1H, d, J ) 8.0), 6.34
(1H, s), 5.30 (2H, s), 4.68 (2H, s), 2.41 (3H, s), 1.62 (9H, s).
7-(4-Ben zyloxy-1H-in d a zol-3-ylm et h ylsu lfa n yl)-4-m e-
th ylcou m a r in (22). A solution of 21 (151 mg, 0.29 mmol) in
CH2Cl2 (4 mL) containing TFA (0.5 mL) was stirred for 4 h at
room temperature. The reaction mixture was neutralized with
saturated NaHCO3 solution and extracted with EtOAc. The
organic layer was washed with brine and dried (Na2SO4), and
the solvent was evaporated. The residue was precipitated from
CHCl3 and hexane, providing 22 (112 mg, 91%) as white
1
provided 16 (36.7 g, 68%) as a white solid. H NMR (DMSO-
1
d6): δ 12.34 (1H, s, br), 9.89 (1H, s), 7.03 (1H, t, J ) 7.8), 6.80
(1H, d, J ) 7.8), 6.32 (1H, d, J ) 7.8), 2.55 (3H, s). MS (EI+):
m/z 148 (M+).
powder. H NMR (DMSO-d6): δ 12.96 (1H, s), 7.66 (1H, d, J
) 8.0), 7.60-7.49 (3H, m), 7.36-7.24 (5H, m), 7.07 (1H, t, J )
8.0), 6.67 (1H, t, J ) 8.0), 6.35 (1H, s), 5.28 (2H, s), 4.72 (2H,
s), 2.43 (3H, s). MS (EI+): m/z 428 (M+). HRMS calcd for
4-(ter t-Bu t yl-d im et h yl-sila n yloxy)-3-m et h yl-1H -in d a -
zole (17). Imidazole (3.08 g, 45.20 mmol) and tert-butyldim-
ethyl-chlorosilane (1.57 g, 10.40 mmol) were added to a cooled
solution (0 °C) of 16 (1.34 g, 9.04 mmol) in DMF (100 mL).
The solution was stirred for 12 h at room temperature. Then
H2O (100 mL) was added, and the mixture was extracted with
EtOAc (3 × 50 mL). The organic extracts were washed with
H2O and brine and dried (Na2SO4). Evaporation of the solvent
and chromatography (hexane/EtOAc 2:1 to 1:1) provided 17
(2.10 g, 89%) as a white solid. 1H NMR (DMSO-d6): δ 12.5
(1H, s, br), 7.12 (1H, t, J ) 7.8), 6.98 (1H, d, J ) 7.8), 6.39
(1H, d, J ) 7.8), 2.57 (3H, s), 1.01 (9H, s), 0.31 (6H, s). MS
(EI+): m/z 262 (M+).
C
25H20N2O3S 428.1195, found 428.1194.
7-[4-(4-ter t-Bu t yl-b en zyloxy)-1H -in d a zol-3-ylm et h yl-
su lfa n yl]-4-m eth ylcou m a r in (23). KF (42 mg, 0.72 mmol)
was added to a suspension of 20 (200 mg, 0.36 mmol) and
4-tert-butylbenzyl bromide (80 µL, 0.43 mmol) in DMF (10 mL).
After being stirred for 4 h at room temperature, H2O (20 mL)
was added and the mixture was extracted with EtOAc. The
combined organic layers were washed with brine and dried
(Na2SO4), and the solvent was evaporated. The residue was
dissolved in CH2Cl2 (5 mL), and TFA (1 mL) was added. After
3 h at room temperature the reaction mixture was neutralized
with saturated NaHCO3 solution and extracted with EtOAc.
The organic layer was once again washed with brine and dried
(Na2SO4), and the solvent was evaporated. The residue was
precipitated from CHCl3 and hexane, providing 23 (158 mg,
4-(ter t-Bu tyl-d im eth yl-sila n yloxy)-3-m eth yl-in d a zole-
1-ca r boxylic Acid ter t-Bu tyl Ester (18). DMAP (43 mg, 0.35
mmol) was added to a solution of 17 (462 mg, 1.76 mmol) and
Et3N (270 µL, 1.94 mmol) in acetonitrile (3 mL). The mixture
was cooled to 0 °C, and a solution of (BOC)2O (461 mg, 2.11
1
90%) as a gray powder. H NMR (DMSO-d6): δ 12.98 (1H, s,
br), 7.69 (1H, d, J ) 8.0), 7.50 (1H, d, J ) 1.2), 7.40 (2H, d, J

Novel Inhibitors of DNA Gyrase
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14 2673
(4) Voss, A.; Milatovic, D.; Wallrauch-Schwarz, C.; Rosdahl, V. T.;
Braveny, I. Methicillin-Resistant Staphylococcus aureus in
Europe. Eur. J . Clin. Microbiol. Infect. Dis. 1994, 13, 50-55.
(5) Wolfson, J . S. Quinolone Antimicrobial Agents: Adverse Effects
and Bacterial Resistance. Eur. J . Clin. Microbiol. Infect. Dis.
1989, 8, 1080-1092.
(6) Nakada, N.; Shimada, H.; Hirata, T.; Aoki, Y.; Kamiyama, T.;
Watanabe, J .; Arisawa, M. Biological characterization of cy-
clothialidine, a new DNA gyrase inhibitor. Antimicrob. Agents
Chemother. 1993, 37, 2656-2661.
(7) Lewis, R. J .; Singh, O. M. P.; Smith, C. V.; Skarzynski, T.;
Maxwell, A.; Wonacott, A. J .; Wigley, D. B. The nature of
inhibition of DNA gyrase by the coumarins and the cyclothia-
lidines revealed by X-ray crystallography. EMBO J . 1996, 15,
1412-1420.
(8) Wigley, D. B.; Davies, G. J .; Dodson, E. J .; Maxwell, A.; Dodson,
G. Crystal structure of an N-terminal fragment of the DNA
gyrase B protein. Nature 1991, 351, 624-629.
(9) Kim, O. K.; Ohemeng, K., A. Patents on DNA Gyrase inhibi-
tors: J anuary 1995 to March 1998. Exp. Opin. Ther. Patents
1998, 8, 959-969.
(10) Lambert, H. P.; O’Grady, F. W. Antibiotic and Chemotherapy.
In Coumarins, 6th ed.; Lambert, H. P., O’Grady, F. W., Eds.;
Churchill Livingstone: Edinburgh, 1992; pp 140-141.
(11) Stieger, M.; Angehrn, P.; Wohlgensinger, B.; Gmuender, H. GyrB
mutations in Staphylococcus aureus strains resistant to cy-
clothialidine, coumermycin, and novobiocin. Antimicrob. Agents
Chemother. 1996, 40, 1060-1062.
(12) Angehrn, P.; Go¨tschi, E. Glucuronidation of Cyclothialidines.
Unpublished results.
(13) Goetschi, E.; Angehrn, P.; Gmuender, H.; Hebeisen, P.; Link,
H.; Masciadri, R.; Nielsen, J . Cyclothialidine and its congeners:
a new class of DNA gyrase inhibitors. Pharmacol. Ther. 1993,
60, 367-380.
(14) Goetschi, E.; Angehrn, P.; Gmuender, H.; Hebeisen, P.; Link,
H.; Masciadri, R.; Nielsen, J .; Reindl, P.; Ricklin, F. The DNA
gyrase inhibitor cyclothialidine: progenitor of a new class of
antibacterial agents. In Med. Chem.: Today Tomorrow, Proc.
AFMC Int. Med. Chem. Symp. - Tokyo, September 1995;
Yamazaki, M., Blackwell Science Ltd.: Oxford, U.K., 1997; pp
263-270.
(15) For all described DNA gyrase proteins and protein fragments
we refer to the E. coli numbering scheme.
(16) Boehm, H. J . The computer program LUDI: a new method for
the de novo design of enzyme inhibitors. J . Comput.-Aided Mol.
Des. 1992, 6, 61-78.
(17) Boehm, H.-J . On the use of LUDI to search the Fine Chemicals
Directory for ligands of proteins of known three-dimensional
structure. J . Comput.-Aided Mol. Des. 1994, 8, 623-632.
(18) Sprague, P. Automated chemical hypothesis generation and
database searching with Catalyst. Perspect. Drug Discovery Des.
1995, 3, 1-20.
(19) Gasteiger, J .; Rudolph, C.; Sadowski, J . Automatic generation
of 3D atomic coordinates for organic molecules. Tetrahedron
Comput. Methodol. 1990, 3, 537-547.
) 9.0), 7.34 (1H, t, J 1 ) 8.0, J 2 ) 1.2), 7.28 (1H, t, J ) 8.0),
7.13 (2H, d, J ) 9.0), 7.06 (1H, d, J ) 8.0), 6.65 (1H, d, J )
8.0), 6.34 (1H, s), 5.20 (2H, s), 4.73 (2H, s), 2.42 (3H, s), 1.16
(9H, s). MS (ISP): m/z 485 ((M + H)+). HRMS calcd for
C29H28N2O3S 484.1821, found 484.1823.
3-Mer ca p to-ben zoic Acid Meth yl Ester (24). A solution
of 3-mercapto-benzoic acid (1.00 g, 6.48 mmol) in 4 M metha-
nolic HCl was refluxed 4 h. After evaporation, the residue was
dissolved in ethyl acetate and extracted with NaHCO3. The
organic layer was washed with brine, dried (MgSO4), and
concentrated to dryness yielding 24 as a dark yellow oil (1.09
1
g, quant.). H NMR (DMSO-d6): δ 7.91 (1H, s), 7.70 (1H,d, J
) 7.6), 7.56 (1H, d, J ) 7.8), 7.40 (1H, t, J ) 7.7), 5.88 (1H, s),
3.85 (3H, s). MS (EI+): m/z 168 (M+).
4-(ter t-Bu t yl-d im et h yl-sila n yloxy)-3-(3-m et h oxyca r b -
on yl-p h en ylsu lfa n ylm eth yl)-in d a zole-1-ca r boxylic Acid
ter t-Bu tyl Ester (25). A solution of 19 (3.50 g, 7.92 mmol)
and 24 (1.40 g, 8.32 mmol) in CH2Cl2 (25 mL) containing Et3N
(1.16 mL, 8.32 mmol) was stirred for 4 h at room temperature.
The mixture was diluted with CH2Cl2 and washed with brine.
The dried (MgSO4) and evaporated organic layer was flash
chromatographed (hexane/EtOAc 7:3), providing 25 as a
slightly yellow oil (3.72 g, 89%). 1H NMR (CDCl3): δ 8.01 (1H,
s), 7.86 (1H, d, J ) 7.0), 7.71-7.64 (2H, m), 7.39-7.30 (2H,
m), 6.66 (1H, d, J ) 7.9), 4.57 (2H, s), 3.90 (3H, s), 1.67 (9H,
s), 1.03 (9H, s), 0.37 (6H, s). MS (ISP): m/z 529.3 ((M+H)+).
4-Ben zyloxy-3-(3-m et h oxyca r b on yl-p h en ylsu lfa n yl-
m eth yl)-in d a zole-1-ca r boxylic Acid ter t-Bu tyl Ester (26).
To a solution of 25 (1.29 g, 2.44 mmol) and benzyl bromide
(350 µL, 2.93 mmol) in DMF (20 mL) was added KF (283 mg,
4.88 mmol). After the mixture was stirred for 2 h at room
temperature, H2O (80 mL) was added. The mixture was
extracted with EtOAc. Flash chromatography (hexane/EtOAc
4:1) provided 26 (765 mg, 62%) as a white foam.1H NMR
(DMSO-d6): δ 7.82 (1H, s), 7.78 (1H, d, J ) 7.8), 7.63 (1H, d,
J ) 7.9), 7.61 (1H, s), 7.56-7.21 (4H, m), 7.35-7.28 (3H, m),
6.98 (1H, d, J ) 8.3), 5.29 (2H, s), 4.58 (2H, s), 3.82 (3H, s),
1.60 (9H, s). MS (EI+): m/z 504 (M+).
3-(4-Ben zyloxy-1H-in d a zol-3-ylm eth ylsu lfa n yl)-ben zo-
ic Acid (27). To a solution of 26 (757 mg, 1.49 mmol) in THF/
MeOH/H2O 2:1:1 (20 mL) was added LiOH‚H2O (189 mg, 4.50
mmol). After 5 h at room temperature, H2O (50 mL) was added.
The pH was adjusted to pH 2 by adding 3 M HCl. The obtained
suspension was extracted with EtOAc. The combined organic
layers were dried (MgSO4) and concentrated to dryness. The
residue was precipitated from ether and hexane, providing 27
(487 mg, 84%) as a white powder. 1H NMR (DMSO-d6): δ 13.1
(1H, s br), 12.9 (1H, s, br), 7.90 (1H, s), 7.74 (1H, d, J ) 7.5),
7.61 (1H, d, J ) 7.7), 7.56-7.47 (2H, m), 7.41 (1H, t, J ) 8.3),
7.33-7.21 (4H, m), 7.03 (1H, d, J ) 8.4), 6.68 (1H, d, J ) 7.7),
5.24 (2H, s), 4.62 (2H, s). HRMS calcd for C22H18N2O3S
390.1038, found 390.1038. Anal. (C22H17.5Li0.5N2O3S) C, H, Li,
N, S.
(20) Gerber, P. R.; Mueller, K. MAB, a generally applicable molecular
force field for structure modeling in medicinal chemistry. J .
Comput.-Aided Mol. Des. 1995, 9, 251-268.
(21) Daylight Chemical Information Sytsems Inc., Santa Fe, NM.
(22) Gmuender, H.; Kuratli, K.; Keck, W. Effect of pyrimido[1,6-a]-
benzimidazoles, quinolones, and Ca2+ on the DNA gyrase-
mediated cleavage reaction. Antimicrob. Agents Chemother 1995,
39, 163-169.
(23) Nakada, N.; Gmunder, H.; Hirata, T.; Arisawa, M. Characteriza-
tion of the binding site for cyclothialidine on the B subunit of
DNA gyrase. J . Biol. Chem. 1995, 270, 14286-14291.
(24) Dwek, R. A. Monographs on Physical Biochemistry: Nuclear
Magnetic Resonance (NMR) in the Biochemistry. Applications
to enzyme systems. Oxford University Press: New York, 1973;
p 396.
Ack n ow led gm en t. We sincerely thank J .-C. Alt
(chemistry), A. D’Arcy (X-ray), A. Arnet (chemistry), G.
Dale (X-ray), B. Gsell (NMR), J . Kohler (SPR), K.
Kuratli (biology), E. Kusznir (AUC), T. Masquelin
(chemistry), F. Montavon (chemistry), W. Pirson (biol-
ogy), A. Ross (NMR), P. Schmitz (chemistry), H. Senn
(NMR), U. Steiert (biology), M. Stihle (X-ray), B. Takacs
(biology), and D. Wechsler (chemistry) for their excellent
and dedicated work.
(25) Shuker, S. B.; Hajduk, P. J .; Meadows, R. P.; Fesik, S. W.
Discovering high-affinity ligands for proteins: SAR by NMR.
Science 1996, 274, 1531-1534.
(26) Klaus, W. Peak assignment in the heteronuclear 1H/15N cor-
related NMR of the N-terminal 24 kDa fragment of DNA gyrase
subunit B of S. aureus. Unpulished results.
(27) Kostrewa, D.; D’Arcy, A. Novel inhibitors of DNA gyrase: X-ray
structures. Unpublished results.
(28) Dale, G. E.; Kostrewa, D.; Gsell, B.; Stieger, M.; D’Arcy, A.
Crystal engineering: deletion mutagenesis of the 24kDa frag-
ment of the DNA gyrase B subunit from Staphylococcus aureus.
Acta Crystallogr., Sect. D: Biol. Crystallogr. 1999, D55, 1626-
1629.
(29) The triazine 7b was presented by Zeneca as a highly potent DNA
gyrase inhibitor soon after we succeeded validating triazines.
(30) Block, M. H. In Search Of New Gyrase Mediated Antibacterial
Agents. Cambridge, U.K., September, 1997.
Refer en ces
(1) Reece, R. J .; Maxwell, A. DNA Gyrase: Structure and Function.
CRC Crit. Rev. Biochem. Mol. Biol. 1991, 26, 335-375.
(2) Maxwell, A. The interaction between coumarin drugs and DNA
gyrase. Mol. Microbiol. 1993, 9, 681-686.
(3) Ho, P.-L.; Que, T.-L.; Tsang, D. N.-C.; Ng, T.-K.; Chow, K.-H.;
Seto, W.-H. Emergence of Fluoroquinolone Resistance among
Multiply Resistant Strains of Streptococcus pneumoniae in Hong
Kong. Antimicrob. Agents Chemother. 1999, 43, 1310-1313.

2674 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 14
Boehm et al.
(31) Poyser, J . P.; Telford, B.; Timms, D.; Block, M. H.; Hales, N. J .
Preparation of aminotriazine derivatives as antibacterials. PCT
Int. Appl. 1999, 59 pp.
(32) Reusser, F.; Dolak, L. A. Novenamine is the active moiety in
novobiocin. J . Antibiot. 1986, 39, 272-274.
(33) It was known that the mutation Arg136Ile could be induced
rapidly by novobiocin, leading to a highly novobiocin resistant
S. aureus strain. Nevertheless, a primary importance was
assigned to a H-bond between an inhibitor and Arg136, since
the same mutation did not occur or led to minor resistance in
the case of other inhibitors.11,29
(34) Henke, B. R.; Aquino, C. J .; Birkemo, L. S.; Croom, D. K.;
Dougherty, R. W., J r.; Ervin, G. N.; Grizzle, M. K.; Hirst, G. C.;
J ames, M. K.; J ohnson, M. F.; Queen, K. L.; Sherrill, R. G.; Sugg,
E. E.; Suh, E. M.; Szewczyk, J . W.; Unwalla, R. J .; Yingling, J .;
Willson, T. M. Optimization of 3-(1H-Indazol-3-ylmethyl)-1,5-
benzodiazepines as Potent, Orally Active CCK-A Agonists. J .
Med. Chem. 1997, 40, 2706-2725.
(35) Schuck, P. Simultaneous radial and wavelength analysis with
the Optima XL-A analytical ultracentrifuge. Prog. Colloid Polym.
Sci. 1994, 94, 1-13.
(36) Karlsson, R.; Michaelsson, A.; Mattsson, L. Kinetic analysis of
monoclonal antibody-antigen interactions with a new biosensor
based analytical system. J . Immunol. Methods 1991, 145, 229-
240.
(37) Kumar, S.; Gunnarsson, K. Small molecule drug screening based
on surface plasmon resonance. In Adv. Drug Discovery Tech;
Harvey, A. L., Ed.; Wiley: Chichester, U.K., 1998; pp 97-114.
(38) J ohnsson, B.; Loefaas, S.; Lindquist, G. Immobilization of
proteins to a carboxymethyldextran-modified gold surface for
biospecific interaction analysis in surface plasmon resonance
sensors. Anal. Biochem. 1991, 198, 268-277.
(39) Delaglio, F.; Grzesiek, S.; Vuister, G. W.; Zhu, G.; Pfeifer, J .;
Bax, A. NMRPipe:
a multidimensional spectral processing
system based on UNIX pipes. J . Biomol. NMR 1995, 6, 227-
293.
(40) Bartels, C.; Xia, T.-h.; Billeter, M.; Guentert, P.; Wuethrich, K.
The program XEASY for computer-supported NMR spectral
analysis of biological macromolecules. J . Biomol. NMR 1995, 6,
1-10.
(41) Kubota, K.; Funatsu, M.; Kanzawa, K.; Ishikawa, A.; Takeuchi,
M. Preparation of aniline derivatives as calcium release-
activated calcium channel inhibitors and their uses. J pn. Kokai
Tokkyo Koho 1999, J P 98-10147, 9 pp.
(42) Traxler, P. M.; Furet, P.; Mett, H.; Buchdunger, E.; Meyer, T.;
Lydon, N. 4-(Phenylamino)pyrrolopyrimidines: Potent and Se-
lective, ATP Site directed Inhibitors of the EGF-Receptor
Protein Tyrosine Kinase. J . Med. Chem. 1996, 39, 2285-2292.
(43) Borsche, W.; Scriba, W. Zur Kenntnis der Benzisoxazole. (About
Benzisoxazoles.) J ustus Liebigs Ann. Chem. 1939, 540, 83-98.
(44) Crooks, P. A.; Robinson, B. Azaindoles. I. Syntheses of 5-aza-
and 5,7-diazaindoles by the noncatalytic thermal indolization
of 4-pyridyl- and 4-pyrimidylhydrazones, respectively. Can. J .
Chem. 1969, 47, 2061-2067.
(45) Zhong, Z.; Xu, T.; Chen, X.; Qui, Y.; Zhang, Z.; Hu, H. A new
and facile synthesis of 1H-indazoles. J . Chem. Soc., Perkin Trans.
1 1993, 1279-1280.
(46) The herein described synthesis of 16 was easier to perform and
gave a much better yield than the method described by Zhong
et al.45
J M000017S