Design and synthesis of indolo[2,3-a]quinolizin-7-one inhibitors of the ZipA-FtsZ interaction

Article abstract of DOI:10.1016/j.bmcl.2004.01.028

The binding of FtsZ to ZipA is a potential target for antibacterial therapy. Based on a small molecule inhibitor of the ZipA-FtsZ interaction, a parallel synthesis of small molecules was initiated which targeted a key region of ZipA involved in FtsZ bindi

Full text of DOI:10.1016/j.bmcl.2004.01.028

Bioorganic & Medicinal Chemistry Letters 14 (2004) 1427–1431
Design and synthesis of indolo[2,3-a]quinolizin-7-one inhibitors of
the ZipA–FtsZ interaction
Lee D. Jennings,a,* Ken W. Foreman,d Thomas S. Rush, III,d Desiree H. H. Tsao,d
Lidia Mosyak,d Yuanhong Li,d Mohani N. Sukhdeo,a Weidong Ding,b
Elizabeth G. Dushin,c Cynthia Hess Kenny,b Soraya L. Moghazeh,c Peter J. Petersen,c
Alexey V. Ruzin,c Margareta Tuckmanc and Alan G. Sutherlanda
aWyeth Research, Department of Medicinal Chemistry, 401 N. Middletown Rd, Pearl River, NY 10965, USA
bBiophysics/Enzymology, 401 N. Middletown Rd, Pearl River, NY 10965, USA
cInfectious Disease Discovery Research, 401 N. Middletown Rd, Pearl River, NY 10965, USA
dWyeth Research, Departments of Computational Chemistry and Structural Biology, 35 Cambridge Park Dr.,
Cambridge, MA 02140, USA
Received 29 October 2003; revised 9 January 2004; accepted 14 January 2004
Abstract—The binding of FtsZ to ZipA is a potential target for antibacterial therapy. Based on a small molecule inhibitor of the
ZipA–FtsZ interaction, a parallel synthesis of small molecules was initiated which targeted a key region of ZipA involved in FtsZ
binding. The X-ray crystal structure of one of these molecules complexed with ZipA was solved. The structure revealed an
unexpected binding mode, facilitated by desolvation of a loosely bound surface water.
# 2004 Elsevier Ltd. All rights reserved.
The prokaryotic tubulin analogue, FtsZ, and the mem-
brane-anchored protein, ZipA, are essential for Eschercia
coli cell division. ZipA links FtsZ to the cytoplasmic
membrane, forming part of the septal ring that mediates
cell division. Overexpression or depletion of ZipA cre-
ates non-septate filaments, disrupting cell division in E.
coli.1 Accordingly, agents that inhibit the ZipA–FtsZ
interaction are potentially useful in treating bacterial
infections in humans and in animals. Consequently, the
discovery of small-molecule inhibitors of the ZipA–FtsZ
interaction was chosen as a goal of our anti-infective
research program.2
polarization assay.3 The crystal structure of 1 bound to
ZipA was solved at 2.0 A resolution and showed that 1
occupies a hydrophobic cavity on the surface of ZipA
necessary for the binding of the FtsZ peptide.4 The
strongest contribution to binding appears to be the
hydrophobic effect, as the molecule displaces about 600
A2 worth of low entropy water from the site. A smaller
enthalpic contribution to binding stems from both van
der Waals interactions and a p-stacking interaction
between the hexahydroquinolizinone ring system and a
phenylalanine sidechain within the cavity.
Prior to screening of the entire corporate compound
collection, the members of a set of structurally diverse,
lead-like compounds were assayed for their interaction
with ZipA. 1,2,3,4,12,12b-Hexahydro-indolo[2,3-a]qui-
nolizin-7-one (1) was found to inhibit ZipA–FtsZ bind-
ing by surface plasmon resonance and in a fluorescence
A series of searches of our corporate compound
collection for compounds similar to 1 yielded a number
of compounds of interest. Among these were the
dihydrocarbazole 25 and the acridine 36 which, while
Keywords: Antibacterials; Protein–protein interaction; X-ray crystal
structure; Structure based design.
* Corresponding author. Tel.: +1-845-602-3618; fax: +1-845-602-
3045; e-mail: jenninl@wyeth.com
0960-894X/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2004.01.028

1428
L. D. Jennings et al. / Bioorg. Med. Chem. Lett. 14 (2004) 1427–1431
less potent inhibitors than 1, also yielded high quality
X-ray structures of their complex with ZipA. The indole
nucleus of 2 bound in an essentially identical position to
that of 1 but had the added feature of the amine side
chain accessing a neighboring pocket of ZipA which is
not involved in the binding of FtsZ. The acridine ring
system of 3 also bound to a similar region of ZipA but
the longer side chain not only entered the neighboring
pocket but also made a hydrogen bonding interaction
with a histidine residue His46. We therefore sought to
prepare a series of compounds which would retain the
superior binding of 1 to the shallow hydrophobic pocket
but would also bear side-chains which would emulate
the interaction of the second pocket displayed by 2 and
3.7
presence of potassium hydroxide in DMSO at 180 ꢀC
under microwave irradiation to give the carboxylic acids
11a and 11b having chain lengths of 1 and 3 methylene
units. The carboxylic acids were coupled with amines to
form amides using 1-ethyl-3-(3-dimethylaminopropyl)-
carbodiimide (EDCI) as a coupling reagent. Similarly, 1
was alkylated with methyl acrylate in the presence of
potassium tert-butoxide at 150 ꢀC under microwave
irradiation and the methyl proprionate intermediate was
hydrolyzed with acid to give the propionic acid deriva-
tive 11c. In the same way this carboxylic acid inter-
mediate was also coupled with amines to afford amides.
Indoloquinolizinone 18 was prepared following the pub-
lished route to the 9-methyl substituted 1,2,3,4,6,7,12,12b-
octahydro-7-oxo-indolo[2,3-a]quinolizine9 (Scheme 1).
Phenylhydrazine was condensed with 2-acetyl pyridine
in ethanol and the resultant hydrazone was cyclized in
polyphosphoric acid at 130 ꢀC to afford the 2-pyridin-
2-yl-1H-indole 5. The pyridine ring of 5 was reduced
with hydrogen over platinum in acetic acid. The satu-
rated product 6 was alkylated on nitrogen with ethyl
bromoacetate and the resulting ester was cyclized
in methanesulfonic acid at 55 ꢀC to afford 1.
Derivatives of 1 were prepared by the following methods.
In Method A (Scheme 2), 1 was alkylated with acrylo-
nitrile in the presence of lithium bis(trimethylsilyl)amide
at 150 ꢀC under microwave irradiation10 to afford the
nitrile 8. The nitrile was reduced by hydrogenation over
Raney nickel in a methanolic solution of ammonia at 50
PSI H2. The resulting amine (9) was acylated in parallel
with acid chlorides and sulfonyl chlorides to give amides
and sulfonamides 10a–k.11 Products were individually
purified by reverse phase HPLC and checked for purity
and integrity by LC-MS.
Scheme 2. Synthesis of Compounds 10a–k. Conditions: (a) acrylonitrile,
LiHMDS, toluene, 150 ꢀC, 17 min, microwave irradiation; (b) H2, 50
psi, Raney nickel, NH3, MeOH (65%); (c) R1COCl or R1SO2Cl,
Na2CO3, CH2Cl2–H2O.
In Method B (Scheme 3), 1 was alkylated with methyl
4-bromobutyrate or with methyl bromoacetate in the
Scheme 3. Synthesis of compounds 12a–g. Conditions: (a) methyl
bromoacetate or methyl 4-bromobutyrate, KOH, DMSO, 180 ꢀC,
2–17 min, microwave irradiation; (b) methyl acrylate, KOt-Bu,
toluene, 150 ꢀC, 17 min, microwave irradiation; (c) 6N HCl; (d)
R1NH2, EDCI, DIEA, CH2Cl2–H2O.
Scheme 1. Synthesis of Compound 1. Conditions: (a) phenylhydrazine,
EtOH; (b) PPA, 130 ꢀC, (81% for two steps); (c) H2, 50 psi, PtO2,
AcOH (82%); (d) methyl bromoacetate, Na2CO3, n-Bu4N+I- (100%);
(d) MeSO3H, 55 ꢀC (79%).

L. D. Jennings et al. / Bioorg. Med. Chem. Lett. 14 (2004) 1427–1431
1429
Inhibition of the ZipA–FtsZ interaction was assayed on
a Biacore 3000 (Biacore Inc., Piscataway, NJ) using the
peptide CEPDWLDIPAFLRKAD-NH2 immobilized
on a CM5 sensor chip via thiol coupling as described by
the manufacturer. This peptide is an N-terminal cysteine
labeled version of the C-terminal 16 amino acid residues
of E. coli FtsZ. A tryptophan has been substituted for
tyrosine in the fifth position in order to lower the KD of
the peptide binding to ZipA approximately three-fold.
The KD in 5% DMSO was approximately 5 mM and
ZipA was used at this concentration in the assays.12
Compounds were tested at 1 mM concentration. Results
for selected compounds are collected in Table 1. Com-
pounds were also tested for antibacterial activity against
a panel of Gram-positive and Gram-negative bacteria,
including E. coli.13 The minimal inhibitory concen-
tration was found to be greater than or equal to 200 mM
for all compounds tested against all species in the panel.
Table 1 shows that despite significant variation in
the shape and length of the sidechains attached to the
indoloquinolizinone scaffold, no significant improve-
ment in the inhibition of the interaction of ZipA with
the modified FtsZ 16-mer was observed. Consequently,
elucidation of the structure of the ZipA/inhibitor com-
plex was felt to be necessary in order for us to determine
how best to make progress on the optimization of this
series.
The compound 10b was co-crystallized with ZipA14 and
the crystal structure was solved using the molecular
replacement method (AMORE)15 with ZipA monomer
as a search model. The complex was crystallographically
refined with CNS,16 and the process of the refinement
and rebuilding was repeated cyclically several times
until most of the water molecules could be fitted into the
density. The refined models (with no inhibitor added)
Table 1. Inhibition of ZipA–FtsZ N-terminal 16-mer association observed using an optical biosensor systema
Example
R1
H
% Inhibition
at 1 mM
Example
R1
% Inhibition
at 1 mM
1
9
39
47
10j
8
38
24
31
23
39
21
32
22
10k
12a
12b
12c
12d
12e
12f
10a
10b
10c
10d
10e
10f
10g
10h
10i
38
46
33
25
11
12
41
25
30
12g
a Conditions: Compounds were tested at 1 mM concentration in the presence of 5 mM ZipA in 5% DMSO.

1430
L. D. Jennings et al. / Bioorg. Med. Chem. Lett. 14 (2004) 1427–1431
Figure 1.
were then used to calculate the 3Fo-2Fc and Fo-Fc
maps, which showed clear electron density for the
bound compounds. The final model for the ZipA/10b
complex contained 2 protein molecules, one inhibitor
molecule, and 250 molecules of water. No electron density
that could be attributed to the inhibitor bound to the
second ZipA monomer was observed. The correspond-
ing refinement values obtained were Rfactor=22.6% and
Rfree=27.9%.
bound in both the site and the targeted region adjacent
to it. Models for compound 10b outside the site require
significant rearrangement of the waters, whereas the
crystal structure reveals a mode in which the water net-
work is largely undisturbed. In addition, the molecule of
water that makes a hydrogen bond with compound 10b
is mostly shielded from bulk water. This shielding
should lower the local dielectric around the water,
enhancing the water’s binding to the protein. Thus, the
water effectively becomes a structural component of the
protein. Given these considerations, locking inhibitors
into a specific orientation appeared possible, opening
the door to structure-based potency optimization. We
will report the design and synthesis of these subsequent
series of ZipA inhibitors in a forthcoming publication.
The crystal structure shows that, while 10b occupies the
same hydrophobic area on ZipA as 1 [Fig. 1(a)], the
scaffold is inverted in the binding domain and the sul-
fonamide chain extends in the exact opposite direction
as was expected [Fig. 1(b)]. The key hydrophobic inter-
action between the hydrophobic hexahydroquinolizin-
one and the active site Phe269 is maintained.
Additionally, the sulfonamide nitrogen atom makes a
hydrogen bond with a bound water molecule that is also
present in the FtsZ peptide–ZipA structure.4 The pri-
mary association of the bound water with the protein is
with the backbone carbonyl of Asn247. The iso-propyl-
sulfonyl group acts to shield the water molecule, isolat-
ing it from the bulk water which surrounds the protein.
In retrospect, this binding mode might have been
anticipated, as the FtsZ peptide binds in exactly the
same region as does 10b. Because the hydrophobic effect
is non-specific and dominant for the interaction of 1
with the protein, 10b was able to find alternate, more
potent modes within the binding site.
Acknowledgements
We would like to acknowledge Tom McDonagh and
Mark Stahl for providing purified protein for crystal-
lizations, Russell Dushin for helpful discussions, and
George Ellestad for assistance with assay development.
References and notes
1. Hale, C. A.; de Boer, P. A. J. Cell 1997, 88, 175.
2. (a) For small molecule inhibitors of FtsZ polymerization,
see: Wang, J.; Galgoci, A.; Kodali, S.; Herath, K. B.;
Jayashuriya, H.; Dorso, K.; Vicente, F.; Gonzalez, A.;
Cully, D.; Bramhill, D.; Singh, S. J. Biol. Chem. 2003,
278, 44424. (b) White, E. L.; Suling, W. J.; Ross, L. J.;
Seitz, L. E.; Reynolds, R. C. J. Antimicrobial Chemo-
therapy 2002, 50, 111.
3. Kenny, C. H.; Ding, W.; Kelleher, K.; Benard, S.;
Dushin, E. G.; Sutherland, A. G.; Mosyak, L.; Kriz, R.;
Ellestad, G. Anal. Biochem. 2003, 323, 224.
4. For the X-ray crystal structure of ZipA and the C-term-
inal fragment of FtsZ, see: Mosyak, L.; Zhang, Y.;
Glasfeld, E.; Haney, S.; Stahl, M.; Seehra, J.; Somers,
W. S. EMBO Journal 2000, 19, 3179.
The key interactions between 1 and the FtsZ binding
domain on ZipA are hydrophobic. Hydrophobic inter-
actions are not strongly directional and consequently
compounds in this lead series bound in unexpected
ways, making structure based design difficult. Even
though potency is largely unaffected by the additional
hydrogen bonding capacity of these chains, compound
10b is inverted in the binding cavity. This scaffold reor-
ientation illustrates the powerful balance between elec-
trostatics that orient the molecule and the desolvation
costs that prevent significant potency gains from those
hydrogen bonds in this relatively open site. Waters are
5. Freed, M. E.; Hertz, E.; Rice, L. M. J. Med. Chem. 1964,
7, 628.

L. D. Jennings et al. / Bioorg. Med. Chem. Lett. 14 (2004) 1427–1431
1431
6. Denny, W. A. Current Medicinal Chemistry 2002, 9, 1655.
7. For an account of an alternate approach to the optimiza-
tion of the indoloquinolizinone inhibitor lead, see:
Sutherland, A. G.; Alvarez, J.; Ding, W.; Foreman, K. W.;
Kenny, C. H.; Labthavikul, P.; Mosyak, L.; Petersen,
P. J.; Rush, T. S.; Ruzin, A.; Tsao, D. H. H.; Wheless,
K. L. Org. Biomol. Chem. 2003, 1, 4138.
8. Rosenmund, P.; Trommer, W.; Dorn-Zachertz, D.;
Ewerdwalbesloh, U. Liebigs Annalen der Chemie 1979, 11,
1643.
blank channel in which a mock thiol coupling with no
peptide had been performed and the signal from the blank
was subtracted from that with the peptide. Compounds
were added to ZipA, filtered, and then binding to the
peptide was determined. Percent inhibition is calculated
from the average resonance signal for the sensogram col-
lected with inhibitor present and the sensogram collected
in the absence of inhibitor.
13. ‘Methods for Dilution of Antimicrobial Susceptibility
Tests for Bacteria That Grow Aerobically’, Approved
Standards: M7-A5, Vol. 20, National Committee for
Clinical Laboratory Standards, Wayne, PA. Bacterial
species in the panel included S. aureus (ATCC 29213), E.
faecalis, S. pneumoniae, B. subtilis, H. influenzae (ATCC
49247), M. catarrhalis, and E. coli.
9. Grandi, T.; Sparatore, F.; Sparatore, A. Il Pharmaco
1999, 54, 479.
10. EmrysTM Microwave Synthesizer, Personal Chemistry
AB, Uppsala, Sweden.
11. Na2CO3 (0.034 g, 0.32 mmol) in water (0.3 mL) was
added to a solution of propylamine 9 (0.024 g, 0.08 mmol)
in 0.3 mL CH2Cl2. The reaction was shaken for 1 min.
Furoyl chloride (0.021 g, 0.16 mmol) was added and the
reaction was mixed at ambient temperature overnight.
The reaction was diluted with 2 mL CH2Cl2 and washed
with approximately 0.8 mL aqueous NaHCO3. The
organic phase was removed and the solvent was evapo-
rated. The residue was purified by RP-HPLC (Gilson
Semi-Preparative HPLC system with Unipoint Software
v. 1.71, Phenomenex C18 Luna column, 21.2Â100 mm, 5 m
particle size, water–acetonitrile solvent system with added
0.05% NH4OH buffer, at 22.5 mL/min) to give 14.3 mg
(36.5 mmol) of 10a [LC/MS (Hewlett Packard 1100 MSD
with ChemStation Software, Keystone Aquasil C18 50
mmÂ2 mm column, 5m particle size, at 40 ꢀC, 10 mM
NH4OAc–acetonitrile solvent system at 0.8 mL/min flow
rate, 254 nm DAD detection, API-ES scanning mode,
fragmentor 70 mV): m/z 392 (M+H); retention time
1.96 min].
14. 2 mM of 10b was added to ZipA (20 mg/mL), so that the
final mixture contained 2:1 compound versus the protein.
Co-crystals of approximate dimensions 0.5Â0.7Â0.15
mm3 were grown by vapor diffusion from 2 mL droplets
consisting of 1 mL of protein solution and 1 mL of pre-
cipitant (20% PEG 6000, 0.1 M MES pH 6.0). To produce
high quality crystals, streak seeding with the native crys-
tals as seeds was applied. Crystals belonging to space
group P21 (a=52.45 A, b=38.85 A, c=71.52 A, b=106.3
deg) containing two ZipA molecules per asymmetric unit
were obtained. Prior to data collection, the crystals were
flash cooled in liquid nitrogen at 100 K using 25% ethyl-
ene glycol as a cryoprotectant. The 2.1 A data were col-
lected in house using RAXIS IV imaging plate system.
The data were again integrated using DENZO and then
scaled and merged with SCALEPACK.17
15. Navaza, J. Acta Crystallogr. 1994, A50, 157.
16. Brunger, A. T.; Adams, P. D.; Clore, G. M.; DeLano,
W. L.; Gros, P.; Grosse-Kunstleve, R. W.; Jiang, J.-S.;
Kuszewski, J.; Nilges, M.; Pannu, N. S.; Read, R. J.; Rice,
L. M.; Simonson, T.; Warren, G. L. Acta Crystallogr.
1998, D54, 905.
12. E. coli ZipA(185-328) was passed over the immobilized
peptide in Biacore’s HBS-EP buffer (10 mM Hepes, pH
7.4, 150 mM NaCl, 3 mM EDTA, 0.005% P20) with
DMSO added to a final concentration of 5%. The run-
ning buffer was the same. The sample was also run over a
17. DENZO and SCALEPACK: Otwinowski, Z.; Minor, W.
Methods Enzymol. 1997, 276, 307.