A de novo designed inhibitor of D-ala-D-Ala ligase from E. coli

Article abstract of DOI:10.1002/anie.200501662

(Figure Presented) Blueprints for an inhibitor: The de novo molecular design program SPROUT was used in conjunction with the X-ray crystal structures of the bacterial enzymes DdlB and VanA to produce a novel enzyme-selective inhibitor template. Following

Full text of DOI:10.1002/anie.200501662

Angewandte
Chemie
covalent linear glycan polymer that consists of alternating N-
acetylglucosamine (GlcNAc) and N-acetylmuramic acid
(MurNAc) residues, the latter of which have an appended
pentapeptide chain. In addition to the transglycosylase and
transpeptidase enzyme families, the biosynthesis of peptido-
glycan involves a number of ATP-dependant ligases (namely
the Mur ligases). These catalyze the assembly of the
pentapeptide moiety through the successive addition of l-
Ala, d-Glu, m-dpm or l-Lys, and d-Ala–d-Ala to UDP-
MurNac by MurC, MurD, MurE, and MurF enzymes,
respectively (m-dpm = meso-diaminopimelic acid; UDP =
uridine-5’-diphosphoryl). An additional ATP-dependant
enzyme, d-Ala–d-Ala ligase (dd-ligase) is responsible for
supplying the d-Ala–d-Ala dipeptide, which is the substrate
for MurF.
Inhibition of any of these essential enzymes in either
Gram-positive or Gram-negative organisms leads to the loss
of cell shape and integrity followed by bacterial death.[2–4] The
emergence of bacterial strains that are resistant to all
antibiotics in current clinical use has created an urgent need
for the development of new inhibitors of peptidoglycan
biosynthesis enzymes and, in particular, those that act on
targets that have not been previously exploited.
As part of a structure-based design and synthesis program
for the discovery of new enzyme inhibitors, we previously
described the computer-aided molecular design, synthesis,
and biological evaluation of novel inhibitors of the ligase
responsible for d-Glu attachment, MurD.[5] Other research
groups, using either library-screening methods or substrate-
inspired approaches, have also reported useful inhibitors of
the MurC[6–8] and MurD[6,9,10] ligases. Aphosphinate-based
transition-state isostere of dd-ligase has also been
reported.[10]
Enzyme Inhibitors
DOI: 10.1002/anie.200501662
The structure-based design of small-molecule enzyme
inhibitors is a powerful tool in drug discovery, particularly if
an X-ray crystal structure of the target enzyme is available.
For cases in which the X-ray crystal structure includes a
substrate or a known inhibitor, a substrate-inspired approach
can be adopted; classical molecular modeling techniques can
be used to alter the structure of the co-crystallized substrate,
and a new molecule with enhanced enzyme-binding affinity
can be produced. To identify different inhibitor structure
types, the technique of virtual high-throughput screening
(VHTS) can also be applied to the crystal structure of the
enzyme in an attempt to identify new “hits” by docking the
3D structures of molecules contained in a database.[11]
However, both techniques have a number of drawbacks.
The substrate-inspired approach requires a suitable enzyme–
substrate co-crystal structure. Furthermore, the structures of
the designed molecules will necessarily be biased toward that
of the original co-crystallized small molecule, which, as is the
case for many natural enzyme substrates, may not possess
drug-like properties and may not be an ideal starting point.
Although suitable databases allow the design of molecules by
the VHTS method to focus on drug-like fragments, this
technique is limited by the size and diversity of the small-
molecule database; even databases that consist of millions of
compounds constitute only a small fraction of the available
“diversity space”.
A De Novo Designed Inhibitor of d-Ala–d-Ala
Ligase from E. coli
Gilbert E. Besong, Julieanne M. Bostock,
Will Stubbings, Ian Chopra, David I. Roper,
Adrian J. Lloyd, Colin W. G. Fishwick,* and
A. Peter Johnson*
Peptidoglycan is an essential component of the bacterial cell
wall and provides the structural integrity necessary to resist
internal osmotic pressure and to prevent cell lysis.[1] It is a
[*] G. E. Besong, Dr. C. W. G. Fishwick, Prof. Dr. A. P. Johnson
School of Chemistry, University of Leeds
Leeds, LS29JT (UK)
Fax : (+44)113-343-6565
E-mail: c.w.g.fishwick@leeds.ac.uk
a.p.johnson@chem.leeds.ac.uk
Dr. J. M. Bostock, Dr. W. Stubbings, Prof. Dr. I. Chopra
Department of Biochemistry and Molecular Biology
University of Leeds
Leeds, LS29JT (UK)
Dr. D. I. Roper, Dr. A. J. Lloyd
Department of Biological Sciences, University of Warwick
Coventry, CV47AL (UK)
Angew. Chem. Int. Ed. 2005, 44, 6403 –6406
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
6403

Communications
Recent reports involving the application of de novo
design indicate that this approach is a powerful alternative
to these classical drug discovery approaches.[12–15] Through the
use of only structural features present within the enzyme, new
inhibitor designs can be built up sequentially according to the
requirements of the targeted binding site. Therefore, in
principle, de novo design should be a far more useful
technique than VHTS, because a good de novo design
program will examine structure space many orders of
magnitude larger than that of most virtual libraries currently
in use for this purpose. Herein, we report the application of
the de novo molecular design program SPROUT[16] to the
design of a novel inhibitor of dd-ligase from E. coli. We also
describe the development of an efficient synthesis of this
novel inhibitor and the enzymological characterization of its
interaction with dd-ligase and a related enzyme.
SPROUT[16] is a powerful suite of software modules for
de novo structure-based molecular design. SPROUT has
modules for 1) characterizing regions within a protein,
2) detecting “hot-spots” in which ligand atoms are expected
to form strong interactions with the protein, 3) docking
molecular fragments to these sites, 4) joining these fragments
with a molecular backbone to give a complete ligand, and
5) ranking the set of predicted ligands by criteria that include
predicted binding energy, structural complexity, and synthetic
accessibility.
In E. coli, there are two isoforms of dd-ligase (DdlAand
DdlB) encoded by the genes ddlA and ddlB, with similar
kinetic characteristics. The crystal structure of E. coli DdlB
complexed with a phosphorylphosphinate inhibitor and
resolved to 2.3 Š has been used to propose a catalytic
mechanism for the formation of d-Ala–d-Ala.[17] To use a
de novo approach for the discovery of novel dd-ligase
inhibitors, we applied SPROUT to the X-ray crystal struc-
ture[18] of dd-ligase encoded by ddlB from E. coli.
The active site contains two distinct d-alanine binding
sites; the N-terminal site is a high-affinity site with strict
substrate specificity, whereas the C-terminal site is a lower-
affinity site showing lower substrate specificity.[19] Aphos-
phonate-based transition-state isostere is included in the
crystal structure of DdlB, located within the active site. It
makes direct contacts with essentially all of the residues
implied by the catalytic reaction mechanism[17] (Figure 1).
To design a structurally simple inhibitor which was
predicted to show high affinity for the enzyme, we wished
to use only a small number of these contacts in our SPROUT
design strategy. Examination of this active site with SPROUT
Figure 1. a) The bonding interactions between the phosphonate-based
isostere and DdlB based on the crystal structure; b) view of the DdlB
active site.
(Figure 2) was selected from the many suggestions that arose
from the SPROUT design iterations.
One of us recently reported the X-ray crystal structure of
the closely related ligase VanAco-crystallized with the same
phosphinate-based isostere as described above for DdlB.[20] In
addition to the dd-ligases, VanAis present in certain
vancomycin-resistant enterococci and is responsible for the
formation of d-alanine–d-lactate. This unit is then incorpo-
rated into the pentapeptide chain, and it is the resulting
lactate-containing peptide that binds poorly to vancomycin.
We were aware that although the overall topology of the
active sites of VanAand DdlB were very similar, there are
also important differences, particularly in the hydrophobic
region bordered by L282 and Y216 in DdlB (corresponding to
R317 and H244 in VanA). We therefore used SPROUT to
produce a model of our designed inhibitor 1 within the active
site of VanA(Figure 3). This revealed that the three designed
hydrogen-bonding contacts to E15 (E16 in VanA), R255
(R290 in VanA), and G276 (G311 in VanA) as well as binding
to the magnesium ion were possible for the inhibitor 1 bound
within the active site of VanA. However, VanA possesses a
indicated that use of the side chains of residues R255 and E15,
a
À
the C backbone N H group of G276, and the magnesium ion
Mg331 should result in relatively simple designed molecular
templates. Additionally, SPROUT analysis revealed the
presence of a small hydrophobic region (which is not used
by the phosphonate isostere) between the phenolic ring of
Y216 and the alkyl side chain of L282 that we also wished to
use for inhibitor design. During this work, it became apparent
that the design of a molecular template which allows good
contacts to be made to all these sites would require a rigid
small-ring framework; the cyclopropyl-based amino acid 1
6404 www.angewandte.org
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 6403 –6406

Angewandte
Chemie
Figure 3. SPROUT-designed DdlB inhibitor 1 in the active site of VanA
showing contacting residues. The corresponding residues in DdlB are
superimposed (VanA residues in darker shading, DdlB numbering
shown above VanA numbering).
Figure 2. a) The bonding interactions of SPROUT-designed inhibitor 1;
b) view of inhibitor 1 within DdlB active site.
much narrower hydrophobic region between H244 and R317
than the equivalent region in DdlB (between L282 and Y216)
due to protrusion of the imidazole ring of VanAH244 into the
cavity (Figure 3). In fact, SPROUTanalysis of the inhibitor 1–
VanAcomplex indicated that this inhibitor should bind only
very poorly to VanA. Following synthesis, we therefore
planned to measure the affinity of inhibitor 1 with both
DdlB and VanAand test the selectivity of our inhibitor
predicted by these computational studies.
Ashort synthesis of cyclopropane 1 was developed and is
summarized below (Scheme 1). Ring opening of the readily
available racemic lactone 2[21] followed by halogen exchange
and oxidation afforded aldehyde 3. Olefination with phos-
phorane 4 followed by catalytic reduction of the double bond
with concomitant N-deprotection and final ester hydrolysis
yielded the desired inhibitor 1 as an inseparable 1:1 mixture of
diastereomers.[22]
Scheme 1. Synthesis of inhibitor 1: a) SOBr2, MeOH, 93%; b) NaI,
Me2CO, 98%; c) DMSO, iPr2NEt, 63%; d) 4, DBU, 87%; e) H2, Pd/C,
MeOH, 33%; f) LiOH, MeOH, 63%. DBU=1,8-diazabicyclo-
[5.4.0]undec-7-ene.
had no inhibitory activity against VanA. The apparent Ki
value of cycloserine for DdlB, a known inhibitor of this
enzyme, was 1.4 mm. Clearly, the designed template is for a
single enantiomer of absolute stereochemistry shown in
Figure 2. In these proof-of-principal studies, however, we
chose to perform biological evaluation with a readily avail-
able diastereomeric mixture containing this designed tem-
plate along with equimolar amounts of three other stereiso-
meric versions of this inhibitor. Therefore, although the
measured binding affinity of this mixture to DdlB is very
encouraging, it is likely that the actual designed enantiomer 1
is an even more potent inhibitor of DdlB. It should be noted
that the production of inhibitors that are active at the low
micromolar concentration range after pure de novo design is
quite typical of the current state of this approach; of course,
inhibitors active at this level provide an excellent starting
Enzymological evaluation[23] of designed inhibitor 1 was
performed with recombinant E. coli DdlB and VanA
enzymes. In keeping with our modeling predictions, this
molecule (as a diastereomeric mixture) inhibits the activity of
DdlB with an apparent Ki value of 12.5(Æ0.1) mm, whereas it
Angew. Chem. Int. Ed. 2005, 44, 6403 –6406
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 6405

Communications
[20] D. I. Roper, T. Huyton, A. Vagin, G. Dodson, Proc. Natl. Acad.
Sci. USA 2000, 97, 8921 – 8925.
point for further optimization. Development of more accu-
rate functions for the estimation of binding affinity remains a
key requirement for further progress in the de novo design of
even more potent inhibitors.
[21] S. Shuto, H. Takada, D. Mochizuki, R. Tsujita, Y. Hase, S. Ono,
N. Shibuya, A. Matsuda, J. Med. Chem. 1995, 38, 2964 – 2968.
[22] Diastereoisomer ratio determined by high-resolution NMR
analysis. Attempts to separate the diastereomeric mixture
under a variety of HPLC conditions proved unsuccessful.
[23] Apparent Ki values were estimated with the method of Y. Cheng,
W. H. Prusoff, Biochem. Pharmacol. 1973, 22, 3099 – 3108. For
DdlB assays (Km for the first d-Ala site on DdlB = 3.3 mm),
activity was monitored by detection of orthophosphate using
malachite green; see: P. A. Lanzetta, P. S. Alvarez, P. S. Reinach,
O. A. Candia, Biochemistry 1979, 18, 95 – 98. Assays were
performed at 378C with inhibitor 1 (between 0.003 mm and
32 mm), DdlB (0.32 mm), and enzyme substrate in reaction
buffer: (HEPES (50 mm, pH 8), ATP (1 mm), MgCl2 (10 mm),
KCl (10 mm), d-Ala (2.5 mm)), malachite green-molybdate
(50 mm, malachite green hydrochloride (0.045% w/v) and
ammonium molybdate (4.2% w/v) in 4-n HCl). VanAassays
were performed with the same inhibitor concentration range as
for DdlB by using the method of I. A. Lessard, V. L. Healy, I. S.
Park, C. T. Walsh, Biochemistry 1999, 38, 14006 – 14022, follow-
ing the d-Ala and d-lactate dependent release of ADP.
In summary, it appears that SPROUT-based de novo
design holds tremendous potential for the rapid discovery of
selective inhibitors of DdlB and related enzymes. We believe
that this approach is complementary to the use of high-
throughput screening and is particularly attractive for cases in
which such screening methodology is not available or in which
access to large collections of library compounds of sufficient
molecular diversity is limited.
Received: May 13, 2005
Revised: June 4, 2005
Published online: September 13, 2005
Keywords: inhibitors · ligases · medicinal chemistry ·
.
molecular modeling · molecular recognition
[1] K. H. Schleifer, D. Kandler, Bacteriol. Rev. 1972, 36, 407 – 412.
[2] K. Isono, M. Uramoto, H. Kusakabe, K. Kimura, K. Izaki, C. C.
Nelson, J. A. McCluskey, J. Antibiot. 1985, 38, 1617 – 1620.
[3] K. Kimura, Y. Ikeda, S. Kagami„ M. Yoshihara, J. Antibiot. 1998,
51, 1099 – 1103.
[4] D. W. Green, Expert Opin. Ther. Targets 2002, 6, 1 – 5.
[5] J. R. Horton, J. M. Bostock, I. Chopra, L. Hesse, S. E. V. Phillips,
D. J. Adams, A. P. Johnson, C. W. G. Fishwick, Bioorg. Med.
Chem. Lett. 2003, 13, 1557 – 1560.
[6] Z. Li, G. D. Francisco, W. Hu, P. Labthavikul, P. J. Petersen, A.
Severin, G. Singh, Y. J. Yang, B. A. Rasmussen, Y. I. Lin, J. S.
Skotnicki, T. S. Mansour, Bioorg. Med. Chem. Lett. 2003, 13,
2591 – 2594.
[7] D. E. Ehmann, J. E. Demeritt, K. G. Hull, S. L. Fisher, Biochim.
Biophys. Acta Proteins Proteom. 2004, 1698, 167 – 174.
[8] A. El Zoeiby, F. Sanschagrin, A. Darveau, J. R. Brisson, R. C.
Levesque, J. Antimicrob. Chemother. 2003, 51, 531 – 543.
[9] M. E. Tanner, S. Vaganay, J. van Heijenoort, D. Blanot, J. Org.
Chem. 1996, 61, 1756 – 1760.
[10] L. D. Gegnas, S. T. Waddell, R. M. Chabin, S. Reddy, K. K.
Wong, Bioorg. Med. Chem. Lett. 1998, 8, 1643 – 1648.
[11] K. M. Branson, B. J. Smith, Aust. J. Chem. 2004, 57, 1029 – 1037.
[12] G. Schneider, O. Clement-Chomienne, L. Hilfiger, P. Schneider,
S. Kirsch, H.-J. Bohm, W. Neidhart, Angew. Chem. 2000, 112,
4305 – 4309; Angew. Chem. Int. Ed. 2000, 39, 4130 – 4133.
[13] Q. Han, C. Dominguez, P. F. W. Stouten, J. M. Park, D. E. Duffy,
R. A. Galemmo Jr., K. A. Rossi, R. S. Alexander, A. M. Small-
wood, P. C. Wong, M. M. Wright, J. M. Luettgen, R. M. Knabb,
R. R. Wexler, J. Med. Chem. 2000, 43, 4398 – 4415.
[14] T. Honma, K. Haysashi, T. Aoyama, N. Hashimoto, T. Machida,
K. Fukasawa, T. Iwama, C. Ikeura, M. Ikuta, I. Suzuki-
Takahashi, Y. Iwasawa, T. Hayama, S. Nishimura, H. Morishima,
J. Med. Chem. 2001, 44, 4615 – 4627.
[15] B. A. Grzybowski, A. V. Ishchenko, C.-Y Kim, G. Topalov, R.
Chapman, D. W. Christianson, G. M. Whitesides, E. I. Shakhno-
vich, Biophysics 2002, 99, 1270 – 1273.
[16] V. J. Gillet, W. Newell, P. Mata, G. Myatt, S. Sike, Z. Zsoldos,
A. P. Johnson, J. Chem. Inf. Comput. Sci. 1994, 34, 207 – 217.
[17] C. Fan, P. C. Moews, C. T. Walsh, J. R. Knox, Science 1994, 266,
439 – 443.
[18] Brookhaven Protein Data Bank code 2DLN.
[19] L. E. Zawadzke, T. D. H. Bugg, C. T. Walsh, Biochemistry 1991,
30, 1673 – 1682.
6406 www.angewandte.org
ꢀ 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 6403 –6406