European Journal of Medicinal Chemistry 57 (2012) 185e195

Contents lists available at SciVerse ScienceDirect

European Journal of Medicinal Chemistry

journal homepage: http://www.elsevier.com/locate/ejmech

Original article

Identification of potential drug targets in Yersinia pestis using metabolic pathway

analysis: MurE ligase as a case study
Aditya Sharma 1, Archana Pan*
Centre for Bioinformatics, School of Life Sciences, Pondicherry University, Pondicherry 605014, India

a r t i c l e i n f o a b s t r a c t

Article history: Sporadic outbreaks of plague, lack of a vaccine, emergence of multidrug-resistant strains of Yersinia
Received 19 April 2012 pestis, and its potential use in bioterrorism, call for an urgent need to develop new drugs for plague. We
Received in revised form have used comparative metabolic pathway analysis to identify 245 drug-target candidate enzymes in
7 September 2012
Y. pestis CO92 which are non-homologous to host Homo sapiens and likely to be essential for the path-
Accepted 11 September 2012
Available online 18 September 2012
ogen’s survival. Further analysis revealed that 25 of these are potential choke point enzymes. As a case
study, structure of a choke point enzyme, MurE ligase, was modeled and docking studies performed
against a library of compounds leading to identification of a potential inhibitor. This approach enables
Keywords:
Docking
rapid potential drug-target identification, thereby facilitating search for new antimicrobials.
Drug target Ó 2012 Elsevier Masson SAS. All rights reserved.
Metabolic pathways
MurE ligase
Plague
Yersinia pestis

1. Introduction annually, in absence of any effective vaccine against it. Current

Plague is often dismissed as an ancient disease, with little
relevance in the present day world. The historical prevalence of the
disease has been well documented [1,2]. However, sporadic
outbreaks in the last two decades and the potential use of the
causative organism as an agent of bioterrorism make it a relevant
threat even now [2]. Though the incidence of the disease has
drastically reduced with improvement in health awareness and
availability of antibiotics, it can by no means be considered as
eradicated.
The causative organism of plague, Yersinia pestis is an obligate
parasite. Based on metabolic differences (ability to convert nitrate
to nitrite and glycerol fermentation), it has three biotypes viz.,
antiqua, mediaevalis and orientalis, with no difference in their
virulence [1]. It infects mammals as their primary hosts, with fleas
acting as agents of transmission. Transmission of the pathogen is
also possible through air, making it an even bigger threat as a bio-
logical weapon. Plague continues to afflict thousands of people
* Corresponding author. Tel.: þ91 413 2654584; fax: þ91 413 2655211.
E-mail addresses: aditya2088@gmail.com (A. Sharma), archana@bicpu.edu.in,
archanpan@gmail.com (A. Pan).
1
Present address: Delhi Technological University, Shahbad Daulatpur, Main
Bawana Road, Delhi 110042, India.
treatment relies solely on antibiotics. In such a scenario, emergence
of strains resistant to multiple drugs is a cause of great concern
[3,4]. Also, several thousand new cases are reported each year,
predominantly in Africa [5]. All the cases in recent history have
been attributed to the orientalis biovar. In 2001, the complete
genome of a clinical isolate belonging to the orientalis biovar,
Y. pestis CO92 was sequenced [6]. Endemic plague foci persist in
many countries in Africa; the former Soviet Union; the Americas,
including the southwestern United States; and parts of Asia. From
1989 to 2003, 38,310 human plague cases and 2845 deaths have
been reported to the World Health Organization by 25 countries
[7]. All these factors suggest that plague still poses a grave threat to
human health, and more potent alternatives to current treatment
methods are urgently required. Hence, in the present study we
attempt to find a set of potential drug targets in the Y. pestis CO92
strain of the orientalis biovar, responsible for most of the recent
outbreaks of the disease.
Availability of complete genome sequences of human pathogens
is a crucial asset in obtaining biological information about them. In
silico analysis of these genomes and the information extracted from
them is a useful strategy to develop tools to counter these patho-
gens. One aspect of this strategy is the analysis of the complete
metabolic network of these pathogens to understand their physi-
ology in depth. A deep understanding of the metabolic networks,

0223-5234/$ e see front matter Ó 2012 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.ejmech.2012.09.018
186 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195

intra-cellular processes, and the physiology of these pathogens can Table 2

help us in determining the factors responsible for their pathoge- Metabolic pathways present in both the pathogen Yersinia pestis CO92 and the host
Homo sapiens.
nicity, virulence, and survival. This in turn, can be used to find novel
drugs and vaccines against them. S. No. Metabolic pathway KEGG pathway ID
In this study, comparative metabolic pathway analysis was 1. Glycolysis/gluconeogenesis ype00010
performed between the pathogen Y. pestis CO92 and its host Homo 2. Citrate cycle (TCA cycle) ype00020
3. Pentose phosphate pathway ype00030
sapiens in order to detect the enzymes which are unique to path-
4. Pentose and glucuronate interconversions ype00040
ogen. Enzymes which are specific to the pathogen, and show no 5. Fructose and mannose metabolism ype00051
significant homology to any protein in the host organism can serve 6. Galactose metabolism ype00052
as potential drug targets, as there is little risk of a potential drug 7. Ascorbate and aldarate metabolism ype00053
adversely interacting with a host protein. Also, inactivation of 8. Fatty acid biosynthesis ype00061
9. Fatty acid metabolism ype00071
proteins which are essential for survival could be lethal for the 10. Ubiquinone and other terpenoid-quinone ype00130
pathogen. Based on this a list of potential drug targets for plague biosynthesis
have been compiled. To further refine the list, those enzymes which 11. Oxidative phosphorylation ype00190
either uniquely consume a specific substrate or uniquely produce 12. Purine metabolism ype00230
13. Pyrimidine metabolism ype00240
a specific product in a metabolic network (choke points) [8], were
14. Alanine, aspartate and glutamate metabolism ype00250
shortlisted through choke point analysis. The analysis revealed 15. Glycine, serine and threonine metabolism ype00260
a set of 245 enzymes present in Y. pestis CO92 as potential drug 16. Cysteine and methionine metabolism ype00270
targets for the treatment of plague, with 25 of them as choke 17. Valine, leucine and isoleucine degradation ype00280
points. Many of the enzymes identified as potential drug targets by 18. Valine, leucine and isoleucine biosynthesis ype00290
19. Lysine biosynthesis ype00300
our approach, have also been identified as drug targets or putative 20. Lysine degradation ype00310
drug targets by The Center for Structural Genomics of Infectious 21. Arginine and proline metabolism ype00330
Diseases (CSGID) Consortium (http://csgid.org/csgid/), which 22. Histidine metabolism ype00340
determines the three-dimensional structures of proteins from 23. Tyrosine metabolism ype00350
24. Phenylalanine metabolism ype00360
major human pathogens. Thus, this approach can be used for a fast
25. Tryptophan metabolism ype00380
and accurate assessment of the genomes of human pathogens for 26. Phenylalanine, tyrosine and tryptophan ype00400
detecting potential drug targets. As a case study, the three- biosynthesis
dimensional structure of one of the potential targets, MurE ligase 27. beta-Alanine metabolism ype00410
enzyme was modeled in order to find a suitable inhibitor through 28. Taurine and hypotaurine metabolism ype00430
29. Selenoamino acid metabolism ype00450
docking studies. 30. Cyanoamino acid metabolism ype00460
31. D-Glutamine and D-glutamate metabolism ype00471
32. Glutathione metabolism ype00480
2. Results and discussion 33. Starch and sucrose metabolism ype00500
34. Other glycan degradation ype00511
2.1. Metabolic pathways analysis 35. Amino sugar and nucleotide sugar metabolism ype00520
36. Glycerolipid metabolism ype00561
37. Inositol phosphate metabolism ype00562
Comparative metabolic pathway analysis of H. sapiens and
38. Glycerophospholipid metabolism ype00564
Y. pestis CO92 shows that there are 24 pathways unique to the 39. Arachidonic acid metabolism ype00590
pathogen and 72 pathways common to both, as listed in Tables 1 40. alpha-Linolenic acid metabolism ype00592
and 2, respectively. 41. Sphingolipid metabolism ype00600
42. Pyruvate metabolism ype00620
43. Glyoxylate and dicarboxylate metabolism ype00630
Table 1 44. Propanoate metabolism ype00640
Metabolic pathways present in Yersinia pestis CO92 but not in the host Homo sapiens. 45. Butanoate metabolism ype00650
46. One carbon pool by folate ype00670
S. No. Metabolic pathway KEGG pathway ID 47. Methane metabolism ype00680
1. Geraniol degradation ype00281 48. Thiamine metabolism ype00730
2. gamma-Hexachlorocyclohexane degradation ype00361 49. Riboflavin metabolism ype00740
3. Benzoate degradation via hydroxylation ype00362 50. Vitamin B6 metabolism ype00750
4. Fluorobenzoate degradation ype00364 51. Nicotinate and nicotinamide metabolism ype00760
5. Novobiocin biosynthesis ype00401 52. Pantothenate and CoA biosynthesis ype00770
6. D-Alanine metabolism ype00473 53. Biotin metabolism ype00780
7. Streptomycin biosynthesis ype00521 54. Lipoic acid metabolism ype00785
8. Polyketide sugar unit biosynthesis ype00523 55. Folate biosynthesis ype00790
9. Lipopolysaccharide biosynthesis ype00540 56. Porphyrin and chlorophyll metabolism ype00860
10. Peptidoglycan biosynthesis ype00550 57. Terpenoid backbone biosynthesis ype00900
11. 1- and 2-Methylnaphthalene degradation ype00624 58. Limonene and pinene degradation ype00903
12. 1,4-Dichlorobenzene degradation ype00627 59. Nitrogen metabolism ype00910
13. Fluorene degradation ype00628 60. Sulfur metabolism ype00920
14. Benzoate degradation via CoA ligation ype00632 61. Aminoacyl-tRNA biosynthesis ype00970
15. Trinitrotoluene degradation ype00633 62. Biosynthesis of unsaturated fatty acids ype01040
16. 3-Chloroacrylic acid degradation ype00641 63. ABC transporters ype02010
17. C5-Branched dibasic acid metabolism ype00660 64. Ribosome ype03010
18. Caprolactam degradation ype00930 65. RNA degradation ype03018
19. Biosynthesis of siderophore group ype01053 66. RNA polymerase ype03020
nonribosomal peptides 67. DNA replication ype03030
20. Two-component system ype02020 68. Protein export ype03060
21. Bacterial chemotaxis ype02030 69. Base excision repair ype03410
22. Flagellar assembly ype02040 70. Nucleotide excision repair ype03420
23. Phosphotransferase system (PTS) ype02060 71. Mismatch repair ype03430
24. Bacterial secretion system ype03070 72. Homologous recombination ype03440
 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195
2.2. Identification of potential drug targets
Enzymes with a valid Enzyme Commission number from the
metabolic pathways enlisted in the Tables 1 and 2, as documented
in KEGG database [9], were analyzed in order to find out which
ones are non-homologous to any human protein. Pathogen
enzymes which do not show a significant homology to any host
protein can serve as potential drug targets, as this reduces the risk
of undesirable interactions of a potential drug with the host
organism’s proteins. Furthermore, Y. pestis enzyme which shows
significant homology to proteins proved as essential for the survival
of other bacterial species, as cataloged in the Database of Essential
Genes [10], is more likely to be essential for the survival of the
pathogen. Such enzymes are ideal candidates for drug targets, as
targeting them directly poses a threat to the survival of the path-
ogen. These enzymes are listed in Supplementary Tables 1 and 2 for
pathways unique to the pathogen and the pathways common to the
pathogen and the host, respectively. Out of 24 unique pathways,
only nine pathways were found to contain potential drug targets,
while 57 of the 72 common pathways had potential drug targets
(Fig. 1a, bee). In total, 245 unique enzymes (the enzymes which
participate in more than one metabolic pathway have been counted
only once) were identified as potential drug targets with this
methodology, 66 in unique pathways and 179 in the common
pathways.
Many of the potential targets found by this method are also
under various stages of progress at The Center for Structural
Genomics of Infectious Diseases (CSGID) Consortium for deter-
mining the three-dimensional structures of proteins from major
human pathogens. Out of the 245 unique targets proposed in our
study, 125 are already at some stage of investigation at CSGID,
classified as predicted essential, drug target function, conserved
genes and/or pathway, virulence associated factors etc. This vali-
dates the methodology used in our study with respect to experi-
mental observations. The other 120 enzymes are novel targets
predicted by the methodology adopted. Most of the potential
targets identified have homologs in other pathogenic bacteria as
revealed by the cluster of orthologous group (COG) search [11]
indicating these enzymes could serve as potential broad-
spectrum targets.
2.3. Identification of choke point enzymes
Biochemical reactions which uniquely metabolize a specific
substrate or uniquely produce a specific product in a metabolic
network in such a way that there is always a balancing reaction to
produce or utilize that specific substrate or product respectively,
are called “chokepoint reactions” [8]. As is evident from the name,
inactivation of these reactions can choke that particular meta-
bolic pathway, as there is no other alternate source for that
specific metabolite. This may pose a threat for the fitness or
survival of the organism [12]. Hence, an enzyme catalyzing
a choke point reaction can be a potential drug target, as inhibition
of its activity will effectively shut down the essential biochemical
network.
The 245 potential drug targets identified in Y. pestis CO92 were
analyzed using the Pathway Hunter Tool [13] and the Pathway tools
13.5 [14] for detecting potential choke point enzymes. Out of the
identified potential targets (Supplementary Tables 1 and 2), 25
enzymes were found to have optimally passed the choke point
analysis by both the tools used. The list of choke point enzymes
identified by this approach is given in Table 3.
Eight of these 25 choke point enzymes are found in the pepti-
doglycan biosynthesis pathway, reaffirming the fact that this
pathway is a rich source for potential drug targets in bacteria. The
187
peptidoglycan biosynthetic pathway can be divided into two stages.
The initial stage involves the formation of UDP-N-acetylmuramyl
pentapeptide which is catalyzed by different Mur enzymes, namely
MurABCDEF. The first step involves addition of phosphoenolpyr-
uvate to UDP-N-acetylglucosamine leading to formation of UDP-N-
acetylmuramic acid, catalyzed by MurA and MurB. This is followed
by the sequential addition of L-alanine, D-glutamic acid, dia-
minopimelic acid, and D-alanyl-D-alanine by the four ATP-
dependent amino acid ligases (Mur CeF), thus adding the penta-
peptide moiety to UDP-N-acetylmuramic acid [15]. The glycan
chains so formed are then cross-linked in the second stage of this
pathway. Many antimicrobial compounds have been discovered
which target the cross-linking stage in this pathway, but the early
stages of the cell wall biosynthesis process have not been explored
sufficiently as targets for antimicrobial activity [16]. The current
drugs such as Penicillin, Ampicillin, Vancomycin etc., which target
the later stages of this pathway, have been rendered nearly inef-
fective due to development of resistance in the pathogens against
these drugs. Hence, the enzyme UDP-N-acetylmuramoylalanyl-D-
glutamate-2,6-diaminopimelate ligase, or MurE ligase (6.3.2.13),
which catalyzes the addition of meso-diaminopimelic acid to UDP-
N-acetylmuramoyl-L-alanyl-D-glutamate, was chosen as a model
enzyme for homology modeling and docking studies. Being a choke
point enzyme, inhibition of this enzyme will most likely stop the
progress of the peptidoglycan biosynthesis pathway, hence inhib-
iting cell wall synthesis, which is lethal for bacteria. Due to the
common catalytic mechanism and conserved motifs of the Mur
ligases (CeF), a potential inhibitor against MurE may inhibit one or
more of these enzymes. Thus it provides a potential counter-
mechanism against development of possible resistance, as it
would require mutations in all of the involved genes simulta-
neously [15].
2.4. MurE ligase e a potential broad spectrum target
The Mur enzymes of the peptidoglycan biosynthesis pathway
constitute the majority in the list of choke point enzymes identified
in this study. MurE ligase of Y. pestis CO92 is a 52,235 Da protein
(from UniProtKB entry Q8ZIF4) of 490 amino acids length, most
likely expressed in the cytoplasm. It belongs to the Mur CDEF ligase
family and MurE ligase subfamily. This protein is enzymatically
similar to MurE from Escherichia coli. MurE in Y. pestis CO92 cata-
lyzes the addition of meso-diaminopimelic acid to the nucleotide
precursor UDP-N-acetylmuramoyl-L-alanyl-D-glutamate (UMAG),
leading to the formation of UDP-MurNAc-tripeptide in the
biosynthesis of bacterial cell-wall peptidoglycan. The peptide bond
formation is driven by ATP hydrolysis. Sequence analysis of MurE
ligase of Y. pestis CO92 using the Pfam database [17] shows that it
consists of three major domains: the characteristic Mur ligase
family catalytic domain (Pfam Acc No.PF01225), the Mur ligase
family glutamate ligase domain (Pfam Acc No. PF02875), and the
Mur ligase middle domain (Pfam Acc No. PF02875). The N and C
termini have the cytoplasmic peptidoglycan synthetase signatures
with InterPro domain database accession numbers IPR000713 and
IPR004101, respectively, along with the Mur ligase central domain
with InterPro domain database accession number IPR013221. All
the three domains are also found in the UDP-N-acetylmur-
amoylalanyl-D-glutamate-2,6-diaminopimelate ligase entry of
InterPro domain database [18] with accession number IPR005761.
Based on sequence analysis and similarity with homologous
enzymes (such as MurE ligase of E. coli 77% identity), this enzyme
is deduced to have ATP binding and amino acid ligase activity, and
is believed to be involved in cellular processes like cell cycle, cell
division, cell wall organization, peptidoglycan biosynthesis process
and regulation of cell.
188 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195

Fig. 1. Number of drug targets: (a) which do not have human homolog and have a DEG hit, present in the nine pathways unique to the pathogen w.r.t. the host, (bee) which do not
have human homolog and have a DEG hit, present in the 57 pathways common to both the pathogen and the host.

2.5. Homology modeling of MurE ligase and energy minimization
Since the structure of Y. pestis CO92 MurE ligase has not been
determined experimentally till now, homology modeling was used
to construct a three-dimensional model of the protein. A BLASTp
search of the Y. pestis CO92 MurE ligase sequence (Accession No.
YP_002345621) against the PDB database revealed MurE ligase
chain A of E. coli (PDB ID 1E8C) as the closest homolog, with 76%
identity and 99% query coverage (e-value ¼ 0.0). Hence the E. coli
protein was used as the template for homology modeling. The
alignment of the target and template sequences by using the online
program Multalin [19] is shown in Fig. 2. Eleven models of MurE
 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195 189

Table 3
Enzymes found to be acting as metabolic choke points in Yersinia pestis CO92.

S. No. Choke point enzymes EC number

1 YPO1056 e UDP-N-acetylglucosamine 2.3.1.129
acyltransferase
2 YPO3569 e UDP-N-acetylglucosamine 2.5.1.7
1-carboxyvinyltransferase
3 YPO3760 e UDP-N-acetylenolpyruvoylglucosamine 1.1.1.158
reductase
4 YPO0556 e UDP-N-acetylmuramate-L-alanine ligase 6.3.2.8
5 YPO0553-UDP-N-acetylmuramoyl-L-alanyl-D-glutamate 6.3.2.9
synthetase
6 YPO0550-UDP-N-acetylmuramoylalanyl-D-glutamate-2, 6.3.2.13
6-diaminopimelate ligase
7 YPO0551-UDP-N-acetylmuramoyl-tripeptide–D- 6.3.2.10
alanyl-D-alanine ligase
8 YPO0649 e Undecaprenyl pyrophosphate phosphatase 3.6.1.27
9 YPO0555 e Undecaprenyldiphospho- 2.4.1.227
muramoylpentapeptide
beta-N-acetylglucosaminyltransferase
10 YPO2294 e Acetolactate synthase 1 regulatory subunit 2.2.1.6
11 YPO2995 e PTS system glucose-specific transporter 2.7.1.69
subunit
12 YPO1608 e PTS system glucose-specific transporter 2.7.1.69
subunits IIBC
13 YPO0439 e Phosphopentomutase 5.4.2.7
14 YPO1298 e Bifunctional PTS system fructose-specific 2.7.1.69
transporter subunit IIA/HPr protein
15 YPO1300 e PTS system fructose-specific transporter 2.7.1.69
subunits IIBC
16 YPO1758 e PTS system, mannose-specific IIAB 2.7.1.69
component
17 YPO4068 e PTS system, mannitol-specific IIABC 2.7.1.69
component
18 YPO0837 e Putative PTS permease protein 2.7.1.69
19 YPO0834 e Putative PTS transport protein 2.7.1.69
20 YPO2569 e Phosphotransferase enzyme II, A component 2.7.1.69
21 YPO2203 e Tryptophan synthase subunit alpha 4.2.1.20
22 YPO2204 e Tryptophan synthase subunit beta 4.2.1.20
23 YPO2628 e PTS system, N-acetylglucosamine-specific 2.7.1.69
IIABC component
24 YPO1441 e Methylglyoxal synthase 4.2.3.3
25 YPO0053 e Phosphopantetheine adenylyltransferase 2.7.7.3

ligase were built by Modeller9v7 [20] using the crystal structure

coordinates of the template structure (1E8C). The Modeller
program uses the spatial constraints determined from the crystal
structure, to build a three-dimensional model of target protein with
the unknown tertiary structure. All the eleven models were then
energy minimized using the Macromodel module of Schrodinger
Software Suite 2009. The quality of the energy minimized protein
models was then validated with the PROCHECK [21], ERRAT [22]
Fig. 3. Model quality assessment: Ramachandran plot of the model generated for
Y. pestis CO92 MurE ligase protein.
and Verify3D [23] programs, available at the SAVES server. The
Ramachandran plot generated by PROCHECK for the best model
(Fig. 3) shows that not a single residue in the generated structure
falls in the disallowed region of the phiepsi dihedral angle plot,

Fig. 2. Sequence alignment for modeling: Alignment of the Y. pestis CO92 MurE ligase protein sequence (Accession No. YP_002345621) with the template sequence, E. coli MurE
ligase chain A (PDB ID 1E8C) using MultAlin.
190 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195

while 88.4% residues fall in the most favored region, thus validating
the quality of the model. The Verify3D program, which checks for
the compatibility of a three-dimensional atomic model with its
own one-dimensional amino acid sequence, validates our selected
model as having 96.95% residues with an averaged 3De1D score
greater than 0.2. The ERRAT server gives the overall quality factor
(expressed as the percentage of the protein for which the calculated
error value falls below the 95% rejection limit) of the model as
88.47. The main chain parameters plots for the model were
generated using the PROCHECK analyses server available at
PDBsum, and are shown in Fig. 4. The plots are for the Ram-
achandran plot quality, peptide bond planarity, bad non-bonded
interactions, Ca tetrahedral distortion, main chain hydrogen bond
energy and the overall G factor. These show how well the model
structure compares with well refined structures at a similar reso-
lution. Based on all these scores, the model selected was found to be
reliable for further docking studies.
Fig. 5 shows the energy minimized modeled structure of the
MurE ligase protein of Y. pestis CO92. Further, the template crystal
Fig. 5. Homology model: Yersinia pestis CO92 MurE ligase model structure generated
by Modeller9v7 program.

structure, 1E8C, was superimposed on the generated model using

the Pymol visualization software. Both the structures were found to
be significantly similar in topology, and aligned well with each
other, with a RMS value of 0.602. The active site of the protein
model, where the reaction to be inhibited takes place, was deter-
mined for docking studies. Since no experimental data is available
for Y. pestis CO92 UDP-N-acetylmuramoylalanyl-D-glutamate-2,6-
diaminopimelate ligase, the active site residues of the closest
homolog, the E. coli MurE ligase protein were used. The corre-
sponding residues were located in our model, as given in Table 4,
and assumed to be a part of the catalytic site of our protein, since
both the proteins share high identity, are evolutionarily related, and
have the same catalytic function. The residues mentioned in Table 4
were selected in the superimposed structures in Pymol to map the
active site of the E. coli protein in our model, and the active sites of
both the proteins were found to be well aligned (Fig. 6). The resi-
dues which are a part of the active site in our model structure are
Lys 115, Thr 153, Glu 178, Asp 205, His 206, Lys 220, Asn 306, Arg
337, Asp 352, His 355 and Arg 385.

2.6. Docking

The natural substrate of the enzyme, UDP-N-acetylmuramoyl-L-

alanyl-D-glutamate (PubChem CID 449538, Fig. 7a), obtained from
the PubChem Compound database, was docked with the MurE
ligase protein model using GLIDE module of Schrodinger [24]. In
GLIDE, initially the search space is narrowed with a rough posi-
tioning and scoring phase. The surviving poses then undergo
a torsionally flexible energy optimization on an OPLS-AA non-
bonded potential grid. Monte Carlo sampling is used to further
refine the best candidate poses by examining nearby torsional
minima. The Glide XP docking score (11.38 kcal/mol) as well as

Table 4
The active site residues of the template structure and the target structure of the
MurE ligase protein of E. coli and Y. pestis CO92 respectively.

Active site residues in E. coli Corresponding residues in

MurE ligase (template)
Lys 119, Thr 157, Glu 182,
His 210, Asp 209, Lys 224,
Asn 310, Arg 341, Asp 356,
Fig. 4. Model quality assessment: Main chain parameters for the Y. pestis CO92 MurE His 359, Arg 389
ligase model generated by Modeller9v7.
Yersinia pestis CO92 MurE ligase
Lys 115, Thr 153, Glu 178,
Asp 205, His 206, Lys 220,
Asn 306, Arg 337, Asp 352,
His 355, Arg 385
 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195 191

Fig. 6. Active site comparison: Active site residues of Yersinia pestis CO92 MurE ligase model superimposed with the active site residues of the template MurE ligase from E. coli (PDB
ID 1E8C) using PyMol. (RMS value of the alignment ¼ 0.602).

Fig. 7. Docked small molecules: (a) Natural substrate of MurE ligase e UDP-N-acetylmuramoyl-L-alanyl-D-glutamate (CID 449538), (b) in vivo inhibitor of Staphylococcus aureus
MurE ligase having residual activity (RA) of 12% in the presence of 1 mM compound (Humljan et al., 2006), (c) The compound obtained from the PubChem library e SET D e having
the best Glide gscore.
192 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195
the Glide emodel score (114.91 kcal/mol) for the natural substrate
were significantly high (Table 5). Docking pose analysis shows
numerous hydrogen bond interactions with the amino acid resi-
dues of the protein, many of which were with amino acids that are
part of the putative active site (Lys 115, Arg 337, His 355, Arg 385) as
mentioned in Table 5. Electrostatic interactions were observed with
many residues, including the active site residues Asp 205, Arg 385,
Lys 115, Arg 337, Glu 178, while hydrophobic interactions were seen
with Phe 302, Tyr 353, Tyr 466, and Leu 468 (Fig. 8a). Jan Humljan
et al. (2006), in their study of peptidosulfonamides as potential
inhibitors of bacterial peptidoglycan biosynthesis enzymes MurD
and MurE, reported a particular compound, designated as
compound 22 (Fig. 7b), which showed in vivo inhibition of the
Staphylococcus aureus MurE ligase activity, with an IC50 value in the
micromolar range [25]. Thus, this compound is a good inhibitor and
represents a promising starting point for further structural modi-
fications. When this inhibitor was docked against the Y. pestis CO92
MurE ligase model, it was found to have a good XP docking score
(7.53 kcal/mol) and emodel score (94.04 kcal/mol) and showed
hydrogen bonding interactions with five amino acid residues, four
of which were part of the putative active site (Lys 115, Thr 153, His
206, Arg 385) (Table 5). This compound showed electrostatic
interactions with active site residues Arg 385, Arg 337, Asp 352, Lys
115 and Glu 178. A few hydrophobic interactions were also seen
(Phe 302, Tyr 353, Ala 359) as shown in Fig. 8b. However, the best
scoring ligand was the one obtained from virtual screening of the
Set D of ligands from PubChem Database, based on similarity search
and superstructures search against the natural substrate. This
compound, (3S)-4-[[(1S,2S)-1-[(3R,4S,5R)-5-(aminomethyl)-3,4-
dihydroxyoxolan-2-yl]oxy-1-[(3S,4R)-5-(2,4-didioxopyrimidin-1-
yl)-3,4-dihydroxyoxolan-2-yl]-3-(hexylamino)-3-oxopropan-2-yl]
methylamino]-3-hydroxy-2 (methylamino) butanoic acid (Fig. 7c),
had an excellent gscore (15.39 kcal/mol) as well as emodel score
(130.94 kcal/mol). It was also involved in nine hydrogen bond
interactions with the protein. While three of these bonds were with
residues part of the putative active site (Thr 153, His 355, Arg 385),
the rest of the interactions were with nearby residues. This mole-
cule showed stronger hydrophobic interactions (Val 139, Pro 149,
Leu 468, Tyr 466, Tyr 353), and a number of electrostatic interac-
tions with the putative active site residues (Arg 385, Asp 205, Arg
337, Glu 178, Lys 115) as shown in Fig. 8c. Hence this molecule
(PubChem Compound CID 22489285) can be investigated further as
a potential inhibitor of the Y. pestis CO92 MurE ligase protein, and
a lead molecule for drug development against plague. The docking
algorithm used in this study was validated by docking the cocrys-
tallized final product, UDP-MurNAc-L-Ala-gamma-D-Glu-meso-
A(2)pm (UAGeAPI), in the template MurE PDB structure (1E8C) in
its binding site, and comparing the original crystal structure with
the docked complex using RMSD. A RMSD of 2.00  A determined
between the two structures with the cocrystallized UAGeAPI and
the redocked UAGeAPI suggests that the docking program was
successful in docking the substrate in its appropriate binding site
(Supplementary Fig. 1).
Table 5
Docking scores and hydrogen bond interactions between protein residues and ligands as determined after docking by Glide.
S. No. Compound name PubChem
compound ID (CID)
1. Natural Substrate: UDP-N-acetylmuramoyl- 449538
L-alanyl-D-glutamate
2. in vitro Inhibitor NA
3. Highest scoring ligand from docking studies 22489285
a
Indicates the residues in bold are part of the putative active site of the protein.
3. Conclusions
Emergence of multidrug resistance in pathogenic bacteria is
a major cause of concern, and there is a great need to develop new
drugs to replace the existing antibiotics. In the present study, we
have identified a set of potential drug targets in the plague-causing
bacterium, Y. pestis CO92, by performing a comparative metabolic
pathway analysis between the pathogen and its host, H. sapiens. The
bacterial enzymes which bear no significant similarity to human
proteins were identified using a BLASTp search at a stringent e-value
threshold of 0.005. Presence of homologs in the prokaryotic data-
base of essential genes suggests that an enzyme could be essential
for survival of the pathogen, and hence was taken as a criterion for
short listing the potential drug targets in Y. pestis CO92. Finally, a list
of 245 potential drug targets was arrived at after checking for
homologs in other pathogens using a cluster of orthologous groups
(COG) search for each enzyme. Of the 245 enzymes, 125 are already
under investigation at The Center for Structural Genomics of Infec-
tious Diseases (CSGID) Consortium, while 120 are novel. These
enzymes can be explored for novel drug development using rational
drug design approaches, especially 25 of them which have been
identified as choke point enzymes (Table 3). Inhibition of these
choke points can lead to choking of vital metabolic pathways in
which they are involved, which include key pathways such as
peptidoglycan biosynthesis, amino sugar and nucleotide sugar
metabolism, phosphotransferase system, biosynthesis and/or
metabolism of amino acids like lysine, glycine, serine, threonine,
aromatic amino acids etc., metabolism of sugars (galactose, pyru-
vate), purines etc. Blocking of these key pathways poses a serious
threat to the survival of the pathogen. Also, as revealed from the COG
search, many of the choke point enzymes are widely represented in
many pathogenic proteobacteria, such as Pseudomonas aeruginosa,
Salmonella typhimurium, Vibrio cholerae, Haemophilus influenzae,
Rickettsia prowazekii etc., making them potential candidates for
development of broad spectrum drugs. As a case study, we built
a homology model of MurE ligase (KEGG ID YPO0550), a choke point
enzyme involved in peptidoglycan biosynthesis as well as lysine
biosynthesis. Molecular docking studies on the protein model
against a virtual library of compounds revealed a compound (Fig. 7c),
which fared better than the natural substrate of the enzyme (Fig. 7a),
as well as an experimentally proved inhibitor of the enzyme (Fig. 7b)
in terms of docking scores as well as molecular interactions with the
enzyme active site. Thus, this ligand can be explored for further
experimental studies and lead development. Computational studies
like these can provide a rapid approach to identify novel drug targets
and develop better drugs against crucial pathogens.
4. Computational protocols
4.1. Metabolic pathway analysis for identifying potential drug targets
All the metabolic pathways of the host H. sapiens and the target
pathogen Y. pestis CO92, as documented in the KEGG database [9],
Glide XP Glide Glide Residues with which H-bond interactions are seena
gscore energy emodel
11.38 84.32 114.91 Asn 113, Lys 115, Glu 151, Asn 152,
Arg 337, His 355, Arg 385, Asn 410, Thr 138,
Arg 412, Asn 471
7.53 58.99 94.04 Lys 115, Asn 152, His 206, Arg 385, Thr 153
15.39 86.94 130.94 Thr112, Thr 153, His 355, Arg 385, Asp 409,
Asn 410, Arg 412, Gly 460, Tyr 466
 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195 193

Fig. 8. Ligand interaction diagrams showing the hydrophobic, electrostatic, hydrogen bonds and other key interactions of the ligand with the protein residues in the docked
complex, for (a) the natural substrate, UDP-N-acetylmuramoyl-L-alanyl-D-glutamate, of the MurE ligase enzyme, (b) in vivo inhibitor of Staphylococcus aureus MurE ligase (Humljan
et al., 2006), and (c) the compound obtained from the PubChem library eSET D e having the best Glide gscore (CID 22489285).

were compared. The metabolic pathways present in Y. pestis CO92
but not found in H. sapiens were identified as the unique pathways
in the pathogen with respect to the host. The pathways common to
both the organisms were analyzed separately. The enzymes
present in the unique pathways identified in Y. pestis CO92 were
then assessed for their potential as drug targets. The genes in each
KEGG pathway database entry having a valid EC number were
selected, and their sequences were retrieved from the RefSeq
sequence database of NCBI [26]. These protein sequences were
then subjected to BLASTp search against the non-redundant
database, with the e-value threshold set at 0.005, the search
being limited to human proteins by selecting H. sapiens in the
194 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195
organism field. Enzymes which have no hits below the e-value
inclusion threshold of 0.005 are thus marked as having no signif-
icant homologs in human beings and hence are shortlisted as
potential drug targets. These potential targets were then further
filtered by checking for homology against proteins present in the
Database of Essential Genes (DEG) [10]. They were subjected to
a BLASTp search against the Prokaryotic protein database of
essential genes at DEG, having essential genes from more than 10
bacterial species. The BLAST search was carried out at the lowest
possible expect value threshold of 0.0001. Any enzyme which
returns a hit in DEG is assumed to be essential for Y. pestis CO92
also, based on homology with a protein already proved to be
essential for another bacterial species. The enzymes in each unique
pathway, which do not have any significantly homologous human
protein, and show homology with an essential bacterial protein,
were then subjected to a Cluster of Orthologous Groups (COG)
search to identify homologs in other pathogens in order to hunt for
broad-spectrum drug targets. The same procedure was repeated
for all the enzymes in the pathways common to both the host and
pathogen.
4.2. Choke point analysis
The Pathway Hunter Tool (PHT) [13] was used to extract all the
choke point enzymes present in Y. pestis CO92, already present in
the tool as a model organism. The tool lists all the enzymes in the
organism ranked as choke points. The enzymes filtered so far
through a BLASTp search, a DEG blast search and COG search were
then looked for in this list of choke points, and the high ranked
enzymes were shortlisted. Another tool, the Pathway Tools 13.5
[14] was also used to extract choke points in Y. pestis CO92. The
enzymes which were designated as choke points by both the tools,
with a high rank in PHT, were selected as final potential drug
targets. One of the top hits, UDP-N-acetylmuramoylalanyl-D-
glutamate-2,6-diaminopimelate ligase (MurE) was selected for
further analysis, as a case study.
4.3. Homology modeling
The three-dimensional structure for the enzyme UDP-N-ace-
tylmuramoylalanyl-D-glutamate-2,6-diaminopimelate ligase (MurE)
of Y. pestis CO92 was built using the Modeller program (Version 9v7)
[20]. A BLASTp search of the Y. pestis CO92 MurE sequence against the
proteins available in the PDB database was carried out. Based on the
sequence identity and query coverage, MurE ligase from E. coli (PDB
ID 1E8C) was chosen as the most suitable template for homology
modeling. The primary sequences of the two proteins were aligned
using the Multalin program [19]. Eleven models were generated, and
all were energy minimized using the Macromodel module of
Schrodinger Software Suite 2009 (Schrödinger LLC, New York, USA).
The quality of all the minimized structures was determined using the
programs PROCHECK [21], ERRAT [22], and Verify3D [23], all avail-
able at the Structural Analysis and VErification Server (SAVES e
http://nihserver.mbi.ucla.edu/SAVES/). Based on these scores and
the Ramachandran Plot generated by PROCHECK, the best model was
selected for docking studies.
4.4. Docking study
4.4.1. Preparation of the target protein model
The three-dimensional model of the UDP-N-acetylmur-
amoylalanyl-D-glutamate-2,6-diaminopimelate ligase (MurE) of
Y. pestis CO92 was prepared for small molecule docking using the
Protein Preparation Wizard module of Schrodinger Software Suite
2009, which involves removal of all water molecules, addition of
hydrogen atoms to the protein structure and assignment of all atom
force field (OPSL-2005) charges and atom types. Minimizations
were performed until the average root mean square deviation of
non-hydrogen atoms reached 0.3  A.
4.4.2. Ligand data and preparation
A library of small molecules compiled from a variety of sources
was used for the docking studies.
a SET A e 1001 compounds were downloaded from the Zinc
database [27], based on similarity search against the structure
of UDP-N-acetylmuramoyl-L-alanyl-D-glutamate: meso-diam-
inopimelate ligase, the natural product of the enzyme MurE
ligase [28].
b SET B e 10,230 compounds obtained from Ligandinfo subsets
[29] after 3D database generation and filtering using Phase
module of Schrodinger Software Suite 2009.
c SET C e Compounds obtained from the Pubchem Compounds
database based on identity search against the core group of
phosphinates and sulphonamides [25].
d SET D e Compounds obtained from Pubchem Compounds
database based on similarity search and superstructures search
against the natural substrate (UDP-N-acetylmuramoyl-L-
alanyl-D-glutamate) of the MurE ligase enzyme.
Set A, Set C and Set D of compounds were also prepared like Set
B by 3D database generation using Phase module of Schrodinger
software, which includes the LigPrep tool. Also, the natural
substrate for the MurE ligase and a compound proven to inhibit the
E. coli MurE ligase in vitro [25] were prepared using the LigPrep
module of Schrodinger Software Suite 2009.
4.4.3. Docking
The prepared ligands were docked into the active site of MurE
ligase using the GLIDE (Grid-based LIgand Docking with Ener-
getics) module of Schrodinger [24]. Since the binding side or active
site residues of the Y. pestis CO92 MurE ligase are not known
experimentally, the catalytic residues of the E. coli MurE ligase [28]
were located in our protein model and used for receptor grid
generation. The grid was generated around this assumed active
site of Y. pestis CO92 MurE ligase model to define the space for
docking of ligand molecules. The large library of compounds was
first docked using the High Throughput Virtual Screening (HTVS),
followed by Standard Precision (SP) docking of all the ligands
having a Glide Score less than 6.0 in HTVS. Glide Score is used to
predict binding affinity and ranking the ligands in database
screens, while a composite scoring function emodel, which is
a combination of energy-grid score, binding affinity predicted by
Glide Score and internal strain energy (for flexible docking) for the
model is used to select the correctly docked pose [24]. The ligand
with the best SP docking scores was then further analyzed by eXtra
Precision (XP) docking. Also, the natural substrate of MurE ligase,
and the known in vitro inhibitor of E. coli MurE were docked at the
active site using XP docking. To validate the docking algorithm
used in the study, the template structure of MurE ligase from E. coli
(PDB ID 1E8C) was considered. This protein structure has the final
product of the reaction, UDP-MurNAc-L-Ala-gamma-D-Glu-meso-
A(2)pm (UAGeAPI) cocrystallized in its binding site. For validation,
this bound product was removed from the 1E8C structure and
docked again in the binding site defined by the residues
mentioned in the right hand column of Table 4. The same docking
parameters were used as in the docking studies mentioned above.
The original 1E8C structure and the docked complex were super-
imposed by the maestro program, and the RMSD of the two
structures determined.
 A. Sharma, A. Pan / European Journal of Medicinal Chemistry 57 (2012) 185e195
Acknowledgments
Authors are indebted to Centre for Bioinformatics, Pondicherry
University, India for providing infrastructure facility to carry on the
work. Research carried out at the laboratory of the Centre for
Excellence in Bioinformatics, Pondicherry University, India is fun-
ded by the Department of Information Technology (DIT) and the
Department of Biotechnology (DBT), Government of India, New
Delhi, India.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.ejmech.2012.09.018.
References
[1] R.D. Perry, J.D. Fetherston, Yersinia pestis- etiologic agent of plague, Clin.
Microbiol. Rev. 10 (1997) 35e66.
[2] B.L. Ligon, Plague: a review of its history and potential as a biological weapon,
Semin. Pediatr. Infect. Dis. 17 (2006) 161e170.
[3] M. Galimand, A. Guiyoule, G. Gerbaud, B. Rasoamanana, S. Chanteau,
E. Carniel, P. Courvalin, Multidrug resistance in Yersinia pestis mediated by
a transferable plasmid, N. Engl. J. Med. 337 (1997) 677e681.
[4] M. Galimand, E. Carniel, P. Courvalin, Resistance of Yersinia pestis to antimi-
crobial agents, Antimicrob. Agents Chemother. 50 (2006) 3233e3236.
[5] N.C. Stenseth, B.B. Atshabar, M. Begon, S.R. Belmain, E. Bertherat, E. Carniel,
K.L. Gage, H. Leirs, L. Rahalison, Plague: past, present, and future, PLoS Med. 5
(2008) e3.
[6] J. Parkhill, B.W. Wren, N.R. Thomson, R.W. Titball, M.T. Holden, M.B. Prentice,
M. Sebaihia, K.D. James, C. Churcher, K.L. Mungall, S. Baker, D. Basham,
S.D. Bentley, K. Brooks, A.M. Cerdeño-Tárraga, T. Chillingworth, A. Cronin,
R.M. Davies, P. Davis, G. Dougan, T. Feltwell, N. Hamlin, S. Holroyd, K. Jagels,
A.V. Karlyshev, S. Leather, S. Moule, P.C. Oyston, M. Quail, K. Rutherford,
M. Simmonds, J. Skelton, K. Stevens, S. Whitehead, B.G. Barrell, Genome sequence
of Yersinia pestis, the causative agent of plague, Nature 413 (2001) 523e527.
[7] World Health Organization, Human plague cases from 1989 to 2003, Wkly
Epidemiol. Rec. 79 (2004) 301e306.
[8] A.S. Rahman, D. Schomburg, Observing local and global properties of meta-
bolic pathways: ‘load points’ and ‘choke points’ in the metabolic networks,
Bioinformatics 22 (2006) 1767e1774.
[9] M. Kanehisa, S. Goto, S. Kawashima, A. Nakaya, The KEGG databases at
GenomeNet, Nucleic Acids Res. 30 (2002) 42e46.
[10] R. Zhang, Y. Lin, DEG 5.0, a database of essential genes in both prokaryotes and
eukaryotes, Nucleic Acids Res. 37 (2009) D455eD458.
[11] R.L. Tatusov, N.D. Fedorova, J.D. Jackson, A.R. Jacobs, B. Kiryutin, E.V. Koonin,
D.M. Krylov, R. Mazumder, S.L. Mekhedov, A.N. Nikolskaya, B.S. Rao,
S. Smirnov, A.V. Sverdlov, S. Vasudevan, Y.I. Wolf, J.J. Yin, D.A. Natale, The COG
database: an updated version includes eukaryotes, BMC Bioinform. 4 (2003).
[12] I. Yeh, T. Hanekamp, S. Tsoka, P.D. Karp, R.B. Altman, Computational analysis
of Plasmodium falciparum metabolism: organizing genomic information to
facilitate drug discovery, Genome Res. 14 (2004) 917e924.
195
[13] S.A. Rahman, P. Advani, R. Schunk, R. Schrader, D. Schomburg, Metabolic
pathway analysis web service (Pathway Hunter Tool at CUBIC), Bioinformatics
21 (2005) 1189e1193.
[14] P.D. Karp, S.M. Paley, M. Krummenacker, M. Latendresse, J.M. Dale, T.J. Lee,
P. Kaipa, F. Gilham, A. Spaulding, L. Popescu, T. Altman, I. Paulsen,
I.M. Keseler, R. Caspi, Pathway tools version 13.0: integrated software for
pathway/genome informatics and systems biology, Brief. Bioinform. 11
(2010) 40e79.
[15] A. El Zoeiby, F. Sanschagrin, R.C. Levesque, Structure and function of the Mur
enzymes: development of novel inhibitors, Mol. Microbiol. 47 (2003) 1e12.
[16] S. Anishetty, M. Pulimi, G. Pennathur, Potential drug targets in Mycobacterium
tuberculosis through metabolic pathway analysis, Comput. Biol. Chem. 29
(2005) 368e378.
[17] R.D. Finn, J. Mistry, J. Tate, P. Coggill, A. Heger, J.E. Pollington, O.L. Gavin,
P. Gunasekaran, G. Ceric, K. Forslund, L. Holm, E.L. Sonnhammer, S.R. Eddy,
A. Bateman, The Pfam protein families database, Nucleic Acids Res. 38 (2010)
D211eD222.
[18] S. Hunter, R. Apweiler, T.K. Attwood, A. Bairoch, A. Bateman, D. Binns, P. Bork,
U. Das, L. Daugherty, L. Duquenne, R.D. Finn, J. Gough, D. Haft, N. Hulo,
D. Kahn, E. Kelly, A. Laugraud, I. Letunic, D. Lonsdale, R. Lopez, M. Madera,
J. Maslen, C. McAnulla, J. McDowall, J. Mistry, A. Mitchell, N. Mulder, D. Natale,
C. Orengo, A.F. Quinn, J.D. Selengut, C.J. Sigrist, M. Thimma, P.D. Thomas,
F. Valentin, D. Wilson, C.H. Wu, C. Yeats, InterPro: the integrative protein
signature database, Nucleic Acids Res. 37 (2009) D211eD215.
[19] F. Corpet, Multiple sequence alignment with hierarchical clustering, Nucleic
Acids Res. 16 (1988) 10881e10890.
[20] N. Eswar, B. Webb, M.A. Marti-Renom, M. Madhusudhan, D. Eramian,
M.Y. Shen, U. Pieper, A. Sali, Comparative protein structure modeling using
MODELLER, Curr. Protoc. Protein Sci. 50 (2007) 2.9.1e2.9.31.
[21] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thorton, PROCHECK -
a program to check the stereochemical quality of protein structures, J. Appl.
Crystallogr. 26 (1993) 283e291.
[22] C. Colovos, T.O. Yeates, Verification of protein structures: patterns of
nonbonded atomic interactions, Protein Sci. 2 (1993) 1511e1519.
[23] R. Lüthy, J.U. Bowie, D. Eisenberg, Assessment of protein models with three-
dimensional profiles, Nature 356 (1992) 83e85.
[24] R.A. Friesner, J.L. Banks, R.B. Murphy, T.A. Halgren, J.J. Klicic, D.T. Mainz,
M.P. Repasky, E.H. Knoll, M. Shelley, J.K. Perry, D.E. Shaw, P. Francis,
P.S. Shenkin, Glide: a new approach for rapid, accurate docking and
scoring. 1. Method and assessment of docking accuracy, J. Med. Chem. 47
(2004) 1739e1749.

[25] J. Humljan, M. Kotnik, A. Boniface, T. Solmajer, U. Urleb, D. Blanot, S. Gobec,
A new approach towards peptidosulfonamides: synthesis of potential inhib-
itors of bacterial peptidoglycan biosynthesis enzymes MurD and MurE,
Tetrahedron 62 (2006) 10980e10988.
[26] K.D. Pruitt, T. Tatusova, D.R. Maglott, NCBI reference sequences (RefSeq):
a curated non-redundant sequence database of genomes, transcripts and
proteins, Nucleic Acids Res. 35 (2007) D61eD65.
[27] J.J. Irwin, B.K. Shoichet, ZINC  a free database of commercially available
compounds for virtual screening, J. Chem. Inf. Model. 45 (2005) 177e182.
[28] E. Gordon, B. Flouret, L. Chantalat, J. van Heijenoort, D. Mengin-Lecreulx,
O. Dideberg, Crystal structure of UDP-N-acetylmuramoyl-L-alanyl-D-
glutamate: meso-diaminopimelate ligase from Escherichia coli, J. Biol.
Chem. 276 (2001) 10999e11006.
[29] M. von Grotthuss, G. Koczyk, J. Pas, L.S. Wyrwicz, L. Rychlewski, Ligand. Info
small-molecule meta-database, Comb. Chem. High Throughput Screen. 7
(2004) 757e761.