NAD+-dependent DNA ligase (Rv3014c) from M. tuberculosis: Strategies for inhibitor design

Abstract

NAD+-dependent DNA ligases (LigA) are essential enzymes found only in bacteria and some virus species. This makes them attractive
drug targets. Based on the crystal structure of the NAD+ binding domain of the M. tuberculosis enzyme (MtuLigA) and virtual screening, we have earlier identified several novel classes of inhibitors for this enzyme. These inhibitors
bind to the adenylation domain and compete with the cofactor NAD+. Recently, we identified that the BRCT domain is essential for the enzyme activity of MtuLigA. We have used virtual screening to identify compounds from the CAP database that should potentially bind to the BRCT
domain. These will now be evaluated as inhibitors of the enzyme with a novel mechanism of action. Challenges faced in designing
specific and potent inhibitors of the enzyme which can distinguish between the human adenosine triphosphate (ATP)-dependent
ligase and MtuLigA are additionally discussed in this report. Proposed strategies for the design of potent inhibitors with desired properties
are also outlined.

Full text

Page 1

Med. Chem. Res.17, 189-198.( 2008)

NAD+ DEPENDENT DNA LIGASE (RV3014C) FROM
M. TUBERCULOSIS: STRATEGIES FOR INHIBITOR DESIGN

Divya Dube1, Vandna Kukshal1, Sandeep Kumar Srivastava1, Rama Pati Tripathi2
& Ravishankar Ramachandran1*

1Molecular & Structural Biology Division, 2Medicinal & Process Chemistry Division,
Central Drug Research Institute, P.O. Box 173, Lucknow – 226001, India
Email: ravi_anitha@yahoo.com

Abstract: NAD+ -dependent DNA ligases (LigA) are essential enzymes found only in
bacteria and some virus species. This makes them attractive drug targets. Based on the
crystal structure of the NAD+ binding domain of the M. tuberculosis enzyme (MtuLigA)
and virtual screening we have earlier identified several novel classes of inhibitors for this
enzyme. These inhibitors bind to the adenylation domain and compete with the co-factor
NAD+. Recently we identified that the BRCT-domain is essential for the enzyme activity
of MtuLigA. We used virtual screening to identify compounds from the CAP database
that should potentially bind to the BRCT domain. These will now be evaluated as
inhibitors of the enzyme with a novel mechanism of action. Challenges faced in
designing specific and potent inhibitors of the enzyme which can distinguish between the
human ATP-dependent ligase and MtuLigA are additionally discussed in this report.
Proposed strategies for the design of potent inhibitors with desired properties are also
outlined.

Introduction
Tuberculosis is one of the worst disease scourges afflicting mankind.
Mycobacterium tuberculosis, the aetiological agent of the disease, kills more than two
million people every year (Duncan 1998). Moreover, multi drug resistant (MDR)
varieties have been identified. These factors necessitate the identification of novel
therapies based on different mechanisms of action (Duncan 1998). DNA ligases use
either ATP or NAD+ as co-factors to catalyze the joining of breaks in double stranded
DNA (Lehman 1974; Engler & Richardson 1982). NAD+ -dependent ligases are
conserved enzymes in bacteria and have recently drawn attention as novel drug targets
(Srivastava et al., 2005a; 2005b). No existing drug is known to target these enzymes and
conceptually therefore designed inhibitors should also be useful against drug resistant
strains of the pathogen. We had recently solved the crystal structure of the adenylation
domain of the enzyme (Srivastava et al., 2005a) (Fig. 1) and have used it in virtual

Page 2

Med. Chem. Res.17, 189-198.( 2008)

Figure 1. Crystal structure of the adenylation domain of MtuLigA (cartoon
representation) superposed onto crystal structure of full length TfiLigA (ribbon
representation). Modeled BRCT domain of MtuLigA used in the reported virtual
screening experiments is also depicted separately.

screening to identify novel inhibitory classes of compounds (Srivastava et al., 2005a;
2005b & 2007). In this report we summarize the ongoing effort in identification of new
inhibitors of the enzyme; the challenges faced and proposed strategies to improve both
the potency and specificity of a designed inhibitor for MtuLigA. Recently, using a
specific LigA deficient E. coli strain we demonstrated (Srivastava et al., 2007) that the
BRCT domain of the enzyme is essential for bacterial viability. We undertook virtual
screening using the modeled domain from MtuLigA and have identified compounds with
the potential to bind to this domain. These molecules will be evaluated in an ongoing
program.

Results and Discussion
As part of a long range program to develop anti-TB therapies based on MtuLigA
inhibition, we have in the first instance searched for diverse compound families which
inhibit MtuLigA with several fold specificity compared to ATP-dependent ligases
including for the human DNA ligase I (Table 1).

Page 3

Med. Chem. Res.17, 189-198.( 2008)
Table 1. MtuLigA inhibitors with corresponding IC50 values in µM. Representative
structures from each compound class are depicted.

S.No Class Structure IC50 (µM)



1.


Arylamino
Compounds


46.0 ± 2.5


2.


Pyrido-
chromanone



0.04-0.1



3.


Glycosyl
ureides




9.65 ± 0.5



4.0 ± 0.3



4.



Glycosylamines



46.2 ± 1


5.


Tetracyclic
indole
derivatives




13.5 ± 0.6


The compound classes numbered 3-5 in the table were identified by our group. These
compounds possess IC50 values in the low µM range. Bacterial growth inhibition studies
using specific LigA deficient strains suggest that their observed antibacterial activity is
most likely due to inhibition of the LigA in the bacteria (Srivastava et al., 2005a; 2005b).

LigA-inhibitor interactions

Page 4

Med. Chem. Res.17, 189-198.( 2008)
No co-crystal structure with any of the compounds is known so far. However
detailed kinetic studies coupled with mutations strongly suggest that all compound
classes, except for aryl amino compounds, are competitive inhibitors and apparently bind
to the NAD+ site. Aryl amino compounds however are suggested to have a different
mode of interaction with the enzyme. They also exhibit some unwanted DNA
intercalation (Ciarrocchi et al., 1999). The highest affinity among the compound classes
is exhibited by pyridochromanones (Srivastava et al., 2007). The other classes of
compounds exhibit IC50 values in the low micro molar range. Detailed molecular docking
studies with Glycosyl uriedes and amines (Srivastava et al., 2005a; 2005b). (Fig. 2)
demonstrated that mimicking the interactions of NAD+ with the enzyme improves the
specificity/distinguishing ability of the compounds for LigA. These studies also
suggested that the inhibitors should interact with conserved residues in the binding site.













(A) (B)
Fig. 2 Glycosyl amines: compound A can distinguish between MtuLigA and ATP-
dependent ligase whereas compound B whose modeled interactions extend beyond
the NAD+ binding site cannot. Some interacting residues are indicated for clarity.
Please see reference (Srivastava et al., 2005) for detailed information.

BRCT domain and inhibitor development

Page 5

Med. Chem. Res.17, 189-198.( 2008)
LigA is a highly modular enzyme with the BRCT domain occurring at the C-
terminus of the enzyme (Fig. 1). More work is necessary to actually elucidate the role of
this domain in detail. Its deletion resulted in 3-fold reduction in activity of E. coli LigA
(Wilkinson et al., 2005) whereas the T. filiformis LigA retains activity even after its
deletion (Jeon et al., 2004). In contrast, no activity, both in vitro and in vivo, is observed
in the case of the M. tuberculosis enzyme. Even in the enzyme from viral sources, the
same situation exists, this domain is absent in the active enzymes from A. Moorei
(Sriskanda et al, 2001) and M. Sanguinipes (Lu et al., 2004), while the LigA from
mimivirus possesses a BRCT domain, essential for its activity (Benarroch & Shuman
2006).
In the quest for identifying better LigA inhibitors, there are two issues at stake
viz. the identified inhibitors should in the first instance be able to distinguish between
NAD+ and ATP -dependent ligases. In the second instance, the inhibitor should possess
high affinity (specificity) for a particular pathogen/bacterial species in contrast to being a
general anti-bacterial compound. The present crop of inhibitors mostly block the binding
of the co-factor and thereby the enzyme activity. Residues in its binding site are well
conserved (Doherty & Suh 2000). In fact, 5 out of 6 conserved sequence motifs form part
of the site (Fig. 3).

Figure 3. Conserved Sequence motifs in DNA ligases. The adenylation domain
contains five of six conserved sequence motifs in NAD+ ligases. The alignment
includes LigA encoded by M. tuberculosis (Pdb code: 1ZAU), T. filiformis (1V9P), B.
stearothermophilus (1B04), E. faecalis (1TAE) and human DNA ligase I (1X9N). The
numbers of amino acid residues separating the motifs are indicated. The active site
lysine is indicated by an arrow.

Given the conserved nature of the co-factor binding site, expectedly most of the
inhibitors exhibit some degree of general anti-bacterial activity too. Better inhibitor
development is focused on improving specificity of the compounds for MtuLigA. Two
such approaches are discussed below; one approach deals with utilizing active site water

Page 6

Med. Chem. Res.17, 189-198.( 2008)
while the other approach involves the development of inhibitors, which can bind to other
regions of the molecule, in this case the BRCT domain, and block subsequent catalytic
steps.

Strategies in inhibitor development
Water clusters in the active site of an enzyme offer a lot of promise in this
context. It is well recognized that inhibitors designed to mimic the interactions of
displaced active site water can have improved affinities of up to 20 times (Ravishankar et
al., 1997; 1999). A superposition of the available LigA structures have led to
identification of conserved water clusters in the NAD+ site (Fig. 4a).


(A) (B)
Figure 4. (A) Conserved water clusters in the active site of LigA. AMP is shown as
sticks and the water clusters which are networked amongst them are indicated with
dashed circles. Some interacting residues are shown for clarity. (B) Some fragments
identified by virtual screening from a commercial library which are predicted to
displace water clusters. These fragments will now be ‘stitched’ together and
evaluated.

Virtual screening using a commercial small compound library, CAP database,
(M/s Accelrys) identified several fragments predicted to displace and mimic the
interactions of active site water. These will now be synthesized and tested.
We had earlier identified that the BRCT domain of MtuLigA is essential for
activity (Srivastava et al., 2007) in contrast to some other characterized LigA including
those from E. coli and T. filiformis. Inhibitors designed to prevent BRCT domain

Page 7

Med. Chem. Res.17, 189-198.( 2008)
interactions with the adenylation site or DNA would conceptually be novel ones which
block other steps of the enzyme’s action. Previously, mutations in a homologous enzyme
(Feng et al, 2004) had identified two glycine residues as being important for its activity.
We accordingly modeled the MtuLigA BRCT domain and used the structure in virtual
screening experiments to identify potential binders. Compounds designed to bind to a
pocket near the corresponding glycine residues in the model (Fig. 5) should be candidates
for further testing by in vitro assays.

(A)

(B) (C)
Fig 5. (A) Sequence alignment of BRCT domains of MtuLigA & Thermus
thermophilus LigA (PDB code: 1L7B). Two conserved and essential glycine residues
are circled and indicated.
(B) Cartoon representation of the modelled MtuLigA BRCT domain. The
essential glycine residues were used as the center-of-search in in silico screening
calculations. Some potential inhibitors in both regions are also depicted.
(C) Potential binding sites on the modeled MtuLigA BRCT domain (black
ribbon) identified through ‘Binding Site Analysis’ (InsightII, M/s Accelrys, San
Diego, CA, USA). The sites are indicated by grey crosses. The modeled BRCT
domain of human pol λ (grey) is also shown to highlight possible differences in the
structures of the BRCT domain. These differences are to be exploited in the design
of species specific inhibitors. Docked monoacetylcurcumin is shown in ball-and-stick
representation.

Page 8

Med. Chem. Res.17, 189-198.( 2008)
Indeed, such compounds could be identified from the CAP database and are being
procured for testing. In another report (Takeuchi et al., 2006) specific inhibitors of the
human pol λ BRCT domain could be identified. Our modeling and docking simulations
involving a curcumin derivative, monoacetylcurcumin (Fig. 5c) suggests that the same
compound should bind to different regions of the BRCT domain in the respective
proteins.


Conclusions

LigA are important enzymes with good potential for development as novel drug
targets. The existing inhibitors, several of which were identified by our group, have
shown that these are capable of distinguishing between the human and pathogen
enzymes; this being an important step in specific inhibitor development. Given the
conserved nature of the NAD+ binding site amongst the enzyme from different bacterial
species, these inhibitors also show general anti-bacterial activity. One approach to
improve their specificity and affinity for an enzyme from a specific bacterial species is to
utilize the spatial dispositions of active site water, where inhibitors designed to mimic the
interactions of displaced water oxygen should be better than the first generation. We had
also identified earlier that the BRCT domain of MtuLigA is essential for enzyme activity
and for bacterial viability. Compounds designed to bind to the domain to prevent the
interactions of its key residues should have better specificity for a given pathogen
because of the suggested finer variations in individual enzyme action. Towards this we
have modeled the MtuLigA BRCT domain and demonstrated that there are potential
regions which can be exploited in novel inhibitor development. The modeling and
docking results involving this domain also suggests structural differences which can also
be exploited in rational inhibitor design.

Materials & Methods
Enzyme Assays:
All assays have been carried out using procedures previously reported by our
group (Srivastava et al., 2005a; 2005b; 2007).
Modeling of the BRCT domain:

Page 9

Med. Chem. Res.17, 189-198.( 2008)
Homology model for the BRCT domain was generated using MODELLER6v2
(Marti-Renom et al., 2000) where T. thermophillus BRCT domain (PDB: 1L7B) was
used as the template. The stereo-chemical quality of the model was verified using
PROCHECK (Laskowski et al., 1993) and WHAT IF (Vriend 1990). Prior to virtual
screening experiments, all the hydrogens were added to the model. The model of human
pol λ BRCT domain was also generated as reported by the group (Takeuchi et al., 2006).
Databases used:
The commercially available Ludi/CAP database (M/s Accelrys Inc.) was used for
all the virtual screening experiments.
Virtual screening protocol:
Ludi module interfaced with InsightII package (Accelrys Inc. 2000) was used to
identify fragments and design ligands that would potentially overlap with the active site
water. Ludi is based on the fragment approach whereby it saturates small fragments into
the clefts of the target sites in such a way that hydrogen bonds can be formed with the
enzyme and hydrophobic pockets are filled with hydrophobic groups. These fragments
might be linked together by using Ludi in "Link mode".
Generating Novel Ligand de novo: The binding pockets within the modelled BRCT
domain and adenylation domain were visually identified using InsightII (M/S Acclerys
Inc.). The Ludi parameters were reviewed, modified and also the Ludi library was
specified. The program was executed in the ‘Targeted_Mode’. The ‘Center of Search’
parameter with a radius sphere of 5Å was specified based on two conserved glycine
residues in the BRCT domain and also water clusters within the adenylation domain
(Pdb: 1ZAU). The docked conformations of the identified ‘Hits’ were analyzed and Ludi
scores tabulated within the modeling and simulation environment of InsightII.
AUTODOCK v. 3.05 (Morris et al., 1998) was also used in several instances for
examining the poses of potential inhibitors with the enzyme.
Hardware used:
A computer cluster consisting of SGI ORIGIN350 servers and SGI OCTANES
was used for the reported work.
Illustrations

Page 10

Med. Chem. Res.17, 189-198.( 2008)
Images were made using PyMol (http://www.pymol.org), Photoshop (Adobe
Systems) and InsightII (M/s Accelrys Inc.).

Acknowledgements
The work was funded by Council of Scientific and Industrial Research, India grant
CMM0017 and SMM0003. DD, VK & SKS acknowledge fellowships from CSIR. This is
communication number 7135 from CDRI.

References
Benarroch D & Shuman S. (2006) Characterization of mimivirus NAD+-dependent DNA
ligase. Virology 353: 135-43
Ciarrocchi G, MacPhee DG, Deady LW & Tilley L. (1999) Specific inhibition of the
eubacterial DNA ligase by arylamino compounds. Antimicrob. Agents Chemother. 43:
2766-2772.
Doherty AJ & Suh SW. (2000) Structural and mechanistic conservation in DNA ligases.
Nucleic Acids Res. 28: 4051–4058.
Duncan K. (1998) The impact of genomics on the search for novel tuberculosis drugs.
Novartis Found. Symp. 217: 228-237.
Engler MJ & Richardson CC. (1982) The Enzymes (Boyer, P. D, ed) 15: 3–29,
Academic Press, New York
Feng H, Parker JM, Lu J & Cao W. (2004) Effects of deletion and site-directed mutations
on ligation steps of NAD+-dependent DNA ligase: a biochemical analysis of BRCA1 C-
terminal domain. Biochemistry 43: 12648- 12659.
Jeon HJ, Shin HJ, Choi JJ, Hoe HS, Kim HK, Suh SW & Kwon ST. (2004) Mutational
analyses of the thermostable NAD+-dependent DNA ligase from Thermus filiformis.
FEMS Microbiol. Lett. 237: 111-118.
Laskowski RA, Rullmannn JA, MacArthur MW, Kaptein R &Thornton JM.( 1996)
AQUA and PROCHECK-NMR: programs for checking the quality of protein structures
solved by NMR. J. Biomol. NMR. 8: 477-86.
Lehman IR. (1974) DNA ligase: structure, mechanism, and function. Science 186: 790-
797.
Lu J, Tong J, Feng H, Huang J, Afonso CL, Rock DL, Barany F & Cao W. (2004)
Unique ligation properties of eukaryotic NAD+-dependent DNA ligase from Melanoplus
sanguinipes entomopoxvirus. Biochim. Biophys. Acta. 1701: 37-48

Page 11

Med. Chem. Res.17, 189-198.( 2008)

Marti-Renom MA, Stuart A, Fiser A, Sánchez R, Melo F & Sali A. (2000) Comparative
protein structure modeling of genes and genomes. Anu. Rev. Biophys. Biomol. Struct.
29: 291-325
Morris GM, Goodsell DS, Halliday RS, Huey R, Hart WE, Belew RK & Olson AJ.
(1998) Automated docking using a Lamarckian genetic algorithm and an empirical
binding free energy function. J. Comput. Chem. 9: 1639–1662
Ravishankar R, Ravindran M, Suguna K, Surolia A & Vijayan M. (1997) Crystal
structure of the peanut lectin-T-antigen complex. Carbohydrate specificity generated by
water bridges. Current Science 72: 855-861. (also: Current Science, (1999) 76: 1393.
Ravishankar R, Suguna K, Surolia A & Vijayan M. (1999) The crystal structures of the
complexes of peanut lectin with Methyl-β-galactose and N-acetyllactosamine and a
comparative study of carbohydrate binding in Gal/GalNAc specific legume lectins. Acta
Crystallogr. D 55: 1375-1382.
Sriskanda V, Moyer RW & Shuman S. (2001) NAD+-dependent DNA ligase encoded by
a eukaryotic virus. J. Biol. Chem. 276: 36100-36109.
Srivastava SK, Tripathi RP & Ravishankar R. (2005) NAD+ -dependent ligase (Rv3014c)
from M. tuberculosis H37Rv: Crystal structure of the adenylation domain and
identification of novel inhibitors. J. Biol. Chem. 280, 30273-30281.
Srivastava SK, Dube D, Tewari N, Dwivedi N, Tripathi RP. & Ravishankar R. (2005)
Mycobacterium tuberculosis NAD+ -dependent DNA ligase is selectively inhibited by
glycosylamines compared with human DNA ligase I. Nucleic Acids Res. 33: 7090-7101.
Srivastava SK, Dube D, Vandna K, Jha AK, Hajela K & Ravishankar R. (2007) NAD+ -
dependent DNA ligase (Rv3014c) from Mycobacterium tuberculosis: Novel structure-
function relationship and identification of novel inhibitors. Proteins 69: 97-111.
Takeuchi T, Ishidoh T, Iijima H, Kuriyama I, Shimazaki N, Koiwai O, Kuramochi K,
Kobayashi S, Sugawara F, Sakaguchi K, Yoshida H, Mizushina Y. (2006) Structural
relationship of curcumin derivatives binding to the BRCT domain of human DNA
polymerase lambda. Genes Cells. 11: 223-35.
Vriend G. (1990) WHAT IF: a molecular modeling and drug design program. J. Mol.
Graph 8: 52-56.
Wilkinson A, Smith A, Bullard D, Lavesa-Curto M, Sayer H, Bonner A, Hemmings A &
Bowater R. (2005) Analysis of ligation and DNA binding by Escherichia coli DNA ligase
(LigA). Biochim. Biophys. Acta 1749: 113-122