Journal Pre-proof

Molecular docking and dynamics simulations studies of OmpATb identifies four

potential novel natural product-derived anti-Mycobacterium tuberculosis compounds

Samuel K. Kwofie, Courage Adobor, Erasmus Quansah, Joana Bentil, Michael

Ampadu, Whelton A. Miller, III, Michael D. Wilson

PII: S0010-4825(20)30178-5
DOI: https://doi.org/10.1016/j.compbiomed.2020.103811
Reference: CBM 103811

To appear in: Computers in Biology and Medicine

Received Date: 29 January 2020

Revised Date: 3 May 2020
Accepted Date: 3 May 2020

Please cite this article as: S.K. Kwofie, C. Adobor, E. Quansah, J. Bentil, M. Ampadu, W.A. Miller III.,
M.D. Wilson, Molecular docking and dynamics simulations studies of OmpATb identifies four potential
novel natural product-derived anti-Mycobacterium tuberculosis compounds, Computers in Biology and
Medicine (2020), doi: https://doi.org/10.1016/j.compbiomed.2020.103811.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier Ltd.

1 Molecular docking and dynamics simulations studies of OmpATb
2 identifies four potential novel natural product-derived anti-
3 Mycobacterium tuberculosis compounds
4

5 Samuel K. Kwofiea, b, c, d *
, Courage Adobora, e, Erasmus Quansaha, Joana Bentila, Michael

6 Ampadua, Whelton A. Miller IIIc,f and Michael D. Wilsone,c

a
7 Department of Biomedical Engineering, School of Engineering Sciences, College of Basic &

8 Applied Sciences, University of Ghana, PMB LG 77, Legon, Accra, Ghana

b
9 West African Centre for Cell Biology of Infectious Pathogens, Department of Biochemistry,

10 Cell and Molecular Biology, College of Basic and Applied Sciences, University of Ghana,

11 Accra, Ghana
c
12 Department of Medicine, Loyola University Medical Center, Maywood, IL, 60153, USA

d
13 Department of Physics and Engineering Science, Coastal Carolina University, Conway, SC
14 29528
e
15 Department of Parasitology, Noguchi Memorial Institute for Medical Research (NMIMR),

16 College of Health Sciences (CHS), University of Ghana, P.O. Box LG 581, Legon, Accra, Ghana
f
17 Department of Chemical and Biomolecular Engineering, School of Engineering and Applied

18 Science, University of Pennsylvania, Philadelphia, PA, USA

19

20 *Corresponding Author: Samuel K. Kwofie; Email: [email protected]

21

22

23

24

1
25 Abstract

26 The outer membrane protein A (OmpATb) of Mycobacterium tuberculosis is a virulence factor

27 that neutralizes the host pH to impede the uptake of hydrophilic antitubercular drugs. Identifying

28 natural compounds with the potential to inhibit OmpATb could allow circumvention of the

29 porin-like activities of OmpATb. Four potential leads comprising ZINC000003958185,

30 ZINC000000157405, ZINC000000001392 and ZINC000034268676 were obtained by virtual

31 screening of 6394 diverse natural products. Characterization of the binding interactions of the

32 potential leads with OmpATb revealed nine critical residues comprising ARG86, LEU110,

33 LEU113, LEU114, ALA115, PHE142, SER145, VAL146, and PHE151. Molecular dynamics

34 simulations also revealed very stable protein-lead complexes. Most residues contributed lower

35 binding energies to the overall molecular mechanics Poisson–Boltzmann surface area (MM-

36 PBSA) binding free energies of the interactions between the molecules and OmpATb protein.

37 Induced Fit Docking (IFD) of the compounds regenerated poses of the molecular docking using

38 AutoDock Vina. These molecules could be starting templates for designing inhibitors to bypass

39 the pore mediating activities of OmpATb. Based on structural similarity, ZINC000034268676

40 was suggested as a potential scaffold for designing efflux pump inhibitors of the gate mediating

41 activities of OmpATb and may enhance the uptake of hydrophilic drugs to reduce the duration

42 time of tuberculosis treatment. Furthermore, structurally similar compounds available in the

43 DrugBank database with a similarity threshold of 0.7 have been reported to exhibit antitubercular

44 and anti-mycobacterial activities. These biomolecules can be further characterized

45 experimentally to corroborate their antitubercular activity. Also, the skeletons of the molecules

46 can be adopted as sub-structures for the design of future anti-mycobacterial drugs.

47

2
48 Keywords: Mycobacterium tuberculosis, OmpATb; natural products; virtual screening;

49 molecular docking; molecular dynamics; MM-PBSA; antitubercular compounds; induced-fit

50 docking

51 1 Introduction

52 Tuberculosis (TB) is one of the oldest known infectious disease dating back to 1000 BC and is

53 still one of the leading causes of deaths among infectious diseases [1,2]. An estimated one-third

54 of the world’s population are latently infected, of which nine million develop the infection

55 further to the active stage and about two million deaths arise each year [3,4]. Of those infected,

56 over one and half million cases occur annually in sub-Saharan Africa [5]. TB treatment protocols

57 are based on a combination of three or more of the first-line antitubercular drugs comprising

58 streptomycin, isoniazid, rifampicin, pyrazinamide and ethambutol [6,7].

59 The etiological agent of tuberculosis is Mycobacterium tuberculosis which is transmitted by

60 aerosol droplets from actively infected persons [3,8]. A person becomes infected through

61 inhalation of these aerosol droplets containing the bacteria, the inhaled bacteria enters the lungs

62 and is engulfed by the macrophages in the alveoli. These macrophages digest the bacteria by the

63 secretion of acid which results from the fusion of phagosome and lysosome inside the

64 macrophage and thereby making their inner environment acidic [1,2]. The host immune system

65 can attenuate the growth of M. tuberculosis at this stage, forming a granuloma around the

66 bacteria and causing it to enter into a state of dormancy. The host becomes latently infected for a

67 period or even a lifetime with the dormant bacteria awaiting reactivation. At this stage, the

68 bacteria cannot be transmitted [9,10]. For persons with efficient cell-mediated immunity, the

69 infection may be arrested permanently at this point [11]. However, if the host immunity cannot

70 control this initial stage of M. tuberculosis or if a latently infected person’s immune system is

3
71 compromised by some factors such as aging, HIV infection, or even by some drugs, the

72 granuloma formed around the bacteria becomes liquefied inside, giving the bacteria a suitable

73 condition to grow and multiply [9,10]. When the growth and multiplication of the bacteria

74 become uncontrollable, they burst out of the granuloma and infect other macrophages. At this

75 stage, the person becomes actively infected, showing signs and symptoms of tuberculosis and

76 can also transmit the infection [10].

77 In the macrophage, there is a lysosome and phagosome fusion which changes the pH of the inner

78 environment of the macrophage from about 6.2 to between 4.5 – 5.0 [12]. M. tuberculosis has

79 over the years developed some mechanisms to resist the pH change of the host macrophages.

80 These mechanisms which are not fully understood helps the bacteria to function and grow within

81 the host. Studies have shown that the outer membrane protein A (OmpATb), which is a pore-

82 forming protein is actively involved in these mechanisms [12-14]. OmpATb has two functions:

83 as a pore-forming protein with properties of a porin, and to help secrete ammonium into the

84 phagosomal environment to neutralize the acid [13,14]. Other studies further suggest that at low

85 pH, the OmpATb might be the only functioning porin and these pores close at this low pH

86 [13,15]. The closure of these hydrophilic pores causes the slow uptake of hydrophilic molecules

87 including current first-line drugs since the outer cell envelop is hydrophobic [13,15]. With this

88 slow drug intake of the bacterial, the time duration for the curing of TB infection is prolonged

89 than expected.

90 OmpATb has two terminal domains: the N-terminal and the C-terminal domains [3,15]. The

91 pore-forming activity and the resistance of the M. tuberculosis to acidic medium in macrophages

92 are mainly mediated by the N-terminal domain of OmpATb [3,15]. Due to the key role played by

93 OmpATb in the survival of the mycobacterium in macrophages, it has been suggested as a

94 plausible drug target [12,16,17]. Under normal conditions, the loss of OmpATb did not affect the

4
 95 growth of M. tuberculosis, but its ability to grow was decreased at reduced pH [13]. The porin-

96 like activity of OmpATb plays a major role in mitigating the low pH environment of the host,

97 making it a worthy drug target. Therefore, any molecules which have the potential to disrupt the

98 normal activity of the porin at low pH could allow the host defense mechanisms to overcome the

99 virulence of M. tuberculosis [13]. The pH mediating mechanisms could be exploited by

100 inhibiting OmpATb, which could attenuate the growth of the bacteria due to host immune

101 mechanisms. Also, once the porin-like activity of OmpATb is impeded, the uptake of other

102 hydrophilic molecules under low pH could increase. The role of outer membrane protein A

103 (OmpA) as a therapeutic target is well corroborated in other bacterial organisms like

104 Acinetobacter baumannii and Pasteurella multocida [18-22]. Therefore, we hypothesized that

105 the binding of compounds (especially from natural extracts) to OmpATb would enable the

106 circumvention of its pH mediated pore-forming activity.

107 Natural products and their derivatives have historically been invaluable as sources of

108 therapeutic agents [23,24]. Also, natural product structures have high chemical diversity,

109 biochemical specificity, and other molecular properties that make them more favorable as lead

110 compounds for drug discovery, and which serve to differentiate them from libraries of synthetic

111 and combinatorial compounds [23]. To combat the dormant phenotype acquired by M.

112 tuberculosis during infection and the reduced bacilli tolerance to front-line drugs, bioassay

113 directed methods have been adopted to screen natural products from higher plant extracts against

114 tuberculosis[25]. Marine natural products have also been screened against M. tuberculosis using

115 luciferase reporter assay and enumerated colony-forming unit (CFU) bioassays[26].

116 Computationally, natural products from the Philippines and those in Ambinter database have

117 been screened against S-adenosyl-L-homocysteine hydrolase (SAHH) protein of M. tuberculosis

118 [27]. Moreover, the structural-based virtual approach has also been used to screen commercial

5
119 libraries from Asinex database against the l-alanine dehydrogenase protein of M. tuberculosis to

120 identify novel compounds [28].

121 This study aimed to identify novel small molecules of natural origin with the potential to inhibit

122 the pore-forming activities of OmpATb by using computational structure-based drug design.

123 This involved computational protocols including virtual screening, molecular docking

124 simulations, pharmacological profiling of hits, and molecular dynamics simulation of potential

125 lead compounds.

126 2 Material and methods

127 2.1 Protein structure retrieval

128 The three-dimensional (3D) molecular structure file of the N-terminal domain of OmpATb was

129 retrieved from Protein Data Bank (PDB; http://www.rcsb.org/pdb/)[29] with PDB ID 2KGS [30].

130 2.2 Small molecules retrieval

131 Ligand molecular structural files were retrieved from eleven catalogues in ZINC15 database [31]

132 comprising AfroDB[32], AnalytiCon Discovery NP[33], Herbal Ingredients In-Vivo

133 Metabolism[34], Herbal Ingredients Targets[35], Interbioscreen Natural Compounds [36],

134 INDOFINE Natural Products[37], NPACT Database[38], NuBBEDB Natural Products[39],

135 Princeton BioMolecular Research Natural Products[40], Specs Natural Products[41] and UEFS

136 Natural Products[42]. The screening libraries were categorized into two A and B. Library A

137 contained 831 African natural products from the AfroDB database, whilst library B contained

138 5,493 small molecules from ten ZINC15 catalogues (Table 1). The compounds in library B were

139 obtained by pre-filtering using Lipinski’s rule of five (log P ≤ 5, hydrogen bond donors ≤ 5,

140 hydrogen bond acceptors ≤ 10 and molecular weight ≤ 500 Daltons) [43]. Compounds that
6
141 violated any of these rules were eliminated from downstream analysis. A total of 6,324 small

142 molecules were generated by combining libraries A and B.

143 2.3 Active site characterization

144 The putative ligand binding sites of OmpATb were computed using KVFinder [44] and

145 confirmed with MetaPocket 2.0 [45]. KVFinder applies a geometry-based method for the

146 identification of protein cavities. MetaPocket combines LIGSITEcs, PASS, Q-SiteFinder and

147 SURFNET methods to enhance the prediction of a protein ligand-binding site.

148 2.4 Protein and ligand preparation

149 The 3D crystal structure of the N-terminal domain of the OmpATb protein was prepared using

150 the AutoDockTools version 1.5.6 [46] utility script “prepare_receptor4.py”, which assigned

151 partial Gasteiger charges and added polar hydrogen atoms to the protein structure. Also,

152 AutoDockTools were used to generate a grid box of size 26Å x 30Å x 30Å around the active site

153 to provide a search space for the small molecules during the virtual screening process.

154 The 3D structures of the 6,324 small molecules were converted to “.pdbqt” file format using

155 AutoDockTools utility script “prepare_ligand4.py”. This utility script allows the addition of

156 polar hydrogens, assignment of Gasteiger charges, and set free rotatable and torsional bonds for

157 the ligands.

158 2.5 Virtual screening of libraries

159 A custom python script was used to automate the screening of the libraries by docking them into

160 the active site of the receptor using AutoDock Vina 1.1.2 [47]. The virtually screened small

161 molecules were ranked based on their binding affinity to the protein in each of the two libraries.

7
162 The results of the binding poses of all ligands were visualized using PyMOL 2.0.6 [48]. Taking

163 into consideration the binding poses and the binding affinity ranking, hits were selected from

164 each screened library for pharmacological profiling.

165 2.6 Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) prediction

166 Derek Nexus version 2.1[49] was used to predict the toxicity of the virtual screening hits for

167 three animal species: mouse, monkey and human [50]. A total toxicity weight (TTW) was

168 computed from the Derek Nexus toxicity results, which was used to assess the toxicological

169 profiles of the compounds. Also, the mutagenicity of the hits was predicted using Sarah Nexus

170 version 2.1 [51]. After combining the results from Derek Nexus and Sarah Nexus, the best

171 scoring compounds in terms of their toxicological profiles were shortlisted for the Absorption,

172 Distribution, Metabolism and Excretion (ADME) testing. The ADME profiles of the selected

173 small molecules were evaluated using ADMET Predictor 8.1[52] and SwissADME [53]. Also,

174 the Lipinski’s rule of five was used for physicochemical profiling of library A via SwissADME.

175 The ligands with reasonably good ADMET profiles were selected for downstream analysis.

176 2.7 Protein-ligand bond interactions

177 The molecular interactions between the hits and OmpATb protein were elucidated for each

178 complex using LigPlot [54].

179 2.8 Molecular dynamics simulations of the predicted lead compounds

180 Molecular dynamics (MD) simulations were performed using GROMACS version 2018 [55]

181 separately on the unbound OmpATb receptor and the OmpATb receptor complexed with the

182 potential leads. PRODRG[56] (settings: Chirality=Yes, Charges=Full and EM=No) was used to

8
183 generate the GROMACS topology files for each compound. The simulations were run using the

184 Gromos 43a1 force field in a 1.0 nm cube water solvent box and five sodium ions were added to

185 the systems to neutralize them. The systems were equilibrated at a temperature of 300 K and a

186 pressure of 1 bar. The production MD was run for 100 ns for each simulation. The root mean

187 square deviation (RMSD) and root mean square fluctuation (RMSF) graphs were generated and

188 plotted using XMGRACE[57] version 5.1.25 after the production run for each simulation.

189 2.9 Free binding energy calculation using the MM-PBSA method

190 The binding free energies of the complexes over the 100 ns MD simulations with a time step of 1

191 ns were calculated using the molecular mechanics Poisson–Boltzmann surface area (MM-PBSA)

192 method implemented in g_mmpbsa [58,59] (Supplementary Table S5). Also, the binding free

193 energy contributions of different OmpATb residues involved in the binding of each compound

194 was estimated. The results were plotted using the R programming package [60].

195 2.10 Induced-Fit Docking and Superimposition

196 Schrödinger Suite 2018-2 was used to generate the Induced Fit-Docking (IFD) [61]. The

197 potential lead compounds were re-docked with the OmpATb structure using the IFD technique.

198 The poses of the ligands from the IFD were superimposed and fitted onto those of AutoDock

199 Vina using LigAlign [62].

200 2.11 Anti-mycobacterial exploration of potential leads

201 Structural similarity searches were done at a threshold of 0.7 via DrugBank 5.0 database [63] to

202 evaluate the potential anti-mycobacterial activity and possible mechanisms of action from similar

9
203 compounds. Also, Prediction of Activity Spectra for Substances (PASS) [64] was used to further

204 evaluate the antituberculosis and anti-mycobacterial activities of the molecules.

205 3 Results and Discussion

206 3.1 Protein structure and characterization of the active site

207 The 3D structure of the N-terminal domain contained three α-helices and six ß-sheets connected

208 by loops (Figure 1a) [30]. The characterization of the active site using KVFinder [44] and

209 MetaPocket 2.0 [45] revealed various binding cavities (Supplementary Table S6). One of the

210 cavities was considered as a plausible binding pocket because of the large volume of 53.57 Å3

211 and surface area of 83.52 Å2 (Figure 1a), which makes it possible for small compounds to dock

212 firmly into it [65,66]. The active site residues comprised LE84, ARG86, ASN89, THR90,

213 VAL91, LEU93, ALA109, LEU110, ASN111, LEU113, LEU114, ALA115, VAL118, ASN119,

214 VAL120, VAL141, PHE142, THR143, SER145, VAL146, PRO147, ILE148 and PHE151

215 (Figure 1b).

216

217

10
218 Figure 1. The (a) Cartoon and (b) Surface representations of the N-terminal domain of OmpATb (PDB

219 ID: 2KGS) retrieved from the Protein Data Bank. The active site residues are in multi-colors and the

220 volume of sphere colored in magenta denotes the buried active site.

221 3.2 Virtual screening and filtering of libraries using physicochemical and ADMET profiles

222 The virtual screening results were compiled and ranked based on binding energy scores.

223 Compounds with the lowest binding energies were ranked as having the highest binding affinity

224 based on the scoring function of AutoDock Vina. The highest-ranked compounds docked firmly

225 in the binding pocket of the OmpATb protein upon visualization using PyMOL (Figure 2). Out

226 of a total of 831 small molecules in library A, 14 docked firmly in the active site pocket of the

227 protein and also did not violate any of the Lipinski’s Rule of Five. For the 5493 small molecules

228 in library B, 870 compounds docked firmly in the active site pocket. The binding affinity and the

229 results of the physicochemical profiling of predicted leads using Lipinski’s Rule of Five are

230 shown in Table 2.

231 For a good ADME profile, a compound should be absorbed easily through the gastrointestinal

232 tract so that it can be available in the systemic circulation. It should be metabolized by more than

233 one metabolism enzymes so that it can be excreted easily from the body without causing

234 undesirable adverse effects. Also, it should not inhibit an enzyme’s functions in the body so as

235 not to impede the normal biological processes [67]. Thus, compounds with logP less than 5 were

236 considered more water-soluble and have high gastrointestinal (GI) absorption. Also, compounds

237 that inhibited not more than one Cytochrome P450 (CYP) metabolism enzymes and were

238 substrate to more than one CYP metabolism enzymes were selected as compounds that were in

239 the range of drug-likeness. Applying these selection criteria, one compound from library A and

240 three compounds from library B passed these pharmacodynamics selection criteria (Table 2).

11
241 A merged hit list composed of the 14 compounds from library A and the first 35 from library B

242 were profiled using ADMET. Therefore, a total of 49 hits were obtained from the virtual

243 screening protocol (Supplementary Table S1).

244

245 Figure 2. PyMOL visualization of the top hit compounds superimposed on each other (multi-colored in

246 the middle) from the virtual screening protocol showing the hits as being docked firmly within the active

247 site pocket of the N-terminal domain of OmpATb protein (grey).

248 3.3 Toxicity prediction of hit compounds

249 The toxicity profiles of the virtual screening hits were evaluated using Derek Nexus and Sarah

250 Nexus [49,51]. Derek Nexus is a structure-activity relationship (SAR) expert knowledge-based

251 system that searches its database to determine whether a compound possesses toxicophores

252 [50,67,68]. Derek Nexus predicts toxicity endpoints into categories (Table 3) for multiple animal

12
253 species [50]. Sarah Nexus predicts the mutagenicity of chemical compounds using the self-

254 organizing hypothesis network (SOHN) based on the presence of structural fragments of these

255 compounds, which have been associated with activity or inactivity in its training dataset [70].

256 The predictions of Sarah Nexus are categorized as either positive (mutagenic), negative (non-

257 mutagenic) or equivocal (equal weights) with prediction percentage confidence.

258 3.3.1 Toxicity weight evaluation

259 The toxicity endpoints of the forty-nine virtual screening hits were predicted into seven

260 categories by Derek Nexus for three animal species (human, monkey and mouse) and were

261 assigned category weights (Table 3). The category weights of the seven categories were

262 interpolated linearly between 0 and 1 (inclusive) [Table 3]. Also, the different animal species

263 were assigned weighted values summing up to ten as follows: human = 5, monkey = 3 and

264 mouse = 2, based on the closeness of the species to human in terms of their coding DNA

265 similarity (~97% with mouse [70,71] and ~98% with monkey [72,73]). The highest weight was

266 assigned to human species, followed by monkey and mouse. For easy interpretation, the assigned

267 category weights and the weighted values of the different animal species (species weight) were

268 used to transform the qualitative toxicity results (Table 4) from Derek Nexus to quantitative

269 results (Table 5). The following conversion formula was used to calculate the total toxicity

270 weight (TTW):

= 1× 1 + 2× 2 + 3× 3 1

271 Where; C1 = mouse toxicity category weight S1 = mouse species weight

272 C2 = monkey toxicity category weight S2 = monkey species weight

273 C3 = human toxicity category weight S3 = human species weight

13
274

275 For example, the compound ZINC000003958185 had a toxicity endpoint prediction of HERG

276 channel inhibition in vitro with category Doubted (= 0.3334) for all the three animal species. By

277 substituting the values into the formula (1), TTW was calculated as follows:

= 0.3334 × 2 + 0.3334 × 3 + 0.3334 × 5 = 3.334

278 Therefore, ZINC000003958185 had a TTW of 3.334 for the toxicity endpoint prediction (Table

279 5). These TTW values which are out of ten were used to assess how toxic the compounds were,

280 with compounds having TTW weights closest to 10 considered to be more toxic and compounds

281 closest to zero considered to be less toxic. Thus, compounds with TTW above 6.0 were

282 considered to be more toxic and eliminated from downstream analysis.

283

284 3.3.2 Mutagenicity prediction

285 The total toxicity weight (TTW) evaluation and the mutagenicity of the virtual screening hits

286 were considered in the selection of compounds for downstream analysis (Table 6). Compounds

287 with prediction confidence above 50% for positive mutagenicity by Sarah Nexus were

288 eliminated. The filtering of compounds using toxicity further reduced the hits in library A to four

289 compounds and those in library B to twelve compounds, making a total of 16 small molecules

290 that complied with the toxicity test (Supplementary Tables S2 and S3).

291

292 3.4 Protein-ligand binding interactions of predicted lead compounds

293 The binding interactions of the four potential lead compounds were elucidated to evaluate the

294 important intermolecular bonds involved in each of the complexes. ZINC000003958185 formed

14
295 hydrophobic bonds with six OmpATb residues comprising ARG86, LEU110, ASN111, LEU113,

296 PHE142 and VAL146, and one hydrogen bond with SER145 of bond length 2.45Å (Figures 3 and

297 4a). ZINC000000157405 formed hydrophobic bonds with eight residues ILE84, LEU114,

298 ALA115, PHE142, SER145, VAL146, ILE148 and PHE151 (Supplementary Figure S1a).

299 ZINC000000001392 formed hydrophobic bonds with six residues ARG86, LEU110, LEU113,

300 VAL118, PHE142 and PHE151 (Supplementary Figure S1b). Additionally, ZINC000034268676

301 formed hydrophobic bonds with six residues ASN89, LEU113, LEU114, THR143, VAL146 and

302 PRO147, and two hydrogen bonds with ARG86 and ALA115 with bond lengths of 3.07Å and

303 3.27Å, respectively (Figure 4b). Of all the binding interactions, nine residues ARG86, LEU110,

304 LEU113, LEU114, ALA115, PHE142, SER145, VAL146 and PHE151 interacted with at least

305 two of the ligands.

306

307

308

15
309 Figure 3. PyMOL visualization of the predicted top lead compound, ZINC000003958185 (colored in

310 brown sticks) docked into the predicted active site (surrounded by sticks) of OmpATb protein (cartoon

311 representation).

16
312

313 Figure 4. Binding interactions between OmpATb protein residues and two of the potential lead

314 compounds (joined with purple bonds): (a) ZINC000003958185 and (b) ZINC000034268676. The red

17
315 dotted lines indicate hydrophobic contacts and the green dotted lines indicate hydrogen bonds with bond

316 lengths in angstroms.

317

318 3.5 Molecular dynamics simulations of the predicted lead compounds

319 The conformational stability of the native OmpATb protein and its complexes with the predicted

320 leads were evaluated over 100 ns simulation time. The root mean square deviation (RMSD)

321 values obtained for all the simulations rose sharply to within a range of 0.4 – 0.07 nm over the

322 first 5 ns and stayed within this range for the rest of the simulation time (Figure 5). This narrow

323 RMSD range indicated the conformational stability of the protein and its complexes over the

324 simulation time. Although, ZINC000000157405 complex experienced fewer fluctuations after 50

325 ns. Generally, there were not many fluctuations in the residues as shown in the root mean square

326 fluctuation (RMSF) plot (Figure 6), except for the two terminal regions and predicted active site

327 cavity. The loop at the C-terminus region of the protein experienced more fluctuations, whereas

328 the loop at the N-terminus region and the central residues (between 115 and 120) in the active

329 site cavity experienced fewer level of fluctuations (Figures 1a and 6). These show that the

330 residues in the C-terminal loop region were more flexible than other residues. Also, the minimal

331 fluctuations of residues in the active cavity reveal fewer residue flexibility indicative of ligand-

332 complex stability [75-78].

18
333

334 Figure 5. RMSD graphs for the backbone atoms of the protein after 100 ns MD simulations for the

335 OmpATb protein (black) and the OmpATb protein complexed with ZINC000003958185 (blue),

336 ZINC000000157405 (green), ZINC000000001392 (red) and ZINC000034268676 (yellow).

337

19
338

339 Figure 6. RMSF graphs for the protein residues after 100 ns MD simulations for the OmpATb protein

340 (black) and OmpATb protein complexed with ZINC000003958185 (blue), ZINC000000157405 (green),

341 ZINC000000001392 (red) and ZINC000034268676 (yellow).

342 3.6 Free binding energy calculation using the MM-PBSA method

343 Molecular mechanics Poisson–Boltzmann surface area (MM-PBSA) method is a popular

344 approach used in estimating the free binding energies of ligands to macromolecules. This method

345 has been applied to several protein-ligand systems with varying degrees of success [79,80]. The

346 binding free energy components for each complex of the OmpATb with the ligands were

347 calculated over 100 ns molecular dynamics simulation at a time step of 1 ns (Table 7). The

348 predicted lead from AfroDB, ZINC000034268676 had the lowest binding free energy (-126.238

349 kJ/mol), which was largely contributed by van der Waal forces. The other compounds,

350 ZINC000003958185, ZINC000000157405 and ZINC000000001392 had binding free energies of

20
351 -31.688 kJ/mol, -118.821 kJ/mol and -102.356 kJ/mol, respectively, which were also largely

352 contributed by van der Waal forces. Therefore, per the MM-PBSA energy calculations,

353 ZINC000034268676 established stronger binding interactions with the OmpATb protein.

354 The MM-PBSA binding free energies of the interactions between the predicted leads and

355 OmpATb protein were decomposed into the energy contribution of each protein residue, to

356 evaluate the critical binding residues, which are residues with higher energy contributions to the

357 overall binding energy. The critical residues were LEU82, ILE84, SER85, ARG86, VAL91,

358 LEU93, PHE97, LEU106, MET107, ALA109, LEU110, ASN111, LEU113, LEU114, ALA115,

359 PRO116, VAL118, ASP122, ILE124, LEU133, ASP134, ALA138, GLU139, PRO140,

360 VAL141, PHE142, THR143, ALA144, SER145, VAL146, ILE148, PRO149, ASP150, PHE151,

361 GLY152, LEU153, LEU162, VAL175, ALA178, ALA179, TRP183, PRO199 and PRO200

362 (Figure 7, Supplementary Table S4 and Supplementary Figure S2). The predicted active site

363 residues contributed the highest energies to the MM-PBSA binding free energies between the

364 OmpATb protein and the four predicted lead compounds. Also, the nine residues comprising

365 ARG86, LEU110, LEU113, LEU114, ALA115, PHE142, SER145, VAL146 and PHE151,

366 which formed binding interactions with at least two of the potential lead compounds from the

367 molecular dockings, were among the highest contributors to the MM-PBSA binding free

368 energies. All the residues contributed highly negative binding energies except residue ARG85,

369 which contributed positive binding energies in all cases. MM-PBSA results were used to validate

370 the molecular docking since it is an efficient metric for distinguishing between good binders and

371 weak ones [81].

372

373

21
374

375 Figure 7. A plot of MM-PBSA binding free energy decomposition of each OmpATb residue contribution

376 to the binding interaction between OmpATb and ZINC000034268676. The predicted active site residues

377 are colored in red.

378 3.7 Comparison of the binding of the predicted leads to those of known TB drugs

379 The binding affinity between the predicted leads and OmpATb were compared to those of known

380 TB drugs and targets. The binding energies of known TB drugs docked against receptors

381 inhibited by these drugs range from -4.2 to -7.3 kcal/mol (Supplementary Figure S4 and Table

382 S7), compared to those of the potential leads that range from -5.3 to -7.3 (Table 2). Rifampicin

383 had the highest binding energy of -7.3 kcal/mol amongst the known drugs, which was the same

384 as ZINC000003958185. The conformational stability of the rifampicin-receptor complex was

385 obtained using RMSD computations over 100 ns. There were fluctuations in RMSD from 0 ns

22
386 until after 30 ns when the complex appeared to be fairly stable (Supplementary Figure S5A). The

387 binding free energy obtained from the MM-PBSA calculations for the complex of rifampicin and

388 DNA-directed RNA polymerase subunit beta was -1345.984 +/- 63.247 kJ/mol (Supplementary

389 Table S8 and Figures S5B and C), which was lower than those of the predicted leads and

390 OmpATb (Table 7). This implies that rifampicin showed stronger binding interactions to its

391 receptor than the predicted leads and OmpATb. Nevertheless, the predicted leads demonstrated a

392 good binding affinity against OmpATb.

393

394 3.8 Induced-fit docking and superimposition of ligands

395 Molecular docking protocol is a plausible means to predict the binding poses and affinities of

396 docked ligands. However, this technique is constrained by efficient modeling of protein

397 flexibility, causing some inaccuracies in predicting the most plausible active site geometries and

398 poses of the docked ligands [82]. To account for these molecular docking constraints, Induced-

399 Fit Docking (IFD) techniques account for the movement of the side chains and consider both the

400 ligands and protein as flexible. IFD was used to re-dock the predicted lead compounds against

401 the active site of OmpATb. The IFD protocol reproduces the most plausible binding pose of a

402 ligand in the active site with very small deviations compared to experimental binding poses

403 [83,84]. The IFD output structures were superimposed onto those of AutoDock Vina to compare

404 the RMSDs of both the ligand poses and the protein atoms. The IFD generated protein structures

405 had minimal atomic deviations compared to the original protein structure (Table 8 and

406 Supplementary Figure S3). Minimal structural deviations were observed between the IFD and

407 the pre-IFD proteins, which suggest that the predicted leads docked firmly into the active pocket.

408 Also, the RMSDs obtained from the overlap of the ligands from both the IFD and AutoDock

23
409 Vina docking protocols showed minimal deviations, as two compounds had RMSD values below

410 2.0 Å (a generally accepted threshold) and the other two compounds had slightly higher RMSD

411 values than the 2.0 Å (Table 8, Figure 8) [85]. This showed that the poses from the molecular

412 docking protocol can easily be regenerated using IFD. However, the superimposition of the

413 ligands from the IFD and virtual screening (Figure 8) showed better fitting of these compounds

414 onto each other despite the slightly higher RMSDs of the other two compounds. These similar

415 binding poses generated by the two different docking protocols provide further confidence to the

416 predicted lead compounds as plausible lead-like molecules worthy of further exploration using

417 wet laboratory assays.

418

419

24
420

421 Figure 8. Superimposition and fitting between the ligand poses obtained with Induced Fit Docking (red)

422 and AutoDock Vina docking (green) for (a) ZINC000003958185, (b) ZINC000000157405, (c)

423 ZINC000000001392 and (d) ZINC000034268676.

424

25
425 3.9 Exploring the anti-mycobacterial potential of the predicted leads

426 To support the computationally identified molecules in this study, chemical structure similarity

427 searches were carried out via DrugBank 5.0 database [63]. Analogs or similar compounds with

428 anti-mycobacterial or antitubercular activities were evaluated. Also, Prediction of Activity

429 Spectra for Substances (PASS) [64] was used to evaluate the antitubercular and anti-

430 mycobacterial activities of compounds.

431 The compound ZINC000003958185, known as 8-Hydroxy-3,4-dihydro-2-quinolinone or 8-

432 Hydroxy-3,4-dihydroquinolin-2(1H)-one, is a quinoline analog. Quinolines and its derivatives

433 are plausible scaffolds for antitubercular drug development [86,87]. Therefore,

434 ZINC000003958185 is predicted as a potentially novel antitubercular quinoline derivative.

435 A search with ZINC000000157405 yielded five compounds belonging to the benzenoids class,

436 among these compounds was an experimental drug known as bis(4-hydroxyphenyl)methanone,

437 with the highest structural similarity score of 0.793 and was known to have activity against

438 Lanosterol 14-alpha demethylase protein of M. tuberculosis [88]. With the high structural

439 similarity between ZINC000000157405 and bis(4-hydroxyphenyl)methanone, this demethylase

440 is likely a target of ZINC000000157405. Therefore, ZINC000000157405 could be targeting both

441 OmpATb and Lanosterol 14-alpha demethylase receptors, making it a more probable potent

442 antitubercular agent.

443 A search with ZINC000000001392 yielded six compounds belonging to the benzenoids class.

444 None of these six similar compounds had been associated with tuberculosis in the DrugBank

445 database. However, an analog of ZINC000000001392 known as p-hydrobenzoate, when

446 combined with other antitubercular drugs had higher potency against Mycobacterium smegmatis,

447 which is a rapidly growing surrogate of M. tuberculosis [89]. ZINC000000001392 is a

26
448 potentially novel hydrobenzoate analog that may also show high potency when combined with

449 other antitubercular drugs to reduce the duration of tuberculosis treatment. Furthermore,

450 ZINC000000001392 was predicted to possess anti-tuberculosis activity with probability of

451 activity (Pa) and probability of inactivity (Pi) of 0.436 and 0.022, respectively, Also,

452 ZINC000000001392 was predicted as an anti-mycobacterial with Pa and Pi of 0.400 and 0.039,

453 respectively. Since Pa > Pi in both cases, ZINC000000001392 is a potential scaffold for the

454 development of anti-tuberculosis and anti-mycobacterial drugs [90].

455 A search with ZINC000034268676 retrieved two similar structures which are terpenoids

456 comprising Farnesol and Squalene with similarity scores of 0.8 and 0.75, respectively. Farnesol

457 is a trien-1-ol derivative whilst ZINC000034268676 is a dien-1-ol derivative. Farnesol is a

458 potential efflux pump inhibitor in M. smegmatis [91], a surrogate model of M. tuberculosis.

459 Therefore, it is worthy to explore ZINC000034268676 as a baseline scaffold to design efflux

460 pump inhibitors of the gating activities of the OmpATb receptor of M. tuberculosis. Also,

461 Farnesol and squalene have been shown to inhibit the growth of M. tuberculosis up to 99% at

462 100 µg/mL concentration in vitro [92]. Squalene has also been shown to possess M. tuberculosis

463 inhibition at 100 µg/mL concentration in vitro [93]. Generally, terpenoid derivatives have anti-

464 mycobacterial activity against M. tuberculosis [94]. With the high structural similarity scores

465 between ZINC000034268676 and Farnesol as well as squalene, ZINC000034268676 could

466 potentially exhibit similar mechanisms of action against M. tuberculosis.

467 Interestingly, the predicted leads and structurally similar compounds have been associated with

468 anti-mycobacterial activity which makes them plausible compounds or scaffolds for

469 antitubercular drug design. Also, this study supports the need to exploit the pore-forming

470 activities of OmpATb for novel antitubercular drug discovery.

27
471 3.10 Drug-likeness evaluations of the predicted leads

472 As previously stated, all compounds which violated Lipinski’s rule of five were eliminated from

473 the virtual screening. We further evaluated the drug-likeness of the predicted leads using other

474 metrics such as those of the Comprehensive medicinal chemistry database (CMC), Modern drug

475 data report (MDDR), World drug index (WDI), Ghose, Veber, and Egan violations

476 (Supplementary Table S9). These drug-likeness metrics have been comprehensively described

477 elsewhere [53,95]. ZINC000003958185, ZINC000000001392, and ZINC000034268676 violated

478 none of the CMC, Veber, Ghost, and Egan rules. ZINC000000157405 violated one of the CMC

479 and Ghose rules, whilst ZINC000034268676 violated two of WDI like rules. All the four

480 compounds violated two of the MDDR like rules. Since the compounds complied with most of

481 the drug-like indices, the predicted leads have the propensity to be drug-like. Nevertheless,

482 compliance of any of these metrics does not necessarily render a compound drug-like [96].

483 3.11 Implications of the study

484 Historically, natural products and their derivatives have been invaluable sources of
485 therapeutic agents because of their high chemical diversity, biochemical specificity, and other
486 essential properties [23,97,98]. Our study has prioritized four novel natural compounds out of
487 over 6,000 different molecules with the potential to inhibit the pore-forming mechanism of
488 OmpATb. The methodology adopted involved carefully chosen pharmacoinformatics techniques
489 that allow the validation of the predicted leads using molecular dynamics protocols including
490 MM-PBSA and IFD. Furthermore, rigorous pharmacological profiling was used to evaluate the
491 physicochemical and pharmacokinetic properties. These molecules have been predicted as
492 anti-mycobacterial as well as antitubercular and are worthy of experimental validation
493 [83,84]. Even though other M. tuberculosis receptors have been exploited as drug targets, this
494 work primarily focused on the screening of natural compounds computationally against
495 OmpATb. OmpATb is a virulence factor that mediates the pH-dependent pore activities of the
496 bacteria. Inhibiting the OmpATb prevents the bacteria from neutralizing the acidic environment
497 of the host cells, thereby impeding the survivability of the bacteria. Blocking the OmpATb could

28
498 allow the passage of inhibitors since they can bypass the difficult impermeable outer membrane
499 [16]. Also, one of the predicted leads ZINC000034268676 shares significant structural similarity
500 with Farnesol which is a potential efflux pump inhibitor in M. smegmatis [91], a surrogate model
501 for M. tuberculosis. Therefore, the potential exists for ZINC000034268676 to be explored as an
502 efflux pump inhibitor of the gating activities of OmpATb receptor of M. tuberculosis.
503 Experimental assays include the screening of these compounds against cell lines of M.
504 tuberculosis to ascertain their inhibition constants and against human cell lines to evaluate
505 their toxicity [99]. The potential lead compounds can be purchased or synthesized for
506 experimental validation. The stringent thresholds applied in the in silico analysis and the
507 promising results obtained so far shows a higher likelihood of the templates being used as
508 antitubercular fragments. Also, these fragments could form scaffolds for the de novo design of
509 new inhibitors against OmpATb. Chemically similar structures, derivatives, or analogs can be
510 explored in experimental assays in cases where the actual compounds are not readily available
511 [100–103]. The reported work consolidates existing efforts towards discovery of
512 antitubercular agents and has the potential to greatly reduce the high cost and time consuming
513 low-throughput experimental assays.

514 4 Conclusion

515 OmpATb is a crucial target since it enables M. tuberculosis to survive in the harsh acidic

516 environment of the macrophages by impeding the uptake of hydrophilic compounds, including

517 some antitubercular molecules. This study predicted four novel natural products

518 ZINC000003958185, ZINC000000157405, ZINC000000001392, and ZINC000034268676,

519 which could be utilized as templates for the design of potential OmpATb inhibitors.

520 ZINC000034268676 was specifically suggested as a potential scaffold for designing efflux pump

521 inhibitors of OmpATb gating. These potential leads have appreciably high binding affinity to

522 OmpATb and exhibited similar active site binding pose when re-docked using the IFD protocol.

523 Also, the compounds had reasonably good pharmacological profiles with negligible toxicity.

524 Molecular dynamics simulations showed minimal deviations and fluctuations in the backbone

29
525 atoms between the unbound OmpATb receptor and its complexes. Most residues contributed to

526 the overall low binding free energies of the interactions of the predicted lead compounds with the

527 OmpATb protein via MM-PBSA. This suggests that the complexes were stable experimentally.

528 Furthermore, structurally similar compounds have been reported to have antitubercular and anti-

529 microbial activities. The scaffolds of the molecules merit further studies for antitubercular drug

530 development.

531 5 List of abbreviations

532 TB: Tuberculosis

533 HIV: Human immunodeficiency virus

534 OMPATB: Outer membrane protein A, Tuberculosis

535 ADMET: Absorption, Distribution, Metabolism, Excretion and Toxicity

536 3D: Three-dimensional

537 PASS: Prediction of activity spectra for substances

538 CFU: Enumerated colony-forming unit

539 SAHH: S-adenosyl-L-homocysteine hydrolase

540 AfroDB: Dataset of natural products from African flora

541 PDBQT: Protein Data Bank, Partial Charge (Q), & Atom Type (T)

542 RCSB: Research Collaboratory for Structural Bioinformatics

543 TTW: Total toxicity weight

544 GI: Gastrointestinal

30
545 PDB: Protein Data Bank

546 RMSD: Root Mean Square Deviation

547 RMSF: Root mean square fluctuation

548 TPSA: Topological Polar Surface Area

549 LogP: Logarithm of the octan-1-ol/water partition coefficient

550 MD: Molecular dynamics

551 MM-PBSA: Molecular mechanics Poisson-Boltzmann surface area

552 IFD: Induced fit docking

553 SAR: Structure-activity relationship

554 SOHN: Self-organizing hypothesis network

555 DNA: Deoxyribonucleic acid

556

557 6 Supplementary Materials

558 Figures S1–S5, and Tables S1–S9 are available as supplementary materials accessible online.

559 7 Author Contributions

560 S.K.K. and M.D.W. conceptualized the research project. Data analysis was predominantly

561 undertaken by S.K.K., M.D.W., C.A., E.Q., J.B., and M.A. with inputs from W.A.M.III on

562 molecular dynamics simulations. S.K.K., M.D.W., C.A., E.Q., J.B., and M.A. co-wrote the first

563 draft. All authors contributed to the revision of the drafts and agreed on the final version of the

564 manuscript before submission.

31
565 8 Funding
566 The study was not funded.

567 9 Acknowledgments
568 The authors express their gratitude to the faculty members of the Department of Biomedical
569 Engineering, University of Ghana, for all advice on the project. The West African Centre for Cell
570 Biology of Infectious Pathogens (WACCBIP) at University of Ghana made available Zuputo, a
571 Dell EMC high performance computing cluster for the molecular dynamics simulations.

572 10 Conflicts of Interest

573 The authors declare no conflicts of interest.

574 11 References

575 [1] S. Brighenti and M. Lerm, “How Mycobacterium tuberculosis Manipulates Innate and Adaptive

576 Immunity : New Views of an Old Topic Manipulates Innate and Adaptive Immunity – New Views

577 of an Old Topic,” 2012.

578 [2] E. A. Talbot and B. J. Raffa, “Mycobacterium tuberculosis,” in Molecular Medical Microbiology:

579 Second Edition, vol. 3, Elsevier Ltd, 2014, pp. 1637–1653.

580 [3] S. Flores-Villalva, E. Rogríguez-Hernández, Y. Rubio-Venegas, J. G. Cantó-Alarcón, and F.

581 Milián-Suazo, “What Can Proteomics Tell Us about Tuberculosis?,” vol. 25, no. 8, pp. 1181–

582 1194, 2015.

583 [4] K. K. Addo, P. Caulley, I. Mensah, M. Minamikawa, C. Lienhardt, and F. A. Bonsu, “SPECIAL

584 ARTICLE DIAGNOSIS OF TUBERCULOSIS IN GHANA : THE ROLE OF LABORA- TORY

585 TRAINING,” vol. 44, no. 1, pp. 31–36, 2010.

586 [5] A. Zumla et al., “Tuberculosis treatment and management--an update on treatment regimens,

587 trials, new drugs, and adjunct therapies.,” Lancet. Respir. Med., vol. 3, no. 3, pp. 220–34, Mar.

32
588 2015.

589 [6] S. Caño-Muñiz, R. Anthony, S. Niemann, and J. W. C. Alffenaar, “New approaches and

590 therapeutic options for mycobacterium tuberculosis in a dormant state,” Clinical Microbiology

591 Reviews, vol. 31, no. 1. American Society for Microbiology, 01-Jan-2018.

592 [7] C. C. Leung, H. L. Rieder, C. Lange, and W. W. Yew, “Treatment of latent infection with

593 Mycobacterium tuberculosis: Update 2010,” European Respiratory Journal, vol. 37, no. 3.

594 European Respiratory Society, pp. 690–711, 01-Mar-2011.

595 [8] R. Cloete, E. Oppon, E. Murungi, W. Schubert, and A. Christoffels, “Resistance related metabolic

596 pathways for drug target identification in Mycobacterium tuberculosis,” BMC Bioinformatics, pp.

597 1–10, 2016.

598 [9] G. Ravi Kr., B. Kumar, B. Deepa, K. Katoch, M. Kalyan, and R. Srivastava, “Differentially

599 Expressed Proteins in Response to Resuscitation of NonCulturable Cells of Mycobacterium

600 tuberculosis H37Rv: Potential New Drug Targets,” 2015.

601 [10] P. K. Drain et al., “Incipient and subclinical tuberculosis: A clinical review of early stages and

602 progression of infection,” Clinical Microbiology Reviews, vol. 31, no. 4. American Society for

603 Microbiology, 01-Oct-2018.

604 [11] P. P. Salvatore and Y. Zhang, “Tuberculosis: Molecular Basis of Pathogenesis,” in Reference

605 Module in Biomedical Sciences, Elsevier, 2017.

606 [12] O. H. Vandal, C. F. Nathan, and S. Ehrt, “Acid Resistance in Mycobacterium tuberculosis ᰔ ,” vol.

607 191, no. 15, pp. 4714–4721, 2009.

608 [13] C. Raynaud et al., “The functions of OmpATb , a pore-forming protein of Mycobacterium

609 tuberculosis,” vol. 46, pp. 191–201, 2002.

610 [14] H. Song et al., “Expression of the ompATb operon accelerates ammonia secretion and adaptation

611 of Mycobacterium tuberculosis to acidic environments,” vol. 80, no. 4, pp. 900–918, 2012.

33
612 [15] A. Alahari, N. Saint, S. Campagna, and V. Molle, “The N-Terminal Domain of OmpATb Is

613 Required for Membrane Translocation and Pore-Forming Activity in Mycobacteria ᰔ ,” vol. 189,

614 no. 17, pp. 6351–6358, 2007.

615 [16] M. Niederweis, O. Danilchanka, J. Huff, C. Hoffmann, and H. Engelhardt, “Mycobacterial outer

616 membranes: in search of proteins.,” Trends Microbiol., vol. 18, no. 3, pp. 109–16, Mar. 2010.

617 [17] V. Molle et al., “pH-dependent pore-forming activity of OmpATb from Mycobacterium

618 tuberculosis and characterization of the channel by peptidic dissection,” Mol. Microbiol., vol. 61,

619 no. 3, pp. 826–837, Aug. 2006.

620 [18] X. Vila-Farrés et al., “Combating virulence of Gram-negative bacilli by OmpA inhibition,” Sci.

621 Rep., vol. 7, no. 1, Dec. 2017.

622 [19] A. M. Viale and B. A. Evans, “Microevolution in the major outer membrane protein OmpA of

623 Acinetobacter baumannii,” bioRxiv, p. 711606, Jul. 2019.

624 [20] A. Kubera, A. Thamchaipenet, and M. Shoham, “Biofilm inhibitors targeting the outer membrane

625 protein A of Pasteurella multocida in swine,” Biofouling, vol. 33, no. 1, pp. 14–23, Jan. 2017.

626 [21] D. Nie et al., “Outer membrane protein A (OmpA) as a potential therapeutic target for

627 Acinetobacter baumannii infection,” Journal of Biomedical Science, vol. 27, no. 1. BioMed

628 Central Ltd., p. 26, 18-Jan-2020.

629 [22] R. Parra-Millán et al., “Synergistic activity of an OmpA inhibitor and colistin against colistin-

630 resistant Acinetobacter baumannii: mechanistic analysis and in vivo efficacy.,” J. Antimicrob.

631 Chemother., vol. 73, no. 12, pp. 3405–3412, 2018.

632 [23] A. A. Siddiqui, F. Iram, S. Siddiqui, and K. Sahu, “Role of natural products in drug discovery

633 process,” Int. J. Drug Dev. Res., vol. 6, no. 2, pp. 172–204, 2014.

634 [24] S. Mushtaq, B. H. Abbasi, B. Uzair, and R. Abbasi, “Natural products as reservoirs of novel

635 therapeutic agents,” EXCLI Journal, vol. 17. Leibniz Research Centre for Working Environment

34
636 and Human Factors, pp. 420–451, 04-May-2018.

637 [25] C. E. Salomon and L. E. Schmidt, “Natural Products as Leads for Tuberculosis Drug

638 Development,” Curr. Top. Med. Chem., vol. 12, no. 7, pp. 735–765, Mar. 2012.

639 [26] C. Rodrigues Felix et al., “Selective killing of dormant Mycobacterium tuberculosis by marine

640 natural products,” Antimicrob. Agents Chemother., vol. 61, no. 8, pp. 1–14, 2017.

641 [27] A. R. B. Sampaco and J. B. Billones, “Virtual screening of natural products, molecular docking

642 and dynamics simulations on m. tuberculosis S-adenosyl-L-homocysteine hydrolase,” Orient. J.

643 Chem., vol. 31, no. 4, pp. 1859–1865, 2015.

644 [28] S. Saxena, P. B. Devi, V. Soni, P. Yogeeswari, and D. Sriram, “Identification of novel inhibitors

645 against Mycobacterium tuberculosis l-alanine dehydrogenase (MTB-AlaDH) through structure-

646 based virtual screening,” J. Mol. Graph. Model., vol. 47, pp. 37–43, 2014.

647 [29] S. K. Burley et al., “RCSB Protein Data Bank: biological macromolecular structures enabling

648 research and education in fundamental biology, biomedicine, biotechnology and energy.,” Nucleic

649 Acids Res., vol. 47, no. D1, pp. D464–D474, Jan. 2019.

650 [30] V. Molle et al., “Structure of the Mycobacterium tuberculosis OmpATb protein: A model of an

651 oligomeric channel in the mycobacterial cell wall ´,” Proteins Struct. Funct. Bioinforma., vol. 79,

652 no. 2, pp. 645–661, 2011.

653 [31] T. Sterling and J. J. Irwin, “ZINC 15--Ligand Discovery for Everyone.,” J. Chem. Inf. Model., vol.

654 55, no. 11, pp. 2324–37, Nov. 2015.

655 [32] F. Ntie-Kang et al., “AfroDb: A Select Highly Potent and Diverse Natural Product Library from

656 African Medicinal Plants,” PLoS One, vol. 8, no. 10, p. e78085, 2013.

657 [33] Y.-W. Chin, M. J. Balunas, H. B. Chai, and A. D. Kinghorn, “Drug discovery from natural

658 sources,” AAPS J., vol. 8, no. 2, pp. E239–E253, 2006.

35
659 [34] H. Kang et al., “HIM-herbal ingredients in-vivo metabolism database,” J. Cheminform., vol. 5, no.

660 5, 2013.

661 [35] H. Ye et al., “HIT: linking herbal active ingredients to targets.,” Nucleic Acids Res., vol. 39, no.

662 Database issue, pp. D1055-9, Jan. 2011.

663 [36] “InterBioScreen ltd.” [Online]. Available: https://www.ibscreen.com/natural-compounds.

664 [Accessed: 27-Feb-2020].

665 [37] “INDOFINE Chemical Company.” [Online]. Available: https://www.indofinechemical.com/.

666 [Accessed: 27-Feb-2020].

667 [38] M. Mangal, P. Sagar, H. Singh, G. P. S. Raghava, and S. M. Agarwal, “NPACT: Naturally

668 Occurring Plant-based Anti-cancer Compound-Activity-Target database.,” Nucleic Acids Res., vol.

669 41, no. Database issue, pp. D1124-9, Jan. 2013.

670 [39] A. C. Pilon et al., “NuBBEDB: an updated database to uncover chemical and biological

671 information from Brazilian biodiversity,” Sci. Rep., vol. 7, no. 1, p. 7215, Dec. 2017.

672 [40] “Princeton BioMolecular Research, Inc.” [Online]. Available: http://www.princetonbio.com/.

673 [Accessed: 27-Feb-2020].

674 [41] “Specs chemistry database.” [Online]. Available:

675 http://www.specs.net/snpage.php?snpageid=home. [Accessed: 27-Feb-2020].

676 [42] “UEFS.” [Online]. Available: http://www.uefs.br/. [Accessed: 27-Feb-2020].

677 [43] C. A. Lipinski, F. Lombardo, B. W. Dominy, and P. J. Feeney, “Experimental and computational

678 approaches to estimate solubility and permeability in drug discovery and development settings,”

679 Adv. Drug Deliv. Rev., vol. 46, no. 1–3, pp. 3–26, Mar. 2001.

680 [44] R. V. Honorato et al., “KVFinder : Steered identification of protein cavities as a PyMOL plugin

681 KVFinder : steered identification of protein cavities as a PyMOL plugin,” no. July, 2014.

36
682 [45] Z. Zhang, Y. Li, B. Lin, M. Schroeder, and B. Huang, “Identification of cavities on protein surface

683 using multiple computational approaches for drug binding site prediction.,” Bioinformatics, vol.

684 27, no. 15, pp. 2083–8, Aug. 2011.

685 [46] G. M. Morris et al., “AutoDock4 and AutoDockTools4: Automated docking with selective

686 receptor flexibility.,” J. Comput. Chem., vol. 30, no. 16, pp. 2785–91, Dec. 2009.

687 [47] O. Trott and A. J. Olson, “AutoDock Vina: Improving the speed and accuracy of docking with a

688 new scoring function, efficient optimization, and multithreading,” J. Comput. Chem., vol. 31, no.

689 2, p. NA-NA, Jan. 2010.

690 [48] PyMOL, “The PyMOL Molecular Graphics System, Version 2.0.6 Schrödinger, LLC.”

691 Schrodinger LLC, 2010.

692 [49] L. Limited, “Derek Nexus.” [Online]. Available: https://www.lhasalimited.org/products/derek-

693 nexus.htm. [Accessed: 27-Feb-2020].

694 [50] P. M. Sancheti and S. P. Pawar, “IN SILICO TOXICITY PREDICTION OF TROGLITAZONE,

695 ROSIGLITAZONE AND PIOGLITAZONE USING DEREK NEXUS,” vol. 3, no. 2, pp. 129–

696 134, 2016.

697 [51] L. Limited, “Sarah Nexus.” [Online]. Available: https://www.lhasalimited.org/products/sarah-

698 nexus.htm. [Accessed: 27-Feb-2020].

699 [52] J. Ghosh, M. S. Lawless, M. Waldman, V. Gombar, and R. Fraczkiewicz, “Modeling ADMET,” in

700 Methods in Molecular Biology, vol. 1425, 2016, pp. 63–83.

701 [53] A. Daina, O. Michielin, and V. Zoete, “SwissADME: a free web tool to evaluate

702 pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules,” Sci.

703 Rep., vol. 7, no. October 2016, p. 42717, 2017.

704 [54] R. A. Laskowski and M. B. Swindells, “LigPlot+: Multiple ligand-protein interaction diagrams for

705 drug discovery,” J. Chem. Inf. Model., vol. 51, no. 10, pp. 2778–2786, 2011.

37
706 [55] M. J. Abraham et al., “Gromacs: High performance molecular simulations through multi-level

707 parallelism from laptops to supercomputers,” SoftwareX, vol. 1–2, pp. 19–25, Sep. 2015.

708 [56] A. W. Schüttelkopf and D. M. F. van Aalten, “PRODRG: a tool for high-throughput

709 crystallography of protein–ligand complexes,” Acta Crystallogr. Sect. D Biol. Crystallogr., vol.

710 60, no. 8, pp. 1355–1363, Aug. 2004.

711 [57] “Grace (plotting tool).” [Online]. Available: http://plasma-gate.weizmann.ac.il/Grace/. [Accessed:

712 27-Feb-2020].

713 [58] N. A. Baker, D. Sept, S. Joseph, M. J. Holst, and J. A. McCammon, “Electrostatics of

714 nanosystems: application to microtubules and the ribosome.,” Proc. Natl. Acad. Sci. U. S. A., vol.

715 98, no. 18, pp. 10037–41, Aug. 2001.

716 [59] R. Kumari, R. Kumar, A. Lynn, and A. Lynn, “g_mmpbsa —A GROMACS Tool for High-

717 Throughput MM-PBSA Calculations,” J. Chem. Inf. Model., vol. 54, no. 7, pp. 1951–1962, Jul.

718 2014.

719 [60] R. C. Team, “R: A Language and Environment for Statistical Computing.” R Foundation for

720 Statistical Computing, Vienna, Austria, 2018.

721 [61] Schrödinger, “Schrödinger Suite 2018-2 Induced Fit Docking protocol; Glide, Schrödinger, LLC,

722 New York, NY, 2018; Prime, Schrödinger, LLC, New York, NY, 2018.” New York, NY, 2018.

723 [62] A. Heifets and R. H. Lilien, “LigAlign: Flexible ligand-based active site alignment and analysis,”

724 J. Mol. Graph. Model., vol. 29, no. 1, pp. 93–101, Aug. 2010.

725 [63] D. S. Wishart et al., “DrugBank 5.0: A major update to the DrugBank database for 2018,” Nucleic

726 Acids Res., vol. 46, no. D1, pp. D1074–D1082, 2018.

727 [64] D. A. Filimonov et al., “Prediction of the Biological Activity Spectra of Organic Compounds

728 Using the Pass Online Web Resource,” Chem. Heterocycl. Compd., vol. 50, no. 3, pp. 444–457,

729 Jun. 2014.

38
730 [65] A. Stank, D. B. Kokh, J. C. Fuller, and R. C. Wade, “Protein Binding Pocket Dynamics,” 2016.

731 [66] M. Cammisa, A. Correra, G. Andreotti, and M. V. Cubellis, “Identification and analysis of

732 conserved pockets on protein surfaces,” BMC Bioinformatics, vol. 14, no. SUPPL7, p. S9, Apr.

733 2013.

734 [67] H. Wan, “What ADME tests should be conducted for preclinical studies?,” Admet Dmpk, vol. 1,

735 no. 3, pp. 19–28, 2013.

736 [68] N. Greene, P. N. Judson, J. J. Langowski, and C. A. Marchant, “Knowledge-Based Expert Systems

737 for Toxicity and Metabolism Prediction: DEREK, StAR and METEOR,” SAR QSAR Environ.

738 Res., vol. 10, no. 2–3, pp. 299–314, Jul. 1999.

739 [69] J. E. Ridings et al., “Computer prediction of possible toxic action from chemical structure: an

740 update on the DEREK system.,” Toxicology, vol. 106, no. 1–3, pp. 267–79, Jan. 1996.

741 [70] C. Barber et al., “Evaluation of a statistics-based Ames mutagenicity QSAR model and

742 interpretation of the results obtained,” Regul. Toxicol. Pharmacol., vol. 76, pp. 7–20, Apr. 2016.

743 [71] R. J. Mural et al., “A Comparison of Whole-Genome Shotgun-Derived Mouse Chromosome 16

744 and the Human Genome,” Science (80-. )., vol. 296, no. 5573, pp. 1661–1671, May 2002.

745 [72] S. Batzoglou, L. Pachter, J. P. Mesirov, B. Berger, and E. S. Lander, “Human and mouse gene

746 structure: comparative analysis and application to exon prediction.,” Genome Res., vol. 10, no. 7,

747 pp. 950–8, Jul. 2000.

748 [73] C. Gunter and D. Ritu, “CHIMPANZEE,” Nature, vol. 437, no. 7055, p. 7055, 2005.

749 [74] K. Prüfer et al., “The bonobo genome compared with the chimpanzee and human genomes.,”

750 Nature, vol. 486, no. 7404, pp. 527–31, Jun. 2012.

751 [75] B. Burton, M. T. Zimmermann, R. L. Jernigan, and Y. Wang, “A Computational Investigation on

752 the Connection between Dynamics Properties of Ribosomal Proteins and Ribosome Assembly,”

39
753 PLoS Comput. Biol., vol. 8, no. 5, p. e1002530, May 2012.

754 [76] H. Shukla, R. Shukla, A. Sonkar, T. Pandey, and T. Tripathi, “Distant Phe345 mutation

755 compromises the stability and activity of Mycobacterium tuberculosis isocitrate lyase by

756 modulating its structural flexibility /631/45/56 /631/92/606 /631/114/2411 /9 /82/83 /82 /82/16

757 /101 article,” Sci. Rep., vol. 7, no. 1, pp. 1–11, Dec. 2017.

758 [77] S. Shukla, K. Bafna, D. Sundar, and S. S. Thorat, “The bitter barricading of prostaglandin

759 biosynthesis pathway: Understanding the molecular mechanism of selective cyclooxygenase-2

760 inhibition by amarogentin, a secoiridoid glycoside from Swertia chirayita,” PLoS One, vol. 9, no.

761 3, p. e90637, Mar. 2014.

762 [78] C. V. Kumar, R. G. Swetha, A. Anbarasu, and S. Ramaiah, “Computational Analysis Reveals the

763 Association of Threonine 118 Methionine Mutation in PMP22 Resulting in CMT-1A,” Adv.

764 Bioinformatics, vol. 2014, 2014.

765 [79] S. Genheden and U. Ryde, “The MM/PBSA and MM/GBSA methods to estimate ligand-binding

766 affinities,” Expert Opin. Drug Discov., vol. 10, no. 5, p. 449, 2015.

767 [80] S. Borkotoky, C. K. Meena, and A. Murali, “Interaction Analysis of T7 RNA Polymerase with

768 Heparin and Its Low Molecular Weight Derivatives - An In Silico Approach.,” Bioinform. Biol.

769 Insights, vol. 10, pp. 155–66, 2016.

770 [81] B. Kuhn, P. Gerber, T. Schulz-Gasch, and M. Stahl, “Validation and use of the MM-PBSA

771 approach for drug discovery,” J. Med. Chem., vol. 48, no. 12, pp. 4040–4048, Jun. 2005.

772 [82] M. Xu and M. A. Lill, “Induced fit docking, and the use of QM/MM methods in docking,” 2013.

773 [83] W. Sherman, T. Day, M. P. Jacobson, R. A. Friesner, and R. Farid, “Novel Procedure for

774 Moldeing Ligand/Receptor Induced Fit Effects,” J Med Chem, vol. 49, no. 2, pp. 534–553, 2006.

775 [84] A. J. Clark et al., “Prediction of Protein-Ligand Binding Poses via a Combination of Induced Fit

776 Docking and Metadynamics Simulations,” J. Chem. Theory Comput., vol. 12, no. 6, pp. 2990–

40
777 2998, 2016.

778 [85] K. Liu and H. Kokubo, “Exploring the Stability of Ligand Binding Modes to Proteins by

779 Molecular Dynamics Simulations: A Cross-docking Study,” J. Chem. Inf. Model., vol. 57, no. 10,

780 pp. 2514–2522, Oct. 2017.

781 [86] S. Singh, G. Kaur, V. Mangla, and M. K. Gupta, “Quinoline and quinolones: Promising scaffolds

782 for future antimycobacterial agents,” J. Enzyme Inhib. Med. Chem., vol. 30, no. 3, pp. 492–504,

783 2015.

784 [87] R. S. Keri and S. A. Patil, “Quinoline: A promising antitubercular target,” Biomed.

785 Pharmacother., vol. 68, no. 8, pp. 1161–1175, 2014.

786 [88] DrugBank, “bis(4-hydroxyphenyl)methanone.” [Online]. Available:

787 https://www.drugbank.ca/drugs/DB07635. [Accessed: 27-Feb-2020].

788 [89] E. A. Abourashed, A. M. Galal, A. M. Shibl, E. A. Abourashed, A. M. Galal, and A. M. Shibl,

789 “Antimycobacterial activity of ferutinin alone and in combination with antitubercular drugs

790 against a rapidly growing surrogate of Mycobacterium tuberculosis,” vol. 6419, 2011.

791 [90] E. Tatar et al., “Design, Synthesis, and Molecular Docking Studies of a Conjugated Thiadiazole–

792 Thiourea Scaffold as Antituberculosis Agents,” Biol. Pharm. Bull., vol. 39, no. 4, pp. 502–515,

793 2016.

794 [91] J. Jin et al., “Farnesol, a Potential Efflux Pump Inhibitor in Mycobacterium smegmatis,”

795 Molecules, vol. 15, no. 11, pp. 7750–7762, Oct. 2010.

796 [92] A. Jiménez, M. Meckes, V. Alvarez, J. Torres, and R. Parra, “Secondary metabolites from

797 Chamaedora tepejilote (Palmae) are active against Mycobacterium tuberculosis,” Phyther. Res.,

798 vol. 19, no. 4, pp. 320–322, 2005.

799 [93] M. A. Tan, H. Takayama, N. Aimi, M. Kitajima, S. G. Franzblau, and M. G. Nonato,

800 “Antitubercular triterpenes and phytosterols from Pandanus tectorius Soland. var. laevis,” J. Nat.

41
801 Med., vol. 62, no. 2, pp. 232–235, Apr. 2008.

802 [94] D. B. dos Reis et al., “Synthesis and biological evaluation against Mycobacterium tuberculosis and

803 Leishmania amazonensis of a series of diaminated terpenoids,” Biomed. Pharmacother., vol. 84,

804 pp. 1739–1747, Dec. 2016.

805 [95] N. Roy and R. Kadam, “Recent trends in drug-likeness prediction: A comprehensive review of In

806 silico methods,” Indian J. Pharm. Sci., vol. 69, no. 5, p. 609, 2007.

807 [96] “Prediction of Drug-Like Properties - Madame Curie Bioscience Database - NCBI Bookshelf.”

808 [Online]. Available: https://www.ncbi.nlm.nih.gov/books/NBK6404/. [Accessed: 19-Apr-2020].

809 [97] P. Wangchuk, “Therapeutic Applications of Natural Products in Herbal Medicines, Biodiscovery

810 Programs, and Biomedicine,” Journal of Biologically Active Products from Nature, vol. 8, no. 1.

811 Taylor and Francis Ltd., pp. 1–20, 02-Jan-2018.

812 [98] D. A. Dias, S. Urban, and U. Roessner, “A Historical overview of natural products in drug

813 discovery,” Metabolites, vol. 2, no. 2. MDPI AG, pp. 303–336, 16-Apr-2012.

814 [99] G. Khare, P. Kumar, and A. K. Tyagi, “Whole-cell screening-based identification of inhibitors

815 against the intraphagosomal survival of Mycobacterium tuberculosis,” Antimicrob. Agents

816 Chemother., vol. 57, no. 12, pp. 6372–6377, Dec. 2013.

817 [100] J. P. Renaud, T. Neumann, and L. Van Hijfte, “Fragment-Based Drug Discovery,” in Small

818 Molecule Medicinal Chemistry: Strategies and Technologies, Hoboken, NJ: Wiley Blackwell,

819 2015, pp. 221–249.

820 [101] D. Joseph-McCarthy, A. J. Campbell, G. Kern, and D. Moustakas, “Fragment-based lead

821 discovery and design,” Journal of Chemical Information and Modeling, vol. 54, no. 3. American

822 Chemical Society, pp. 693–704, 24-Mar-2014.

823 [102] B. J. Davis and S. D. Roughley, “Fragment-Based Lead Discovery,” in Annual Reports in

824 Medicinal Chemistry, vol. 50, Academic Press Inc., 2017, pp. 371–439.

42
825 [103] Y.-C. Lo and J. Z. Torres, “Chemical Similarity Networks for Drug Discovery,” in Special Topics

826 in Drug Discovery, InTech, 2016.

827

43
Highlights
• New insights into the binding mechanisms of OmpATb
• Four scaffolds for antitubercular drug design
• A scaffold for designing efflux pump inhibitors against OmpATb
• Four potential novel anti-tubercular molecules for in vitro activity testing
Authors declare no conflict of interest