Computational Biology and Chemistry 77 (2018) 64–71

Contents lists available at ScienceDirect

Computational Biology and Chemistry

journal homepage: www.elsevier.com/locate/cbac

Research Article

Pharmacophore-based virtual screening for identifying β5 subunit inhibitor T

of 20S proteasome
⁎
Muhammad Arbaa, , Andry Nur-Hidayata, Slamet Ibrahim Surantaadmajab, Daryono H. Tjahjonob
a
Faculty of Pharmacy, Halu Oleo University, Kendari, 93231, Indonesia
b
School of Pharmacy, Bandung Institute of Technology, Bandung, 40312, Indonesia

A R T I C LE I N FO A B S T R A C T

Keywords: Proteasomal system plays an important role in maintaining cell homeostatis. Overexpression of proteasomes
MM-PBSA leads to several major diseases, such as cancer and autoimmune disorder. The β5 subunit of proteasome is a
Molecular docking crucial active site in proteolysis, and targeting proteasome β5 subunit is essential for proteasome inhibition. In
Molecular dynamics simulation the present study, a pharmacophore-based virtual screening and molecular docking were employed to identify
Pharmacophore model
ligands as inhibitors of β5 subunit of proteasome. The pharmacophore features were built with one hydrogen
Proteasome
bond donor, two hydrogen bond acceptors, and one hydrophobic feature using native ligand of proteasome
Virtual screening
(HU10), which was then used to screen ZINC database using ZINCPharmer. The retrieved virtual hits were
subjected to molecular docking analysis using iDock. The best six hits were subjected to molecular dynamics
(MD) simulation and each complex was stable during 40 ns MD simulation as indicated by root-mean-square-
deviation (RMSD) and root-mean-square-fluctuation (RMSF) values. The current study identifies 5 best hits
having better binding potentials than HU10 as predicted by molecular mechanics Poisson-Boltzmann Surface
Area (MM-PBSA) method, i.e. Lig1540/ZINC33356240, Lig1546/ZINC33356235, Lig1522/ZINC20854878,
Lig980/ZINC12391945, and Lig1119/ZINC19865241, which can be used in the development of new proteasome
inhibitors.

1. Introduction 2006; Obaidat et al., 2011).

The main structure of ubiquitin proteasome pathway is proteasome,
The ubiquitin proteasome pathway is crucial in maintaining several which is composed of 20S core particle and two 19S regulatory parti-
cellular processes such as proteolytic degradation, cell cycle progres- cles. The core particle consists of four stacked rings: two outer rings and
sion, DNA repair, and apoptosis. It is upregulated in several major two inner rings. The outer rings contain seven different α subunits,
diseases such as cancer, neurodegenerative disorder, and infectious while the inner rings contain seven different β-subunits; all constitute
diseases, which may be due to the increased metabolism rate in ma- subunit stoichiometry α1-7β1-7β1-7α1-7. The proteasome proteolytic ac-
lignant cells (Edelmann et al., 2011; Zheng et al., 2016). Inhibition of tivity is located on β1 which possesses a caspase-like (C-L) activity, β2
the pathway induces apoptosis activation as a result of misfolded pro- which shows a trypsin-like (T-L) activity, and β5 subunits which possess
tein accumulation (Augello et al., 2018). Thus, targeting proteasome a chymotrypsin-like (ChT-L) activity (Wehmer and Sakata, 2016). More
has been a promising approach in the chemotherapeutic drug dis- importantly, the inhibition of chymotrypsin-like subunit alone was in-
covery. The first-in-class proteasome inhibitor, Bortezomib, was ap- dicated to a have substantial implication for cancer therapy (Britton
proved by Food and Drug Administration (FDA) in 2003, and widely et al., 2009; McBride and Ryan, 2013; Yang et al., 2016; Miller et al.,
applied in the treatment of multiple myeloma. The great success of 2015; Han et al., 2017; Qin et al., 2017; Teicher and Tomaszewski,
bortezomib stimulated the development of the second-generation pro- 2015).
teasome inhibitor, Carfilzomib, which was also approved in 2012 for Proteasome inhibitors such as Bortezomib and Carfilzomib work by
the treatment of multiplemyeloma. Other proteasome inhibitors are still covalent binding and contain highly reactive and unstable chemical
in clinical trials such as marizomib, oprozomib, and delanzomib groups. They are hypothesized to be the main source of toxicity on
(Teicher and Tomaszewski, 2015; Buac et al., 2013; Liu et al., 2015; treated patients. Several side effects were recorded in the use of cova-
Moreau, 2014; Allegra et al., 2014; Moreau et al., 2012; Chauhan et al., lent inhibitors including peripheral neuropathy, thrombocytopenia, and

⁎
Corresponding author at: Faculty of Pharmacy, Halu Oleo Universitya, Kendari, Southeast Sulawesi, 93231, Indonesia.
E-mail address: muh.arba@uho.ac.id (M. Arba).

https://doi.org/10.1016/j.compbiolchem.2018.08.009
Received 11 April 2018; Received in revised form 16 August 2018; Accepted 26 August 2018
Available online 28 August 2018
1476-9271/ © 2018 Elsevier Ltd. All rights reserved.
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

gastrointestinal disorders, which may be due to their less specific The grid box was set to the center of HU10 binding site with a size of
properties to the targeted cell (Kazi et al., 2014; Singh et al., 2011; 22.5 × 22.5 × 22.5 Å. Analysis and visualization of docking results
Orlowski et al., 2002). Therefore, emerging efforts are focusing on the were performed by using Discovery Studio Visualizer. All docked li-
development of non-covalent proteasome inhibitor to take advantage of gands were then ranked based on their binding energies and the top six
their rapid binding and dissociation kinetics, although, it not clear yet ligands having lowest binding energies and best conformations were
whether or not they suffer from the same limitation as covalent in- subjected for further molecular dynamics simulation.
hibitors (Kazi et al., 2014). Among others, ritonavir, aminobenzylstatin Molecular dynamics simulations on six top-scored ligand-protea-
and peptide derivatives have been non-covalent proteasome inhibitors some complexes were carried out using the GPU version of the PMEMD
under intensive investigation (Genin et al., 2010; Beck et al., 2012; engine provided with the Amber 16 package (Case et al., 2016;
Blackburn et al., 2010; Zhang et al., 2016). The current study aims to Salomon-Ferrer et al., 2013). The ff14SB (Maier et al., 2015) and GAFF
identify potential non-covalent β5 subunit of proteasome inhibitors by (Wang et al., 2004) force fields as well as AM1-BCC (Jakalian et al.,
using virtual screening method. 2002) were used to parameterize the complex. The truncated octahe-
Virtual screening methods play a major role in the drug discovery dron TIP3P water model with the minimum distance of 10 Å around the
processes for their time and cost efficiencies. Virtual screening is de- system was used and each system was neutralized by the addition of
scribed as the use of in silico techniques to search for small molecules in Na + counterions using Leap module of Amber 16.
the large compound databases to identify structures which are poten- Minimization and equilibration were carried out using Sander
tially bound to a therapeutic target (Kumalo and Soliman, 2015; Jain, module of Amber 16. The first minimization was performed by 500
2004). Virtual screening is grouped into structural-based approach and steps of steepest descents and 5500 steps of conjugate gradients with
ligand-based screening (Singh and Jana, 2017; Sangeetha et al., 2017). protein were restrained (k = 500 kcal mol−1 Å−2). The same steps
While the ligand-based screening including pharmacophore modelling were applied to the second and third minimizations with the backbone
is designed to compare structural similarity of known and unknown atoms of protein restrained (k = 500 kcal mol−1 Å−2) and without
compounds with active known ligand as a query input (Agrawal et al., restraint, respectively.
2013; Dror et al., 2009), the structure-based virtual screening such as System heating was performed gradually in NVT ensemble from 0 to
molecular docking aims to predict the preferred binding mode and af- 100, 100 to 200, and 200 to 310 K, respectively, for each 50 ps with a
finity of the compounds in a drug target. In the present study, both time step of 0.0005 ps and backbone atom restraints for protein (k
pharmacophore-based virtual screening and molecular docking study = 5 kcal mol−1 Å−2). Then, the system was equilibrated at 310 K
were performed. However, since molecular docking technique did not during two 100 ps subsequent equilibration steps with restraints (k = 5
take into account the full flexibility of protein, molecular dynamics and k = 3 kcal mol−1 Å−2), respectively, which was followed by 100 ps
simulation was also performed to confirm the predicted docked struc- NPT ensemble equilibration without restraint. Furthermore, 40 ns MD
ture while accounting for the effect of receptor flexibility and solvent production was performed for each system in NPT ensemble by using
environment (Arba et al., 2017a, b). The combination of those techni- pmemd.cuda module in Amber 16. The temperature regulation was
ques is important in hit identification of inhibitors of β5 subunit of attained using Langevin thermostat with a collision rate of 1.0 ps−1.
proteasome. The SHAKE algorithm (Ryckaert et al., 1977) was retained to all
bonds involving hydrogen atoms with 2 fs integration time step. The
2. Computational methods particle-mesh Ewald method was used to treat long-range electrostatics
interactions (Darden et al., 1993) with a non-bonding cutoff distance of
2.1. Pharmacophore modelling and database screening 9.0 Å, applying cubic periodic boundary conditions. The coordinate
files were saved every 1 ps and analyses were performed with the
The pharmacophore model was obtained with the ZINCPharmer CPPTRAJ module (Roe and Cheatham, 2013) in Amber16, while vi-
web server (http://zincpharmer.csb.pitt.edu/) (Koes and Camacho, sualization of the results was carried out using the Visual Molecular
2012), using crystallographic structure of proteasome (PDB ID: 3SHJ) Dynamics software (Humphrey et al., 1996).
containing HU10 inhibitor. ZINCPharmer performed searching of hits
compounds of ZINC database (Irwin et al., 2012) using a pharmaco- 2.3. Binding free energy calculations
phore model of HU10. A pharmacophore is a spatial arrangement of the
essential features of ligand necessary for optimal interaction with the The binding free energy was calculated using the Molecular
target protein (Dror et al., 2009). The pharmacophore hypotheses of the Mechanics-Poisson Boltzmann solvent accessible surface area (MM-
present work were chosen for having one hydrogen bond donor, two PBSA) method (Kollman et al., 2000; Arba and Tjahjono, 2015) on 200
hydrogen bond acceptors, and one hydrophobic feature. In that scheme, snapshots taken from 20 to 40 ns simulation trajectories with the py-
not all pharmacophore features were selected in order to retrieve more thon modules supplied with the Amber 16 (Miller et al., 2012). Ac-
diverse small molecules of ZINC database. Each hydrogen bond donor/ cording to MM-PBSA method, the binding free energy of complex
acceptor feature has a radius of 0.5 Å, while the hydrophobic feature (ΔGbind) is the difference between the free energies of the complex
has radius of 1 Å. The chosen pharmacophore features were then sub- (ΔGcomplex) and the unbound receptor (ΔGrec) and the free ligand
mitted to the ZINCPharmer and the screening result of small molecules (ΔGlig), which is theoretically described as follows:
was then saved in SDF file format for further analysis. ΔG bind = Gcomplex −Grec−Gligand (1)

2.2. Molecular docking and molecular dynamics simulations ΔG bind = ΔEMM + ΔGsol−TΔS (2)

All compounds resulted from the virtual screening of ZINCPharmer ΔEMM = ΔEbond + ΔEangle + ΔEtorsion + ΔE vdw + ΔEEEL (3)
were submitted for molecular docking on β5/β6 subunit of the pro-
ΔGsol = ΔGPB + ΔGSA (4)
teasome using iDock software (Li et al., 2012) to predict their affinities
and binding modes in the active site of the proteasome. The structures The terms of ΔEMM, ΔGsol, and TΔS denote the gas phase binding en-
of β5 and β6 subunits of the proteasome were taken from the Protein ergy, solvation free energy, and the conformational entropy change
Data Bank with PDB ID 3SHJ (Gallastegui et al., 2012). Native ligand upon ligand binding at temperature T, respectively. The ΔEMM consists
(HU10) was also redocked to validate the docking protocol by mea- of bonded energy terms (bond energy, angle energy, and torsion en-
suring the root mean-square deviation (RMSD) between docked and X- ergy) and non-bonded terms (van der Waals energy and electrostatic
ray conformations. energy). The ΔGsol consists of the polar contribution to solvation free

65
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Fig. 1. 3D pharmacophore features generated on ZINCPharmer. The hydro-

phobic, hydrogen bond donor, and hydrogen bond acceptors were shown in
green, white and gold spheres, respectively. The arrows show the constraint
direction. (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article). Fig. 2. Receiver operating characteristic (ROC) curve.

energy (ΔGPB) calculated by solving the Poisson-Boltzmann (PB) residues. The sulfonamide group interacted through hydrogen bond (H-
equation using a grid size of 0.5 Å, and the non-polar contribution bond) with Thr21, while oxygen atom with hydroxyl group of Ser118.
(ΔGSA) which was calculated using the solvent accessible surface area Electrostatic interaction (pi-cationic) was observed between phenyl
(SASA) with solvent-probe radius set to 1.4 Å.
group of the ligand with guanidino group of Arg125, while pi-pi in-
teraction was observed between ligand phenyl ring with imidazole
3. Results and discussion group of His98.
The replacement of chloride atom in Lig1546 with methyl group
The virtual screening of ZINCPharmer using the chosen pharmaco- produced the second best hit, Lig1522, which had a docking score of
phore features yielded 1565 compounds out of 22,723,923 small mo- −11.52 kcal/mol and established interactions in the similar way to
lecules of ZINC database. Fig. 1 shows the pharmacophore features Lig1546. The H-bonds were established between sulfonamide group
selected for virtual screening using ZINCPharmer. with Thr1 and Thr21, and the ligand carbonyl group with hydroxyl
Furthermore, molecular docking on all 1565 compounds on β5/β6 group of Ser118. The phenyl and guanidino groups of ligand and
subunit of the proteasome using iDock resulted in conformations and Arg125 residue, respectively, were detected to established pi-cationic
binding energies ranging from −5.46 to −11.68 kcal/mol. Redocking interaction, while pi-pi interaction was observed between ligand phenyl
of native ligand (HU10) resulted in a pose with a binding energy of ring and imidazole group of His98.
−8.85 kcal/mol. A total of 868 compounds exhibited better affinities In the Lig980 pose, oxygen atom of ligand sulfonamide group in-
compared to that of HU10 and top 6 molecules were selected based on teracted through two H-bonds with side chains of Ser112 and Arg125.
their binding interactions and energies: Lig1546/ZINC33356235 The ligand naphthalene group interacted through pi-pi and pi-cationic
(E = −11.53 kcal/mol), Lig1522/ZINC20854878 (E = −11.52 kcal/ interactions with imidazole group of His98 and guanidine group of
mol), Lig980/ZINC12391945 (E = −11.48 kcal/mol), Lig1540/ Arg125, respectively.
ZINC33356240 (E = −11.4 kcal/mol), Lig1180/ZINC13039962 In the Lig1540 docked pose, ligand benzodioxole group established
(E = −11.35 kcal/mol), Lig1119/ZINC19865241 (E = −11.31 kcal/ two H-bonds with the backbones of Ala49 and Ala50. In addition, one
mol). Fig. 2 displays chemical structures of best 6 hits. H-bond between side chain hydroxyl group of Ser118 and ligand car-
Validation of docking protocol was achieved by redocking of HU10 bonyl group was observed. Electrostatic interaction (pi-cationic) was
to assess the credibility of docking protocol used. The structure of both also observed between phenyl group of the ligand with guanidino group
conformations was essentially identical with RMSD 1.103 Å, indicating of Arg125. The phenyl group was also established pi-stacking interac-
the reliability of docking protocol (Morris et al. 1998). In the docked tion with imidazole group of His98. Fig. 4 shows the interactions of
conformation, several important interactions were noted: the hydroxyl Lig1546, Lig1522, Lig980, and Lig1540 in the β5/β6 subunits of the
group of HU10 interacted through hydrogen bond with Gly47, oxygen proteasome. Lig1180 formed H-bond through ligand sulfonyl group
atom of carbonyl group interacted through two hydrogen bonds with with side chain hydroxyl group of Ser118. In addition, ligand quina-
Thr1 and Thr21, and amino group made interaction with Thr21 residue. zolinone group interacted through two H-bonds with guanidino group
It is noted that hydrogen bond with Thr21 was also observed in the of Arg125 and indole group of Trp25. Lig1119 established two H-bonds
crystallographic structure of HU10. In addition, van der Waals inter- through ligand morpholine group with side chain hydroxyl group of
action was observed between HU10 with Ala20, Al27, Val31, and Thr1, and between ligand imidazolidinone group with guanidino group
Ala49. Fig. 3 depicts an overlay of docked and crystallographic con- of Arg125. Fig. 5 displays the interaction of Lig1180 and Lig1119 in the
formations of the native ligand (HU10). β5/β6 subunits of the proteasome.
Furthermore, it is observed that the docked poses of six best hits Overall, the virtual screening using iDock showed that the interac-
established several important interactions with proteasome. The tions of six best hits to proteasome were dominated by the H-bond and
Lig1546/ZINC33356235 which exhibited the highest docking score of van der Waals interactions.
−11.53 kcal/mol displayed interactions with important amino acid

66
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Fig. 3. The chemical structures of best 6 hits obtained from virtual screening performed by iDock software.

3.1. Molecular dynamics simulations all six complexes have a similar pattern in all regions. The carbonyl end
of β5 subunit (Gly212) and loop region of β6 subunit (Thr14I) showed
Each proteasome complexed with top six hits and HU10 was sub- high fluctuation, while residues of loop regions including those around
jected to MD simulations to confirm the stability of each pose and in- β6 Asp69 residues slightly fluctuated. Interestingly, residues of β5
teraction. Verification of system stability throughout 40-ns simulation Thr21 and β6 Ser118, which were observed to have interactions with
period was analysed by measuring the root mean square deviation ligand, showed high rigidity, indicating that the ligand binding yielded
(RMSD) and the root mean square fluctuation (RMSF). Fig. 6 displays minor fluctuation on the protein.
RMSD value of each ligand-proteasome complex during 40 ns dynamics
runs, while Fig. 7 depicts RMSF plot of each amino acid residue during
3.2. Ligand interactions
40 ns dynamics simulation. Fig. 6 clearly shows that Lig980, Lig1119,
and Lig1546, each complexed with proteasome, are more stable than
The stability of the ligand binding pose was checked by analysing
the HU10-proteasome complex. In addition, the complexes of Lig1522
the dynamics of number of the non-bond interactions between ligand
and Lig1180 have comparable stability as HU10, while Lig1540 slightly
and proteasome. Fig. 8 depicts the dynamics of number of non-bond
fluctuated compared to HU10. An inspection on RMSF plot showed that
interaction during 40 ns, which showed that most of ligand had more

Fig. 4. An overlay of docked (blue) and crystallographic (green) conformations of the native ligand (HU10). The hydrogen bond of docked conformation was shown
in green-colored dashed line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

67
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Fig. 5. The interaction of Lig1546, Lig1522, Lig980 and Lig1540 in the proteasome β5/β6 subunits. The hydrogen bond, pi-cationic and pi-stacking interactions are
represented in green, orange and purple colored dashed lines, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article).

number of non-bond interactions compared to HU10. It was consistent dynamics of H-bond interaction throughout simulation time was also
with the predicted binding free energy by MM-PBSA method (discussed monitored. Table S1 shows the occupancy of H-bond of each ligand
later), in which binding affinity of HU10 was lower than those of most during 40-ns MD simulation time. For HU10, it was observed that H-
ligands. Additionally, it reveals that the number of non-bond interac- bonds with Gly47, Thr1, and Thr21 which were observed in docked
tions in Lig1540 seems to be superior than that of other ligands, which pose showed occupancies of 33.79%, 19.61%, and 8.3%, respectively.
aligns with its lowest binding free energy (Table 1). The new H-bonds were formed including those with β6 Asp114 and β6
Further, as an important part of non-bond interactions, the Ser118 with the occupancies of 1.17% and 0.19%, respectively, which

Fig. 6. The interaction of Lig1180 and Lig1119 in the proteasome β5/β6 subunits. The hydrogen bonds are represented in green colored dashed lines. (For inter-
pretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

68
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Fig. 7. RMSD value of protein heavy atom of each ligand-proteasome complex

during 40 ns MD calculated for (a) HU10 (red), Lig980 (green) and Lig1119 Fig. 9. Number of non-bond interaction versus time of each ligand-proteasome
(blue); (b) HU10 (red), Lig1522 (green) and Lig1180 (blue); (c) HU10 (red), complex during 40 ns MD simulation for (a) HU10 (red), Lig980 (green) and
Lig1546 (green) and Lig1540 (blue). (For interpretation of the references to Lig1119 (blue); (b) HU10 (red), Lig1522 (green) and Lig1180 (blue); (c) HU10
colour in this figure legend, the reader is referred to the web version of this (red), Lig1546 (green) and Lig1540 (blue). (For interpretation of the references
article). to colour in this figure legend, the reader is referred to the web version of this
article).

2.07%, respectively, while those between Lig1180 with Trp25 and

Arg125 are 5.7% and 1.18%, respectively, which were significantly
unstable.
In Lig1546 binding dynamics, the H-bonds originated from docked
pose displayed low occupancies, including those with Thr21 and Ser118
(each with 3.68% and 4.19%, respectively). Meanwhile, new H-bonds
with β6 Phe113 and β6 His98 showed moderate stabilities with the
occupancies of 43.66% and 13.66%, respectively. In the Lig980
binding, the H-bonds with Ser112 and Arg125 showed occupancies of
22.51% and 0.15%, respectively, while new H-bonds were formed with
Ser28, Ala27, and Thr21 which displayed the occupancies of 21.19%,
14.42%, and 6.38%, respectively. The low occupancies of H-bonds were
noted between Lig1119 and β6 Arg125 (2.85%) and Thr1 (0.01%),
respectively. The new H-bonds are formed between Lig1119 and β6
Ser112 and β6 His98 with the occupancies of 45.56% and 22.98%,
respectively.
Overall, Lig1540, Lig1522, and Lig1180 formed H-bonds with
Fig. 8. RMSF change during 40 ns MD simulation: (a) HU10 (red), Lig980
(green) and Lig1119 (blue); (b) HU10 (red), Lig1522 (green) and Lig1180
Ser118 with high occupancies, which indicate their persistence during
(blue); (c) HU10 (red), Lig1546 (green) and Lig1540 (blue). (For interpretation 40-ns simulation time, while those with other residues displayed minor
of the references to colour in this figure legend, the reader is referred to the web stabilities.
version of this article).
3.3. MM-PBSA binding free energy
were essentially unstable (Fig. 9).
In Lig1540 binding dynamics, the occupancy of H-bond with β6 Table 1 shows the binding free energies and separate energy terms
Ser118 reached 94.38%, while those with Ala49 and Ala50 showed (kcal/mol) calculated using the MM-PBSA method. The predicted
significantly lower occupancies (0.07% and 1.79%, respectively). New binding free energy (ΔEPBTOT) of HU10 was −19.91 kcal/mol, which
H-bonds were formed with β6 Ser112 and β6 Asp114 with the occu- was about two times lower than that of the experimental data
pancies of 14.14% and 3.43%, respectively. The high stabilities of H- (ΔGexp = −10.25 kcal/mol, ΔG = RT ln Ki, where Ki = 34 nM)
bonds with β6 Ser118 were also noted in Lig1522 and Lig1180, with the (Gallastegui et al., 2012). The excluded entropy term in the calculation
occupancies of 93.88% and 92.64%, respectively. In the meantime, H- could be the reason for the discrepancy between the experimental and
bonds occupancies of Lig1522 and Thr1 and Thr21 are 0.58% and predicted binding free energy (Homeyer and Gohlke, 2012; Arba et al.,
2017b; Kollman et al., 2000; Foloppe and Hubbard, 2006).

Table 1
The binding free energies and separate energy terms (kcal/mol) calculated using the MM-PBSA method.
Comp ΔEELE ΔEVDW ΔEPBCAL ΔEPBSUR ΔEPBTOT

HU10 −12.48 ± 7.27 −37.25 ± 4.24 33.98 ± 8.91 −4.16 ± 0.29 −19.91 ± 4.99
Lig1119 −15.11 ± 6.44 −43.13 ± 6.83 41.32 ± 11.40 −4.73 ± 0.38 −21.65 ± 3.91
Lig980 −21.14 ± 6.02 −48.96 ± 3.77 51.80 ± 6.43 −4.73 ± 0.31 −23.04 ± 4.31
Lig1180 −19.62 ± 5.28 −48.86 ± 3.39 56.25 ± 5.63 −5.08 ± 0.21 −17.32 ± 3.97
Lig1522 −19.54 ± 6.51 −46.47 ± 3.94 44.75 ± 7.28 −4.67 ± 0.23 −25.93 ± 4.01
Lig1546 −12.76 ± 4.45 −39.42 ± 3.77 28.85 ± 5.49 −4.11 ± 0.27 −27.44 ± 3.07
Lig1540 −27.15 ± 5.49 −50.59 ± 4.10 54.70 ± 7.11 −5.02 ± 0.28 −28.06 ± 4.47

69
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Furthermore, the binding affinity of Lig1540 was strongest than T.J., Gohlke, H., Goetz, A.W., Homeyer, N., Izadi, S., Janowski, P., Kaus, J.,
those of other ligands as indicated by the total binding free energy Kovalenko, A., Lee, T.S., LeGrand, S., Li, P., Lin, C., Luchko, T., Luo, R., Madej, B.,
Mermelstein, D., Merz, K.M., Monard, G., Nguyen, H., Nguyen, H.T., Omelyan, I.,
(ΔElig1540 = −28.06 kcal/mol). While the electrostatic energy (ΔEELE) Onufriev, A., Roe, D.R., Roitberg, A., Sagui, C., Simmerling, C.L., Botello-Smith,
terms were favourable, the polar energies of desolvation (ΔEPBCAL) were W.M., Swails, J., Walker, R.C., Wang, J., Wolf, R.M., Wu, X., Xiao, L., Kollman, P.A.,
disadvantageous, which yielded unfavourable net electrostatic con- 2016. Amber 2016. University of California, San Francisco.
Chauhan, D., Hideshima, T., Anderson, K.C., 2006. A novel proteasome inhibitor NPI-
tributions (ΔEELE + ΔEPBCAL). It is noted that Lig1180 has highest net 0052 as an anticancer therapy. Br. J. Cancer 95, 961–965.
electrostatic energy (ΔEPBCAL = 56.25 kcal/mol) which causes the Darden, T., York, D., Pedersen, L., 1993. Particle mesh Ewald: an N∙log(N) method for
lowest binding affinity (ΔEPBTOT = −17.32 kcal/mol). On the other Ewald sums in large systems. J. Chem. Phys. 98, 10089–10092.
Dror, O., Schneidman-Duhovny, D., Inbar, Y., Nussinov, R., Wolfson, H.J., 2009. Novel
hand, van der Waals (ΔEvdw) and the nonpolar solvation energies approach for efficient pharmacophore-based virtual screening: method and applica-
(ΔEPBSUR) were both advantageous for ligand binding, indicating their tions. J. Chem. Inf. Model. 49, 2333–2343.
main contribution to negative binding free energy. Edelmann, M.J., Nicholson, B., Kessler, B.M., 2011. Pharmacological targets in the ubi-
quitin system offer new ways of treating cancer, neurodegenerative disorders and
infectious diseases. Expert Rev. Mol. Med. 13, 1–17.
4. Conclusion Foloppe, N., Hubbard, R., 2006. Towards predictive ligand design with free-energy based
computational methods? Curr. Med. Chem. 13 (29), 3583–3608.
This study identifies five small molecules as potential inhibitors of Gallastegui, N., Beck, P., Arciniega, M., Huber, R., Hillebrand, S., Groll, M., 2012.
Hydroxyureas as noncovalent proteasome inhibitors. Angew. Chem. Int. Ed. 51,
β5 subunit of proteasome, i.e. Lig1540/ZINC33356240, Lig1546/ 247–249.
ZINC33356235, Lig1522/ZINC20854878, Lig980/ZINC12391945, and Genin, E., Reboud-Ravaux, M., Vidal, J., 2010. Proteasome inhibitors: recent advances
Lig1119/ZINC19865241, by using pharmacophore based virtual and new perspectives in medicinal chemistry. Curr. Top. Med. Chem. 10 (3),
232–256.
screening, molecular docking, and molecular dynamics simulation. All Han, L.-Q., Yuan, X., Wu, X.-Y., Li, R.-D., Xu, B., Cheng, Q., Liu, Z.-M., Zhou, T.-Y., An, H.-
six compounds bound to the β5 subunit of the proteasome were stable Y., Wang, X., Cheng, T.-M., Ge, Z.-M., Cui, J.-R., Li, R.-T., 2017. Urea-containing
for 40 ns as indicated by RMSD and RMSF values. The binding of ligand peptide boronic acids as potent proteasome inhibitors. Eur. J. Med. Chem. 125,
925–939.
was mainly supported by van der Waals and electrostatic interactions as Homeyer, N., Gohlke, H., 2012. Free energy calculations by the molecular mechanics
predicted by MM-PBSA method. The results of this study provide five poisson Boltzmann surface area method. Infect. Genet. Evol. 31, 114–122.
potential ligands to be considered in further discovery and development Humphrey, W., Dalke, A., Schulten, K., 1996. VMD-visual molecular dynamics. J. Mol.
Graph. 14, 33–38.
research on β5 subunit inhibitor of proteasome. Irwin, J.J., Sterling, T., Mysinger, M.M., Bolstad, E.S., Coleman, R.G., 2012. ZINC: a free
tool to discover chemistry for biology. J. Chem. Inf. Model. 52 (7), 1757–1768.
Acknowledgement Jain, A.N., 2004. Virtual screening in lead discovery and optimization. Curr. Opin. Drug
Discov. Devel. 7, 396–403.
Jakalian, A., Jack, D.B., Bayly, C.I., 2002. Fast, efficient generation of high-quality atomic
This work is supported in part by the Hibah Penelitian Pasca Doktor charges. AM1-BCC model: II: parameterization and validation. J. Comput. Chem. 23,
2018 from Ministy of Research, Technology, and Higher Education, 1623–1641.
Republic of Indonesia. The authors wish to thank Dr. Sandra Megantara Kazi, A., Ozcan, S., Tecleab, A., Sun, Y., Lawrence, H.R., Sebti, S.M., 2014. Discovery of
PI-1840, a novel noncovalent and rapidly reversible proteasome inhibitor with anti-
for helping in the analysis of pharmacophore model. tumor activity. J. Biol. Chem. 289 (17), 11906–11915.
Koes, D.R., Camacho, C.J., 2012. ZINCPharmer: pharmacophore search ZINC database.
Appendix A. Supplementary data Nucleic Acids Res. 40, 409–414.
Kollman, P.A., Massova, I., Reyes, C., Kuhn, B., Huo, S., Chong, L., Lee, M., Lee, T., Duan,
Y., Wang, W., Donini, O., Cieplak, P., Srinivasan, J., Case, D.A., Cheatam, T.E., 2000.
Supplementary data associated with this article can be found, in the Calculating structures and free energies of complex molecules: combining molecular
online version, at https://doi.org/10.1016/j.compbiolchem.2018.08. mechanics and continuum models. Acc. Chem. Res. 33, 889–897.
Kumalo, H.M., Soliman, M.E., 2015. Per-residue energy footprints-based pharmacophore
009. modeling as an enhanced in silico approach in drug discovery: a case study on the
identification of novel b-Secretase1 (BACE1) inhibitors as anti Alzheimer agents. Cell.
References Mol. Bioeng. 9 (1), 175–189.
Li, H., Leung, K.-S., Wong, M.-H., 2012. idock: a multithreaded virtual screening tool for
flexible ligand docking. Proceedings of the 2012 IEEE Symposium on Computational
Agrawal, R., Jain, P., Dikshit, S.N., Bahare, R.S., Ganguly, S., 2013. Ligand-based phar- Intelligence in Bioinformatics and Computational Biology (CIBCB). pp. 77–84.
macophore detection, screening of potential pharmacophore and docking studies, to Liu, J., Shaik, S., Dai, X., Wu, Q., Zhou, X., Wang, Z., Wei, W., 2015. Targeting the ubi-
get effective glycogen synthase kinase inhibitors. Med. Chem. Res. 22 (11), quitin pathway for cancer treatment. Biochim. Biophys. Acta 1855, 50–60.
5504–5535. Maier, J.A., Martinez, C., Kasavajhala, K., Wickstrom, L., Hauser, K.E., Simmerling, C.,
Allegra, A., Alonci, A., Gerace, D., Russo, S., Innao, V., Calabro, L., Musolino, C., 2014. 2015. ff14SB: improving the accuracy of protein side chain and backbone parameters
New orally proteasome inhibitors in multiple myeloma. Leukemia Res. 38, 1–9. from ff99SB. J. Chem. Theory Comput. 11 (8), 3696–3713.
Arba, M., Tjahjono, D.H., 2015. The binding modes of cationic porphyrin-anthraquinone McBride, A., Ryan, P.Y., 2013. Proteasome inhibitors in the treatment of multiple mye-
hybrids to DNA duplexes: in silico study. J. Biomol. Struct. Dyn. 33 (3), 657–665. loma. Expert Rev. Anticancer Ther. 13 (3), 339–358.
Arba, M., Ihsan, S., Ramadhan, L.O.A.N., Tjahjono, D.H., 2017a. In silico study of por- Miller, B.R., McGee, T.D., Swails, J.M., Homoeyer, N., Gohlke, H., Roitberg, A.E., 2012.
phyrin-anthraquinone hybrids as CDK2 inhibitor. Comput. Biol. Chem. 67, 9–14. MMPBSA.py: an efficient program for end-state free energy calculations. J. Chem.
Arba, M., Ruslin, Ihsan, S., Wahyudi, S.T., Tjahjono, D.H., 2017b. Molecular modeling of Theory Comput. 8, 3314–3321.
cationic porphyrin-anthraquinone hybrids as DNA topoisomerase IIβ inhibitors. Miller, Z., Kim, K.-S., Lee, D.-M., Kasam, V., Baek, S.-E., Lee, K.H., Zhang, Y.-Y., Ao, L.,
Comput. Biol. Chem. 71, 129–135. Carmony, K., Lee, N.-R., Zhou, S., Zhao, Q., Jang, Y., Jeong, H.-Y., Zhan, C.-G., Lee,
Augello, G., Modica, M., Azzolina, A., Puleio, R., Cassata, G., Emma, M.R., Sano, C.D., W., Kim, D.E., Kim, K.-B., 2015. Proteasome inhibitors with pyrazole scaffolds from
Cusimano, A., Montalto, G., Cervello, M., 2018. Preclinical evaluation of antitumor structure-based virtual screening. J. Med. Chem. 58 (4), 2036–2041.
activity of the proteasome inhibitor MLN2238 (ixazomib) in hepatocellular carci- Moreau, P., 2014. The emerging role of carfilzomib combination therapy in the man-
noma cells. Cell Death Dis. 9 (28), 1–13. agement of multiple myeloma. Expert Rev. Hematol. 7 (2), 265–290.
Beck, P., Dubiella, C., Groll, M., 2012. Covalent and non-covalent reversible proteasome Moreau, P., Richardson, P.G., Cavo, M., Orlowski, R.Z., Miguel, J.F.S., Palumbo, A.,
inhibition. Biol. Chem. 393 (10), 1101–1120. Harousseau, J.-L., 2012. Proteasome inhibitors in multiple myeloma: 10 years later.
Blackburn, C., Barrett, C., Blank, J.L., Bruzzese, F.J., Bump, N., Dick, L.R., Fleming, P., Blood 120, 947–959.
Garcia, K., Hales, P., Hu, Z., Jones, M., Liu, J.X., Sappal, D.S., Sintchak, M.D., Tsu, C., Obaidat, A., Weiss, J., Wahlgren, B., Manam, R.R., Macherla, V.R., McArthur, K., Chao,
Gigstad, K.M., 2010. Optimization of a series of dipeptides with a P3 threonine re- T.-H., Palladino, M.A., Lloyd, G.K., Potts, B.C., Enna, S.J., Neuteboom, S.T.C.,
sidue as non-covalent inhibitors of the chymotrypsin-like activity of the human 20S Hagenbuch, B., 2011. Proteasome regulator marizomib (NPI-0052) exhibits pro-
proteasome. Bioorg. Med. Chem. Lett. 20, 6581–6586. longed inhibition, attenuated efflux and greater cytotoxicity than its reversible ana-
Britton, M., Lucas, M.M., Downey, S.L., Screen, M., Pletnev, A.A., Verdoes, M., Tokhunts, logs. J. Pharmacol. Exper. Ther. 337 (2), 479–486.
R.A., Amir, O., Goddard, A.L., Pelphrey, P.M., Wright, D.L., Overkleeft, H.S., Kisselev, Orlowski, R.Z., Stinchcombe, T.E., Mitchell, B.S., Shea, T.C., Baldwin, A.S., Stahl, S.,
A.F., 2009. Selective inhibitor of proteasome’s caspase-like sites sensitizes cells to Adams, J., Esseltine, D.-L., Elliott, P.J., Pien, C.S., et al., 2002. Phase I trial of the
specific inhibition of chymotrypsin-like sites. Chem. Biol. 16, 1278–1289. proteasome inhibitor PS-341 in patients with refractory hematologic malignancies. J.
Buac, D., Shen, M., Schmitt, S., Kona, F.R., Deshmukh, R., Zhang, Z., Dudas, C.N., Mitra, Clin. Oncol. 20, 4420–4427.
B., Dou, Q.P., 2013. From Bortezomib to other inhibitors of the proteasome and Qin, J.-M., Huang, R.-Z., Yao, G.-Y., Liao, Z.-X., Pan, Y.-M., Wang, H.-S., 2017. Terminal
beyond. Curr. Pharm. Des. 19, 4025–4038. functionalized thiourea-containing dipeptides as multidrug resistance reversers that
Case, D.A., Betz, R.M., Cerutti, D.S., Cheatham III, T.E., Darden, T.A., Duke, R.E., Giese, target 20S proteasome and cell proliferation. Eur. J. Med. Chem. 126, 259–269.

70
M. Arba et al. Computational Biology and Chemistry 77 (2018) 64–71

Roe, D.R., Cheatham III, T.E., 2013. PTRAJ and CPPTRAJ: software for processing and Teicher, B.A., Tomaszewski, J.E., 2015. Proteasome inhibitors. Biochem. Pharmacol.
analysis of molecular dynamics trajectory data. J. Chem. Theory Comput. 9, 96, 1–9.
3084–3095. Wang, J., Wolf, R.M., Caldwell, J.W., Kollman, P.A., Case, D.A., 2004. Development and
Ryckaert, J.–P., Ciccotti, G., Berendsen, H.J.C., 1977. Numerical integration of the car- testing of a general amber force field. J. Comput. Chem. 25, 1157–1174.
tesian equations of motion of a system with constraints: molecular dynamics of n- Wehmer, M., Sakata, E., 2016. Recent advances in the structural biology of the 26S
alkanes. J. Comput. Phys. 23, 327–341. proteasome. Int. J. Biochem. Cell Biol. 79, 437–442.
Salomon-Ferrer, R., Goetz, A.W., Poole, D., Grand, S.L., Walker, R.C., 2013. Routine Yang, L., Wang, W., Sun, Q., Xu, F., Niu, Y., Wang, C., Liang, L., Xu, P., 2016.
microsecond molecular dynamics simulations with AMBER-Part II: particle Mesh Development of novel proteasome inhibitors based on phthalazinone scaffold. Bioorg.
Ewald. J. Chem. Theory Comput. 9 (9), 3878–3888. Med. Chem. Lett. 26 (12), 2801–2805.
Sangeetha, K., Sasikala, R.P., Meena, K.S., 2017. Pharmacophore modeling, virtual Zhang, J., Gao, L., Xi, J., Sheng, L., Zhao, Y., Xu, L., Shao, Y., Liu, S., Zhuang, R., Zhou, Y.,
screening and molecular docking of ATPase inhibitors of HSP70. Comput. Biol. Chem. Li, J., 2016. Design, synthesis and biological evaluation of novel non-covalent pi-
70, 164–174. peridine-containing peptidyl proteasome inhibitors. Bioorg. Med. Chem. 24 (23),
Singh, A., Jana, N.K., 2017. Discovery of potential Zika virus RNA polymerase inhibitors 6206–6214.
by docking-based virtual screening. Comput. Biol. Chem. 71, 144–151. Zheng, Q., Zhang, L., Zhou, Y., Luo, H., Wang, X., Huang, T., Xu, H., 2016. Dysregulation
Singh, J., Petter, R.C., Baillie, T.A., Whitty, A., 2011. The resurgence of covalent drugs. of ubiquitin-proteasome system in neurodegenerative diseases. Front. Aging
Nat. Rev. Drug Discov. 10, 307–317. Neurosci. 8, 1–10.

71