Вы находитесь на странице: 1из 11

Journal of Molecular Structure 1198 (2019) 126893

Contents lists available at ScienceDirect

Journal of Molecular Structure


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

Mechanism of inhibition of drug-resistant HIV-1 protease clinical


isolates by TMC310911: A molecular dynamics study
Suchetana Gupta, Sanjib Senapati*
BJM School of Biosciences and Department of Biotechnology, Indian Institute of Technology Madras, Chennai, 600036, India

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

Article history: HIV-1 protease continues to be a major target for therapy against AIDS. Although there are ten FDA
Received 19 June 2019 approved drugs available, long term use of these drugs elicit drug resistant mutations leading to major
Received in revised form challenges in therapy. Recently there have been reports of a new inhibitor, TMC310911 (TMC) that has
29 July 2019
shown significant activity against a wide spectrum of HIV clinical isolates that are resistant to even
Accepted 1 August 2019
darunavir (DRV), the best HIV-1 protease drug so far. However the mechanism of action of TMC is un-
Available online 2 August 2019
known. In this work, we have employed all-atom molecular dynamics simulation to understand how
TMC can be a potential drug candidate against mulidrug-resistant protease variants. Our results suggest
Keywords:
HIV-1 protease
that TMC has a dual mode of action. It acts as a conventional peptidomimetic inhibitor as well as a
Darunavir dimerisation inhibitor. It can bind to the active site cavity of dimeric protease in an extended confor-
TMC310911 mation similar to the available crystal structure pose. In parallel, it can also bind to the monomeric
Dimerisation protease and block the dimerisation interface of variant monomers. The detailed mechanism of action of
Molecular dynamics simulations TMC and the underlined mode of interaction could pave way for designing other potential HIV-1 protease
Protein-ligand interactions inhibitors.
© 2019 Elsevier B.V. All rights reserved.

1. Introduction undergo reversal in flap handedness to attain the closed state [8].
However, entry of ligand requires the flaps to be in an open state.
Human immunodeficiency virus 1 (HIV-1) continues to be a Thus, the ligand binding process involves two-steps: formation of a
worldwide threat to the human population. Among the various loose collision complex between open state protease and ligand,
targets tested, protease remains to be one of the most lucrative followed by ligand induced flap closure. This results in the forma-
targets against HIV. The enzyme, HIV-1 Protease plays an indis- tion of a tight complex between protease and ligand [9e11].
pensable role in the life cycle of the virus and is responsible for the The crystal structures for these inhibitor bound proteases have
cleavage of the polypeptides into functional proteins [1]. Small revealed a wealth of information regarding inhibitor binding. In
peptidomimetic drug molecules have been developed and used in most structures, the inhibitor has been found to occupy an
therapy. Presently, there are ten FDA approved drug molecules extended conformation within the active site cavity and the
against HIV-1 Protease [2e4]. The mode of action of these drugs hydrogen bonded interactions have also been seen [12]. Over the
primarily lies on binding to the active site of the protease. The course of time, the virus is reported to develop resistance against
active site of the homodimeric enzyme is gated by two flexible beta most of the FDA approved drugs and a number of resistant muta-
hairpin loops, commonly known as flaps. The flaps are present in tions have arisen in the protease molecule [2,3]. To overcome drug
multiple conformations at different states of the comformational resistance, a number of second generation inhibitors have been
equilibria [5]. Depending on the presence and absence of ligands at developed using the lead molecules obtained from various exper-
the active site, the enzyme is described to be in open, semi-open imental and computational studies [4]. While most of these mol-
and closed states [6,7]. Protease exists in a semi-open form pre- ecules have shown promising results, cases of resistance against
dominantly in the unliganded state; on ligand binding the flaps many clinical isolates have been reported [13e15]. Thus, the need
has arrived to develop better drugs that would be effective against
the drug resistant clinical isolates too.
Of the available HIV-1 protease drugs, the most recently
* Corresponding author. approved darunavir (Fig. 1a) still continues to be the most potent
E-mail address: sanjibs@iitm.ac.in (S. Senapati).

https://doi.org/10.1016/j.molstruc.2019.126893
0022-2860/© 2019 Elsevier B.V. All rights reserved.
2 S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893

Fig. 1. Molecular structures of the two inhibitors studied: (a) Darunavir (DRV) and (b) TMC310911 (TMC). The functional groups which are dissimilar in the two inhibitors are
circled in red.

against many of the clinical isolates [16e18]. Reports suggest that Protease is possible only in the open conformation of the dimeric
this can be attributed to the dual mode of action of DRV: it binds to enzyme. Hence, for our present study, we chose the open confor-
the active site like other peptidomimetic drugs as well as it pre- mation of the dimeric protease (PDB ID: 2PC0) [8] as the starting
vents dimerisation [19e22]. The X-ray structures of two indepen- structure for WT protease simulation. The two variant systems,
dent HIV-1 PR mutants show that DRV binds at the active site and PR20 and DRV01-48w were modelled using Swiss PDB viewer [28]
also at an exosite on the surface of one of the flaps (PDB IDs: 2HS1 using the WT open structure as the template. The structures of DRV
for V32I and 2HS2 for M46I) [23]. The latter mode of binding was and TMC were obtained from the corresponding crystal structures
initially thought to be an artefact of crystallization. However, (PDB ID: 4LL3 and 3R4B) [25,29] and docked to the open confor-
simulation reports suggested that indeed there are two distinct mation crystal structures. These structures were used to subse-
binding sites for DRV [21]. Apart from the active site in the dimeric quently run triplicate sets of 100ns simulation. For the simulation
protease, the inhibitor can also bind to the flap region in mono- of the monomer protease bound to the two inhibitors, the protocol
meric protease. ESI-MS data confirmed that DRV binds to WT mentioned in the work of Danzhi et al. [21] was followed. According
monomeric HIV-1 protease in a one-to-one molar ratio [19]. These to this protocol, we removed a single monomer from dimeric in-
studies hypothesised that in presence of DRV, the inhibitor binds in hibitor bound X-ray structures and used the resultant inhibitor
the proximity of the active site interface of the monomer and bound monomers for running triplicate sets of 100ns simulation. A
prevents a second monomer from binding. However, in absence of list of all simulated systems are detailed in Table 1.
the inhibitor, the two monomers interact at the active site interface
and form unstable dimerised protease subunits. Further in-
teractions at the terminal residues then give rise to stable and 2.2. Details of MD simulation
functional dimer units. FRET results of protease variants containing
some of the most commonly known mutations in DRV resistant MD simulations were carried out using AMBER12 [30] with
variants, viz. V32I, L33F, I54M and I84V mutations also showed that AMBER ff99 S B forcefield [31] and generalized Amber forcefield
in presence of 0.1 and 1.0 mM DRV, dimerisation of protease failed (GAFF) [32] for protein and drug molecules respectively. These are
[24]. well established forcefields used in protein-inhibitor complex
Recently, a new inhibitor, TMC (Fig. 1b) has been developed that simulations [33,34]. LEaP module of AMBER12 was used to add the
has shown highly promising results to inhibit the activity of pro- missing hydrogen atoms. One of the two catalytic Asp residues was
tease from DRV resistant clinical isolates [25]. This inhibitor has protonated according to NMR reports [35]. The protein molecules
presently completed phase 2 of clinical trial. Darunavir is the parent were energy minimised for 2000 steps using steepest descent and
molecule, based on which the TMC310911 has been designed. As conjugate gradient algorithm. The energy minimised structure was
seen from the structures of the two inhibitors in Fig. 1, Darunavir then hydrated in a cubic water box comprising of TIP3P water
and TMC310911 share chemical similarity up to the sulfonyl group. molecules [36]. Charge neutralisation was performed by adding
Beyond that, an additional benzothiazol-2-yl-1-(1-cyclopentyl- requisite number of Cl-ions. SHAKE algorithm [37] was used to
piperidin-4-yl)-amine moiety exists in TMC310911 (marked in red
circle), corresponding to the phenyl amine moiety in Darunavir Table 1
(marked in red circle). List of systems studied. Each system has been simulated in triplicate.
In this work, we attempt to understand how TMC acts as a
Type Description Inhibitor Time
potent inhibitor against the DRV resistant clinical isolates. To study
this, we started with MD simulations of two DRV resistant clinical HIV-1 Protease Dimer WT Darunavir 100ns
ISO1 Darunavir 100ns
isolates, viz PR20 (ISO1) [26] and DRV01-48w (ISO2) [27] bound to ISO2 Darunavir 100ns
TMC. Our results have shown that TMC can bind to the active site of WT TMC310911 100ns
dimeric protease, in addition to the strong binding to the flap re- ISO1 TMC310911 100ns
gion of monomeric protease similar to DRV. Moreover, the differ- ISO2 TMC310911 100ns
HIV-1 Protease Monomer WT Darunavir 100ns
ential binding observed between TMC and DRV, with the clinical
ISO1 Darunavir 100ns
isolates tested here, has provided new working hypothesis to un- ISO2 Darunavir 100ns
derstand the better activity of TMC over DRV and other protease WT TMC310911 100ns
drugs/inhibitors. ISO1 TMC310911 100ns
ISO2 TMC310911 100ns
HIV-1 Protease Dimer WT Apo 100ns
ISO1 Apo 100ns
2. Methods ISO2 Apo 100ns
HIV-1 Protease Monomer WT Apo 100ns
2.1. System preparation ISO1 Apo 100ns
ISO2 Apo 100ns
HIV-1 Protease Monomer PDB ID: 2HS1 Darunavir 50ns
It is known from previous studies that ligand entry into HIV-1
S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893 3

constraint bonds involving hydrogen atoms. Electrostatic in- obtained the coordinates for the two inhibitors. We introduced the
teractions were calculated with a cut-off distance of 12 Å using necessary mutations in these structures to generate starting
Particle mesh Ewald method [38]. Initial minimisation sets were structures for ISO1 and ISO2 proteases bound to the two inhibitors.
performed to avoid bad contacts. This was followed by equilibration These systems were then run for 100 ns to obtain stable trajectories
using NVT ensemble at 300 K for about 500 ps. The density of the in triplicate.
systems were then equilibrated using NPT ensemble at 1 atm
pressure for 1 ns. The time step used was 2 fs throughout the 3.1. Wild type protease attains closed conformation in presence of
minimisation-equilibration-production run phase. After the energy both DRV and TMC
values and density were seen to converge, the systems were sub-
jected to 100 ns production using NPT ensemble at 300 K and 1 atm In HIV-1 protease, the flap conformations are extremely
pressure. The coordinates were saved every 2 ps. The trajectories important for ligand entry. Once the substrate (or inhibitor) enters
were visualised using VMD [39] and analyses were performed us- the active site pocket, the flaps undergo a closure induced by ligand
ing CPPTRAJ [40] module of AMBER12. binding. When the flaps are in a proper closed state, the protease
cleaves the substrate into shorter peptides. The strategy for
designing peptidomimetic inhibitors is that the inhibitor should be
2.3. Calculation of free energy of ligand binding
able to keep the protease in the proper closed state. In addition,
once the inhibitor binds to the active site cavity, the flaps have to
To predict the free energies of binding of ligands to receptor, the
remain in a closed state to prevent exit of inhibitor and entry of
two most commonly used methods are the molecular mechanics
natural substrate. Thus, flap dynamics play a crucial role in the
generalized Born surface area (MM-GBSA) and molecular me-
mechanism of inhibition by protease. As a result, it is essential to
chanics PoissonBoltzmann surface area (MMPBSA) methods. Even
understand the conformation of the flaps when a particular in-
though the PoissonBoltzmann (PB) method is theoretically much
hibitor is binding to protease to understand the inhibitor's efficacy.
more rigorous than the generalized Born (GB) method, both the
The time-averaged structures obtained from our MD simula-
methods are equally efficient to predict the correct binding affin-
tions were superposed to the corresponding closed state crystal
ities [41e45]. Here, we used the MMGBSA method to calculate the
structures of inhibitor bound WT protease and have been shown in
relative free energies for binding of ligands to HIV-1 protease. For a
Fig. S1. While for the WT protease bound to both the inhibitors, the
given complex, the free energy of binding is calculated as
flaps of protease reached proper closed state as evident in the
DGbind ¼ DH  TDS figure, for both the variants, the flaps were never properly closed.
The close matching of the simulated structures for WT protease
with the corresponding crystal structures of WT protease bound to
DH ¼ DEelec þ DEvdW þ DGpolar þ DGnonpolar
the two inhibitors validates the simulation protocol and forcefield
used in this study. On superposing the structures, we note that
where DEelec and DEvdW are the electrostatic and van der Waal's
there exists significant deviation in flap and hinge regions in the
contributions respectively, and DGpolar and DGnonpolar are the polar
variants (more pronounced in ISO1) with respect to the closed state
and nonpolar solvation terms, respectively. The nonpolar energy is
crystal structures. The RMSD values of the entire protein-inhibitor
calculated by the solvent accessible surface area (SASA) while the
complex with respect to the closed state crystal structures for all
polar contribution is estimated using the GB model with an
systems were plotted and shown in Fig. 3a. While the WT protease
external dielectric constant of 80 and an internal dielectric constant
complexed with both the inhibitors showed the least RMSD values,
of 4. As our calculations involve binding of similar types of ligands
in case of the variants, the deviation was quite significant. The
to the receptor, the entropic contributions are neglected. Thus,
complex of ISO1 bound to TMC exhibited the largest change. Also,
these computed values will be referred to as relative binding free
our simulation of the ISO1 protease bound to DRV has shown the
energies.
typical distortion in the hinge region because of the presence of the
M36I mutation (Fig. S1c), in consistence with previous reports [46].
3. Results and discussion Similar observations have also been seen in ISO1 bound to TMC
(Fig. S1d). However, the absence of this mutation in ISO2 has led to
We selected two DRV resistant clinical isolates, viz PR20 (ISO1) the preservation of the overall structure of this hinge region in the
and DRV01-48 W (ISO2) to examine the binding motif of TMC ISO2 bound to DRV and TMC inhibitors.
relative to DRV. Both the variants had V32I, L33F, I54M and I84V One of the common metrices to measure the flap closure is to
mutations which are known to occur in most of the DRV resistant calculate the distance between the flap tips. In this case, the dis-
clinical isolates. The multi-drug resistant variant PR20 contains 20 tance between CA atoms of Gly48 and Gly48’ residues are
mutations on each monomer, the mutations being evenly distrib- measured on each system. The reference value was chosen as the
uted between the active and non-active sites. The mutations are distance noted in the respective closed state crystal conformations
D30N, V32I and I84V at the active site and Q7K, L10F, I13V, I15V, of wild type HIV-1 protease bound to DRV and TMC. This value is
L33F, E35D, M36I, S37N, I47V, I54L, Q58E, I62V, L63P, A71V, N88D, approximately 6 Å in both the closed states. In our simulations, the
L89T, L90 M at the non-active site and are indicated in Fig. 2a. On dimeric wild type HIV-1 Protease bound to both inhibitors have
the other hand, DRV01-48w, another clinical isolate consists of 16 attained a closed state. The distance between the flap tips have
mutations on each monomer. These mutations are located on both matched very well with the reported crystal values of 6.10 Å and
the active site and non-active sites as represented in Fig. 2b. The 5.68 Å respectively (Fig. 3b), thereby validating our simulation
active site mutations are V32I and I84V. The non-active site mu- protocol. The inset plot in Fig. 3b shows that over the course of the
tations comprise of L10F, V11I, I13V, G16E, L24I, L33F, M36I, M46I, simulation, the distance between the flap tips were highly fluctu-
I47V, I54 M, Q58E, L63P, I64V, N83D. We started our simulation ating and the value never reached the crystal value in either of the
from the open conformation of HIV-1 protease by taking the crystal clinical isolates, ISO1 or ISO2. This can be attributed to the curling of
structure, PDB ID: 2PC0 for the WT simulation. We aligned this one of the flaps observed in the course of simulation. However,
structure to the crystal structure of WT dimeric protease bound to experiments show that the inhibitor TMC is indeed potent against
DRV (PDB ID: 4LL3) and TMC (PDB ID: 3R4B) respectively and mutant species, including some of the ones resistant to DRV.
4 S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893

Fig. 2. Mutations in HIV-1 Protease clinical isolates. The locations of mutations in clinical isolates (a) PR20 (ISO1) and (b) DRV01-48w (ISO2) are shown. Red spheres show the
active site mutations and blue spheres show non-active site mutations.

Fig. 3. Differential dynamics in WT and mutant HIV proteases bound to inhibitors. (a) Time evolution of protease backbone RMSD in the complexes of DRV-WT (black solid),
DRV-ISO1 (black dotted), DRV-ISO2 (cyan dotted), TMC-WT (red solid), TMC-ISO1 (red dotted), and TMC-ISO2 (violet dotted) with respect to the crystal structure of corresponding
inhibitor bound WT protease. (b) Time evolution of flap tip - flap tip distances in WT protease bound to DRV (black) and TMC (red). The blue and green solid lines indicate the crystal
values form DRV and TMC bound proteases (PDB IDs: 4LL3 and 3R4B respectively). Inset shows the variations in the mutants: ISO1 (solid) and ISO2 (dotted).

Perhaps, some alternate mechanism exists in this case which will for the other monomer complexes.
be discussed subsequently. The time averaged structures of all the six monomer complexes
studied (see Table 1) have been shown in Fig. 4. The mode of
binding of the inhibitors to the monomer is strikingly different than
3.2. Both inhibitors bind to the monomeric protease effectively that of the dimer. While in dimer the inhibitor fully occupies the
active site of the protease, parallel to the active site in an extended
As discussed in the Introduction, darunavir was proposed to way, in case of monomer, the inhibitors are in a position perpen-
have two distinct binding modes. It binds to the active site in the dicular to the active site. While the binding motif for DRV in both
dimeric protease like other protease inhibitors (PDB ID: 4LL3) as the variants, ISO1 and ISO2 are similar to that reported earlier [21],
well as to an exosite near the flaps (PDB ID: 2HS1, 2HS2). Since TMC it is slightly different in WT. In WT protease, the DRV molecule flips
is an extension of the Darunavir molecule, it might also have a dual from its initial position by almost 180 . The complexes studied in
mode of action. In order to study this, the monomeric WT and the previous report [21] also had a mutation as in ISO1 and ISO2
variant protease were simulated bound to both inhibitors. To do so, and the binding pose reported in that study matches the binding
we followed the protocol mentioned in the study by Danzhi et al. pose of DRV in our simulation results in ISO1 and ISO2. Since WT
[21] where a monomer unit from DRV bound dimer protease (PDB does not have favourable interactions due to similar residues in
ID: 2HS1) was removed and taken forward for simulation. Thus, we mutants, the ligand flips. Due to the presence of mutations near the
took the DRV bound V32I mutant protease (PDB ID 2HS1) and binding site, in both the variants, the DRV molecule maintains its
removed a monomer, thereby generating a mutant monomer binding pose similar to the starting structure unlike in WT. The
bound to DRV. This mutant complex was simulated for 50 ns and its inhibitor molecule TMC is longer than DRV. As a result, it binds to
binding mode was determined. The average structure obtained the protease monomer perpendicular to the active site and while
from this simulation matched very well with that reported in doing so, it occupies the dimerisation interface fully covering the
Ref. 21, as shown in Fig. S2a. Also, most of the interacting residues of active site and also the flaps to some extent. The greater length and
protease in close proximity to DRV matched with those reported in bulkier shape of the inhibitor prevent it from flipping in the WT and
that study [21], viz. Val82, Ile84, Ile32, Ala28, Ile47 and Ile50 variant protease complexes. So, both the WT and the two variants
thereby validating our simulation protocol (shown in Fig. S2b). have a similar binding mode for TMC (Fig. 4). The four-stranded
Thus, the same protocol was followed to carry out our simulations
S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893 5

Fig. 4. Time-averaged structures of inhibitor bound monomeric protease. Results are shown for (a) DRV-WT protease, (b) DRV-ISO1 protease, (c) DRV-ISO2 protease, (d) TMC-
WT protease, (e) TMC-ISO1 protease, (f) TMC-ISO2 protease.

beta sheet that is present at the terminal regions in dimer proteases protease was found to be 56.373 kcal/mol whereas that for DRV
disappeared in the inhibitor-monomer complexes. As a result, the bound to the two variants, ISO1 and ISO2 were 47.174 kcal/mol
terminal residues (residues 1e4 and 95e99) in monomeric com- and 48.170 kcal/mol respectively. These two variants are reported
plexes become highly flexible and exhibited large variations. to be resistant to DRV, and our calculated binding affinities for the
We have also compared the B-factor values of protease residues mutant protease reflected the similar trend with lower binding
in dimer and monomer complexes and the results are shown in energy values. On the other hand, when the inhibitor is TMC,
Fig. 5. It is evident from this figure that the binding of inhibitor to the binding energy values for all the proteases (including WT and
the protease monomer reduces the flexibility of most of the pro- both the variants) show comparable values (65.919 kcal/
tease residues more prominently than in the dimer, indicating mol, 62.355 kcal/mol and 59.265 kcal/mol respectively). It is
stronger binding of these inhibitors to the monomeric protease interesting to note that these values of binding free energies for the
than their binding to the protease dimer. A deviation to this trend is inhibitor TMC to all variants of protease are much higher than that
observed in the terminii due to the absence of second monomer. for DRV bound to even WT protease. Stronger binding coupled with
Moreover, unlike the DRV bound monomer, the flap residues in closed flap conformation could explain the observed better inhib-
TMC bound monomer show considerably greater B-factor than in itory effect of TMC over DRV.
the dimer. This is because, the binding mode of TMC is such that it In monomeric proteases, the presence of mutations in isolates
occupies the active site and terminal residues more than the flaps ISO1 and ISO2 also decreases the binding affinity of DRV compared
(to be validated in subsequent sections through residue level to WT (30.516 kcal/mol and 28.438 kcal/mol in the variants
interaction data). On the other hand, since binding of DRV extends versus 37.938 kcal/mol in WT). This is in consistence with a pre-
till the flap region, the B-factor values for flap residues in DRV vious ESI-MS study [19], which has reported that the presence of
bound monomer is significantly lower. the four mutations V32I, L33F, I54M and I84V (present in both our
isolates) decreases the binding affinity of DRV to monomeric pro-
tease. In this present study, as seen in Table 2, the binding affinity of
3.3. TMC binds to the protease variants more strongly than DRV
DRV to ISO1 decreases by ~20% compared to WT while in ISO2, the
reduction is by ~25%. However, from the same table we note that in
In order to understand the differential effects of the two in-
case of TMC, the binding of the inhibitor to monomeric protease
hibitors on HIV-1 Protease, we used binding energy analyses. The
(WT as well as variants) is strong. In fact, the binding affinities for
binding energy values were calculated for each inhibitor against
both WT and the two variants are comparable (44.359 kcal/
WT and mutant protease systems using MMGBSA and are tabulated
mol, 42.623 kcal/mol and 40.467 kcal/mol respectively). A
in Table 2. The binding energy for Darunavir to WT dimeric HIV-1
6 S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893

the monomeric protease. We also calculated the respective H-bond


lifetimes. To define a H-bond, the geometric criteria of H-bond were
used where a radial distance of 3 Å between the donor and acceptor
atoms and donor-H-acceptor angle greater than 135 were
considered. The number of residues involved in long lasting
hydrogen bonds is always greater for TMC than Darunavir, for both
WT and the two variants. For instance, in DRV and TMC bound to
WT protease, the numbers are nine and thirteen respectively, as
seen in Table 3. In case of the inhibitors bound to the variant pro-
teases, the numbers are ten and twelve for ISO1 and eight and
eleven for ISO2 when the inhibitors are DRV and TMC respectively,
as shown in Table 3.
In case of WT monomeric protease, both Darunavir and TMC
form hydrogen bonds with some of the most important function-
ally relevant residues of protease, including those of the active site
(residues 25, 28), flaps (residues 49e51) and the 80s loop (residues
80e81) and the % lifetime is quite significant. A detailed list of
residues have been tabulated in Table 3 with the % lifetime of H-
bond. Across all complexes studied (WT and the two variants), DRV
shows a greater lifetime of H-bonds formed with the flap residues
(residues 49e51) than TMC as shown in Table 3. However, for the
other functionally relevant residues like the active site (residues
25e29) and the 80s loop (residues 80e81), TMC forms either
comparable or stronger hydrogen bonds having greater lifetime in
all three complexes as seen in Table 3. This again corroborates our
observation of the binding pose of the two inhibitors in the
monomer complexes as discussed in the previous section. When
the inhibitor is TMC, due to the extended length of the inhibitor,
certain additional residues from the N-terminus are also involved
in forming H-bonds having appreciable lifetime in all three com-
plexes. For instance, in WT protease, Thr4, Leu5, Val32, Ile47 resi-
dues form additional hydrogen bonds with TMC compared to DRV.
In ISO1 and ISO2, similarly, the residues forming additional
hydrogen bonds are Ala28, Gly49, Leu54, Thr80, Pro81, and Val82.
The greater number of hydrogen bonds of TMC with some of the
additional residues of protease may account for the greater binding
Fig. 5. Inhibitor binding to protease monomer is stronger than the dimer. The B-
factor values of protease residues bound to (a) DRV and (b) TMC310911. Color code: WT
energy of this inhibitor compared to DRV.
(black), ISO1 (blue), and ISO2 (red). Solid lines stand for the dimeric and dotted lines Apart from the hydrogen bonds, hydrophobic contacts also play
for the monomeric proteases. The data for the terminal residues are omitted. an important role in binding of the ligands to the protease active
site. To explain the stronger binding of TMC over DRV, the hydro-
phobic interactions of the monomeric protease with the inhibitors
have also been examined. These have been identified from the time
detailed discussion for its causes will be taken up in subsequent averaged structures generated from the last 10 ns frames of simu-
sections. It is also to be noted from Table 2 that the binding energy lation trajectories. The hydrophobic interactions have been
of both inihibitors to monomeric protease is larger than half the captured by finding the pairs of heavy atoms of protease and ligand
binding energy values of the inhibitors to the dimer (dimer counts that reside within cut-offs of 4 Å. The network of interactions of
the same interaction twice). This corroborates very well with the B- protease with the two inhibitors have been shown graphically in
factor profiles in Fig. 5. Fig. 6. The number of protease residues showing hydrophobic in-
teractions with TMC is greater than the DRV bound protease, both
WT and the two variants. Also, the interacting residues are dis-
3.4. Differential H-bonding interactions are observed for DRV
similar in case of the two inhibitors. The increased length of TMC
versus TMC in monomeric protease
compared to DRV makes contacts with certain additional residues
of protease as seen in Fig. 6. These residues include some of the
To understand the basis of stronger binding affinity of TMC over
most functionally important residues in protease, viz. Gly27, Ala28,
DRV, we counted the number of hydrogen bonded interactions in

Table 2
Binding energy values of the inhibitors against dimeric and monomeric WT and variant proteases. Standard deviations are included in parentheses.

System Inhibitor Binding Energy in Dimer (kcal/mol) Binding Energy in Monomer (kcal/mol)

WT DRV 56.373 (0.070) 37.938 (0.021)


ISO1 DRV 47.174 (1.944) 30.516 (1.460)
ISO2 DRV 48.170 (0.064) 28.438 (1.630)
WT TMC310911 65.919 (2.610) 44.359 (2.247)
ISO1 TMC310911 62.355 (0.431) 42.623 (2.217)
ISO2 TMC310911 59.265 (3.761) 40.467 (3.156)
S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893 7

Table 3
List of monomeric protease residues that form H-bonds with the inhibitors (first column). Also shown are the % lifetime of each of these H-bonds. Residues in bold indicate
positions of mutations.

Protease Residue DRV-WT TMC-WT DRV-ISO1 TMC3-ISO1 DRV-ISO2 TMC-ISO2

2 18.77 13.67
4 12.3
5 84.15
8 29.16 35.2
25 76.64 84.09 30.91 27.67 41.13 38.76
28 31.41 26.99 15.5 18.34
29 64.53 24.99 29.4 20.3 44.54
30 52.02 49.75 23.98 37.62 35.96 68
32 20.55
47 16.75
48 12.38 10.94
49 33.81 14.61 23.79 22.87
50 86.8 13.67 20.26 17.79 30.63 16.38
51 44.78 43.63 61.8 13.87 50.21 15.64
52 22.9 23.08
54 13.81 10.56
80 31.86 32.64 19.24 24.98
81 38.69 36.01 16.49 26.67
82 37.37 18.22 11.23
87 17.32

Asp29 (active site), Gly48 (flap), Val82 (80 S loop) and are mostly bonds formed by DRV are stronger in WT than the variants, as
conserved in nature as highlighted in Fig. 6. The interacting resi- expected since these variants are clinical isolates, resistant to DRV.
dues are mainly located along the dimerisation interface and thus On the other hand, when the inhibitor is TMC, the % lifetime values
the inhibitor blocks the space for the second monomer to bind in for WT, ISO1 and ISO2 are 38.03, 47.04 and 42.55 respectively, in
the mutant monomer. For Darunavir also, we notice similar in- consistence with comparable binding energy values as reported
teractions, though the number of such interacting residues is not earlier in Table 2.
that sharp. This could be one of the plausible reasons of why TMC is At a closer view, we see that the hydrogen bonds of DRV in the
potent against some of the DRV resistant isolates. WT protease are mostly restricted to the active site in both the
monomers and one flap. On the other hand, in case of the variants,
3.5. Extended interactions with more conserved residues are the active site residues (residues 126e131) and flap residues (res-
observed for TMC in dimeric protease idues 147 and 149) also form hydrogen bonds with appreciable
lifetime as listed in Table S1. For TMC bound systems, the distri-
To explain the reported stronger efficacy of TMC over DRV [25], bution of hydrogen bonds is through both the monomers (active
we also compared the interactions of the two inhibitors with the site and flaps) in WT and both the variants as seen in Table S1. In the
protease dimer. The crystal structures of WT dimeric protease complexes of the variants with TMC, we note a very strong H-bond
bound to DRV and TMC have reports of hydrogen bonds with res- formed with the flap residue Ile50 in both the monomers. In
idues Asp25, Asp29 and Asp30 in both the monomers. Our simu- addition, we also observe stronger hydrogen bonds formed be-
lations of inhibitors bound to dimeric protease have been able to tween the active site residues of variants with TMC in comparison
capture these hydrogen bonds in WT as listed in Table S1, which to DRV. These stronger and additional H-bonds account for the
further validates our simulation protocol and the forcefield used in greater binding affinity of TMC to the dimeric protease complexes.
the study. For instance, the lifetimes of hydrogen bonds of Asp25 It is also interesting to note that TMC forms hydrogen bonds with
with DRV and TMC are 45.76% and 48.7% respectively. For Asp29, most of the conserved residues of protease like the active site and
these values are 54.1% and 51.76% respectively. Asp30 from WT flap residues. However, it forms a strong hydrogen bond with
protease forms hydrogen bonds with DRV and TMC having lifetime Asn30 in both the variants, which is a site of mutation. But we have
of 67.8% and 62.64% respectively (Table S1). Strong hydrogen observed that this hydrogen bond is formed through the main
bonding with these residues in the second monomer is also chain. As a result, the effect of mutation is compensated. H bond
observed, as shown in Table S1. formations with the conserved residues help in maintaining sta-
For WT proteases, when the inhibitor is TMC, due to its addi- bilising contacts with the protease molecule, irrespective of the
tional length, the number of H-bonds formed with significant mutations arising as a result of drug resistance. This could be a
lifetime (more than 10%) is greater compared to DRV bound pro- plausible reason why this inhibitor is effective even against certain
tease as shown in Table S1 (eight versus ten). Apart from the drug resistant variants of protease as reported experimentally.
common hydrogen bonds with residues Asp25, Asp29, Asp30 and Apart from hydrogen bonding, another significant contribution
Ile50 as mentioned above, the specific protease residues that form to strong binding of ligands can be explained on the basis of the
H bonds with TMC are located on or near the active site in the hydrophobic interactions of dimeric protease residues with the
second monomer, viz. Ala127 and Val131. Although the number of inhibitors. These hydrophobic contacts are calculated from the time
residues showing appreciable H-bond interactions is much more in averaged structures generated from the last 10 ns frames of our
all the variants compared to WT for both inhibitors, when we simulation trajectories and have been plotted graphically using
consider the key residues (residues 25, 50, 124 and 149), the Ligplot [47] and are shown in Fig. S3. The hydrophobic interactions
average % lifetime values present a different scenario. For DRV have been captured by finding the pairs of heavy atoms of protease
bound to WT, ISO1 and ISO2, these values are 34.23, 16.07 and 20.66 and ligand that reside within cut-offs of 4 Å respectively. These
respectively, in consistence with similar trend in binding energy hydrophobic contacts play a significant role in explaining the dif-
values as reported earlier in Table 2. This shows that the key H- ferential binding affinities of the inhibitors. The longer inhibitor
8 S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893

Fig. 6. 2D representation of the interactions of monomeric protease with the inhibitors. Results are shown for (a) WT-DRV (b) WT-TMC (c) ISO1-DRV (d) ISO1-TMC (e) ISO2-
DRV (f) ISO2-TMC. The inhibitors are shown in stick representation and the interacting protease residues are shown in red semi-circles. Some of the key interacting residues are
encircled.

TMC shows much greater number of hydrophobic contacts with the and the two variants, we see that in WT the key interacting residues
protease compared to DRV in both WT and the two variants, are more in number (18) than the two variants (ISO1: 10 and ISO2:
explaining the significantly greater binding energies as shown in 15). This is also reflected in the stronger binding energy values of
Table 2. Comparing the number of contacts for DRV bound to WT DRV-WT complex than the two variants bound to DRV (Table 2). On
S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893 9

Table 4 in Table 2, we observe that the stability of the dimeric protease-


Energy of dimerisation of WT and variant protease monomers. Standard inhibitor complexes are greater than the dimerisation energies.
deviations are shown in parentheses.
Thus, binding of the inhibitors to the dimer is a favourable process,
System Dimerisation Energy (kcal/mol) irrespective of the presence or absence of mutations.
WT 34.677 (1.790) On the other hand, for monomer-inhibitor complexes, the sce-
ISO1 26.099 (0.338) nario is slightly different. We note from the observed values listed
ISO2 27.595 (3.030) in Tables 2 and 4 that the binding affinity for the variant protease is
greater towards an inhibitor than the binding affinities of two
monomers. This effect is more pronounced when the inhibitor is
the other hand, when the inhibitor is TMC, the number of hydro- TMC as compared to DRV. This means that the dimerisation energy
phobic contacts for all three complexes (WT, ISO1 and ISO2) are is less than the binding affinity of protease variant monomer to
almost comparable which is on similar lines as the binding energy TMC. Thus, we may hypothesise (Fig. 7) that when a solution
values for the three complexes. From these protein-ligand inter- contains two molecules of mutant monomers and one molecule of
action profiles, we notice that the hydrophobic contacts are mostly inhibitor, before one monomer can bind to another monomer
centered around the active site. Also, some of the residues that forming a stable dimer, the inhibitor blocks the flap, active site and
interact with the additional cyclopentylpiperidine moiety of TMC, terminii region (as noted in previous sections). As a result, the
as shown in Fig. S3 are located near the 80s loop and flap residues, dimer forming interface is blocked (validated from our residue-
all of which are functionally relevant for protease. Thus, from our level findings as discussed in previous sections), leading to non-
residue-level interaction data, it is evident that the cyclo- appearance of a stable and fully functional dimeric protease. Thus
pentylpiperidine group present in the longer inhibitor TMC plays a TMC and similar inhibitors are active even against the drug-
significant role in increasing its binding affinity with the protease. resistant clinical variants and are susceptible by their dual mode
of action. They bind to active site like other peptidomimetic drugs
3.6. TMC prevents the formation of dimeric protease variant as well as bind and block the dimerisation interface, thereby dis-
rupting the dimer formation. The very comparable dimerisation
From our discussions above, we observe that the binding modes energy and the energy of DRV binding of the tested clinical isolates
of the inhibitors are different to the monomeric and dimeric pro- are consistent with the fact that these variants are DRV-resistant.
tease. In dimer, the inhibitors lie at the active site in an extended
conformation, a pose matching that of the crystal structures. 4. Conclusion
However, for monomeric protease, the inhibitors assume a position
perpendicular to that of the crystal pose. As a result, the inhibitors Even though long-term usage of the FDA approved protease
fully cover the residues at the dimerisation interface (active site and inhibitors elicit drug resistance mutations, there are a number of
terminal residues) of the protease monomer. Also, we have inhibitors that are currently in clinical trials exhibiting potent ac-
observed that the binding of the smaller DRV to the monomeric tivities against drug resistant clinical isolates. TMC is one such in-
protease is different than the longer TMC. The additional length of hibitor that has shown promising results against isolates that are
TMC accounts for interactions with additional residues of protease. resistant to even darunavir, the best protease inhibitor till date. In
As a result, the binding energy values are significantly different as this work, we attempted to unravel the mechanism of action of
seen in Table 2. TMC. Our results from all-atomistic MD simulations suggest that
In order to propose a mechanism of stronger binding of TMC the inhibitor has dual mode of action e it binds to the active site of
to the variant HIV proteases, we also calculated the dimerisation dimeric protease, as well as it binds at the monomer-monomer
energies of WT protease and the dimerisation energies of interface strongly. Key crystal contacts and hydrogen bond in-
two protease variants: ISO1 and ISO2. The respective values teractions observed in the available dimeric crystal structure were
were 34.677 kcal/mol, 26.099 kcal/mol and 27.595 kcal/mol as maintained throughout our simulation in WT protease. Similar
included in Table 4. These values have been calculated by taking the interactions have also been observed in darunavir-resistant clinical
difference between the free energy of the dimer and the sum of the isolates, explaining the greater potency of TMC over darunavir and
free energies of the two monomers. Comparing these with the other protease inhibitors. This binding motif of TMC to dimeric
binding energies for the two inhibitors with the dimeric proteases protease makes it a potent peptidomimetic inhibitor. Interestingly,

Fig. 7. Proposed mechanism of action of TMC to HIV protease variants. It has a dual mode of action. It can bind to the active site cavity of dimeric WT protease, and also to the
variant monomer blocking the protease dimerisation.
10 S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893

TMC also binds to the monomeric protease in a pose that blocks the inhibition by darunavir, Proc. Natl. Acad. Sci. 111 (2014) 12234e12239.
https://doi.org/10.1073/pnas.1400027111.
monomer-monomer interactions. The additional cyclo-
[20] E.S. Furfine, T. Spector, D.J.T. Porter, et al., Two-step binding mechanism for
pentylpiperidine moiety in TMC is responsible for forming addi- HIV protease inhibitors, Biochemistry 31 (1992) 7886e7891. https://doi.org/
tional contacts and interactions compared to DRV in the clinical 10.1021/bi00149a020.
isolates studied. In summary, our results suggest that the binding of [21] D. Huang, A. Caflisch, How does darunavir prevent HIV-1 protease dimer-
ization? J. Chem. Theory Comput. 8 (2012) 1786e1794. https://doi.org/10.
TMC to the mutant protease monomer is more favourable than the 1021/ct300032r.
dimerisation of two variant monomers. As a result, this inhibitor [22] M. A, D. D, H. H, et al., Mechanism of Darunavir (DRV)’s high genetic barrier to
could produce sufficient potency against the clinical isolates with HIV-1 resistance: a key V32I substitution in protease rarely occurs, but once it
occurs, it predisposes HIV-1 to develop DRV resistance, mBio 9 (2018)
multiple mutations. The obtained knowledge of detailed in- e02425-17, https://doi.org/10.1128/mBio.02425-17.
teractions between protease and inhibitors could help in designing [23] A.Y. Kovalevsky, F. Liu, S. Leshchenko, et al., Ultra-high resolution crystal
future HIV protease inhibitors. structure of HIV-1 protease mutant reveals two binding sites for clinical in-
hibitor TMC114, J. Mol. Biol. 363 (2006) 161e173. https://doi.org/10.1016/J.
JMB.2006.08.007.
Acknowledgements [24] Y. Koh, M. Aoki, M.L. Danish, et al., Loss of protease dimerization inhibition
activity of darunavir is associated with the acquisition of resistance to dar-
unavir by HIV-1, J. Virol. 85 (2011) 10079e10089. https://doi.org/10.1128/JVI.
We thank DST, Govt. of India, Project No. DST/INT/SWD/VR/P- 05121-11.
08/2016 for funding. The authors acknowledge the computing fa- [25] I. Dierynck, H. Van Marck, M. Van Ginderen, et al., TMC310911, a novel human
cility of the P. G. Senapathy Centre, IIT Madras. immunodeficiency virus type 1 protease inhibitor, shows in vitro an improved
resistance profile and higher genetic barrier to resistance compared with
current protease inhibitors, Antimicrob. Agents Chemother. 55 (2011)
Appendix A. Supplementary data 5723e5731. https://doi.org/10.1128/aac.00748-11.
[26] J. Agniswamy, C.-H. Shen, A. Aniana, et al., HIV-1 protease with 20 mutations
exhibits extreme resistance to clinical inhibitors through coordinated struc-
Supplementary data to this article can be found online at tural rearrangements, Biochemistry 51 (2012) 2819e2828. https://doi.org/10.
https://doi.org/10.1016/j.molstruc.2019.126893. 1021/bi2018317.
[27] S. De Meyer, E. Lathouwers, I. Dierynck, et al., Characterization of virologic
failure patients on darunavir/ritonavir in treatment-experienced patients,
References AIDS 23 (2009) 1829e1840. https://doi.org/10.1097/QAD.0b013e32832cbcec.
[28] N. Guex, M.C. Peitsch, SWISS-MODEL and the Swiss-Pdb Viewer: an envi-
[1] N.E. Kohl, E.A. Emini, W.A. Schleif, et al., Active human immunodeficiency ronment for comparative protein modeling, Electrophoresis 18 (1997)
virus protease is required for viral infectivity, Proc. Natl. Acad. Sci. 85 (1988) 2714e2723. https://doi.org/10.1002/elps.1150181505.
4686e4690. https://doi.org/10.1073/pnas.85.13.4686. [29] M. Ko  skova
zísek, M. Lepsík, K. Grantz Sa , et al., Thermodynamic and structural
[2] D. Leung, G. Abbenante, D.P. Fairlie, Protease inhibitors: current status and analysis of HIV protease resistance to darunavir - analysis of heavily mutated
future prospects, J. Med. Chem. 43 (2000) 305e341. https://doi.org/10.1021/ patient-derived HIV-1 proteases, FEBS J. 281 (2014) 1834e1847. https://doi.
jm990412m. org/10.1111/febs.12743.
[3] Z. Lv, Y. Chu, Y. Wang, HIV protease inhibitors: a review of molecular selec- [30] D.A. Case, T.A. Darden, T.E. Cheatham III, C.L. Simmerling, J. Wang, R.E. Duke,
tivity and toxicity, HIV AIDS Res. Palliat. Care 7 (2015) 95e104. https://doi.org/ R. Luo, R.C. Walker, W. Zhang, K.M. Merz, B. Roberts, S. Hayik, A. Roitberg,
10.2147/HIV.S79956. G. Seabra, J. Swails, A.W. Go €tz, I. Kolossvary, K.F. Wong, F. Paesani, J. Vanicek,
[4] A.K. Ghosh, H.L. Osswald, G. Prato, Recent progress in the development of HIV- R.M. Wolf, J. Liu, X. Wu, S.R. Brozell, T. Steinbrecher, H. Gohlke, Q. Cai, X. Ye,
1 protease inhibitors for the treatment of HIV/AIDS, J. Med. Chem. 59 (2016) J. Wang, M.-J. Hsieh, G. Cui, D.R. Roe, D.H. Mathews, M.G. Seetin, R. Salomon-
5172e5208. https://doi.org/10.1021/acs.jmedchem.5b01697. Ferrer, C. Sagui, V. Babin, T. Luchko, S. Gusarov, A. Kovalenko, P.A. Kollman,
[5] S. Karthik, S. Senapati, Dynamic flaps in HIV-1 protease adopt unique ordering AMBER 12, University of California, San Francisco, 2012.
at different stages in the catalytic cycle, Prot. Struct. Funct. Bioinform. 79 [31] K. Lindorff-Larsen, S. Piana, K. Palmo, et al., Improved side-chain torsion po-
(2011) 1830e1840. https://doi.org/10.1002/prot.23008. tentials for the Amber ff99SB protein force field, Prot. Struct. Funct. Bioinform.
[6] Viktor Hornak, Asim Okur, Robert C. Rizzo, Carlos Simmerling, HIV-1 protease 78 (2010). NA-NA, https://doi.org/10.1002/prot.22711.
flaps spontaneously close to the correct structure in simulations following [32] J. Wang, R.M. Wolf, J.W. Caldwell, et al., Development and testing of a general
manual placement of an inhibitor into the open state. https://doi.org/10.1021/ amber force field, J. Comput. Chem. 25 (2004) 1157e1174. https://doi.org/10.
JA058211X, 2006. 1002/jcc.20035.
[7] G. Toth, A. Borics, Closing of the flaps of HIV-1 protease induced by substrate [33] A. Pabis, I. Geronimo, D.M. York, P. Paneth, Molecular dynamics simulation of
binding: a model of a flap closing mechanism in retroviral aspartic proteases. nitrobenzene dioxygenase using AMBER force field, J. Chem. Theory Comput.
https://doi.org/10.1021/bi060188k, 2006. 10 (2014) 2246e2254.
[8] H. Heaslet, R. Rosenfeld, M. Giffin, et al., Conformational flexibility in the flap [34] A. Manglik, H. Lin, D.K. Aryal, J.D. McCorvy, D. Dengler, G. Corder, A. Levit,
domains of ligand-free HIV protease, Acta Crystallogr. Sect. D Biol. Crystallogr. R.C. Kling, V. Bernat, H. Hübner, X.-P. Huang, M.F. Sassano, P.M. Gigue re,
63 (2007) 866e875. https://doi.org/10.1107/S0907444907029125. S. Lo € ber, Da Duan, G. Scherrer, B.K. Kobilka, P. Gmeiner, B.L. Roth,
[9] G. To th, A. Borics, Flap opening mechanism of HIV-1 protease, J. Mol. Graph. B.K. Shoichet, Structure-based discovery of opioid analgesics with reduced
Model. 24 (2006) 465e474. https://doi.org/10.1016/J.JMGM.2005.08.008. side effects, Nature 537 (2016) 185e190.
[10] M.D. Altman, E.A. Nalivaika, M. Prabu-Jeyabalan, et al., Computational design [35] R. Ishima, D.I. Freedberg, Y.X. Wang, et al., Flap opening and dimer-interface
and experimental study of tighter binding peptides to an inactivated mutant flexibility in the free and inhibitor-bound HIV protease, and their implica-
of HIV-1 protease, Prot. Struct. Funct. Bioinform. 70 (2008) 678e694. https:// tions for function, Structure 7 (1999) 1047e1055.
doi.org/10.1002/prot.21514. [36] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, et al., Comparison of simple
[11] J.R. Collins, S.K. Burt, J.W. Erickson, Flap opening in HIV-1 protease simulated potential functions for simulating liquid water, J. Chem. Phys. 79 (1983)
by “activated” molecular dynamics, Nat. Struct. Biol. 2 (1995) 334e338. 926e935. https://doi.org/10.1063/1.445869.
[12] J. Vondrasek, C.P. Van Buskirk, A. Wlodawer, Database of three-dimensional [37] J.-P. Ryckaert, G. Ciccotti, H.J. Berendsen, Numerical integration of the carte-
structures of HIV proteinases, Nat. Struct. Biol. 4 (1997) 8. sian equations of motion of a system with constraints: molecular dynamics of
[13] I.T. Weber, J. Agniswamy, HIV-1 Protease: structural perspectives on drug n-alkanes, J. Comput. Phys. 23 (1977) 327e341. https://doi.org/10.1016/0021-
resistance, Viruses 1 (2009) 1110e1136. https://doi.org/10.3390/v1031110. 9991(77)90098-5.
[14] K. Strisovsky, U. Tessmer, J. Langner, et al., Systematic mutational analysis of [38] U. Essmann, L. Perera, M.L. Berkowitz, et al., A smooth particle mesh Ewald
the active-site threonine of HIV-1 proteinase: rethinking the “fireman's grip” method, J. Chem. Phys. 103 (1995) 8577e8593. https://doi.org/10.1063/1.
hypothesis, Protein Sci. 9 (2009) 1631e1641. https://doi.org/10.1110/ps.9.9. 470117.
1631. [39] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol.
[15] D.D. Richman, 2017update of Drug Resistance Mutatons in HIV-1, 2017, Graph. 14 (1996) 33e38. https://doi.org/10.1016/0263-7855(96)00018-5.
pp. 132e141. [40] D.R. Roe, T.E. Cheatham, PTRAJ and CPPTRAJ: software for processing and
[16] A.K. Ghosh, Z.L. Dawson, H. Mitsuya, Darunavir, a conceptually new HIV-1 analysis of molecular dynamics trajectory data, J. Chem. Theory Comput. 9
protease inhibitor for the treatment of drug-resistant HIV, Bioorg. Med. (2013) 3084e3095. https://doi.org/10.1021/ct400341p.
Chem. 15 (2007) 7576e7580. https://doi.org/10.1016/j.bmc.2007.09.010. [41] T. Hou, J. Wang, Y. Li, W. Wang, Assessing the performance of the MM/PBSA
[17] R.D. MacArthur, Darunavir: promising initial results, Lancet 369 (2007) and MM/GBSA methods. 1. The accuracy of binding free energy calculations
1143e1144. https://doi.org/10.1016/S0140-6736(07)60499-1. based on molecular dynamics simulations, J. Chem. Inf. Model. 51 (2011)
[18] K. McKeage, C.M. Perry, S.J. Keam, Darunavir. Drugs 69 (2009) 477e503. 69e82. https://doi.org/10.1021/ci100275a.
https://doi.org/10.2165/00003495-200969040-00007. [42] D.P. Oehme, R.T.C. Brownlee, D.J.D. Wilson, Effect of atomic charge, solvation,
[19] H. Hayashi, N. Takamune, T. Nirasawa, et al., Dimerization of HIV-1 protease entropy, and ligand protonation state on MM-PB(GB)SA binding energies of
occurs through two steps relating to the mechanism of protease dimerization HIV protease, J. Comput. Chem. 33 (2012) 2566e2580. https://doi.org/10.
S. Gupta, S. Senapati / Journal of Molecular Structure 1198 (2019) 126893 11

1002/jcc.23095. in HIV-1 protease toward inhibitors probed by molecular dynamics simula-


[43] H. Tzoupis, G. Leonis, T. Mavromoustakos, M.G. Papadopoulos, A comparative tions and binding free energy predictions, RSC Adv. 6 (2016) 58573e58585.
molecular dynamics, MMePBSA and thermodynamic integration study of https://doi.org/10.1039/C6RA09201B.
saquinavir complexes with wild-type HIV-1 PR and L10I, G48V, L63P, A71V, [46] Z. Liu, X. Huang, L. Hu, et al., Effects of hinge-region natural polymorphisms on
G73S, V82A and I84V single mutants, J. Chem. Theory Comput. 9 (2013) human immunodeficiency virus-type 1 protease structure, dynamics, and
1754e1764. https://doi.org/10.1021/ct301063k. drug pressure evolution, J. Biol. Chem. 291 (2016) 22741e22756. https://doi.
[44] J. Chen, X. Wang, T. Zhu, et al., A comparative insight into amprenavir resis- org/10.1074/jbc.M116.747568.
tance of mutations V32I, G48V, I50V, I54V, and I84V in HIV-1 protease based [47] A.C. Wallace, R.A. Laskowski, J.M. Thornton, LIGPLOT: a program to generate
on thermodynamic integration and MM-PBSA methods, J. Chem. Inf. Model. schematic diagrams of protein-ligand interactions, Protein Eng. 8 (1996)
55 (2015) 1903e1913. https://doi.org/10.1021/acs.jcim.5b00173. 127e134.
[45] J. Chen, Drug resistance mechanisms of three mutations V32I, I47V and V82I

Вам также может понравиться