Structural and binding insights into HIV-1 protease and P2-ligand interactions through molecular dynamics simulations, binding free energy and principal component analysis

Abstract

HIV-1 protease (HIV-1-pr) plays an important role in viral replication and maturation, making it one of the most attractive targets for anti-retroviral therapy. To design new effective inhibitors able to combat drug resistance in mutant HIV-1-pr variants, it is essential to gain further understanding about the mechanisms by which the recently proposed inhibitors deactivate the mutant HIV-1-pr variants. In the present work, we explored the interactions between two P2-ligands (DRV, and one new derivative, 4UY) with wild type (WT) and two multiple mutant HIV-1-pr variants (p20 and p51) with all atom molecular dynamics (MD) simulations, binding free energy calculations, and principal component analysis (PCA). The trajectories of MD simulations show that both 4UY and DRV primarily bind with the active sites, flap and 80s loop regions of HIV-1-pr variants through either hydrogen bonds or hydrophobic interactions. More hydrogen bonds and hydrophobic interactions were located for 4UY/HIV-1-pr complexes than for DRV/HIV-1-pr counterparts. More importantly, 4UY was found to have an extra hydrogen bond with the backbone of Gly48′ in the flap region of the HIV-1-prs. The flap tip-tip distance (I50-I50′) and flap tip-active site distance (I50-D25 and I50′-D25′) indicate that the flaps turn more closed in 4UY bound HIV-1-prs than DRV bound ones, and the former also have more compact hydrophobic cavities than the latter. Further, the vector projections from PCA indicate that 4UY/DRV inhibitor binding projects the closing of flap in HIV-1-pr variants. In line with the above trajectory analysis, the thermodynamics calculation with MM-PBSA method suggests much stronger binding affinity for 4UY/HIV-1-pr than DRV/HIV-1-pr by 4.3-6.4 kcal/mol. Although p20 and p51 also induce weaker binding due to multiple mutants for 4UY inhibitor by 1.9-1.8 kcal/mol, their bindings to the new P2 ligand (4UY) are indeed significantly enhanced as compared to DRV. The thermodynamic components responsible for the binding differences and the contribution from key residues to the binding were also discussed in detail.

Graphical Abstract

1. Introduction

Human immunodeficiency virus type 1 protease (HIV-1-pr), a retroviral aspartyl protease, is responsible for the production of viral enzymes and structural proteins critical to HIV viral maturation and infectivity. Thus, HIV-1-pr is an important target for HIV/AIDS therapy.^1,2,3 The protease is a homodimer of two 99 amino acid subunits, and each monomer consists of one α-helix and two antiparallel β-sheets. Catalytic aspartic acid residues from two monomers, Asp25 and Asp25′, meet at the dimer interface and form the catalytic active sites.^4,5 The active sites can be characterized as a channel of eight subsites (S4-S1 and S1′-S4′), each of which corresponds to an amino acid of the substrate (P4-P1 and P1′-P4′ from N to C terminals).⁶ One strategy to combat AIDS is to use protease inhibitors (PIs) to deactivate the actives sites and close the flap region of HIV-1-pr.

The second generation PI, DRV also known as TMC114, is one of the most potent one among nine FDA approved PIs. It has high antiviral activity and high genetic barrier to the generation of resistant viruses.^7,8 It is speculated that the bis-THF oxygen atoms may form hydrogen bonds with the backbone N-H groups of residues Asp30 (Asp30′) of HIV-1-pr, enhancing the binding of the PI with HIV-1-pr.⁹ Thus, one of the major strategies to combat protease drug resistance is to design PIs that are able to promote hydrogen bonds with the backbone atoms in the active sites and/or flap region of HIV-1-prs.¹⁰ However, similar to the previous PIs drug resistance issues still exist on DRV for HIV-1-pr mutant variants, such as the single mutants like D30N and I50V.^11,12 Recently, several HIV-1 patient-derived protease variants bearing multiple mutations like p20 and p51 also show high level of resistance to DRV.^{13,14,15,16,17}

Very recently, a novel P2-ligand 4UY (Figure 1), amino-bis-tetrahydrofuran derivative chemically similar to DRV was suggested to improve the drug resistance of DRV.^16,10 4UY was modified from DRV by introducing methylamine at P2 site C4 of bis-THF moiety and methoxy group replacing amine group in P2′ region. The new P2 ligand exhibits not only hydrogen bonds with S2 subsites (Asp29 and Asp30) but also hydrogen bond interactions with the backbone atoms in the flap region (Gly48/Gly48′) of wild type HIV-1-pr.¹⁶ The stronger interaction may enhance the potency for the new PI to deal with drug resistances. For this end the two multiple mutant HIV-1-pr variants in Figure 2 with highlighted mutated residues were intensively explored using two P2-ligands (DRV and 4UY), ¹⁶ where HIV-1_DRV^R_P20, also known as p20, consists of 11 residue substitutions, and the other HIV-1_DRV^R_P51 (p51) has 14 residue substitutions. 4UY shows active against two DRV-resistant HIV-1 variants (HIV-1_DRV^R_P20 and HIV-1 _DRV^R_P51) that were resulted from using the mixture of eight multidrug resistant HIV-1 variants as a starting viral population.¹⁸.

Figure 1.

(a) The HIV-1-pr structure in 4UY inhibitor-bound state (PDB ID: 5BRY), chains A and B in cyan and magenta colored cartoons, respectively. (b) Molecular structure of P2-ligands amino-bistetrahydrofuran derivative (4UY) and Darunavir (DRV/TMC114) with labelled P1, P2, P1′ and P2′ moieties.

Figure 2.

DRV-resistant HIV-1 variant p20 (HIV-1_DRVR_P20) with highlighted mutated residues.

In the present study, all atom molecular dynamics (MD) simulations were performed to provide a comprehensive understanding about drug resistance mechanism at a molecular level and to obtain detailed information about the binding of two chemically similar P2-ligands DRV and 4UY to wild type (WT) and mutant HIV-1-pr variants p20 and p51. In addition, to quantitatively binding free energies of DRV and 4UY to the WT HIV-1-pr and mutant variants were calculated using molecular mechanics/Poisson-Boltzmann surface area (MM-PBSA) method. The P2-ligands (DRV and 4UY) primarily bind with HIV-1-pr via hydrophobic interaction and hydrogen bonding. Further, principal component analysis (PCA) was performed to probe the difference in conformational changes induced by inhibitor binding. The combination of various computational techniques, such as MD simulation, MM-PBSA, PCA and decomposition analysis of interactions is able to offer new insights into the dynamic behavior, principal interactions, and binding mode of DRV and 4UY to the HIV-1-pr WT and mutant variants. This study can not only reveal the binding mode of P2-ligands against WT and mutant HIV-1-pr variants, but also provide significant dynamical information for designing potent inhibitors inhibiting the WT and mutant HIV-1-prs. Many groups explored the HIV-1 protease drug resistance using MD simulations and binding free energy analysis.^19,20

2. System Preparation and Theoretical Methods

2.1. System Preparation and Molecular Dynamics Simulations

The crystal structures of WT HIV-1-pr complexed with DRV (PDB ID: 1T3R) and 4UY (PDB ID: 5BRY) were obtained from Protein Data Bank (PDB). Two DRV-resistant mutant variants (HIV-1_DRV^R_P20 and HIV-1_DRV^R_P51), the p20 variant with 11 substitutions (L10I/I15V/K20R/L24I/V32I/M36I/M46L/L63P/A71T/V82A/L89M) and p51 with 14 substitutions (L10I/I15V/K20R/L24I/V32I/L33F/M361/M46L/I54M/L63P/K70Q/V82I/I84V/L89M), were manually adapted from the crystal structures using PyMOL.²¹ In total, we have created three apo HIV-1-pr structures, three HIV-1-pr/DRV complexes, and three HIV-1-pr/4UY complexes. Because of the important effect of Asp25/Asp25′ on protease-ligand binding,²² a monoprotonated state was assigned to the oxygen atom OD2 of Asp25′ of the protease.^23,24,25 AMBER14SB force field was assigned for HIV-1-prs.²⁶ The structures of DRV and 4UY were optimized with Gaussian 09 program at HF/6-31G* level.²⁷ Restrained electrostatic potential (RESP) charges and the general Amber force field (GAFF) were then assigned to the optimized structures with antechamber.²⁸ Each system was solvated in a truncated octahedral box of TIP3P water molecules with a 12.0 Å buffer along each dimension.²⁹ An applicable number of Cl⁻ counterions were added to neutralize the systems.

For each system, energy minimization and MD simulation were performed using the GPU-supported pmemd module in AMBER16.^30,31,32,33 Energy minimization was carried out in two steps. Water molecules were firstly minimized for 1000 steps of energy minimization with steepest descent method while keeping force constants over proteins and protein-ligand complexes. In the second stage of minimization, the entire system was energy minimized without positional restraints for 2000 steps with conjugate gradient method. Minimized systems were slowly heated to bring system’s temperature from 0K to 300K in NVT ensemble with time step of 2 fs. Systems were then equilibrated until pressure and density of systems were stabilized in NPT ensemble. Final production MD simulations were performed for 500 ns under NPT conditions. Berendsen thermostat was used in NPT ensemble with target pressure of 1 bar and pressure coupling constant of 2ps. During the MD simulations, all bonds involving hydrogen atoms were constrained using the SHAKE algorithm,³⁴ and a time step of 2 fs was adopted. The Particle mesh Ewald (PME) method was used to calculate all of the long-range electrostatic interactions.³⁵ The cutoff distances for the long-range electrostatic and van der Waals interactions were set to 10.0 Å. For temperature scaling, langevin dynamics was used with a collision frequency of 2 ps.

2.2. Principal Component Analysis (PCA)

PCA is an important tool for probing conformation changes of proteins.^36,37 The cpptraj module was used for performing PCA.³⁸ The collective motions of protein-ligand complexes were studied by using the positional covariance matrix C based on atomic coordinates and its eigenvectors. The elements of the positional covariance matrix C were determined by the equation as following.

$$C_i=<(q_i-<q_i>)(q_j-<q_j>)>(i,j=1,2,\mathrm{\dots .},3N)$$ (1)

where q_i is Cartesian coordinate of the ith C_α atom, and N is the number of C_α atom considered. The average was over the equilibrated trajectory after superimposition on a reference structure to remove overall translations and rotations using a least-square fit procedure. The matrix C is symmetric and can be diagonalized by an orthogonal coordinate transformation matrix T, which transforms the matrix C into a diagonal matrix Λ of eigenvalues λ_i :

$$\Lambda =T^T{C}_{ij}T$$ (2)

where the columns represent the eigenvectors corresponding to the direction of motion of relative to <q_i>, and each eigenvector associated with an eigenvalue that describes the total mean-square fluctuation of the system along the corresponding eigenvector.

2.3. Binding Free Energy Calculations

The binding free energy of protein-ligand complexes were evaluated using MM/PBSA program in AMBER16. For each complex, 200 snapshots were extracted along MD trajectory from the last 100ns MD simulations with an interval of 500ps. The binding energy (ΔG) in condensed phase can be simply defined by the following equations.^39,40

$$\Delta G=\Delta H-T\Delta S$$ (3)

$$\Delta H=\Delta E_{\mathrm{MM}}+\Delta G_{\mathrm{sol}}$$ (4)

Where ΔG is the binding free energy in solution that consists of the molecular mechanics energy in the gas phase (ΔE_MM), the solvation free energy (ΔG_sol) and the conformational entropy effect due to the binding (TΔS)

$$\Delta E_{\mathrm{MM}}=\Delta E_{\mathrm{vdw}}+\Delta E_{\mathrm{ele}}$$ (5)

Where ΔE_vdw and ΔE_ele correspond to the van der Waals and electrostatic interactions in gas phase, respectively.

$$\Delta G_{\mathrm{sol}}=\Delta G_{\mathrm{pb}}+\Delta G_{\mathrm{np}}$$ (6)

Where ΔG_pb and ΔG_np are the polar and non-polar contributions to the solvation free energy, respectively. The ΔG_sol is calculated with the PBSA module, where the dielectric constant is set to 1 inside the solute and 80.0 in the solvent. The nonpolar contribution of the solvation free energy is calculated as a function of the solvent accessible surface area (SASA), as follows:

$$\Delta G_{\mathrm{np}}=\gamma (\text{SASA})+\beta$$ (7)

Where, SASA was estimated using the MSMS program, with a solvent probe radius of 1.4 Å. The values of empirical constants γ and β were set to 0.000542 kcal mol⁻¹ Å⁻² and 0.92 kcal/mol, respectively.

The contribution of entropy (TΔS) to the binding free energy arises from changes in the translational, rotational and vibrational degrees of freedom, and was calculated by using classical statistical thermodynamics and normal mode analysis using the AMBER mmpbsa_py_nabnmode program.

Experimental binding free energies for the WT proteases were calculated from published inhibition constants K_i by the following equation:

$$\Delta G_{\exp }=-RT\ln K_{\mathrm{i}}$$ (8)

The inhibition constants for HIV-1-pr/DRV and HIV-1-pr/4UY were obtained from King et al. and Ghosh et al. respectively.^41,16,42 However, inhibition constants for the mutant variants were not available.

2.4. Residue-Inhibitor Interaction Decomposition

In order to understand the inhibitor-residue interaction in more detail, the interaction energy was further decomposed into the contributions from each residue of the protease. On account of the huge demand of computational resources for PB calculations, the interaction between DRV/4UY and each HIV-1-pr residue was computed using the MM-GBSA decomposition process applied in the mm_pbsa module in AMBER16.

Further, hydrogen bonds formed between HIV-1-pr and the inhibitors were analyzed using cpptraj in AMBER 16.³⁸ A hydrogen bond was identified when the distance between proton donor and acceptor atoms is less than 3.5 Å and the angle of donor-H-acceptor greater than 120°. The occupancy of a hydrogen bond was computed by dividing the number of snapshots showing the hydrogen bond by the total number of snapshots along the MD trajectory. These methodologies were used for HIV-1-pr and inhibitors in previous publications.^42,43

3. Results and Discussion

To discuss structural variation due to the inhibitors, 500ns explicit solvent MD simulations were initially performed for three apo HIV-1-pr variants (WT, p20 and p51), and WT and mutant HIV-1-pr variants complexed with 4UY/DRV were then explored.

3.1. Stability of apo HIV-1-pr and its interaction with 4UY/DRV inhibitor

To obtain information about the stability and evaluate the reliability of MD simulation, the root-mean-square deviation (RMSD) of the backbone atoms relative to the corresponding minimized structure was calculated. Figure 3a shows the RMSD over 500ns MD trajectories for apo proteases. WT has a slightly higher mean value (1.79 Å) than mutant variants, p20 (1.76 Å) and p51 (1.77 Å). The RMSDs of WT and mutant HIV-1-pr variant complexed with 4UY/DRV were plotted in Figures 3b and 3c, respectively. The ligand molecules remain in the active site without significant conformational changes in the HIV-1-prs. The p20 mutant variant binding with DRV shows higher variation and less stable as compared to other complexes, whereas the RMSD of p20/4UY reflects that the binding of p20 with 4UY tends to be stable throughout simulation. Mean RMSD value for p20/4UY does show improvement (1.65 Å) over p20/DRV (1.81Å), and is similar to those of WT/4UY and WT/DRV(~1.61 Å), and slightly better than p51/4UY and p51/DRV (1.67 Å). Overall, the results indicate that the binding of the inhibitors, especially 4UY stabilizes the conformation of the HIV-1-pr.

Figure 3.

Root-mean-square deviation (RMSD) plot for backbone atoms of WT and mutant HIV-1-pr variants and inhibitor complexes relative to their initial minimized structure as a function of time. (a) apo HIV-1-pr variants (b) WT and mutant HIV-1-pr variants bound to 4UY and (c) HIV-1-pr variants bound to DRV.

3.2. Flap flexibility in apo HIV-1-pr and its interaction with 4UY/DRV inhibitor

In order to check the flap flexibility in apo HIV-1-pr variants as well as the influence from 4UY/DRV inhibitors, the root mean square fluctuations (RMSFs) were calculated for individual residues (Figure 4). RMSF of backbone atoms may provide direct insights into the structural fluctuation and flexibility of different regions of HIV-1-pr variants. In line with the fact reflected by the above RMSD, Figure 4 shows that the apo structures have larger RMSF values than 4UY/DRV bound HIV-1-prs. The residues at active sites 24(24′)-29(29′) stay relatively rigid in both apo and inhibitor bound structures, whereas the flap region around 50(50′) in apo structures show high flexibility, but their RMSFs are significantly decreased in the 4UY/DRV bound structures due to the binding with the inhibitors. The 80s loop in chain A around residue 84 shows more rigid in 4UY/DRV bound structures than apo structures. The RMSF height of flap region of 4UY/DRV bound structures decreases by 1.1-1.2Å, indicating that inhibitor interaction (especially 4UY) with amino acid residue Gly48/48′ backbone binding considerably stabilizes the flap region.

Figure 4.

Root-mean-square fluctuation (RMSF) vs residue index for the WT HIV-1-pr and its mutant variants p20 and p51 (a) apo structures (b) HIV-1-pr/4UY complexes and (c) HIV-1-pr/DRV complexes. The residues in active site, flap region, and 80s loop region are highlighted in red, black and blue circles, respectively.

Active Site to Flap Tip Distances.

The inter-residue distances of D25-I50, D25′-I50′, and I50-I50′ have often been used to reflect the vertical and horizontal motion of flap regions. The active site to flap tip distance frequency distribution and time series plots were depicted in Figure 5 and Figure S1, respectively. Figure 5 shows that the distributions are significantly narrower in both 4UY (solid line) and DRV (dot line) bound structures than those of apo (dash line) structures. The binding of the inhibitors with HIV-I-prs compress the hydrophobic cavity. For chain A, the mean distances of the catalytic aspartate-flap tip (D25-I50) for DRV bound structures locate at around 13.8-14.1Å, smaller than those of 4UY bound complexes by 0.3-0.9Å oppositely in chain B the DRV bound structures locate at around 15.1-15.8 Å, much higher than 4UY bound complexes by 0.9-1.5Å. The results indicate that 4UY may bind more strongly to the residues in chain B as compared to apo and DRV bound HIV-1-pr structures.

Figure 5.

Histogram distribution of (a) D25-I50 Cα distance, and (b) D25′-I50′ Cα distance for the WT and mutant HIV-1-pr and its interaction with 4UY and DRV

Flap Tip-Flap Tip Distances.

To explore the relative motion of the flap tips, the distance between the two Cα atoms of I50 and I50′ is calculated. The frequency distribution and time series plot for I50-I50′ distance were shown in Figure 6 and Figure S2, respectively. The frequency distribution plot clearly shows that 4UY bound complexes have only single narrow peak around 6.0Å however, apo and DRV bound structures show two small peaks, one also being around 6.0Å and the other beyond 7.0 Å. The mean values of I50-I50′ for the apo structure are 7.35, 6.10, and 6.20 Å, whereas DRV bound structures show 6.78, 6.10, and 6.36 Å. There is a substantial decrease in distance for 4UY bound structures with mean values of 6.12, 6.10, and 6.20Å for WT, p20 and p51, respectively. The frequency distributions in Figure 6 as well as the mean values for I50-I50′ suggest that the binding of 4UY to HIV-1-prs leads to the flap tips less opened and tighter binding as compared to DRV bound and apo structures.

Figure 6.

Histogram distribution of I50-I50′ Cα distance for the WT and mutant HIV-1-pr and its interaction with 4UY/DRV.

3.3. Analysis of TriCa angle of apo HIV-1-pr and its interaction with 4UY/DRV inhibitor

To explain the flap dynamics behavior of proteins, Scott and Schiffer introduced the flap curling in and out that makes the flap open and close, respectively, in order to access the substrate/inhibitor.⁴⁴ The frequency distribution and time series plots of TriCa angle (Gly48-Gly49-Ile50) were shown in Figure 7 and Figure S3, respectively. Analyzing the TriCa angle frequency distribution plot, it shows that the TriCa angle is narrower in 4UY bound structures compared to apo and DRV bound structures. The TriCa angle mean values are 135.7°, 137.8° and 138.7° for apo WT, p20 and p51, respectively. The TriCa angle 48-49-50 mean values are 135.0°, 137.9° and 141.0° for DRV bound structures, whereas 4UY bound structures shows 137.9°, 139.7° and 141.9° for WT, p20 and p51, respectively. The results indicate the stability of flap tip-flap tip distance in 4UY bound structures compared to apo and DRV bound structures.

Figure 7.

TriCa angle (Gly48-Gly49-Ile50) for the WT and mutant HIV-1-pr variants and its interaction with 4UY/DRV.

3.4. Principal Component Analysis (PCA)

To identify the overall patterns of motion on HIV-1-pr variants interacting with P2-ligands, we have used PCA to see the dominant motions observed during a simulation by visual inspection. The dominant movements defined by the first eigenvector are obtained by performing the PCA on the last 100ns MD trajectory of each system. The porcupine plots were drawn using the projections on principal component PC1 was shown in Figure 8. The arrow in each Cα atom represents the direction of motion, while the length of the arrow characterizes the movement strength. As seen in Figure 8, WT-4UY, p20-4UY, WT-DRV, and p20-DRV complexes show projecting mode towards closing the flap region; however, the p51-4UY and p51-DRV complexes project to opening the flap. In the WT-4UY and WT-DRV complexes, the flap elbows experience the largest movements among the different segments of the protease, as indicated by length of the vectors. Whereas in mutant variants bound to 4UY and DRV, the protease move as a whole in and outward/inward manner, as indicated by the vectors. The results signify that 4UY/DRV inhibitor binding project the closing of flap in HIV-1-pr.

Figure 8.

Collective motions corresponding to PC1 obtained by performing principal component analysis on MD simulation trajectory from the last 100 ns. (a) WT HIV-1-pr/4UY (b) p20 HIV-1-pr/4UY (c) p51 HIV-1-pr/4UY (d) WT HIV-1-pr/DRV (e) p20 HIV-1-pr/DRV (f) p51 HIV-1-pr-DRV complexes.

3.5. Binding Mode Analysis

Hydrogen bonds and hydrophobic contacts were shown in Figure 9 for the final snapshots from MD trajectories of complexes of DRV with WT and mutant HIV-1-prs. WT HIV-1-pr/DRV complex shows hydrogen bonds with Asp25, Asp25′, and Asp30′. Only two hydrogen bonds with Asp25 and Asp30′ exist for p20 and p51 mutant HIV-1-prs with DRV, and Asp25′ does not show clear hydrogen bond with DRV.

Figure 9.

The WT and mutant HIV-1-pr variants binding with DRV. (a) WT HIV-1-pr/DRV (b) p20 HIV-1-pr/DRV, and (c) p51 HIV-1-pr/DRV complexes. Ligplot showing hydrogen bonding and hydrophobic contacts of DRV with HIV-1-pr.

Both 3-dimensional and 2-dimensional plots for the 4UY complexed with WT, p20 and p51 HIV-1-prs were shown in Figures 10-12. There are considerably more hydrogen bonds and hydrophobic contacts than DRV complexes. The MD simulations and binding mode analysis clearly show that 4UY could be more potent against DRV-resistant HIV-1-pr variants. Specifically, according to Figure 10b the binding mode of 4UY with WT HIV-1-pr shows six hydrogen bonding interaction with residues Asp25/Asp25′, Gly27′, Asp29′, Asp30′, and Gly48′, and sixteen hydrophobic contacts with residues Asp25′, Arg8, Leu23, Gly27, Ala28, Asp30, Val32, Ile47, Gly48, Gly49, Ile50, Leu76, Val82, Ile84, Ala28′, and Pro81′, which locate at the active sites, flap and 80s loop regions, respectively. As shown in Figure 10a, the methoxy group in 4UY, which replaces the amine group of DRV at P2′ side, directly interact with the side group of Asp30 through van der Waals interaction, while the newly introduced methylamine as a donor forms a hydrogen bond with the backbone of Gly48′, which supports the binding mode analysis from the experimental studies.¹⁶ Table S2 confirms the hydrogen bond (N-H⋯O) between the methylamine of 4UY with the backbone oxygen of Gly48′ in an occupancy of 67% and an average distance of 2.95Å. In spite of the binding enhancement, the other advantage of the backbone binding lies in the minimal distortion of the protease backbone around the enzyme’s active site.¹⁶

Figure 10.

(a) The WT HIV-1-pr binding with 4UY. The amino acids interactions with 4UY are highlighted. (b) Ligplot showing hydrogen bonding and hydrophobic contacts of 4UY with WT HIV-1-pr.

Figure 12.

(a) The p51 HIV-1-pr binding with 4UY. The amino acids interactions with 4UY are highlighted. (b) Ligplot showing hydrogen bonding and hydrophobic contacts of 4UY with p51 HIV-1-pr.

In p20/4UY complex (Figure 11) the hydrogen bond with Gly27′ is absent, but the other five with Asp25, Ile50, Asp29′, Asp30′, and especially Gly48′ still remain. The detailed hydrophobic contacts, such as Arg8, Leu23, Gly27, Ala28, Asp30, Ile47, Gly48, Gly49, Leu76, Pro81, Ala82, Ile84, Leu23′, Asp25′, Gly27′, Ala28′, and Ile50′, are slightly different from WT-4UY complex. Figure 11a again clearly illustrates that 4UY interacts with the backbone of residue Gly48′ in p20 through a hydrogen bond of N-H⋯O, which has lower occupancy than that in the complex of 4UY and WT HIV-1-pr. In p51/4UY complex (Figure 12) there are only three hydrogen bonds with Asp25, Ile50, and Asp29′, and 14 hydrophobic contacts with Arg8, Gly27, Asp30, Ile47, Gly48, Gly49, Ile82, Leu23′, Asp25′, Gly27′, Ala28′, Gly48′, Ile50′, and Val84′. As compared with WT-4UY and p20-4UY complexes (Figures 10 and 11), in Figure 12 the hydrogen bond of the methylamine group of the P2-ligand with the backbone atoms of residue Gly48′ of p51 is not clearly shown. Hydrogen bond analysis does indicates that N-H⋯O between 4UY and Gly48′ of p51 has rather long distance (3.30Å) and low occupancy (43%).

Figure 11.

(a) The p20 HIV-1-pr binding with 4UY. The amino acids interactions with 4UY are highlighted. (b) Ligplot showing hydrogen bonding and hydrophobic contacts of 4UY with p20 HIV-1-pr.

In view of the important role of hydrogen bond, the hydrogen bond analysis between two P2-ligands (DRV and 4UY) and binding site residues of WT and mutant HIV-1-pr variants was performed. The type, distance, and occupancy of hydrogen bonds for DRV and 4UY were summarized in Tables S1 and S2, respectively. The water that makes bridges between inhibitor and flap tip residues Ile50/Ile50′ was observed and their distance and occupancy of hydrogen bonds were calculated. The amino acid residues Asp25, Asp30′, and Gly27′ show hydrogen bonds with DRV (Table S1). The amino residue Asp30′ show >90% occupancy with DRV. The hydrogen bonds with a high percentage of occupancy indicate that the interactions between amino acid residues and inhibitors are relatively strong. The hydrogen bond interactions between the HIV-1-pr variants and 4UY was shown in Table S2. Apart from binding in the active site region, it is observed that 4UY is shown stable hydrogen bond in the flap region, with the Gly48/Gly48′ of WT and mutant HIV-1-pr variants.

3.6. Binding Free Energy Analysis

The binding free energies for all DRV and 4UY bound complexes performed with the MM-PBSA method were summarized in Table 1 and Figure 13. The binding free energies (ΔG) of complexes WT-DRV, p20-DRV and p51-DRV are −13.77, −12.22, and −10.2 kcal/mol, respectively. Similar to a number of HIV-1-pr single mutants, the two groups of mutants also show binding affinity decrease, 1.55 kcal/mol for p20 and a significant decrease of 3.57 kcal/mol for p51. The weaker binding due to the mutations is responsible to their drug resistance. Consistent with the above binding mode analysis that more hydrogen bonds and more hydrophobic contacts than DRV complexes, the binding free energies of WT-4UY, p20-4UY, and p51-4UY complexes also tend to be −18.41, −16.52, and −16.63 kcal/mol, respectively, much stronger than DRV counterparts by 4-6 kcal/mol. The binding affinities of WT HIV-1-pr to DRV and 4UY are reasonably close to the available experimental values −15.20 and −16.23 kcal/mol.¹⁶ Though the two group mutants also induce weaker binding for inhibitor 4UY, the binding affinities decrease by only 1.89 kcal/mol for p20 and 1.78 kcal/mol for p51. And their binding to 4UY is even much stronger than the binding of WT to DRV, implying that the performance of 4UY may be improved in the aspect of the drug resistance for p20 and p51. A comparison of the free energy components between 4UY and DRV complexes in Table 1 indicates that the binding enhancement for inhibitor 4UY mostly arises from van der Waals energy (ΔE_vdw) for all three complexes in a range of 6-13 kcal/mol. The polar contributions including electrostatic energy and polar solvation energy components(ΔE_ele + ΔG_pol) oppositely weaken the binding, by ~3.4 kcal/mol for WT yet dramatically for p20 and p51 variants by ~5.2, and 9.2 kcal/mol, respectively. The nonpolar terms (ΔG_nonpol) for all of the DRV and 4UY systems are similar within ~−5.7 to −6.2 kcal/mol. As a result (ΛH=ΔE_vdw+ΔG_ele+ΔG_pol+ΔG_nonpol), the binding enthalpies for three 4UY complexes are stronger than those of DRV counterparts by 4.1, 3.1, and 3.7 kcal/mol for WT, p20 and p51 HIV-1-pr, respectively.

Table 1.

Binding free energies of HIV-1 PR-DRV and HIV-1 PR-4UY complexes computed by the MM-PBSA method. All energies are reported in kcal/mol.

Componenta	Wild-DRV	p20-DRV	p51-DRV	Wild-4UY	p20-4UY	p51-4UY
ΔEvdw	−58.11 (0.23)d	−56.32 (0.25)	−54.67 (0.27)	−64.51 (0.24)	−64.09 (0.27)	−67.21 (0.26)
ΔEele	−53.02 (0.35)	−57.34 (0.43)	−55.58 (0.43)	−60.91 (0.42)	−56.98 (0.50)	−40.39 (0.33)
ΔGpol	73.54 (0.29)	77.62 (0.33)	76.85 (0.34)	84.02 (0.35)	82.42 (0.40)	70.89 (0.31)
ΔGnonpol	−5.80 (0.01)	−5.71 (0.01)	−5.75 (0.01)	−6.15 (0.01)	−6.14 (0.01)	−6.17 (0.01)
ΔH	−43.42 (0.27)	−41.74 (0.30)	−39.14 (0.33)	−47.56 (0.35)	−44.79 (0.35)	−42.88 (0.34)
−TΔS	29.65 (0.32)	29.52 (0.38)	28.94 (0.38)	29.15 (0.41)	28.27 (0.44)	26.25 (0.39)
ΔG	−13.77	−12.22	−10.2	−18.41	−16.52	−16.63
Gexp	−15.20b −13.20c	N/A	N/A	−16.23 b	N/A	N/A

^aComponent: ΔE_vdw is the van der Waals free energy; ΔE_ele is the electrostatic free energy; ΔG_pol is the polar solvation energy; ΔG_nonpol is the nonpolar solvation energy; ΔH = ΔE_vdw + ΔE_ele + ΔG_pol + ΔG_nonpol; −TΔS is the entropic contribution.

^bThe experimental value G_exp is obtained from the reference King et al. and Ghosh et al.^41,16 using equation ΔG_exp = RT lnK_i, K_i inhibition constant.

^cThe experimental value G_exp is obtained from the reference Kovalevsky et al.¹¹

^dStandard error of the mean (SEM) values are enclosed in bracket ( ).

Figure 13.

Contributions of the binding free energy components of inhibitor DRV (a) and 4UY (b) to WT and mutant HIV-1-pr variants.

An inhibitor that binds a protein becomes less mobile, and the resulting loss in configurational entropy (−TΔS) opposes the attractive forces (ΔH) that drive binding.⁴⁵ In spite of strong binding enthalpies, the penalties in binding affinity due to entropy loss for inhibitor 4UY are unexpectedly smaller than those for DRV by 0.5-2.7kcal/mol. The results for WT HIV-1 variants matched our expectations and the experimental binding energy values. All three 4UY/HIV-1-pr variants show higher binding free energy than DRV/HIV-1-pr variants.

3.7. Decomposition Analysis of Binding Free Energies

In order to determine the contribution of individual amino acid residues of HIV-1-pr variants to the overall binding affinity of the DRV and 4UY, a binging free energy decomposition analysis was performed using MM-GBSA method. The interaction enthalpies are decomposed into each protein residue and ligand pair, and the decomposition of ΔG_bind values on a per-residue basis has contributions from van der Waals energy, electrostatic interaction, polar and non-polar solvation free energy. The per-residue decomposition method is extremely useful to explain the drug-resistant mechanism at atomic detail and also helpful to reveal the contribution of individual residues in protein-ligand interactions.⁴⁰ The 4UY/HIV-1-pr variants inhibitor-residue interaction spectrum is shown in Figure 14 and energy contributions of each residue are shown in Table 2, whereas DRV/HIV-1-pr variants inhibitor-residue interaction spectrum and energy contributions for each residue are shown in Figure S4 and Table S3, respectively. A lot of residues directly participate in hydrogen bond and hydrophobic interactions. The major contributions with more than −1.0 kcal/mol for 4UY interacting with HIV-1-pr variants mainly come from amino acid residues Asp25/Asp25′, Ala28/Ala28′, Ile47, Gly48′, Asp29′, Gly49/Gly49′, Ile50/Ile50′ and Ile84/Ile84′, whereas the key contributing residues in DRV interacting with HIV-1-pr variants are Asp25, Ala28/Ala28′, Ile50/Ile50′, Ile84/Ile84′, Gly27′, Asp30′ and Ile47′. Comparing the contribution of all residues, 4UY/HIV-1-pr complexes show stronger interactions than DRV/HIV-1-pr complexes. The binding energy of Gly48′ that interacts with the modified methylamine group of 4UY is in the range from −0.5 to – 0.7 kcal/mol. In 4UY/HIV-1-pr complexes, the binding energy of active site residue Asp25 in the range from −4.09 to −4.81 kcal/mol, and flap region residue Ile50 in the range from −2.33 to −2.99 kcal/mol. The active site residues and flap region residues are shown more stable interaction with 4UY than DRV. The 4UY/HIV-1-pr variant 80s loop residue Ile84 in WT variant is shown −1.73 kcal/mol, and p20 variant is shown −1.57 kcal/mol, whereas in p51 the Ile84 is mutated residue Val84 is shown almost similar interaction (−1.15 kcal/mol). The other important residue in 80s loop Val82 in WT, mutated Ala82 in p20, and mutated Ile82 in p51 variants are shown more stable interaction with 4UY than DRV. The mutation in 80s loop region doesn’t decrease the binding affinity of 4UY, whereas DRV binding is decreased significantly.

Figure 14.

Decomposition of ΔG on a per-residue basis for the WT and mutant HIV-1-pr/4UY complexes: (a) WT, (b) p20 and (c) p51 variants

Table 2.

Decomposition of binding energies for the WT HIV-1-pr/4UY on a per-residue basis. All energies are reported in kcal/mol.

	WT						P20						P51
	Tvdw	Tele	Tsol	Tsur	Tot		Tvdw	Tele	Tsol	Tsur	Tot		Tvdw	Tele	Tsol	Tsur	Tot
Asp25	0.94	−13.4	8.46	−0.06	−4.09	Asp25	0.61	−15.84	10.52	−0.10	−4.81	Asp25	−0.28	−10.82	−6.40	−0.06	−4.77
Ala28	−1.67	0.79	−0.49	−0.15	−1.53	Ala28	−1.28	0.66	−0.38	−0.19	−1.19	Ala28	−1.69	0.51	−0.19	−0.22	−1.60
Ile47	−1.29	0.31	−0.18	−0.17	−1.33	Ile47	−0.90	0.25	−0.20	−0.16	−1.02	Ile47	−1.10	0.15	−0.02	−0.15	−1.18
Gly49	−0.80	−1.76	1.60	−0.09	−1.04	Gly49	−0.72	−2.09	1.48	−0.06	−1.39	Gly49	−0.67	−2.21	1.57	−0.08	−1.40
Ile50	−2.03	−0.56	0.44	−0.18	−2.33	Ile50	−2.24	−1.19	0.67	−0.22	−2.99	Ile50	−2.20	−1.08	0.55	−0.25	−2.99
Ile84	−1.63	0.01	0.04	−0.15	−1.73	Ile84	−1.48	0.01	0.06	−0.16	−1.57	Val84	−1.07	−0.17	0.20	−0.11	−1.15
Asp25′	0.44	−4.69	1.88	−0.08	−2.45	Asp25′b	−0.86	−0.91	1.10	−0.07	−0.74	Asp25′b	−1.38	1.05	−0.33	−0.09	−0.77
Ala28′	−2.11	−2.13	1.12	−0.13	−3.26	Ala28′	−2.17	−2.10	1.20	−0.13	−3.20	Ala28′	−2.21	−2.13	0.92	−0.16	−3.58
Asp29′b	−1.25	3.04	−2.60	−0.14	−0.95	Asp29′	−1.44	3.55	−3.20	−0.13	−1.22	Asp29′	−1.28	−0.45	0.85	−0.15	−1.03
Gly48′ b	−0.79	−1.79	2.04	−0.23	−0.77	Gly48′ b	−0.56	−1.50	1.75	−0.18	−0.50	Gly48′ b	−1.14	−1.22	1.84	−0.24	−0.76
Gly49′	−1.29	−1.56	1.79	−0.14	−1.19	Gly49′ b	−0.97	−1.41	1.54	−0.11	−0.95	Gly49′	−1.27	−1.61	1.91	−0.13	−1.10
Ile50′	−2.67	−1.21	1.15	−0.20	−2.93	Ile50′	−2.66	−1.14	1.16	−0.24	−2.89	Ile50′	−2.77	−1.54	1.37	−0.29	−3.21
Ile84′	−1.34	0.09	−0.05	−0.11	−1.41	Ile84′	−1.58	0.15	−0.11	−0.11	−1.66	Val84′	−0.68	0.17	−0.11	−0.05	−0.63

^aEnergies are shown as contributions from van der Waals energy (T_vdw), electrostatic energy (T_ele), polar solvation energy (T_sol), nonpolar solvation energy (T_sur), and sum of them (Tot) of WT, p20 and p51 HIV-1-pr/4UY complexes. Residues of ∣ ΔG ∣ ≥1.0 kcal/mol were listed.

^bResidues with energies that did not meet ∣ ΔG ∣ ≥1.0 kcal/mol.

4. Conclusion

In the present study, we have explored the interactions between HIV-1-prs (WT and mutant variants p20 and p51) and P2-ligand (4UY and DRV) using a combination of all atom MD simulations, binding free energy calculation, decomposition and PCA. From the MD simulation results, the RMSD and RMSF profiles of HIV-1-pr variants interacting P2-ligand (4UY/DRV) obtained proved the stability of 4UY/HIV-1-pr variants over DRV/HIV-1-pr variants during simulation. The binding free energies between the P2-ligands (4UY/DRV) and WT HIV-1-pr are in good agreement with the experimental data, suggesting that the molecular force fields used here are suitable for the description of protease-inhibitor interactions. The binding free energy calculations reveal the high stability of WT and mutant 4UY/HIV-1-pr variants than DRV/HIV-1-variants. The results of PCA proved the closing of flap regions in HIV-1-pr variants on P2-ligand binding. Further analysis of per-residue interaction energies in 4UY/HIV-1-pr variants shows stable hydrogen bond and hydrophobic interactions with active site (Asp25/Asp25′, Ala28/Ala28′, Asp29′), flap region (Ile47, Gly48′, Gly49/Gly49′, Ile50/Ile50′) and 80s loop region (Ile84/Ile84′). The basic amine in 4UY shown flap region backbone binding with Gly48′ makes it gain in binding affinity of the system, which agrees with experimental result. 4UY that has P2-moiety with strong flap backbone binding may confer desirable features in the future design of potent HIV-1-pr inhibitors. This study gives mechanistic insights into the mutational effect on HIV-1-pr/P2-ligand (4UY/DRV) interaction and sheds light on the design of potent inhibitors for the drug resistant HIV-1-pr variants.

1. In General, 4UY has more hydrogen bonds and hydrophobic contacts with HIV-1-prs including both WT and multiple mutants (p20 and p51), bring about smaller hydrophobic cavity than in the case of DRV.

2. 4UY binds stronger to HIV-1-prs by 4.3-6.4 kcal/mol that DRV. It was also found that although p20 and p51 also induce weaker binding due to multiple mutants for 4UY inhibitor by 1.9-1.8 kcal/mol, their bindings to the new P2 ligand (4UY) are indeed significantly enhanced as compared to DRV. This well explain why 4UY has better performance to be against drug resistance.

3. The thermodynamic components responsible for the binding differences and the contribution from key residues to the binding were also discussed in detail, and a few key residues were located.