How Mutations Can Resist Drug Binding Yet Keep HIV-1 Protease

dynamical equilibrium of flaps is perturbed, selectively resist the drug ... inhibitors to WT and HIV-1 protease mutants to understand how mutations c...
0 downloads 0 Views 2MB Size
Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Article

How Mutations Can Resist Drug Binding Yet Keep HIV-1 Protease Functional Rajeswari Appadurai, and Sanjib Senapati Biochemistry, Just Accepted Manuscript • Publication Date (Web): 15 May 2017 Downloaded from http://pubs.acs.org on May 17, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Biochemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

How Mutations Can Resist Drug Binding Yet Keep HIV-1 Protease Functional Rajeswari Appadurai and Sanjib Senapati* Bhupat and Jyoti Mehta School of Biosciences, Department of Biotechnology, Indian Institution of Technology Madras, Chennai 600 036, India. Tel: +91-44-22574122, Fax: +91-44-22574102, E-mail: [email protected]

Keywords: HIV protease, Mutations, Drug resistance, Functional protease variant, MD simulations, Free energy calculations

1 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract HIV-1 protease is an important drug target for AIDS therapy. Nearly ten small molecule drugs have been approved by FDA. However, prolonged usage of these drugs produced protease mutants that are insusceptible to many of these drugs. The mutated proteases however continue to cleave the substrate peptides and thus remain largely functional. This poses a major challenge in the treatment strategies. Thus it has become imperative to understand how these mutations induce drug resistance while maintaining the enzymatic activity of this protein. Here, we carry out a comprehensive study of wild type (WT) and clinically relevant mutated protease bound to a series of FDA approved drugs and substrates of varying sequences to unravel the mechanism of unhindered activity of the drug-resistant protease variants. Our results from large molecular dynamics (MD) simulation data suggest that while the substrate binding to WT and protease mutants involves multiple H-bonding interactions between substrate subsites and protease’s main chain atoms, the drug binds primarily through the hydrophobic interactions with the side chains of protease’s active site and flap residues. This implies that any side chain variations due to mutations in protease could greatly modulate the binding affinity of inhibitors, but not of the substrates. The significantly weaker free energy of binding of the drugs could also be attributed to the limited number of interaction subsites present in the inhibitor structures compared to the substrates. These findings in combination with the identified protease flap and active site residues that contribute in ligand recognition and strong binding can help designing future resistance-evading HIV-1 protease inhibitors.

2 ACS Paragon Plus Environment

Page 2 of 38

Page 3 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

Introduction HIV protease is an effective drug target for anti-retroviral therapy. The enzyme catalyzes the cleavage of gag and gag-pol polyproteins for the maturation of HIV virus. The cleavage begins with the auto-proteolysis of HIV protease itself from the precursor polyproteins. The mature enzyme in its dimeric form subsequently cleaves the polyproteins in an ordered and sequential manner.1 Based on the cleavage sites of the substrates, short peptide-like drug molecules have been developed and successfully used in the treatment of AIDS. However, the long-term usage of these drugs elicits mutant versions of HIV protease that become insusceptible to multiple drugs.2 This poses a major challenge in the treatment strategies. The HIV protease inhibitors inhibit the substrate cleavage in protease active site by competitive binding. The enzyme active site is gated by two flexible beta hairpin loops, known as flaps. The flaps adopt unique ordering at different stages of ligand binding.3 While the flaps exist predominantly in semi-open conformation in apo enzyme, the flaps in ligand-bound protease adopt a closed state with reversed flap handedness. However, both these conformations are not suitable for ligand entry. The flaps need to open wider for the ligand to access the active site.4,5 Subsequently, the enzyme binds to the ligand following a two-step mechanism as shown in Scheme I.6 First a loose collision complex is formed between the open HIV protease and the ligand (E.L). Later, the loosely bound ligand at the active site induces the flap closing and thus produces a tight protease – ligand complex (E.L*). 







 +  ⇌ .  ⇌ . ∗

Scheme. I

Here, k1 and k-1 represent the association and dissociation rate constants for the formation and breaking of the collision complex; whereas, k2 and k-2 represent the equilibrium rate constants for the interconversion of ligand-bound protease from open to closed conformations and vice versa. While the collision of free enzyme and ligand occurs relatively faster, the latter step of ligand-induced conformational transition of the protease to closed state occurs rather slowly, acting as the rate-limiting step.6 3 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Studies by solution NMR spectroscopy suggested that the internal dynamics of flap region in ligand-bound enzyme differs significantly from that in the free protease.5,7,8 While in free protease the flaps display a range of dynamics in the timescale from nanoseconds to micro seconds,7,8 the binding of ligand imposes rigidity to the flaps.5,7 This suppression of flap dynamics in the bound state helps substrate to attain proper orientation in the active site for effective catalysis. Although the NMR and X-ray crystallographic results could show the structural differences between the apo and ligand bound proteases, these techniques could not effectively describe the molecular events associated with the conformational transitions. Computational methods such as molecular dynamics (MD) simulations have provided useful information about protease conformational changes that occur during the ligand binding process.9–11 In these MD simulation studies, to mimic the first step of ligand binding, the ligand was manually placed into the active site of the open conformation of WT HIV-1 protease. During the course of simulations, the bound ligand induced conformational changes to produce the correct closed state as seen in the crystal structures of ligand bound HIV-1 protease. The role of structural water molecule in inducing such transition was also captured, when the simulation was carried out in explicit water.11 Since the flap dynamics is crucial in ligand binding,3,7 any alteration in flap dynamics could result in changes in ligand association/dissociation kinetics. Unsurprisingly, alteration in flap structure/dynamics is a common attribute noted in the multi drug resistant protease variants such as MDR769,12,13 PR20,14 V613 and CA84179.15 These multi drug resistant variants harbor multiple active and non-active sites mutations to resist the drugs profoundly. Interestingly, despite their poor drug susceptibility, these multi drug resistant (MDR) variants ensure the vitality of the virus by performing proper cleavage of the substrates.16–18 It is not clear how these MDR variants, in which the dynamical equilibrium of flaps is perturbed, selectively resist the drug binding without compromising the binding of substrates. Hence, it is of tremendous importance to perceive how the ligand binding differs from substrate to inhibitor and even more critically, how that varies from WT to the mutant proteases.

4 ACS Paragon Plus Environment

Page 4 of 38

Page 5 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

Here we report a comprehensive MD simulation study of the binding of substrates and inhibitors to WT and HIV-1 protease mutants to understand how mutations can resist drug binding yet keep HIV-1 protease functional. We considered the clinically relevant mutations by studying the drug-resistant HIV isolate CA84179 that contains three activesite and nine non-active-site mutations in each monomer (Fig. S1).19,20 As ligands, we have chosen three natural heptapeptide substrates of HIV-1 protease, which possess distinctly different cleavage sites with the side chains of cleaving residues spanning from small aliphatic to bulky aromatic groups. Additionally, three FDA approved drugs have been chosen which show significantly less susceptibility to the mutant isolate under study (Fig. S2, Table S1). These drugs, although mimic the HIV-1 protease substrates, are smaller in length than the substrates. Thus, a total 12 systems were simulated in this study (Table 1). Our results show that the WT protease undergoes a conformational transition from open to closed state upon binding both the substrates and inhibitors, while the mutant protease was unable to close the flaps upon binding the inhibitors. This was primarily due to the differential interactions of protease flap and active site residues with substrates and inhibitors in WT and mutants. While there existed a strong correlation between substrate-flap interactions (substrate recognition) and protease’s flap closing event in both WT and mutant proteases, such a correlation was missing specifically in the inhibitor-mutant protease complexes. Subsequently, we identified a set of crucial interactions between the flap residues and ligand subsites that contributed significantly to the flap closing event in WT protease. The flap residues G48–I50 (G147–I149) and substrate/inhibitor subsites p2, p1′ and p2′ played the critical role in inducing the transitions. While these critical interactions persisted in substrate recognition for the variant protease, majority of the interactions were lost in inhibitor–variant complexes. Thus, the study could successfully explain the long-prevailing question of why mutant HIV-1 protease is efficient enough to hydrolyze the substrate, while it fails to bind the drugs effectively. It is worth mentioning here that the ligands discussed in this manuscript represent the substrates and substrate-mimicking protease inhibitors. However, there could be other protease inhibitors, which may enter the binding site without the need for the flaps to fully open.21

5 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 38

Table 1: List of systems studied. Cleavage sites in the protease substrates are indicated by the arrow. System No. 1 2 3 4 5 6

WT protease complexed with p2-NC heptapeptide (ATIM↓MQR) Ca-p2 heptapeptide (ARVL↓AEA) RT-RH heptapeptide (AETF↓YVD) Saquinavir Indinavir Nelfinavir

System No. 7 8 9 10 11 12

Variant protease complexed with p2-NC heptapeptide (ATIM↓MQR) Ca-p2 heptapeptide (ARVL↓AEA) RT-RH heptapeptide (AETF↓YVD) Saquinavir Indinavir Nelfinavir

Methodology Selection of protease variant To start with, a thorough literature search revealed that the combination of primary mutations, V32I, M46I, V82A and L90M, is one of the most frequently occurring mutational patterns in AIDS patients. Patients with this mutational pattern are reported to exhibit more than 10-fold decreased susceptibility to as many as six FDA approved protease drugs.22 Subsequently, for this pattern, we searched for accompanying secondary mutations that show more than 25% occurrences in AIDS patients and listed in HIV drug resistance database (hivdb.stanford.edu). A clinical isolate, CA84179,20 which fulfills these criteria with secondary mutations, I15V, K20R, M36I, F53L, I54V, L63P, A71I, I93L, is thus chosen for the investigations. The drug resistance data of this isolate is included in Table S1, which shows a fold increase in Ki from 1.4 for tipranavir to 78 for nelfinavir with respect to the WT protease.19,20 System preparation Wild type HIV-1 protease with open flap conformation (PDB ID: 2PC023) was chosen as the starting structure based on the previous reports that ligand can enter into the active site only when protease flaps are widely open.4,5 The structure of the clinical isolate, CA84179 was generated by comparative protein modeling with Swiss-Pdb Viewer24 using WT protease in open conformation as the template. For the ligand, we have chosen 6 ACS Paragon Plus Environment

Page 7 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

three natural hepta-peptide substrates that have varying cleavage site sequences and three least susceptible drugs to CA84179 as listed in Table 1. Following the reported mechanism of complex formation (Scheme I),6 the ligand was placed loosely in the active site pocket of the enzyme. Subsequently, each loosely bound ligand–protease complex was simulated in explicit water for 50 ns to observe the conformational transitions. Details of MD simulations All MD simulations were carried out using AMBER12 package25 with AMBERff99SB26 and the Generalized Amber Force Field (GAFF)27 for protein/substrate and drug molecules, respectively. Hydrogens for the heavy atoms were added using LEAP module of AMBER12. Protonation state of the HIV-1 protease catalytic site was depicted based on a NMR report that suggested one of the two catalytic aspartates of HIV protease will be protonated over the pH range of 2.5–7.0.27 Accordingly, in our simulations at neutral pH, we assigned D25′ as protonated and D25 as deprotonated. Mutated residues and hydrogens were energy minimized by 1000 steps of the steepest descent followed by 1000 steps of the conjugate gradient algorithms. After the added atoms were relaxed, the structures were hydrated in cubic periodic box extending 9Å outside the protein on all sides with explicit water molecules. Three-site TIP3P model was chosen to describe the water molecules.28 For charge neutralization, requisite number of Cl- ions was placed randomly in the simulation box. The number of water molecules and Cl- ions included in each system are tabulated in Table S2. Bonds involving hydrogens were constrained using SHAKE algorithms.29 Particle mesh Ewald30 with 12Å cut off was used to calculate electrostatics. The initial minimization and thermalization were performed to avoid the bad contacts created by mutations and addition of hydrogens and water. All the systems were then equilibrated in NVT ensemble at 300K temperature for about 500 ps and subsequently in NPT ensemble at 1 atmospheric pressure and 300K for 1 ns. A time step of 2fs was used. After the density and potential energy were seen to converge during equilibration, 50 ns of production run was performed for each system. Coordinates were saved every 2ps. Trajectories were visualized and rendered using VMD.31 Trajectory analysis, e.g. root mean squared deviations (RMSD) of protease structures, distance between secondary elements etc. were performed using CPPTRAJ module32 of AMBER. 7 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Calculation of free energy of ligand binding: The molecular mechanics generalized-Born surface area (MM-GBSA) and molecular mechanics Poisson-Boltzmann surface area (MMPBSA) are the two commonly used methods for predicting the binding free energy of ligands to the receptor.33–37 Although the Generalized Born (GB) method is theoretically less rigorous than the Poisson Boltzmann (PB) method, often both the methods perform equally good in predicting the correct binding affinities.36,37 Here, we used the MMGBSA method for calculating the relative binding free energies of ligands to HIV protease in open and closed conformations. For a given complex, the free energy of binding is calculated as,  =  −   ! !   +  +  + "#  =    where,  and  are the electrostatic and van der Waals contributions, ! ! respectively, and  and "# are the polar and nonpolar solvation terms,

respectively. While the non-polar energy is estimated by solvent accessible surface area (SASA), the polar contribution is calculated using GB model with an external dielectric constant of 80 and internal dielectric constant of 4. The conformational entropy was computed by normal-mode analysis of the harmonic vibrational frequencies of the open and closed state of protease, using the Amber nmode module.38 The protein conformations were subjected to minimization for 10000 steps with a distance-dependent dielectric constant of 4 and the convergence threshold for the energy gradient of 104 kcal/mol/Å.

Results and Discussion Ligand binding in HIV protease is reported to follow a two-step process (Scheme I6). In the first step, the ligand diffuses into protease active site through its open flap conformation that produces a loose collision complex. In the second step, the flaps in the loosely bound complex undergo significant conformational changes and produce the tight protease-ligand complex. In our study, to mimic the initial loosely bound collision complex, we placed the ligand into the open conformation of HIV-1 proteases. It is worth 8 ACS Paragon Plus Environment

Page 8 of 38

Page 9 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

mentioning here that, although the semiopen conformation is prevalent in free HIV protease, it does not permit ligand to access the active site. The protease needs to open widely for ligand entry. Hence the apo protein with open flap conformation23 is chosen as the starting structure. Ligand structures were obtained from the corresponding ligand bound WT HIV-1 protease crystal structures and were placed in open conformation by superposing all protease residues other than those in the flaps (K45–K55 from each monomer). The loosely bound enzyme-ligand complexes were simulated for 50 ns each in explicit water. The convergence of the simulation data was assessed by calculating the RMS average correlation (RAC), an autocorrelation function of the protein RMSD values.39 For a given time interval, RAC measures the average RMSD of all running averaged structures over that time interval. Fig. S3 shows the decay in the RMS average correlation as a function of increasing time interval for all the simulations of proteaseligand complexes. As the time interval increases, RAC reaches a plateau beyond 30 ns of simulation time, suggesting the structural convergence. After the simulations were fully converged, the equilibrium properties of conformational transitions were analyzed over the final 20 ns data and the results are presented below. Ligand-induced flap closing in WT HIV-1 protease reproduced the correct crystal conformation The visual inspection of the simulation trajectories revealed that the binding of ligand to the open conformation of WT HIV-1 protease (system 1-6 in Table 1) readily induces conformational changes that produced closed flap protease conformations, which are very similar to the crystal structures of ligand bound HIV protease complexes. To verify, we superposed and compared the simulation generated closed structures of WT HIV-1 protease with the corresponding closed crystal conformation. Fig. 1 shows the timeaveraged structures of p2-NC- and saquinavir-bound WT HIV protease (generated over last 20 ns simulation data), superposed onto the corresponding crystal conformations. Similar comparisons for other substrates and inhibitors are shown in Fig. S4. Both Fig. 1 and Fig. S4 clearly show that the simulation-generated closed structures of WT proteaseligand complexes match very well with the X-ray structures. Moreover, as we have shown in a previous study,11 a key water molecule that appears spontaneously from the 9 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

solvent during the simulations and bridges the flaps with ligand in closed conformation is also reproduced precisely. Notably, this water molecule mediating the flap–ligand interaction is present in all protease–ligand closed crystal structures. Deviations in closed protease structure from crystal conformations are calculated in terms of RMSD of the protein backbone Cα atoms and found to be less than 1.3Å in all six complexes as listed in Table 2. These results clearly indicate that WT HIV-1 protease attains correct closed state upon binding to both substrates and inhibitors.

Figure 1

10 ACS Paragon Plus Environment

Page 10 of 38

Page 11 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

Table 2: Time-averaged RMSD values of the simulation generated conformations of the ligand-bound protease complexes with respect to the corresponding closed crystal structures. RMSDs are calculated with respect to the main chain atoms of the protein residues and expressed in Å. Indicated error bars are the standard deviations obtained from the final 20 ns simulation data. Ligand

WT

Variant

∆RMSD (Variant– WT)

p2-NC

1.17 ± 0.10

1.34 ± 0.11

0.17

CA-p2

1.04 ± 0.10

1.24 ± 0.10

0.20

RT-RH

1.18 ± 0.12

1.33 ± 0.09

0.15

Saquinavir

1.11 ± 0.12

1.97 ± 0.16

0.86

Indinavir

1.20 ± 0.13

1.70 ± 0.10

0.50

Nelfinavir

1.29 ± 0.11

1.86 ± 0.11

0.57

Substrates could induce flap closing in protease variant too, but inhibitors failed We performed similar structural analysis on the substrate and inhibitor bound protease variant, CA84179 and the results are shown in Table 2. As this table indicates, simulation generated closed structures of variant–ligand complexes exhibited larger RMSD compared to the WT–ligand complexes. However, the most noteworthy feature of Table 2 is that, drug bound variant protease exhibited significantly higher RMSD compared to the substrate bound variant protease (e.g. 1.97Å for saquinavir complex versus 1.34Å for p2-NC complex). We further noted that the major contribution to the large RMS deviations in drug–variant complexes was due to the flexible flaps that deviated significantly more compared to the other protein regions, as shown in Fig. S5. Hence in the subsequent analysis, we devote a special emphasis on flap dynamics and its role in protease open to closed state transition. To start with, we compare the time evolution of flap RMSDs in WT and variant protease complexes. Flap RMSD was calculated by superposing all protein residues onto the reference closed crystal structure, leaving out residues K45 – K55 that constitute the flap in each monomer. Thus, Fig. S6 shows the evolution of flaps from open to closed state in 11 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

substrate and inhibitor bound WT protease. As the figure indicates, the RMSD of both flaps, flap A and flap B was deviated about 5Å in the starting open structure. During the course of simulations, however, this large RMSD gradually decreases to a very small value suggesting the appearance of closed conformation, in all WT protease–ligand complexes, as shown in Figs. S6a-c, S6g-i. This complete process of flap closing took place within 2–10 ns simulation time. Once the closed conformation is attained, the RMSD remained stable for the rest of the simulation time. Thus the binding of substrate or the drugs to the WT protease shifts the flaps from open to stable closed state, in consistent with the earlier reports.9–11 Very similar to the WT protease, the CA84179 variant also attained the correct closed conformation when bound to the substrates. As shown in Figs. S6d-e, the flap RMSDs in p2-NC and CA-p2 bound complexes decreased rapidly, representing the transition of the substrates bound variant protein from open to the closed state within 10 ns. The RT-RH bound variant also attains the correct closed conformation, as shown in Fig. S6f, although it takes little longer. Interestingly, on binding with any of the three inhibitors, the flaps of the variant never reach to the closed state during the entire stretch of 50ns simulation (Figs. S6j-l), although the WT-inhibitor complexes closed down readily (Figs. S6g-i). The time evolution depicted in Fig. S6j for the variant–saquinavir complex shows that the RMSD increased up to 7 Å in the initial stages of simulation and later fluctuated around 3–5 Å during the rest of the simulation time. Similar larger deviation is observed for variant– indinavir and variant–nelfinavir complexes (Figs. S6k-l). These results clearly imply that the protease variant on binding to the inhibitors fails to attain the closed state, and thus can promote faster dissociation of the bound drugs. On the other hand, the substrate bound protease variant shifts the conformational equilibrium towards the canonical closed conformation (Figs. S6d-f), and thus can maintain sufficient catalytic activity. In crystal structures, the open and closed conformations of HIV protease are distinguished by flap tip to catalytic site distances and flap tip to flap tip distances. The flap tip to catalytic site distance, defined between the CA atoms of flap tip residue Ile50 and the catalytic residue Asp25′, is reported to be 18 Å in the X-ray structure of the open 12 ACS Paragon Plus Environment

Page 12 of 38

Page 13 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

state (PDB ID: 2PC0), and 12.5 Å in the closed state (PDB ID: 2AVO). Similarly, the flap tip to tip distance between the Ile50 of each flaps is measured to be 12 Å and 4.6 Å in the open and closed crystal structures, respectively. Hence to further validate the conformational transition during our MD simulations, we measured these distances as a function of time for all the protease-ligand complexes. The calculated distance profiles for WT and variant proteases bound to p2-NC and saquinavir are shown in Fig. 2. The profiles for other protease-substrate and protease-inhibitor complexes are shown in supporting information (Fig. S7 and Fig. S8). As the figures depict, the reduction in flap-tip to catalytic site distance and flap-tip to flap-tip distance (shown in the inset) takes place simultaneously. However, the distances reduced faster for one of the flaps, suggesting an asymmetric flap closing mechanism, in consistence with the reports from earlier MD simulations10,11 and X-ray structural intermediates.40 Notably in all the WT–substrate and WT–inhibitor complexes, both the flap-tip to catalytic site and flap tip-tip distances reach to the reference value of the closed crystal structure (Figs. 2a, 2b, S7a-b, S8a-b). Similar to these, the variant– substrate complexes have also reached to the closed state, although the RT-RH bound complex shows larger fluctuations (Figs. 2c, S7c-d). On contrary, none of the CA84179 variant–drug complexes reached to the reference closed structure value (Figs. 2d, S8c-d). Taken together, the analyses revealed that the flaps in mutant protease fail to close upon drug binding, while they continue to reach to the correct closed state upon binding to the substrates.

13 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 2

Free energy of flap closing correlates very well with the drug resistance data To quantify the relative efficiency of WT versus protease variant, CA84179 in inducing flap closing, we calculated the free energy of binding of the substrates and drugs in both the closed and open conformations and the results are shown in Table 3. The free energy of ligand binding was calculated by Molecular Mechanics and Generalised Born Surface Area (MMGBSA) approach on the initial 1 ns and final 20 ns data that represent the open and closed state of protease, respectively. The conformational entropy was calculated by the normal-mode analysis of the harmonic vibrational frequencies of the open and closed state of protease. The difference in binding energy between the open and closed state ensembles, therefore, would signify the free energy of conformational change of protease due to ligand binding ( $%& ). As a corollary, the difference in free energies of () conformational change between the variant and WT protease complexes ' $%& −

14 ACS Paragon Plus Environment

Page 14 of 38

Page 15 of 38

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biochemistry

* $%& ) would indicate the relative efficiency of ligand in inducing the flap closing in

variant versus WT protein. In WT protease, as shown in Table 3, the binding free energy of all the three substrates to closed state ( $* ) is much larger than the open state * ( &* ) and thus the free energy of conformational transition ( $%& ) is highly

favorable. This means that the binding of substrates to protease could induce the closed state very easily. Although the inhibitor binding in open state of WT protease also could shift the protease to closed state, the free energy of conformational transition is not as favorable as that of the substrates (See Table 3). The implication of this small free energy difference is that any mutations that reduce the binding energy of inhibitor in closed state could shift the conformational equilibrium towards open state, and consequently could result in higher rate of inhibitor dissociation from the protease binding pocket. The free energy values of the conformational transitions for the CA84179 variant () ' $%& ) suggest that, all the three substrates could induce conformational transition

with similar effectiveness as they did in the WT protease (Table 3). The loss in free () energy of conformational change in the variant protease with respect to WT ' $%& − * $%& ) is only 4.67% for p2-NC binding and Indinavir > Saquinavir. Strikingly, this order of free energy loss of flap closing matches very well with the fold resistance order of this isolate to these drugs, as measured by PhenoSense assay20 (shown in last column of Table 3). Thus the results suggest a direct relationship between the loss in free energy of flap closing and drug resistance efficacy of the protease variant. Interestingly, a similar correlation between flap closing event and change in inhibitor IC50 values was observed by Vera et al. in another multidrug resistant variant, MDR769.42 Taken together, our result provides compelling evidence that the higher order drug resistance of the protease mutants is primarily due to the interrupted flap closing of the enzyme. 15 ACS Paragon Plus Environment

Biochemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 38

Table 3: Free energy of flap closing in WT and CA84179 isolate of HIV protease bound with various ligands. ∆G values for open and closed states are calculated from the first 1 ns and last 20 ns frames of the simulation trajectory, respectively. All free energy values are in kcal/mol. Contributions of enthalpy and entropy in the free energy values are listed in Table S3.

Ligand

p2-NC

,-./0

-17.40 ± 1.92 CA-p2 -30.18 ± 2.07# RT-30.92 RH ± 1.75 Saq -15.62 ± 1.28 Ind -27.25 ± 2.39 Nel -18.13 ± 1.66 * indicates values

,-1/0 11.72 ± 2.23 -10.07 ± 2.37# 6.27 ± 2.62 -3.26 ± 1.93 -11.13 ± 2.64 -0.55 ± 2.37 obtained

Relative activity 34 /fold resistance ,,-2%1 /0 of the protease − ,,-2%1 variant (VR) (% loss) -18.02 9.74 -27.76 1.36 0.84 (Ref. 41) ± 2.56 ± 3.25 (4.67%) -28.51 -10.89 -17.62 2.49 1.00 (Ref. 41) ± 1.65# ± 2.16# (12.38%) -21.47 8.83 -30.30 6.89 0.07 (Ref. 41) ± 3.09 ± 2.89 (18.52%) -8.48 -2.28 -6.20 6.16 14 (Ref. 20) ± 2.20* ± 3.72 (49.84%) -19.19 -11.34 -7.85 8.27 49 (Ref. 20) ± 2.89* ± 2.82 (51.30%) -9.22 -2.26 -6.96 11.02 78 (Ref. 20) ± 2.03* ± 3.22 (61.29%) final state that may not correspond to closed conformation. In second

/0 ,,-2%1

-29.12 -20.11 -37.19 -12.36 -16.12 -17.98 with the

,-234

,-134

34 ,,-2%1

last column, the % loss in free energy of conformational change in the variant protease with respect to the 34 /0 WT is defined as (,,-2%1 − ,,-2%1 /6679: .%8 ) ∗ ;