SEARCH

SEARCH BY CITATION

Keywords:

  • allosteric inhibitor;
  • crystal structure;
  • fragment screen;
  • HIV protease;
  • multidrug resistance

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

We have employed a fragment-based screen against wild-type (NL4-3) HIV protease (PR) using the Active Sight fragment library and X-ray crystallography. The experiments reveal two new binding sites for small molecules. PR was co-crystallized with fragments, or crystals were soaked in fragment solutions, using five crystal forms, and 378 data sets were collected to 2.3–1.3 Å resolution. Fragment binding induces a distinct conformation and specific crystal form of TL-3 inhibited PR during co-crystallization. One fragment, 2-methylcyclohexanol, binds in the ‘exo site’ adjacent to the Gly16Gly17Gln18loop where the amide of Gly17is a specific hydrogen bond donor, and hydrophobic contacts occur with the side chains of Lys14and Leu63. Another fragment, indole-6-carboxylic acid, binds on the ‘outside/top of the flap’ via hydrophobic contacts with Trp42, Pro44, Met46, and Lys55, a hydrogen bond with Val56, and a salt-bridge with Arg57. 2-acetyl-benzothiophene also binds at this site. This study is the first fragment-based crystallographic screen against HIV PR, and the first time that fragments were screened against an inhibitor-bound drug target to search for compounds that both bind to novel sites and stabilize the inhibited conformation of the target.


Abbreviations:
EPR

electron paramagnetic resonance

FDA

Food and Drug Administration

HAART

highly active anti-retroviral therapy

HIV-1

human immunodeficiency virus type 1

IPTG

isopropyl β-d-1-thiogalactopyranoside

KSCN

potassium thiocyanate

MD

molecular dynamics

MDR

multi-drug-resistant

MW

molecular weight

NMR

nuclear magnetic resonance

NRTI

nucleoside RT inhibitor

NNRTI

non-nucleoside RT inhibitor

PEG

polyethylene glycol

PR

protease

RT

reverse transcriptase

SSRL

Stanford Synchrotron Radiation Lightsource

UNAIDS

Joint United Nations Program on HIV/AIDS

WHO

World Health Organization

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

According to UNAIDS and the WHO, approximately 33 million people are currently living with an HIV infection, and 2.7 million people became newly infected with HIV in 2007, of which 370 000 were children under the age of 15.a From 1.8 to 2.3 million people died of AIDS in 2007, and throughout this epidemic, 25 million people have died of HIV-related causes.

When HIV protease inhibitors were combined with other anti-AIDS drugs to form the highly active anti-retroviral therapy (HAART) ‘cocktails,’ the mortality associated with HIV infections drastically decreased. However, a rapidly growing percentage of new HIV infections involve strains that are resistant to at least one of the HAART drugs, and the percentage of these mutant strains that are able to resist multiple different drugs simultaneously have been increasing (1). A primary HIV infection with a drug-resistant strain ensures the dissemination and establishment of cellular reservoirs of that resistant strain, and the resistance pattern persists for years (2,3). Although a temporary interruption in treatment will allow wild-type strains to increase their relative proportion of the viral load in sera, the drug-resistant strain immediately re-emerges and becomes dominant once treatment is resumed (2). Recent estimates indicate that ∼14% of new HIV infections in America are caused by a resistant strain (4). Even with future prevention of HIV infections, for decades to come there will be millions of patients requiring treatment for an HIV infection they already have. Because of the current difficulty of treating MDR (multidrug-resistant) HIV mutants, and the necessity of suppressing the emergence of even more drug-resistant mutants, new drugs with novel mechanisms of action are needed.

Comparison of crystal structures of apo-PR to drug-bound PR (5), biophysical methods including NMR (6,7) and EPR (8–10), and MD simulations (11–15) all demonstrate that mobility of the flaps, which open to admit substrates and close during catalysis, is essential for PR function. Further, MDR mutants often exhibit greater flap conformational variability (5,16–21). For example, conformational change and loss of twofold symmetry increase progressively in a series of 1×, 3×, and 6× mutants versus wild-type PR, where the mutants exhibit increased resistance to the peptidomimetic inhibitor, TL-3 (22). In this case, the 6× residues include mutations commonly observed in clinical isolates of MDR strains.

Molecular Dynamics (MD) simulations of both wild-type HIV-1b PR and the V82F/I84V MDR mutant revealed that the flaps opened farther and displayed substantially more flexibility in the MDR mutant (21). Comparative analyses suggested that a greater preference by MDR mutants for semi-open conformations requires active site inhibitors to overcome a larger enthalpic penalty to bind (21). Consistent with the MD analysis, seven FDA-approved HIV protease inhibitors exhibit reduced enthalpy of binding to another MDR mutant, I50L/A71V (23). The MD simulations further revealed a dynamic relationship within the PR structure for both wild-type and the V82F/I84V mutant (21). Flap motion was anti-correlated with the width of a cleft in the PR fold formed between a loop (residues 38–42) and a β strand (residues 59–63): residues 43–58 form the anti-parallel β hairpin that comprises the flap. The solvent-exposed cleft, termed the ‘exo site’, was compressed when the flaps were open, and it expanded when the flaps were closed. This ‘exo site’ (or ‘allosteric groove’) is composed of the regions that include the elbow, fulcrum, and cantilever sections of HIV PR [i.e. the region between the ‘ear flap’ and the ‘cheek’ (21)]. Additional MD simulations imposing restraints on compression of the exo site in wild-type PR and the V82F/I84V mutant (11), and coarse-grained simulations of much longer durations (12), supported the effect of exo site restraints on suppressing flap mobility. These observations suggested that non-active site binding pockets could exist in HIV PR, dependent on protein conformation, and in particular, that small molecules binding in the exo site could inhibit PR by suppressing the motion of the flaps (21).

Fragment-based lead discovery is a high-throughput, empirical screening strategy (24–28) not yet applied to HIV PR. The approach employs a library of chemically diverse molecules (MW ∼150), which because of their small size, seek out specific hydrogen bonds while limiting non-specific hydrophobic interactions (29–31). Fragment hits can be developed into larger molecules by exploiting adjacent interactions in the protein binding site (32–35). The fragment-based approach has led to the development of high-affinity lead molecules against a number of target proteins (27,28). We have carried out a fragment screen against HIV PR; the results identify two binding sites outside the active site of the PR dimer in its inhibitor-bound, closed conformation.

Methods and Materials

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

Expression and purification

Cloning and expression of wild-type NL4-3 HIV-1 PR has been described previously (36) The HIV-1 PR construct, containing Q7K mutation for decreased autolysis, was cloned into pET-21a(+) vector, and the recombinant plasmid was transformed into the Rosetta 2(DE3)pLysS (Novagen, Inc.) strain of E. coli for protein expression. When the cell culture reached an OD600 of 0.8 for 5 h at 37 °C, PR expression was induced with 1 mm isopropyl β-d-1-thiogalactopyranoside. Inclusion bodies containing PR were isolated by centrifugation at 10,000 × g for 30 min and further purified by several washings with deionized water. Insoluble material was removed by centrifugation followed by filtration through 0.45-μm membrane. The purified inclusion bodies were solubilized in 8 m urea containing 20 mm Tris, 5 mm EDTA, pH 8.0, 5 mm DTT, and subsequently purified by ion-exchange chromatography as described previously (5). The PR solution was mixed with Whatman DE52 anion exchanger and incubated for 1 h and filtered through 0.45-μm membrane. The filtered solution containing PR was then applied onto RQ column that was equilibrated with 8 m urea containing 20 mm Tris, 5 mm EDTA, pH 8.0. The column flow through PR was dialyzed and refolded in 25 mm phosphate buffer pH 7.2, containing 25 mm NaCl and 0.1% 2-mercaptoethanol, overnight followed by dialysis against 50 mm sodium acetate buffer, pH 4.5, 0.1% 2-mercaptoethanol, for 2 h. The refolded PR was clarified by centrifugation and filtration through 0.45-μm membrane. The sample was then concentrated to 2–3 mg/mL and buffer exchanged for crystallization using Amicon Ultracel-10K. The purified PR was separated using SDS–PAGE and verified by immunoblot using a specific antibody against HIV-1 PR. The TL-3 complex was prepared by 1:10 dilution of saturated TL-3 in DMSO into 2–3 mg/mL PR in 50 mm sodium acetate, pH 4.5, followed by incubation on ice for 30 min and centrifugation to remove unbound, insoluble TL-3.

Crystallization and fragment addition

Crystallization was performed by vapor diffusion in 24-well plates at 4 °C using the reservoir compositions and drop ratios given in Table 1. For crystallization of apo-PR or PR:TL-3, a protein solution at 2–3 mg/mL in 50 mm sodium acetate, pH 4.5, was diluted 9:10 with DMSO, except for the P41 form which was crystallized without DMSO. For fragment soaking experiments, crystals were transferred to 4-μL drops of a cryoprotectant consisting of the reservoir solution with 25% glycerol and containing 10% DMSO with the fragment concentrations given in Table 2. Following soaking, crystals were mounted and frozen at 100 K in liquid N2. For co-crystallization in the presence of fragments, aliquots of apo-PR (C2221 form) or PR:TL-3 (P6122, P21212, P212121 forms) at 2–3 mg/mL in 50 mm sodium acetate, pH 4.5, were diluted 9:10 with fragment solutions in DMSO, centrifuged, and used to make drops (Table 2). For screening the fragment library, ninety-six 9-μl protein aliquots were combined with ninety-six 1-μL aliquots of mixtures of four fragments at 50 mm each prepared from 200 mm stock solutions in DMSO. Diffraction quality crystals were obtained in ∼60% of drops. Co-crystals were mounted by diluting the drop with an equal volume of cryoprotectant consisting of the reservoir solution at 10% DMSO and containing 30% glycerol, and frozen at 100 K in liquid N2.

Table 1.   HIV PR crystal forms for fragment screening
Space groupMorphologyNL4-3 wild-type PRCrystallization conditionsDrop ratioaUnit cell ÅResolution ÅAsymmetric unitFlaps
  1. aProtein:reservoir volume ratio used to mix crystallization drops.

  2. bResidues 48–52 (GGIGG) at the tip of each flap disordered.

  3. cAcetate bound in the exo site; TL-3 bound form used for 20 mm 4D9 soaks (Table 3).

  4. dDMSO bound in the exo site (Table 3).

  5. eFragment 4D9 bound in the exo site (Table 3).

  6. fFragment 1F1 or 2F4 bound to the outside/top of the flap (Table 3).

P41OctahedralApo20% w/v PEG 8K 0.3 m MgCl2 0.1 m Tris-HCl, pH 8.51:149.38  49.38  106.022.1DimerOpen, lattice contact
C2221Rectangular rodsApo0.5 m KSCN 0.1 m MES-HCl, pH 5.8  10% DMSO1:149.09 68.53 58.112.1MonomerOpen, disorderedb
P21212cSquare platesApo/+TL-31.0 m KSCN 0.1 m bis-Tris-HCl, pH 7.0 10% DMSO4:158.29 86.26 46.291.6DimerClosed
P6122dHexagonal rods+TL-30.5 m KSCN 0.1 m MES-HCl, pH 5.8 10% DMSO1:162.39 62.39 82.061.8MonomerClosed
P21212eSquare plates+TL-30.5 m KSCN 0.1 m MES-HCl, pH 5.8 10% DMSO1:159.52 85.91 46.441.6DimerClosed
P212121fRectangular blocks+TL-30.5 m KSCN 0.1 m MES-HCl, pH 5.8 10% DMSO1:128.72 65.57 92.361.3DimerClosed
Table 2.   Fragment screening data collection
Crystal formFragment(s) (conc.)MethodCrystals screenedData setsaCoverageb (%)Comment
  1. aSpecific data sets listed in Table 3 are identified in parentheses.

  2. bIncludes redundant data sets for fragment mixtures or individual fragments in the Active Sight library of 384 small molecules, i.e. approximately 3.3 fragments screened per data set.

  3. cOne hit was identified: 5-nitroindole bound to the ‘eye site’ of a semi-open conformation of PR (see Figure S1).

P41Mixtures of 4 2.5 mm10-min soak821818.8>2.5 Åc
P41Individual 10 mm10-min soak11111.3>2.5 Å
C2221Individual 10 mmCo-crystal97466.8No hits
P21212Mixtures of 4 2.5 mm10-min soak7663 (e.g. 1B8)46.9No hits
P21212Mixtures of 4 2.5 mm4-hr soak1265546.9No hits
P6122Individual 10 mmCo-crystal23111.6No hits
P6122Mixtures of 4 2.5 mmCo-crystal365161 (e.g. E5)88.5No hits
P21212D9mix 2.5 mmCo-crystal52 (H2, H6)1.04D9
P212121D9, 2D9, 4D9 3.3 mmCo-crystal41 (K3)0.84D9
P212124D9 20 mm10-min soak32 (E1, E2)0.34D9
P61224D9 20 mm10-min soak32 (D4)0.3Not 4D9
P212121F1mix 2.5 mmCo-crystal52 (J1, K5L)1.01F1
P2121211F1 10 mmCo-crystal52 (K8, L2)0.31F1
P212121F4mix 2.5 mmCo-crystal32 (F1, F2)1.02F4
All  80837897.94 hits

Crystallography and structure analysis

Data sets for the P41 crystal form (Table 2) were collected at 100 K at Active Sight on a Saturn 92 CCD system mounted on a FR-E X-ray generator equipped with VariMax optics (λ 1.54 Å) and processed with d*TREK (37). All other data sets were collected at 100 K at SSRL protein crystallography beam lines using the Stanford Automated Mounting system (38) and blu-ice hutch control software (39) via the Internet (40). Data sets were processed with mosflm (41) and CCP4 programs (42). Structures were solved by molecular replacement using HIV-1 NL4-3 PR structures in the corresponding open (2PC0) or closed (2AZ8) conformations and refined with refmac (43). Electron density map interpretation and model building were performed with xtalview (44) implemented as mifit (45). For data sets in the P21212 crystal form (Table 2), an automated script developed at SSRL for the Joint Center for Structural Genomics was used to process data collection frames to partially refined protein co-ordinates (46). Difference map interpretation of these data sets was assisted by use of mapdock (47). Data sets in the P6122 and P212121 crystal forms (Table 2) were processed and scaled individually, and refined using a ccp4 program script implemented in version 8 of mifit (45) (Tables 3 and S1). Protein structures and fragment binding sites were compared and analyzed using mifit, pmv (48), pymol (49), and autoligand (35), molecular graphics programs.

Table 3.   Fragment screening results
ProteaseExperimentData setCrystal formaResolution ÅR/RfreeBinding siteConformationPDB code
Exo site
Monomer AMonomer BMonomer A
Gly17 NHLeu63 NHGly17 NHLeu63 NH 
  1. aDefined by space group (Table 1).

  2. bMixture of four fragments at 2.5 mm each (Table 2).

  3. cAcetate bound in all P21212 form crystals soaked with fragment mixtures at 2.5 mm each (Table 2).

  4. dFor soaking TL-3 bound PR in this crystal form, crystals were grown with conditions using 4:1 drops (Table 1).

  5. eDMSO or H2O bound in all co-crystallization experiments with fragment mixtures at 2.5 mm that yielded the P6122 crystal form (Table 2).

apo-PRD9mixb soak (2.5 mm)1B8P212121.540.181/0.213AcetatecH2OH2OH2OLeu63 ‘down’3FK1
PR:TL-3D9mix co-crystal (2.5 mm)H2P212121.800.206/0.2464D9H2OH2OH2OLeu63 ‘up’3KF0
PR:TL-3D9mix co-crystal (2.5 mm)H6P212122.000.188/0.2324D9H2OH2OH2OLeu63 ‘up’ 
PR:TL-34D9 co-crystal (3.3 mm)K3P212121.600.191/0.2394D9H2OH2OH2OLeu63 ‘up’ 
PR:TL-34D9 soakd (20 mm)E1P212121.770.214/0.2744D9H2ODMSOH2OLeu63 ‘up’3KFN
PR:TL-34D9 soak (20 mm)E2P212122.200.223/0.3104D9H2ODMSOH2OLeu63 ‘up’ 
PR:TL-3B8mix co-crystal (2.5 mm)E5P61222.000.228/0.299DMSOeDMSO  Leu63 ‘down’ 
PR:TL-34D9 soak (20 mm)D4P61221.770.226/0.313DMSODMSO  Leu63 ‘down’3KFP
      Outside/top of flap  
      Monomer AMonomer BMonomer B 
PR:TL-3F1mix co-crystal (2.5 mm)J1P2121211.300.183/0.207Arg57–Trp421F1Arg57–1F13KFR
PR:TL-3F1mix co-crystal (2.5 mm)K5LP2121211.300.189/0.227Arg57–Trp421F1Arg57–1F1 
PR:TL-31F1 co-crystal (10 mm)K8P2121212.000.217/0.264Arg57–Trp421F1Arg57–1F1 
PR:TL-31F1 co-crystal (10 mm)L2P2121211.770.185/0.235Arg57–Trp421F1Arg57–1F1 
PR:TL-3F4mix co-crystal (2.5 mm)F1P2121211.900.206/0.275Arg57–Trp422F4Arg57–Glu35 
PR:TL-3F4mix co-crystal (2.5 mm)F2P2121211.800.203/0.250Arg57–Trp422F4Arg57–Glu353KFS

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

Crystallography can detect low-affinity interactions: the electron density not only identifies the ligand but also reveals details of the surrounding protein structure. HIV PR is readily crystallized, and many high-resolution structures are known. At the same time, the application of robotics at synchrotron beam lines for high-throughput structural genomics (38–40) enables an empirical approach based on crystal structure determination. The HIV PR screen employed Active Sight fragment library 1 comprised of 384 commercially available, structurally diverse small molecules soluble at 200 mm in DMSO, with average molecular weight 142 Da (95% in the range of 90–220 Da) (45). The majority of these fragments consist of a single rigid core decorated with one to three short side groups; consequently, they have limited torsional degrees of freedom. Scaffolds include benzenes, pyridines, pyrimidines, indoles, indazoles, benzimidazoles, purines, pyrazoles, pteridines, cyclohexanes, napthalenes, quinolines, isoquinolines, quinazolines, piperidines, isoxazoles, oxazoles, thiazoles, and furans (45). The library is designed to be assembled into 96 mixtures, in which the four fragments per mixture are maximally shape diverse.

Crystal forms

For fragment screening experiments we have utilized five crystal forms of HIV PR with inhibitor-bound (TL-3) or without an active site ligand (apo) (Table 1). A tetragonal crystal form in space group P41 was grown using PEG 8K and MgCl2; this crystal form has a PR dimer in the asymmetric unit and the flaps in an open conformation via lattice contacts (6). However, because of the sensitivity of this crystal form to soaking in DMSO (the solvent used for delivery of the fragments), alternative crystallization conditions including 10% DMSO were screened. When the crystals were grown without the presence of DMSO, the DMSO gradient caused by the addition of the dissolved fragments often degraded the quality of the crystals. When similar DMSO concentrations were used for both growing the crystals and for delivering the fragments, the lack of this DMSO gradient improved the production of crystals that diffracted well. New conditions were also sought that avoided PEGs as precipitating agents, as an informal survey of the Protein Data Bank showed that >300 protein structures have ordered segments of polyethylene glycol bound in surface clefts and cavities. In addition, electron density consistent with repeating ethylene glycol units was observed in maps for several of the P41 form crystals (Table 1). Conditions were found for apo-PR using potassium thiocyanate as the precipitating agent; this crystal form has a dimer on a crystallographic twofold axis in space group C2221. Variation of the protein and KSCN concentrations, and pH, yielded a better diffracting crystal form in space group P21212 for both apo-PR, and PR bound to the C2 symmetric, nanomolar affinity inhibitor, TL-3 (23). Conditions for the C2221 and P21212 forms were further refined using TL-3 bound PR, yielding the P6122 crystal form. Subsequently, the P21212 form, and another form in space group P212121, were obtained using conditions for the P6122 form, but only when particular fragments bound (Table 1). Thus, the change in crystal form from P6122 to either P21212 or P212121 can be a macroscopic signature indicating the possibility that an allosteric fragment bound to the PR drug target.

Data collection

We screened 808 crystals and collected 378 data sets using co-crystallization, soaking with fragment mixtures, or soaking with individual fragments (Table 2). The first set of experiments employed the P41 crystal form. Analysis of data sets collected at Active Sight revealed one hit, 5-nitroindole, bound to the inside of the flap via hydrophobic contacts with Val32, Ile47, Ile54, Ile84, and Pro81, and via hydrogen bonds between amides at the tip of the flap (Ile50Gly51Gly52) and the nitro group. This crystallographic fragment site overlaps with the predicted binding mode of a chemically similar molecule, designated ‘Damm compound 1′, discovered in virtual screens (50), and it displays similar interactions with the ‘eye site’ of the semi-open form of wild-type PR (5,13) (Figure S1). ‘Damm compound 2,’ the para-methoxy derivative of the (auto-fluorescent) ‘Damm compound 1,’ displayed an IC50 of 18 ± 3 μm in the standard FRET-based HIV PR activity assay (50). Unfortunately, most crystals of the P41 form lost diffraction quality when soaked in 10% DMSO, and only 29 diffracted to 2.5 Å resolution or better. Subsequent experiments employed crystal forms grown in 10% DMSO (Table 1), which were used for data sets collected at SSRL beam lines (Table 2) (38–40).

Individual fragment soaks at 10 mm in the C2221 crystal form yielded no hits (Table 2). Screening then employed the higher resolution P21212 crystal form in soaking experiments, but no hits were found in 118 data sets collected from 202 crystals. Experiments then focused on co-crystallization under conditions for the P6122 crystal form with individual fragments at 10 mm, and with fragment mixtures at 2.5 mm, but no fragment sites were observed. However, two other crystal forms were obtained. In one case, the P21212 crystal form grew in the presence of mixture D9 because of the binding of the fragment 4D9; in the other case, the P212121 crystal form grew in the presence for mixtures F1 and F4 because of the binding of fragments 1F1 and 2F4, respectively.

Additional data sets were collected to confirm the hits by co-crystallization or soaking.

Exo site

HIV PR structures were refined for all data sets, to examine the difference Fourier maps for possible fragment binding. Table 3 summarizes refined structures for fragment hits and three representative structures as controls. In the structure from the 1B8 data set, representative of many soaking experiments in the P21212 form, the exo site is occupied with acetate from the buffer and H2O molecules (Figures 1A and S2). In the structure for the E5 data set, representative of many co-crystallization experiments in the P6122 form, the exo site is occupied by DMSO and H2O molecules (Figure 1B). In these structures, the amides of Gly16, Gly17, Leu63, and Gly68 are hydrogen bond donors to acetate, DMSO, or H2O, while aliphatic atoms of Lys14 and Leu63 provide hydrophobic contacts.

image

Figure 1.  Unbiased difference electron density for small molecule binding in the exo site of HIV PR contoured at 2, 3, 4, 5 σ. The amide of Gly17 acts a hydrogen bond donor to acetate, DMSO, or 4D9 (2-methylcycolhexanol); the amides of Gly16, Leu63, and Gly68 act as hydrogen bond donors to DMSO or H2O. (A) Acetate ion and H2O molecules in the structure derived from the 1B8 data set (Table 3) where the P21212 crystal form was soaked with fragment mixture D9 containing 4D9 at 2.5 mm; at this concentration, 4D9 does not displace acetate. (B) DMSO and H2O in the structure derived from the E5 data set where the P6122 crystal form was co-crystallized in the presence of mixture B8 containing fragments at 2.5 mm. In this space group, the side chain of Leu63 is restricted from changing conformation because of the contact with Ile72; no fragment binding hits were observed in the P6122 form for 174 data sets following co-crystallization or soaking. (C) 4D9 in the structure derived from the H2 data set where the P21212 crystal form was co-crystallized in the presence of fragment mixture D9 containing 4D9 at 2.5 mm; the side chain of Leu63 flips up to accommodate binding of the fragment, and is unrestricted because of the conformer of Ile72 in this space group. (D) 4D9 in the structure derived from the E1 data set where the P21212 crystal form was soaked with 4D9 at 20 mm; at this concentration acetate is displaced while the crystal lattice allows the side chain of Leu63 to flip up to accommodate fragment binding. In addition to packing with Leu63 and accepting a hydrogen bond from Gly17, 4D9 packs with the side chain of Lys14.

Download figure to PowerPoint

For the H2 data set, the P21212 form grew during co-crystallization with the fragment mixture D9, containing 4D9 (2-methylcyclohexanol) at 2.5 mm, under conditions that otherwise yield the P6122 crystal from. Thus, binding of 4D9 shifted PR from the P6122 crystal from to the P21212 form. In this structure, the fragment is bound in the exo site with a hydrogen bond from the Gly17 amide to the 4D9 hydroxyl group and with hydrophobic contacts between the methylcyclohexyl group and Lys14 and Leu63 (Figure 1C). At 1.8 Å resolution, the cyclohexanol ring best fits the electron density in the boat conformation (Figure S2) with the 2-methyl group of the trans stereoisomer in contact with Leu63. Binding of 4D9 occurs with rotations of both the χ1 and χ2 torsions of Leu63, flipping the side chain ‘up’, when compared to the ‘down’ conformer that occurs when acetate or DMSO is bound (Figures 1C–D versus 1A–B). Binding of 4D9 by co-crystallization into the P21212 form was confirmed in replicate experiments (H6 and K3 data sets). In each structure, 4D9 is bound in the exo site of Monomer A, while the Monomer B exo site contains H2O molecules (Table 3).

Subsequently, crystals of the P21212 form were soaked with 20 mm 4D9, confirming binding in the Monomer A exo site (E1, E2 data sets, Table 3; Figures 1D and S2). Hence, soaking at high concentration can displace acetate from the exo site, and even at 20 mm, 4D9 only binds to the exo site of the inhibitor-bound, closed PR. However, the complementary experiment, in which the P6122 form crystals were soaked in 20 mm 4D9, did not result in fragment binding, and only DMSO was observed in the exo site (D4 data set, Table 3; Figure S2). This result is consistent with packing interactions in the P21212 versus P6122 crystal forms. In the PR, monomer Leu63 is in contact with Ile72 on the adjacent β strand. In the P21212 form, Ile72 of subunit A faces solvent (except for a 4.2 Å contact with Gly40 by symmetry), and its orientation allows the side chain of Leu63 to change conformation, as required for 4D9 binding (i.e. Figures 1A–B versus 1C–D; 5A, S3). In the P6122 form, Ile72 interacts with itself on a crystallographic twofold, and its orientation and packing environment prevent rearrangement of Leu63 (Figure S3). This implies that during co-crystallization, 4D9 binding requires conformations of Leu63 and Ile72 that are accommodated in the P21212 crystal form. In other words, appearance of the P21212 crystals under conditions for the P6122 form only occurs when 4D9 binds in the exo site. These structures illustrate the binding potential of the exo site for small molecules, when particular conformational states of PR are sampled.

Outside/top of flap

As with 4D9, appearance of a new crystal form (P212121) during co-crystallization with mixtures F1 and F4 correlates with fragment binding (Tables 1 and 2). From the J1 data set (Table 3), the fragment 1F1 (indole-6-carboxylic acid) binds in a pocket on the surface of the flap in Monomer B, flanked by side chains from the anti-parallel β strands of the flap (Figures 2A and S2). A new salt bridge is formed between 1F1 and Arg57, disrupting the interactions of Arg57 with Glu35, Met36, Tyr59, and Trp42. The 1F1 indole also donates a hydrogen bond to the backbone carbonyl of Val56, stacks with Trp42, and has hydrophobic contacts with Pro44, Met46, and Lys55. Binding of 1F1 was confirmed in co-crystallization experiments at 2.5 mm and 10 mm (K5L, K8, L2 data sets). The P212121 form also exhibits a lattice contact in which Gln61 is inserted into the exo site of Monomer B (Figure S4). The fragment 2F4 (2-acetyl-benzothiophene) binds like 1F1 in the outside/top of the flap, where it stacks with Trp42 and makes hydrophobic contacts with Pro44, Met46, and Lys55, but it cannot form a hydrogen bond with Val56 (Figures 2B, S2; F1, F2 data sets). Flap residues are rearranged similarly, except that the larger acetyl moiety prevents pairing with Arg57, which retains interactions with Glu35, Met36, and Tyr59 when 2F4 is bound.

image

Figure 2.  Unbiased difference electron density contoured at 2, 3, 4 σ for fragment binding to the outside/top of the HIV PR flap. (A) Fragment 1F1 (indole-6-carboxylic acid) in the structure derived from the J1 data set (Table 3) where the P212121 crystal form was co-crystallized in the presence of fragment mixture F1 containing 1F1 at 2.5 mm. In the rearranged structure, 1F1 salt-bridges to Arg57, hydrogen bonds to Val56, and packs with Trp42, Pro44, Met46, and Lys55. (B) Fragment 2F4 (2-acetyl-benzothiophene) in the structure derived from the F2 data set where the P212121 crystal form was co-crystallized in the presence of fragment mixture F4 containing 2F4 at 2.5 mm. The structure is similar to that with 1F1 bound, but Arg57 interacts with Glu35 in the rearranged segment of residues 35–41, and no other hydrogen bonds are formed.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

Three fragment hits at two sites in the inhibited, closed form of HIV PR have been identified (Figure 3). The hit rate of ∼1% may result from the limited number of co-crystallization conditions used to screen the fragment library (Table 1), as well as the structurally diverse nature of the fragments in this Active Sight library. Intuitively, one might expect that specific small molecule binding to a solvent-exposed surface cleft on a protein would be less favorable compared to a more buried cavity. Nevertheless, a virtue of co-crystallization has been the 1:1 correspondence between the appearance of new crystal forms and fragment binding hits: 4D9, 1F1, and 2F4 were the only instances of the phenomenon of change in space group in the entire screen (Table 2). These hits were confirmed in replicate experiments and by soaking, if the crystal form was compatible, i.e. for 4D9 in the P21212 form. In the case of 4D9 and the P6122 form, soaking failed, because the lattice restricts conformational changes associated with fragment binding. For 1F1 and 2F4, growth of the P212121 form requires the fragment to bind, i.e. P212121 form crystals were not available for soaking. Interestingly, soaking the P6122 form with 10 mm 1F1 completely destroys the crystals, consistent with the observed rearrangement of flap residues. The inference from these crystallographic results, where fragment binding is associated with distinct conformational changes in PR and a concomitant change in the crystal form, is that the hits are not false positives, but rather indicative of novel sites on HIV PR generally capable of binding small molecules in a specific manner.

image

Figure 3.  Schematic view normal to the local twofold axis of the HIV PR dimer with the C2 symmetric inhibitor, TL-3, bound in the active site. Positions of two sites on the protein surface identified for fragment binding, the exo site, and the outside/top of the flap, are indicated by space filling models for fragments 4D9 and 1F1, respectively; 2F4 also binds in the 1F1 site. In each case, fragments bind to only one monomer, while the PR dimer occupies the asymmetric unit of the crystal.

Download figure to PowerPoint

PR binding sites

Each fragment site is distal from the local twofold axis and present in one monomer of the PR dimer (Figure 3). The exo site is flanked by a pair of loops (residues 14–18 and 38–42) and a β strand (residues 59–65) (Figure 4A). As a preexisting shallow groove on the protein surface, the only conformational change required for 4D9 to bind to the closed form of PR is rotation of the χ1 and χ2 torsion angles of Leu63 (Figure 5A). The amide of Gly17 donates a hydrogen bond to the 4D9 hydroxyl group, and the cyclohexyl ring and methyl group fit within a pocket defined by carbon atoms of Lys14, Gly16, Gly17, and Leu63, making seven potential van der Waals contacts. In the other monomer, or in other crystal structures, the Gly17 amide hydrogen bonds to acetate, DMSO, H2O, or the carboxamide moiety of Gln61 by symmetry. Hence, the Gly17 amide appears to be a determinant for specific binding of small molecules in the exo site. Similarly, the adjacent amide of Gly16 and carbonyl and amide of Leu63 form hydrogen bonds with DMSO and H2O.

image

Figure 4.  Space filling rendering of the exo site (A) and outside/top of the flap (B) showing the solvent-exposed clefts on the PR surface into which the fragments bind. The exo site is a preexisting feature of the protein fold while the outside/top of the flap rearranges to accommodate binding of 1F1 or 2F4. N, O, and S atoms are colored blue, red, and yellow, respectively.

Download figure to PowerPoint

image

Figure 5.  Conformational changes associated with fragment binding in the exo site (A) and outside/top of the flap (B). In (A) the structures derived from the H2 (gold, 4D9 bound) and 1B8 (gray, acetate bound), data sets are superposed (rms deviation 0.27 Å). Leu63 is repositioned, but other residues in the exo site are essentially unchanged. In (B) subunit B in the structure derived from the J1 data set (gold, 1F1 bound) is superposed onto subunit A (gray) (rms deviation 0.85 Å). Significant conformational rearrangement occurs in three anti-parallel β strands at residues 45–47, 53–56, and 78–81, and the side chains of residues 42, 44, 46, 55, and 57 in contact with 1F1 are shifted. A similar rearrangement occurs in the PR structure with 2F4 bound.

Download figure to PowerPoint

The 1F1 and 2F4 binding site at the outside/top of the flap involves residues 42–46 and 55–57 (Figure 4B) and entails both disruption of side chain interactions and shifts in the main chain (Figure 5B). Hydrophobic interactions with Trp42, Pro44, Met46, and Lys55 in the pocket induced by fragment binding appear important for recognition of the indole or benzothiophene rings when complemented by electrostatic interactions between the fragment’s carboxyl group and the side chain of Arg57. Significant conformational changes occur: the anti-parallel β strand segments at residues 45–47, 53–56, and 78–81 are shifted laterally, and residues 35–41 undergo rearrangement (1.7 Å on average for seven Cα’s) (Figure 5B). Comparison of PR monomers at the base of the flap with and without 1F1 bound, to ‘wide-open’ versus closed structures (2PC0 versus 2AZ8), shows that the β strands shift in the same direction for unbound[RIGHTWARDS ARROW]1F1 as for closed[RIGHTWARDS ARROW]open. In other words, the PR conformation with 1F1 bound resembles in part the open form, even though TL-3 is bound in the active site and the flaps are closed. In addition, the difference in structure is greater for 1F1 binding than for flap opening (0.99 Å versus 0.52 Å for 11 Cα’s on the three β strands).

Prediction that a MW 293 molecule, ‘Damm compound 1’, binds to the inside of the flap in the ‘eye site’ of the semi-open, ‘curled’ form of HIV PR (50), is consistent with the fragment hit 5NI in the P41 crystal form (Figure S1). Similarly, two molecules of a C2 symmetric pyrrolidine-based inhibitor bind on the twofold axis of the PR dimer, interact with Asp25 residues in the active site, and contact the inside surfaces of the flaps in a wide-open conformation (51).

Allosteric inhibitors

As hypothesized in previous manuscripts, a large enough allosteric ligand bound in the exo site could restrict compression of residues 38–42 and 59–63, thereby inhibiting the flaps from opening, as a result of the anti-correlated motion of these chain segments (11,21). In the previous restrained MD simulations with wild-type versus V82F/I84V MDR mutant, the exo site distance restraints that were shown to regulate the motion of the flaps and to stabilize the closed conformation of PR also shifted the estimated free energy of binding of the inhibitor, JE-2147, in ‘Relaxed Complex’ docking calculations (11). By restraining the exo site to be slightly expanded, which favored flap closure, near wild-type flap behavior was restored to the multidrug-resistant mutant system; i.e. the greater dynamic mobility of the MDR mutant’s flaps was suppressed while the binding affinity for an active site inhibitor was enhanced.

Although no new simulations were performed to generate the data presented herein, these crystallographic data do support the conclusions that were based on previously published simulations (11,21). Thus, this fragment-based screen against HIV PR provided experimental confirmation of the existence of allosteric binding sites that were predicted in previous simulations. As the 1F1 and 4D9 fragments each only bind to a specific site in particular crystal forms of the inhibitor-bound PR, and as these fragments induce distinct changes in the conformational preferences of PR, these crystallographic results support the idea that various conformations of the flaps can be specifically targeted to control PR activity, demonstrating the possibility for allosteric control of HIV protease. In accord with the previous conventional, restrained, and coarse-grained MD simulations (11,12,21), the presence of 4D9 in the exo site when PR has a flipped conformation of Leu63 and a closed conformation of the flaps supports the hypothesis that allosteric inhibitors could be developed against HIV PR. Inhibitors of this nature could synergistically improve affinities of current FDA-approved active site PR inhibitors, in particular for multidrug-resistant mutants (11,21).

Conclusions and Future Directions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

All current FDA-approved PR inhibitors target the active site (52). Strategies to suppress the evolution of resistance in HIV PR include the development of inhibitors that fit the peptide substrate envelope (53), contribute more main chain hydrogen bonds (54), or provide a polar ‘solvent anchor’ (55). To improve the long-term treatment of patients infected with HIV, especially those infected by MDR mutant strains, new types of drugs with novel mechanisms of action are urgently needed. In concept, trapping inactive protein conformations with allosteric inhibitors offers greater selectivity in drug design (56). The potential for targeting distinct conformational states of the flaps has been addressed experimentally (50,51) and computationally (50), and has been discussed in a review article (57).

Non-active site, allosteric inhibitors would have the virtue of avoiding selective pressure for either the PR active site or the Gag-Pol substrate to mutate (58). In combination with active site inhibitors, these allosteric inhibitors would likely increase the number of PR mutations required for significant clinical resistance to HAART. The fragment hits provide a starting point for the development of larger, higher affinity molecules that bind in the exo site, or the outside/top of the flap, as prospective allosteric inhibitors. Thus, these results lay the structural foundation for the development of a new class of HIV PR inhibitors: allosteric inhibitors that might rescue the efficacy of the current HIV PR drugs against the increasingly prevalent multidrug-resistant mutants. The development of these allosteric fragments into high-affinity allosteric inhibitors will be pursued in the next stage of this research project, with guidance provided by new virtual high-throughput screens that are currently being performed as part of the FightAIDS@Home project.

Footnotes

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

We thank Drs. Rodney Harris and David S. Goodsell for insightful discussions and assistance in using the AutoLigand and AutoDock programs. We thank Dr. John Badger at Active Sight for assistance with MIFit custom software, and Drs. Qingping Xu and Ashley Deacon at SSRL for calculations using JCSG program scripts. We thank Assoc. Prof. Heather Carlson for co-ordinates of ‘Damm compound 1’ docked to apo-PR. This research was carried out to support and extend the FightAIDS@Home project (a part of IBM’s ‘World Community Grid’). This research was supported by NIH grant P01 GM083658-01. Portions of this research were carried out at the Stanford Synchrotron Radiation Lightsource, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program, and the National Institute of General Medical Sciences.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Materials
  5. Results
  6. Discussion
  7. Conclusions and Future Directions
  8. Acknowledgments
  9. References
  10. Supporting Information

Figure S1: Superimposition of the crystallographic complex of the 5-nitroindole fragment bound to the “eye site” of a semi-open conformation of PR (green ribbon; P41 crystal form, Tables 1, 2) with the recently-published, predicted complex of “Damm compound 1” (ref. 50) docked against a semi-open PR crystal structure (purple ribbon; PDB code 1HHP = ref. 59).

Figure S2: Unbiased difference electron density maps from fragment screening experiments.

Figure S3: Crystal packing interactions in the P6122 form prevent the conformational change necessary for 4D9 binding in the exo site, as occurs in the P21212 form.

Figure S4: Insertion of Gln61 from Monomer A of a symmetry related PR dimer (cyan) into the exo site of Monomer B (green) in the P212121 crystal form (Table 1).

Table S1: Refinement Criteria for Six PDB Depositions in Table 3

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
CBDD_943_sm_FigsS1-S4_TableS1.pdf326KSupporting info item
CBDD_943_sm_FiguresS1-S4.docx2239KSupporting info item
CBDD_943_sm_TableS1.docx20KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.