New Active HIV-1 Protease Inhibitors Derived from 3-Hexanol: Conformation Study of the Free Inhibitors in Crystalline State and in Complex with the Enzyme


  • Natasza E. Ziółkowska,

    1. Institute of Technical Biochemistry, Technical University of Łódź, Stefanowskiego 4/10, 90-924 Łódź, Poland
    Search for more papers by this author
    • Present address: Cell Biology and Biophysics Unit, National Institute of Neurological Disorders and Stroke, Bethesda, MD, USA.

  • Anna Bujacz,

    1. Institute of Technical Biochemistry, Technical University of Łódź, Stefanowskiego 4/10, 90-924 Łódź, Poland
    Search for more papers by this author
  • Ramnarayan S. Randad,

    1. Division of Chemistry I, Office of Generic Drugs, FDA, Rockville, MD 20855, USA
    Search for more papers by this author
  • John W. Erickson,

    1. Sequoia Pharmaceuticals, Inc., 401 Professional Drive, Suite 200, Gaithersburg, MD 20879, USA
    Search for more papers by this author
  • Tereza Skálová,

    1. Institute of Macromolecular Chemistry, Czech Academy of Science, Heyrovskeho nam. 2, 162 06 Prague 6, Czech Republic
    Search for more papers by this author
  • Jindřich Hašek,

    1. Institute of Macromolecular Chemistry, Czech Academy of Science, Heyrovskeho nam. 2, 162 06 Prague 6, Czech Republic
    Search for more papers by this author
  • Grzegorz Bujacz

    Corresponding author
    1. Institute of Technical Biochemistry, Technical University of Łódź, Stefanowskiego 4/10, 90-924 Łódź, Poland
      Corresponding author: Grzegorz Bujacz,
    Search for more papers by this author

Corresponding author: Grzegorz Bujacz,


Four novel linear non-peptidic HIV-1 protease inhibitors derived from 2,5-diamino-1,6-diphenyl-3-hexanol were synthesized and characterized. All of them exhibit tight binding to HIV-1 protease, with inhibition constants Ki in the range 20 pm–5 nm. The investigated inhibitors were crystallized, and their crystal structures were determined by X-ray diffraction. In all cases, the conformations found in the crystalline state differ significantly from the conformations obtained by computational docking of the inhibitor in the binding cleft of native HIV-1 protease. Owing to the prevalence of hydrophobic substituents in all these inhibitors, the conformational mobility in water solution is restricted to their compact forms. The spectrum of low-energy conformations in solution dramatically changes during the formation of inhibitor crystals (phenyl ring stacking as a leading motif) or during the formation of a complex with HIV-1 protease (elongated conformation suitable to fit the enzyme pockets as a factor responsible for tight binding). High conformational flexibility and low conformational stress in the molecules of these inhibitors most likely increase their biological activity in comparison with more rigid compounds.

Acquired immunodeficiency syndrome (AIDS) is the final and most serious stage of the disease caused by the human immunodeficiency virus (HIV) (1–3). The main therapeutic strategies for intervention in cases of HIV-1 infections involve the inhibition of enzymes that play critical roles in the life cycle of the virus: reverse transcriptase (RT), protease (PR), and integrase (IN) (4–6). The crystal structure of HIV-1 protease was solved in 1989 (7–9). Structural studies have become the main basis for the investigation of inhibitors of this enzyme and rational drug design (10–13). Many of the potential drugs against AIDS, including inhibitors of HIV-1 protease, have received FDA approval; they include saquinavir (14–16), indinavir (17,18), ritonavir (19,20), nelfinavir (21,22), lopinavir (23,24), amprenavir (25–27), kaletra (combination of lopinavir and ritonavir) (28,29), atazanavir (30,31), tipranavir (32,33), and darunavir (34,35).

The aim of continuing the search for new HIV-1 protease inhibitors is to create wide-spectrum drugs, active against the mutations in the HIV-1 genome which rapidly develop in patients under the selection pressure of all of the clinically approved drugs (36–40). Among the inhibitors that have been characterized to date, the substrate-mimic inhibitors with hydroxyethylene (41), hydroxyethylamine (42,43), ethylamine (44), and reduced amide (45) replacement of the cleaved amide bond are the best known.

The inhibitors studied in this work can be classified as competitive, substrate based but non-peptidic, with the hydroxyethylamine isostere replacing the cleavable peptide bond. The structures of HIV-1 protease complexes with hydroxyethylene inhibitors form a subset of the most frequently determined structures of the HIV-1 protease. However, the inhibitors designed as peptide substrate analogs have a common disadvantage of possible enzymatic cleavage and other adverse effects in the body. The inhibitors studied here have a central fragment (isostere of the cleavage site in substrate) resistant to hydrolysis (hydroxyethylene) and non-peptidic substituents imitating the side chains of the substrate. Design and synthesis of these inhibitors were performed with the intention to optimize the interactions between the P2/P2′ substituents of inhibitors and the S2/S2′ pockets of HIV-1 protease.

Despite the fact that over one hundred experimentally determined structures of complexes of different mutants of the HIV protease with inhibitors belonging to several different classes are already available in the PDB (46), and over 300 in HIVdb (47), little attention has been paid to the conformational variability in the structures of inhibitor molecules in different environments. This study of conformational behavior of inhibitors in different environments brings a better understanding of the nature of ligand–enzyme interaction.

The structures and biological activity of four new peptidomimetic inhibitors of HIV-1 protease, derivatives of 1,6-diphenyl-3-hexanol based on the hydroxyethylene isostere (48), are reported here. The synthesized compounds were recrystallized to obtain single crystals suitable for X-ray analysis. The conformation of free inhibitors in crystalline state is compared with their conformation after binding into the HIV-1 protease active site.

Methods and Materials

Chemical methods

Although the reported series includes only four inhibitors, it is not possible to write only a single general procedure of obtaining them. Condensation of the core unit with P2/P2′ substituents was carried out using, in most cases, the standard peptide coupling reagents 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate (TBTU) and 1-hydroxybenzo-triazole hydrate (HOBt) in the presence of N,N-diisopropylethylamine (DIPEA) in dimethylformamide (DMF) as a solvent. Detailed procedures for the preparation of (2S,3S,5S)-2-N-(3-hydroxy-2-methylbenzoyl)amino-5-N-benzoylamino-1,6-diphenyl-3-hexanol (1), (2S,3S,5S)-2,5-N,N’-bis-(2-methylbenzoyl)amino-1,6-diphenyl-3-hexanol (2), (2S,3S,5S)-2-N-(3-hydroxy-2-methylbenzoyl)amino-5-N-[(3S)-tetrahydrofuryloxy-carbonyl]-amino-1,6-diphenyl-3-hexanol (3), and (2S,3S,5S)-2-N-[(2-pyridinylmethoxycarbonyl) valine]amino-5-N-(tert-butoxycarbonyl)-amino-1,6-diphenyl-3-hexanol (4) are shown in Appendix S1.

Biological activity

Ki values for 14 were determined by a fluorometric assay with the fluorogenic substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(DABCYL)-Arg and the chromogenic substrate Lys-Ala-Arg-Val-Tyr-Phe(p-NO2)-Glu-Ala-NleONH2, as previously described (49,50). The antiviral activity of 14 is expressed as 50% inhibitory concentration IC50 (Table 1) and was measured as previously described (50).

Table 1.   Structural formulae and biological activity of 14Thumbnail image of

X-ray crystallographic studies

Single crystals of compounds 1–4 were obtained by recrystallization using vapor diffusion method. The diffraction data sets for crystals of 1 and 2 were collected at room temperature on a CAD4 diffractometer equipped with graphite monochromatized CuKα radiation. An empirical absorption correction was applied by the use of the ψ-scan method [eac program (51)] for 1 and 2. The intensity correction was applied only for 2 [decay program (52)]. Data correction for 1 and 2 was carried out using the Enraf-Nonius SDP crystallographic computing package (52). The diffraction data sets for crystals of 3 and 4 were collected using synchrotron radiation at EMBL beamline BW7B (Desy, Hamburg, Germany). Diffraction images were recorded using a Mar 165 mm CCD detector at 100 K. Two data collection runs were made: high-resolution data run to 0.815 Å and low-resolution run to 1.21 Å. The data were processed by denzo (53) and scaled with scalepack (53).

All observed reflections with I > 0 were used to solve the structures of 14 by direct methods and to refine it by full matrix least-squares using F’s (54). Structures’ solution was performed using the program shelxs (54), and shelxl (55) was used for structure refinement. For molecule visualization and preparing Figures 1 and 2, the XPa program was used. Crystal data and experimental details are shown in Table S1. The torsion angles, the dihedral angles between planes passing through the atoms of six-membered and five-membered rings, and the lengths and angles of the hydrogen bonds for 14 were calculated using the program CSU (56).

Figure 1.

 Thermal ellipsoidal view of the molecules of 14 with the numbering scheme of atoms and P substituents. Thermal ellipsoids are shown with 50% probability.

Figure 2.

 Unit cells of 2 and 3. The molecules of crystallization solvent bound into the crystal lattice are indicated by the arrows.

Theoretical calculations

Two experimental structures of the complexes deposited in PDB (46), with inhibitors very similar to the investigated compounds and with data collected at medium resolution, were used as templates:

• HIV-1 protease complex with inhibitor L700,417b, pdb code: 4PHV (a = 58.88, b = 86.80, c = 46.79, P21212, resolution 2.1 Å) (57) for 1, 2, 3;

• HIV-1 protease complex with inhibitor SB206343, pdb code: 1HPS (a = 63.00, c = 83.30, P61, resolution 2.3 Å) for 4.

The inhibitors 1–4 were docked into the protease active site using program O (58) and minimized in discover in insightIIb using force field cff91. Solvent was considered as a dielectric continuum. Only one water molecule between inhibitor carbonyls (at P1, P1′ sites) and Ile49, Ile149 amines at tips of the HIV-1 protease flaps was included into the optimization analogous with many similar experimentally determined structures. Classical ionization schemes of amino acids corresponding to pH 5 were used. Non-bonded interactions were cut off at a distance of 14.00 Å. Energy and geometry optimization was carried out in three steps: (i) only hydrogen atoms were optimized, (ii) all hydrogen atoms of protease and all atoms of inhibitor were optimized, (iii) all atoms of the whole complex were optimized. The structures were minimized until maximum derivatives were <0.01 kcal/Å. The torsion angles and the dihedral angles between planes of the aromatic rings of molecules 14 were calculated using CSU (56).


Synthesis and design of the inhibitors

The synthesis of four protease HIV-1 inhibitors was performed: (2S,3S,5S)-2-N-(3-hydroxy-2-methylbenzoyl)amino-5-N-benzoylamino-1,6-diphenyl-3-hexanol (1), (2S,3S,5S)-2,5-N,N’-bis-(2-methylbenzoyl)amino-1,6-diphenyl-3-hexanol (2), (2S,3S,5S)-2-N-(3-hydroxy-2-methylbenzoyl)amino-5-N-[(3S)-tetrahydrofuryloxy-carbonyl]-amino-1,6-diphenyl-3-hexanol (3), and (2S,3S,5S)-2-N-[(2-pyridinylmethoxycarbonyl)valine]amino-5-N-(tert-butoxycarbonyl)-amino-1,6-diphenyl-3-hexanol (4).

The structural formulae of 14 are shown in Table 1. The substituents P1 and P1′ of all four inhibitors investigated here are hydrophobic. Phenyl rings are attached to the C1 and C6 atoms of the hexanol chain. These moieties imitate phenylalanine, which interacts with S1 and S1′ pockets of the protease. P2 substituents are identical for both 1 and 3: 3-hydroxy-2-methylbenzoyl. 2 is a symmetric inhibitor with 2-methylbenzoyl group as P2 and P2′. Inhibitor 4 additionally has a 2-pyridinylmethoxycarbonyl group as P3, which interacts with S3 pocket of protease. All investigated inhibitors differ in their P2′ substituents.

The compounds discussed here were selected for crystallization from a large group of HIV-1 protease inhibitors (59,60) (the synthesis and biological activity of the inhibitors from that group will be published elsewhere) in order to study their conformation in a crystal structure. The investigated inhibitors are based on the central fragment (2S,3S,5S)-2,5-diamino-1,6-diphenyl-3-hexanol, which was reported previously (33). Syntheses of this series of inhibitors were started from (2S,3S,5S)-5-amino-2-N-dibenzylamino-1,6-diphenyl-3-hexanol after protection on 5-N by Boc group gave (2S,3S,5S)-2-amino-5-N-(tert-butoxycarbonyl)-amino-1,6-diphenyl-3-hexanol (1s). The procedure of obtaining the final inhibitors 1–4 is shown on the Scheme 1, and experimental procedures are described in the Appendix S1. All final compounds were >99.9 pure, which was proved by standard chemical analyses, including HPLC, TLC, MS, and 1H NMR.

Figure Scheme 1:.

 The procedure of obtaining inhibitors 14. a,e – 1 m, HCl in dioxane, b – 2-methyl-benzoic acid, c – 2-pyridinylmethoxycarbonyl-(S)-val, d – 2-methyl-3-hydroxybenzoic acid, f – benzoic acid, g – (S)-3-tetrahydrofurylo-N-succinimidyl carbonate, CH3CN, ET3N.

Crystal structures of the inhibitors

Crystals suitable for data collection were obtained by the vapor diffusion method. CAD4 diffractometer was used for diffraction data collection for the crystals of 1 and 2. Synchrotron radiation was needed to obtain useful diffraction data for 3 and 4. Thermal ellipsoidal view of the molecules of 14 is shown in Figure 1.

The investigated inhibitors crystallize in various space groups and crystallographic systems (triclinic, monoclinic, orthorhombic, and trigonal – Table S1) and show significant differences of packing in the unit cells. The structures of all investigated compounds reveal good agreement between the measured and calculated data. The best structural model was obtained for 1 where Robs is <0.05 (Table S1). Diffraction data for 2 show that for 4146 unique reflections, there are only 1966 reflections observed with I > 2σ(I), whereas the structure contains 392 refined parameters. Additionally, the presence of a loosely bound solvent molecule in the crystal lattice introduces a partial disorder in the structure of 2. However, the structure was solved and refined to Robs < 0.07.

The molecules of the inhibitors under study are generally very flexible. However, the bulky phenyl substituents have restricted mobility and number of low-energy conformations in water, in solid state, and also after binding to the HIV-1 protease-binding site. The conformation of the investigated inhibitors in crystalline state is largely determined by crystal packing. The exact position and orientation of the molecules is then influenced by a net of hydrogen bonds and π-stacking of aromatic substituents. The inhibitor molecules are interconnected by hydrogen bonds in the crystalline state and form molecular layers mutually connected by van der Waals interactions and by phenyl rings stacking. In addition, the presence of planar solvent molecules influences the conformation of molecules 2 and 3 (Figure 2). The network of hydrogen bonds in the crystalline state of inhibitors 2–4 presented in Table 3 is very similar. The molecules are bound into chains parallel to the y-axis for 2 and into chains parallel to the z-axis for 4. A two-dimensional network of hydrogen bonds is created for 1 and 3: a pair of bridges binds the molecules of 1 into chains parallel to y-axis and a second pair of bridges binds these chains in the direction of the z-axis. The molecules of 3 are bound into chains parallel to the x-axis by three hydrogen bridges, and these chains are connected by one strong hydrogen bond in the direction of z-axis. The details of hydrogen bonding geometry in the crystal lattices of 1–4 are presented in Table 2.

Table 2.   Hydrogen bonding geometry in the crystal lattices of inhibitors 1–4
  1. Only hydrogen bonds with the H…Y distance shorter than 2.80 Å are listed.

X -H....YX…Y (Å)X-H…Y (°)Symmetry codesX -H....YX…Y (Å)X-H…Y (°)Symmetry codes
O1 – H...O32.679 (3)170.2 (31)1 − x, 0.5 + y, 0.5 − zO1 – H...O22.884 (6)146.0 (38)x, 1 + y, z
N2 – H...O22.965 (3)158.0 (25)1 − x, 0.5 + y, 0.5 − zN1 – H...O22.915 (5)160.0 (5)x, 1 + y, z
O4 – H...O12.708 (3)179.6 (39)2 − x, −0.5 + y, 0.5 − zN2 – H...O32.929 (5)161.3 (5)x, −1 + y, z
N1 – H...O43.086 (3)159.5 (26)2 − x, 0.5 + y, 0.5 − z
O1 – H...O22.761 (3)163.2 (36)1 + x, y, zO1 – H...O22.769 (3)176.0 (2)x, y, 1 + z
N1 – H...O22.966 (3)151.2 (31)1 + x, y, zN1 – H...O22.900 (2)160.5 (2)x, y, 1 + z
N2 – H...O32.929 (2)149.6 (30)−1 + x, y, zN2 – H...O32.921 (2)168.2 (2)x, y, −1 + z
O4 – H...O62.743 (3)162.4 (33)x, y, 1 + zN3 – H...O42.960 (3)158.6 (2)x, y, −1 + z

Molecules of benzene (crystallization solvent) are present in the crystal lattices of 2 and 3 (Figure 2). Two locations of that molecule are shown for 2 in the disordered model of solvent, with occupation refined to 56% and 44%. The presence of a large solvent molecule (the benzene ring) is an important factor for determining the conformation and crystal packing of 2 and 3, especially for the location of P1, P1′, P2, and P2′. A very small unit cell parameter c (about 5 Å) in comparison with a and b parameters (about 41 Å) corresponds to interactions in the crystal lattice of 4. There are hydrophobic intermolecular interactions around the threefold symmetry axis between tert-butyl groups and phenyl and pyridinyl rings in the unit cell of 4. There are also π-stacking interactions between the aromatic rings of the molecules from neighboring unit cells.

The values of selected torsion angles of 14 are shown in Table 3. For all the investigated compounds, there are very similar torsion angles in the rotation around the external bonds of the peptidomimetic inhibitor core chain: N1-C2, N1-C19, C2-C3, N2-C5, and N2-C26(C31). The conformation of 1 varies considerably in comparison with the other compounds, where the substituents P1′ and P2′ have different orientation. A noticeable difference is observed for the torsion angles of rotation around bonds C3-C4 and C4-C5 (in the central part of inhibitor) and for the torsion angles of rotation around the bonds C1–C2 and C5–C6. Contrarily, these torsion angles vary only by a few degrees for inhibitors 2, 3, and 4. Analysis of the X-ray structures presented in Table 4 shows that the dihedral angles between planes 1 and 1′ (passing through the atoms of P1, P1′ substituents) are very similar for compounds 2 and 3. The location and orientation of these substituents is mostly determined by the presence of solvent molecules in the crystal lattices. Dihedral angles between the planes 1′/2 for 2, 3 and the planes 1′/3 for 4 are in the narrow range of about 71°–80°. This fact suggests that attaching another substituent interacting with S3 pocket of the enzyme, when P2 has rather small size (l-valine), does not affect the mutual location of the aromatic rings of the considered chemical groups.

Table 3.   Comparison of selected torsion angles (°) of 14 in the crystal structures of free inhibitors (in crystal) and in the molecular models of complexes with HIV-1 protease (in the complex with HIV protease)
InhibitorIn crystalIn the complex with HIV-1 proteaseIn crystalIn the complex with HIV-1 protease
Table 4.   Orientation of planar substituents of inhibitors 14 in crystal and after binding to HIV protease Thumbnail image of

Inhibitor conformations in crystalline state and in the complex with HIV-1 protease

Based on experience in X-ray structure determination of several HIV-1 protease complexes with substrate-analog inhibitors (41–43,61–65), we docked the 14 inhibitors into the binding site of HIV-1 protease and optimized their geometry as described in Methods and Materials. The X-ray structures of 14 were used as input data for molecular modeling of their complexes with HIV-1 protease. The hydroxyl group of each inhibitor forms hydrogen bonds to catalytic aspartates Asp 25A and Asp 25B. The side chains of inhibitors fill the respective HIV-1 protease pockets S1 and S2. The geometry of complexes was optimized using the program insightII.b

Figure 3 [prepared using Pymol (66)] shows a comparison of conformations of inhibitors in the forms of complexes with HIV-1 protease. The selected torsion angles and dihedral angles between the planes of the aromatic rings of 1–4 before and after docking of an inhibitor are presented in Tables 3 and 4. The best similarity of these structures is observed for 1 where the atoms of the main chain and the P2, P2′ substituents of the inhibitor are mostly in the same positions as in the crystal structures and only the phenyl rings of P1, P1′ are rotated around the bonds C1–C2 and C5–C6. The difference in the position of P1′ and P2′ is significant for 24 where the largest disagreement is noticeable for rotation around the bonds C3–C4 and C4–C5 (about 100°). Additional differences between the X-ray structures and the modeled structures in complex are the orientation of P1′ phenyl ring of 2 (rotation round the bond C5–C6), the orientation of P1 phenyl ring of 3 (rotation round the bond C1–C2), and the orientation of P3 pyridinyl ring of 4 (rotation around the bond O5–C25). The dihedral angles of the planes of the aromatic rings of 1–3 vary considerably between the crystal structures and the molecular models. Similarities are observed only for the angles between the planes 1 and 2 of 1 (describing mutual orientation of rings of P1, P2) and the planes 1′ and 2 of 2 and 3 (showing mutual orientation of rings of P1′, P2). The angles between the planes of the aromatic rings of 4 for the free inhibitor in the crystal structure and in the molecular model of the enzyme-inhibitor complex are very similar except for the angle between the planes 1 and 3 (passing through the atoms of rings of P1 and P3, respectively).

Figure 3.

 The top and side views of the inhibitors 14 modeled into the native HIV-1 protease [carbon atoms colored blue (1), light gray (2), yellow (3), green (4)]. The inhibitor oxygens are red, nitrogens are dark blue). The side chains of the Asp25 and Asp25′ in the active site are shown in thin lines. The side view is rotated about 20° around y-axis to make a view more transparent.

Inhibitor binding

The binding scheme of hydroxyethylene isostere inhibitors is well known [see for example several structures of complexes with HIV-1 protease (1FQX, 1IIQ, 1LZQ, 1M0B, 1Z8C, 1ZBG, 1ZJ7, 1ZLF, 1ZPK, 1ZSF, 1ZSR) (41–43,62–65)]. A significant difference between conformation of those inhibitors and compounds 14 is in the antiparallel orientation of the amide groups at the opposite sides of the inhibitors’ main chains and in the non-peptidic nature of inhibitors’ side chains. The starting geometry of the HIV-1 protease used for molecular modeling of enzyme–inhibitor complexes was taken from the structures deposited in PDB (46) (pdb codes: 4PHV, 1HPS).

The modeled inhibitors 14 are connected with the HIV-1 protease-binding site by networks of hydrogen bonds with lengths in the range 2.66–3.65 Å [see Figure 4 prepared using ligplot (67)]. There are two hydrogen bridges mediated by the buried water molecule connecting Ile 50A and Ile 50B and the O1, O2 atoms of the inhibitors 14. The hydroxyethylene isostere binds to the catalytic residues Asp 25A and Asp 25B via three hydrogen bonds in 1 and 3 and via four hydrogen bonds in 2 and 4. Additionally, the isostere hydroxyl group creates a hydrogen bond to Ala 28B for 3. The hydrogen bridges formed by the other main chain atoms of the investigated inhibitors are between N1 and Gly 27A (1); N2 and Gly 27B (2); and N1 and Asp 25B, N1 and Gly 27B, N2 and Gly 27A, N3 and Gly48B, O4 and Arg8A, O4 and Asp29B (4). Hydrogen bonds also exist between the P substituents and the protease residues: the hydroxyl group of P2 and Asp 30A (1), the hydroxyl group of P2 and Asp 29B (3), and N4 atom of pyridinyl ring of P3 and Gly 48B (4).

Figure 4.

 Comparison of conformations of free inhibitors 14 in crystalline state (black) and conformations of inhibitors after docking in the complexes with HIV-1 protease (gray).

Figure 5 shows the superposition of inhibitors 14 docked into the native HIV-1 protease. The overlap is based on fitting of all the protease Cα atoms [calculated by program O (58)]. It is evident and not surprising that the chemical differences in inhibitors are reflected by conformational changes of their molecules and also by the conformation of protease side chains in the neighborhood of the inhibitors. HIV-1 protease dimer has twofold symmetry but the asymmetry of inhibitors (the central group –COH-CH2– and in the inhibitor side chains) induces asymmetry of the inhibitor–protease complexes and thus the direction is important. Respecting the used nomenclature, the shift in direction from P1 to P1′ is called here the shift to the right and vice versa.

Figure 5.

 The network of hydrogen bonds (within 3.6 Å) and non-bonded interactions (within 3.8 Å) formed between protease HIV-1 and the inhibitors 14 in the enzyme–inhibitor complexes.

The differences in inhibitors’ conformation in the complex with HIV-1 protease shown in the Figure 5 are to some degree compensated by small rotational movements of the protease side chains and by the movement of water molecule trapped usually between the P1 and P1′ carbonyls and the amine groups at the tips of the protease flaps. It means that the atom positions of the overlapped inhibitors should not be compared absolutely but should be evaluated for each inhibitor individually. In this sense, the most important differences influencing biological activity of inhibitors could be as follows:

  •  a shift of the molecule of the most active inhibitor 4 to the right, invoking a shift of the hydrophobic side chains in P2, P1, and P1′ positions by almost 2 Å,
  •  different conformation of the central part of inhibitor 4 between the P2 and P2′ sites in comparison with 1–3,
  •  the right shift of the Ile in 4 with respect to the 2-methylphenol moieties in P2 position of 1–3,
  •  the aromatic residue or purely hydrophobic residue in the P2 position seems to play a negative role.

Both Figure 5 comparing different inhibitors in the active site and Figure 3 showing large differences in inhibitor conformations in the inhibitor crystals and in the protease active site evidence that the stacking of aromatic substituents and their conformational mobility play an important role, especially the perpendicular orientation of phenyl ring in P1 position in 4. It is related also to different torsion angle C7-C1-C2-C3 for the inhibitor 4 (Table 3) and to the dihedral angle between phenyl planes in P1 and P1′ (see Table 4), showing a correlation with Ki (20°–∼5 nm, 40°–∼5 nm, 61°– 0.32 nm, 73°– 0.02 nm) for inhibitors 1–4.


Crystal and molecular structures of the crystalline inhibitors are good basis for a glimpse on inhibitor flexibility, factors responsible for their conformation, and the subsequent modeling of HIV-1 protease–inhibitor interactions. Comparison of the structures of the molecules of free inhibitors in the crystal and models of the inhibitors in complex with HIV-1 protease is presented here. It provides detailed information about conformational changes caused by the influence of external forces invoked by interactions with the surrounding molecules in the inhibitor crystals and after creating the complex with HIV-1 protease with moieties creating enzyme-binding sides. The water interaction with inhibitor plays an important role in changing of conformation upon complex formation. The hydrophobicity of mimicking side chains substituents force compact ‘spherical’ conformation in crystal form and in aqueous fluids is compensated by favorable interaction with HIV-1 protease after conformational fitting. All of these interactions have to be taken into account in in silico modeling of HIV-1 protease–inhibitor complexes.

1, 3, and 4 exhibit very good biological activity (Table 1) because of the possibility of creating many favorable contacts between substituents P, P’ and the S, S’ pockets of the enzyme (Figure 4). Compound 3, with (S)-3-tetrahydrofurane ring as P2′, combining hydrophobic and hydrophilic properties, is a better inhibitor of HIV-1 protease (Ki = 0.32 nm) than 1 and 2, with phenyl and 2-methylphenyl rings as P2′, respectively. The best inhibitor is 4 (K= 0.02 nm), with l-valine as P2, and 2-pyridinylmethoxy moiety as P3; these groups form a large number of hydrogen bonds in the binding site of the protease.

The difficulties in crystallization suggested significant conformational flexibility of the molecules of the investigated compounds. The use of synchrotron radiation (not standard for diffraction data collection for crystals of small molecules) was necessary to obtain useful data for 3 and 4. The unit cell parameters for crystals of 4 were relatively large (Table 2), comparable with the parameters typical for protein crystals. X-ray structural data reveal that in the crystalline state, none of the 1–4 inhibitors is in the β-extended conformation – the conformation characteristic for inhibitors bound in their complexes with the enzyme. Two main factors, a network of hydrogen bonds and the hydrophobic interactions, determine the conformation of the inhibitor both in its crystalline state and in complex with HIV-1 protease. Because the compounds contain only a small number of potential hydrogen bond donors, only a few such bonds are seen in the crystal structures, mostly between symmetrically related molecules.

The known macromolecular structure of a protease–inhibitor complex (37) deposited in PDB (46) and the crystal structures of free inhibitors were used to model the interactions in the active site of HIV-1 protease. The investigated compounds acting as competitive inhibitors bind tightly to the enzyme’s active site. In the enzyme–inhibitor complex, the chemical groups of the inhibitor molecule can serve as hydrogen bond donors and hydrogen bond acceptors in interactions with the amino acids creating the binding pocket.

The entropy terms seem to be related to rather flexible aromatic substituents in P1, P1′ positions. Dihedral angle between the P1 and P1′ phenyl rings is correlated with the measured inhibition constants. The optimized models show that hydrophobic interactions with the enzyme are crucial not only in the S1 and S1′ pockets but also in the S2 and S2′ pockets. The flexibility of the central part of inhibitors (i.e., rotations around the bonds C2–C3 and C3–C4) determines the elasticity of the inhibitor and fitting to the HIV-1 PR active site.

In spite of the fact that all four inhibitors fit virtually identical positions in the HIV-1 protease-binding site, the networks of hydrogen bonds, the strain in the main chain, the mobility of side chains, and the van der Waals interactions vary significantly. The resulting affinity of these inhibitors depends mainly on fitting the hydrophobic side chains into the binding pockets of HIV-1 protease with an excessive strain in the inhibitor main chain and the reduction in the side chains’ mobility. The binding site of HIV-1 protease, rich in hydrogen bond donors and acceptors, is able to create many types of networks that stabilize the exact position(s) of inhibitors under the HIV-1 protease flaps. Thus, the numerous hydrogen bonds are important for binding affinity, but not decisive.

Conclusions and Future Directions

The experimental data reported here show that the molecules of the inhibitors under study are very flexible. Large rotations around the bonds of the main chain and the side chains of the inhibitors are necessary to bind into the enzyme and/or to form the crystalline state. Thus, the flexible nature of the inhibitors and their ability to adopt an energetically most favorable conformation while binding to the enzyme increase their biological activity in comparison with more rigid compounds. The structural data presented here extend the knowledge of HIV-1 protease inhibitors and are a good basis for the design of more active compounds.


  • a

    XP-Interactive Molecular Grafics (1992) Siemens Analytical Xray Inst.

  • b

    insightII. (2000) Accelrys, Molecular Simulations Inc, available at:


The crystallographic work was supported by the Polish Committee for Scientific Research, KBN (Grant No 4P05F 00819), Czech-Polish bilateral program CONTAC (Grant No 8/2002-2003) and GA ČR 305/07/1073 and GA AV ČR IAA500500701. The authors are grateful to Wiesław Majzner for his assistance in the data collection on CAD4 and to Dr Alexander Wlodawer and Jerry Alexandratos for editorial suggestions. The EMBL – Hamburg sponsored the data collection on DESY synchrotron. The experimental parts including chemistry and biological tests were performed in the Structural Biochemistry Program, National Cancer Institute – Frederick Cancer Research and Development Center, SAIC, Frederick, MD, USA.