Crystal structures of HIV protease V82A and L90M mutants reveal changes in the indinavir-binding site Bhuvaneshwari Mahalingam 1, *, Yuan-Fang Wang 1 , Peter I. Boross 1,2 , Jozsef Tozser 2 , John M. Louis 3 , Robert W. Harrison 1,4 and Irene T. Weber 1,5 1 Department of Biology, Georgia State University, Atlanta, GA, USA; 2 Biochemistry and Molecular Biology Department, University of Debrecen, Hungary; 3 Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases, The National Institutes of Health, Bethesda, MD, USA; 4 Department of Computer Science, Georgia State University, Atlanta, GA, USA, 5 Department of Chemistry, Georgia State University, Atlanta, GA, USA The crystal structures of the wild-type HIV-1 protease (PR) and the two resistant variants, PR V82A and PR L90M ,have been determined in complex with the antiviral drug, indi- navir, to gain insight into the molecular basis of drug resistance. V82A and L90M correspond to an active site mutation and nonactive site mutation, respectively. The inhibition (K i )ofPR V82A and PR L90M was 3.3- and 0.16- fold, respectively, relative to the value for PR. They showed only a modest decrease, of 10–15%, in their k cat /K m values relative to PR. The crystal structures were refined to resolutions of 1.25–1.4 A ˚ to reveal critical features associ- ated with inhibitor resistance. PR V82A showed local changes in residues 81–82 at the site of the mutation, while PR L90M showed local changes near Met90 and an additional inter- action with indinavir. These structural differences concur with the kinetic data. Keywords: aspartic protease; crystal structure; drug resist- ance; HIV-1. Inhibitors of the HIV-1 protease are effective antiviral drugs for the treatment of acquired immune-deficiency syndrome (AIDS). However, their therapeutic efficacy is limited owing to the rapid selection of drug-resistant mutants of the protease. Analysis of clinical isolates has revealed extensive mutations in 45 residues of the 99-residue protease that are associated with resistance to protease inhibitors [1,2]. Indi- navir was one of the first protease inhibitors used as an antiviral agent to treat AIDS. Resistance to indinavir arises by a combination of different mutations in the protease gene [3,4]. A high level of resistance is associated with substitu- tions of up to 11 residues in the protease, although different combinations of these mutations have been observed [5]. Mutations of the conserved residues V82 and L90 are among those most commonly observed in protease inhibitor treat- ments [2], and are frequently observed, even in indinavir monotherapy [4]. Drug-resistant mutants of HIV protease are expected to show reduced sensitivity to specific inhibitors, while maintaining sufficient enzymatic activity and specific- ity for viral maturation and infectivity. However, single protease mutations and specific combinations can have lower viral infectivity than wild-type HIV. Drug-exposed HIV with multiple protease mutations, including resistant substitutions of M46I/G48V/L90M and F53L/A71V/V82A, produce defects in polyprotein processing and reduced viral infectivity [6]. Therefore, it is important to understand the molecular basis for the altered activity and structural changes of these resistant mutants as compared to the wild-type protease, in order to understand the molecular mechanism of resistance and to develop new antiviral therapies. Crystal structures show that HIV protease forms a binding site that consists of subsites S3–S4¢,whichspan about seven residues (P3–P4¢) of a peptide substrate [7]. The clinical inhibitors primarily bind in subsites S2–S2¢.Struc- tural and kinetic studies of resistant protease mutants have shown a range of effects that depend on the specific combination of mutation with substrate or inhibitor, as well as the assay conditions. Mutations observed in drug resistance have been classified either as substitutions in the active site (inhibitor-binding site) that directly influence inhibitor binding, or as substitutions of nonactive site residues with indirect influences. Previously, mutants with either increased or decreased catalytic activity, inhibition constants, and stability relative to the wild-type enzyme were observed, independently of the location of the mutation [8–14]. We have analyzed high-resolution crystal structures of the mature HIV-1 protease bearing either single or double substitution mutations bound to substrate analogs [13,15]. Some of these mutants showed structural changes consistent with differences in their enzymatic activity. Crystal structures of an inactive mature protease bearing the mutations D25N and V82A, in complex with inhibitors ritonavir or saquinavir, and substrates, show differences in the interactions of inhibitors as compared to Correspondence to I. T. Weber, Department of Biology, Georgia State University, PO Box 4010, Atlanta, GA 30302-4010, USA. Fax: + 1 404 651 2509, Tel.: + 1 404 651 0098, E-mail: iweber@gsu.edu Abbreviations: PR, wild-type HIV-1 protease; RMS, root mean square. *Present address: Renal Unit, Massachusetts General Hospital, Charlestown, MA 02129, USA. (Received 19 December 2003, revised 19 February 2004, accepted 27 February 2004) Eur. J. Biochem. 271, 1516–1524 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04060.x substrates [16]. The structural differences compared to the wild-type enzyme help to explain the resistant phenotype. Previously, the crystal structure of the HIV protease in complex with indinavir was reported at a resolution of 2.0 A ˚ [17,18]. The crystal structures of several multiple mutant drug-resistant HIV proteases with indinavir have also been reported at 2–2.5 A ˚ resolution [18–20]. In these structures it is difficult to interpret the effect of a single mutation. Hence, we describe the crystal structures of an optimized wild-type HIV-1 protease (termed PR, see the Experimental proce- dures), and drug-resistant mutants, PR V82A and PR L90M ,in complex with indinavir refined at 1.30-, 1.4- and 1.25 A ˚ resolution, respectively. Atomic details from these very high-resolution structures will be essential for the design of second-generation inhibitors against HIV-1 protease, to offset drug resistance. Experimental procedures Purification of HIV-1 protease constructs The wild-type mature HIV-1 protease (PR) [21], optimized for structural and kinetic studies, bears five mutations: Q7K, L33I, and L63I, which minimize the autoproteolysis of the protease; and C67A and C95A, which prevent cysteine-thiol oxidation. The kinetic parameters and stabil- ity of this mutant are indistinguishable from those of the mature enzyme [12,21]. Plasmid DNA (pET11a; Novagen) encoding the PR was used, together with the appropriate oligonucleotide primers, to generate the constructs PR V82A and PR L90M . All constructs were generated using the Quick- Change mutagenesis protocol (Stratagene) and verified by DNA sequencing and mass spectrometry. Escherichia coli BL21(DE3) was grown in LB (Luria–Bertani) medium at 37 °C and induced for expression. Proteins were prepared using an established protocol, as described previously [13]. Kinetic parameters, inhibition constants and urea denaturation assay The chromogenic substrate Lys-Ala-Arg-Val-Nle-p-nitro- Phe-Glu-Ala-Nle-amide (Sigma) was used to determine the kinetic parameters. Protease, at a final concentration of 70–120 n M , was added to varying concentrations of sub- strate (25–400 l M ), maintained in 50 m M sodium acetate, pH 5.0, containing 0.1 M NaCl, 1 m M EDTA, 1 m M 2-mercaptoethanol, and assayed by monitoring the decrease in absorbance at 310 nm using a PerkinElmer Lambda 35 or a Hitachi U-3000 UV/Vis spectrophotometer. The absor- bances were converted to substrate concentration via a calibration curve. The enzyme concentrations were based on active site titration data. The Michaelis–Menten curves were fitted using SIGMAPLOT 8.0.2 (SPSS Inc., Chicago, IL, USA). The K i values were obtained from the 50% inhibitory concentration (IC 50 ) values, estimated from an inhibitor dose)response curve with the spectroscopic assay and using the following equation: K i ¼½ðIC 50 À½E=2Þ=ð1 þ½S=K m Þ where [E] and [S] are the protease and substrate concentra- tions, respectively [22]. The urea denaturation assay was carried out by measur- ing the protease activity in the presence of 0–4.0 M urea, using the spectroscopic assay. Initial velocities were plotted against urea concentration and fitted to a curve for solvent denaturation using SIGMAPLOT 8.0.2. Crystallographic analysis Crystals were grown at room temperature by vapor diffusion using the hanging drop method. The protein (5–10 mgÆmL )1 ) was preincubated with a 5- or 10-fold molar excess of inhibitor. The crystallization drops con- tained a 1 : 1 ratio, by volume, of reservoir solution and protein. The final drop was typically 2 lL, and crystals grew in 2–7 days. The wild-type and PR V82A crystals were obtained using 0.05 M citrate/phosphate buffer, pH 5.0– 5.6, with 0.2–0.4 M NaCl as precipitant, while the PR L90M crystals were obtained using 0.1 M sodium acetate buffer, pH 4.6, with 2 M ammonium sulfate as precipitant. The crystals were frozen with a cryoprotectant of 20–30% glycerol. X-ray diffraction data were collected on beamline X26C at the National Synchrotron Light Source at Brookhaven. Data were processed using the HKL suite [23]. K45I (Protein Data Bank accession code 1DAZ) protease coordinates were used as the starting model for molecular replacement using AMORE [24]. The structures were refined using SHELX [25] and modeled using O [26]. Alternate conformations for residues were modeled where appropriate. The solvent was modeled with more than 150 water molecules, and ions were present in the crystallization solutions. Anisotropic B factors were refined for all the structures. Hydrogen atom positions were included in the last stage of refinement using all data once all other parameters, including disorder, had been modeled. The structures have been submitted to the Protein Data Bank with accession codes 1SDT(PR), 1SDU (PR L90M )and 1SDV (PR V82A ). Crystal structures were superimposed on all Ca atoms using an implementation of the algorithm described previously [27]. Results and discussion Kinetics and stability The PR and two variants, PR V82A and PR L90M , harboring single mutations that commonly appear in drug resistance, including indinavir treatment [2,4], were chosen for this study. While V82A alters a residue in the active site of the protease that is critical to inhibitor binding, L90M is distal to the inhibitor-binding site and is located near the dimerization interface. The kinetic parameters of protease- catalyzed hydrolysis of the chromogenic substrate, the inhibition constants for the hydrolytic reaction by the inhibitor indinavir, and the sensitivity to urea denaturation, are shown in Table 1. PR exhibited the highest k cat value compared to PR V82A and PR L90M , respectively. The k cat /K m values of PR V82A and PR L90M were % 85% and 81%, respectively, relative to PR. The PR L90M hydrolysis of three other peptide substrates has shown k cat /K m values of % 40%, relative to PR, albeit under different assay condi- tions [12]. The K i values for the three enzymes showed a greater variation than the k cat /K m values. The apparent Ó FEBS 2004 HIV protease crystal structures with indinavir (Eur. J. Biochem. 271) 1517 binding affinity of indinavir to PR V82A was about three-fold lower than that of PR. Klabe et al. [28], using different assay conditions, reported a six-fold lower K i value for the PR V82A –indinavir interaction relative to that observed with PR. The six-fold stronger inhibition of PR L90M by indinavir is consistent with its enhanced inhibition by peptide analog inhibitors [12]. The DDG values calculated for indinavir binding were 0.79 and )1.19 kcal/mol for PR V82A and PR L90M , respectively. These DDG values correspond to the loss in PR V82A ,orgaininPR L90M , of about one hydrogen bond or van der Waals contact relative to the PR interaction with indinavir. The stability of PR V82A , as assessed by urea denaturation, wassimilartothatofPR,whereasPR L90M was significantly more sensitive to urea. In general, PR V82A ,whichaltersa residue in the inhibitor-binding site, showed similar activity and stability, and lower inhibition, relative to PR. However, the nonactive site mutant, PR L90M , had a slightly lower activity, % 50% less stability, and improved inhibition by indinavir. Therefore, the single mutations showed inde- pendent effects on catalytic activity, inhibition, and stability, consistent with our previous findings [12]. Crystal structures The crystal structures are described for PR, PR V82A and PR L90M in complexes with indinavir at 1.25–1.4 A ˚ resolution. These are the highest resolution complexes with indinavir reported to date [17–20]. Data collection and refinement statistics are given in Table 2. The crystallo- graphic asymmetric unit had a dimer of HIV protease, with the residues in the two subunits numbered 1–99 and 1¢)99¢. In each structure the inhibitor was observed to have a single orientation, and 168–183 water molecules were included. The indinavir atoms are in well defined electron density, with the exception of the pyridyl group that showed higher B-values and disordered density, as shown for the PR complex in Fig. 1. The solvent for the two structures in space group P2 1 2 1 2 included two chloride ions, while the L90M structure in space group P2 1 2 1 2 1 included a sulfate and an acetate ion. All the ions were observed on the surface of the protein. Among the three complexes, the quality of the electron density maps decreased in the order: PR>PR L90M > PR V82A . The distribution of the mean B factors for the main chain atoms showed similar peaks at the termini and the variable surface loops of residues 16–18, 37–41, and 67– 69 of both subunits in all structures, except for the larger values at 50–52 and 51¢)54¢ in PR, and 80–81 in PR L90M (Fig. 2). In PR L90M , the main chain atoms of residues 79–81 in one subunit have relatively high B factors and anisotropic density. This anisotropy could be caused by the crystal lattice because close van der Waals contacts (% 3.5 A ˚ )are observed between Pro81 and symmetry related Tyr6¢ in PR L90M . Although this region has similar crystal contacts in both subunits, Pro79–Pro81 and the symmetry related Tyr6¢ are disordered, unlike the equivalent residues in the other subunit. However, the benzyl ring of indinavir, which is close to these residues, had very good electron density. Higher main chain B factors were observed for flap residues 50–52 and 51¢)54¢ in PR relative to the values in the two mutants. Ile50¢ was ordered in all three structures and the side chain fitted well in the S2 binding pocket created by the t-butyl group of indinavir. Ile50, on the other hand, was disordered in all the structures and had poorer comple- mentarity for the van der Waals interaction with the indanyl group. Analysis of the anisotropic displacement parameters, using PARVATI [29], showed a distribution typical for Table 1. Kinetic data. Kinetic parameters for hydrolysis of the spectroscopic substrate (Lys-Ala-Arg-Val-Nle-p-nitroPhe-Glu-Ala-Nle-amide), inhibition by indinavir and sensitivity to urea. The DDG values are calculated from RTlnKi. UC 50 , urea concentration at 50% activity. Protease K m (l M ) k cat (min )1 ) k cat /K m (min )1 Æl M )1 ) K i (p M ) DDG (kcalÆmol )1 ) UC 50 ( M ) PR 55.0 ± 7.0 285.0 ± 9.5 5.2 ± 0.2 540 ± 70 0.00 1.95 a PR V82A 44.0 ± 6.5 194.5 ± 7.6 4.4 ± 0.1 1810 ± 270 0.79 1.77 PR L90M 21.7 ± 1.9 91.2 ± 1.4 4.2 ± 0.4 86 ± 8 ) 1.19 1.00 a a Taken from Mahalingam et al. [12]. Table 2. Crystallographic data statistics. RMS, root mean square. Protease mutant PR PR V82A PR L90M Space group P2 1 2 1 2P2 1 2 1 2P2 1 2 1 2 1 Unit cell dimensions (A ˚ ) a 85.8 85.8 51.3 b 58.8 58.7 58.4 c 46.6 46.6 61.6 Unique reflections 54 269 41 646 50 225 R merge (%) 4.3 5.8 6.3 Overall (final shell) (25.8) (19.6) (37.4) I/sigma(I) 28.68 24.35 23.14 Overall (final shell) (7.96) (8.17) (4.41) Resolution range for refinement (A ˚ ) 10–1.30 10–1.40 10–1.25 R work (%) 15.51 15.97 14.11 R free (%) 19.04 20.54 18.32 No. of waters 169 183 168 Completeness (%) 92.7 88.7 96.8 Overall (final shell) (87.8) (91.5) (78.0) RMS deviation from ideality Bonds (A ˚ ) 0.012 0.010 0.012 Angle distance (A ˚ ) 0.029 0.028 0.030 Average B-factors (A ˚ 2 ) Main chain 12.5 10.4 10.9 Side chain 17.5 14.5 16.5 Inhibitor 15.9 13.5 13.9 Solvent 26.1 24.7 25.8 1518 B. Mahalingam et al. (Eur. J. Biochem. 271) Ó FEBS 2004 proteins with a mean anisotropy of 0.43–0.45 for protein atoms. Atoms that display large anisotropy are typically disordered or exhibit alternate conformations. The Cc and Od2 atoms of Asp25 in PR and PR V82A , and both the Od atoms of Asp25¢ in PR L90M , exhibited large anisotropy. This anisotropy probably represents different states of charge distribution and/or protonation of the catalytic aspartates. Several residues showed disordered density for the side chains and/or had alternate conformations (Table 3). Ile50, Met46 and Met46¢ in the flaps were disordered in all three structures. Disordered density has been reported previously for hydrophobic protease residues that interact with substrate analog inhibitors [13]. The side-chain of Val82 exhibited alternate conformations in both PR L90M and PR, suggesting that its mobility is intrinsic. In PR, the alternate conformations of Val82¢ appeared to be associated with those of Arg8¢,Glu21¢ and two intervening water molecules. This is an example of how residues that are further away from the binding pocket can be associated with the conformation of active-site residues. However, mutations of Arg8 and Glu21 are rarely observed in resistant isolates [2]. One of the conformations of the side chain of Lys45 in PR forms hydrogen bonds with the Od2 of Asp30, which is expected to stabilize the flap. The other conformation of Lys45 forms hydrogen bonds to the carbonyl oxygen of Met46 through a water molecule. ThesidechainsofMet90andMet90¢ in the PR L90M structure showed two conformations (Fig. 3), as described previously for complexes with peptide analog inhibitors [13,15]. Met90 showed one conformation with occupancy of 0.34, with the Ce atom at a short distance of 3.53 A ˚ from the carbonyl oxygen of Asp25. The other conformation of Met90, with 0.66 occupancy, had the Ce atom at 5.46 A ˚ from the carbonyl oxygen of Asp25. In the other subunit, Met90¢ had one conformation at occupancy of 0.45 in which the Ce atom was closer (3.43 A ˚ ) to the carbonyl oxygen of Asp25¢, while the other conformation had the Ce atom at 5.51 A ˚ from the carbonyl oxygen of Asp25¢. In comparison, the Leu90 and Leu90¢ in the PR showed the closest distances of 3.76 and 3.78 A ˚ to the carbonyl oxygen of the catalytic Asp25 and Asp25¢, respectively. Alternate conformations were not observed for Leu90 or Leu90¢ in either PR or PR V82A . Therefore, the shorter van der Waals contact between the minor conformation of Met90/90¢ and the carbonyl oxygen of the catalytic Asp was proposed to account for the lowered catalytic activity and stability of the PR L90M mutant compared to the PR [13]. Similar close contacts were reported for Met90 in the crystal structure of the mutant G48V/L90M with saquinavir [30]. Presumably the presence of the Met90 conformation in close contact with Asp25 arises from an unusual electron distribution around the catalytic residues that is required for the proteolytic reaction. Protease–indinavir interactions One molecule of indinavir bound to protease residues from both subunits (Table 4). The inhibitors in all the three structures superpose very well, except for a small change in position of the pyridyl end in PR L90M (Fig. 4A). The pyridyl group of indinavir is accessible to the surface, while all the other groups in indinavir are shielded either by a network of water molecules or protease residues. Residues Arg8¢ and Val82¢, which surround the partly disordered pyridyl group, exhibit alternate side chain conformations in PR. The Pro81¢ ring is puckered away to avoid unfavorable interactions with C36 of the pyridyl group. Pro81, which contacts the benzyl group, has a ring pucker towards the Fig. 1. Omit map for indinavir in the wild-type HIV-1 protease (PR) crystal structure. The contour level is 3.5. The polar atoms and pyridyl group of indinavir are labeled. Fig. 2. Mean B-values for main chain atoms of the wild-type HIV-1 protease (PR) and PR L90M . The mean B-values (A ˚ 2 ) are plotted for the residuesofPR( )andPR L90M (––). The residues in the two subunits are numbered 1–99 and 1¢)99¢.TheB-values for PR V82A are not shown because they are lower than in the other structures. Table 3. Amino acid side-chains with conformational flexibility. Subunit PR PR V82A PR L90M AB AB AB Alternate conformation Glu21 Arg8¢, Met46¢ Met46 Lys45 Glu21¢ Val82¢ Ile33¢ Ile84 Val82 Val82 Met90 Met90¢ Disordered density Leu23 Lys45¢ Met46 Met46¢ Met46 Met46¢, Met46 Ile50 Phe53¢ Ile50 Ile50 Val82¢ Ó FEBS 2004 HIV protease crystal structures with indinavir (Eur. J. Biochem. 271) 1519 inhibitor. Thus, subtle conformational changes of residues interacting with the inhibitor play a role in the kinetics. The number of protease–indinavir van der Waals contacts showed only a small variation among the three crystal structures. PR had 96 van der Waals contacts, with interatomic distances of < 4.0 A ˚ for the major conformation of the side chains and 98 contacts for the minor conformations. PR V82A showed 95 van der Waals contacts with indinavir, similar to the PR. However, PR L90M showed fewer van der Waals contacts with indinavir than PR: 93 for the major and 92 for the minor side chain conformations. The three crystal structures showed a very similar arrangement of proteaseÆindinavir hydrogen bond inter- Fig. 3. Interaction of Met90¢ and Asp25¢ in PR L90M . (A) The 2Fo–Fc electron density map showing Met90¢,Asp25¢ and Thr26¢ in the PR L90M structure. The side chain of Met90¢ has two conformations, and one conformation has a short separation from the carbonyl oxy- gen of the catalytic Asp25¢.(B)ComparisonofMet90¢ in PR L90M and Leu90¢ in the wild-type HIV-1 protease (PR) relative to Asp25¢.The PR residues are in black and the PR L90M residues are in gray. Hydrogen bonds are indicated by dashed lines, with the distances shown in A ˚ . Table 4. Protease residues with van der Waals interactions with indi- navir. Interatomic distances of 3.3–4.2 A ˚ indicate van der Waals contacts. Subunit A Subunit B Arg8 a Arg8¢ b Leu23 Leu23¢ c Asp25 a Asp25¢ a Gly27 a Gly27¢ a Ala28 Ala28¢ Asp29¢ a Asp30 Asp30¢ Val32 Val32¢ Ile47 Ile47¢ Gly48 Gly48¢ Gly49 Gly49¢ Ile50 a Ile50¢ a Pro81 Pro81¢ Val/Ala82 Val/Ala82¢ Ile84 Ile84¢ a Residues with hydrogen bond or water–mediated interactions (Fig. 4 and Table 5). b Hydrogen bond interaction only in PR L90M . c Interatomic distance of > 4.3 A ˚ in PR L90M . Fig. 4. Protease hydrogen bond interactions with indinavir. The stereo figures were prepared using MOLSCRIPT [31]. Hydrogen bonds are indicated by dashed lines, with the distances shown in A ˚ .(A)Com- parison of indinavir interactions with Arg8¢ in the wild-type HIV-1 protease (PR) (black) and PR L90M (gray). (B) PR interactions with indinavir. The indinavir bonds are in black, the protease bonds are in gray, and water molecules are represented as spheres and labeled A–D. 1520 B. Mahalingam et al. (Eur. J. Biochem. 271) Ó FEBS 2004 actions, including the same water-mediated interactions, except for one new interaction in PR L90M (Fig. 4 and Table 5). Similar hydrogen bond and water-mediated interactions were observed in the previous crystal structure of wild-type protease with indinavir (PDB code 1HSG), except that water OD was not observed [17]. Four water molecules that mediate interactions between indinavir and the protease were observed in all the high resolution crystal structures. The O4 atom of indinavir formed hydrogen bond interactions with the amide and OD2 of Asp29¢,and interacted via water OB with the carbonyl oxygen of Gly27, the OD1 of Asp29¢ and the NE atom of Arg8. The N4 atom of indinavir interacted with the carbonyl oxygen of Gly27¢. The indinavir O1 and O3 atoms formed hydrogen bond interactions through a water molecule (OA) to the amides of Ile50 and 50¢. Indinavir N2 showed a water-mediated interaction with the amide nitrogen of Asp29. Indinavir N1 formed a water-mediated hydrogen bond to the carbonyl oxygen of Gly27; this interaction was not observed in the 1HSG structure [17]. The O2 hydroxyl group of indinavir formed hydrogen bonds with the four carboxylate oxygens of Asp25 and 25¢. The four O2 to carboxylate oxygen distances ranged from 2.7 to 2.9 A ˚ in PR and PR V82A . However, PR L90M showed greater asymmetry, with two shorter distances of 2.5 and 2.7 A ˚ and two longer distances of 2.9 and 3.2 A ˚ . It is possible that the asymmetrical interaction of indinavir O2 with the catalytic aspartates is associated with the short van der Waals interaction of the Met90/90¢ side chains with the carbonyl oxygen of Asp25/25¢ (Fig. 3B). In PR L90M ,the pyridyl N5 of indinavir formed a hydrogen bond with the NH 2 of Arg8¢ (3.1 A ˚ ). The corresponding distances in PR and PR V82A were 4.0 and 4.2 A ˚ , respectively. This new interaction could explain the better inhibition of PR L90M by indinavir. Structural differences between mutants and PR The PR V82A structure is very similar to the PR structure, with a root mean square (RMS) deviation of 0.12 A ˚ for all main chain atoms, as both crystal structures were obtained in the same space group. Only the main chain atoms of residues 81–82 and 81¢)82¢ showed larger RMS differences of 0.34–0.59 A ˚ , while the catalytic triplets of 25–27 and 25¢)27¢ showed very low RMS differences, of 0.03–0.05 A ˚ for main chain atoms in both subunits (Fig. 5). The estimated main chain errors (calculated from the B-values) were 0.13–0.21 A ˚ for residues 81–82 and 81¢)82¢,and 0.08–0.12 A ˚ for the catalytic triplets. There was a small movement of the main chain atoms of residues 81–82 towards indinavir in PR V82A , which partly compensated for the change from Val82 in PR to the smaller side chain of Ala (Fig. 6A). In PR, the two Cc atoms of Val82 formed van der Waals contacts with indinavir, of 3.8 A ˚ .In PR V82A , the change in the position of the main chain atoms placed the Cb atom of Ala82 within reasonable van der Waals distance of indinavir (4.1 A ˚ ), resulting in a loss of only one contact compared to PR. In contrast, studies of an inactive protease containing the mutations D25N/ V82A showed that Ala82 had no van der Waals contacts with the drugs saquinavir or ritonavir [16]. The structural changes observed in residues 81–82, which tend to compensate for the smaller Ala82 compared to Val in PR, were consistent with the small reduction in k cat /K m and the three-fold increased K i for indinavir with PR V82A relative to PR. PR L90M showed an RMS difference of 0.61 A ˚ compared to the PR, mainly owing to differences in lattice contacts in the two space groups. The catalytic triplet residues 25–27 showed values of 0.08–0.15 A ˚ for comparison of main chain atoms, consistent with the highly conserved core structure. Differences of > 1.0 A ˚ were observed for the main chain atoms of residues 16¢)18¢, 37–41/37¢)41¢,43¢)46¢,70¢)71¢ and 81 (Fig. 5). The estimated main chain errors were 0.37 A ˚ for Pro81 and 0.16–0.32 A ˚ for the other residues, suggesting that differences of > 1.0 A ˚ are significant. However, these large changes reflect variation in surface residues owing to the different space groups, except for Pro81 that forms part of the inhibitor-binding site. The main chain atoms of residues 79–81 in PR L90M have Table 5. Protease–indinavir hydrogen bond interactions. Indinavir Atoms Distance (A ˚ ) Water Protease PR PR V82A PR L90M 1HSG Direct interactions O4 OD2 Asp29¢ 3.3 3.3 3.1 3.1 O4 N Asp29¢ 3.0 3.0 3.0 3.2 N4 O Gly27¢ 3.1 3.1 3.2 3.0 OH OD1 Asp25 2.8 2.8 2.9 2.8 OD2 Asp25 2.7 2.7 2.5 2.9 OD1 Asp25¢ 2.9 2.8 2.7 2.6 OD2 Asp25¢ 2.8 2.8 3.2 3.0 N5 NH2 Arg8¢ 3.1 Water-mediated interactions O3 OA 2.6 2.7 2.6 2.8 O1 OA 2.8 2.8 2.8 2.7 OA N Ile50 3.0 2.9 2.9 3.1 OA N Ile50¢ 2.9 2.9 2.9 2.9 O4 OB 3.3 3.2 3.3 2.8 OB O Gly27¢ 2.7 2.7 2.6 3.2 OB OD1 Asp29¢ 2.8 2.8 2.8 2.7 OB NE Arg8 3.1 3.1 3.1 2.8 N2 OC 3.1 3.1 3.0 3.3 OC N Asp29 2.9 2.9 2.9 3.0 N1 OD 3.0 3.1 3.0 OD O Gly27 3.1 2.9 2.9 Fig. 5. Structural differences in main chain atoms. The root mean square (RMS) differences (A ˚ ) per residues are plotted for main chain atoms of PR V82A (––) and PR L90M (- - -) compared with the wild-type HIV-1 protease (PR). Ó FEBS 2004 HIV protease crystal structures with indinavir (Eur. J. Biochem. 271) 1521 anisotropic electron density. The density is ordered in the plane of the interaction with indinavir, shown in Fig. 6B, and extended/disordered in the perpendicular direction. Although the main chain atoms of 80–81 in PR L90M have moved 0.6–1.2 A ˚ further from indinavir compared to PR, the Cc of P81 has maintained similar van der Waals contacts with C19 of indinavir (3.8 and 3.9 A ˚ in PR and PR L90M , respectively) (Fig. 6B). The closest Cc atom of Val82 is 4.1 A ˚ from indinavir, only a little farther than the 3.9 A ˚ separation in PR. The positions of the main chain atoms, of 80–82, relative to indinavir, were consistent with the smaller number of van der Waals contacts between the protease and indinavir observed for PR L90M compared to PR (93 compared to 96). Residues 81¢ and 82¢ interact with the pyridyl group of indinavir. There appear to be small correlated changes in the position of the side chain of Pro81¢, the pyridyl group, Arg8¢ and Phe53 in PR L90M , relative to their positions in PR and PR V82A . The carboxylate groups of the catalytic Asp25 and Asp25¢ also showed a small shift relative to their positions in the other complexes, and less symmetrical interactions with the hydroxyl of indinavir (Table 5), as well as the close contacts between the carbonyl oxygen atoms and Met90/90¢ (Fig. 3). It is probable that these small structural changes result in the lowered activity and stability of PR L90M relative to PR (Table 1). The increase in affinity for indinavir may arise from the new hydrogen bond interaction between the pyridyl of indinavir and Arg8¢, and the structural changes associated with the close contact between Met90/190 and the catalytic aspartates. The new hydrogen bond interaction is consistent with the observed DDGof)1.19 kcal/mol for the inhibition of PR L90M compared to PR. The structural changes described for PR V82A and PR L90M , relative to PR, differ from those reported in previous studies with other mutations. Munshi et al.[19] suggested that the 80 s loop is intrinsically flexible; however, mutations can influence the conformation of this loop and its interactions with indinavir. PR with mutations M46I/ L63P/V82T/I84V showed structural changes in the flaps near the mutated Ile46, and in the interactions of the mutated Thr82 and Val84 with indinavir [18]. Local changes in the mutated residues were also observed in the crystal structure of the L63P/V82T/I84V mutant with indinavir [20]. Our structures of PR V82A and PR L90M showed opposite changes in the main chain atoms of residues 81–82, and PR L90M also had changes in the conformation of the catalytic aspartates, probably associated with the close contact of Met90, and in the side chains interacting with the pyridyl group of indinavir. The mutation V82A produced local changes around residue 82, while L90M showed both local and more distal changes propagating to the inhibitor- binding site. Conclusions The optimized wild-type HIV-1 protease (PR) and the drug- resistant mutants, PR V82A and PR L90M , were compared by using crystallographic and kinetic analysis. The two mutants showed slightly decreased k cat /K m values as compared to PR. PR V82A and PR L90M had increased and decreased K i values for indinavir, respectively, compared to PR. Most of the interactions with indinavir were similar for the three high resolution crystal structures. Small differences were observed in the van der Waals contacts with indinavir for the mutants compared to PR. The active site mutant, PR V82A , showed changes in the positions of the main chain atoms of residues 81–82 in both subunits that partially compensated for the mutation by improved interactions with indinavir. In contrast, PR L90M showed fewer van der Waals contacts with indinavir, the main chain atoms of residues 80–82 were further from the indinavir, and the side chain of Met90 and Met90¢ had altered interactions with the catalytic Asp25 and Asp25¢. However, there is a new polar interaction between the pyridyl N5 of indinavir and the side chain of Arg108, which may account for the apparent decreased K i of PR L90M for indinavir. Both the mutants showed small structural changes around the indinavir that must be interpreted, together with kinetic and stability data, in order to understand the effect of the mutation. The lower stability of PR L90M is consistent with the observed small structural changes in Asp25 and Asp25¢ at the dimer interface. The DDG values for binding of indinavir corres- pond to the observed loss in PR V82A of about one van der Waals contact and gain in PR L90M of one hydrogen bond relative to the PR interaction with indinavir. The structural and kinetic data suggest that the resistant mutation, V82A, acts directly to reduce the affinity for indinavir, while L90M appears to act indirectly by lowering the dimer stability, despite the apparent higher affinity for indinavir. The changes in protease structure and interactions with indinavir must be considered during the design of new inhibitors for resistant HIV. Fig. 6. Structural variation in residues 81–82 near indinavir. Stereoview showing the benzyl group of indinavir interacting with residues 81–82, using the major conformation of Val82. The wild-type HIV-1 protease (PR) structure is in black and the mutant is in gray bonds. Interatomic distances are given in A ˚ .(A)PR V82A superimposed on PR. (B) PR L90M superimposed on PR. 1522 B. Mahalingam et al. (Eur. J. Biochem. 271) Ó FEBS 2004 Acknowledgements We thank Xianfeng Chen for assistance with discussion of protease– indinavir interactions. We thank Merck & Co. for providing the indinavir used for the crystallographic analysis. The X-ray diffraction data were recorded at the beamline X26C of the National Synchrotron Light Source at Brookhaven National Laboratory, which is supported by the US Department of Energy, Division of Materials Sciences and Division of Chemical Sciences, under Contract No. DE-AC02- 98CH10886. The research was supported, in part, by the Georgia Research Alliance, National Institute of Health grants GM62920, AIDS-FIRCA TW01001, and Hungarian OTKA F35191 and T43482. References 1. Hertogs, K., Bloor, S., Kemp, S.D., Van den Eynde, C., Alcorn, T.M., Pauwels, R., Van Houtte, M., Staszewski, S., Miller,V.&Larder,B.A.(2000)Phenotypicandgenotypic analysis of clinical HIV-1 isolates reveals extensive protease in- hibitor cross-resistance: a survey of over 6000 samples. AIDS 14, 1203–1210. 2. Wu, T.D., Schiffer, C.A., Gonzales, M.J., Taylor, J., Kantor, R., Chou, S., Israelski, D., Zolopa, A.R., Fessel, W.J. & Shafer, R.W. (2003) Mutation patterns and structural correlates in human immunodeficiency virus type 1 protease following different pro- tease inhibitor treatments. J. Virol. 77, 4836–4847. 3. Condra, J.H., Holder, D.J., Schleif, W.A., Blahy, O.M., Danov- ich, R.M., Gabryelski, L.J., Graham, D.J., Laird, D., Quintero, J.C., Rhodes, A., Robbins, H.L., Roth, E., Shivaprakash, M., Yang,T.,Chodakewitz,J.A.,Deutsch,P.J.,Leavitt,R.Y.,Mas- sari, F.E., Mellors, J.W., Squires, K.E., Steigbigel, R.T., Teppler, H. & Emini, E.A. (1996) Genetic correlates of in vivo viral resistance to indinavir, a human immunodeficiency virus type 1 protease inhibitor. J. Virol. 70, 8270–8276. 4. Sa-Filho, D.J., Costa, L.J., de Oliveira, C.F., Guimaraes, A.P., Accetturi, C.A., Tanuri, A. & Diaz, R.S. (2003) Analysis of the protease sequences of HIV-1 infected individuals after indinavir monotherapy. J. Clin. Virol. 28, 186–202. 5. Brown, A.J., Korber, B.T. & Condra, J.H. (1999) Associations between amino acids in the evolution of HIV type 1 protease sequences under indinavir therapy. AIDS Res. Hum. Retroviruses 15, 247–253. 6. Zennou,V.,Mammano,F.,Paulous,S.,Mathez,D.&Clavel,F. (1998) Loss of viral fitness associated with multiple Gag and Gag- Pol processing defects in human immunodeficiency virus type 1 variants selected for resistance to protease inhibitors in vivo. J. Virol. 72, 3300–3306. 7. Louis, J.M., Weber, I.T., Tozser, J., Clore, G.M. & Gronenborn, A.M. (2000) HIV-1 protease: maturation, enzyme specificity, and drug resistance. Adv. Pharmacol. 49, 111–146. 8. Gulnik, S.V., Suvorov, L.I., Liu, B., Yu B., Anderson, B., Mits- uya, H. & Erickson, J.W. (1995) Kinetic characterization and cross-resistance patterns of HIV-1 protease mutants selected under drug pressure. Biochemistry 34, 9282–9287. 9. Ermolieff, J., Lin, X. & Tang, J. (1997) Kinetic properties of saquinavir-resistant mutants of human immunodeficiency virus type 1 protease and their implications in drug resistance in vivo. Biochemistry 36, 12364–12370. 10. Ridky, T.W., Kikonyogo, A., Leis, J., Gulnik, S., Copeland, T., Erickson, J., Wlodawer, A., Kurinov, I., Harrison, R.W. & Weber, I.T. (1998) Drug-resistant HIV-1 proteases identify enzyme residues important for substrate selection and catalytic rate. Biochemistry 37, 13835–13845. 11. Xie, D., Gulnik, S., Gustchina, E., Yu B., Shao, W., Qoronfleh, W., Nathan, A. & Erickson, J.W. (1999) Drug resistance mutations can effect dimer stability of HIV-1 protease at neutral pH. Protein Sci. 8, 1702–1707. 12. Mahalingam,B.,Louis,J.M.,Reed,C.C.,Adomat,J.M.,Krouse, J., Wang, Y.F., Harrison, R.W. & Weber, I.T. (1999) Structural and kinetic analysis of drug resistant mutants of HIV-1 protease. Eur. J. Biochem. 263, 238–245. 13. Mahalingam, B., Louis, J.M., Hung, J., Harrison, R.W. & Weber, I.T. (2001) Structural implications of drug resistant mutants of HIV-1 protease: high resolution crystal structures of the mutant protease/substrate analog complexes. Proteins 43, 455–464. 14. Feher, A., Weber, I.T., Bagossi, P., Boross, P., Mahalingam, B., Louis, J.M., Copeland, T.D., Torshin, I.Y., Harrison, R.W. & Tozser, J. (2002) Effect of sequence polymorphism and drug resistance on two HIV-1 Gag processing sites. Eur. J. Biochem. 269, 4114–4120. 15. Mahalingam, B., Boross, P., Wang, Y F., Louis, J.M., Fischer, C., Tozser, J., Harrison, R.W. & Weber, I.T. (2002) Combining mutations in HIV-1 protease to understand mechanisms of resistance. Proteins 48, 107–116. 16. Prabu-Jeyabalan, M., Nalivaika, E.A., King, N.M. & Schiffer, C.A. (2003) Viability of a drug-resistant human immunodeficiency virus type 1 protease variant: structural insights for better antiviral therapy. J. Virol. 77, 1306–1315. 17. Chen, Z., Li, Y., Chen, E., Hall, D.L., Darke, P.L., Culberson, C., Shafer, J.A. & Kuo, L.C. (1994) Crystal structure at 1.9 angstroms resolution of human immunodeficiency virus (HIV) II protease complexed with l-735,524, an orally bioavailable inhibitor of the HIV proteases. J. Biol. Chem. 269, 26344–26348. 18. Chen, Z., Li, Y., Schock, H.B., Hall, D., Chen, E. & Kuo, L.C. (1995) Three-dimensional structure of a mutant HIV-1 protease displaying cross-resistance to all protease inhibitors in clinical trials. J. Biol. Chem. 270, 433–436. 19. Munshi, S., Chen, Z., Yan, Y., Li, Y., Olsen, D.B., Schock, H.B., Galvin, B.B., Dorsey, B. & Kuo, L.C. (2000) An alternate binding site for the P1–P3 group of a class of potent HIV-1 protease inhibitors as a result of concerted structural change in the 80s loop of the protease. Acta Crystallogr. D56, 381–388. 20. King, N.M., Melnick, L., Prabu-Jeyabalan, M., Nalivaika, E.A., Yang,S.S.,Gao,Y.,Nie,X.,Zepp,C.,Heefner,D.L.&Schiffer, C.A. (2002) Lack of synergy for inhibitors targeting a multi-drug- resistant HIV-1 protease. Protein Sci. 11, 418–429. 21. Louis, J.M., Clore, G.M. & Gronenborn, A.M. (1999) Auto- processing of HIV-1 protease is tightly coupled to protein folding. Nat. Struct. Biol. 6, 868–875. 22. Maibaum, J. & Rich, D.H. (1988) Inhibition of porcine pepsin by two substrate analogues containing statine: the effect of histidine at the P2 subsite on the inhibition of aspartic proteinases. J. Med. Chem. 31, 625–629. 23. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac- tion data in oscillation mode. Methods Enzymol. 276, 307–326. 24. Navaza, J. (1994) AMoRe: An automated package for molecular replacement. Acta Crystallogr. D50, 157–163. 25. Sheldrick, G.M. & Schneider, T.R. (1997) High resolution refinement. Methods Enzymol. 277, 319–343. 26. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47, 110–119. 27. Ferro, D.R. & Hermans, J. (1977) A different best rigid-body molecular fit routine. Acta Crystallogr. A33, 345–347. 28. Klabe, R.M., Bacheler, L.T., Ala, P.J., Erickson-Vitanen, S. & Meek, J.L. (1998) Resistance to HIV protease inhibitors: a comparison of enzyme inhibition and antiviral potency. Bio- chemistry 37, 8735–8742. Ó FEBS 2004 HIV protease crystal structures with indinavir (Eur. J. Biochem. 271) 1523 29. Merritt, E.A. (1999) Expanding the model: anisotropic displace- ment parameters in protein structure refinement. Acta Crystallogr. D55, 1109–1117. 30. Hong, L., Zhang, X.C., Hartsuck, J.A. & Tang, J. (2000) Crystal structure of an in vivo HIV-1 protease mutant in complex with saquinavir: insights into the mechanisms of drug resistance. Protein Sci. 9, 1898–1904. 31. Kraulis, P.J. (1991) MOLSCRIPT : a program to produce both detailed and schematic plots of protein structures. J. Appl. Cryst. 24, 946–950. 1524 B. Mahalingam et al. (Eur. J. Biochem. 271) Ó FEBS 2004 . Crystal structures of HIV protease V82A and L90M mutants reveal changes in the indinavir-binding site Bhuvaneshwari Mahalingam 1, *, Yuan-Fang. observed in the crystal structure of the L63P/V82T/I84V mutant with indinavir [20]. Our structures of PR V82A and PR L90M showed opposite changes in the main