Critical differences in HIV-1 and HIV-2 protease specificity for clinical inhibitors

Abstract

Clinical inhibitor amprenavir (APV) is less effective on HIV-2 protease (PR₂) than on HIV-1 protease (PR₁). We solved the crystal structure of PR₂ with APV at 1.5 Å resolution to identify structural changes associated with the lowered inhibition. Furthermore, we analyzed the PR₁ mutant (PR_1M) with substitutions V32I, I47V, and V82I that mimic the inhibitor binding site of PR₂. PR_1M more closely resembled PR₂ than PR₁ in catalytic efficiency on four substrate peptides and inhibition by APV, whereas few differences were seen for two other substrates and inhibition by saquinavir (SQV) and darunavir (DRV). High resolution crystal structures of PR_1M with APV, DRV, and SQV were compared with available PR₁ and PR₂ complexes. Val/Ile32 and Ile/Val47 showed compensating interactions with SQV in PR_1M and PR₁, however, Ile82 interacted with a second SQV bound in an extension of the active site cavity of PR_1M. Residues 32 and 82 maintained similar interactions with DRV and APV in all the enzymes, whereas Val47 and Ile47 had opposing effects in the two subunits. Significantly diminished interactions were seen for the aniline of APV bound in PR_1M and PR₂ relative to the strong hydrogen bonds observed in PR₁, consistent with 15- and 19-fold weaker inhibition, respectively. Overall, PR_1M partially replicates the specificity of PR₂ and gives insight into drug resistant mutations at residues 32, 47, and 82. Moreover, this analysis provides a structural explanation for the weaker antiviral effects of APV on HIV-2.

Introduction

Acquired immunodeficiency syndrome (AIDS) is a major health challenge with a global estimate of over 30 million people infected with HIV and 1.8 million deaths in 2009.¹ Anti-retroviral treatment has increased survival of HIV-infected patients, however, long-term therapy is compromised by the selection of drug resistance mutations and the high genetic diversity of the virus. There are two major types: HIV-1 and HIV-2; the more common HIV-1 is subdivided into the four groups of M, N, O, and P as well as several subtypes. HIV-2 infections account for more than 1 million people, or about a third of the HIV prevalence in West Africa^{2, 3} and are spreading into other continents.^{4, 5} Treatment of HIV-2 infections employs the drugs developed for HIV-1, however, several drugs are less effective on HIV-2.^{2, 3} A further therapeutic quandary is posed by the drug resistant mutations arising in HIV-2 and co-infections of HIV-1 and HIV-2.^{2, 6}

HIV-1 protease (PR₁) is a very effective drug target for AIDS treatment because its activity is essential for hydrolyzing the viral Gag and Gag-Pol precursor polyproteins during the maturation of infectious virus.⁷ PR₁ inhibitors illustrate the success of structure-guided drug designs. Several hundred crystal structures are available for wild type and mutant PR₁ complexes with the clinical drugs and many other inhibitors.⁸ Currently, nine FDA approved PR₁ inhibitors are used in Highly Active Antiretroviral Therapy (HAART). Some of these clinical inhibitors, such as amprenavir (APV) and nelfinavir (NFV), show lower efficacy on HIV-2 infections and weaker inhibition of HIV-2 protease (PR₂).^{2, 7, 9} PR₁ and PR₂ share 39–48% amino acid sequence identity depending on the strain of virus and similar overall structure.³, 10–12 The two enzymes differ in their cleavage site sequences in the viral precursors and in their specificity for peptide substrates and inhibitors, especially at the P2 positions of peptide substrates.^{13, 14} The sequence differences between PR₁ and PR₂ are expected to be responsible for the differences in efficacy of inhibitors and include substitutions observed in resistance of HIV-1 to the current drugs (Fig. 1).¹⁵ In particular, the binding site for clinical inhibitors differs only in the conservative substitution of hydrophobic residues Val32, Ile47, and Val82 in PR₁ by Ile32, Val47, and Ile82 in PR₂. Earlier studies showed that PR₁ bearing the substitutions, V32I, I47V, and V82I, altered the inhibition but not the binding mode of a tripeptide inhibitor.^{16, 17} These residues are the sites of drug resistance mutations V32I, I47V, and various substitutions of Val82 in HIV-1 (Fig. 1).¹⁵

Figure 1.

A: Amino acid sequences of HIV PR₁ (upper line) and PR₂ (lower line). PR₂ residues associated with drug resistance in PR₁ are underlined. Arrows indicate the amino acid differences in the inhibitor binding site Val/Ile32, Ile/Val47, and Val/Ile82. B: Location of residues 32, 47, and 82 (magenta) in PR dimer with SQV (cyan). C: Chemical structures of APV, DRV, and SQV with groups P2-P2′ labeled for APV. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

In contrast to PR₁, very few crystal structures are available for PR₂ complexes with clinical inhibitors. We have shown that DRV, which maintains antiviral potency on HIV-1 and HIV-2 infections, demonstrates similar binding mode in PR₁ and PR₂ crystal structures, as does indinavir (IDV).^{11, 12} Here, we report the crystal structure of PR₂ with APV, which by comparison with our PR₁–APV structure¹⁸ helps explain the lower efficacy of this inhibitor on HIV-2 infections. Furthermore, we constructed the PR₁ mutant with substitutions of the three PR₂ residues that differ in the inhibitor-binding site (V32I, I47V, and V82I; designated PR_1M) to investigate the importance of these residues in the substrate specificity and binding of clinical inhibitors. The inhibitors APV, DRV and SQV were selected due to their distinct effects on the two types of virus. HIV-2 strains were shown to be susceptible to DRV¹⁹ and to SQV,^{20, 21} while natural resistance to APV was found for several HIV-2 strains.20–22 Thus, crystallographic and kinetic analysis of PR_1M, PR₁ and PR₂ will improve our understanding of the differences in inhibitor potency. Furthermore, this knowledge can be exploited in the design of broader-spectrum inhibitors targeting the natural variants of PR₁, PR₂ and their drug resistant mutants.

Results

Substrate specificity and inhibition

The three enzymes were assessed for hydrolysis of peptides representing natural cleavage sites of HIV-2 precursor polyproteins. Also, peptides were tested with variants of the P2 and P4 positions of the HIV-1 MA-CA cleavage site (between the MA and CA proteins in the precursor) that distinguish the substrate specificities of retroviral PRs.^{14, 23} Two peptides represent the HIV-2 cleavage sites CA/p2 (KARLM↓AEALK, where ↓ indicates the position of the cleaved peptide bond) and p2/NC (IPFAA↓AQQRK). Four peptides were selected with different amino acids (Val and Leu) at the P2 and P4 positions in the HIV-1 MA/CA cleavage site (VSQNY↓PIVQ) to explore the variation due to the substitutions of residues 32, 47, and 82 that differ in the substrate binding cavities of PR₁ and PR₂ (Fig. 1). Kinetic parameters are summarized in Table I. The K_m values showed low variation ranging from 0.07 to 0.57 mM over the measured PRs and substrates. Significant differences were observed only for hydrolysis of the HIV-2 p2/NC peptide where the K_m values were identical for PR_1M and PR₂ and 4-fold lower than that for PR₁. The k_cat values for different substrates exhibited a wider range covering two orders of magnitude from 0.06 to 5.1 s⁻¹. Values of k_cat for PR_1M were closer to those of PR₂ for substrates (2), (4), and (6), however, the three PRs had k_cat values in the same range for substrates (1) and (3).

Table I.

Kinetic Parameters Obtained for Oligopeptide Substrates

	Peptide	Protease	Km (mM)	kcat (s−1)	kcat/Km (mM−1s−1)	Relative kcat/Km
(1) HIV-2 CA/p2	KARLM↓AEALK	PR1a	0.13	0.3	2.3	1.0
		PR2a	0.08	1.2	15.0	6.5
		PR1M	0.09	0.6	6.6	2.9
(2) HIV-2 p2/NC	IPFAA↓AQQRK	PR1a	0.28	0.3	1.1	1.0
		PR2a	0.07	0.8	11.4	10.4
		PR1M	0.07	1.2	16.3	14.8
(3) HIV-1 MA/CA	VSQVY↓PIVQ	PR1	0.24	2.6	10.8	1.0
		PR2b	0.43	2.5	5.8	0.5
		PR1M	0.35	3.0	8.6	0.8
(4) HIV-1 MA/CA	VSQLY↓PIVQ	PR1b	0.12	0.4	3.3	1.0
		PR2b	0.17	3.4	20.0	6.1
		PR1M	0.38	5.1	8.6	2.6
(5) HIV-1 MA/CA	VVQNY↓PIVQ	PR1	0.19	0.16	0.8	1.0
		PR2c	N.D.	N.D.	0.6	0.8
		PR1M	0.23	0.18	0.8	1.0
(6) HIV-1 MA/CA	VLQNY↓PIVQ	PR1c	0.40	0.06	0.2	1.0
		PR2c	0.57	0.6	1.0	5.0
		PR1M	0.32	0.26	0.8	4.0

^aData taken from Reference³⁵.

^bData taken from Reference¹³.

^cData taken from Reference⁴⁸.

Residues that were substituted in the HIV-1 MA/CA cleavage site VSQNY↓PIVQ are in bold and underlined.

The k_cat/K_m value is the most useful for comparing the specificity of different enzymes. PR₂ had significantly greater catalytic efficiency by 5- to 10-fold compared to PR₁ for substrates (1), (2), (4), and (6). Similarly, PR_1M mutant showed 2.6–15-fold higher k_cat/K_m values relative to PR₁. PR_1M and PR₂ shared similar k_cat/K_m values for substrate (2) representing the HIV-2 p2/NC site and substrate (6). However, PR_1M showed k_cat/K_m values intermediate between the values for PR₁ and PR₂ for substrates (1) and (4). The three enzymes had indistinguishable k_cat/K_m values (less than two-fold difference) for substrates (3) and (5). The PR_1M mutant showed no preference for the Val or Leu at P2 in peptides (3) and (4), although P2 Leu was favored by PR₂ and Val by PR₁. Leu at the P4 position was preferred over Val in peptides (5) and (6) for PR₂ and PR_1M, although the opposite trend was observed for PR₁. Overall, the kinetic parameters of PR_1M mutant were more comparable to those of PR₂ rather than PR₁.

The catalytic activities of the HIV proteases were inhibited by APV, DRV and SQV (Table II). The results for PR₁ and PR₂ are similar to those reported using a chromogenic substrate based on the HIV-1 CA-p2 site.⁹ APV exhibited good inhibition for PR₁ while PR_1M and PR₂ were more poorly inhibited with K_i values of 15 and 19-fold higher than for PR₁, respectively. DRV showed similar inhibition for PR₁ and PR₂, and 6-fold weaker inhibition of the mutant. SQV showed similar inhibition constants for all three enzymes with differences of no more than 2-fold. Both SQV and DRV retain antiviral potency for HIV-2 and HIV-1 infections. Overall, APV is the least effective among three inhibitors for PR_1M and PR₂, which is consistent with the observations of natural resistance of HIV-2 to APV, and the 10-30 fold higher IC₅₀ values of APV for HIV-2 infected compared to HIV-1 infected cells.^{3, 21}

Table II.

K_i Values for Inhibitors (nM)

	DRV	SQV	APV
PR1	0.04	0.11	0.17
PR1M	0.23 (6)	0.24 (2)	2.53 (15)
PR2	0.04 (1)	0.07 (0.6)	3.24 (19)

Values relative to PR₁ are shown in parentheses.

Crystallographic analysis

Crystal structures were solved for the PR₂ with APV and of PR_1M complexes with clinical inhibitors DRV, SQV, and APV (Table III). The asymmetric units contained one PR dimer with residues numbered 1–99 and 1′–99′. Three datasets were collected with high resolutions of 1.26–1.51 Å and were refined to final R-factors of 0.16–0.18. The PR_1M-SQV complex had the lowest resolution of 1.88 Å and was refined to an R-factor of 0.19. Clear electron density was observed for all the residues in PR_1M and PR₂, inhibitor, solvent molecules and ions in all structures. The APV complex showed a single conformation of inhibitor in both PR_1M and PR₂, although the P2′ aniline group had two alternate positions in PR_1M (Fig. 2). Two alternate orientations of DRV were refined in the active site cavity of PR_1M-DRV, as described previously.²⁴ An extra SQV molecule was observed in an extension of the usual binding site in the structure of PR_1M-SQV, as described later. The crystal structure of PR_1M-APV, obtained from crystals grown in potassium iodine solution, was refined with 21 iodide ions, while one sodium and two chloride ions were observed in the PR_1M-DRV crystals grown with sodium chloride as precipitant. Iodide ions were identified by the high peaks in electron density maps, even at partial occupancy, abnormal B factors, and van der Waals contacts of 3.4–3.8 to nitrogen atoms, as noted in Reference¹⁸. The solvent in the PR₂-APV structure comprised one sodium, eight chloride, seven zinc ions, and six imidazole molecules from the crystallization solution, as described for PR₂-DRV.¹²

Table III.

Crystallographic Data Collection and Refinement Statistics

Structures	PR1M-DRV	PR1M–SQV	PR1M-APV	PR2-APV
Space group	P21212	P212121	P21212	C2
Unit cell parameters
a (Å)	58.6	29.2	58.4	106.0
b (Å)	86.2	67.4	86.6	31.0
c (Å)	46.2	92.8	46.3	56.2
β (°)	90.0	90.0	90.0	91.66
Resolution range (Å)	50–1.42	50–1.88	50–1.26	50–1.51
Unique reflections	45,155	15,355	58,771	25,917
Rmerge (%) overall (final shell)	8.4 (33.6)	10.8 (40.8)	6.3 (36.1)	8.1 (25.5)
I/σ(I) overall (final shell)	16.2 (3.0)	13.7 (2.2)	14.5 (2.1)	12.8 (2.9)
Completeness (%) overall (final shell)	90.9 (50.5)	98.3 (87.5)	91.3 (59.9)	90.3 (63.3)
Data range for refinement (Å)	10–1.42	10–1.88	10–1.26	10–1.51
R (%)	15.7	18.8	15.9	18.2
Rfree (%)	20.8	23.4	19.1	23.8
No. of solvent atoms (total occupancies)	178 (138.3)	100 (83.0)	162 (103.5)	144 (124.7)
RMS deviation from ideality
Bonds (Å)	0.010	0.032	0.012	0.008
Angle distance (Å)	0.030	2.388*	0.031	0.027
Average B-factors (Å2)
Main-chain atoms	17.8	17.3	17.0	17.7
Side-chain atoms	24.3	22.5	23.8	23.0
Inhibitor	15.6	19.5	17.5	20.8
Solvent	27.9	20.3	24.6	24.6

^*RMS deviation in °

Figure 2.

Omit electron density map for APV and mutated residues Ile32, Val47, and Ile82 in the two subunits of the PR_1M dimer. The map was contoured at 2.2σ level. Ile82 and 82′ in the S1/S1′ subsites interact with P1′ and P1 groups of the inhibitor. Ile32/32′ and Val47/47′ contribute to the S2/S2′ subsites and interact with the P2 and P2′ groups of APV. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Second binding site for SQV in PR_1M-SQV

The structure of PR_1M in complex with SQV contained an extra SQV molecule (designated SQV-B) bound in an extension of the regular inhibitor binding pocket. This second SQV binding site has not been reported before. The P3 quinoline of SQV-B interacted with the aromatic P3 and P1 groups of SQV-A bound in the usual position in the active site cavity [Fig. 3(A)]. Hence, the SQV molecules alternate their orientation continuously though the adjacent binding pockets of the PR dimers packed in the crystal lattice [Fig. 3(B)]. Half of SQV-B interacted with the PR_1M dimer while the other half interacted with two symmetry related dimers [Fig. 3(C)]. SQV-B formed a hydrogen bond with the side chain of Glu34′, a water-mediated interaction with the side chain of Arg8′, and hydrophobic interactions with Leu10′, Glu21′, Leu23′, Pro81′, and Ile82′. The residues Arg8, Pro81, Ile82, Gly48′, and Phe53′ from symmetry dimer 1 and Trp42, Pro44 from symmetry dimer 2 formed a binding site for the other part of SQV-B. Two direct hydrogen bonds were formed between SQV-B oxygen atoms and the side chain of Arg8 of symmetry dimer 1 and a network of two water molecules linked it to Asn83. No hydrogen bond interactions were observed between SQV-B and symmetry related dimer 2. The minimal interactions of SQV-B with three symmetry related dimers suggested that it was likely a result of crystal packing. Interestingly, a second DRV binding site in a different location on the flap on one subunit was reported in HIV-1 PR_V32I and PR_M46L complexes²⁵ in the same space group and similar cell parameters. The SQV-B interacts closely with mutated Ile82 in PR_1M, which raises the question of whether SQV can bind to a similar second site in PR₂. The majority of residues contacting SQV-B are identical in PR₁ and PR₂ [Fig. 1(A)], however, the hydrogen bond interaction of SQV-B with the side chain of Glu34 cannot occur for Ala34 in PR₂, which lowers the probability of SQV binding at the equivalent site.

Figure 3.

SQV-B in the extended binding site of PR_1M. A: Omit map for SQV-A and SQV-B. SQV-B interacts with SQV-A in an extension of the regular active site cavity. B: SQV-B molecule is surrounded by three PR_1M dimers in the crystal lattice, colored in green, cyan, and magenta. The SQV-A occupying the regular binding pocket is colored by atom type, whereas the extra SQV-B is shown in red. The arrows represent the alternating orientations of SQV-A and SQV-B molecules. C: SQV-B interactions with PR_1M dimer, symmetry-related dimer 1, and symmetry-related dimer 2. Mutated residues are in pink surface representation, SQV-A is in golden color, and the surfaces of the residues involved in polar or hydrophobic interactions are shown in green and blue, respectively. Water molecules are shown as red spheres. Hydrogen bonds are indicated by red dotted lines. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Comparison of PR-inhibitor structures

The interactions of the individual inhibitors were analyzed in their complexes with PR_1M, wild-type PR_1, and PR₂ to understand the structural effects of the three mutations. The following structures were chosen for the comparison: PR₁-DRV (2IEN at 1.30 Å resolution),²⁴ PR₁-APV (3NU3 at 1.02 Å resolution),¹⁸ PR₁-SQV (2NMY at 1.10 Å resolution),²⁶ and PR₂-DRV (3EBZ at 1.20 Å resolution).¹² The PR_1M and PR₁ dimers are almost identical with small RMSDs of ∼0.2 Å for the pairs of structures in the same space group and unit cell, whereas larger differences of 0.7–0.8 Å were apparent between structures with non-isomorphous unit cells. The largest difference of 1.1 Å was shown by PR₂-DRV and PR₂-APV structures compared with the corresponding PR_1M complexes, due in part to the different space groups. The majority of interactions with each inhibitor were conserved in the three enzymes. Therefore, the detailed structural analysis focusing on differences around residues 32, 47, and 82 is described separately for each inhibitor.

SQV complexes with PR_1M and PR₁

The PR_1M and PR₁ structures superimposed with the relatively large RMSD of 0.67 Å with the largest shifts of 2–3 Å for the surface residues 35–40. No suitable crystals were obtained for PR₂ with SQV. Comparison of the SQV interactions is complicated by the second SQV-B found in PR_1M (Fig. 3). SQV-A was bound within subsites S3 to S2′ of PR_1M and showed almost identical hydrogen bond interactions to those described for PR₁-SQV.²⁶ The minor exception was that SQV had a shorter hydrogen bond to the main chain carbonyl oxygen of Gly27 in PR_1M with a distance of 3.2 Å compared to the weaker interaction with longer separation of 3.6 Å for PR₁. A similar shorter interaction was described in the SQV complex with another mutant PR_L76V,²⁷ however, the interaction in PR_1M may be influenced by the aromatic-aromatic interactions of the adjacent P1 and P3 groups of SQV-A with the quinoline rings of SQV-B [Fig. 3(A)].

The hydrophobic interactions at the sites of the three mutations are illustrated in Figure 4. The larger side chain of Ile32 added two hydrophobic interactions with the P2 Asn group of SQV-A in PR_1M, whereas both Ile47 and Val47 formed hydrophobic contacts with the P2 group [Fig. 4(A)]. Slight differences were seen in the other subunit resulting in one less hydrophobic contact of Ile32′ with the t-butyl group at P2′ of SQV-A in the mutant compared to the wild-type PR [Fig. 4(B)]. Ile82 in PR_1M formed more contacts with the P1′ decahydroisoquinoline ring compared to Val82 in PR₁ [Fig. 4(C)]. In the other subunit, the methyl of the Ile82′ has shifted to make van der Waals contacts with the benzene ring at P1 of the second SQV-B [Fig. 3(D)]. The quinoline ring of SQV-A adjusted by about 15 degrees rotation to fit the SQV-B molecule. Pro81′ had shifted to interact with the P3 group of SQV-B while retaining hydrophobic contacts with the P1 and P3 groups of SQV-A [Fig. 4(D)]. Overall, the structural adjustment of the triple mutant to accommodate inhibitor binding was consistent with the similar inhibition constants observed for PR_1M and PR₁ with SQV of 0.24 and 0.11 nM, respectively.

Figure 4.

Comparison of PR₁ and PR_1M interactions with SQV. The PR₁ structure is colored by atom type, while that of the triple mutant is shown in green bonds. Only the major conformation is shown for SQV and protein residues with alternate conformation in the PR₁-SQV structure. SQV-A is the molecule bound in the regular active site cavity. Hydrophobic interactions (distances of 3.3–4.2 Å between non-hydrogen atoms) are indicated as dashed lines. Interactions with aromatic groups (CH…π) are indicated by dashed arrows. (A) Residues 32 and 47 of subunit A; (B) Residues 32′ and 47′ of subunit B; (C) Residue 82; (D) Residue 82′. SQV-B (cyan bonds) indicates the extra SQV molecule. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

DRV complexes with PR_1M, PR_1, and PR₂

DRV was bound in the active site cavity of PR_1M in two alternate conformations with relative occupancy of 0.55/0.45 in the active site cavity. This structure was very similar to that of wild type PR₁-DRV in the same space group with an overall RMSD on Cα atoms of only 0.16 Å. The PR_1M-DRV complex superimposed on PR₂-DRV (space group C2) with RMSD value of 1.1 Å, due to a large shift of 5–6 Å around residues 38–40 and 38′–40′, as described for comparison of the PR₁-DRV and PR₂-DRV crystal structures.¹² Notably, the PR₂-DRV crystal structure contained zinc ions and imidazole from the crystallization solution, although these solvent molecules were not close to the regions with large deviations.

The PR-DRV hydrogen bond interactions were essentially identical in the three enzymes. The minor exception was a slightly longer (3.4–3.5 Å) hydrogen bond interaction between one bis-THF oxygen of DRV and the amide of Asp30′ in PR_1M compared with distances of 3.1–3.3 Å in the other structures. Structural changes at the mutation sites are illustrated in Figure 5. Only the major conformation of DRV is shown for PR₁-DRV. Ile32 and 32′ in PR_1M had alternate conformations of the side chains with relative occupancies of 0.6/0.4 and 0.5/0.5, respectively. The side chains of Val32 and the minor conformation of Ile32 exhibited hydrophobic interactions with the aniline ring of DRV. Val47 in PR_1M had no contact with DRV, although Ile47 made two hydrophobic interactions with the aniline in PR₁ [Fig. 5(A)]. In the other subunit, the minor conformation of Ile32′ in PR_1M and Val32′ in PR₁ formed similar hydrophobic contacts with the bis-THF group, while the side chain of Val47′ rotated to make hydrophobic contacts with the bis-THF group unlike Ile47′ of PR₁ [Fig. 5(B)]. In comparison, the single mutant PR_V32I-DRV²⁵ showed similar interactions to those of the minor conformation of Ile32 in PR_1M. The side chain of Ile82 rotated to form comparable hydrophobic interactions with the P1 group of DRV as seen for Val82 in PR₁ [Fig. 5(C)]. Also, Ile82′ and Val82′ formed similar contacts with DRV in all structures.

Figure 5.

Structural comparison of DRV complexes with PR_1M, PR_1, and PR₂. The PR₁ structure is colored by atom type, PR_1M is shown in green bonds with the minor conformations of Ile32 and 32′ in cyan, and PR₂ is colored magenta. Only the major conformation of DRV is shown for PR₁-DRV. Hydrophobic interactions are indicated as dashed lines. CH…π interactions are indicated by dashed arrows. PR_1M is compared with PR₁ in the upper panels and with PR₂ in the lower panels. A: Residues 32 and 47 interactions with P2′ aniline of DRV; B: Residues 32′ and 47′ interactions with bis-THF at P2; and C: Residue 82 interactions with P1 of DRV. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

In comparison of the PR₂ and PR_1M-DRV complexes, the side chain conformations of residues 32, 47, and 82 were preserved, except that the minor conformations of Ile32 and 32′ in the mutant most closely resembled the single conformation in PR₂. Ile32 formed CH…π interactions with the aniline ring of DRV in both complexes. The majority of the hydrophobic contacts for residues 32 and 82 were essentially identical in PR₁, PR_2, and PR_1M. Differences were observed in the DRV interactions of residue 47 [Fig. 5(B)]. Val47 had no hydrophobic interactions with the aniline of DRV in PR₂ and PR_1M, whereas two contacts were seen in PR₁. The opposite was seen for the other subunit, where Val47′ showed three van der Waals contacts with bis-THF in PR₂ and PR_1M and no contacts in PR₁. Overall, the slightly longer hydrogen bond interaction of the triple mutant with DRV was the only change that could help explain the 6-fold poorer inhibition of PR_1M compared to PR₂ and PR₁.

APV complexes with PR_1M, PR_1, and PR₂

The APV complex of PR_1M was very similar in overall conformation to PR₁-APV as shown by the low RMSD of 0.23 Å, whereas it had larger differences of 1.1 Å with the new structure of PR₂-APV solved in the C2 space group. APV was seen in a single orientation in PR₂ and PR_1M, although the aniline group in PR_1M-APV had two alternate conformations with an occupancy ratio of 0.6/0.4, as described for the PR_V82A-DRV structure.²⁴ Most of the PR₂ and PR_1M interactions with inhibitor were very similar to those of the major (0.7) occupancy conformation in the structure of PR₁ with APV,¹⁸ with the exception of the interactions of the P2′ aniline group [Fig. 6(A)]. In PR₂ and PR_1M, Ile32 formed CH…π interactions with the aniline ring of APV, as occurred for the DRV complexes. In PR_1M-APV, the major conformation of the aniline group retained the hydrogen bond interactions seen in the wild type complex with distances of 3.0–3.1 Å, however, no van der Waals contacts were formed with Val47 unlike the favorable hydrophobic interactions with Ile47 seen in the wild type PR₁. The rotation of the minor conformation of the aniline restored favorable van der Waals interactions with Val47, concomitant with loss of the hydrogen bond interactions with the carbonyl oxygen and amide of Asp30 in PR_1M (interatomic distances of more than 4.0 Å). In the other subunit, the THF group of APV retained van der Waals contacts with residues 32′ and 47′ in the three structures [Fig. 6(B)]. The hydrophobic interactions of Ile32 with APV are similar in PR₂, PR_1M and the single mutant PR_V32I.¹⁸ The side chain of Ile82 formed good hydrophobic interactions with the aromatic ring of APV similar to those in the DRV complexes [Fig. 5(C)]. Residue 82′ in the other subunit showed similar hydrophobic contacts with APV in all structures, although Ile82′ is closer to APV in the PR₂ complex [Fig. 6(C)]. In PR_1M-APV, the most significant differences arose from the alternate conformations of the aniline ring of APV, which were accompanied by the loss of hydrophobic contacts with Val47 for the major conformation and the loss of two hydrogen bonds for the minor conformation. The loss of these interactions with APV is likely to contribute to the 15-fold weaker inhibition by APV observed for the PR_1M compared with PR₁.

Figure 6.

Comparison of PR₁, PR_2, and PR_1M interactions with APV. The PR₁ structure is colored by atom type, PR_1M is shown in green bonds with the minor conformations of Ile32 and 32′ in cyan, and PR₂ is colored magenta. Only the major (0.7 occupancy) conformation is shown for APV and PR₁ residues in the wild type complex. Hydrogen bond interactions are shown as dotted lines, hydrophobic interactions are indicated as dashed lines, and CH…π interactions are indicated by dashed arrows. A: Asp30, residues 32 and 47 interactions with aniline of APV. The aniline group and Asp30 have two alternate conformations in PR_1M -APV. The major conformations closely resemble the wild type structure with strong hydrogen bond interactions. The minor conformation (in cyan) has only one hydrogen bond and more hydrophobic interactions with Val47. B: Residue 32′ and 47′ interactions with the THF group of APV. C: Residue 82′ interactions with the P1′ group of APV. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

The crystal structure of PR₂-APV reveals notable changes in the hydrogen bond interactions of the aniline NH₂ with Asp30 where two interatomic distances are significantly lengthened by about 0.5 Å to 3.6 and 3.7 Å (Fig. 7), which is outside of the normal range of 2.6–3.2 Å for strong hydrogen bonds.²⁸ The elongated hydrogen bond is associated with a 1.2 Å shift of the carbonyl oxygen of Asp30 in the PR₂-APV complex relative to its position in the PR₁-APV complex. The altered interactions of the aniline group are consistent with the observed 19-fold weaker inhibition of PR₂ relative to PR₁.

Figure 7.

Hydrogen bond interactions of P2 and P2′ groups in PR₁ and PR₂ complexes with APV (A) and DRV (B). PR₁ complexes are shown in grey and PR₂ complexes in green bonds. The inhibitors are shown with Asp29 and 30. Dotted lines indicate hydrogen bonds with distances up to 3.4 Å, and broken lines indicate longer distances. The red arrow indicates the 1.2 Å shift of the carbonyl group of Asp30. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Discussion

Recognition and selectivity in enzyme/substrate and enzyme/inhibitor systems are a complex combination of strong local effects and weaker, but still significant, distal effects. This is clearly demonstrated in our results where the three mutations in PR_1M which were chosen to mimic the inhibitor binding site of PR₂ changed the specificity and inhibition from the PR₁ toward the PR₂ values, but did not match them exactly. The relative k_cat/K_m values in Table I show that the specificity of the PR_1M enzyme roughly mirrors that of PR₂, and the K_i values in Table II reproduce the trend seen in PR₂, although the inhibition is measurably different for DRV. The biochemical similarities of PR_1M and PR₂ demonstrate the importance of residues 32, 47, and 82 for recognition. The differences between PR_1M and PR₂ are assumed to arise from distal effects due to over 50% difference in sequence (Fig. 1). Previously, differences in residues 31–37 were shown to make a major contribution to the inhibition and unusual mode of binding of a tripeptide analog in studies of chimeric enzymes.¹⁷ In fact, residues 33–44 vary in sequence and conformation among different groups and subtypes of PR₁ although little effect on inhibition has been reported.29–31

Structurally, PR_1M resembles PR₁ more than PR₂. However, PR_1M and PR₂ share very similar side chain conformations for Ile32, Val47, and Ile82 reflecting the importance of local effects for recognition. Our crystallographic analysis of the weak inhibitor APV with PR₂ and the PR_1M mutant representing the inhibitor binding site of PR₂ helps explain the lower antiviral effectiveness of APV in HIV-2 infections compared to DRV or SQV. DRV and APV are chemically related and differ only in their P2 groups [Fig. 1(C)]. The bis-THF P2 group of DRV bears two oxygens and forms four hydrogen bonds with Asp29′ and Asp30′, while the THF group in APV forms only two hydrogen bonds²⁴(Fig. 7). Despite these differences, the P2 groups of APV and DRV retain similar interactions in PR₁ and PR₂. In contrast, the P2′ group at the opposite end of DRV maintains similar interactions in the enzymes, whereas the identical P2′ aniline of APV shows differences [Figs. 6(A) and 7]. The altered interactions of the P2′ aniline group of APV appear to reflect less stable anchoring of the P2 THF group relative to the larger bis-THF of DRV in the binding sites of PR₁ and PR₂

These new structures show how drug resistant mutations of V32I, I47V, and V82I can alter the interactions with the P2 and P2′ groups of inhibitors. The diminished interactions of APV with Asp30 and Val47 in PR_1M and PR₂ are consistent with the observed 15- and 19-fold weaker inhibition, respectively, relative to PR₁, and provide a structural explanation for the low antiviral potency on HIV-2 infections. In contrast, few structural changes were apparent for DRV and SQV complexes consistent with their sub-nanomolar inhibition of both enzymes and equivalent antiviral potency on HIV-1 and −2.³ Therefore, this analysis suggests a strategy for improved inhibitors of HIV-2 and drug resistant HIV-1 by introducing P2 or P2′ groups with enhanced interactions in the S2 and S2′ subsites.

Materials and Methods

Preparation of proteases

The optimized HIV-1 PR clone with mutations Q7K, L33I, and L63I to diminish the autoproteolysis of the PR₁, as well as mutations C67A and C95A to prevent cysteine-thiol oxidation was used as the initial template for adding mutations.³² This optimized PR₁ had almost identical kinetic parameters and stability as the mature PR. Plasmid DNA (pET11a, Novagen, Gibbstown, NJ) encoding PR₁ was utilized to construct mutant PR_1M with substitutions V32I/I47V/V82I by the Quick-Change mutagenesis kit (Stratagene, Santa Clara, CA). The PR_1M mutant was expressed in Escherichia coli BL21 (DE3) and the protein was purified from inclusion bodies as described.³³ The presence of the appropriate mutations was confirmed by DNA sequencing. PR₂ was prepared as described.³⁴

Enzyme kinetic assays

Assays were performed at 37°C using purified PRs and chemically synthesized oligopeptides. The reaction was initiated by the mixing of 5 μL (0.05–8 μM) purified wild-type or mutant PR with 10 μL incubation buffer [0.5 m potassium phosphate buffer, pH 5.6, containing 10% glycerol, 2 mM ethylenediaminetetraacetic acid (EDTA), 10 mM dithiothreitol, 4 m NaCl] and 5 μL 0.5–7 mM substrate. The reaction mixture was incubated at 37°C for 1 h and terminated by the addition of 180 μL 1% trifluoroacetic acid. Substrates and the cleavage products were separated using a reversed-phase HPLC (High-performance liquid chromatography) method described previously.³² Kinetic parameters were determined by fitting the data obtained at less than 20% substrate hydrolysis to the Michaelis–Menten equation using SigmaPlot 8.02 (San Jose, CA). The standard errors of the kinetic parameters were below 20%.

Active site titration of PR with SQV, APV, and DRV

The amount of active and correctly folded enzyme used in the assays was determined by active site titration using the PR₁ inhibitor DRV. Active site titrations were performed by using the HPLC method with substrate VSQLYPIVQ (peptide 4) as described,³⁵ except that 0.2 μL aliquot of the inhibitor (0–10 μM in dimethylsulfoxide) was added to the reaction mixture. K_i values were obtained from the IC₅₀ values estimated from an inhibitor dose-response curve using the equation K_i = (IC₅₀−[E]/2)/(1 + [S]/K_m), where [E] and [S] are the PR and substrate concentrations, respectively.³⁶

Crystallographic analysis

Crystals were grown at room temperature by the hanging drop vapor diffusion method. The protein (about 3.5 mg/mL) was preincubated with the clinical inhibitors at a molar ratio of 1:5. Each crystallization drop contained 1 μL protein and 1 μL reservoir solution. Crystals of suitable size for diffraction were obtained within 1 week. PR₂-APV crystals grew from 1.5 m NaCl with 0.6 m imidazole/0.12 m zinc acetate buffer at pH 6. For PR_1M-SQV, 0.1 m sodium acetate buffer, pH 5.0, 0.4 m potassium chloride as precipitant; for PR_1M-APV, 0.1 m sodium citrate, phosphate buffer, pH 5.4, 4% dimethyl sulfoxide (DMSO) and 0.175 m potassium iodine as precipitant; For PR_1M-DRV, the crystal was grown from 0.1 m sodium acetate buffer, pH 4.6 and 2M NaCl as precipitant. Crystals were cryo-cooled in liquid nitrogen after soaking in 30% glycerol to prevent freezing.

X-ray diffraction data for all the complexes were collected on the SER-CAT 22ID beamline of the Advanced Photon Source, Argonne National Laboratory (Argonne, IL). Data were processed using HKL-2000.³⁷ The structures were solved by molecular replacement based on our published structures: PR₂-DRV (3EBZ), PR₁-SQV (2NMW), PR_D30N-GRL98065 (2QCI), and PR₁-DRV (2IEN) using AMoRe³⁸ in CCP4i.^{39, 40} The lowest resolution structure of PR_1M-SQV was refined using Refmac5 and isotropic B factors.⁴¹ The other structures were refined by SHELX-97.⁴² Structures were refitted using O⁴³ and COOT.⁴⁴ Alternate conformations for residues were modeled according to the electron density maps. Anisotropic B factors were refined and hydrogen atom positions were included in the last stage of refinement for the structures at better than 1.5 Å resolution. Structural figures were made using Bobscript^{45, 46} and PyMOL.⁴⁷

Protein data bank accession numbers

The atomic coordinates and structure factors were deposited in the Protein Data Bank with accession codes: 3S56 for PR_1M-SQV, 3S54 for PR_1M-DRV (P2₁2₁2), 3S43 for PR_1M-APV, and 3S45 for PR₂-APV.

Glossary

Abbreviations:

AIDS

acquired immunodeficiency syndrome

APV

amprenavir

DRV

darunavir

HAART

Highly Active Antiretroviral Therapy

HIV-1

human immunodeficiency virus type 1

HIV-2

human immunodeficiency virus type 2

IDV

indinavir

PI

protease inhibitor

PR

HIV protease

PR₁

HIV-1 protease

PR₁M

HIV-1 protease with V32I, I47V and V82I mutations

PR₂

HIV-2 protease

RMSD

root mean square deviation

SQV

saquinavir

THF

tetrahydrofuran