Unique Thermodynamic Response of Tipranavir to Human Immunodeficiency Virus Type 1 Protease Drug Resistance Mutations

Abstract

Drug resistance is a major problem affecting the clinical efficacy of antiretroviral agents, including protease inhibitors, in the treatment of infection with human immunodeficiency virus type 1 (HIV-1)/AIDS. Consequently, the elucidation of the mechanisms by which HIV-1 protease inhibitors maintain antiviral activity in the presence of mutations is critical to the development of superior inhibitors. Tipranavir, a nonpeptidic HIV-1 protease inhibitor, has been recently approved for the treatment of HIV infection. Tipranavir inhibits wild-type protease with high potency (K_i = 19 pM) and demonstrates durable efficacy in the treatment of patients infected with HIV-1 strains containing multiple common mutations associated with resistance. The high potency of tipranavir results from a very large favorable entropy change (−TΔS = −14.6 kcal/mol) combined with a favorable, albeit small, enthalpy change (ΔH = −0.7 kcal/mol, 25°C). Characterization of tipranavir binding to wild-type protease, active site mutants I50V and V82F/I84V, the multidrug-resistant mutant L10I/L33I/M46I/I54V/L63I/V82A/I84V/L90M, and the tipranavir in vitro-selected mutant I13V/V32L/L33F/K45I/V82L/I84V was performed by isothermal titration calorimetry and crystallography. Thermodynamically, the good response of tipranavir arises from a unique behavior: it compensates for entropic losses by actual enthalpic gains or by sustaining minimal enthalpic losses when facing the mutants. The net result is a small loss in binding affinity. Structurally, tipranavir establishes a very strong hydrogen bond network with invariant regions of the protease, which is maintained with the mutants, including catalytic Asp25 and the backbone of Asp29, Asp30, Gly48 and Ile50. Moreover, tipranavir forms hydrogen bonds directly to Ile50, while all other inhibitors do so by being mediated by a water molecule.

Previously, we have studied the binding thermodynamics of all protease inhibitors in clinical use to wild-type (WT) human immunodeficiency virus type 1 (HIV-1) protease and their response to the most common mutations (10, 25, 26, 37-39). In this paper, we report the response of tipranavir (TPV). TPV is a recently approved nonpeptidic protease inhibitor of HIV-1 for highly treatment experienced patients or those infected with HIV strains resistant to multiple protease inhibitors (9, 10). It belongs to the class of 4-hydroxy-5,6-dihydro-2-pyrone sulfonamides, a chemical scaffold different from that of other protease inhibitors. TPV was developed with the goal of overcoming broad protease inhibitor cross-resistance and has demonstrated excellent antiviral activity against HIV-1 clinical isolates resistant to other protease inhibitors (1, 16, 30).

Early protease inhibitors have generally been optimized against the WT enzyme and tend to be less potent against mutant forms of the protease. A desired characteristic for superior next-generation inhibitors is to exhibit high potency against the WT and to lose as little affinity as possible when confronting mutant forms of the protease. At the thermodynamic level, extremely high affinity is achieved when both the enthalpy and entropy changes contribute favorably to binding. On the other hand, a small loss in binding affinity against a mutant protease may originate from three possible mechanisms: (i) a small loss in both binding enthalpy and binding entropy, (ii) a loss in binding enthalpy partially compensated for by a gain in binding entropy, or (iii) a loss in binding entropy partially compensated for by a gain in binding enthalpy. The latter two mechanisms represent alternative forms of enthalpy/entropy compensation.

In previous publications, we have reported the mechanism of two experimental inhibitors that exhibit low susceptibility to mutations (KNI-764 and TMC-126) (25, 35). Both inhibitors were able to maintain high affinity towards the mutations by compensating for losses in binding enthalpy with actual gains in binding entropy. Analysis of the crystal structures of KNI-764 with the WT and the resistant mutant V82F/I84V (35) revealed that the presence of a hydrophobic asymmetric functionality joined to the rest of the scaffold by rotatable bonds allowed the inhibitor molecule to bury itself more deeply into the mutant protease, thereby gaining desolvation entropy. An analog of this inhibitor lacking the adaptable moiety failed to bind deep into the pocket and lost an additional order of magnitude of binding affinity (35). In this paper, we show that TPV also maintains high affinity towards a broad range of mutated enzymes but by an opposite mechanism: compensating for entropy losses by actual enthalpy gains or by limiting enthalpy losses. A thorough characterization of the mechanisms by which TPV achieves extremely high affinity and responds to mutations is of fundamental importance. Here, we present a thermodynamic characterization of the response of the protease inhibitors TPV, atazanavir (ATV), lopinavir (LPV), amprenavir (AMP), indinavir (IDV), and darunavir (DRV) to the common drug-resistant mutants V82F/I84V and I50V, the multidrug-resistant mutant MDR-HM (L10I/L33I/M46I/I54V/L63I/V82A/I84V/L90M) (27), and the TPV in vitro-selected mutant TRM (I13V/V32I/L33F/K45I/V82L/I84V) (7). The inhibitors included in this study are representative of the various generations of inhibitors currently in clinical use or under development.

MATERIALS AND METHODS

Protein expression and purification.

The HIV-1 protease enzymes (WT and V82F/I84V, I50V, MDR-HM, and TRM mutants) used in these studies were expressed and purified in the laboratory as described before (23). A well-characterized protective autocatalytic site mutation (Q7K) was introduced in all constructs (except MDR-HM) for solution and crystallographic experiments (21, 31). This single mutant was favored over the triply stabilized mutant Q7K/L33I/L63I (21) to eliminate any potential influence of the combined L→I mutation at residues 33 and 63 on protease inhibitor binding. Throughout the text, the WT protease contains the protective mutation Q7K.

Inhibitors.

ATV, LPV, APV, and IDV were purified from commercial capsules by high-performance liquid chromatography (Waters). Purified inhibitors were lyophilized and stored at −20°C in the crystalline form. Purity of the inhibitors was determined by mass spectrometry. TPV was provided by Boehringer Ingelheim (Canada) Ltd., Research & Development. Inhibitors were dissolved in 100% dimethyl sulfoxide (DMSO) to a concentration of 15 mM, and several aliquots at different concentrations were prepared by diluting the stock solution in 100% DMSO.

Enzymatic assays.

The inhibition constant (K_i) for TPV and the other protease inhibitors was determined by measuring the change in fluorescence associated with the cleavage of the fluorogenic substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(DABCYL)-Arg. Enzymatic assays were performed in 10 mM sodium acetate buffer, pH 5.0, with 40 μM substrate and 1 to 400 nM protease at 25°C. The inhibitor concentration varied between 2 nM and 1 μM, depending on the type of inhibitor and mutant used. Several aliquots at different concentrations were prepared by diluting the stock solution (15 mM) in 100% DMSO. The final concentration of DMSO was 2% (vol/vol) in all reaction mixtures. Fluorescence was measured on a CytoFluor fluorescence multiwell plate reader (Applied Biosystems, Foster City, CA) with an excitation wavelength of 360 nm and an emission wavelength of 508 nm. Hydrolysis rates were obtained from the initial portion of the data, where at least 80% of the substrate remained nonhydrolyzed.

The catalytic rate constant, k_cat, and the Michaelis constant, K_m, were measured for the WT and all of the mutants considered in this study. The catalytic parameters were determined using the chromogenic substrate Lys-Ala-Arg-Val-Nle-nPhe-Gln-Ala-Nle-NH₂, which mimics the KARVL/AEAM sequence between the capsid and nucleocapsid. All experiments were carried out in 10 mM sodium acetate buffer, pH 5.0, at 25°C in the presence of 1 M NaCl. These experiments were performed to confirm the viability of the recombinant protease mutants. Table 1 summarizes the kinetic parameters for WT and mutant HIV-1 proteases. As previously observed, drug-resistant mutants are generally less active than the WT protease (40).

TABLE 1.

Kinetic parameters for WT and mutant HIV-1 proteases^a

Protease type	Km (μM)	kcat (s−1)	kcat/Km (s−1 μM−1)
WT	12.9 ± 0.9	10.8 ± 0.3	0.84
I50V	29 ± 2	1.90 ± 0.04	0.07
V82F/I84V	13 ± 2	3.0 ± 0.1	0.23
MDR-HM	32 ± 3	1.5 ± 0.1	0.05
TRM	31 ± 1	7.0 ± 0.1	0.23

^aThe catalytic parameters were determined using the chromogenic substrate Lys-Ala-Arg-Val-Nle-nPhe-Gln-Ala-Nle-NH₂, which corresponds to KARVL/AEAM sequence between the capsid and nucleocapsid. All of the experiments were carried out in 10 mM sodium acetate buffer, pH 5.0, at 25°C in the presence of 1 M NaCl.

ITC.

Isothermal titration calorimetry (ITC) experiments were carried out using a high-precision VP-ITC titration calorimeter system (Microcal Inc.). Direct titration experiments were performed to determine the binding enthalpy of inhibitors. In these experiments, inhibitor placed in the syringe was titrated into the cell containing protease solution. Since TPV has a limited solubility in 2% DMSO, reverse titration experiments were performed in order to determine binding enthalpies. In this reverse titration experiment, the heat associated with the binding reaction was determined by injecting 10 μl of HIV-1 protease (∼100 μM) in 10 mM sodium acetate buffer, pH 5.0, containing 2% DMSO from the syringe into the cell containing inhibitor (8.0 μM) dissolved in the same buffer. The heat evolved after each injection was calculated from the integral of the calorimetric signal. The heat due to the binding reaction between the inhibitor and the enzyme was obtained as the difference between the heat of reaction and the corresponding heat of dilution. Data were analyzed using Origin 5.0 (Microcal Software, Inc., Northampton, MA). Also, due to the poor solubility of TPV, the binding affinity, K_a, was estimated by determining K_i, the inhibition constant (which is K_d or 1/K_a) under similar solvent conditions. The following relationship was used to determine the Gibbs energy of binding (17, 36, 38): ΔG = −RT ln K_a, where K_a is equal to 1/K_d or 1/K_i, R is the gas constant, and T is the absolute temperature.

The inhibition constant (K_i) was determined by measuring the change in fluorescence associated with the cleavage of the fluorogenic substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(DABCYL)-Arg under the same conditions as the calorimetric experiments. Binding entropy was determined by using the following relationship: ΔG = ΔH − TΔS.

Crystallography.

Two to five microliters of inhibitor at 0.1 mg/μl in DMSO was added to about 130 μl of protease solution at ∼6 mg/ml, and this mixture was incubated on ice for 1 h. Any precipitated inhibitor was removed by centrifugation. Crystals were grown by the hanging drop vapor diffusion method at room temperature. Drops comprised equal volumes of protein/inhibitor complex and reservoir precipitant solution. Precipitant solutions consisted of NaCl (0.75 to 2.0 M) at pHs ranging from 4.8 to 5.8 (0.1 M acetate or citrate buffer). Crystals appeared within 24 h and grew to dimensions of around 0.20 by 0.15 by 0.02 mm after 2 to 3 days.

Crystals were briefly soaked (5 to 10 s) in a cryoprotectant prepared by making the reservoir solution 20% (vol/vol) in glycerol. Crystals were either frozen in liquid nitrogen prior to mounting or frozen in the cryostream upon mounting. Intensity data for the WT/TPV, WT/ATV, and TRM/TPV complexes were collected at 100 K on an R-AXIS IV Image Plate detector system using CuK_α radiation from a rotating copper anode source. Diffraction data for the WT/LPV and ISDV/TPV complexes were collected at beamline X4A of the National Synchrotron Light Source (Brookhaven, NY) on a Quantum-4 charge-coupled device detector (ADSC) using X-ray radiation tuned to 1.01 Å.

All diffraction data were indexed, integrated, and scaled using either the HKL or HKL2000 software suite (28). Model refinement was performed using the programs Crystallography and NMR Systems (CNS) (2) and Refmac5 (22) as implemented in the CCP4 software suite (6). All final rounds of refinement were performed with Refmac5. Initial models were obtained either by direct rigid-body refinement or by molecular replacement using the program CNS or Molrep (34). PDB file 1MSM was used as the search molecule. Residue mutations and rebuilding guided by σΑ-corrected 2mFo-DFc electron density maps were performed using the program O (11). Inhibitors were sketched using either the Sketcher module in the program Quanta (Accelrys, Inc.) or the Monomer Library Sketcher provided in CCP4i, the graphical user interface to the CCP4 collection of software. Necessary parameter and topology files were generated using the program XPLO2D (14). Inhibitors were manually positioned in the binding pocket of the protease based on mFo-DFc density maps. The presence of alternative ligand orientations in the structures of the WT protease complexed with LPV and ATV and of the I50V protease mutant complexed with TPV were detected in mFo-DFc omit maps. These were modeled, and occupancies were adjusted until refined B-factors were similar for equivalent atoms. The presence of such alternative ligand orientations has been observed previously (5). Analyses were performed using the conformation refined with the highest occupancy as interactions between inhibitor and protease were similar between orientations. Water molecules were placed using either the program CNS or ARP/wARP (29). Final structures were obtained by multiple rounds of rebuilding and refinement. Statistics for data collection and refinement are shown in Table 2.

TABLE 2.

Statistics for crystallographic data collection and refinement

Parameter	Result fora:
	WT			Mutant
	TPV	ATV	LPV	I50V-TPV	TRM-TPV
Crystal parameters	P2(1)2(1)2	P2(1)2(1)2	P2(1)2(1)2	P2(1)2(1)2	P2(1)2(1)2
a (Å)	58.6	58.7	59.7	59.0	59.1
b (Å)	86.3	85.8	85.4	86.4	86.5
c (Å)	45.9	46.2	46.2	46.2	46.4
Data collection
Resolution (Å)	48.5	23.82	25.7	24.35	19.6
Outermost shell (Å)	1.86-1.80	1.66-1.60	1.60-1.54	1.38-1.33	2.07-2.00
Total no. of reflections	89,779	143,789	184,209	270,803	61,511
No. of unique reflections	21,285	30,983	35,717	46,728	15,839
Completeness (%)	95.6 (90.2)	98.3 (95.7)	99.8 (98.6)	85.0 (88.5)	94.7 (84.2)
Redundancy	4.2 (3.0)	4.6 (3.8)	5.2 (3.9)	5.8 (5.1)	3.9 (3.1)
I/σ (I)	15.26 (1.42)	24.51 (2.88)	34.55 (3.37)	44.47 (5.62)	16.90 (3.07)
Rsymb	0.09 (0.55)	0.05 (0.47)	0.09 (0.70)	0.08 (0.40)	0.09 (0.31)
Refinement
Rworkc	0.20	0.18	0.18	0.19	0.19
Rfreec,d	0.23	0.21	0.21	0.22	0.23
Stereochemistry
R.M.S. bond length (Å)	0.015	0.013	0.014	0.010	0.018
R.M.S. angles (Å)	1.57	1.31	1.46	1.26	1.64
B-factor
Protein	26.1	23.3	11.0	13.9	24.2
Inhibitor	26.6	22.5	6.1	12.9	21.3
Water	38.3	31.4	29.4	29.6	33.4
Model composition
No. of amino acids (no. of atoms)	198 (1,516)	198 (1,516)	198 (1,516)	198 (1,514)	198 (1,520)
Ligands (no. of atoms)
Inhibitor	1 (42)	1 (51)e	1 (46)e	1 (42)e	1 (42)
Occupancy	1.00	0.75/0.25	0.75/0.25	0.70/0.30	1.00
Water	124	256	296	374	191
Glycerol	1 (6)	0	2 (12)	1 (6)	1 (6)
Chloride ion	0	1	4 (4)	2	0

^aData collection statistics given in parentheses are for the highest-resolution shell.

^bR_sym = ΣΣ_j|I_j − <I>|/Σ<I> where I_j is the measured intensity of reflection j and <I> is the mean over multiple measurements.

^cR_free is the R value as calculated below for a randomly selected test set of the intensity data not used during refinement. The test set consisted of 10% of the total intensity data.

^dR_{work, free} = Σ‖F_o − F_c‖/|F_o|.

^eTwo ligand orientations were modeled. Relative occupancies are shown in the “Occupancy” row.

ASA calculations.

Changes in accessible surface area (ASA) upon inhibitor binding were calculated according to the Lee and Richards algorithm (18). In all calculations, a solvent radius of 1.4 Å and a slice width of 0.25 Å were used. In order to better define small changes in solvent accessibility, 64 different molecular orientations of the inhibitors and protease with respect to the slicing plane are considered in the ASA calculation. The orientations are generated by rotating the molecule around the x, y, and z axes. The solvent accessibility for each atom was obtained as the numerical average of the values calculated for all molecular orientations (19).

MD.

The free-state conformations of the inhibitors were calculated by sampling the conformational space of each inhibitor during 15 ns of molecular dynamics (MD) analysis using a continuum solvation model within MOE 2004.03 (Chemical Computing Group, Inc., Montreal, Canada). To ensure adequate sampling of phase space for each inhibitor, 30 conformationally diverse structures were first constructed by running high-temperature MD simulations. The structures were inspected for conformational diversity by overlaying them, calculating root mean square deviation (RMSD) values, and by visual inspection. Each conformation was then submitted to energy minimization with gradually vanishing restraints followed by 10 ps of equilibration MD and 0.5 ns of production MD at 300 K with implicit solvent and no non-bonded interaction cutoff. The time step was set to 2 fs, and all hydrogen atom bond vibrations were frozen using the bond length constraint in the MD module of MOE. MMFF94 charges were used, as implemented within MOE 2004.03. Snapshots were recorded at 0.5-ps intervals. The accumulated free-state conformations were then overlaid with the respective bioactive conformation, obtained from X-ray structures, and RMSD values were calculated.

Protein structure accession numbers.

The coordinates for the structures listed in Table 2 have been deposited in the Protein Data Bank (2O4P, 2O4N, 2O4L, 2O4K, and 2O4S).

RESULTS AND DISCUSSION

Inhibition of WT HIV-1 protease.

The inhibition constants (K_i) of protease inhibitors were determined in enzyme inhibition assays utilizing the fluorescence substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(DABCYL)-Arg as described in Materials and Methods. TPV inhibited the WT protease with a K_i of 19 pM. ATV, LPV, and DRV (TMC-114) have comparable K_is at 35, 31, and 10 pM respectively. Earlier inhibitors APV and IDV had significantly higher K_is at 170 and 250 pM, respectively. The experimental inhibitor KNI-764 inhibited the protease with an affinity of 11 pM. Table 3 summarizes the binding thermodynamics of the seven protease inhibitors studied to the WT HIV-1 protease. Figure 1 shows the structures of all inhibitors considered in these studies.

TABLE 3.

Thermodynamic dissection of the binding affinities of protease inhibitors to WT HIV-1 protease^a

Inhibitor	Kd (pM)	ΔG (kcal/mol)	ΔH (kcal/mol)	−TΔS (kcal/mol)
TPV	19	−14.6	−0.7	−13.9
ATV	35	−14.3	−4.2	−10.1
LPV	31	−14.3	−2.4	−11.9
APV	170	−13.3	−6.7	−6.6
IDV	250	−13.1	1.3	−14.4
KNI-764b	11	−14.9	−8.0	−6.9
Darunavirb	10	−15.0	−12.7	−2.3

^aK_d values for TPV, ATV, LPV, APV and IDV are obtained from inhibition kinetics assay (from K_i). The inhibition constant (K_i) was determined by measuring the change in fluorescence associated with the cleavage of the fluorogenic substrate Arg-Glu(EDANS)-Ser-Gln-Asn-Tyr-Pro-Ile-Val-Gln-Lys(DABCYL)-Arg. Enzymatic assays were performed in 10 mM sodium acetate buffer, pH 5.0, at 25°C.

^bValues from references 24 and 40.

FIG. 1.

Chemical structure of the HIV-1 protease inhibitors included in the study.

TPV binds to the WT protease with only a slightly favorable binding enthalpy (−0.7 kcal/mol at 25°C), and its high binding affinity is driven by the entropy change. The extremely large binding entropy of TPV is unique and can be attributed to a strong contribution from the hydrophobic effect, as evidenced by its calculated log P (where P is the partition between octanol and water) value of 7.6, as well as to its small polar surface area (105.6 Ų) and its extremely low water solubility (which also precluded the performance of standard calorimetric titrations with this inhibitor). At the same temperature, ATV, LPV, and DRV, which also exhibit high affinity, demonstrated binding enthalpies of −4.2, −2.4, and −12.7 kcal/mol respectively. APV and IDV, on the other hand, exhibit binding enthalpies of −6.7 and +1.3 kcal/mol. KNI-764 also exhibits strong favorable binding enthalpy of −8.0 kcal/mol. TPV was identified as early as 1995 (33), and as such it displays a characteristic similar to that of the early inhibitor IDV (i.e., a binding affinity dominated by large favorable entropy change). The much higher potency of TPV over IDV originates from enthalpic differences unfavorable for IDV and slightly favorable for TPV. Similar binding thermodynamics has been observed with saquinavir and nelfinavir, another two of the earlier protease inhibitors (38). The situation is different with the newer inhibitors ATV, LPV, and DRV or other inhibitors (38), all of which gain potency by improved enthalpic interactions. It appears that TPV represents the potency limit of what can be achieved without invoking a strong favorable binding enthalpy before completely losing water solubility. The thermodynamic signatures of all the inhibitors considered in these studies are shown schematically in Fig. 2, which clearly indicates the entropic similarity of TPV and IDV.

FIG. 2.

Enthalpic and entropic contributions to the binding affinity of protease inhibitors. The blue, green, and red bars represent free energy change (ΔG), enthalpy change (ΔH), and entropy change (−TΔS) at 25°C, respectively, associated with inhibitor binding to the WT HIV-1 protease.

Response to multidrug-resistant mutant V82F/I84V.

V82F/I84V is a well-studied resistance mutation combination that affects all HIV-1 protease inhibitors in clinical use (35, 40). These mutations are located at the edges of the active site of the HIV-1 protease and act by distorting the three-dimensional geometry of the binding site without changing its polarity. Figure 3 shows the position of these mutations in the active site of the protease. V82 is located in the subsites P2 and P2′, whereas I84 is located in the subsites P1 and P1′. The replacement of valine by phenylalanine at the 82 position causes a volume reduction in the binding cavity, whereas the replacement of isoleucine by valine at position 84 creates room by eliminating a methyl group. The combination of a volume increase (82V→F) and decrease (84I→V) results in a distorted geometry compared to the WT cavity. The inhibition constants, K_i, were determined to be 0.3, 0.4, 1.0, 21, and 32 nM for TPV, ATV, LPV, APV, and IDV, respectively. Inhibition constants of 0.3 and 0.4 nM have been previously reported for KNI-764 and DRV against the same mutant (24, 40).

FIG. 3.

Structure of HIV-1 protease mutants showing the location of amino acid mutation sites. (A) V82F/I84V. (B) I50V. (C) I13V/V32I/L33F/K45I/V82L/I84V (TRM). (D) L10I/L33I/M46I/I54V/L63I/V82A/I84V/L90M (MDR-HM). Most of the mutations are located near the active site of HIV-1 proteases.

As summarized in Table 4, all inhibitors lose affinity against the mutant, however, not all to the same degree; some inhibitors lose potency by more than 100-fold. Invariably, the inhibitors that lose significant potency do so because they suffer both binding enthalpy and binding entropy losses. Two mechanisms are observed among the inhibitors that respond well to the mutation. In the first case, ATV, KNI-764, and DRV experience a small loss in ΔG because they partially compensate for a loss in binding enthalpy with an actual gain in binding entropy. This escape mechanism has been reported before (25, 35, 38) and is structurally accomplished when the inhibitor is flexible enough to adapt to the distorted binding cavity and gains entropy by additional burial from the solvent and/or a gain in conformational degrees of freedom. The second mechanism is that of TPV, which loses little affinity to the mutant by partially compensating for a substantial binding entropy loss by an actual enthalpy gain. To our knowledge, this type of behavior has not been documented before and is unique to TPV. It must be noted, however, that Yanchunas et al. (42) reported that all inhibitors, including acetyl pepstatin, bind more exothermically to the I50L/A71V protease mutant.

TABLE 4.

Thermodynamic dissection of the potency loss of protease inhibitors to multidrug-resistant V82F/I84V protease^a

Inhibitor	Kd ratio	ΔΔG (kcal/mol)	ΔΔH (kcal/mol)	−TΔΔS (kcal/mol)
TPV	16	1.6	−1.0	2.6
ATV	11	1.4	4.2	−2.8
LPV	32	2.1	2.2	−0.1
APV	124	2.9	2.1	0.8
IDV	128	2.9	2.2	0.7
KNI-764	26	1.9	2.3	−0.4
Darunavir	41	2.2	3.6	−1.4

^aExperiments were performed in 10 mM sodium acetate, pH 5.0, and 2% DMSO, at 25°C. The K_d ratio was determined using the relationship K_d ratio = K_{d V82F/I84V}/K_{d WT}. Free energy change associated with mutation, ΔΔG, was calculated by using the relationship ΔΔG = ΔG_V82F/I84V − ΔG_WT. Enthalpy change (ΔΔH) and entropy change (−TΔΔS) associated with mutation were determined using similar relationships.

At the thermodynamic level, a broad antiviral response to mutations translates into a small loss in the Gibbs energy of binding. In turn, a small loss in Gibbs energy is achieved when enthalpic and entropic losses are very small or when enthalpic or entropic losses partially compensate for each other. Most inhibitors lose binding enthalpy when facing a mutation due to the fact that they are highly optimized to the WT enzyme, and any changes in the binding cavity tends to weaken van der Waals (vdW) interactions and hydrogen bonds. TPV is the first and so far the only inhibitor that responds to the V82F/I84V mutation by gaining favorable binding enthalpy as reflected by the negative ΔΔH of −1.0 kcal/mol (Table 4).

The response to a TPV in vitro-selected resistant mutant.

While V82F/I84V is a common resistance combination, it is not specific to TPV. Thus, it was important to verify whether the response described above was representative of a more general behavior. In order to investigate further the response of TPV to mutations, we studied the responses of the different inhibitors to the TPV in vitro-selected mutant I13V/V32I/L33F/K45I/V82L/I84V (TRM) (7). These mutations are located within and outside the binding cavity as shown in Fig. 3. Table 5 shows the results for the inhibitors of this study. Contrary to V82F/I84V, the combination of mutations in TRM elicits a much more pronounced reduction in potency, ranging from 2 to 3 orders of magnitude in all inhibitors. Since TRM contains a set of mutations selected specifically by TPV, it is not surprising it loses potency by about 800-fold while LPV and DRV lose potency by less than 200-fold. On the other hand, the most pronounced reduction in response is that of APV, which loses potency by a factor of 1,400. Against this mutant, all inhibitors lose binding enthalpy, except TPV, which actually gains a substantial 3 kcal/mol of binding enthalpy. This is further evidence that TPV exhibits a unique thermodynamic response to mutations. In fact additional experiments with another common mutation, I50V, localized at the tips of the flaps (Fig. 3) also resulted in a binding enthalpy gain for TPV, while the other inhibitors studied sustained significant losses (Table 6).

TABLE 5.

Thermodynamic dissection of the potency loss of protease inhibitors to multidrug-resistant TRM protease^a

Inhibitor	Kd ratio	ΔΔG (kcal/mol)	ΔΔH (kcal/mol)	−TΔΔS (kcal/mol)
TPV	790	4.0	−3.0	7.0
ATV	630	3.8	6.9	−3.1
LPV	110	2.8	2.4	0.4
APV	1,400	4.3	4.5	−0.2
IDV	380	3.5	3.5	0.0
KNI-764	390	3.5	9.4	−5.9
Darunavir	170	3.0	7.3	−4.3

^aExperiments were performed in 10 mM sodium acetate, pH 5.0, and 2% DMSO at 25°C. The K_d ratio was determined using the relationship K_d ratio = K_{d TRM}/K_{d WT}. Free energy change associated with mutation, ΔΔG, was calculated by using the relationship ΔΔG = ΔG_TRM − ΔG_WT. Enthalpy change (ΔΔH) and entropy change (−TΔΔS) associated with mutation were determined using similar relationships.

TABLE 6.

Thermodynamic dissection of the potency loss of protease inhibitors to I50V drug-resistant protease^a

Inhibitor	Kd ratio	ΔΔG (kcal/mol)	ΔΔH (kcal/mol)	−TΔΔS (kcal/mol)
TPV	32	2.1	−3.6	5.7
ATV	94	2.7	1.3	1.4
LPV	87	2.6	1.5	1.1
APV	65	2.5	2.6	−0.1
IDV	26	1.9	1.1	0.8
KNI-764b	18	1.7	2.5	−0.8
Darunavirb	52	2.3	5.2	−2.9

^aExperiments were performed in 10 mM sodium acetate, pH 5.0, and 2% DMSO at 25°C. The K_d ratio was determined using the relationship K_d ratio = K_{d I50V}/K_{d WT}. Free energy change associated with mutation, ΔΔG, was calculated by using the relationship ΔΔG = ΔG_I50V − ΔG_WT. Enthalpy change (ΔΔH) and entropy change (−TΔΔS) associated with mutation were determined using similar relationships.

^bValues from reference 24.

The response to multidrug-resistant hexamutant MDR-HM.

Of all the HIV-1 protease mutants previously studied in this laboratory, MDR-HM (L10I/L33I/M46I/I54V/L63I/V82A/I84V/L90M) was the one that consistently elicited the largest drops in potency to existing inhibitors (27). These mutations are located within and outside the binding cavity, including the flap as shown in Fig. 3. The thermodynamic parameters for all inhibitors against MDR-HM are summarized in Table 7. Against the MDR-HM mutant, TPV experienced a loss in binding enthalpy. Although this is the only occurrence of an enthalpy loss observed for TPV, it nevertheless was the smallest loss of all inhibitors studied (Fig. 4).

TABLE 7.

Thermodynamic dissection of the potency loss of protease inhibitors to MDR-HM protease^a

Inhibitor	Kd ratio	ΔΔG (kcal/mol)	ΔΔH (kcal/mol)	−TΔΔS (kcal/mol)
TPV	53	2.4	3.0	−0.6
ATV	170	3.0	5.9	−2.9
LPV	210	3.2	6.7	−3.5
APV	180	3.1	8.3	−5.2
IDV	720	3.9	7.1	−3.2
KNI-764b	23	1.9	5.6	−3.7
Darunavirb	23	1.9	8.7	−6.8

^aExperiments were performed in 10 mM sodium acetate, pH 5.0, and 2% DMSO at 25°C. The K_d ratio was determined using the relationship K_d ratio = K_{d MDR-HM}/K_{d WT}. Free energy change associated with mutation, ΔΔG, was calculated by using the relationship ΔΔG = ΔG_TRM − ΔG_WT. Enthalpy change (ΔΔH) and entropy change (−TΔΔS) associated with mutation were determined using similar relationships.

^bValues from reference 24.

FIG. 4.

The enthalpic response of HIV-1 protease inhibitors to different mutants associated with drug resistance. A positive value indicates a loss in binding enthalpy, and a negative value indicates a gain in binding enthalpy.

In summary, these experiments indicate that contrary to other inhibitors which predominantly lose enthalpic interactions when facing mutations, TPV instead loses binding entropy and is able to gain binding enthalpy even against mutations like TRM, against which it loses significant potency.

Crystal structure of the TPV/WT protease complex.

TPV is tightly bound to the WT protease interacting with 11 residues from chain A and 6 from chain B through direct and water-mediated hydrogen bonds (H-bonds) and vdW interactions. Figure 5 illustrates the bound conformation of TPV, and Fig. 6 illustrates its hydrogen bonding interactions with the protease. Except for the terminal benzyl moiety and sulfonamide group, TPV is completely wrapped by protease residues. Considering the pyran ring as the center of the TPV molecule, the half containing the sulfonamide and trifluoro groups is largely hydrophilic, and the other is largely hydrophobic. In contrast, the protein environments of each half are almost identical, being related by an approximate two-fold symmetry. Relative to LPV and ATV, TPV does not bind in an extended conformation, but rather adopts a conformation resembling an “S” in which the hydrophilic portion is compressed (Fig. 5). This conformation of the ligand leaves a small cavity within the binding pocket on the hydrophobic side of the ligand which is occupied by a glycerol molecule (glycerol was used as a cryoprotectant). TPV makes a total of 17 vdW contacts with 7 residues from chain A and 4 from chain B of the protease, as determined by the program LIGPLOT (41). Contacts are distributed uniformly over the inhibitor with one notable exception: the pyran ring makes no vdW contacts. The WT protease is involved in two close C = O … H-C interactions with TPV. In making a strong hydrogen bond with the ligand sulfonamide nitrogen, the backbone carbonyl oxygen of residue Gly48A is brought to within 3.05 Å of the aromatic carbon atom located ortho to the sulfonamide and pyran substituents. Also the backbone carbonyl oxygen of residue Gly27A is at a distance of 3.15 Å from the aromatic carbon atom located between the pyridine nitrogen and the CF₃ group of TPV.

FIG. 5.

TPV in the binding site of HIV-1 protease. The protein surface is colored according to the chain. Residues 46 to 62 in the flap region of each chain have been omitted to show clearly the ligand. The mFo-DFc omit map is contoured at 2 σ and carved about the ligand at a distance of 1.7 Å. This figure was created with PyMol.

FIG. 6.

Hydrogen bond interactions of TPV with WT HIV-1 protease and mutant proteases. (A) WT. (B) TRM. (C) I50V. The structures were obtained to resolutions of 1.8 Å, 2.0 Å, and 1.33 Å, respectively.

TPV is involved in seven direct hydrogen bonds with the WT protease. The two oxygen atoms of the TPV pyran ring participate in H-bond interactions with equivalent residues on both chains of the protease: the carbonyl oxygen atom is recognized by the backbone nitrogen atoms of Ile50A and Ile50B, while the enol oxygen atom is recognized by the side-chain oxygen atoms of Asp25A and Asp25B. All other hydrogen bonds are located on the hydrophilic side of the ligand and involve only one monomer (chain A) of the protease dimer. The nitrogen atom of the sulfonamide interacts with the backbone oxygen of Gly48A, and the nitrogen atom in the terminal pyridine ring participates in an H-bond interaction with the backbone nitrogen of Asp29A. Additionally, one of the sulfonamide oxygens is involved in an H-bond interaction with the backbone nitrogen of Asp30A. A potential eighth direct H-bond interaction between the same sulfonamide oxygen and a side-chain oxygen of Asp30A fails to meet the necessary geometric criteria. Finally, one water-mediated H-bond between the other sulfonamide oxygen and the NH backbone of Gly48A is present.

The most important binding features of the in vitro-selected TRM-mutant protease with TPV are similar to those of the WT. Similarly, the mutation I50V has little effect on protease or TPV conformation. An overlay of Cα atoms results in a Cα RMSD of only 0.25 Å for I50V relative to the WT, and an overlay of inhibitors results in an all-atom RMSD of 0.20 Å for bound inhibitors. The only noticeable structural difference between overlaid inhibitors is a 26° rotation of the n-propyl substituent of the pyran ring. This readjustment allows for packing to be optimized in the absence of the Ile50A CD1 atom.

The structural origin of the TPV response.

When bound to the HIV-1 protease, TPV establishes a strong hydrogen bond network involving only conserved residues or backbone atoms within the active site cavity (catalytic Asp25 and backbone of Asp29, Asp30, Gly48, and Ile50) (Fig. 6 and 7A). These hydrogen bonds are maintained with all of the mutants studied, as also shown in Fig. 6. In the case of the TRM mutant, an additional water-mediated hydrogen bond is made between the sulfonamide group and Asp30A. Unlike all inhibitors in clinical use, TPV forms hydrogen bonds directly to the backbone NH atoms of Ile50 in both chains of the protease (Fig. 6 and 7A). For all other inhibitors, this interaction with those two residues in the flaps of the protease is mediated by a water molecule. The direct hydrogen bond interaction is considered energetically more favorable than the water-mediated one due to the entropic gain associated with the release of the ordered water molecule into the bulk solvent (rather than the immobilization of a water molecule by all the other inhibitors as shown in Fig. 7). In general, TPV relies on fewer water-mediated hydrogen bonds than other inhibitors (Fig. 7). There is only one water-mediated hydrogen bond between TPV and the protease, while there are six for ATV, three for LPV, four for IDV (PDB no. 1HSG) (4), two for APV (PDB no. 1HPV) (12), three for KNI-764 (PDB no. 1MSM) (35), and three for DRV (PDB no. 1T3R) (32). Interestingly, it was also found that the absolute number of water molecules within 6 Å of the wild-type-bound TPV, ATV, and LPV (17, 29, and 21 water molecules, respectively) was not only smallest for TPV but also remained the lowest after removing the molecular size bias by dividing by the vdW surface area of the inhibitor (data not shown). While the resolutions of the three structures are not exactly the same, all of them are of sufficient quality to permit this analysis. The smaller number of water molecules immobilized at the TPV-protein interface will certainly contribute to a more favorable binding entropy for TPV (while decreasing the binding enthalpy) (20).

FIG. 7.

Image of X-ray structure of WT HIV protease complexes with TPV (A) ATV (B), and LPV (C). For the protease complex with TPV, H-bonds are formed with only conserved residues or backbone atoms within the active site cavity. A water-mediated H-bond is observed between the sulfonamide group and the backbone of Gly48. TPV forms direct H-bonds with the backbone NH atoms of Ile50 in both chains of the protease. For LPV and ATV, this interaction with those two residues in the flap of the protease is mediated by a water molecule. TPV also has fewer water-mediated hydrogen bonds compared to other inhibitors. H-bond distances in angstroms are shown.

The entropy penalty associated with the ordering of one water molecule has been estimated to be as high as 7 cal/K·mol, or equivalently close to 2 kcal/mol at room temperature (8). The large favorable binding entropy of TPV originates primarily from the burial of hydrophobic surface area (827 Ų) combined with the entropic gain associated with the release of the water molecule hydrogen bonded to Ile50. On the other hand, IDV achieves a similar entropic contribution to the binding energy, but primarily due to a much higher hydrophobic burial (1,053 Ų). KNI-764 binding is associated with the burial of 889 Ų of hydrophobic surface, APV with 799 Ų, DRV with 760 Ų, LPV with 995 Ų, and ATV with 1,038 Ų. It is apparent from this analysis that the large binding entropy observed for TPV is higher than expected based on the burial of hydrophobic surface alone. In fact, a correlation analysis between the buried hydrophobic surface areas described above and the entropy change for the WT (Table 3) yields a correlation coefficient of 0.92 if TPV is excluded but a correlation coefficient of 0.65 if TPV is included (Fig. 8). Thus, the entropic contribution of TPV is close to 8.2 kcal/mol more favorable than that predicted on hydrophobicity alone. This analysis underlines the unique behavior of TPV among existing HIV-1 protease inhibitors, which can be partly attributed to the distinctive feature of ordering fewer water molecules than all of the other inhibitors. Moreover, MD simulations of TPV, ATV, and LPV in the free-state show that TPV is able to adopt the bound conformation more readily than ATV and LPV, as determined by the areas under the curve (AUCs) below the 3.0-Å RMSD between free-state and bioactive conformations from X-ray structures (Fig. 9). TPV has an AUC of 8375 below 3.0 Å, compared to AUCs of 654 for ATV and 2,532 for LPV. This behavior of TPV reflects its highly hydrophobic character that tends to limit exposure to bulk water. This in turn further contributes to a lower conformational entropy loss for TPV upon binding to protease.

FIG. 8.

Correlation analysis between buried hydrophobic surface areas (change in apolar accessible surface area, ΔASA_apolar) and entropy change for WT protease and different protease inhibitors. Changes in ASA_apolar for IDV, APV, KNI-764, and DRV were calculated from the PDB files 1HSG, 1HPV, 1MSM, and 1T3R, respectively. Excluding TPV, there is a correlation coefficient of 0.92 between buried ASA_apolar and entropy changes associated with inhibitor binding to WT protease. The binding entropy for TPV is higher than expected (∼8.2 kcal/mol) based on the burial of hydrophobic surface only. The higher entropy change for TPV may be attributed to the entropy gained from the release of buried water molecule from the active site of HIV-1 protease on binding.

FIG. 9.

RMSD between free-state conformations obtained by MD simulations of TPV, ATV, and LPV and their bioactive conformations from X-ray structures.

Although favorable, the binding enthalpy of TPV is small compared to those of other newer inhibitors (Table 3). The small binding enthalpy of TPV cannot be attributed solely to fewer hydrogen bond interactions than other inhibitors with more favorable enthalpies. Other binding characteristics appear to contribute to the low binding enthalpy of TPV. First, while a smaller number of ordered water molecules contribute favorably to the binding entropy, this carries a compensating enthalpy penalty that lowers the overall binding enthalpy (20). Second, ligands that carry a hydrogen-bonded and solvent-buried sulfonamide group have been observed to elicit highly favorable binding enthalpies (3). Among the existing HIV-1 protease inhibitors, DRV (15), APV, and TPV have a sulfonamide moiety. In TPV, the sulfonamide group participates in an H-bond with the backbone NH atom of Asp30A (3.2 Å) and a water-mediated bond with Gly48A (2.8 Å). In APV and DRV, the sulfonamide participates in a water-mediated H-bond with Ile50 (2.8 Å and 2.9 Å, respectively). A fundamental difference between the sulfonamide moieties of these inhibitors is that in DRV and APV, the sulfonamide group is completely buried from the solvent, while in TPV it is partially exposed (13). The partial solvent exposure is consistent with a lower binding enthalpy and also provides important clues to the enthalpic response of TPV to mutations. Contrary to other inhibitors, TPV either gains enthalpy or loses little enthalpy against the mutants studied. As mentioned above, all of the hydrogen bonds found in the WT complex are maintained in the complexes with the mutants, and in the case of the resistant mutant TRM, an additional water-mediated H-bond between the sulfonamide and Asp30 is formed. Furthermore, in other mutants, the sulfonamide moiety becomes less exposed to the solvent, either directly or indirectly via an additional structured water molecule.

Conclusions.

The study presented here reveals a new thermodynamic mechanism by which an HIV protease inhibitor can minimize the deleterious effects of resistance mutations. Previously, inhibitors maintaining activity in the setting of mutations have partially compensated for enthalpy losses with entropy gains (16). TPV is the only HIV protease inhibitor known to date to exhibit the opposite behavior: it partially compensates for entropy losses by actual enthalpy gains or by sustaining minimal enthalpy losses. Although the origin of this unique mechanism remains to be fully understood and explored, the structural and thermodynamic analysis are providing clues as to why TPV may behave in this distinctive manner. Structurally, TPV makes an extensive network of seven direct hydrogen bonds with conserved elements of the protease (backbone atoms and catalytic residues) that cannot undergo mutation. Also, TPV establishes fewer water-mediated hydrogen bonds compared to other protease inhibitors. In particular, TPV is the only protease inhibitor in clinical use that makes direct H-bonds to the Ile50 residues in the flap, while all other protease inhibitors do so through water-mediated H-bonds.

This and previous studies (16) delineate alternative mechanisms by which protease inhibitors can minimize the effects of mutations by eliciting enthalpy/entropy compensation. TPV can be considered as an enthalpically restrained inhibitor; i.e., an inhibitor that binds to the WT protease with a barely favorable enthalpy but that contains the potential to enhance its enthalpic interactions when facing protease mutants. This inhibitor retains some flexibility within the WT binding pocket, as suggested by ordering fewer water molecules. Under those circumstances, drug resistance mutations that make the binding cavity smaller usually induce a conformational entropy loss which is enthalpically compensated for by improved vdW interactions. This response to mutations may also have kinetic consequences. For example, the balance between k_on and k_off of the inhibitors might be different for molecules that compensate for entropic losses by enthalpic gains compared with those that exhibit the opposite response, potentially representing a beneficial mechanism for responding to mutations from a clinical point of view. In summary, the discoveries identified in this study provide additional evidence supporting the high genetic barrier to mutations of TPV as well as the thermodynamic basis for its potent antiviral activity against protease inhibitor-resistant mutants.