PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Biochemistry. Author manuscript; available in PMC 2017 April 26.
Published in final edited form as:
PMCID: PMC5013724
NIHMSID: NIHMS813828

Binding of clinical inhibitors to a model precursor of a rationally selected multidrug resistant HIV-1 protease is significantly weaker than to the released mature enzyme

Abstract

We have systematically validated the activity and inhibition of a HIV-1 protease (PR) variant bearing 17 mutations (PRS17), selected to represent high resistance by machine learning on genotype-phenotype data. Three of five mutations in PRS17 correlating with major drug resistance, M46L, G48V and V82S, and five of eleven natural variations, differ from two clinically derived extreme mutants, PR20 and PR22 bearing 19 and 22 mutations, respectively. PRS17, which forms a stable dimer (<10 nM), is ~10- and 2-fold less efficient in processing the Gag polyprotein relative to the wild-type and PR20, respectively, but maintains the same cleavage order. Isolation of a model precursor of PRS17 flanked by the 56 amino acid transframe region (TFP-p6pol) at its N-terminus, which is impossible when expressing an analogous PR20 precursor, allowed systematic comparison of inhibition of TFP-p6pol-PRS17 and mature PRS17. Resistance of PRS17 to 8 protease inhibitors (PIs) relative to PR ranges from 1.5 to 5 orders of magnitude increase in Ki from 0.01 to 8.4 μM. Amprenavir, darunavir, atazanavir and lopinavir, the most effective of the 8 PIs, inhibit precursor autoprocessing at the p6pol/PR site with IC50 ranging from ~7.5 to 60 μM. Thus this process, crucial for stable dimer formation, shows ~200 to 800-fold weaker inhibition than the mature PRS17. TFP/p6pol cleavage, which occurs faster, is inhibited even more weakly by all PIs except darunavir (IC50 of 15 μM); amprenavir shows a 2-fold increase in IC50 (~15 μM), and atazanavir and lopinavir show increased IC50 of >42 μM and >70 μM, respectively.

Keywords: HIV protease, drug resistance, autoprocessing, inhibitor binding, enzyme kinetics

Graphical abstract

An external file that holds a picture, illustration, etc.
Object name is nihms813828u1.jpg

INTRODUCTION

The human immunodeficiency virus type 1 protease (HIV-1 PR) plays a pivotal role by processing the viral polyproteins into mature proteins essential for the assembly and maturation of the virus in its replication cycle.1 The mature PR is a homodimeric aspartic protease in which each 99 amino acid monomer contributes one of the two aspartic acid residues (Asp25) required for catalytic function. A single copy of PR is synthesized in the pol open-reading frame of the Gag-Pol precursor. The PR promotes its own release via transient dimerization and cleavages at its N- and C-termini, termed autoprocessing. The initial intramolecular cleavage at the N-terminus, between the transframe region (TFR, Fig. 1A) and PR, is essential for stable dimer formation and appearance of mature-like catalytic activity. Subsequent intermolecular cleavage at the C terminus of PR (PR/RT junction) yields the mature PR.2 Since its initial discovery, PR has been a target for therapeutic intervention, and several drugs designed specifically to bind to the active site of PR are in current use.3,4 Protease inhibitors (PIs) are used in combination antiretroviral therapy (cART) with other antiretroviral agents that block reverse transcriptase and integrase activities, and viral entry5 (https://aidsinfo.nih.gov/contentfiles/lvguidelines/adultandadolescentgl.pdf). However, the short replication cycle of 1–2 days6 combined with error-prone nature of RT (~2×10−5 errors/nucleotide per replication cycle)7 favor rapid evolution of mutations selected under drug pressure thus resulting in extreme drug resistance.811

Figure 1
Organization of the Gag-Pol polyprotein and description of the drug resistant mutants. (A) Domain organization of HIV-1 Gag-Pol polyprotein. Abbreviations are MA, matrix; CA, capsid; SP1, spacer peptide 1; NC, nucleocapsid; TFR, transframe region; PR, ...

Several multi-drug resistant mutants of PR have been isolated and characterized structurally to explain weaker binding of clinical PIs and enable the design of future inhibitors targeting such mutants to be used in cART.4,8,9,1214 One well characterized extremely drug resistant isolate of PR bearing 19 mutations (PR20)10,1517 exhibits 3–4 orders of magnitude weaker binding to clinical PIs relative to the wild-type enzyme. However, it mediates Gag processing only about 4-fold more slowly and retains the same order of cleavage as the wild-type enzyme. Importantly, it undergoes efficient N-terminal autoprocessing, a requisite to release the fully active stable mature PR20. Consistent with the weak binding of SQV and DRV to mature PR20, autoprocessing of the PR20 model precursor (TFR-PR20) when expressed in E. coli is not inhibited up to a practical limit of 150 μM by these PIs in contrast to the 1–2 μM IC50 observed for the wild type precursor.16

Exponentially growing databases in recent years have led to the application of machine learning and computational methods for the evaluation of inhibitors for use in drug therapy to optimize virological outcomes in HIV-1 infected patients.18 A new predictive method for drug resistance uses machine learning with a unified encoding of the sequence and 3-D structure of the target protein, such as HIV-1 PR.19 Superior results were obtained by including structural and sequence information as compared to other methods based only on sequence, with classification accuracies in the 93–99% range for PR instead of 60–87% accuracy.20 The unified encoding allowed accurate prediction of drug resistance from genomic data with several different machine learning algorithms. It also gave excellent correlation between predicted and observed resistance in cross-validated regression to predict not just resistance but the degree of resistance.20 The success of linear regression demonstrated that the unified encoding accurately encapsulated the structure and sequence information. This approach can be used to cluster genomic sequence data and select representative samples for further analysis.

Representative sequences were extracted from the genotype sequence data in the Stanford HIV drug resistance database by mean shift clustering with regression analysis, in a procedure designed to identify mutants with the common features characterizing high level resistance.21 The clustering selected a small number of sequences that are able to quantitatively reproduce the resistance phenotype for the entire dataset with high accuracy in the regression. The method also selected sequences that are in the center of the distribution and therefore characteristic of a wide range of other mutations. Some sequences represented mutants resistant to more than one inhibitor. Analyzing such overlaps resulted in a single sequence predicted to have high level resistance to six of the inhibitors: atazanavir (ATV), amprenavir (APV), nelfinavir (NFV), indinavir (IDV), lopinavir (LPV), and SQV. This sequence, bearing 17 mutations, was termed PRS17 (S for selected) and chosen for experimental validation of the inhibition and further characterization.

Here we evaluate the dimer dissociation constant of PRS17, and its catalytic efficiency by monitoring cleavage of a peptide substrate as well as the Gag polyprotein. The binding affinity of 8 clinical inhibitors to mature PRS17 was estimated by enzyme kinetics and isothermal titration calorimetry. Very little is known about the folding landscape, inhibition of autoprocessing and associated drug resistance for the PR precursor. The latter two are predicted to have distinct differences from the properties of the mature PR described in a majority of studies. Even though isolation of the wild-type model precursor is feasible, a systematic comparison of inhibition of autoprocessing (IC50) in vitro with Ki values for inhibition of the mature PR was impractical because of the detection limit of the precursor below 1–2 μM protein concentration by SDS-PAGE. On the other hand, it was also impossible to isolate and purify a model precursor (TFR-PR20) of the extremely drug resistant mutant PR20 because the precursor does not accumulate during expression due to efficient autoprocessing coupled with weak binding of PIs even at 150 μM.16 Fortuitously, the model precursor of PRS17, TFR-PRS17, provided an ideal window that is between the 2 extremes typified by PR and PR20. Thus, we were able to compare the Ki values for the inhibition of mature PRS17 by 8 PIs with the IC50 values for inhibition of autoprocessing of TFR- PRS17 in vitro by the same PIs.

Our results indicate that PIs bind 1–5 orders of magnitude weaker to mature PRS17 relative to the wild type and PIs with higher affinity to the mature PRS17 show larger differences when compared with IC50 values for inhibition of autoprocessing, providing the first such comparison. Hence, the PRS17 construct provided an important model for a comprehensive in vitro evaluation of drug resistance of the mature PR and its precursor.

MATERIALS AND METHODS

Cloning, protein expression and purification

The cloning, expression and isolation of TFR-PR, PR and PR20 have been described before.16,22 Inserts coding for TFR-PRS17 and PRS17 were custom synthesized (DNA 2.0, Menlo Park, CA), cloned in pJ414 flanked by Nde1 and BamH1 sites and transformed into E. coli BL-21 (DE3). Induction for protein expression, isolation of inclusion bodies and protein purification followed previously described protocols.23,24 Recombinant constructs were verified both by DNA sequencing and electrospray ionization mass spectrometry.

The clinical isolates, PR20 containing 14 PI resistance-associated mutations including those specific to DRV, and PR22 from a pediatric patient with high viral load despite HAART including PIs are described in references 15 and 11, respectively.

Dimer dissociation and kinetic parameters

The dimer dissociation constant (Kdimer) for PRS17 was determined in 50 mM sodium acetate buffer, pH 5.0, containing 250 mM NaCl (buffer B) from the dependence of initial rate for hydrolysis of a chromogenic substrate (Lys-Ala-Arg-Val-Nle-[4-nitrophenylalanine]-Glu-Ala-Nle-NH2, California Peptide Research, Napa, CA)25 as a function of dimeric PRS17 concentration at 28 °C. Assaying at concentrations below 10 nM was impractical and only an upper limit for dimer dissociation could be estimated. Km and k cat for substrate hydrolysis by mature PRS17 were determined spectrophotometrically in the same buffer at an enzyme concentration of 0.5 μM and 20 - 400 μM of the chromogenic substrate (Δε at 310 nM = 1797 Abs M−1cm−1) at 28 °C. Data were analyzed and plotted (Fig. S1) by use of the Enzyme Kinetics module of Sigmaplot 10.0 (Systat Software, Inc).

Inhibitor binding

Protein–inhibitor association constants (Kassoc) were measured by isothermal titration calorimetry (Microcal iTC200, Malvern, Westborough, MA). Freshly folded26 PRS17 (~15 μM as dimer) in 50 mM sodium acetate buffer, pH 5.0 (buffer A), was titrated with 105 μM inhibitors in the same buffer at 28 °C. APV, IDV and DRV exhibited adequate thermal response upon titration and the data were processed with the instrument’s Origin software (Fig. S2). For competitive inhibitors that bind at only one site, dissociation constants (1/Kassoc) are equivalent to the inhibition constants measured by enzyme kinetics (Ki).

Ki values for six inhibitors were determined kinetically in buffer B, 28 °C, at a concentration of 0.5 μM enzyme and 360 μM chromogenic substrate. Initial rates (vi) of absorbance decrease at 310 nm were converted to μM/min as described for the determination of kinetic parameters. Values of 1/vi plotted versus inhibitor concentration (Fig. S3) gave linear regressions from which the slope and y-axis intercept were obtained; for competitive inhibitors, the negative x-axis intercept corresponds to the IC50 for inhibition and was determined from the relationship, x-intercept = – (y-intercept)/(slope). Since the IC50 is substrate dependent, the Ki value was obtained from the relationship Ki = IC50/(1 + [substrate]/Km) using the measured value of Km = 193 μM under the same conditions. Errors were estimated from the sum of the percent errors in the slope, y-intercept and independently measured Km.

Autoprocessing

A stock solution of the TFR-PRS17 precursor (1.86 mg/ml, 54 μM in dimer concentration) maintained in 12 mM HCl was diluted with 2.3 volumes of 5 mM sodium acetate buffer at pH 6 in the presence or absence of inhibitor, followed immediately by the addition of 3.3 volumes of 100 mM acetate buffer at pH 5. In the absence of inhibitor, aliquots (10 μl) were drawn at the indicated times from 0.5–60 min, mixed with 4 μl sample loading buffer and subjected to SDS-PAGE on 20% homogeneous PhastGels (GE Healthcare). Similarly, autoprocessing reactions were carried out with varying concentrations of the inhibitor for 60 min, followed by SDS-PAGE.

Gag Processing

Assays (20 μl) to monitor the order and rate of cleavage of ΔGag, a truncated Gag construct spanning the matrix, capsid, SP1 and nucleocapsid domains, were carried out as described previously.10,27 ΔGag (50 μM) in 20 mM sodium phosphate, pH 6.5, 300 mM NaCl, 0.1 mM ZnCl2 and 1 mM tris(2-carboxyethyl)phosphine was mixed with the appropriate protease to give a final concentration of 1 μM at room temperature. Aliquots (2 μl) were drawn at the indicated times and mixed with 3 μl of the above assay buffer and 2 μl of SDS-PAGE sample loading buffer, heated to 85 °C for 2 min and subjected to electrophoresis on 20% homogeneous PhastGels using 8 lane × 1 μl applicators (GE Healthcare). Gels were stained with PhastGel Blue R, destained and digitized.

RESULTS AND DISCUSSION

Description of mutations in PRS17

Organization of the precursor Gag-Pol polyprotein and alignment of three mature protease sequences, PR, PR20 and PRS17 are shown in Figure 1A and 1B, respectively. The positions of all mutations in PRS17 are shown in comparison with the extremely drug resistant clinical isolate PR20 (3UCB17) in Figure 1C. In a recent study, we ascribed the autoprocessing, catalytic efficiency and extreme drug resistance of PR20 partly to cooperative effects of mutations in three clusters comprising residues 30–37, 60–67 and 87–94 (red stripes in Fig. 1B).10 These clusters in PR20 only partially overlap regions containing mutations in PRS17, such that 3 out of 6 residues corresponding to cluster 1, 2 out of 3 in cluster 2, and 1 out of 3 in cluster 3 are represented in PRS17. Eleven of all mutations either in PR20 or in PRS17 correspond to regions of natural variation (sequence not highlighted in Fig. 1B, see Fig. S1 in reference 16, and reference 28 for details). Major mutations associated with drug resistance (DRMs; Fig. 1B) are localized between regions that are highly conserved, essential for dimerization and enzyme function, and regions showing natural variation. Highly conserved and regions of natural variation account for about 28% and 46% of the total length of the PR monomer, respectively.

Mutations in PR20 and another highly drug resistant mutant, PR22,11 compared with PRS17 and their associated resistance to specific clinical PIs are listed in Table S1 (http://hivdb.stanford.edu/DR/PIResiNote.html and reference 29). Mutations in PRS17 associated with resistance to multiple drugs are M46L, G48V, I54V, V82S and L90M. With the exception of L90M these mutations are not present in PR20. In contrast, mutations D30N, V32I, L33F, I47V, I54L, and I84V are present in PR20 but not in PRS17. Thus, despite the large number of mutations in both PR20 and PRS17, similarity between mutations at specific sites in the two constructs is limited to 2 conservative substitutions shown in blue and 6 exact matches (in red). Within this group of 8 mutations, the conservative substitutions I54L (in PR20) and I54V (in PRS17) have similar but not identical drug-resistance profiles. In addition L90M, which confers major resistance to multiple drugs, is common to both PR20 and PRS17. Unlike mutations V82S, proximal to the active site pocket, and G48V and I54V in the flap regions, non-active site mutations L90M and A71V may indirectly influence inhibitor or substrate binding. The latter mutations were proposed to distort the binding cavity due to increased volume of the side chains buried in the core region beneath the active site.30

Dimer dissociation constant (Kdimer) and kinetic parameters of mature PRS17

In earlier studies, we observed no strict correlation between Kdimer and stability of the tertiary fold, as measured by Tm. We were unable to make this comparison for PRS17 because no distinct peak by differential scanning calorimetry denoting the transition from a folded to an unfolded state was observed even in the presence of inhibitor DRV which is expected to enhance the overall stability of the dimer. Mature PRS17 exhibits a strictly linear relationship between protein concentration and catalytic activity, as measured by the initial rate of cleavage of a chromogenic substrate, from ~10 to >200 nM dimer concentration (Fig. S1A) consistent with a very low dissociation constant (Kdimer) that is comparable to the wild type PR (Kdimer <10 nM 2,3,4). This observation strongly suggests that most of the substitutions selected in PRS17, which map to the periphery of the dimer, with the exception of a few residues (see Fig. 1C), exert no long range effect on the dimer interface contacts to alter Kdimer.

Mature PRS17 shows ~3-fold more favorable affinity (Km = 193 ± 24 μM, Fig. 1B) for the chromogenic substrate than PR20 (Km = 617 ± 84 μM).16 However, kcat for PRS17 is about half the value estimated for PR20 [107 ± 6 (Fig. S1B) compared with 215 ± 19 min−1].16 Because of compensation between kcat and Km, the overall catalytic efficiency of both proteases is comparable (kcat/Km = 0.55 ± 0.1 and 0.35 ± 0.08 μM−1min−1 for PRS17 and PR20, respectively). Both these mutants, however, are ~10-fold less efficient catalytically than PR which exhibits a Km and kcat of 48 ± 3 μM and 173 ± 3 min−1 for the same substrate under identical conditions.31

Gag Processing mediated by mature PRS17

PR mediated processing of the full length Gag polyprotein to release the structural proteins required for viral assembly occurs via a stepwise process with initial cleavage between SP1 and NC to give two fragments, MA-CA-SP1 and NC-SP2-P61. Thus, an assay utilizing Gag polyprotein substrate provides a realistic assessment of the catalytic efficiency of the proteases by monitoring the rate and order of multiple cleavages in a single reaction. We made use of a simplified recombinant Gag construct (termed ΔGag, Fig. 2) lacking the SP2 and P6 regions in which the SP1-NC cleavage (step 1) is followed by a cleavage between MA and CA (step 2) to give mature MA and CA-SP1 intermediate.10,27 The CA-SP1 is converted by a third, final cleavage (step 3) to CA and SP1. As indicated in Figure 2, this order of cleavage is essentially the same for processing by PR and the two drug resistant variants PR20 and PRS17. SP1-NC cleavage monitored by the appearance of NC occurs most rapidly with PR. PR20 used for comparison with PRS17, exhibits ~4-fold lower activity than PR consistent with a previous study,10 based on the initial SP1-NC cleavage. The corresponding cleavage catalyzed by PRS17 is ~10-fold slower than for PR. MA/CA cleavage catalyzed by PR20 to release CA (and CA-SP1) is essentially complete after 30–40 min (middle gel, lanes 5–6) whereas significant quantities of MA-CA are still clearly detectable with PRS17 even up to 3.5 h. The third cleavage (CA-SP1, step 3) is barely discernible as a doublet at 60 min only in the digest with PRS17 (bottom panel) under the conditions employed and thus no comparison is made with the other digests using PR and PR20. These observations suggest that catalytic efficiency for both SP1-NC and MA-CA cleavages is roughly 2-fold lower for PRS17 relative to PR20. The difference in the catalytic efficiency for hydrolysis of a chromogenic substrate (~1.5-fold higher for PRS17 than PR20) does not exactly match the rate estimate for Gag processing by these 2 enzymes, possibly due to differences in substrate recognition when using a synthetic substrate and a natural polyprotein substrate. In this context, some mutations in these two constructs could play a role in initiating transient encounter complexes between Gag and PR leading to substrate recognition and cleavage. Similarly, a 10–35 fold decrease in the cleavage efficiency of peptide substrates with PR20, relative to PR, is reflected by only ~4-fold decrease in the rate of Gag processing with no change in cleavage order.11

Figure 2
ΔGag processing mediated by mature PR, PR20 and PRS17. (Left) Schematic representation of the sequential cleavages of ΔGag with the order of cleavage (blue), cleavage site positions and calculated molecular weight of the products indicated. ...

Binding affinity of clinical inhibitors to mature PRS17

Estimations of binding constants (Kassoc) of APV, IDV and DRV to PRS17 by isothermal titration calorimetry are shown in Figure S2. The dissociation constants (1/Kassoc) are equivalent to Ki for competitive inhibitors. Titration with SQV under the same experimental conditions did not give an interpretable sigmoidal isotherm because of a low c-value (product of PR concentration times Kassoc),32 due largely to its lower Kassoc of ~105 M−1 as compared to values ranging from 106–108 M−1 for the other three inhibitors. When satisfactory ITC data could not be obtained because of weak inhibitor binding or inadequate heat response, Ki values were determined kinetically (Fig. S3) from inhibition of hydrolysis of the chromogenic substrate in buffer B as described in Materials and Methods.

Experimental results for inhibition of PRS17 by PIs are summarized in Table 1 and Figure 3 and compared with previously published values for the wild type PR. As expected, there is no correlation between Ki values for inhibition of PRS17 and PR. The so-called first generation inhibitors, IDV, NFV, SQV, and RTV, designed to fit sterically into the active site cavity of PR, are hydrophobic, and their affinity is largely determined by their shape and favorable binding entropy.33 Newer, second-generation inhibitors such as DRV and LPV also take advantage of binding enthalpy due to specific interactions such as hydrogen bonds with the peptide backbone or residues in the binding cavity, and exhibit higher affinity to wild-type PR. The smallest difference (55-fold) in binding to PRS17 relative to PR is observed for APV, in part because it exhibits the tightest binding to PRS17 along with relatively weaker binding affinity (high pM) to PR. Although IDV, NFV and SQV show similar high pM binding to PR, these inhibitors exhibit a larger difference between PR and PRS17 because of their very poor binding (μM) to PRS17. Similarly ATV and LPV show Ki for PR in the mid-pM range, and 2000 to 4000-fold weaker inhibition of PRS17. We note that PI resistance of PRS17 relative to PR (PRS17/PR) increases with a similar trend to the Ki for binding of each inhibitor to PRS17 (see also Figure 3B and C), with the exceptions of DRV and RTV.

Figure 3
Plots comparing the Ki values for the inhibition of PR and PRS17 and the relative resistance (Ki for PRS17/Ki for PR) for each clinical inhibitor. Numerical values are listed in Table 1.
Table 1
Comparison of the inhibition of mature PRS17 and PR, and autoprocessing of TFR-PRS17 precursor at the p6pol/PRS17 site, with a panel of clinical inhibitors. The inhibitors are listed in the order of increasing Ki (decreasing affinity) for PRS17. Estimated ...

Comparison of PRS17 with PR20

PR20 and PRS17 have six mutations in common including L90M in cluster 3 associated with resistance to all PIs used in this study except TPV and DRV. For comparisons between the two proteases, we chose second-generation inhibitors APV and DRV, and the first-generation inhibitor SQV, which differ in their binding affinity to both wild type and drug resistant proteases. Notably DRV, which exhibits pM affinity to wild type PR, binds to both PR20 and PRS17 with 4 orders of magnitude poorer affinity (40–50 nM), approximately the same for both drug resistant proteases. Although PR20 bears major DRV resistance mutations V32I, I47V, I54L, there are no corresponding mutations in PRS17 to explain the similar Ki values shown by these 2 enzymes. In contrast, APV binds to PRS17 with 16-fold higher affinity (Ki ~11 nM) relative to PR20 (Ki ~178 nM34). This difference may relate to major APV drug resistance mutations V32I, I47V, I84V present in PR20 but not in PRS17 (see Table S1). Similarly SQV binds to PRS17 with ~10-fold poorer affinity relative to PR20, possibly also reflecting effects of SQV resistance-associated mutations G48V and I54V in PRS17 which are not present in PR20.

PR20 cluster 1 encompasses three mutations contiguous to the active site, D30N, V32I and L33F, that are not present in PRS17. Other mutations in this region, E35D and M36I in both mutants, and S37D in PRS17, are located in the flap hinge loop and likely contribute to altered flap conformation and dynamics.17 Mutations I62V and L63P in cluster 2 are common to PR20 and PRS17, and interestingly, also occur naturally in PR group O.35 L63P has been identified in drug resistant PR mutants30,36 and is classified as a minor drug resistance mutation, possibly associated with resistance to LPV.29 However a basis for relating the structural effects of these two mutations to drug resistance is lacking. Outside of the clusters, mutations L10F in PR2010,1517 and possibly L10I in PRS17 could disrupt the ion pair between R8 and D29’ of the opposite subunit,37 resulting in altered conformation of the binding cavity. I54L (in PR20) and I54V (in PRS17) are major mutations associated with high-level resistance to most PIs, with the possible exception of DRV (for I54V).

Comparison of PRS17 with PR22

Another clinically derived, extremely drug resistant PR, bearing 22 mutations (termed PR22 for consistency with the present numbering system) has been described.11 Interestingly, comparing the location of mutations in PR20, PRS17 and PR22 indicates that clinical isolates PR22 and PR20 are more alike than PRS17 (Fig. S4). Mutations corresponding to clusters 1 and 3 of PR20 are evident in PR22 whereas cluster 2 and a partial cluster 1 (residues 35–37) in PRS17 are in common with PR20. Only five of the mutations in PR22 are identical with those in PRS17, namely, E35D, M36I, S37D, I54V, and L90M. Of these, I54V and L90M are major drug resistance mutations. Conservative substitutions L10V and L10I in PR22 and PRS17 respectively, are classified as minor resistance mutations. Interestingly, in spite of the differences between the two PR sequences, a similar trend of high resistance (large Ki) to SQV, RTV, IDV and NFV in comparison with second-generation inhibitors LPV, ATV, APV and DRV was observed for both PRS17 (Table 1 in this study) and PR22.11 Whether the extremely high resistance of PRS17 is indeed a consequence of an open-closed flap equilibrium shifted towards an open conformation as in PR2017,38 and MDR 76939, relative to the mainly closed conformation of wild type PR,17,38 is currently under investigation. Furthermore, multiple accessory mutations (13 out of 22 in PR22) that are not classified as DRMs may contribute to the viability of highly evolved proteases through indirect effects on the structure and dynamics of the mature protease and its precursor.

Autoprocessing of PRS17 precursor and inhibition by clinical inhibitors

The protease catalyzes its own release at its N terminus from its precursor Gag-Pol, which precedes the C-terminal cleavage, termed autoprocessing (Figure 1A and and44).40 The protease is flanked at the N terminus by the 56 amino acid transframe region (TFR) comprising transframe peptide (TFP) and 48 amino acid p6pol, both separated by a cleavage site. N-terminal autoprocessing is characterized by two pH-dependent cleavages at the TFP/p6pol and p6pol/PR sites occurring independently. At pH 4, the ionization of a group, either in the active site or the EDL sequence within TFP, with a pKa of 3.8 makes the cleavage at the TFP/p6pol site proceed faster than the N-terminal p6pol/PR cleavage.22,41,42 At pH 6, the reverse is evident. However, cleavage at the N-terminal p6pol/PR site is crucial for stable dimer formation (Kdimer < 10 nM) and ensuing catalytic activity.22,42 Thus, at pH 4, there is a lag in the appearance of catalytic activity coinciding with the accumulation of the intermediate p6pol/PR. Subsequent cleavage at the p6pol/PR site is concomitant with mature-like catalytic activity 22,42 At pH 5, both cleavages occur simultaneously, p6pol/PR cleavage being slightly faster. A similar pH dependent stepwise N-terminal autoprocessing from a model precursor was also described for the protease from HIV-1 group N.23

Figure 4
Inhibition of N-terminal autoprocessing of TFP-p6pol-PRS17 precursor by clinical PIs. The precursor mimetic spanning the 2 cleavage sites at the N-terminus of PR and its cleavage products during autoprocessing is shown schematically on top. In wild type ...

Here we have assessed the approximate IC50 values for inhibition of N-terminal autoprocessing by the same 8 PIs used to monitor inhibition of mature PRS17. Partial accumulation of the TFR-PRS17 occurs when expressed in the presence of ~25 μM DRV. Thus, the TFR-PRS17 precursor was purified from inclusion bodies similar to the protocol described for mature PRS17. Inhibition of autoprocessing was evaluated by folding the precursor to a final concentration of 8 μM to initiate autoprocessing in the presence of various PIs. This methodology provided an ideal window to monitor the disappearance of the full length precursor and appearance of products by SDS-PAGE and staining as compiled and shown in Figure 4. Our kinetics and NMR analysis indicate that only a small fraction of precursor exhibits a folded structure, consistent with a transient process of folding/dimerization concomitant with N-terminal autoprocessing in the early phase of the reaction.22,42,43 Thus the effective concentration of folded precursor at the onset of the assay is expected to be at least one order of magnitude lower than 8 μM.

TFR-PRS17 undergoes efficient autoprocessing similar to the wild-type precursor construct (Fig. 4, compare wild type control and 4A).22,42 A majority of precursor is converted to products PRS17, p6pol and TFP within 10 min. Figure 4B and C show the initial screening of inhibition of TFR-PRS17 by 8 PIs at a concentration of 50 μM, except for the maximum permitted solubility of ATV and NFV at 42 μM under these conditions. After 60 min, APV, ATV and DRV showed almost complete inhibition of autoprocessing. IDV showed partial inhibition whereas LPV, NFV, SQV and RTV showed almost no inhibition. RTV at 105 μM, IDV at 50 μM and LPV at 70 μM (limit of solubility) showed ~50% inhibition of the p6pol/PR site and insignificant inhibition of the TFP/p6pol site while SQV even at 105 μM showed almost no inhibition. Next, dose- response experiments were carried out for a total reaction time of 60 min in the presence of 2–40 μM APV, ATV and DRV (Fig. 4E to 4G) and the approximate IC50 values are summarized in Fig. 5A and Table 1. APV, DRV and ATV elicit inhibition with an IC50 in the range of 7.5 μM for APV and 15 μM for DRV and ATV as judged by monitoring the appearance of product bands PRS17 and p6pol. Interestingly, however, cleavage of the TFP/p6pol site is less susceptible to inhibition by APV and ATV (higher IC50 value) evident by the accumulation of the p6pol/PR intermediate (purple arrows, Fig. 4E and 4F). The IC50 for the inhibition of TFP/p6pol cleavage by APV and ATV is estimated to be 15 and >42 μM, respectively. The exclusive lack of accumulation of the p6pol/PR intermediate suggests that DRV inhibits both these cleavage steps equally efficiently with an IC50 of ~15 μM (Fig. 4G). A plot of the differential inhibition of autoprocessing at the TFP/p6pol and p6pol/PR sites by APV, ATV and DRV is shown in Fig. 4H. Thus, cumulative inhibition of autoprocessing steps of PRS17 precursor is most effective by APV and DRV followed by ATV. However, such differentiation of inhibition by PIs is not feasible to monitor for the wild-type precursor because of lower IC50 in the range of 1–2 μM.16 LPV ranks fourth with an IC50 of ~60 μM and >70 μM for blocking autoprocessing at the p6pol/PRS17 and TFP/p6pol sites, respectively.

Figure 5
Plots comparing the inhibition of precursor autoprocessing at p6pol-PRS17 and TFP-p6pol sites and with mature PRS17. Plots of approximate IC50 values for inhibition of autoprocessing of TFP-p6pol-PRS17 at the p6pol/PRS17 (A) and TFP/p6pol (C) sites by ...

Overall, observed IC50 values for inhibition of autoprocessing by PIs are significantly larger than the corresponding Ki values for inhibition of substrate hydrolysis by the mature enzymes. The tightest binders (APV, DRV, ATV and LPV) exhibit differences of ~200 to 800-fold and 300 to 1400-fold for the inhibition of autoprocessing at the TFP/p6pol and p6pol/PRS17 sites, respectively, relative to mature PRS17 (Fig 5 and Table 1). The inhibition of autoprocessing relative to mature activity by the weaker binders, as exemplified by SQV, differs by only ~10 to 100- fold. This suggests that different classes of inhibitors differ in their relative affinity for PRS17 and for its precursor depending on the effects of mutations on differences in the active site environment of the mature PR relative to the corresponding precursor. With the exception of DRV and NFV, the rest of the PIs exhibit differential inhibition of the 2 autoprocessing sites, with weaker inhibition of the TFP-p6pol site. It is worth noting that even though DRV is 2-fold less efficient than APV for blocking cleavage at the p6pol/PRS17 site, it is unique in blocking both sites with equal efficiency. It is unclear if this phenomenon relates to the second binding site of DRV,44 by targeting the flap of the precursor monomer, because APV, a structural analogue of DRV shows kinetics of inhibition of mature PR similar to DRV consistent with a second binding site.44

The initial phase of the N-terminal autoprocessing reaction was shown to be first-order in protein concentration indicating an intramolecular process.40,42 As the reaction progresses, when sufficient active mature protease accumulates, an intermolecular process, involving the released mature protease cleaving the precursor, that becomes competitive with the intramolecular process could prevail.40 Notably, complete autoprocessing (Figure 4E and F) occurs even in the presence of 5 μM APV and 10 μM ATV, ~500- and 100-fold higher in concentration than the estimated Ki for the binding of APV and ATV, respectively, to mature PRS17, strongly suggesting that N-terminal autoprocessing proceeds even when the action of the released mature protease is inhibited. The TFP/p6pol site, 48 amino acids upstream to the protease, is inhibited even more weakly, relative to the p6pol/PRS17 site. In this context, it is worth noting that in vitro inhibition of the SP1/NC site of the wild type Gag-Pol precursor which was shown to precede the N-terminal autoprocessing sites, TFP/p6pol and p6pol/PR sites, is inhibited by RTV ~10,000-fold weaker than the corresponding mature protease,1,45 and similarly, significantly higher concentrations of PIs exert inhibition of a chimeric wild type TFR-PR precursor autoprocessing in cell culture.46 These observations are consistent with the much weaker inhibition (IC50 of 1–2 μM) of autoprocessing of the wild type TFR-PR precursor by DRV and SQV during expression in E.coli,16 drastic increases in IC50 of 8 PIs with the drug resistant TFR-PRS17 precursor described in this study, and no inhibition as in extreme instances of drug resistance such as the PR20 precursor, up to 150 μM DRV or SQV.16

CONCLUDING REMARKS

This study systematically validates the functionality of a rationally selected HIV-1 protease PRS17 predicted to have high level drug resistance. In spite of a different mutation profile compared to the extremely drug resistant clinical isolates with multiple mutations, PR20 and PR22, PRS17 exhibits high resistance to 8 of the potent clinical inhibitors tested thus highlighting the malleability of the protease. Resistance of PRS17 ranges from 1.5 to 5 orders of magnitude increase in Ki relative to PR. A long term focus of our studies has been to compare the potency of PIs on the N-terminal autoprocessing reaction, a process crucial for shifting the equilibrium from a largely unfolded population of the precursor to a stably folded dimer, essential for mature-like catalytic activity. The accumulation of the TFP-p6pol-PRS17 precursor in the presence of PIs during expression permitted its isolation for in vitro comparison of inhibition of the autoprocessing reaction with the corresponding mature enzyme. PRS17 precursor presented a fortuitous window between the wild type precursor and the PR20 precursor ideal for estimation of the IC50 values for inhibition of autoprocessing with inhibitor concentrations maintained well above the precursor concentration to exclude active site titration. The weaker binding of PIs to the precursor points to an active site environment mediating autoprocessing that is different from the mature PR, as well as an unfavorable entropy for PIs to compete with an intramolecular process.

Structural studies by crystallography and NMR would help to reveal if extreme resistance of mature PRS17 correlates with an open flap conformation as noted for PR20 and MDR 769, contrary to the mainly closed conformation of the wild type PR.17,38,39 Design of inhibitors with high affinity to extremely drug-resistant mutants such as the three described in this study presents a significant challenge because of the continuously evolving mutations under drug pressure. A database created by comparing the relative binding affinity of a panel of PIs to mutants such as PR2016, P5116 and PR2211 may aid in selecting the drug regimen for treatment as exemplified by this study. Future attempts in defining inhibitors that block the N-terminal autoprocessing with IC50 values in the low nM range, at least 1-order of magnitude better than those described here, will benefit by exploring structures of the protease bound to drugs prior to its cleavage at its N-terminus such as the model precursor TFR-PR and analogues thereof. Such studies will be challenging, however, because precursor maturation occurs via transient dimer formation from a mainly unfolded population.43,47

Supplementary Material

SI

Acknowledgments

Funding: This research was supported by the Intramural Research Program of the NIDDK, National Institutes of Health and the Intramural AIDS-Targeted Program of the Office of the Director, NIH, the National Institutes of Health grant GM062920 (ITW and RWH) and by a fellowship from the Georgia State University Molecular Basis of Disease Program (XY).

We thank Lalit Deshmukh for providing recombinant ΔGag polyprotein and discussions regarding the in vitro assay. Clinical protease inhibitors (PIs) used in this study were obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. We acknowledge use of the National Institute of Diabetes and Digestive and Kidney Diseases Advanced Mass Spectrometry Core Facility.

Abbreviations

HIV-1
human immunodeficiency virus type 1
PR
pseudo wild-type mature HIV-1 protease containing the mutations Q7K, L33I, L63I, C67A, C95A
PI(s)
protease inhibitor(s)
APV
were amprenavir
DRV
darunavir
ATV
atazanavir
LPV
lopinavir
IDV
indinavir
NFV
nelfinavir
RTV
ritonavir
SQV
saquinavir
TPV
tipranavir

Footnotes

SUPPORTING INFORMATION

One table comparing drug resistance mutations in PRS17 with PR, PR20 and PR22 and four figures pertaining to the estimation of kinetic parameters kcat and Km of PRS17, Ki measurement by enzyme kinetics and ITC of PRS17, and structure representations of mature PRS17, PR20 and PR22 showing the positions of mutations matching with table details.

References

1. Lee SK, Potempa M, Swanstrom R. The choreography of HIV-1 proteolytic processing and virion assembly. J Biol Chem. 2012;287:40867–40874. [PMC free article] [PubMed]
2. Louis JM, Weber IT, Tozser J, Clore GM, Gronenborn AM. HIV-1 protease: maturation, enzyme specificity, and drug resistance. Adv Pharmacol. 2000;49:111–146. [PubMed]
3. Louis JM, Ishima R, Torchia DA, Weber IT. HIV-1 protease: structure, dynamics, and inhibition. Adv Pharmacol. 2007;55:261–298. [PubMed]
4. Weber IT, Agniswamy J. HIV-1 Protease: Structural Perspectives on Drug Resistance. Viruses. 2009;1:1110–1136. [PMC free article] [PubMed]
5. Menendez-Arias L. Molecular basis of human immunodeficiency virus type 1 drug resistance: overview and recent developments. Antiviral Res. 2013;98:93–120. [PubMed]
6. Perelson AS, Neumann AU, Markowitz M, Leonard JM, Ho DD. HIV-1 dynamics in vivo: virion clearance rate, infected cell life-span, and viral generation time. Science. 1996;271:1582–1586. [PubMed]
7. Hu WS, Hughes SH. HIV-1 reverse transcription. Cold Spring Harb Perspect Med. 2012;2 [PMC free article] [PubMed]
8. Weber IT, Kneller DW, Wong-Sam A. Highly resistant HIV-1 proteases and strategies for their inhibition. Future Med Chem. 2015;7:1023–1038. [PMC free article] [PubMed]
9. Fun A, Wensing AM, Verheyen J, Nijhuis M. Human Immunodeficiency Virus Gag and protease: partners in resistance. Retrovirology. 2012;9:63. [PMC free article] [PubMed]
10. Louis JM, Deshmukh L, Sayer JM, Aniana A, Clore GM. Mutations Proximal to Sites of Autoproteolysis and the alpha-Helix That Co-evolve under Drug Pressure Modulate the Autoprocessing and Vitality of HIV-1 Protease. Biochemistry. 2015;54:5414–5424. [PubMed]
11. Kozisek M, Henke S, Saskova KG, Jacobs GB, Schuch A, Buchholz B, Muller V, Krausslich HG, Rezacova P, Konvalinka J, Bodem J. Mutations in HIV-1 gag and pol compensate for the loss of viral fitness caused by a highly mutated protease. Antimicrob Agents Chemother. 2012;56:4320–4330. [PMC free article] [PubMed]
12. Quiñones-Mateu ME, Weber J, Rangel HR, Chakraborty B. HIV-1 Fitness and Antiviral Drug Resistance. AIDS Rev. 2001;3:223–234.
13. Foulkes-Murzycki JE, Rosi C, Kurt YN, Shafer RW, Schiffer CA. Cooperative effects of drug-resistance mutations in the flap region of HIV-1 protease. ACS Chem Biol. 2013;8:513–518. [PMC free article] [PubMed]
14. Cai Y, Myint W, Paulsen JL, Schiffer CA, Ishima R, Kurt YN. Drug Resistance Mutations Alter Dynamics of Inhibitor-Bound HIV-1 Protease. J Chem Theory Comput. 2014;10:3438–3448. [PMC free article] [PubMed]
15. Dierynck I, De Wit M, Gustin E, Keuleers I, Vandersmissen J, Hallenberger S, Hertogs K. Binding kinetics of darunavir to human immunodeficiency virus type 1 protease explain the potent antiviral activity and high genetic barrier. J Virol. 2007;81:13845–13851. [PMC free article] [PubMed]
16. Louis JM, Aniana A, Weber IT, Sayer JM. Inhibition of autoprocessing of natural variants and multidrug resistant mutant precursors of HIV-1 protease by clinical inhibitors. Proc Natl Acad Sci U S A. 2011;108:9072–9077. [PubMed]
17. Agniswamy J, Shen CH, Aniana A, Sayer JM, Louis JM, Weber IT. HIV-1 protease with 20 mutations exhibits extreme resistance to clinical inhibitors through coordinated structural rearrangements. Biochemistry. 2012;51:2819–2828. [PMC free article] [PubMed]
18. Prosperi MC, De Luca A. Computational models for prediction of response to antiretroviral therapies. AIDS Rev. 2012;14:145–153. [PubMed]
19. Yu X, Weber IT, Harrison RW. Sparse Representation for Prediction of HIV-1 Protease Drug Resistance. Proc SIAM Int Conf Data Min. 2013;2013:342–349. [PMC free article] [PubMed]
20. Yu X, Weber IT, Harrison RW. Prediction of HIV drug resistance from genotype with encoded three-dimensional protein structure. BMC Genomics. 2014;15(Suppl 5):S1, 1–13. [PMC free article] [PubMed]
21. Yu X, Weber IT, Harrison RW. Identifying representative drug resistant mutants of HIV. BMC Bioinformatics. 2015;16(Suppl 17):S1, 1–11. [PMC free article] [PubMed]
22. Louis JM, Clore GM, Gronenborn AM. Autoprocessing of HIV-1 protease is tightly coupled to protein folding. Nat Struct Biol. 1999;6:868–875. [PubMed]
23. Sayer JM, Agniswamy J, Weber IT, Louis JM. Autocatalytic maturation, physical/chemical properties, and crystal structure of group N HIV-1 protease: relevance to drug resistance. Protein Sci. 2010;19:2055–2072. [PubMed]
24. Louis JM, Ishima R, Aniana A, Sayer JM. Revealing the dimer dissociation and existence of a folded monomer of the mature HIV-2 protease. Protein Sci. 2009;18:2442–2453. [PubMed]
25. Richards AD, Phylip LH, Farmerie WG, Scarborough PE, Alvarez A, Dunn BM, Hirel PH, Konvalinka J, Strop P, Pavlickova L. Sensitive, soluble chromogenic substrates for HIV-1 proteinase. J Biol Chem. 1990;265:7733–7736. [PubMed]
26. Ishima R, Torchia DA, Louis JM. Mutational and structural studies aimed at characterizing the monomer of HIV-1 protease and its precursor. J Biol Chem. 2007;282:17190–17199. [PubMed]
27. Deshmukh L, Ghirlando R, Clore GM. Conformation and dynamics of the Gag polyprotein of the human immunodeficiency virus 1 studied by NMR spectroscopy. Proc Natl Acad Sci U S A. 2015;112:3374–3379. [PubMed]
28. Ceccherini-Silberstein F, Erba F, Gago F, Bertoli A, Forbici F, Bellocchi MC, Gori C, d’Arrigo R, Marcon L, Balotta C, Antinori A, Monforte AD, Perno CF. Identification of the minimal conserved structure of HIV-1 protease in the presence and absence of drug pressure. AIDS. 2004;18:F11–F19. [PubMed]
29. Wensing AM, Calvez V, Gunthard HF, Johnson VA, Paredes R, Pillay D, Shafer RW, Richman DD. 2014 Update of the drug resistance mutations in HIV-1. Top Antivir Med. 2014;22:642–650. [PMC free article] [PubMed]
30. Muzammil S, Ross P, Freire E. A major role for a set of non-active site mutations in the development of HIV-1 protease drug resistance. Biochemistry. 2003;42:631–638. [PubMed]
31. Ishima R, Torchia DA, Lynch SM, Gronenborn AM, Louis JM. Solution structure of the mature HIV-1 protease monomer: insight into the tertiary fold and stability of a precursor. J Biol Chem. 2003;278:43311–43319. [PubMed]
32. Wiseman T, Williston S, Brandts JF, Lin LN. Rapid measurement of binding constants and heats of binding using a new titration calorimeter. Anal Biochem. 1989;179:131–137. [PubMed]
33. Velazquez-Campoy A, Muzammil S, Ohtaka H, Schon A, Vega S, Freire E. Structural and thermodynamic basis of resistance to HIV-1 protease inhibition: implications for inhibitor design. Curr Drug Targets Infect Disord. 2003;3:311–328. [PubMed]
34. Agniswamy J, Shen CH, Wang YF, Ghosh AK, Rao KV, Xu CX, Sayer JM, Louis JM, Weber IT. Extreme multidrug resistant HIV-1 protease with 20 mutations is resistant to novel protease inhibitors with P1’-pyrrolidinone or P2-tris-tetrahydrofuran. J Med Chem. 2013;56:4017–4027. [PMC free article] [PubMed]
35. Kane CT, Montavon C, Toure MA, Faye MA, Ndiaye AG, Diallo AG, Ndoye I, Liegeois F, Delaporte E, Mboup S, Peeters M. Full-length genome sequencing of HIV type 1 group O viruses isolated from a heterosexual transmission cluster in Senegal. AIDS Res Hum Retroviruses. 2001;17:1211–1216. [PubMed]
36. Koh Y, Amano M, Towata T, Danish M, Leshchenko-Yashchuk S, Das D, Nakayama M, Tojo Y, Ghosh AK, Mitsuya H. In vitro selection of highly darunavir-resistant and replication-competent HIV-1 variants by using a mixture of clinical HIV-1 isolates resistant to multiple conventional protease inhibitors. J Virol. 2010;84:11961–11969. [PMC free article] [PubMed]
37. Louis JM, Ishima R, Nesheiwat I, Pannell LK, Lynch SM, Torchia DA, Gronenborn AM. Revisiting monomeric HIV-1 protease. Characterization and redesign for improved properties. J Biol Chem. 2003;278:6085–6092. [PubMed]
38. Roche J, Louis JM, Bax A. Conformation of inhibitor-free HIV-1 protease derived from NMR spectroscopy in a weakly oriented solution. Chembiochem. 2015;16:214–218. [PMC free article] [PubMed]
39. Martin P, Vickrey JF, Proteasa G, Jimenez YL, Wawrzak Z, Winters MA, Merigan TC, Kovari LC. “Wide-open” 1.3 A structure of a multidrug-resistant HIV-1 protease as a drug target. Structure. 2005;13:1887–1895. [PubMed]
40. Louis JM, Nashed NT, Parris KD, Kimmel AR, Jerina DM. Kinetics and mechanism of autoprocessing of human immunodeficiency virus type 1 protease from an analog of the Gag-Pol polyprotein. Proc Natl Acad Sci U S A. 1994;91:7970–7974. [PubMed]
41. Louis JM, Dyda F, Nashed NT, Kimmel AR, Davies DR. Hydrophilic peptides derived from the transframe region of Gag-Pol inhibit the HIV-1 protease. Biochemistry. 1998;37:2105–2110. [PubMed]
42. Louis JM, Wondrak EM, Kimmel AR, Wingfield PT, Nashed NT. Proteolytic processing of HIV-1 protease precursor, kinetics and mechanism. J Biol Chem. 1999;274:23437–23442. [PubMed]
43. Tang C, Louis JM, Aniana A, Suh JY, Clore GM. Visualizing transient events in amino-terminal autoprocessing of HIV-1 protease. Nature. 2008;455:693–696. [PMC free article] [PubMed]
44. Kovalevsky AY, Ghosh AK, Weber IT. Solution kinetics measurements suggest HIV-1 protease has two binding sites for darunavir and amprenavir. J Med Chem. 2008;51:6599–6603. [PMC free article] [PubMed]
45. Pettit SC, Everitt LE, Choudhury S, Dunn BM, Kaplan AH. Initial cleavage of the human immunodeficiency virus type 1 GagPol precursor by its activated protease occurs by an intramolecular mechanism. J Virol. 2004;78:8477–8485. [PMC free article] [PubMed]
46. Huang L, Chen C. Autoprocessing of human immunodeficiency virus type 1 protease miniprecursor fusions in mammalian cells. AIDS Res Ther. 2010;7:27. [PMC free article] [PubMed]
47. Sadiq SK, Noe F, De Fabtittis G. Kinetic characterization of the critical step in HIV-1 protease maturation. Proc Natl Acad Sci U S A. 2012;109:20449–20454. [PubMed]
48. Ohtaka H, Schon A, Freire E. Multidrug resistance to HIV-1 protease inhibition requires cooperative coupling between distal mutations. Biochemistry. 2003;42:13659–13666. [PubMed]
49. King NM, Prabu-Jeyabalan M, Bandaranayake RM, Nalam MN, Nalivaika EA, Ozen A, Haliloglu T, Yilmaz NK, Schiffer CA. Extreme entropy-enthalpy compensation in a drug-resistant variant of HIV-1 protease. ACS Chem Biol. 2012;7:1536–1546. [PMC free article] [PubMed]
50. Muzammil S, Armstrong AA, Kang LW, Jakalian A, Bonneau PR, Schmelmer V, Amzel LM, Freire E. Unique thermodynamic response of tipranavir to human immunodeficiency virus type 1 protease drug resistance mutations. J Virol. 2007;81:5144–5154. [PMC free article] [PubMed]