Effects of PRE and POST Therapy Drug-Pressure Selected Mutations on HIV-1 Protease Conformational Sampling

Abstract

Conformational sampling of pre- and post-therapy subtype B HIV-1 protease sequences derived from a pediatric subject infected via maternal transmission with HIV-1 were characterized by double electron-electron resonance spectroscopy. The conformational ensemble of the PRE construct resembles native-like inhibitor bound states. In contrast, the POST construct, which contains accumulated drug-pressure selected mutations, has a predominantly semi-open conformational ensemble, with increased populations of open-like states. The single point mutant L63P, which is contained in PRE and POST, has decreased dynamics, particularly in the flap region, and also displays a closed-like conformation of inhibitor-bound states. These findings support our hypothesis that secondary mutations accumulate in HIV-1 protease to shift conformational sampling to stabilize open-like conformations, while maintaining the predominant semi-open conformation for activity.

1. Introduction

HIV-1 protease (HIV-1 PR) is a homodimeric aspartyl protease that plays a fundamental role in the HIV-1 viral life cycle and infection [1]. Specifically, this enzyme cleaves the polyproteins gag and gag-pol into functional viral proteins necessary for mature infections viral particles. As such, HIV-1 PR is an active target in antiretroviral therapy [2]. Protease inhibitors (PIs) are designed as competitive inhibitors that bind within the active site cavity. Figure 1 shows a ribbon diagram of the structure of HIV-1 PR highlighting the active site aspartic acid residues and the β-hairpin “flaps” that modulate access to this pocket. HIV-1 PR is known to undergo significant conformational changes during the catalytic cycle, with various mechanisms of flap motion being described for substrate and inhibitor entry, as well as product release [3,4].

Figure 1.

Ribbon diagram of HIV-1PR (PDB ID 2PBX) with “flaps” and K55R1 reporter sites shown in tan, active site catalytic D25 residues in purple, the hinge regions in red, and L63 and A71 residues as green and pink; respectively.

We have been pioneering the usage of pulsed electron double resonance (PELDOR), also commonly referred to as double electron-electron resonance (DEER), in combination with site-directed spin-labeling (SDSL), to characterize the conformational sampling of HIV-1 PR under various conditions [5–13]. DEER measures the strength of the dipolar coupling between two electron spins spaced at distances between 20–60 Å, where the coupling is proportional to r⁻³, with r representing the interspin distance [14,15].

Using the spin label (1-oxyl-2,2,5,5-tetramethyl-Δ3-pyrroline-3-methyl) methane-thiosulfonate (MTSL) at site K55, referred to as K55R1, distinct ensembles for HIV-1 PR have been described. These are designated as semi-open, closed (or closed-like), wide-open and, tucked/curled (or curled-open); where ensemble assignments are based upon structural insights from molecular dynamic simulations and X-ray crystallography [4,12,13,16].

Here we investigate conformational sampling for three protease variants. The variants referred to as PRE and POST represent subtype B alleles derived from a pediatric subject before and after PI treatment; respectively, where HIV infection was transmitted via maternal transmission [17]. The gag-pol alleles were isolated from serial blood samples obtained over 7 years, starting before therapy initiation (PRE) and after the development of multiple drug resistance following 77 weeks of PI therapy with Ritonavir (RTV) and an additional 16 weeks with Indinavir (IDV) (POST) [17,18]. The effects of the single secondary mutations, L63P, on the conformational sampling of subtype B (LAI) sequence was also investigated because it is a common natural polymorphism contained within the PRE sequence.

2. Materials and Methods

2.1 Cloning and mutagenesis

DNA that encodes for E. coli codon optimized subtype B PRE and POST sequences were purchased from DNA 2.0 (Menlo Park, CA). Each construct was cloned into pET-23a vector (Novagen, Madison WI) under the control of a T7 promoter. The L63P construct was generated using site-directed mutagenesis kit (Strategene) and made within the subtype B background of a pentamutated protease containing the stabilizing mutations Q7K, L33I, L63I and substitution of native cysteines to alanine: C67A, C95A. PRE and POST also contain the C67A and C95A mutations. All samples for DEER spectroscopy also contain the D25N mutation and K55C for labeling. The fidelity of the HIV-1 PR genes was confirmed by Sanger DNA sequencing (ICBR Genomics Facility, University of Florida). L63P for NMR studies did not contain the K55C mutation. All sequences are given in Table S1.

2.2 Protein Expression, purification and spin-labeling

Protein expression for EPR and NMR measurements proceeded as described previously [5,6,12,13] with the following modifications: the inclusion body resuspension buffer pH used for anion exchange depends upon the isoelectric point of a given construct. Buffers were adjusted to pH 9.06 for PRE and POST and pH 9.39 for L63P. For spin-labeling, MTSL dissolved in ethanol is added in three- to four-fold molar excess to protein in 10 mM Tris-HCl pH 6.9 buffer. The reaction is allowed to proceed in the dark for 6–12 hours. Excess spin-label is removed by buffer exchange into 2 mM NaOAC pH 5.0 using HiPrep 26/10 desalting columns. Electrospray mass spectrometry results and obtained masses of labeled proteins are given in Figures S7–9 and Table S2.

2.3 Sample preparation, DEER data collection and analysis

For DEER spectroscopy, samples were prepared at 100 mM HIV-1 PR homodimer in 20 mM D₃-NaOAc/D₂O, pH 5.0 with 30% v/v D₈-glycerol as described previously [10,12]. Details of data collection and sample analysis have been described in detail elsewhere [10,12,19]. Complete data analysis is given in Supporting Information.

2.4 Sample preparation, NMR data collection and analysis

All spectra were obtained at 293 K. All 600 MHz measurements were made on a Bruker Avance spectrometer with a 5 mm TXI cryoprobe (AMRIS Facility, University of Florida). All 700 MHz data were collected on a Bruker Avance system with a 5 mm TCI 700S4 h-C/N-D-05Z Cryoprobe (NHMFL facility, Department of Chemistry and Biochemistry, Florida State University,). NMRPipe [20] and Sparky (Goddard and Kneller, Sparky 3, UCSF, San Francisco) were used for processing and analysis of all NMR data. Given that the L63P construct contains only one mutation compared to subtype B, assignments were made based upon previously published results [21].

Relaxation measurements were performed on ¹⁵N L63P at 600 MHz and 700 MHz as described previously [12]. Experimental errors for NOE values were evaluated by error propagation using resonance intensity errors. Consistency of measurements performed at different magnetic field strengths was confirmed as proposed by Morin et al. [22]. Model-free analysis was performed using relaxGUI (http://www.nmr-relax.com/) [23–25].

3. Results and Discussion

3.1. Altered conformational sampling in PRE and POST

The PRE construct differs from subtype B-LAI sequence by four polymorphisms including S37T, P39T, I62V, and L63P, likely the result of natural genetic drift in the mother of the infected child. The POST construct contains 8 additional drug-selected polymorphisms, including L10I, I15V, E34Q (negative to uncharged), M36I, T37N, I54A, R57E (positive to negative), and V82A. The locations of these mutations are shown in Figure S2 and sequences given in Table S1.

The DEER results in Figure 2. Figure 2 shows the background corrected dipolar modulated echo curves (Fig 2A) and corresponding TKR distance profiles (Figure 2B). Data were analyzed with Deer Analysis 2011 [15] and subsequent Gaussian population analysis as described in detail elsewhere [10,12]. Full details of data analysis are given in the Supporting information. It is clear from these data that the PRE construct has a most probable distance of 33 Å, which corresponds well with a closed-like ensemble, typically seen upon addition of inhibitors [6–9]. In contrast, a most probable distance of 37 Å is seen for POST, similar to the most probable distance observed in subtype C and the drug resistant construct MDR769 [9,11,13]. It is noteworthy that both distance profiles differ markedly from that of subtype B, which has a most probable distance near 35 Å. Table 1 summarizes the DEER results for these four variants.

FIGURE 2.

(A) Background corrected DEER echo curves for PRE (grey), L63P (red), A71V (blue), and POST (orange) constructs. Solid lines through data are fits from Tikhonov regularization (TKR) procedures. (B) Distance profiles from DEER2011, where dotted lines indicate the distances consistent with the closed-like (33 Å) and semi-open (36 Å) populations. (C) Gaussian populations used reconstructing the TKR distance profiles for L63P, showing contributions from individual conformations described as curled/tucked (black), closed-like (red), semi-open (blue), and wide-open (green). In (A) and (B), data are offset for clarity.

Table 1.

Summary of distance parameters obtained from DEER distance profiles of HIV-1PR constructs.

Construct	range (span) (± 1 Å)	most prob. dist. (± 0.2 Å)	avg. dist. (± 0.2 Å)
Subtype B*	24–45 (21)	35.2	35.2
PRE	25–45 (20)	33.2	31.9
POST	27–41 (14)	37.3	34.5
L63P	31–43 (12)	33.2	34.4
A71V§	31–43 (12)	33.0	33.2

^*Data from ref. [13]

^§Data taken from ref.[10]

To interrogate what residue(s) may be responsible for the shift in conformation to a close-like state, we generated the single mutant, L36P, which is a common natural polymorphism and a residue that has been studied with molecular dynamics simulations suggesting this site may affect the conformational flexibility [26]. Results in Figure 2 show that L63P adopts a conformation with a most-probable distance similar to that of PRE. Recently, we observed similar results for A71V in the absence of inhibitors [10], and those data are included here for comparison. Interestingly, as discussed more below, both L63P and A71V are located in the same region of the protein (Figure 1) identified as a “cantilever” region [27] [28] located near a hydrophobic cluster, which may have important roles in modulating flexibility and conformational changes of HIV-1 PR through the hydrophobic sliding mechanism [29].

3.2 Conformational Sampling Ensembles

Regeneration of the TKR distance profiles via linear combinations of Gaussian populations provides flap conformational ensembles profiles for HIV-1 PR [10,12,13]. Figure 2C shows the results of the population analysis of L63P HIV-1 PR conformational sampling based upon a Gaussian reconstruction of the data in Fig 2B. These analyses provide details of fractional occupancy of conformers in the ensemble; comprised of states assigned to tucked/curled, closed-like, semi-open and wide-open conformers. Full details of data analysis for all constructs are given in Figures S4–6. Based upon previous characterizations, the average distances for the preceding populations are 24–28, 33, 36–37, and 42 Å; respectively [4, 6, 12, 13, 16, 30–33]. Shown in Figure 3A are the relative populations of each distinct flap conformation that comprise the states in the conformational ensemble of HIV-1PR. Pictorial models of these states, with spin-label added in silico via MMM 2011.11 (http://www.epr.ethz.ch/software/), are given in Figure 3B. Models for closed and semi-open states come from crystal structures; whereas those for open-like and tucked/curled are from molecular dynamic simulation poses [34].

FIGURE 3.

(A) Fractional occupancy of the closed-like, semi-open, tucked/curled and wide-open conformations for each variant. Error is approximated at (5–8%). (B) Models of HIV-1 PR in the various conformational poses, with the MTSL label at position 55 shown in blue.

The PRE sequence adopts an ensemble with > 50% fractional occupancy of a closed-like state; a conformation typically seen upon interaction with inhibitor. [6,8,9] Because no inhibitor was added to these samples, we term this a closed-like conformation (in contrast to closed) because, as discussed previously [10], we do not expect the structure to be identical to an inhibitor-bound structure. The POST sequence, which results from the accumulation of drug-pressure selected mutations after exposure to antiviral treatment, adopts a conformational sampling profile with > 45% fractional occupancy of the semi- open population, with concomitant decreases in the percentages of the closed-like and tucked/curled conformations and an increase in the wide-open relative population. This observation is similar to what we recently observed for MDR769 and V6 [13] as well as for accumulating D30N/M36I/A71V mutations, where cross drug resistance is associated with increases in open-like conformations [10].

Given the conformational ensemble for L63P resembles that of PRE, results shows that it is not the absence of the stabilizing mutations, namely Q7K, L33I, and L63I [35] that are present in our subtype B constructs but not in PRE/POST, that cause the alteration to the more closed-like state in PRE. Finally, the shift from reported 35 Å in native subtype B to 33 Å in L63P and A71V provides evidence that a single amino acid substitution is sufficient to cause a substantial shift in conformations of HIV-1 PR.

3.3 L63P backbone dynamics

Pulsed EPR experiments provide data regarding the population of states. To interrogate backbone dynamics, high-resolution transverse relaxation rate (R₂), longitudinal relaxation rate (R₁), and (¹H)-¹⁵N heteronuclear Nuclear Overhauser Enhancement (hNOE) for peptide nitrogen of the L63P were collected and compared to subtype B. [3,12,21,36,37] NMR results are given in Figures S10–12 and Table S2. The single point substitution construct was chosen for NMR relaxation measurements because of minimal chemical shift perturbations in the HSQC spectra relative to subtype B; thus, resonance assignment could be easily inferred from published HSQC data. Order parameter, S², values, which characterize the amplitude of motion of the amide bond and range from 0 to 1 in increasing rigidity, were determined from model free analysis [38–40]. Figure 4A compares S² values for L63P and subtype B. Differences between S² for subtype B and L63P are given in Figure 4B. Order parameter values are graphically depicted on the ribbon diagrams of HIV-1 PR in Figure 4C. From the difference plots in Figure 4B, one can readily see that, on average, L63P is more rigid than subtype B. This effect is particularly pronounced in the flap region (residues 47–55) and dimer interface. DEER results show that L63P shifts the conformational ensemble to a more closed-like conformation with a most probable distance of ~33 Å. The NMR measurements indicate that the overall flexibility of the closed-like state is more rigid than the semi-open conformation. Comparison of our L63P S² values to published results on inhibitor-bound dynamics [3,37] shows greater flexibility compared to the inhibitor-bound closed conformation.

FIGURE 4.

Results from NMR relaxation measurements providing (A) order parameter, S2, values for L63P (B) and differences compared to subtype B. (C) Graphical representation S2 on ribbon backbone of HIV-1 PR.

3.4 Effects of single point mutation L63P on conformational ensemble and comparison to A71V

Both L63P and A71V are located away from the active site cavity (Figure 1) in a region of the protein that has been defined by MD simulations as a “cantilever” (residues 59–75), which has anti-correlated motion with the flaps [27] and where mutations in this region are predicted from MD simulations to alter dynamics and flexibility [26,41]. L63P is a substitution within the PRE sequence and our results show that this single point mutation may account for the more closed-like conformation of the PRE construct; an ensemble we have seen previously for A71V [10], with decreased flexibility. A71V is a common secondary mutation seen in multi-drug resistant constructs [42,43], and studies suggest this mutation stabilizes the dimer when primary mutations arise [41,44].

Structurally, 63 and 71 are in close contact with residues 62, 64, 66, 89 and 93 (Figure S13), defined as being within the hydrophobic core [29]. In crystal structures of consensus subtype B, A71 has a side-chain that protrudes into the hydrophobic packing region; hence the substitution with the larger valine residue may impact the communication between the flap and cantilever domain. A recent structure of PR20 [45], a highly resistant construct that contains A71V and L63P, shows similar packing of these residues compared to WT. Although PR20 does not show a predominant closed flap orientation in the absence of inhibitor, this construct also contains the following mutations considered to be within the hydrophobic core: L10F/I13V/I15V/I33F/M36I/-I62V/L90M. Comparing structures reveals a slight rotation of the side chain of I64 and change in side chain of I62V (Figure S13). L90M may also participate in subtle changes in hydrophobic packing of this region. Interestingly, the POST sequence contains hydrophobic core mutations L10I/I15V/M36I, which are mutated positions shared by PR20.

Recently, we showed by DEER studies that addition of M36I/D30N to A71V restores a predominant semi-open conformation with additional flap curling and open-like states [10]; conformations seen in crystal structures of PR20 [45]. To our knowledge, no crystal structures for single mutations A71V/L63P or PRE/POST sequence exist to date; but it is likely that substitution at these sites may alter the hydrophobic packing in this region, thus providing the molecular level reason for the altered flexibility and conformational sampling.

3.5 Relationship of conformational sampling, drug resistance and enzymatic activity

The conformational ensemble of PRE, giving a predominantly closed-like conformation, was unexpected because, except for A71V, a ligand-free construct with a most probable distance corresponding with the inhibitor induced closed flap conformation has not been reported. X-ray crystal structures of ligand-free HIV-1 PR primarily exhibit the semi-open conformation [4], and previous DEER studies have shown that apo subtypes B, C, CRF01_A/E, F and multi-drug resistant constructs V6 and MDR769 exhibit predominantly semi-open conformational states [13].

Inhibitor-bound HIV-1PR has been shown to primarily adopt the closed conformation, caused at least in part by interaction between the Ile50/50′ amide group and ligand via a conserved H₂O [6–9,16,46]. Intra-flap hydrogen bonding, which may stabilize the closed flap conformation, has been observed in MD simulations of apo enzymes, demonstrating that the average flap distance in the closed state is similar for ligand-bound and unbound protease [16]. This supports our experimental findings that despite having no flap-inhibitor interactions, ligand-free protease may adopt a closed-like conformation.

Kinetic studies of A71V suggest that stabilization of the closed-like state lowers catalytic turnover rate [10,42]. In parallel, the work of Clemente et al. also showed minimal impact of the A71V mutation on K_i values for Ritonavir, Indinavir and Nelfinavir compared to WT [42]. Although the drug-pressure selected mutations in POST compared to PRE were shown to affect replicative fitness and drug-response [17]; to date, in vitro kinetic or inhibition studies exists for comparison to our DEER data.

Recent work from our labs has showed that correlations exist among conformational sampling and enzymatic parameters and inhibition constants [10]. Results from that study indicate that drug resistance correlates with higher percentages of open-like conformations where stabilization of the closed-like state impairs enzymatic activity. These findings suggest a mechanism for how mutations combine; they maintain wild-type occupancy of the semi-open conformation while increasing the population of the open-like states and destabilizing the closed conformation. The work reported here adds to our growing understanding of how natural polymorphisms and drug-pressure selected mutations alter conformational sampling of HIV-1 PR, which may be communicated through subtle changes in the hydrophobic packing of key residues implicated in the hydrophobic sliding mechanism [29].

Conclusions

Taken together, these results add evidence to support our hypothesis that a role of combined drug pressure selected mutations may be to stabilize the semi-open conformation and induce a larger percentage of open-like states [10]. This work shows that drug-induced mutations, even single mutations distal to the active site pocket, can alter conformational sampling of HIV-1PR, which may impact drug-resistance and viral efficacy. The results indicate a possible mechanism for combining primary and secondary mutations to affect the conformational ensemble equilibria. Note that results from DEER experiments do not provide information regarding flexibility. NMR investigations of L63P clearly demonstrate decreased flexibility of this construct. Solution NMR assignments and relaxation experiments are underway for PRE and POST constructs.

Highlights.

The accumulation of mutations in POST induces a substantial shift of the conformational sampling of the PRE protease, from primarily closed to semi-open. The conformational sampling of the single point mutation L63P resembles that of PRE and the backbone flexibility of L63P, on average, is more rigid than subtype B, particularly in the flap region. Many of the accumulating mutations in POST reside in clusters of hydrophobic residues implicated in the hydrophobic sliding mechanism, providing a possible route for how these additional mutations communicate changes in conformational flexibility.

Abbreviations

PR

protease

SDSL

site-directed spin labeling

DEER

double electron-electron resonance

PELDOR

pulsed electron double resonance

EPR

electron paramagnetic resonance

NMR

nuclear magnetic resonance

TKR

Tikhonov regularization