Hydrophobic core flexibility modulates enzyme activity in HIV-1 protease

Abstract

Human immunodeficiency virus Type-1 (HIV-1) protease is crucial for viral maturation and infectivity. Studies of protease dynamics suggest that the rearrangement of the hydrophobic core is essential for enzyme activity. Many mutations in the hydrophobic core are also associated with drug resistance and may modulate the core flexibility. To test the role of flexibility in protease activity, pairs of cysteines were introduced at the interfaces of flexible regions remote from the active site. Disulfide bond formation was confirmed by crystal structures and by alkylation of free cysteines and mass spectrometry. Oxidized and reduced crystal structures of these variants show the overall structure of the protease is retained. However, cross-linking the cysteines led to drastic loss in enzyme activity, which was regained upon reducing the disulfide cross-links. Molecular dynamics simulations showed that altered dynamics propagated throughout the enzyme from the engineered disulfide. Thus, altered flexibility within the hydrophobic core can modulate HIV-1 protease activity, supporting the hypothesis that drug resistant mutations distal from active site can alter the balance between substrate turnover and inhibitor binding by modulating enzyme activity.

1. INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1) protease is a homodimeric, aspartyl protease with 99 residues in each monomer. Protease is essential for post-translational processing of the viral polyprotein Gag-Pro-Pol during the assembly and maturation of viral particles^1,2. There is an internal twofold symmetry in the protease dimer. However, ligands are asymmetric and introduce asymmetry in the dimer upon binding. Under protease inhibitor (PI) therapy, drug resistance mutations arise within the active site of the enzyme. These active site mutations directly interfere with the inhibitor binding and are primary cause of drug resistance to PIs. Additional mutations occur throughout the enzyme, and although they have often been shown to contribute to drug ressistance^3-5, the mechanism by which they do so is poorly understood.

Ligand binding in protease involves conformational changes in the protease flaps and the hydrophobic core residues. While the flaps of the unliganded protease are highly dynamic^6-11 and exhibit large conformational changes during ligand binding, hydrophobic core regions also exhibit subtle conformational changes¹². Of the 19 residues that comprise the hydrophobic core of HIV-1 protease in each monomer, 13 sites are associated with drug resistance¹² (Figure. 1). The role of these core residues, which do not contact either substrates or inhibitors directly, in substrate binding or drug resistance is not obvious and remains largely unexplored. Molecular dynamic simulations on unliganded HIV protease suggested the conformational changes occurring during side chain repacking in the hydrophobic core or “Hydrophobic sliding” as possible mechanism affecting drug susceptibility¹². The hydrophobic core domains of each protease monomer rearranges through correlated sliding motions facilitated by near-isoenergetic exchange of van der Waals contacts between hydrophobic side-chains such that the hydrogen bonding network within the core, primarily with the backbone, is conserved¹². While these simulations indicate that the rearrangement of the core facilitates conformational changes in the protease, much remains unknown about the role of this rearrangement in catalytic function and drug resistance.

Figure 1.

Drug resistance mutations in HIV-1 protease. Active site and the primary active site residues causing drug resistance (D30N, G48V,I50L/V, V82A/F/T, I84V) are colored in red. Hydrophobic core residues associated with drug resistance (I13V, I15V, L24I, L33F, M36I, I62V, I64V, I66F, V77I, I85V, L89M, L90M, I93L) are colored cyan. Remaining hydrophobic core residues (L5, V11, A22, L38, V75, L97) and rest of the protease are in yellow and gray, respectively. Catalytic Asps and the loops containing residues 11-22, 31-38 and 58-78 are displayed in black.

The contributions of protein dynamics in both protein structure and function have been investigated in other systems. Studies on the multidrug efflux pump AcrB¹³ and caspase7¹⁴ have shown the functional importance of flexible regions, but there has been little emphasis on studying the dynamics of the hydrophobic core of the proteins. The reasons are two fold: 1) difficulty in accessing the hydrophobic core in folded proteins and, 2) limited resolution of currently available experimental techniques to view protein dynamics. While NMR, Mass spec-trometry and other single molecule techniques have furthered our understanding of protein dynamics and conformational heterogeneity of proteins in bulk solution^15-18, structural characterization of conformations other than those found in the crystal structures, remains challenging.

We experimentally investigated the role of hydrophobic core flexibility in HIV-1 protease function using site-directed cysteine cross-linking. Disulfide bonds are usually either structural and stabilize a protein or catalytic and modulate the function of proteins involved in cellular pathways^19-22. Disulfide bonds which modulate activity are both useful in functional description of proteins and provide a method to evaluate the regulation of function in proteins^13,23,24. In this study, site-directed cysteine cross-linking provided evidence for the requirement of conformational changes in the hydrophobic core of HIV-1 protease for efficient cleavage of its substrates. The reversibility of disulfide chemistry worked as a molecular switch to modulate protease activity confirming that cross-linking caused the reduced activity (Figure. 2). These results confirm the role of hydrophobic core flexibility in protease function.

Figure 2.

Schematic of cysteine redox chemistry used in this study. The cysteine substituted and non-cross-linked protease is flexible and as active as WT enzyme. Upon oxidation of cysteines, protease gets cross-linked via disulfide bond. Resulting loss of hydrophobic core flexibility is accompanied by loss of catalytic activity that is reversible upon reduction of the disulfide bond.

Since conformational flexibility in HIV-1 protease has also been implicated in drug resistance^12,25,26, we hypothesize that the non-active site mutations may generate novel conformational heterogeneity in the protease, such that there is a preferential selection for conformers that process substrates than bind inhibitors. Therefore, intrinsic protein dynamics can mechanistically connect protein function, molecular recognition and drug resistance. More generally, hydrophobic core rearrangement may represent a mechanism by which proteins undergo the necessary conformational changes required for function.

2. RESULTS

2.1 Rationale for cysteine engineering and chosen cysteine pairs

Analysis of previous MD simulations¹² revealed that 19 core residues in each protease monomer had limited solvent accessibility throughout the 14ns simulations (Figure 1). During the simulations these hydrophobic core residues undergo conformational changes concurrent with the opening of the active site cavity. We hypothesized that this hydrophobic sliding in HIV-1 protease is critical for proteolytic activity. To test this hypothesis, specific pairs of cysteines were introduced into the protease in a cysteine-free background. MD simulation trajectories¹² were analyzed at various time points during the simulations and cysteine pairs were chosen at protease positions to restrict the movement of the hydrophobic domains using the C_β-C_β distance criterion for allowing disulfide bond formation²⁷ (Table SI1). The selection of residue positions was based on their proximity to the rearranging hydrophobic core and was not limited to the 19 positions comprising the core. Three distinct regions containing residues (11-21), (31-38) and (58-78) were seen to be tethered to the core and exhibit extensive rearrangement. Specifically, G16, R14 and E65 residues, although not hydrophobic themselves, are contiguous with the hydrophobic core and along with L38 met the distance criteria. These solvent accessible sites were chosen as suitable to probe the hydrophobic core dynamics and are amenable to reduction-oxidation chemistry. Under appropriate oxidation conditions, the introduced cysteines should form disulfide cross-links, likely restricting hydrophobic core movement and potentially enzymatic activity.

Two protease variants, each with a unique cysteine pair, were constructed (Figure 3a). The first pair (G16C/L38C) would establish an intra-monomeric cross-link that should severely restrict hydrophobic core rearrangements. An alanine control variant G16A/L38A was generated to ensure that these sites could tolerate substitutions without compromising protease activity. The second cysteine pair (R14C/E65C) was designed as a control to form cross-links at sites that should not restrict hydrophobic core rearrangements, as the relative distance between these residues did not significantly change during the simulations¹² (Table SI1). Cross-linking in both these variants should occur under oxidizing conditions, but loss of protease function was predicted, based on our hypothesis, in only the G16C/L38C variant. Cross-linking in G16C/L38C should restrict the hydrophobic core dynamics and hinder activity. In R14C/E65C variant, protease function should be unaffected under both oxidizing and reducing (cross-linked (oo) and non-cross-linked (rr)) conditions because core rearrangements should be unaffected (Figure 2).

Figure 3.

Ribbon diagrams of crystal structures of protease with engineered cysteines (a) Sites for engineering cysteines in HIV-1 protease are shown with backbone C_α (G16C/L38C) in red and (R14C/E65C) in blue. (b) The backbone structural superposition of DRV complexes of (G16C/L38C)_rr and (R14C/E65C)_rr on WT in black. Under reducing conditions, no disulfide bonds were observed. Alternate conformations for cysteine side chains were, however, seen for 50% of substituted cysteines in each pair analyzed. (c) _apo(G16C/L38C)_oo with disulfide bonds on both sides of the dimer shown in orange color.

2.2 Modulation of protease activity

By exploiting the reversibility of disulfide bonding, the effect of restricted flexibility on protease function was evaluated. Protease mediated cleavage of a fluorogenic peptide substrate derived from the native CA-p2 cleavage site on Gag polyprotein was measured in the presence and absence of reducing agent. The WT, (G16C/L38C)_rr, (G16A/L38A)_ctrl, and (R14C/E65C)_rr proteases all showed similar catalytic activity (Table 1). In contrast, the cross-linked protease (G16C/L38C)_oo showed 146-fold lower activity than the non-cross-linked (G16C/L38C)_rr, demonstrating dramatic loss of function. 70% of this function was recovered upon addition of the reducing agent TCEP, demonstrating that the loss of activity was due to the G16C/L38C disulfide bond. The R14C/E65C variant exhibited comparable catalytic activity under both oxidizing and reducing conditions, as expected. These results confirm that the activity of disulfide-crosslinked variants depends on the location of the cross-link. Protease activity was only sensitive to the cross-link that restricted the flexibility of HIV-1 protease hydrophobic core.

Table 1.

Kinetic values for the WT, cross-linked, non-cross-linked variants of G16C/L38C and R14C/E65C and G16A/L38A variants of HIV-1 protease.

Protease Variants	Catalytic Activity, Kcat/Km (μm−1s−1)	Normalized
Wild-type	6.8×10−2 ± 0.002	1
G16C_L38C (rr)	7.6 ×10−2 ± 0.004	1.1
G16C_L38C (oo)	5.2 ×10−4 ± 2.5 ×10−6	7.6 ×10−3
G16C_L38C(oo) + 5mM TCEP (regain of function)	4.7 ×10−2 ± 0.006	0.69
G16A_G38A	6.2 ×10−2 ± 0.002	0.91
R14C_E65C (rr)	6.5 ×10−2 ± 0.003	0.96
R14C_E65C (oo)	7.1 ×10−2 ± 0.004	1.04

2.3 Structural analysis

Six crystal structures of the engineered protease variants were determined (Table SI2). Five structures were in complex with DRV, a 4 pM inhibitor of WT²⁸, including the oxidized and reduced forms of G16C/L38C, reduced R14C/E65C, WT and G16A/L38A. The final structure was an apo structure of the oxidized G16C/L38C variant. In the oxidized G16C/L38C variant in complex with DRV, perhaps due to the high affinity of DRV and/or the asymmetry of the crystal packing, although contiguous electron density is observed, one of the disulfide bridges is in a highly strained geometry. However, in the apo form of G16C/L38C variant both disulfides are clearly observed, (Figure SI1) although this structure was excluded from all comparative structural analyses because of the relatively low resolution.

The overall structural fold of all cysteine substituted protease complexes was well preserved when compared to the WT, with RMSD ranging from 0.06 to 0.09 (Figure. 3b and c). Hydrogen bond analysis revealed that all structurally important hydrogen bonds were conserved for all DRV complexes confirming that the known mode of ligand binding was intact in these proteases. The disulfide bonds observed in DRV(G16C/L38C)_or and apo(G16C/L38C)_oo structures fall into geometric categories (-RHHook and –LHHook respectively) commonly observed in natural proteins¹⁹.

2.4 Molecular dynamic simulation analysis of cross-linked and non-crosslinked structures

To evaluate the ability of engineered cross-links to restrict hydrophobic core dynamics, MD simulations were carried out for 20 ns on the cross-linked and non-crossliked protease. DRV was removed from the starting structures, DRV(G16C/L38C)_rr and a model of DRV(G16C/L38C)_oo. Extensive structural differences were observed in the protease backbone for the (oo) compared to the (rr) structures of G16C/L38C protease between 0 ns and 20 ns (Figure. 4a and 4b). These differences map to backbone shifts of greater than 1Å and are most pronounced for the protease flaps (residues 45-55) as indicated by the double difference plots. The presence of disulfide bonds affected the dynamics in (G16C/L38C)_oo protease simulations leading the flaps to rearrange and begin to open after 20 ns. In comparison, for the (G16C/L38C)_rr simulation, the unrestricted hydrophobic core region showed subtle motions, with no significant motion elsewhere in the protease dimer on the time scale of these simulations. The RMSF (root mean square fluctuation) for each protease residue over the 20ns MD simulations for (G16C/L38C)_oo and (G16C/L38C)_rr structures further confirmed these results (Figure 4c). This suggests that flap movements can be modulated in response to the restricted hydrophobic core dynamics as caused by the formation of a disulfide between G16C and L38C.

Figure 4.

Molecular dynamics simulation analyses. (a) Backbone superposition of (G16C/L38C)_oo from 0ns (cyan) and at 20ns (pale cyan) and (G16C/L38C)_rr from 0ns (light pink) to 20ns (magenta) in MD simulations. The side chains of active site aspartic acids and the engineered cysteines are displayed. (b) Differences in internal Cα-Cα distances between the 0ns and 20ns snapshots of the cross-linked (oo) and non-crosslinked (rr) forms of (G16C/L38C) variant are shown in the double difference plots. Each contour line represents a deviation by 0.5Å. Black, green, blue and red distinguish the contour ranges −1.0 Å and below, −1.0 to −0.5 Å, 0.5-1.0 Å and 1 Å and above, respectively. (c) Average RMSF of protease residues in (G16C/L38C)_oo and (G16C/L38C)_rr proteases from 20ns MD simulation trajectories. Protease molecules from 5, 10, 15 and 20ns simulations were superposed on to the 0ns crystal structure using the most invariant residues, 24 to 26 and 85 to 95. The average C_α RMSFs were calculated and mapped on to a representative protease molecule with the most variable regions depicted in red and the most invariant regions depicted in blue.

The structural regions containing residues (11-21), (31-38) and (58-78) that were previously implicated in the hydrophobic core sliding¹² (Figure 1) were further analyzed for the effect of physical 16C/38C cross-linking on the side chain conformational rearrangement in the core during the MD simulations. While quantifying the restricted sliding is challenging, structural analysis of MD trajectories at various time points revealed that side chain repacking within the hydrophobic core was restricted in the (G16C/L38C)_oo simulations compared to the (G16C/L38C)_rr simulations. In (G16C/L38C)_rr simulations, the hydrophobic core side chains sampled a variety of conformations, exchanging van der Waals contacts with neighboring residues. However in the (G16C/L38C)_oo simulations, most of the hydrophobic core interactions were maintained. These results suggest that the restricted hydrophobic core dynamics observed in (G16C/L38C)_oo simulations is a direct consequence of disulfide cross-linking and likely impair catalysis.

2.5 Conclusions

In this study site-directed cysteine engineering was used to reversibly restrict the internal dynamics in protease and evaluate the effects on function. The double-cysteine mutant (G16C/L38C) was selected as a pivotal pair of residues that varied significantly in separation distance depending on the conformational state of the protein. Thus cross linking them would severely restrict the internal core dynamics. The double-cysteine mutant (R14C/E65C) was used as internal controls, as these residues, though on separate loops, were not seen to change their separation distance significantly. Reduction-oxidation of the disulfide bridges was used as a switch to regulate hydrophobic core dynamics and eventually protease function (Figure. 2). The formation of disulfide (G16C/L38C) cross-links in fact exerted conformational restraints on the hydrophobic core, compromising protease activity. This mechanism of core rearrangement explains the results in the observed inactivation of (G16C/L38C)_oo that was largely restored upon reversing the disulfide bonds to (G16C/L38C)_rr (Table 1). Activity was lost and regained only in the (G16C/L38C) variant, where hydrophobic core movements were expected to be crucial for protease function. In the control case of (R14C/E65C) variant crosslinking did not alter enzyme activity. These results strongly support the mechanistic role of hydrophobic core rearrangement in protease function.

3. DISCUSSION

HIV-1 protease undergoes large conformational changes to bind ligands. We show that conformational changes are essential to protease activity. Flap motion and hydrophobic core repacking have been previously implicated in modulating pro-tease activity. However, the inherent challenges of manipulating protein dynamics, make testing these motions experimentally difficult. Our results provide strong support that hydro-phobic core flexibility contributes to the efficiency of substrate cleavage by HIV-1 protease.

When HIV protease processes substrates, the enzyme undergoes large conformational changes upon substrate recognition, cleavage and product release. Competitive active site inhibitors lock down these conformational changes by tightly binding to the active site of the enzyme, mimicking a transition state. Drug resistance occurs when the balance between sub-strate recognition and turnover and inhibitor binding is perturbed. In the case of the active site, resistance-conferring mutations occur in such a manner that directly preserves substrate recognition and turnover while compromising inhibitor binding²⁹. However many mutations associated with resistance occur outside the active site within the protease hydro-phobic core^3,5,12,30 (Figure 1). Mutations that modulate the dynamics of the hydrophobic core of HIV-1 could modulate the functional properties of the active site, and thus the relative binding affinities for substrates and inhibitors. We hypothesize that distal mutations outside the active site cause drug resistance when the balance between substrate recognition and turnover, a dynamic process, and inhibitor binding, a locking down of the enzyme, is altered in a manner to favor substrate turnover by modulating the flexibility of the core of HIV-1 protease.

This study demonstrates the essential role of the hydrophobic core in modulating the activity of HIV-1 protease. The use of specific cross-links to probe the impact of dynamics in pro-tease activity provides a new tool to investigate this enzyme. Such engineered cross-links may be useful to probe our hypothesis of the mechanism of action of mutations outside the active site modulating drug resistance in HIV-1 protease.

ABBREVIATIONS

HIV-1

Human Immunodeficiency virus type-1

TCEP

tris 2-carboxyethyl phosphine

WT

wild type

RMSD

root mean square deviation

MD

molecular dynamics

DRV

darunavir