Conformational asymmetry of human immunodeficiency virus‐1 protease flaps increases along the darunavir drug resistance pathway

Abstract

Under selective drug pressure, the structure and dynamics of HIV‐1 protease (PR) evolve to confer resistance to the drug. Here, we report on the conformational changes in PR flaps for drug‐resistance mutations using ¹⁹F nuclear magnetic resonance (NMR) spectroscopy. We prepared wild‐type (wt) PR and its three mutant variants containing 4, 10, or 11 mutations, namely 4Mut, 10Mut, and 11Mut, that were previously found in a viral passage study against darunavir (DRV). The proteins were uniformly labeled with ¹⁵N and with ¹⁹F amino‐acid specifically in the 5‐indole positions of two conserved tryptophan residues, W6 and W42, which are located near the dimer interface and at the edge of the flap region of PR, respectively. In solution, ¹⁹F NMR spectra of these proteins in the inhibitor‐free forms showed a downfield shift of the W42 resonance as the number of mutations increased. In the DRV‐bound form, the W42 resonance again shifted with increasing number of mutations and, importantly, split into two peaks, reflecting different local electronic environments of the two flaps of the dimer. Solid‐state ¹⁹F magic angle spinning (MAS) NMR spectra of wt‐PR and 10Mut also exhibited two W42 resonances in DRV‐bound crystalline forms. These observations indicate that drug‐resistance mutations increase the asymmetry between the two flaps in the PR‐inhibitor interaction in the inhibitor‐bound forms.

1. INTRODUCTION

Human immunodeficiency virus‐1 (HIV‐1) protease (PR) is an enzyme essential for HIV‐1 replication (Coffin et al., 1997; Kaplan et al., ¹⁹⁹³; Kohl et al., ¹⁹⁸⁸; Krausslich, ¹⁹⁹¹; Oroszlan & Luftig, ¹⁹⁹⁰; Seelmeier et al., ¹⁹⁸⁸) and has long been a target for drug development (Erickson & Burt, 1996; Galiano et al., ²⁰⁰⁹; Ghosh et al., ²⁰⁰⁸; Quinones‐Mateu et al., ²⁰⁰⁸; Surleraux et al., ²⁰⁰⁵; Wlodawer & Erickson, ¹⁹⁹³). The mature HIV‐1 PR is a homodimeric aspartyl protease, with each subunit comprising 99 amino acids that together form an active site with two aspartic acid residues, one from each of the two subunits (Lapatto et al., 1989; Navia et al., ¹⁹⁸⁹; Wlodawer et al., ¹⁹⁸⁹). The flap region of the PR (residues 43–58) is known to be dynamic, exhibiting an ensemble of open and closed conformations in solution (Figure 1a); the closed conformation predominates upon binding to competitive inhibitors (Figure 1b). For characterizing PR flap dynamics, nuclear magnetic resonance (NMR) spectroscopy has made significant contributions (Blackburn et al., 2009; Freedberg et al., ²⁰⁰²; Galiano et al., ²⁰⁰⁹; Ishima et al., ¹⁹⁹⁸; Ishima et al., ¹⁹⁹⁹; Ishima & Louis, ²⁰⁰⁸; Louis et al., ²⁰⁰⁷; Nicholson et al., ¹⁹⁹⁵; Roche et al., ²⁰¹⁵), and findings from NMR studies have been supported by a large number of PR crystal‐structure‐based computational studies that helped to provide insight into the motion of PR in the free and inhibitor‐bound forms (Chetty et al., 2016; Collins et al., ¹⁹⁹⁵; Deng et al., ²⁰¹¹; Ding et al., ²⁰⁰⁸; Gardner & Abrams, ²⁰¹⁹; Harrison & Weber, ¹⁹⁹⁴; Harte et al., ¹⁹⁹⁰; Hou & Yu, ²⁰⁰⁷; Karthik & Senapati, ²⁰¹¹; Leidner et al., ²⁰²¹; Paulsen et al., ²⁰¹⁷; Perryman et al., ²⁰⁰⁶; Ragland et al., ²⁰¹⁷; Shen et al., ²⁰¹⁵; Silva et al., ¹⁹⁹⁶; Toth & Borics, ²⁰⁰⁶; Tzoupis et al., ²⁰¹⁴; Ung et al., ²⁰¹⁴; Venkatachalam et al., ²⁰²⁵; Wang & Zheng, ²⁰²¹; York et al., ¹⁹⁹³).

FIGURE 1.

Superposition of HIV‐1 wt‐PR structures in ribbon presentation for (a) inhibitor‐free and (b) DRV‐bound forms, and (c) mutation sites (in ball representation) for 4Mut (K45I, M46I, V82F, and I84V) and (d) additional mutation sites for 10Mut (I13V, G16E, V32I, L33F, A71V, and L76V). Note, 11Mut has one additional mutation, I54L, on 10Mut. Panel (a) was generated with two PR structures, one is closed (cyan color ribbons, PDB code 4EJL; Tiefenbrunn et al., ²⁰¹³), and the other is more open (red color ribbons, PDB code 1HHP; Spinelli et al., ¹⁹⁹¹), to give a flavor of dynamics of the flap region. Panels (b–d) were generated using PDB 6DGX (Henes et al., 2019). In panel (b), two chains of HIV‐PR are shown by green and pink, with DRV (red color) between the two. In panels (c) and (d), mutation sites and Trp residues are shown in ball representation using either the standard CPK colors or yellow color, respectively.

Differences in the structure and dynamics of PR in the presence of mutations that impart multidrug resistance also have been well studied: most mutations generally increase flap mobility in inhibitor‐free forms of the protein, reduce packing at the inhibitor‐PR interface compared to the wild‐type protein, and enhance asymmetry of the two subunits (Agniswamy et al., 2016; Cai et al., ¹⁹⁹⁷; Cai et al., ²⁰¹²; Cai et al., ²⁰¹⁴; Galiano et al., ²⁰⁰⁹; Henes et al., ²⁰¹⁹; Nakashima et al., ²⁰¹⁶; Rajendran et al., ²⁰²³; Roche et al., ²⁰¹⁵; Wong‐Sam et al., ²⁰²²; Zhang et al., ²⁰¹⁴). These general conclusions are based on analyses of crystal structures, electron spin resonance data, NMR data, and molecular dynamics (MD) simulations, which have provided a consistent picture of wt‐PR dynamics. However, in studies of the inhibitor‐bound forms of the drug‐resistance mutants, NMR analysis of protein backbone amides point to minor conformational asymmetry between the two flaps in the mutants (Cai et al., 2014; Khan et al., ²⁰¹⁸; Khan et al., ²⁰²¹; Persons et al., ²⁰¹⁸), which is in contrast to the significant structural and dynamical changes of the inhibitor‐bound PRs suggested by MD simulations (Agniswamy et al., 2016; Cai et al., ²⁰¹⁴; Galiano et al., ²⁰⁰⁹; Henes et al., ²⁰¹⁹; Nakashima et al., ²⁰¹⁶; Rajendran et al., ²⁰²³; Wong‐Sam et al., ²⁰²²; Zhang et al., ²⁰¹⁴).

The absence of significant conformational asymmetry in the flaps was deduced from small effects on backbone amide signals in the inhibitor‐bound PR and relates to the physics inherent to the specific spectroscopy approach. Amide ¹H and ¹⁵N chemical shifts of diamagnetic proteins reflect electron densities in the local environment, at a level that these can be utilized for secondary‐structure prediction of proteins (Wishart et al., 1992). NMR relaxation, which has been used to study HIV‐1 PR dynamics, is sensitive to local rotational motion but not to translational motion (Dayie et al., 1996; Ishima & Torchia, ²⁰⁰⁰; Lipari & Szabo, ¹⁹⁸¹; Lipari & Szabo, ¹⁹⁸²; Palmer III, ¹⁹⁹⁷). In addition, NMR residual dipolar couplings (RDCs) of the protein backbone atoms are sensitive to relative changes of the orientation of the bond vector against the alignment tensor of the molecule (Roche et al., 2015). When these average properties dominate the data, it is difficult to detect small changes. Therefore, we hypothesized that NMR experiments using nuclei more sensitive to both chemical and local environments may help detect conformational changes in PRs upon drug‐resistance mutations, particularly in the inhibitor‐bound forms.

We tested this hypothesis using ¹⁹F NMR. ¹⁹F is extremely electronegative, rendering its chemical shifts very sensitive to small changes in the electronic environment. ¹⁹F NMR has been applied to characterize molecular interactions, conformational changes, and surface solvent effects in biomolecules (Emsley & Phillips, 1971; Gerig, ¹⁹⁹⁴; Gronenborn, ²⁰²²; Kitevski‐LeBlanc & Prosser, ²⁰¹²). We prepared wt‐PR and three drug‐resistance mutant proteins previously identified and characterized in a viral passage study against darunavir (DRV). The three mutant proteins contained either (i) four mutations (K45I, M46I, V82F, and I84V, Figure 1c); (ii) 10 mutations (those in [i] plus I13V, G16E, V32I, L33G, A71V, and L76V, Figure 1d) or (iii) 11 mutations (those in [ii] plus I54L, Table 1) (Henes et al., 2019). These proteins, referred to as 4Mut, 10Mut and 11Mut, respectively, along with wt protein were uniformly labeled with ¹⁵N and amino acid‐specifically labeled with ¹⁹F at the 5‐indole position (5F‐Trp). Two conserved Trp residues in PR reside at positions 6 and 42, ideally positioned to report on changes in the dimerization interface and the flap disposition (Figure 1c). Based on the crystal structure of PR (Henes et al., 2019), position 5 of W6 is near the dimer interface and solvent‐exposed. Position 5 of W42 is at the end of the flap β‐hairpin structure (residues from 43 to 58) and is partially buried in the protein toward the β‐hairpin (Figures 1c and S1, Supporting Information). Importantly, both Trp residues are at least 14 Å away from DRV and thus are expected to sense changes in the structure and/or dynamics of PR but not a direct effect from DRV.

TABLE 1.

Mutations in the pathway to generate multi‐drug resistance. ^a

Variant	Mutations
WT
4Mut			K45I M46I			V82F I84V
10Mut	I13V G16E	V32I L33F	K45I M46I		A71V L76V	V82F I84V
11Mu	I13V G16E	V32I L33F	K45I M46I	I54L	A71V L76V	V82F I84V

^a Details are available in Henes et al. (2019).

Our 1D ¹⁹F NMR spectra of the PRs, both in the inhibitor‐free and DRV‐bound forms, showed that the W42 resonance in the mutants progressively shifted to a lower field, toward the cleaved‐product position. In addition, in the DRV‐bound form, the W42 resonance of 10Mut and 11Mut is split into two. ¹⁹F magic‐angle spinning (MAS) NMR spectra of DRV‐bound crystals of wt‐PR and 10Mut also exhibited three resonances, similar to the pattern in solution: one W6 resonance and two W42 resonances, with a lower‐field shift for one of the W42 resonances of 10Mut. Taken together, our results clearly demonstrate asymmetric flap‐inhibitor interaction in inhibitor‐bound forms, both in solution and in microcrystalline forms.

2. RESULTS

2.1. Folding of PRs and backbone dynamics in solution

5F‐Trp/U‐¹⁵N labeled PRs were expressed in BL21 (DE3) cells grown in modified M9 medium containing ¹⁵NH₄Cl as the sole nitrogen source supplemented with 5‐fluoroindole before PR induction. Mass spectrometry analysis of the purified PRs revealed one major peak (85%–90% yield) and one minor peak (10%–15% yield). For each protein, the major fraction exhibited a molecular mass corresponding to 99% ¹⁵N labeling and 100% ¹⁹F labeling of the two Trp residues (Table S1). The minor fraction corresponded to 99% ¹⁵N labeling and ¹⁹F labeling for one of the two Trp positions (Figure S2). We concluded that this labeling efficiency was sufficient for our study.

Proper folding of the PRs was confirmed by recording ¹H‐¹⁵N HSQC spectra in solution (Figure 2a). All PR spectra were similar to those known from previous work on PR dimers, with ¹H resonances for Q92 and I93 amides at 7 ppm or below, reflecting the known inter‐subunit ring‐current effect (Louis et al., 2003). Very weak indole NH resonances of the Trp residues were also observed (not shown in Figure 2a), consistent with the presence of a small fraction of protein having only one 5F‐Trp labeled (Figure S1). DRV binding altered a subset of the PR resonances in the ¹H‐¹⁵N HSQC spectra, as observed previously (Figure 2b) (Khan et al., 2018; Persons et al., ²⁰¹⁸).

FIGURE 2.

Superposition of ¹H‐¹⁵N HSQC spectra of 5F‐Trp/U‐¹⁵N labeled wt‐PR, 4Mut, 10Mut, and 11Mut in (a) inhibitor‐free and (b) DRV‐bound forms. The protein concentrations were at 20–25 μM (as dimer) in 20 mM phosphate buffer at 293 K.

Backbone dynamics of DRV‐bound wt‐PR and 11Mut were evaluated by recording ¹⁵N transverse relaxation rate (R₂, Figure 3a,b), ¹⁵N longitudinal relaxation rate (R₁, Figure 3c,d), and {¹H}‐¹⁵N heteronuclear NOE (Figure 3e,f) data. To enable comparison of wt‐PR and 11Mut relaxation data, backbone resonance assignments were performed for DRV‐bound wt‐PR and 11Mut, and resonances were tentatively assigned to one of the two chains (for details, see footnotes of Tables S2, S3). The overall relaxation profile of wt‐PR was similar to that in a previous relaxation study, with only a few minor differences (Figure 3) (Cai et al., 1997). The presence of subtle differences is not surprising, since here the NL4‐3 wt‐PR (Henes et al., 2019) was used rather than the NMR‐optimized wt‐PR of the previous study (Cai et al., 2014) (see Materials and Methods).

FIGURE 3.

NMR relaxation data of 5F‐Trp/U‐¹⁵N labeled wt‐PR (black circles) and 11Mut (white rectangles) in DRV‐bound forms in solution: (a, b) transverse relaxation rate (R₂), (c, d) longitudinal relaxation rate (R₁), and (e, f) {¹H}‐¹⁵N NOE values, each for subunit 1 (a, c, e) and subunit 2 (b, d, f). Note, some amino acids are tentatively assigned to either subunit 1 or 2 to enable comparison between wt‐PR and 11Mut (see Tables S2, S3). For residues with identical amide resonances in the two subunits, identical relaxation rates are plotted for both subunits.

Comparison of the wt‐PR relaxation profiles with those of 11Mut showed nearly identical data, with few exceptions (Figure 3). Residue 80 is positioned between two proline residues, forming a P79‐T80‐P81 stretch, and displayed slightly lower R₂ and NOE in 11Mut compared to the wt‐PR, consistent with previous observations (Cai et al., 1997). In the current data, G40, located in a loop region called “elbow” (residues 38–42, Zoete et al., ²⁰⁰²), showed slightly lower R₂ and NOE in the wt‐PR, compared to 11Mut (Figure 3). In the flap region (residues 43–58), no substantive differences in relaxation values between wt‐PR and 11Mut were noted (Figure 3). Overall, aside from a few sites, the NMR relaxation data generally reflect a rigid, inhibitor‐bound form of all PRs studied, even the PR containing the 11 mutations.

2.2. ¹⁹F NMR spectra reveal chemical shift changes for the PR mutants

¹⁹F NMR spectra were recorded for 5F‐Trp/U‐¹⁵N labeled PRs in their inhibitor‐free forms. The ¹⁹F NMR spectra of 4Mut, 10Mut, and 11Mut exhibited two resonances, one from W6 and the other from W42 (details below), while the spectrum of wt‐PR exhibited three resonances (Figure 4a). The middle resonance in the wt‐PR spectrum, at −48.4 ppm, increased in intensity over time, and similar signals were also observed over time for 4Mut, 10Mut, and 11Mut (Figure 4a, dashed lines), indicating that the middle resonance resulted from autoproteolysis of PRs. Considering that the k_cat/K_m value of wt‐PR is approximately double that of 4Mut and more than 10 times higher than that of 11Mut (Henes et al., 2019), observation of autoproteolysis already in the fresh sample of wt‐PR was not surprising. The other two resonances, at −47.8 ppm and at −48.6 ~ −49.3 ppm showed similar peak intensities or areas across the various protein samples (Table S4), indicating that these two resonances were from W6 and W42.

FIGURE 4.

1D ¹⁹F NMR spectra of 5F‐Trp/U‐¹⁵N labeled wt‐PR (black), 4Mut (blue), 10Mut (green), and 11Mut (red) (a) without inhibitor and (b) with DRV in solution. In both panels, spectra of fresh samples are shown in solid lines, and those recorded 2–3 days later are shown in dashed lines. “f” indicates the resonance of proteolytic fragments.

Assignment of each 5F‐Trp resonance was based on the mutation sites and their locations relative to Trp resonances. For example, comparing the ¹⁹F NMR spectrum of wt‐PR with that of 4Mut, the resonance at −49.3 ppm in wt‐PR shifted downfield to −48.7 ppm in the 4Mut protein, while the resonance at −47.8 ppm remained essentially unchanged (Figure 4a and Table S4). The mutations in 4Mut are K45I, M46I, V82F, and I84V (Table 1, and Figure 1c), with residue 45 being the closest to W42 at 5.5 Å. Indeed, K45I and M46I are located within the same β‐strand as W42. In contrast, among the 4Mut mutations, the closest residue to W6 is residue 84 (17 Å). We therefore reasoned that the higher field resonance belongs to W42. Consistent with this, the ¹⁹F NMR spectra of 10Mut and 11Mut—the latter containing an I54L mutation on 10Mut—showed a slight difference in the chemical shift of the W42 resonance, at −48.62 ppm and −48.58 ppm, respectively, with no change in the W6 resonance position.

Importantly, the ¹⁹F chemical shift of the W42 resonance of PR changed with increasing numbers of amino acid changes, moving from −49.3 ppm to −48.6 ppm, toward the fragment peak position (Figure 4a and Table S4), while the chemical shift of the W6 resonance did not change much. Since PRs may weakly bind substrate/product generated by autoproteolysis, one might consider the possibility that the observed shift was due to the equilibrium of the bound/unbound forms. However, the chemical shift of the W42 resonance did not change by possible autoproteolysis (Figure 4a, dashed lines), making this possibility unlikely. The over 0.5 ppm change in the W42 resonance in 4Mut compared to wt‐PR may have been due to an electron‐shielding difference, since in the K45I mutation, a positive charge is removed.

2.3. ¹⁹F NMR reveals conformational asymmetry of the two flaps in the DRV‐bound form

¹⁹F NMR spectra recorded for 5F‐Trp/U‐¹⁵N labeled PRs in the presence of DRV exhibited patterns similar to those of the inhibitor‐free forms, that is, deshielding of the W42 resonance as the number of mutated residues increased (Figure 4b). In addition, the W42 resonance clearly split into two separate peaks in 10Mut and 11Mut (Figure 4b). The intensity ratio of W42 peaks was 0.25:0.28 and the area ratio was 0.23:0.32 (Table S5). Although the resonance intensity of the downfield W42 peak was slightly less than that of the upfield W42, the intensity ratio did not exhibit temperature‐dependent changes over a range of temperatures, from 283 K to 305 K (Figure S3 and Table S6). Therefore, the two W42 resonances stem from different environments of W42 in the two subunits, rather than two‐state dynamics of PR.

Since the chemical shift difference of the two W42 resonances in the DRV‐bound form was not observed in wt or 4Mut and observed only in 10Mut and 11Mut, the shift was not due to a direct effect of the inhibitor's electron distribution or induced magnetic field but instead arose from a structural difference in the PR subunits. This finding unambiguously revealed conformational asymmetry of the two flaps in the dimer. Unfortunately, the degree of conformational asymmetry cannot be determined from our data, considering the pin‐point observation of W42 and the sensitivity of ¹⁹F NMR to small environmental changes. At the very least, our observation of a conformational difference is qualitatively consistent with results from previous MD simulations, which suggest asymmetric flap structures in the inhibitor‐bound forms (Agniswamy et al., 2016; Cai et al., ²⁰¹⁴; Galiano et al., ²⁰⁰⁹; Henes et al., ²⁰¹⁹; Nakashima et al., ²⁰¹⁶; Rajendran et al., ²⁰²³; Wong‐Sam et al., ²⁰²²; Zhang et al., ²⁰¹⁴).

2.4. ¹⁹F solid‐state MAS NMR spectra of PR‐DRV crystals reveal a similar spectral pattern to that in solution

In order to investigate ¹⁹F NMR spectral pattern of 5F‐Trp in the crystalline state, we prepared crystals of 5F‐Trp/U‐¹⁵N labeled wt‐PR and 10Mut in their DRV‐bound forms and performed ¹⁹F MAS NMR experiments. 11Mut was not used in this crystallization study due to its instability (Figure 4b, dashed lines). Since the inhibitor‐free form of PR undergoes autoproteolysis, crystallization was only performed for the DRV‐bound and not the inhibitor‐free forms of the proteins.

Direct‐polarization MAS (DPMAS) spectra of the DRV‐bound wt‐PR and 10Mut crystals exhibited three resonances (Figure 5a). The resonance at −45 ppm was the same in both wt‐PR and 10Mut, suggesting that this resonance belongs to the W6 indole. The resonance at −48 ppm in the DRV‐bound wt‐PR moved in DRV‐bound 10Mut, similar to what was observed in solution, and was therefore assigned to W42. The ¹H‐¹⁹F cross‐polarization (CPMAS) spectra showed a loss of intensity for the peak at −45 ppm (Figure 5a), but not for the other two resonances (described in more detail below). We concluded that the resonance at −45 ppm originated from the W6 indole, whereas the other two resonances were from W42.

FIGURE 5.

1D ¹⁹F MAS NMR spectra of 5F‐Trp/U‐¹⁵N labeled wt‐PR (black and gray) and 10Mut (red and pink) in crystalline DRV‐bound forms. (a) DPMAS (bold lines) and CPMAS (thin lines) spectra recorded at 310 K, and (b) DPMAS spectra recorded at 310 K (bold lines) and 289 K (dashed lines). In panel (b), spectra at 310 K (bold lines) are the same as those in (a).

The relative intensity of the W6 resonance, compared to that of the W42 resonance, in the ¹⁹F DPMAS spectrum of DRV‐bound wt‐PR differed from that of 10Mut (Figure 5a). Along with the decrease in intensity seen in the CPMAS data (Figure 5a), this indicates that the W6 region experiences internal motion. The dynamic nature of W6 was further confirmed by comparing ¹⁹F DPMAS spectra recorded at 310 K and 289 K. At 289 K, W6 intensity dropped below that of W42, and the peak nearly disappeared (Figure 5b). Similarly, DRV‐bound 10Mut showed a comparable spectral change at these two temperatures (Figure 5b). The observed dynamics of W6 align with the L5‐W6 peptide bond being the primary site of autoproteolysis (Rose et al., 1993) and with a previous NMR chemical exchange analysis of the L5‐W6 loop's dynamics in HIV‐1 PR (Ishima et al., 1999).

In addition, the ¹⁹F MAS NMR spectra of the crystalline 10Mut in the DRV‐bound form exhibited a resonance pattern similar to that in solution, and the W42 shift direction upon mutation was also similar, indicating different environments for the two W42 sites in 10Mut (Figure 4b). Significantly, the ¹⁹F MAS NMR spectra of DRV‐bound wt‐PR exhibited two resonances for W42, similar to the case of 10Mut in the crystalline state (Figure 5a), whereas the W42 resonance of DRV‐bound wt‐PR in solution did not split (Figure 4b). Interestingly, the −48 ppm resonance of crystalline wt‐PR in the DRV‐bound form at 310 K shifted upfield by 0.2 ppm at 289 K, a change not seen in the 10Mut (Figure 5b), indicating that the dynamic nature of the W42 site affects the chemical shift.

3. DISCUSSION

In this study, conformational dynamics of PR flaps in response to drug‐resistance mutations were investigated using ¹⁹F NMR spectroscopy, which is sensitive to small changes in chemical environment (Emsley & Phillips, 1971; Gerig, ¹⁹⁹⁴; Gronenborn, ²⁰²²). W42 is located at the edge of the flap region, with the 5‐position of the indole ring facing the β‐hairpin structure. Thus, this fluorine is ideally suited to report the flap conformation (Figure S1c). In contrast, W6 is located near the dimer interface and exposed to the solution (Figure S1b). Our results from inhibitor‐free forms of these proteins revealed a progressive downfield shift of the W42 fluorine resonance as the number of mutation sites increased (Figure 4a), demonstrating a correlation between the fluorine chemical shift and the mutations along the drug‐resistance pathway. Given that a correlation between ¹⁹F chemical shift and folding is a standard technique in the protein folding field (Bann et al., 2002; Danielson & Falke, ¹⁹⁹⁶; Ropson & Frieden, ¹⁹⁹²), observation of such “directionality” in the ¹⁹F chemical shift changes as the PR drug‐resistance pathway evolves may suggest a small increase in flap dynamics or solvent exposure of the 5‐position of W42. This interpretation is consistent with previous observations that flap dynamics increase as drug‐resistance mutations accumulate (Roche et al., 2015; Zhang et al., ²⁰¹⁴). In addition, the observation of similar shift changes in both the inhibitor‐free and bound forms (Figure 4) suggests that the ¹⁹F shift reflects electronic changes introduced by the mutations.

In the inhibitor‐free form, ¹⁹F chemical shifts of the W6 in 4Mut, 10Mut, and 11Mut changed only slightly, by ~0.1 ppm, compared to that of wt‐PR (Figure 4a and Table S4). 4Mut contains four mutations: K45I and M46I are located in the flap region, and V82F and I84V are located in a loop adjacent to the flap region (Figure 1c). Given that residues 45 and 46 are located in the flap region in PR, changes in the ensemble of flap conformations in 4Mut could affect the dimer interface that is reported on by the W6 resonance. Indeed, it is well known that flap regions interact and contribute to dimerization, and we previously showed that methyl chemical shifts for amino acids at the tips of the flaps differ between dimer and monomer (Ishima & Louis, 2008). Flap‐flap interaction in the PR dimer was also observed in NMR RDC experiments (Roche et al., 2015). Taken together with these previous observations, we conclude that the ¹⁹F chemical shifts of W6 sensitively detect a remote conformational change in the flap region at the dimer interface.

The mechanism by which a symmetric homodimer recognizes an asymmetric substrate or inhibitor has been a major focus in HIV‐1 PR structural biology (Dreyer et al., 1993; Prabu‐Jeyabalan et al., ^{2000, 2002}). In the current study, we observed that the W42 resonance clearly split into two in the DRV‐bound forms of PR in solution for 10Mut and 11Mut (Figure 4b and Table S5). This observation is qualitatively consistent with results of previous MD simulations, which suggested asymmetric flap structures in the inhibitor‐bound forms (Agniswamy et al., 2016; Cai et al., ²⁰¹⁴; Galiano et al., ²⁰⁰⁹; Henes et al., ²⁰¹⁹; Nakashima et al., ²⁰¹⁶; Rajendran et al., ²⁰²³; Wong‐Sam et al., ²⁰²²; Zhang et al., ²⁰¹⁴). In particular, interdependence between subsites throughout HIV‐1 PR upon inhibitor binding has been repeatedly detected in simulations (Henes et al., 2019; Paulsen et al., ²⁰¹⁷). Among the additional mutations in 10Mut compared to 4Mut, V32I and L76V directly interact with the flap region, and residue 32 directly interfaces with the inhibitor. Thus, we consider that the asymmetry in the W42 resonance is not present in 4Mut but is present in 10Mut and 11Mut.

Interestingly, an almost identical spectral pattern was observed in solution and in the crystalline state of DRV‐bound 10Mut. Although various experimental conditions differed between the solid and the solution state studies, the crystalline and solution ¹⁹F NMR spectra consistently showed two W42 resonances. Thus, it is reasonable to assume that the similarity of the spectral patterns stems from similar local structure characteristics, presumably flap asymmetry in the DRV‐bound form. A pair of W42 resonances was also observed in the crystalline state of wt‐PR, but not in solution. Given that dynamic averaging is more restricted in crystal forms than in solution, it comes as no surprise that the degree of conformational asymmetry was more pronounced in the crystal than in solution for the wt‐PR complex with DRV.

This study presents both ¹⁵N relaxation of the protein backbone and ¹⁹F chemical shifts at selected sites in HIV‐1 PRs. Our use of two NMR observables provided insight into the conformational characteristics of DRV‐bound WT and 11Mut (or 10Mut). Backbone ¹⁵N relaxation showed very similar dynamics throughout the protein between the two subunits, except for a few residues. Importantly, there was no significant reduction in the {¹H}‐¹⁵N NOE in the flap region, indicating that the flap β‐hairpin structure did not melt, even in 11Mut (Figure 3). This observation is consistent with the flaps being predominantly closed in the DRV‐bound form in solution (Cai et al., 2014; Nicholson et al., ¹⁹⁹⁵). Thus, the conformational asymmetry probed by ¹⁹F NMR cannot be due to significant differences in spatial motion. Previous studies using crystal structures and simulations also revealed very subtle changes in dynamic propagation through the protease upon mutation (Henes et al., 2019; Paulsen et al., ²⁰¹⁷). Given that ¹⁹F NMR of proteins often provides only pinpoint information at the labeled sites, complementary protein NMR data are critical to avoid overinterpretation.

4. CONCLUSIONS

Structural asymmetry in HIV‐1 PR flaps for mutants along the drug resistance pathway was monitored using ¹⁹F NMR spectroscopy: the W42 resonance progressively shifted downfield with increasing numbers of amino acid changes, sensitively reporting on the asymmetric flap conformation in the inhibitor‐bound forms. The chemical shift difference was observed not only in solution but also, and more pronouncedly, in crystals of the fluorinated proteins in the presence of DRV. Overall, our data demonstrate that ¹⁹F NMR spectroscopy is a sensitive means of detecting subtle yet pivotal conformational differences in HIV‐1 PR associated with drug‐resistance mutations.

5. MATERIALS AND METHODS

5.1. Protein expression and purification

Genes encoding the following wt‐PR, and its three mutants, were prepared as described previously (Henes et al., 2019): PQITLWKRPL VTIKIGGQLK EALLDTGADD TVLEEMNLPG RWKPKMIGGI GGFIKVRQYD QILIEICGHK AIGTVLVGPT PVNIIGRNLL TQIGCTLNF. Compared to a previously used protein that was sequence‐optimized for NMR experimentation (Khan et al., 2018), the wt sequence used here was different, and the protein contained V32, R41, C67, and C95. 5F‐Trp/U‐¹⁵N labeled PRs were expressed in BL21 (DE3) cells transformed with the corresponding pET‐11a expression plasmids containing the relevant PR coding sequences and grown in modified M9‐defined medium with ¹⁵NH₄Cl as a sole nitrogen source. Cells were grown at 37°C at 225 rpm up to OD 600 nm reached >0.9. The 5‐fluoroindole (CAS# 399‐52‐0, Millipore Sigma) was added 30 min before induction with 0.5 M β‐D‐1‐thiogalactopyranoside (IPTG). After 3–4 h incubation at 37°C, cells were harvested by centrifugation at 4000 rpm for 20 min at 4°C. The harvested cells were lysed by sonication, and PRs were purified from inclusion bodies, dissolved in 8 M urea, followed by fractionation in 8 M urea over gel filtration and HPLC using a previously established protocol (Ishima et al., 2007; Louis et al., ²⁰⁰⁹). Identity and purity of final proteins were assessed by SDS‐PAGE and mass spectrometry (a Bruker LC‐ESI‐TOF system, Bruker Daltonics Inc., Billerica, MA). Purified PRs were stored in the unfolded condition at −80°C. 5F‐Trp/U‐¹⁵N/U‐¹³C labeled wt‐PR and 11Mut were expressed and purified similarly, but by using modified M9 media containing ¹³C glucose.

PRs were folded before NMR experiments using the following protocol. Proteins were diluted with 20 mM formic acid at less than 0.3 mg/mL on ice, followed by further dilution with over threefold volume of 10 mM sodium acetate at pH 6.0 at room temperature. The PRs were dialyzed with 20 mM phosphate buffer at pH 5.8, containing 5 mM DTT. For the drug‐bound proteins, two‐fold excess DRV was added during folding, with an additional equimolar amount of DRV added before transferring the sample to the NMR tube.

5.2. Solution NMR experiments

All solution NMR experiments were performed at 293 K, unless indicated otherwise. 1D ¹⁹F NMR and 2D ¹H‐¹⁵N HSQC spectra of 5F‐Trp/U‐¹⁵N labeled wt‐PR, 4Mut, 10Mut, and 11Mut were recorded on a Bruker 600 AVANCE spectrometer with a CP TXO F/C‐H‐D cryoprobe and a 5‐mm triple‐resonance cryoprobe, respectively. Proteins were prepared at 20–25 μM dimer protein concentration in the absence or presence of DRV, in 20 mM phosphate buffer at pH 5.8, containing 5 mM DTT. NMR experiments in the absence of inhibitor were performed in the following order: (1) 1D ¹⁹F NMR experiment, (2) 2D ¹H‐¹⁵N HSQC experiments, (3) NMR 1D ¹⁹F NMR experiment. A set of two ¹⁹F NMR spectra was used to assess autoproteolytic cleavage of PRs. NMR experiments in the presence of DRV were performed in the following order: (1) 1D ¹⁹F experiment and (2) 2D ¹H‐¹⁵N HSQC experiments. ¹⁹F NMR chemical shifts were referenced to trifluoracetic acid.

¹⁵N relaxation experiments for 5F‐Trp/U‐¹⁵N labeled wt‐PR and 11Mut were carried out at ~200 μM PR concentration in the presence of DRV on a Bruker 600 AVANCE spectrometer with a 5‐mm triple‐resonance cryoprobe. Resonance assignments of backbone amides of 5F‐Trp/U‐¹⁵N/U‐¹³C labeled wt‐PR and 11Mut were performed at ~200 μM PR concentration in the presence of DRV, using HNCACB experiments on a Bruker 800 AVANCE spectrometer, aided by comparison with available assignments in biological magnetic resonance bank (BMRB, 27727, 27731, 27377; see Tables S2, S3). Solution NMR data were processed and analyzed using NMRPipe (Delaglio et al., 1995), ccpNMR (Vranken et al., 2005), and exponential fits were carried out using Matlab software (The MathWorks Inc., Natick, MA).

5.3. Crystallization and MAS NMR experiments

Samples of 5F‐Trp/U‐¹⁵N labeled wt‐PR and 10Mut in the presence of DRV were prepared similarly to those for solution NMR, but with dialyzing PRs with 20 mM acetate buffer at pH 5.5, containing 5 mM DTT. Seed crystals were generated in a 24‐well plate (Hampton Research, Aliso Viejo, CA) using freshly prepared ~10 mg/mL PR complexed with DRV in 50 mM sodium citrate pH 5.2, 300 mM ammonium sulfate, 5 mM DTT. This condition was optimized from previous crystallization experiments (Adachi et al., 2009; Agniswamy et al., ²⁰¹²; Nalam et al., ²⁰¹³; Prabu‐Jeyabalan et al., ²⁰⁰²; Weber et al., ¹⁹⁹⁷). The final microcrystalline samples were prepared by mixing 5 μL of the 10 mg/mL protein, 5 μL of the precipitant solution and 2.5 μL of the micro seeds mixture containing disruption glass beads at 0.5 mm size (Research Products International, Mount, Prospect, IL), and kept at 289 K. Microcrystals were packed into a 1.3 mm rotor for MAS NMR experiments.

¹⁹F MAS NMR experiments were performed with DRV‐bound crystals of wt‐PR and 10Mut on a 500 MHz Bruker AVIII spectrometer outfitted with a 1.3 mm HFX MAS probe. 1D ¹⁹F DPMAS and ¹H‐¹⁹F CPMAS spectra were recorded at a MAS frequency of 60 kHz and T of 310 K (the actual sample temperature was calibrated using KBr as an external temperature sensor; Thurber & Tycko, ²⁰⁰⁹). 1D ¹⁹F MAS NMR spectra were also recorded at a MAS frequency of 40 kHz and T of 289 K. ¹⁹F chemical shifts were referenced using the high‐field mefloquine CF₃ peak set at 8.8 ppm, used as a secondary reference (Guo et al., 2021).

AUTHOR CONTRIBUTIONS

Michel Guerrero: Investigation; formal analysis; writing ‐ review and editing. Roman Zadorozhnyi: Investigation; formal analysis; writing – review and editing. Guowu Lin: Investigation; writing – review and editing. Caitlin M. Quinn: Investigation; writing – review and editing; methodology. Guillermo Calero: Investigation, wriging – review and editing; methodology; Nese Kurt‐Yilmaz: Conceptulation, resources, funding acquisition; writing ‐review and editing. Celia A. Schiffer: Conceptualization; resources; funding acquisition; supervision; writing – review and editing. Tatyana Polenova: Methodology; writing – review and editing; conceptualization; funding acquisition. Angela M. Gronenborn: Conceptualization; funding acquisition; supervision; methodology; writing – review and editing. Rieko Ishima: Conceptulation; investigation; supervision; writing – review and editing; writing – original draft.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing financial interests.

Supporting information

Table S1: Mass numbers of 5F‐Trp/U‐¹⁵N labeled wt‐PR, 4Mut, 10Mut, and 11Mut.

Table S2: Backbone signal assignments of 5F‐Trp/U‐¹⁵N labeled wt‐PR in the DRV bound form, used to analyze the relaxation data in Figure 3.

Table S3: Backbone signal assignments of 5F‐Trp/U‐¹⁵N labeled 11Mut in the DRVbound form, used to analyze the relaxation data in Figure 3.

Table S4: ¹⁹F chemical shifts, intensities and areas of 5F‐Trp resonances of wt‐PR, 4Mut, 10Mut, and 11Mut in solution.

Table S5: ¹⁹F chemical shifts, intensities and areas of 5F‐Trp resonances of wt‐PR, 4Mut, 10Mut, and 11Mut in the DRV‐bound forms in solution.

Table S6: Temperature‐dependence of ¹⁹F chemical shifts and resonance intensities of 11Mut in the DRV‐bound form in solution.

Figure S1: Cartoon of (a) HIV‐1 wt‐PR showing positions of (b) W6 and (c) W42. In all panels, tryptophans are shown by sticks, and position 5 is marked in yellow. W6 indole is exposed to solution (panel b). W42 indole is partially buried in the protein core with R57 and Y59 on one side of the ring and P44 the other side.

Figure S2: Typical mass fractions of 5F‐Trp/U‐¹⁵N labeled wt‐PR, 4Mut, 10Mut and 11Mut, showing 10%–15% of the total having ¹⁹F‐labeling at only one Trp.

Figure S3: Temperature dependence of 11Mut in the presence of DRV in solution. “f” indicates fragment resonance. Note, chemical shift reference was adjusted at 293 K spectrum only.

Guerrero M, Zadorozhnyi R, Lin G, Quinn CM, Calero G, Kurt‐Yilmaz N, et al. Conformational asymmetry of human immunodeficiency virus‐1 protease flaps increases along the darunavir drug resistance pathway. Protein Science. 2026;35(9):e70736. 10.1002/pro.70736

DATA AVAILABILITY STATEMENT

The data that supports the findings of this study are available in the supplementary material of this article.