Site-Directed Nitroxide Spin Labeling of p66.

Nitroxide spin labels (R1) were introduced by conjugation of S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate to surface exposed cysteines via a disulfide linkage (11). Two cysteines, Cys38 and Cys280, are naturally occurring and were mutated to Ala in constructs in which R1 was not attached to these residues (17). Twenty-one sites, depicted on the structures of the p66 and p51 subunits of mature RT (Fig. 1B), were employed for spin labeling: 2 in the finger domain, 2 in the palm domain, 6 in the thumb domain, 7 in the connection domain, and 4 in the RNase H domain.

Three different spin-labeling schemes were used to obtain inter- and intramolecular distances (Fig. 1C). Intermolecular (intersubunit) distances between spin labels located on the same residue in both subunits of the dimer were obtained using single-site labeled p66 (Fig. 1 C, Top). Intramolecular (intrasubunit) distances between spin labels located on different residues were obtained using double-site labeling combined with spin dilution in which double spin-labeled p66 and unlabeled p66 were mixed in a 1:10 ratio (Fig. 1 C, Middle). If the dimer is symmetric, the latter labeling scheme will yield only a single distance, but if asymmetric, 2 distances will be observed. Finally, intermolecular distances between spin labels located on different residues in the 2 subunits of the dimer were obtained from samples comprising an equal mixture of 2 differently single-site labeled p66 chains. Such samples can nominally yield 4 distances, 2 between equivalent residues (known from the single-site labeled samples) and 2 between nonequivalent residues. It should be noted that, since any given p66 homodimer only contains a pair of spin labels, the current DEER data present no issues from potential ghost peaks in DEER-derived distance distributions that can arise in samples containing 3 or more spin labels (18).

Spin-labeled p66 homodimer is readily cleaved by HIV-1 protease to yield the p66/p51 heterodimer. This is illustrated in SI Appendix, Fig. S1 for 2 single-site spin labels within the RNase H domain. In the homodimer, 2 spin labels are present, one on the RNase H domain of each subunit (SI Appendix, Fig. S1A), which gives rise to a modulated DEER echo curve (SI Appendix, Fig. S1C). Upon cleavage of one of the RNase H domains, only a single spin label remains, resulting in a large decrease in modulation depth with any remaining modulation due to residual uncleaved p66 homodimer (SI Appendix, Fig. S1C).

Q-Band DEER.

Q-band pulsed DEER EPR data were acquired on p66 homodimer samples at 50 K. The samples were fully deuterated thereby optimizing spin-label phase memory relaxation times and enabling data to be acquired for long dipolar evolution times (19). This not only permits longer distances between spin labels to be accessed (14, 15), but also increases the accuracy of the distance probability distributions P(r), extracted from the data (16) either through Gaussian analysis (20) or through Tikhonov regularization (21). Examples of DEER echo curves and the corresponding P(r) distributions using the 3 labeling schemes described in the previous section (Fig. 1C) are shown in Figs. 2–4 (as well as SI Appendix, Figs. S2 and S3). A summary of the distances between spin labels is provided in Table 1.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Examples of experimental DEER data and intermolecular P(r) distance distributions obtained for single-site spin-labeled samples of the p66 homodimer RT precursor in the absence and presence of the NNRTI delavirdine. The labeling scheme is shown in Fig. 1 C, Top. (A) Examples of background-corrected DEER echo curves (recorded at Q band and 50 K) in the absence (blue) and presence (red) of the NNRTI. (B) Corresponding DEER-derived P(r) distance distributions (blue, free; red, with NNRTI) obtained by Gaussian fitting using the program DD (20). Raw and background-corrected DEER data for all single-site spin-labeled mutants together with P(r) distributions derived by both Gaussian fitting (20) and validated Tikhonov regulation (21) are provided in SI Appendix, Fig. S2.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Examples of experimental DEER data and intramolecular P(r) distance distributions obtained from double-site spin-labeled samples of the p66 homodimer RT precursor spin diluted 10-fold with unlabeled p66 in the absence and presence of the NNRTI delavirdine. The labeling scheme is shown in Fig. 1 C, Middle. The Left show examples of background-corrected DEER echo curves (recorded at Q band and 50 K) in the absence (blue) and presence (red) of the NNRTI. The corresponding DEER-derived P(r) distance distributions (blue, free; red, with NNRTI) obtained by Gaussian fitting using the program DD (20) are shown on the Right. The positions of the peaks corresponding to the distances in the p66 and p66′ subunits (which correspond to p66 and p51, respectively, in the mature RT heterodimer) for the ligand-free samples are indicated by the dashed lines. The latter are unaffected by the addition of the NNRTI. Raw and background-corrected DEER data for all double-site spin-labeled mutants together with P(r) distributions derived by both Gaussian fitting (20) and validated Tikhonov regularization (21) are provided in SI Appendix, Fig. S3.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Examples of experimental DEER data and intermolecular P(r) distance distributions obtained from mixed spin-labeled samples of the p66 homodimer RT precursor in the absence of a NNRTI. The mixed spin-labeling scheme is shown in Fig. 1 C, Bottom. The DEER data (recorded at Q band and 50 K) and corresponding P(r) distributions for the mixed spin-labeled samples are shown in the Middle row, and the results using the constituent single-site-labeled samples are shown in the Top and Bottom rows. The raw and background-corrected DEER echo curves are shown in red and green, respectively, with the best-fit background shown in black and the best-fit DEER echo curve curves obtained by DD analysis (20) (blue) superimposed on the raw experimental curves. The P(r) distributions obtained by Gaussian analysis using DD (20) and by validated Tikhonov regularization using DeerAnalysis 2016 (21) are displayed in red and gray, respectively. The dashed lines indicate the intermolecular distances arising between spin labels attached to the same residue in the p66 and p66′ subunits. The arrows indicate intermolecular distances arising between spin labels attached to different residues in the p66 and p66′ subunits. While 2 such distances are expected (see Fig. 1B), only a single distance is observed in each case corresponding to the distance between A304C-R1 of p66 and Q394C-R1 of p66′ and between E308C-R1 of p66 and T400C-R1 of p66′; in both cases, comparison with the crystal structure of the mature p66/p51 heterodimer suggest that the second distance (i.e., between Q394C-R1 of p66 and A304C-R1 of p66′ and between T400C-R1 of p66 and E308C-R1 of p66′) overlaps with the shorter distance in the single-site-labeled samples (Bottom). It should also be noted that the intensities of the peaks in the P(r) distributions of the mixed samples (Middle) deviate quite significantly from the expected statistical distribution, presumably reflecting variations in the equilibrium dimer dissociation constants for the 4 dimer species (see Fig. 1C) present in each mixed sample.

Comparison of the intermolecular P(r) distributions for the single-site spin-labeled p66 homodimer indicates that only distances between the thumb domains and between the palm domains are affected upon addition of the NNRTI delavirdine (Fig. 3 and SI Appendix, Fig. S2). In the mature p66/p51 heterodimer, the addition of delavirdine effects a conformational change from a closed thumb/finger conformation within the p66 subunit to an open one but has no effect on domain orientations within the p51 subunit (6, 17).

Examination of the intramolecular P(r) distributions obtained for double-site nitroxide-labeled p66 homodimer with spin dilution (with labels located in the finger and thumb domains) reveals 2 sets of distances (Fig. 3 and SI Appendix, Fig. S3) indicating the presence of asymmetry in the p66 homodimer. Comparison of these distances (Table 1) to the corresponding calculated ones from crystal structures of mature p66/p51 RT (SI Appendix, Table S2) reveals that one of the distances corresponds to the p66 subunit and the other to the p51 subunit. Furthermore, only the distance corresponding to the p66 subunit in mature RT is affected by the addition of the NNRTI delavirdine, providing further proof of structural asymmetry in the p66/p66′ homodimer precursor. (Note the prime designation for the second subunit of the homodimer is used to indicate the subunit whose RNase H domain is cleaved by protease.)

Additional intermolecular distances between spin labels at different sites can be obtained by mixing 2 different single-site-labeled samples and comparing the P(r) distributions of the mixed spin-labeled sample with the 2 corresponding single-site spin-labeled samples (Fig. 4). Given that the DEER echo curve of the mixed spin-label samples may reflect up to 4 distances (2 intramolecular and 2 intermolecular; see Fig. 1C), it is important to note that the P(r) distributions obtained by model-dependent analysis fitting a sum of Gaussians (20) or model free analysis by Tikhonov regularization (21) yields essentially the same results (Fig. 4). In the 2 examples involving spin labels located in the thumb and connection domains, only a single intermolecular distance is observed (Fig. 4, Middle). The P(r) distributions, therefore, do not allow one to ascertain a priori whether the single intermolecular distance peak in the 2 P(r) distributions arises from a single spin pair or 2 spin pairs with similar distances. Comparison with crystal structures of the mature p66/p51 heterodimer, however, predicts that, in both instances, only a single intermolecular distance in the P(r) distribution would be observable as the other intermolecular distance overlaps with one of the intramolecular distances.

A comparison of distances involving the finger, palm, thumb, connection domain, as well as the RNase H domain corresponding to the p66 subunit in the mature p66/p51 heterodimer, reveals excellent agreement between the DEER-derived distances in the p66/p66′ homodimer and the calculated distances in 22 p66/p51 crystal structures (Table 1 and SI Appendix, Table S2, Fig. 5 and SI Appendix, Figs. S4 and S5). Moreover, the DEER-derived distances obtained for free p66/p66′ are much better correlated with the corresponding distances in the crystal structures of p66/p51 in the closed I (free) state, than in open (ligand-bound) states, while the reverse is true of p66/p66′ in the presence of the NNRTI delavirdine (Fig. 5 and SI Appendix, Figs. S4 and S5). The differences in fit are largely a consequence of differences in the intra- and intermolecular distances involving the thumb domain of the p66 subunit of the p66/p66′ homodimer (Fig. 5A and SI Appendix, Figs. S4 and S5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Correlation between experimental DEER-derived distances between spin labels in the homodimeric p66/p66′ RT precursor in the absence and presence of the NNRTI delavirdine and the corresponding calculated distances from crystal structures of the mature p66/p51 RT heterodimer. (A) Correlation plots between the experimental DEER-derived distances and the corresponding calculated distances from closed [PDB 2IAJ (33)] and open [PDB 4B3O (29)] crystal structures. The color coding is as follows: red circles, intra- and intermolecular distances involving the thumb domain; green filled-in circles, intramolecular distances involving the p51 domain; blue circles, all remaining distances. (B) Plot of the R² Pierson correlation coefficient between the experimental DEER-derived distances obtained in the absence (blue) and presence (red) of the NNRTI delavirdine and the calculated distances from 22 crystal structures of the mature p66/p51 heterodimer, providing examples of closed I, closed II, partially open, open I, and open II conformations of the finger/palm/thumb domains of the p66 subunit. The mean distances between spin labels from the crystal structures were obtained from P(r) distributions calculated using the program MMMv2013.2 (34). The numbering of the crystal structures corresponds to the numbering and PDB codes provided in SI Appendix, Table S1.

Modeling the p66 Subunit of the p66/p66′ Precursor from DEER-Derived Distances Based on Crystal Structures of p66/p51.

To model the spatial configuration of the domains within the p66 subunit of the p66/p66′ homodimer, we made use of conjoined rigid body/torsion angle simulated annealing (22) in Xplor-NIH (23, 24) driven by the DEER-derived distances between spin labels (see SI Appendix, SI Structural Modeling Section for details). The distance restraints were represented by square-well potentials which reflect the uncertainty in the mean distances obtained from the DEER-derived P(r) distributions (SI Appendix, Table S3). In these calculations, we excluded distances involving the RNase H domain of p66′ (the domain that is cleaved by HIV-1 protease), and the coordinates of p51 and the connection domain (residues 316–324) of p66 were held fixed, while the finger (residues 1–100 and 113–152), palm (residues 103–110 and 155–241), thumb (residues 244–313), and RNase H (residues 428–600) domains of p66 were allowed to move as rigid bodies with only a few linker residues connecting the domains given torsional degrees of freedom (specifically residues 101–102, 111–112, 153–154, 242–243, 314–315, and 425–427). The rationale for fixing the coordinates of p51 and the connection domain of p66 is that a comparison of the crystal structures of p66/p51 in various free and liganded states (including complexes with nucleic acids and/or NNRTIs) reveals that these regions are structurally invariant within the uncertainties of the present modeling calculations (SI Appendix, Fig. S6). Moreover, the DEER-derived intermolecular distances obtained from samples with a single-site spin label within the connection domain (Table 1) are in near-quantitative agreement with those calculated from crystal structures (SI Appendix, Table S2). In addition, based on the comparison of the DEER-derived experimental distances and the distances in the crystal structures presented in the previous section, upper limits of 20, 8, and 4 Å were placed on backbone rms displacements of the thumb, palm, and RNase H domains, respectively, to prevent undue large-scale domain rearrangements. The modeling calculations were carried out starting from representative crystal structure coordinates of the 5 conformational classes of p66/p51 RT: closed I [1DLO (25)], closed II [3DLK (26)], partially open [3KK3 (27)], open I [1IKW (28)], and open II [4B3O (29)].

The resulting 5 lowest energy structures (1 structure from each set of starting coordinates) obtained with the DEER-derived distances in the absence and presence of the NNRTI delavirdine are shown in Fig. 6 A and B, respectively, and the agreement with the experimental distance restraints is provided in SI Appendix, Table S4. The overall configuration of the domains within the p66 subunit of the p66/p66′ homodimer is the same as that in the mature p66/p51 heterodimer. The major difference between structures in the absence and presence of the NNRTI delavirdine lies in the position of the thumb domain which corresponds to the closed I conformation in the free state (Fig. 6A) and, surprisingly, the partially open conformation in the NNRTI-liganded state (Fig. 6B). Thus, the binding of a NNRTI to the p66/p66′ homodimer induces a smaller structural displacement of the thumb domain toward an open conformation than that observed in crystal structures of mature p66/p51 heterodimer complexed to NNRTIs which are in the open I conformation (SI Appendix, Fig. S7).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Modeling of the spatial disposition of the domains of p66 within the homodimeric p66/p66′ HIV-1 RT precursor in the absence and presence of the NNRTI delavirdine. (A) Unliganded closed state and (B) NNRTI-bound partially open state. Modeling was carried out by conjoined rigid body/torsion angle simulated annealing in Xplor-NIH (23) driven by DEER-derived intra- and intermolecular distances between spin labels starting from 5 different crystal structures, closed I [PDB 1DLO (25)], closed II [PDB 3DLK (26)], partially open [PDB code 3KK3 (27)], open I [PDB code 1IKW (28)], and open II [PDB 4B3O (29)]. The lowest energy structure from each starting coordinate is displayed; note that the backbone rms displacement between structures starting from the same initial coordinates is significantly smaller than that starting from different initial structures (displayed). This is due to small differences in the domain coordinates between the various crystal structures, especially evident in linkers and loops. These calculations do not include distances involving the RNase H domain of the p66′ subunit. The p66′ subunit without the RNase H domain (i.e., equivalent to p51) is displayed as a gray surface representation and held fixed together with the connection domain of p66 since these structural elements constitute a single, essentially invariant unit in the mature p66/p51 heterodimer (SI Appendix, Fig. S6). The domains of p66 (with the exception of the connection domain) were allowed to move as rigid bodies and 2 to 3 residues within the linkers (white) connecting them are given torsional degrees of freedom. The finger, palm, thumb, connection, and RNase H domains are shown as blue, green, cyan, orange, and lilac ribbons, respectively. The superpositions on the Left, displaying all of p66, are best fitted to the p51 portion of the p66′ subunit and the connection domain of the p66 subunit. The superpositions on the Right display the finger, palm, and thumb domains of the p66 subunit best fitted to the finger and palm domains. The red ribbons in the Left show the superposition with the closest crystal structures [1DLO (25) in A, 3KK3 (27) in B]; the equivalent superpositions with other crystal structures are shown in SI Appendix, Fig. S7.

Conformational Space Sampled by the RNase H Domain of the p66′ Subunit of the p66/p66′ Homodimer.

Examples of DEER echo curves, together with the corresponding P(r) distributions for double-site spin-labeled samples with spin dilution (intramolecular interactions) and single-site spin-labeled samples (intermolecular interactions) involving, at least, a single spin label within the RNase H domain are shown in Fig. 7 (as well as SI Appendix, Fig. S3). For all 3 single-site spin-labeled samples (A454C-R1, G504C-R1, and Q547C-R1, Fig. 7, Right), 2 distinct distance distributions are observed. Given that the position of the RNase H domain of the p66 subunit is well defined (Fig. 6), these data indicate that the RNase H domain of p66′ exists in 2 major conformers. Two sets of intramolecular distances for p66′ are also observed for one of the double-site spin-labeled samples (T49C-R1/A454C-R1, Fig. 7, Top Left).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Examples of experimental DEER data and corresponding intra- and intermolecular P(r) distance distributions involving the RNase H domains of the p66/p66′ homodimer RT precursor. The Left-hand panels show examples of double-site spin labeling with spin dilution (1 in 10) giving rise to intramolecular distances (see Fig. 1 C, Middle row); the Right-hand panels show examples of single-site spin labeling giving rise to intermolecular distances (see Fig. 1 C, Top row). The raw and background-corrected DEER echo curves (recorded at Q band and 50 K) are shown in red and green, respectively, with the best-fit background shown in black, and the best-fit DEER echo curve curves obtained by DD analysis (20) (blue) superimposed on the raw experimental curves. The P(r) distributions obtained by Gaussian analysis using DD (20) and by validated Tikhonov regularization using DeerAnalysis 2016 (21) are displayed in red and gray, respectively. In the case of intramolecular distances, the identity of the distance corresponding to the p66 subunit of the mature p66/p51 heterodimer is easily ascertained by comparison with the p66/p51 crystal structures (SI Appendix, Table S2). In one instance involving an intramolecular distance between T39C-R1 and A454C-R1 in the p66′ subunit, and in all 3 examples of intermolecular distances, 2 distances are observed indicating that the RNase H domain of the p66′ subunit exists in 2 conformer ensembles (each characterized by a distribution of states).

To structurally model the RNase H domain of the p66′ subunit, two-membered ensemble calculations were carried out in Xplor-NIH (23, 24) starting from the structural models of the closed conformation of p66/p66′ in the free state. The linker region (residues 426–440) between the connection domain and the RNase H domain of p66′ was given Cartesian degrees of freedom, the RNase H domain of p66′ was allowed to move as a rigid body, all sidechains with the exceptions of those with a spin label were allowed torsion angle motion, and all remaining portions of the p66/p66′ homodimer were held fixed. In the 4 instances where 2 distinct P(r) distributions were observed for the p66′ subunit (see above), a double-well potential was employed (SI Appendix, SI Structural Modeling Section and Eq. S1) which allows distance assignments to switch between ensemble members as it was not possible to associate the distances with the corresponding substates a priori. The results are illustrated in Fig. 8 and the agreement with the DEER-derived experimental distance restraints is given in SI Appendix, Table S4. The 2 conformational states of the p66′ RNase H domain are distinct from one another with the RNase H domain of state I associated with the shorter distances between spin labels, making contact with the p66 subunit (Fig. 8, Left), while that of state II, characterized by the longer distances between spin labels is fully solvent accessible (Fig. 8, Right). It is important to stress that the positions of the p66′ RNase H domain displayed in Fig. 8 represent mean positions and do not report on the extent of the conformational space sampled by the 2 states. This is because the distance restraints employed represent distances between the centroids of pairs of nitroxide spin labels. Clearly, the p66′ RNase H domain of state II must sample a large conelike conformational ensemble, given that the domain makes no contact with the rest of the protein.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

Modeling the conformational states of the RNase H domain of the p66′ subunit within the p66/p66′ homodimer RT precursor. Modeling was carried out by simulated annealing driven by DEER-derived intra- and intermolecular distances between spin labels involving the RNase H domain of the p66′ subunit. Two views are displayed. A superposition of 5 structures (ribbons) is shown for each p66′ RNase H conformer with state I in blue (Left) and state II in red (Right); the linker connecting the RNase H domain to the connection domain is shown in white and was given Cartesian degrees of freedom while the RNase H domain was allowed to move as a rigid body. The remainder of the p66′ subunit and the p66 subunit are shown as surface representations in dark and light gray, respectively. Note that only the mean distance for each spin label/spin label interaction is employed and the spin labels are represented by single centroid positions (SI Appendix); hence, the two computed states of the p66′ RNase H domain represent the mean positions of the ensemble of conformers. While the mean positions of the p66′ RNase H domain of the 2 conformers are fairly well defined, the linker connecting the RNase H domain to the connection domain of p66′ is ill-defined as expected since the DEER-derived distance restraints provide no structural information on the linker other than end-to-end distance.

It seems likely that HIV-1 protease probably cleaves the p66′ RNase H domain preferentially in state II, given that the RNase H domain in that state is fully solvent accessible, but in all likelihood, states I and II rapidly interconvert in solution.

Concluding Remarks.

The current DEER-derived distances between spin labels attached to surface engineered cysteine residues demonstrate unambiguously that the p66/p66′ homodimer RT precursor is asymmetric and that the spatial arrangement and relative orientations of the domains within the 2 subunits (excluding the RNase H domain of p66′) is the same as that in the mature p66/p51 heterodimer with the p66′ subunit serving as a scaffold to stabilize the polymerase active site of the p66 subunit. The data also show that the spatial proximity of the thumb and finger domains of the p66 subunit in the homodimer precursor is modulated by interaction with a NNRTI: In the free state, the thumb and finger domains of the p66 subunit form a closed state that blocks access to double-stranded nucleic acid (Fig. 6A); upon binding of a NNRTI to the palm domain that constitutes the floor of the cleft enclosed by the finger and thumb domains (Fig. 1), a more open conformation is favored that is very similar to that of the partially open state of the mature p66/p51 heterodimer (Fig. 6B). This observation implies that both the p66/p66′ homodimer precursor and the mature p66/p51 heterodimer are targets for NNRTIs. While unusual for a homodimer, structural asymmetry of the p66/p66′ homodimer RT precursor is crucial on several fronts: First, it ensures that the precursor is enzymatically active as a RT (4); second, it ensures that only a single RNase H domain is cleaved by HIV-1 protease to generate the mature p66/p51 heterodimer; and third, no conformational rearrangements are required subsequent to protease cleavage. In addition, the structural asymmetry of the p66/p66′ homodimer precursor accounts for the observation that the p66 homodimer precursor is enzymatically active, albeit less so than the mature p66/p51 heterodimer (30) owing to reduced stability of the precursor homodimer relative to the mature heterodimer (31, 32).

In the p66/p66′ homodimer precursor, only the cleavage site located at the N terminus of the first β-strand of the p66′ RNase H domain is accessible to proteolytic attack, while the same site in the p66 subunit is buried and, hence, resistant to proteolysis. The RNase H domain of the p66′ subunit samples, at least, 2 major conformers (Fig. 8), one of which interacts with the p66 subunit (state I), while the other is fully solvent accessible (state II). We speculate that interaction with and cleavage by HIV-1 protease only involves state II and that, in solution, states I and II rapidly interconvert between one another. Moreover, it is also likely that the presence of an excited form of state I involving transient partial unfolding at the N-terminal end of the β1 strand of the p66′ RNase H domain is required to fully unmask the cleavage site and permit binding to HIV-1 protease. Such an excited form is likely to be favored in state II by the presence of the long disordered linker between the connection and the RNase H domains of p66′, as well as the absence of any stabilizing interactions between the RNase H domain of p66′ and the other domains in the p66/p66′ precursor.