- Email: [email protected]
Structure of unliganded HIV-1 reverse transcriptase at 2.7 å resolution: implications of conformational changes for polymerization and inhibition mechanisms
Research Article
853
Structure of unliganded HIV-1 reverse transcriptase at 2.7 Å resolution: implications of conformational changes for polymerization and inhibition mechanisms Y Hsiou1, J Ding1, K Das1, AD Clark, Jr1, SH Hughes2 and E Arnold1* Background: HIV-1 reverse transcriptase (RT) is a major target for anti-HIV drugs. A considerable amount of information about the structure of RT is available, both unliganded and in complex with template-primer or nonnucleoside RT inhibitors (NNRTIs). But significant conformational differences in the p66 polymerase domain among the unliganded structures have complicated the interpretation of these data, leading to different proposals for the mechanisms of polymerization and inhibition. Results: We report the structure of an unliganded RT at 2.7 Å resolution, crystallized in space group C2 with a crystal packing similar to that of the RT–NNRTI complexes. The p66 thumb subdomain is folded into the DNAbinding cleft. Comparison of the unliganded RT structures with the DNA-bound RT and the NNRTI-bound RT structures reveals that the p66 thumb subdomain can exhibit two different upright conformations. In the DNA-bound RT, the p66 thumb subdomain adopts an upright position that can be described as resulting from a rigid-body rotation of the p66 thumb along the ‘thumb’s knuckle’ located near residues Trp239 (in strand b14) and Val317 (in b15) compared with the thumb position in the unliganded RT structure. NNRTI binding induces an additional hinge movement of the p66 thumb near the thumb’s knuckle, causing the p66 thumb to adopt a configuration that is even more extended than in the DNA-bound RT structure.
Addresses: 1Center for Advanced Biotechnology and Medicine (CABM) and Rutgers University Chemistry Department, 679 Hoes Lane, Piscataway, NJ 08854-5638, USA and 2ABLBasic Research Program, NCI-Frederick Cancer Research and Development Center, PO Box B, Frederick, MD 21702-1201, USA. *Corresponding author. E-mail: [email protected] Key words: AIDS, drug design, mechanism of polymerization, non-nucleoside inhibition mechanism, polymerase structure Received: 4 April 1996 Revisions requested: 2 May 1996 Revisions received: 28 May 1996 Accepted: 30 May 1996 Structure 15 July 1996, 4:853–860 © Current Biology Ltd ISSN 0969-2126
Conclusions: The p66 thumb subdomain is extremely flexible. NNRTI binding induces both short-range and long-range structural distortions in several domains of RT, which are expected to alter the position and conformation of the template-primer. These changes may account for the inhibition of polymerization and the alteration of the cleavage specificity of RNase H by NNRTI binding.
Introduction The reverse transcriptase (RT) of human immunodeficiency virus type 1 (HIV-1) is an attractive target for antiviral drugs (see recent reviews in [1–4]). HIV-1 RT is a heterodimer that consists of a 66 kDa subunit (p66) and a 51 kDa subunit (p51) (Fig. 1). The p66 subunit is composed of a polymerase domain and an RNase H domain. The p51 subunit contains only the polymerase domain. Both the p66 and p51 polymerase domains contain four subdomains (fingers, palm, thumb and connection) which are arranged differently in the two subunits. Only the p66 polymerase domain has a DNA-binding cleft, a functional polymerase active site, and a site for binding non-nucleoside inhibitors (NNRTIs). Structures of HIV-1 RT in complexes with various nonnucleoside RT inhibitors (NNRTIs) [5–10], and in complex with a 19-mer/18-mer double-stranded DNA and an antibody Fab fragment [11] have been reported. Structures of unliganded HIV-1 RT have been solved in three
different crystal forms ([12,13], EA and coworkers, unpublished data). It is reasonable to expect that comparing the structures of unliganded HIV-1 RT with those of HIV-1 RT complexed with NNRTIs or nucleic acid should yield valuable insights into the mechanisms of polymerization and inhibition. This information should be useful in the development of new or improved inhibitors of HIV-1 RT. However, in the unliganded HIV-1 RT structures already reported, there are significant conformational differences in the p66 polymerase domain, particularly in the conformation of the b6–b10–b9 sheet and the position of the p66 thumb subdomain ([12,13], EA and coworkers, unpublished data). This has led to controversy regarding the role of the p66 thumb in the polymerization reaction, and different interpretations of the inhibition mechanism of HIV-1 RT by NNRTIs. Here, we report the structure of an unliganded HIV-1 RT crystallized in space group C2, a crystal form closely related to, but not isomorphous with, that of a number of
854
Structure 1996, Vol 4 No 7
Figure 1 Stereo Ca trace of the unliganded HIV-1 RT, drawn using the program RIBBONS [32]. The heterodimer is colored by subdomains: fingers, blue; palm, red; thumb, green; connection, yellow; and RNase H, orange. Every 50th residue is labeled: p66, 1–556 and p51, 1001–1427.
HIV-1 RT–NNRTI complexes reported previously. This unliganded HIV-1 RT structure has a molecular packing in the crystal lattice similar to that of the NNRTI-bound HIV-1 RT structures crystallized in the C2 crystal form. Comparison of these HIV-1 RT structures should therefore help to distinguish conformational changes induced by NNRTI binding from changes caused by a different crystal packing arrangement.
Results and discussion Overview of HIV-1 RT structure and flexibility
Although the unliganded HIV-1 RT structure reported here was determined in a crystal form similar to that present in crystals of the HIV-1 RT–NNRTI complexes, the cell parameters differ considerably. The unit cell of the unliganded RT crystals is approximately 4% (10 Å) longer along the a axis and 10% (10 Å) shorter along the c axis relative to that of the NNRTI-bound RT. This suggests that the repositioning of the p66 thumb (see later discussion) causes a considerable change in crystal packing. In the structures of RT–NNRTI complexes with the C2 crystal form, two symmetry-related molecules juxtapose in such a way that the p66 fingers subdomain of one molecule contacts both the RNase H domain and the tip of the p66 thumb subdomain of the other. In the crystal lattice of the unliganded RT reported here, the RT molecules are arranged in a manner similar to that in the RT–NNRTI crystals. Even though the p66 thumb subdomain is folded into the DNA-binding cleft, the same regions of the p66 thumb subdomain and the fingers subdomain of the symmetry-related molecule interact with each other, although with different specific contacts. Compared with the relative positions of symmetry-related molecules in the RT–NNRTI structures, the symmetry-related molecules in this unliganded RT structure have a 20 Å translation in the xz plane relative to each other. This translation permits the RT molecules in this structure to pack more closely along the c axis, and to expand slightly along the a axis,
which would account for the changes in the observed cell parameters. The structure of unliganded HIV-1 RT has been determined in several crystal forms. Rodgers et al. [12] reported the structure of an unliganded HIV-1 RT in the space group C2 with 4 molecules per asymmetric unit. We have solved the structure of an unliganded HIV-1 RT complexed with an antibody Fab fragment, which was crystallized in space group P3212 with unit cell dimensions isomorphous with those of the RT–DNA–Fab complex (EA and coworkers, unpublished data). In these unliganded HIV-1 RT structures, the p66 thumb subdomain is folded down into the DNA-binding cleft. In the structure of unliganded RT reported here, the p66 thumb subdomain is also folded down into the DNA-binding cleft. Therefore, with dramatically different crystal packing arrangements, the RT molecules in these crystal lattices exhibits strikingly similar folding. The rms deviations between the structures of RT reported here and by Rodgers et al. are 1.17 Å, 1.13 Å, and 0.97 Å based on the superposition of Ca atoms for the whole RT molecule (925 Ca atoms), the p66 subunit (475 Ca atoms), and the p51 subunit (393 Ca atoms), respectively. The largest difference between subdomains is that the p66 thumb subdomain in the unliganded RT structure reported here is rotated further into the DNA-binding cleft by 7° relative to the structure reported by Rodgers et al. This displacement could be the result of a relatively closer crystal packing arrangement in the current crystal form. Esnouf et al. [13] also described the structure of an unliganded HIV1 RT that was crystallized in space group P212121. In this structure, the p66 thumb subdomain is in an upright conformation similar to that seen in the structures of RT–NNRTI complexes and the RT–DNA complex. Moreover, the b6–b10–b9 sheet of the p66 palm subdomain moves toward the p66 thumb subdomain in a manner different from that seen in the other unliganded
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
RT structures ([12], EA and coworkers, unpublished data). In the structure of unliganded HIV-1 RT reported here, the position and conformation of the b6–b10–b9 sheet agree very well with that found in the unliganded RT structures reported earlier ([12], EA and coworkers, unpublished data). The different conformation of the p66 thumb subdomain and the repositioning of the b6–b10–b9 sheet in the unliganded RT reported by Esnouf et al. [13] may be a consequence of the method they used to prepare the unliganded HIV-1 RT crystals. In that study, crystals were obtained by soaking out a weakly-bound NNRTI inhibitor (HEPT) from crystals of the HIV-1 RT–HEPT complex [13]. Examination of the molecular packing in the structure of unliganded HIV-1 RT reported by Esnouf et al. indicates that the p66 thumb subdomain is interlocked between the p66 thumb and RNase H subdomains of the symmetryrelated RT molecule. It appears that the p66 thumb subdomain in this structure of unliganded RT is restricted by crystal packing in such a way that it cannot fold down into the DNA-binding cleft despite the NNRTI HEPT having been soaked out. At any rate, the differences in the position of the p66 thumb subdomain in different HIV-1 RT structures provides ample evidence for high flexibility of the p66 thumb subdomain; a flexibility that may play a vital role in the binding and translocation of the nucleic acid substrate.
855
The structures of RT complexed with the NNRTIs show that the NNIBP is formed primarily by amino-acid residues of the b5–b6 loop, b6, the b9–b10 hairpin, the b12–b13 hairpin, and b15 of p66, and the b7–b8 connecting loop of p51 [8]. In the structure of the unliganded RT reported here and that reported by Rodgers et al., and in the structure of the RT–DNA–Fab complex, the region corresponding to the NNIBP in the RT–NNRTI structures is occupied by the side chains of Tyr181, Tyr188, Phe227, and Trp229. When an NNRTI binds, the NNIBP is created by the reorientation of the side chains of Tyr181 and Tyr188, and by the repositioning of the b12–b13–b14 sheet that contains the ‘primer grip’. In the unliganded RT structure reported by Esnouf et al., a small cavity is present at the NNIBP region and the b6–b10–b9 and b12–b13–b14 sheets (using our nomenclature for secondary structural elements [11]) are shifted, along with the side chains of residues Tyr181, Tyr183, and Tyr188, toward this small residual cavity [13]. This reorientation of the side chain of residue Tyr183 has not been seen in other unliganded RTs ([12], EA and coworkers, unpublished data). Because the p66 thumb subdomain in the Esnouf et al. structure is apparently constrained in a conformation similar to that in the RT–NNRTI complex, the displacement of the side chain of Tyr183 and the altered positions of the b6–b10–b9 and b12–b13–b14 sheets may be caused by an incomplete collapse of the empty drugbinding pocket relative to the other HIV-1 RT structures not containing NNRTIs.
Polymerase active site and NNRTI-binding pocket (NNIBP)
The polymerase active site is located at the b6–b10–b9 sheet of the p66 palm subdomain. The position of the b9–b10 hairpin turn in the unliganded form is nearly identical to that in the NNRTI-bound HIV-1 RT, suggesting that the position of the b6–b10–b9 sheet is not significantly affected by the binding of an NNRTI. The Tyr–Met–Asp–Asp motif (residues 183–186) in both the p66 and p51 subunits is in an unusual bII′ turn conformation that has been observed in several other HIV-1 RT structures [7–10,12,13]. Clear density peaks were found to have good coordination geometry with Od1 of Asp185 (2.8 Å) and Od1 of Asp186 (2.8 Å) at the p66 polymerase active site, and with the Od1 of Asp443 (3.1 Å), the Oε2 of Glu478 (2.4 Å), and the Od1 and Od2 of Asp498 (2.5 Å and 2.4 Å, respectively), at the RNase H active site. Due to the current resolution limit, we cannot distinguish whether these peaks correspond to water or magnesium ions. Structures of the Moloney leukemia virus (MMLV) RT fragment [14] and the fingers and palm subdomains of the p66 subunit in this unliganded HIV-1 RT (rms deviation of 1.4 Å based on 57 Ca atoms in the fingers and 56 Ca atoms of the palm subdomain) were superposed. The results indicate that the position of the electron density peak at the polymerase active site in this unliganded HIV-1 RT is quite close to that of the Mn2+ ion seen in the MMLV RT structure.
Calculation of the solvent-accessible surface area based on a 1.6 Å probe for the current unliganded HIV-1 RT structure reveals two surface depressions at the base of the NNRTI-binding pocket that are plausible candidates for the entrance to the pocket (Fig. 2). These two surface depressions were also observed in the RT–DNA–Fab complex. One surface depression is located at the proposed entrance near Lys101, Lys103, Val179 (p66), and Glu138 (p51) (site 1) [9]. The other one is near the base of the thumb subdomain between two adjacent polypeptide segments: the b5–b6 connecting loop (containing Lys101 and Lys103) and the b13–b14 hairpin (containing Pro236 and Lys238) (site 2). Both sites are exposed to solvent. It is possible that an NNRTI can approach the binding pocket via either site and displace the Tyr181 and Tyr188 side chains. Once the NNRTI is bound, however, only the putative entrance near site 1 remains accessible [9]. Site 2 disappears in the NNRTI-bound HIV-1 RT structure as a result of a hinge-like movement of the b12–b13–b14 sheet relative to the b6–b10–b9 sheet. The exact inhibitor entrance and the mechanism(s) of entry are still unclear. Conformational distortions induced by NNRTI binding
As noted by Rodgers et al., the displacement of the p66 thumb subdomain is the most dramatic change between the structures of the unliganded RT, the RT–DNA–Fab
856
Structure 1996, Vol 4 No 7
Figure 2 Stereoview of the solvent-accessible surface of the p66 palm subdomain in the unliganded RT structure showing the two surface depressions (indicated by arrows) near the NNIBP that are potential entrances to the inhibitor-binding pocket. One (site 1) is located at the proposed entrance near residues Lys101, Lys103, Val179 (p66) and Glu138 (p51). The other (site 2) is between Lys101/Lys103 and Pro236/Lys238, that is, at the base of the p66 thumb subdomain. The latter one disappears after binding of an NNRTI. Water molecules (cyan) are seen in both surface depressions.
complex, and the RT–NNRTI complexes. The structures of RT in this unliganded form and in the DNA-bound form superimposed quite well with an rms deviation of 1.3 Å based on all the Ca atoms, excluding the p66 thumb subdomain. Superposition of the structure of the RT–DNA–Fab complex and the unliganded form of HIV-1 RT, based on the Ca atoms of the b6–b10–b9 sheet, reveals that the p66 thumb subdomain, in the RT–DNA–Fab complex, is rotated 30° away from the p66 fingers and the polymerase active site (Fig. 3a), similar to that noted by Rodgers et al. This rotation functions like a thumb’s knuckle and the axis of the rotation is located at the base of thumb (around residue Trp239 and Val317). The other subdomains of the polymerase domain and the RNase H domain in p66 remain in similar positions in the unliganded RT and the RT–DNA–Fab complex.
the p66 connection and RNase H subdomains move away relative to the fingers subdomain with 13° and 15° rotations, respectively, compared with the corresponding positions in the unliganded RT (Fig. 3b). The position of the p66 fingers subdomain in the HIV-1 RT–NNRTI complexes is relatively unchanged (less than 4° rotation when compared with the unliganded RT). The subdomains in the p51 subunit of the NNRTI-bound RT also move (10°, 12°, 15°, and 17° for fingers, palm, thumb, and connection subdomains, respectively). As a result, the intersubunit interactions at the heterodimer interfaces are maintained in both the unliganded RT and the RT–NNRTI complex, suggesting that the heterodimer interface is relatively stable. It is interesting to note that the binding of a relatively small inhibitor molecule is accompanied by such dramatic rearrangements of the subdomains.
The structures of the unliganded RT and the NNRTIbound RT agree quite nicely within individual subdomains (rms deviations are within 1.5 Å, based on the Ca atoms in each subdomain). However, superposition of the whole molecules does not show as good agreement (rms deviation of 2.1 Å). If the structures are superimposed based on the Ca atoms of the b6–b10–b9 sheet, binding an NNRTI causes a hinge-like movement between the b6–b10–b9 sheet and the b12–b13–b14 sheet. The hinging is near, but not coincident with the axis of the thumb’s knuckle motion (Fig. 3b). This hinge-like movement, which is similar to the internal swivel motion suggested by Jäger et al. [15], extends the position of the p66 thumb subdomain beyond the upright position seen in the RT–DNA complex. In the inhibitor-bound HIV-1 RT structures, the p66 thumb subdomain is rotated approximately 40° relative to the p66 fingers subdomain compared with its position in the unliganded RT. This rotation is significantly larger than the corresponding 30° rotation in the RT–DNA–Fab complex. Comparing the unliganded RT and the DNA-bound RT shows that the hinge movement is coupled with the binding of NNRTIs. Associated with the hinge movement of the p66 thumb,
Mechanistic implications for the inhibition by NNRTIs
Comparison of the available structures of HIV-1 RT indicates that the binding of an NNRTI causes both shortrange and long-range distortions of the HIV-1 RT structure. The short-range distortions include conformational changes of the amino acid residues and/or structural elements that form the NNIBP, such as reorientation of the side chains of Tyr181 and Tyr188, and displacement of the b12–b13–b14 sheet. These distortions affect the precise geometry of the polymerase active site. The long-range distortions involve a hinging movement of the p66 thumb subdomain that results in the displacement of the p66 connection subdomain, the RNase H domain, and the p51 subunit relative to the polymerase active site. The observation that the chemical step of DNA polymerization is slowed by NNRTI binding [16,17] may be explained by the short-range structural distortions induced by NNRTI binding, including the conformational changes of the primer grip. Both short-range and long-range structural distortions induced by NNRTI binding alter the interactions between the nucleic acid substrates and HIV-1 RT, and affect the precise positioning of the template-primer relative to the polymerase and the RNase H active sites. NNRTI-induced changes in RT
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
857
Figure 3 Superposition of (a) unliganded RT and RT–DNA–Fab complex and (b) unliganded RT and RT–a-APA (a-anilinophenylacetamide) complex based on 89 Ca atoms in the p66 palm subdomain, including the b6–b10–b9 region. The unliganded RT is shown in red, RT–a-APA in blue, and RT–DNA–Fab in green. A comparison of the two superpositions reveals that NNRTI binding appears to be accompanied by a long-range distortion that is coupled with a hinge motion (indicated by curved arrows) between the b6–b10–b9 and b12–b13–b14 sheets at the p66 palm subdomain (within the circle). The different positions of the thumb in different HIV-1 RT structures supports the idea that this subdomain could play a role during polymerization.
interactions with nucleic acid substrates may partly account for the observations that NNRTI binding can alter the affinity of HIV-1 RT for the template-primer [16,18] as well as the specificity of RNase H cleavage [19]. HIV-1 RT mutations near the polymerase active site that had a greater effect on RNase H activity than on polymerase activity [20–22] have been described, supporting the idea that communication between distant sites on HIV-1 RT (e.g., the polymerase and RNase H active sites, the NNIBP, and other sites that interact with template-primer) can occur via the template-primer [8,11,22,23]. It is difficult to be certain, however, that repositioning of the template-primer would be solely responsible for the effect of NNRTI binding on RNase H cleavage specificity, as binding does cause other long-range movements of subdomains of both p51 and p66 relative to each other. The long-range distortions caused by NNRTIs have the potential to affect other RT activities such as strand transfer, strand displacement, and recognition of the tRNA used for priming reverse transcription.
Biological implications The cumulative biological, genetic, and structural information about HIV-1 reverse transcriptase (RT) has considerably enhanced our understanding of the mechanisms of polymerization and inhibition. However, the structures of unliganded HIV-1 RT previously reported differ
significantly in the conformation of the polymerase domain of the p66 subunit. This has led to controversy concerning the involvement of the p66 thumb subdomain in the mechanism of DNA polymerization and the mechanism of inhibition by non-nucleoside inhibitors (NNRTIs). We report here the crystal structure of an unliganded HIV-1 RT crystallized in space group C2, with molecular packing similar to that of the NNRTI-bound HIV-1 RT complexes. The overall folding of this enzyme is similar to that determined by Rodgers et al. [12]. In these two structures, the p66 thumb subdomain is folded into the DNA-binding cleft. The differences observed among the unliganded HIV-1 RTs in different crystal forms demonstrates the considerable flexibility of the p66 thumb subdomain, which may play an important role in the recognition of nucleic acid substrates and the translocation of the enzyme along the nucleic acid after incorporation of each new nucleotide. It is likely that the flexibility of the p66 thumb allows it to facilitate the repositioning (and ‘reloading’) of RT during polymerization. A comparison of the structures of the NNRTI-bound RT complexes with those of unliganded RT and the HIV-1 RT–DNA–Fab complex [11] shows that the binding of
858
Structure 1996, Vol 4 No 7
an NNRTI causes both short-range and long-range structural distortions. A hinge movement at the base of the p66 thumb subdomain (near the ‘thumb’s knuckle’) appears to constrain the thumb in a wide-open configuration in the HIV-1 RT–NNRTI complexes, which is extended beyond the upright configuration seen in the HIV-1 RT–DNA–Fab structure. Biochemical and structural data suggest that the conformational changes accompanying NNRTI binding alter the positioning of the template-primer relative to the polymerase and RNase H active sites. This could potentially account both for the inhibition of RT activity (short-range effects) and the alteration of RNase H cleavage specificity (longrange effects) by NNRTIs.
Materials and methods Crystallization and data collection The unliganded HIV-1 RT was crystallized ([24] and references therein) using modified conditions reported by Kohlstaedt et al. for crystallizing the HIV-1 RT/nevirapine complex [5]. The initial crystals were grown in the absence of NNRTIs by streak-seeding with crushed RT/a-APA crystals. After crystals of unliganded HIV-1 RT were obtained, they were used for further seeding. Crystals grew to an average size of 0.1 mm × 0.2 mm × 0.4 mm after two weeks and had the symmetry of space group C2, which is the same as that of the HIV-1 RT–NNRTI complexes. The cooled diffraction data were collected at the F1 beam line (l = 0.91 Å) of the Cornell High Energy Synchrotron Source (CHESS) from seven crystals cooled to –10°C. Data were recorded on Fuji imaging plates with an overall Rmerge of 11.8% and a completeness of 76.6% to 3.3 Å resolution. Another diffraction dataset was collected
at the CHESS F1 beam line from a single crystal at –165°C using the flash-cooled technique [25,26]. This dataset, which was used for structural determination and refinement, has an overall Rmerge of 5.5%, a completeness of 99.5% to 2.7 Å resolution (no s cutoff was used, and the average measurement redundancy was 3.5). The HKL package [27,28] was initially used to process, reduce, scale, and postrefine the data. Final data were scaled using SCALA (CCP4) [29]. A TAMM derivative diffraction dataset was collected at the CHESS A1 beam line and was scaled to the native dataset using FHSCAL of CCP4. The statistics of all diffraction data are summarized in Table 1.
Structure determination and refinement The structure of the unliganded HIV-1 RT was determined with the molecular replacement (MR) technique as implemented in X-PLOR [30], using the diffraction data from the frozen crystal and the coordinates of the unliganded RT provided by Rodgers et al. as the starting model [12]. The rotation function search and translation function search yielded one distinct solution. The correctness of the MR solution was confirmed by the positions of mercury atoms in a difference Fourier map computed between the TAMM derivative dataset and the native dataset phased with the initial MR model. A subset of 6% of the frozen dataset was separated for calculating the free R-factor to monitor the progress of refinement. No subset of data was generated for the free R-factor calculation for the cooled diffraction dataset due to the relative incompleteness of the dataset. Iterative refinement and model building were performed using the programs X-PLOR [30] and O [31]. Omit maps were calculated for both the frozen and the cooled datasets and then averaged between the frozen and cooled datasets to improve the map quality and reduce model bias (KD et al. unpublished data). The final refinement converged to an R-factor of 24.9% and a free R-factor of 33.6% for the frozen crystal dataset, and an Rfactor of 23.0% for the cooled dataset. The fit of the atomic model into the final 2mFobs–Fcalc electron density map in the region of the p66 connection subdomain is shown in Figure 4. A total of 267 water
Table 1 Summary of crystallographic data and refinement statistics. Datasets
Native (cooled)
Native (frozen)
TAMM (frozen)
7 79 –10 240.3,71.5,95.5 104.4 3.3 24 727 18 345 76.6 11.8
1 112 –165 235.5,70.3,93.3 106.1 2.7 142 770 41 049 99.5 5.5
1 92 –165 234.0,69.7,93.0 106.2 3.7 46 901 15 521 99.4 11.2 21.8
Diffraction data statistics No. of crystals No. of images Temperature (°C) a,b,c (Å) b (°) Resolution (Å) Observed reflections (%) Unique reflections (%) Completeness (%) Rmerge* (%) Rderiv† (%) Refinement statistics Resolution range (Å) No. of reflections‡ R factor# (%) Free R factor (%) No. of protein atoms No. of water molecules Average B factor (Å2) Rms deviations Bond (Å) Angles (°)
8.0–2.7 38 310 24.9% 33.6% 7692 267 44 0.01 2.0
*Rmerge =S|Iobs–|/S.†Rderiv =S|IPH–IP|/SIP. ‡No. of reflections used in refinement have F≥2s(F). #R=S||Fobs|–|Fcalc||/S|Fobs|.
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
859
Figure 4 Stereoview of a portion of a (2mFobs–Fcalc) difference Fourier map at the p66 connection subdomain, at 2.7 Å resolution. The phases were computed from the current atomic model and the map is contoured at 1.4s. The side chain and the carboxyl groups are well defined in the electron density map.
molecules were found that have good stereochemistry. Some residues are apparently disordered and have very weak electron density. These residues have been modeled as alanines, including residues 22, 28, 64, 70–71, 134–136, 139, 169, 177, 194, 203, 218, 224, 242, 278, 281, 291, 297, 305, 311–312, 334, 336, 356–359, 404, 407, 413–414, 424, 448, 451, 512, 514, 527, 549–551, and 556 in the p66 subunit, and residues 36, 67–69, 88–89, 173, 197, 238, 281–282, 295, 297, 305–306, 308, 356–358, 361–362, 407, 424–425, and 427 in the p51 subunit. A part of the connecting loop between aF and b13 in the p51 subunit (residues 219-231) has weak electron density and thus has not been modeled. Final refinement statistics are listed in Table 1.
Accession numbers Coordinates for the final model of the native structure refined using the frozen dataset have been deposited with the Brookhaven Protein Data Bank (entry code1DLO).
Acknowledgements We thank David Rodgers, Steve Gamblin, and Steve Harrison for providing the coordinates of their unliganded HIV-1 RT structure prior to publication. We thank the other members of the Arnold and Hughes laboratory, including Pat Clark, Peter Frank, Andrew Holmes, Wade Huber, Karen Lentz, and Dave Miller, Dawn Resnick, Chris Tantillo and Wanyi Zhang for their assistance and helpful discussion; the members of CHESS staff for their help with X-ray data collection; the Frederick Biomedical Supercomputing Center for providing Cray-YMP computing time; the Keck Structural Biology Computing Center at CABM; and Andrew Holmes for critical reading of the manuscript. The research in EA’s laboratory has been supported by NIH grants (AI 27690 and AI 36144), Janssen Research Foundation, and Hoechst General Pharma Research, a grant from NCI-ABL and the Keck Foundation, and SHH’s laboratory is sponsored in part by the National Cancer Institute, DHHS, under contract with ABL and by NIGMS.
References 1. Larder, B.A. (1993). Inhibitors of HIV reverse transcriptase as antiviral agents and drug resistance. In Reverse Transcriptase. (Skalka, A.M. & Goff, S.P., eds), pp. 205–222, Cold Spring Harbor Laboratory Press, New York. 2. Tantillo, C., et al., & Arnold, E. (1994). Locations of anti-AIDS drug binding sites and resistance mutations in the three-dimensional structure of HIV-1 reverse transcriptase: implications for mechanisms of drug inhibition and resistance. J. Mol. Biol. 243, 369–387. 3. De Clercq, E. (1995). Toward improved anti-HIV chemotherapy: Therapeutic strategies for intervention with HIV infections. J. Med. Chem. 38, 2491–2517. 4. Arnold, E., Das, K., Ding, J., Yadav, P.N.S., Hsiou, Y., Boyer, P.L. & Hughes, S.H. (1996). Targeting HIV reverse transcriptase for antiAIDS drug design. Drug Design and Discovery. 13, 29–47.
5. Kohlstaedt, L.A., Wang, J., Friedman, J.M., Rice, P.A. & Steitz, T.A. (1992). Crystal structure at 3.5 Å resolution of HIV-1 reverse transcriptase complexed with an inhibitor. Science 256, 1783–1790. 6. Smerdon, S.J., et al., & Steitz, T.A. (1994). Structure of the binding site for non-nucleoside inhibitors of the reverse transcriptase of human immunodeficiency virus type 1. Proc. Natl. Acad. Sci. USA 91, 3911–3915. 7. Ren, J., et al., & Stammers, D. (1995). High resolution structures of HIV-1 RT from four RT-inhibitor complexes. Nat. Struct. Biol. 2, 293–302. 8. Ding, J., et al., & Arnold, E. (1995). Structure of HIV-1 reverse transcriptase in a complex with the non-nucleoside inhibitor a-APA R 95845 at 2.8 Å resolution. Structure 3, 365–379. 9. Ding, J., et al., & Arnold, E. (1995). Structure of HIV-1 RT/TIBO R 86183 reveals similarity in the binding of diverse non-nucleoside inhibitors. Nat. Struct. Biol. 2, 407–415. 10. Ren, J., et al., & Stuart, D. (1995). The structure of HIV-1 reverse transcriptase complexed with 9-chloro-TIBO: lessons for inhibitor design. Structure 3, 915–926. 11. Jacobo-Molina, A., et al., & Arnold, E. (1993). Crystal structure of human immunodeficiency virus type 1 reverse transcriptase complexed with double-stranded DNA at 3.0 Å resolution shows bent DNA. Proc. Natl. Acad. Sci. USA 90, 6320–6324. 12. Rodgers, D.W., et al., & Harrison, S.C. (1995). The structure of unliganded reverse transcriptase from the human immunodeficiency virus type 1. Proc. Natl. Acad. Sci. USA 92, 1222–1226. 13. Esnouf, R., Ren, J., Ross, R., Jones, Y., Stammers, D. & Stuart, D. (1995). Mechanism of inhibition of HIV-1 reverse transcriptase by nonnucleoside inhibitors. Nat. Struct. Biol. 2, 303–308. 14. Georgiadis, M.M., Jessen, S.M., Ogata, C.M., Telesnitsky, A., Goff, S.P. & Hendrickson, W.A. (1995). Mechanistic implications from the structure of a catalytic fragment of Moloney murine leukemia virus reverse transcriptase. Structure 3, 879–892. 15. Jäger, J., Smerdon, S.J., Wang, J., Boisvert, D.C. & Steitz, T.A. (1994). Comparison of three different crystal forms shows HIV-1 reverse transcriptase displays an internal swivel motion. Structure 2, 869–876. 16. Spence, R.A., Kati, W.M., Anderson, K.S. & Johnson, K.A. (1995). Mechanism of inhibition of HIV-1 reverse transcriptase by nonnucleoside inhibitors. Science 267, 988–993. 17. Rittinger, K., Divita, G. & Goody, R. (1995). Human immunodeficiency virus reverse transcriptase substrate-induced conformational changes and the mechanism of inhibition by non-nucleoside inhibitors. Proc. Natl. Acad. Sci. USA 92, 8046–8049. 18. Debyser, Z., et al., & De Clercq, E. (1991). An antiviral target on reverse transcriptase of human immunodeficiency virus type 1 revealed by tetrahydroimidazo-[4,5,1-jk][1,4]benzodiazepinone-2(1H)one and -thione derivatives. Proc. Natl. Acad. Sci. USA 88, 1451–1455. 19. Palaniappan, C., Fay, P.J. & Bambara, R.A. (1995). Nevirapine alters the cleavage specificity of ribonuclease H of human immunodeficiency virus 1 reverse transcriptase. J. Biol. Chem. 270, 4861–4869.
860
Structure 1996, Vol 4 No 7
20. Boyer, P.L., Ferris, A.L. & Hughes, S.H. (1992). Cassette mutagenesis of the reverse transcriptase of human immunodeficiency virus type 1. J. Virol. 66, 1031–1039. 21. Boyer, P.L., Ferris, A.L. & Hughes, S.H. (1992). Mutational analysis of the fingers domain of human immunodeficiency virus type-1 reverse transcriptase. J. Virol. 66, 7533–7537. 22. Boyer, P.L., et al., & Hughes, S.H. (1994). Mutational analysis of the fingers and palm subdomains of human immunodeficiency virus type-1 (HIV-1) reverse transcriptase. J. Mol. Biol. 243, 472–483. 23. Kleim, J.-P., et al., & Riess, G. (1996). Selective pressure of a quinoxaline class non-nucleoside inhibitor of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) on HIV-1 replication results in the emergence of nucleoside RT inhibitor specific (RT L74→V/I, V75→L/I) HIV-1 mutants. Proc. Natl. Acad. Sci. USA 93, 34–38. 24. Clark, A.D., Jr., Jacobo-Molina, A., Clark, P., Hughes, S.H. & Arnold, E. (1995). Crystallization of human immunodeficiency virus type-1 reverse transcriptase with and without nucleic acid substrates, inhibitors and an antibody Fab fragment. In Methods in Enzymology. (Campbell, J.L., ed.), pp. 171–185, Academic Press, New York. 25. Teng, T.Y. (1990). Mounting of crystals for macromolecular crystallography in a free-standing thin film. J. Appl. Cryst. 24, 387–391. 26. Rodgers, D.W. (1994). Cryocrystallography. Structure 2, 1135–1140. 27. Otwinowski, Z. (1993). Oscillation Data Reduction Program. In Proceeding of the CCP4 Study Weekend: Data Collection and Processing. (Sawyer, L., Issacs N., Bailey, S., eds), pp 55–62, SERC Daresbury Laboratory, UK. 28. Miner, W. (1993). XDISPLAYF Program, Purdue University. 29. Collaborative Computational Project, No. 4. (1994). The CCP4 Suite: Programs for Protein Crystallography. Acta Cryst. D 50, 760–763. 30. Brünger, A.T. (1993). X-PLOR Manual Version 3.1: A system for Xray crystallography and NMR. Yale University Press, New Haven and London. 31. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Cryst. A 47, 110–119. 32. Carson, M. (1987). Ribbon models of macromolecules. J. Mol. Graphics 5, 103–106.
853
Structure of unliganded HIV-1 reverse transcriptase at 2.7 Å resolution: implications of conformational changes for polymerization and inhibition mechanisms Y Hsiou1, J Ding1, K Das1, AD Clark, Jr1, SH Hughes2 and E Arnold1* Background: HIV-1 reverse transcriptase (RT) is a major target for anti-HIV drugs. A considerable amount of information about the structure of RT is available, both unliganded and in complex with template-primer or nonnucleoside RT inhibitors (NNRTIs). But significant conformational differences in the p66 polymerase domain among the unliganded structures have complicated the interpretation of these data, leading to different proposals for the mechanisms of polymerization and inhibition. Results: We report the structure of an unliganded RT at 2.7 Å resolution, crystallized in space group C2 with a crystal packing similar to that of the RT–NNRTI complexes. The p66 thumb subdomain is folded into the DNAbinding cleft. Comparison of the unliganded RT structures with the DNA-bound RT and the NNRTI-bound RT structures reveals that the p66 thumb subdomain can exhibit two different upright conformations. In the DNA-bound RT, the p66 thumb subdomain adopts an upright position that can be described as resulting from a rigid-body rotation of the p66 thumb along the ‘thumb’s knuckle’ located near residues Trp239 (in strand b14) and Val317 (in b15) compared with the thumb position in the unliganded RT structure. NNRTI binding induces an additional hinge movement of the p66 thumb near the thumb’s knuckle, causing the p66 thumb to adopt a configuration that is even more extended than in the DNA-bound RT structure.
Addresses: 1Center for Advanced Biotechnology and Medicine (CABM) and Rutgers University Chemistry Department, 679 Hoes Lane, Piscataway, NJ 08854-5638, USA and 2ABLBasic Research Program, NCI-Frederick Cancer Research and Development Center, PO Box B, Frederick, MD 21702-1201, USA. *Corresponding author. E-mail: [email protected] Key words: AIDS, drug design, mechanism of polymerization, non-nucleoside inhibition mechanism, polymerase structure Received: 4 April 1996 Revisions requested: 2 May 1996 Revisions received: 28 May 1996 Accepted: 30 May 1996 Structure 15 July 1996, 4:853–860 © Current Biology Ltd ISSN 0969-2126
Conclusions: The p66 thumb subdomain is extremely flexible. NNRTI binding induces both short-range and long-range structural distortions in several domains of RT, which are expected to alter the position and conformation of the template-primer. These changes may account for the inhibition of polymerization and the alteration of the cleavage specificity of RNase H by NNRTI binding.
Introduction The reverse transcriptase (RT) of human immunodeficiency virus type 1 (HIV-1) is an attractive target for antiviral drugs (see recent reviews in [1–4]). HIV-1 RT is a heterodimer that consists of a 66 kDa subunit (p66) and a 51 kDa subunit (p51) (Fig. 1). The p66 subunit is composed of a polymerase domain and an RNase H domain. The p51 subunit contains only the polymerase domain. Both the p66 and p51 polymerase domains contain four subdomains (fingers, palm, thumb and connection) which are arranged differently in the two subunits. Only the p66 polymerase domain has a DNA-binding cleft, a functional polymerase active site, and a site for binding non-nucleoside inhibitors (NNRTIs). Structures of HIV-1 RT in complexes with various nonnucleoside RT inhibitors (NNRTIs) [5–10], and in complex with a 19-mer/18-mer double-stranded DNA and an antibody Fab fragment [11] have been reported. Structures of unliganded HIV-1 RT have been solved in three
different crystal forms ([12,13], EA and coworkers, unpublished data). It is reasonable to expect that comparing the structures of unliganded HIV-1 RT with those of HIV-1 RT complexed with NNRTIs or nucleic acid should yield valuable insights into the mechanisms of polymerization and inhibition. This information should be useful in the development of new or improved inhibitors of HIV-1 RT. However, in the unliganded HIV-1 RT structures already reported, there are significant conformational differences in the p66 polymerase domain, particularly in the conformation of the b6–b10–b9 sheet and the position of the p66 thumb subdomain ([12,13], EA and coworkers, unpublished data). This has led to controversy regarding the role of the p66 thumb in the polymerization reaction, and different interpretations of the inhibition mechanism of HIV-1 RT by NNRTIs. Here, we report the structure of an unliganded HIV-1 RT crystallized in space group C2, a crystal form closely related to, but not isomorphous with, that of a number of
854
Structure 1996, Vol 4 No 7
Figure 1 Stereo Ca trace of the unliganded HIV-1 RT, drawn using the program RIBBONS [32]. The heterodimer is colored by subdomains: fingers, blue; palm, red; thumb, green; connection, yellow; and RNase H, orange. Every 50th residue is labeled: p66, 1–556 and p51, 1001–1427.
HIV-1 RT–NNRTI complexes reported previously. This unliganded HIV-1 RT structure has a molecular packing in the crystal lattice similar to that of the NNRTI-bound HIV-1 RT structures crystallized in the C2 crystal form. Comparison of these HIV-1 RT structures should therefore help to distinguish conformational changes induced by NNRTI binding from changes caused by a different crystal packing arrangement.
Results and discussion Overview of HIV-1 RT structure and flexibility
Although the unliganded HIV-1 RT structure reported here was determined in a crystal form similar to that present in crystals of the HIV-1 RT–NNRTI complexes, the cell parameters differ considerably. The unit cell of the unliganded RT crystals is approximately 4% (10 Å) longer along the a axis and 10% (10 Å) shorter along the c axis relative to that of the NNRTI-bound RT. This suggests that the repositioning of the p66 thumb (see later discussion) causes a considerable change in crystal packing. In the structures of RT–NNRTI complexes with the C2 crystal form, two symmetry-related molecules juxtapose in such a way that the p66 fingers subdomain of one molecule contacts both the RNase H domain and the tip of the p66 thumb subdomain of the other. In the crystal lattice of the unliganded RT reported here, the RT molecules are arranged in a manner similar to that in the RT–NNRTI crystals. Even though the p66 thumb subdomain is folded into the DNA-binding cleft, the same regions of the p66 thumb subdomain and the fingers subdomain of the symmetry-related molecule interact with each other, although with different specific contacts. Compared with the relative positions of symmetry-related molecules in the RT–NNRTI structures, the symmetry-related molecules in this unliganded RT structure have a 20 Å translation in the xz plane relative to each other. This translation permits the RT molecules in this structure to pack more closely along the c axis, and to expand slightly along the a axis,
which would account for the changes in the observed cell parameters. The structure of unliganded HIV-1 RT has been determined in several crystal forms. Rodgers et al. [12] reported the structure of an unliganded HIV-1 RT in the space group C2 with 4 molecules per asymmetric unit. We have solved the structure of an unliganded HIV-1 RT complexed with an antibody Fab fragment, which was crystallized in space group P3212 with unit cell dimensions isomorphous with those of the RT–DNA–Fab complex (EA and coworkers, unpublished data). In these unliganded HIV-1 RT structures, the p66 thumb subdomain is folded down into the DNA-binding cleft. In the structure of unliganded RT reported here, the p66 thumb subdomain is also folded down into the DNA-binding cleft. Therefore, with dramatically different crystal packing arrangements, the RT molecules in these crystal lattices exhibits strikingly similar folding. The rms deviations between the structures of RT reported here and by Rodgers et al. are 1.17 Å, 1.13 Å, and 0.97 Å based on the superposition of Ca atoms for the whole RT molecule (925 Ca atoms), the p66 subunit (475 Ca atoms), and the p51 subunit (393 Ca atoms), respectively. The largest difference between subdomains is that the p66 thumb subdomain in the unliganded RT structure reported here is rotated further into the DNA-binding cleft by 7° relative to the structure reported by Rodgers et al. This displacement could be the result of a relatively closer crystal packing arrangement in the current crystal form. Esnouf et al. [13] also described the structure of an unliganded HIV1 RT that was crystallized in space group P212121. In this structure, the p66 thumb subdomain is in an upright conformation similar to that seen in the structures of RT–NNRTI complexes and the RT–DNA complex. Moreover, the b6–b10–b9 sheet of the p66 palm subdomain moves toward the p66 thumb subdomain in a manner different from that seen in the other unliganded
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
RT structures ([12], EA and coworkers, unpublished data). In the structure of unliganded HIV-1 RT reported here, the position and conformation of the b6–b10–b9 sheet agree very well with that found in the unliganded RT structures reported earlier ([12], EA and coworkers, unpublished data). The different conformation of the p66 thumb subdomain and the repositioning of the b6–b10–b9 sheet in the unliganded RT reported by Esnouf et al. [13] may be a consequence of the method they used to prepare the unliganded HIV-1 RT crystals. In that study, crystals were obtained by soaking out a weakly-bound NNRTI inhibitor (HEPT) from crystals of the HIV-1 RT–HEPT complex [13]. Examination of the molecular packing in the structure of unliganded HIV-1 RT reported by Esnouf et al. indicates that the p66 thumb subdomain is interlocked between the p66 thumb and RNase H subdomains of the symmetryrelated RT molecule. It appears that the p66 thumb subdomain in this structure of unliganded RT is restricted by crystal packing in such a way that it cannot fold down into the DNA-binding cleft despite the NNRTI HEPT having been soaked out. At any rate, the differences in the position of the p66 thumb subdomain in different HIV-1 RT structures provides ample evidence for high flexibility of the p66 thumb subdomain; a flexibility that may play a vital role in the binding and translocation of the nucleic acid substrate.
855
The structures of RT complexed with the NNRTIs show that the NNIBP is formed primarily by amino-acid residues of the b5–b6 loop, b6, the b9–b10 hairpin, the b12–b13 hairpin, and b15 of p66, and the b7–b8 connecting loop of p51 [8]. In the structure of the unliganded RT reported here and that reported by Rodgers et al., and in the structure of the RT–DNA–Fab complex, the region corresponding to the NNIBP in the RT–NNRTI structures is occupied by the side chains of Tyr181, Tyr188, Phe227, and Trp229. When an NNRTI binds, the NNIBP is created by the reorientation of the side chains of Tyr181 and Tyr188, and by the repositioning of the b12–b13–b14 sheet that contains the ‘primer grip’. In the unliganded RT structure reported by Esnouf et al., a small cavity is present at the NNIBP region and the b6–b10–b9 and b12–b13–b14 sheets (using our nomenclature for secondary structural elements [11]) are shifted, along with the side chains of residues Tyr181, Tyr183, and Tyr188, toward this small residual cavity [13]. This reorientation of the side chain of residue Tyr183 has not been seen in other unliganded RTs ([12], EA and coworkers, unpublished data). Because the p66 thumb subdomain in the Esnouf et al. structure is apparently constrained in a conformation similar to that in the RT–NNRTI complex, the displacement of the side chain of Tyr183 and the altered positions of the b6–b10–b9 and b12–b13–b14 sheets may be caused by an incomplete collapse of the empty drugbinding pocket relative to the other HIV-1 RT structures not containing NNRTIs.
Polymerase active site and NNRTI-binding pocket (NNIBP)
The polymerase active site is located at the b6–b10–b9 sheet of the p66 palm subdomain. The position of the b9–b10 hairpin turn in the unliganded form is nearly identical to that in the NNRTI-bound HIV-1 RT, suggesting that the position of the b6–b10–b9 sheet is not significantly affected by the binding of an NNRTI. The Tyr–Met–Asp–Asp motif (residues 183–186) in both the p66 and p51 subunits is in an unusual bII′ turn conformation that has been observed in several other HIV-1 RT structures [7–10,12,13]. Clear density peaks were found to have good coordination geometry with Od1 of Asp185 (2.8 Å) and Od1 of Asp186 (2.8 Å) at the p66 polymerase active site, and with the Od1 of Asp443 (3.1 Å), the Oε2 of Glu478 (2.4 Å), and the Od1 and Od2 of Asp498 (2.5 Å and 2.4 Å, respectively), at the RNase H active site. Due to the current resolution limit, we cannot distinguish whether these peaks correspond to water or magnesium ions. Structures of the Moloney leukemia virus (MMLV) RT fragment [14] and the fingers and palm subdomains of the p66 subunit in this unliganded HIV-1 RT (rms deviation of 1.4 Å based on 57 Ca atoms in the fingers and 56 Ca atoms of the palm subdomain) were superposed. The results indicate that the position of the electron density peak at the polymerase active site in this unliganded HIV-1 RT is quite close to that of the Mn2+ ion seen in the MMLV RT structure.
Calculation of the solvent-accessible surface area based on a 1.6 Å probe for the current unliganded HIV-1 RT structure reveals two surface depressions at the base of the NNRTI-binding pocket that are plausible candidates for the entrance to the pocket (Fig. 2). These two surface depressions were also observed in the RT–DNA–Fab complex. One surface depression is located at the proposed entrance near Lys101, Lys103, Val179 (p66), and Glu138 (p51) (site 1) [9]. The other one is near the base of the thumb subdomain between two adjacent polypeptide segments: the b5–b6 connecting loop (containing Lys101 and Lys103) and the b13–b14 hairpin (containing Pro236 and Lys238) (site 2). Both sites are exposed to solvent. It is possible that an NNRTI can approach the binding pocket via either site and displace the Tyr181 and Tyr188 side chains. Once the NNRTI is bound, however, only the putative entrance near site 1 remains accessible [9]. Site 2 disappears in the NNRTI-bound HIV-1 RT structure as a result of a hinge-like movement of the b12–b13–b14 sheet relative to the b6–b10–b9 sheet. The exact inhibitor entrance and the mechanism(s) of entry are still unclear. Conformational distortions induced by NNRTI binding
As noted by Rodgers et al., the displacement of the p66 thumb subdomain is the most dramatic change between the structures of the unliganded RT, the RT–DNA–Fab
856
Structure 1996, Vol 4 No 7
Figure 2 Stereoview of the solvent-accessible surface of the p66 palm subdomain in the unliganded RT structure showing the two surface depressions (indicated by arrows) near the NNIBP that are potential entrances to the inhibitor-binding pocket. One (site 1) is located at the proposed entrance near residues Lys101, Lys103, Val179 (p66) and Glu138 (p51). The other (site 2) is between Lys101/Lys103 and Pro236/Lys238, that is, at the base of the p66 thumb subdomain. The latter one disappears after binding of an NNRTI. Water molecules (cyan) are seen in both surface depressions.
complex, and the RT–NNRTI complexes. The structures of RT in this unliganded form and in the DNA-bound form superimposed quite well with an rms deviation of 1.3 Å based on all the Ca atoms, excluding the p66 thumb subdomain. Superposition of the structure of the RT–DNA–Fab complex and the unliganded form of HIV-1 RT, based on the Ca atoms of the b6–b10–b9 sheet, reveals that the p66 thumb subdomain, in the RT–DNA–Fab complex, is rotated 30° away from the p66 fingers and the polymerase active site (Fig. 3a), similar to that noted by Rodgers et al. This rotation functions like a thumb’s knuckle and the axis of the rotation is located at the base of thumb (around residue Trp239 and Val317). The other subdomains of the polymerase domain and the RNase H domain in p66 remain in similar positions in the unliganded RT and the RT–DNA–Fab complex.
the p66 connection and RNase H subdomains move away relative to the fingers subdomain with 13° and 15° rotations, respectively, compared with the corresponding positions in the unliganded RT (Fig. 3b). The position of the p66 fingers subdomain in the HIV-1 RT–NNRTI complexes is relatively unchanged (less than 4° rotation when compared with the unliganded RT). The subdomains in the p51 subunit of the NNRTI-bound RT also move (10°, 12°, 15°, and 17° for fingers, palm, thumb, and connection subdomains, respectively). As a result, the intersubunit interactions at the heterodimer interfaces are maintained in both the unliganded RT and the RT–NNRTI complex, suggesting that the heterodimer interface is relatively stable. It is interesting to note that the binding of a relatively small inhibitor molecule is accompanied by such dramatic rearrangements of the subdomains.
The structures of the unliganded RT and the NNRTIbound RT agree quite nicely within individual subdomains (rms deviations are within 1.5 Å, based on the Ca atoms in each subdomain). However, superposition of the whole molecules does not show as good agreement (rms deviation of 2.1 Å). If the structures are superimposed based on the Ca atoms of the b6–b10–b9 sheet, binding an NNRTI causes a hinge-like movement between the b6–b10–b9 sheet and the b12–b13–b14 sheet. The hinging is near, but not coincident with the axis of the thumb’s knuckle motion (Fig. 3b). This hinge-like movement, which is similar to the internal swivel motion suggested by Jäger et al. [15], extends the position of the p66 thumb subdomain beyond the upright position seen in the RT–DNA complex. In the inhibitor-bound HIV-1 RT structures, the p66 thumb subdomain is rotated approximately 40° relative to the p66 fingers subdomain compared with its position in the unliganded RT. This rotation is significantly larger than the corresponding 30° rotation in the RT–DNA–Fab complex. Comparing the unliganded RT and the DNA-bound RT shows that the hinge movement is coupled with the binding of NNRTIs. Associated with the hinge movement of the p66 thumb,
Mechanistic implications for the inhibition by NNRTIs
Comparison of the available structures of HIV-1 RT indicates that the binding of an NNRTI causes both shortrange and long-range distortions of the HIV-1 RT structure. The short-range distortions include conformational changes of the amino acid residues and/or structural elements that form the NNIBP, such as reorientation of the side chains of Tyr181 and Tyr188, and displacement of the b12–b13–b14 sheet. These distortions affect the precise geometry of the polymerase active site. The long-range distortions involve a hinging movement of the p66 thumb subdomain that results in the displacement of the p66 connection subdomain, the RNase H domain, and the p51 subunit relative to the polymerase active site. The observation that the chemical step of DNA polymerization is slowed by NNRTI binding [16,17] may be explained by the short-range structural distortions induced by NNRTI binding, including the conformational changes of the primer grip. Both short-range and long-range structural distortions induced by NNRTI binding alter the interactions between the nucleic acid substrates and HIV-1 RT, and affect the precise positioning of the template-primer relative to the polymerase and the RNase H active sites. NNRTI-induced changes in RT
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
857
Figure 3 Superposition of (a) unliganded RT and RT–DNA–Fab complex and (b) unliganded RT and RT–a-APA (a-anilinophenylacetamide) complex based on 89 Ca atoms in the p66 palm subdomain, including the b6–b10–b9 region. The unliganded RT is shown in red, RT–a-APA in blue, and RT–DNA–Fab in green. A comparison of the two superpositions reveals that NNRTI binding appears to be accompanied by a long-range distortion that is coupled with a hinge motion (indicated by curved arrows) between the b6–b10–b9 and b12–b13–b14 sheets at the p66 palm subdomain (within the circle). The different positions of the thumb in different HIV-1 RT structures supports the idea that this subdomain could play a role during polymerization.
interactions with nucleic acid substrates may partly account for the observations that NNRTI binding can alter the affinity of HIV-1 RT for the template-primer [16,18] as well as the specificity of RNase H cleavage [19]. HIV-1 RT mutations near the polymerase active site that had a greater effect on RNase H activity than on polymerase activity [20–22] have been described, supporting the idea that communication between distant sites on HIV-1 RT (e.g., the polymerase and RNase H active sites, the NNIBP, and other sites that interact with template-primer) can occur via the template-primer [8,11,22,23]. It is difficult to be certain, however, that repositioning of the template-primer would be solely responsible for the effect of NNRTI binding on RNase H cleavage specificity, as binding does cause other long-range movements of subdomains of both p51 and p66 relative to each other. The long-range distortions caused by NNRTIs have the potential to affect other RT activities such as strand transfer, strand displacement, and recognition of the tRNA used for priming reverse transcription.
Biological implications The cumulative biological, genetic, and structural information about HIV-1 reverse transcriptase (RT) has considerably enhanced our understanding of the mechanisms of polymerization and inhibition. However, the structures of unliganded HIV-1 RT previously reported differ
significantly in the conformation of the polymerase domain of the p66 subunit. This has led to controversy concerning the involvement of the p66 thumb subdomain in the mechanism of DNA polymerization and the mechanism of inhibition by non-nucleoside inhibitors (NNRTIs). We report here the crystal structure of an unliganded HIV-1 RT crystallized in space group C2, with molecular packing similar to that of the NNRTI-bound HIV-1 RT complexes. The overall folding of this enzyme is similar to that determined by Rodgers et al. [12]. In these two structures, the p66 thumb subdomain is folded into the DNA-binding cleft. The differences observed among the unliganded HIV-1 RTs in different crystal forms demonstrates the considerable flexibility of the p66 thumb subdomain, which may play an important role in the recognition of nucleic acid substrates and the translocation of the enzyme along the nucleic acid after incorporation of each new nucleotide. It is likely that the flexibility of the p66 thumb allows it to facilitate the repositioning (and ‘reloading’) of RT during polymerization. A comparison of the structures of the NNRTI-bound RT complexes with those of unliganded RT and the HIV-1 RT–DNA–Fab complex [11] shows that the binding of
858
Structure 1996, Vol 4 No 7
an NNRTI causes both short-range and long-range structural distortions. A hinge movement at the base of the p66 thumb subdomain (near the ‘thumb’s knuckle’) appears to constrain the thumb in a wide-open configuration in the HIV-1 RT–NNRTI complexes, which is extended beyond the upright configuration seen in the HIV-1 RT–DNA–Fab structure. Biochemical and structural data suggest that the conformational changes accompanying NNRTI binding alter the positioning of the template-primer relative to the polymerase and RNase H active sites. This could potentially account both for the inhibition of RT activity (short-range effects) and the alteration of RNase H cleavage specificity (longrange effects) by NNRTIs.
Materials and methods Crystallization and data collection The unliganded HIV-1 RT was crystallized ([24] and references therein) using modified conditions reported by Kohlstaedt et al. for crystallizing the HIV-1 RT/nevirapine complex [5]. The initial crystals were grown in the absence of NNRTIs by streak-seeding with crushed RT/a-APA crystals. After crystals of unliganded HIV-1 RT were obtained, they were used for further seeding. Crystals grew to an average size of 0.1 mm × 0.2 mm × 0.4 mm after two weeks and had the symmetry of space group C2, which is the same as that of the HIV-1 RT–NNRTI complexes. The cooled diffraction data were collected at the F1 beam line (l = 0.91 Å) of the Cornell High Energy Synchrotron Source (CHESS) from seven crystals cooled to –10°C. Data were recorded on Fuji imaging plates with an overall Rmerge of 11.8% and a completeness of 76.6% to 3.3 Å resolution. Another diffraction dataset was collected
at the CHESS F1 beam line from a single crystal at –165°C using the flash-cooled technique [25,26]. This dataset, which was used for structural determination and refinement, has an overall Rmerge of 5.5%, a completeness of 99.5% to 2.7 Å resolution (no s cutoff was used, and the average measurement redundancy was 3.5). The HKL package [27,28] was initially used to process, reduce, scale, and postrefine the data. Final data were scaled using SCALA (CCP4) [29]. A TAMM derivative diffraction dataset was collected at the CHESS A1 beam line and was scaled to the native dataset using FHSCAL of CCP4. The statistics of all diffraction data are summarized in Table 1.
Structure determination and refinement The structure of the unliganded HIV-1 RT was determined with the molecular replacement (MR) technique as implemented in X-PLOR [30], using the diffraction data from the frozen crystal and the coordinates of the unliganded RT provided by Rodgers et al. as the starting model [12]. The rotation function search and translation function search yielded one distinct solution. The correctness of the MR solution was confirmed by the positions of mercury atoms in a difference Fourier map computed between the TAMM derivative dataset and the native dataset phased with the initial MR model. A subset of 6% of the frozen dataset was separated for calculating the free R-factor to monitor the progress of refinement. No subset of data was generated for the free R-factor calculation for the cooled diffraction dataset due to the relative incompleteness of the dataset. Iterative refinement and model building were performed using the programs X-PLOR [30] and O [31]. Omit maps were calculated for both the frozen and the cooled datasets and then averaged between the frozen and cooled datasets to improve the map quality and reduce model bias (KD et al. unpublished data). The final refinement converged to an R-factor of 24.9% and a free R-factor of 33.6% for the frozen crystal dataset, and an Rfactor of 23.0% for the cooled dataset. The fit of the atomic model into the final 2mFobs–Fcalc electron density map in the region of the p66 connection subdomain is shown in Figure 4. A total of 267 water
Table 1 Summary of crystallographic data and refinement statistics. Datasets
Native (cooled)
Native (frozen)
TAMM (frozen)
7 79 –10 240.3,71.5,95.5 104.4 3.3 24 727 18 345 76.6 11.8
1 112 –165 235.5,70.3,93.3 106.1 2.7 142 770 41 049 99.5 5.5
1 92 –165 234.0,69.7,93.0 106.2 3.7 46 901 15 521 99.4 11.2 21.8
Diffraction data statistics No. of crystals No. of images Temperature (°C) a,b,c (Å) b (°) Resolution (Å) Observed reflections (%) Unique reflections (%) Completeness (%) Rmerge* (%) Rderiv† (%) Refinement statistics Resolution range (Å) No. of reflections‡ R factor# (%) Free R factor (%) No. of protein atoms No. of water molecules Average B factor (Å2) Rms deviations Bond (Å) Angles (°)
8.0–2.7 38 310 24.9% 33.6% 7692 267 44 0.01 2.0
*Rmerge =S|Iobs–|/S.†Rderiv =S|IPH–IP|/SIP. ‡No. of reflections used in refinement have F≥2s(F). #R=S||Fobs|–|Fcalc||/S|Fobs|.
Research Article Structure of unliganded HIV-1 reverse transcriptase Hsiou et al.
859
Figure 4 Stereoview of a portion of a (2mFobs–Fcalc) difference Fourier map at the p66 connection subdomain, at 2.7 Å resolution. The phases were computed from the current atomic model and the map is contoured at 1.4s. The side chain and the carboxyl groups are well defined in the electron density map.
molecules were found that have good stereochemistry. Some residues are apparently disordered and have very weak electron density. These residues have been modeled as alanines, including residues 22, 28, 64, 70–71, 134–136, 139, 169, 177, 194, 203, 218, 224, 242, 278, 281, 291, 297, 305, 311–312, 334, 336, 356–359, 404, 407, 413–414, 424, 448, 451, 512, 514, 527, 549–551, and 556 in the p66 subunit, and residues 36, 67–69, 88–89, 173, 197, 238, 281–282, 295, 297, 305–306, 308, 356–358, 361–362, 407, 424–425, and 427 in the p51 subunit. A part of the connecting loop between aF and b13 in the p51 subunit (residues 219-231) has weak electron density and thus has not been modeled. Final refinement statistics are listed in Table 1.
Accession numbers Coordinates for the final model of the native structure refined using the frozen dataset have been deposited with the Brookhaven Protein Data Bank (entry code1DLO).
Acknowledgements We thank David Rodgers, Steve Gamblin, and Steve Harrison for providing the coordinates of their unliganded HIV-1 RT structure prior to publication. We thank the other members of the Arnold and Hughes laboratory, including Pat Clark, Peter Frank, Andrew Holmes, Wade Huber, Karen Lentz, and Dave Miller, Dawn Resnick, Chris Tantillo and Wanyi Zhang for their assistance and helpful discussion; the members of CHESS staff for their help with X-ray data collection; the Frederick Biomedical Supercomputing Center for providing Cray-YMP computing time; the Keck Structural Biology Computing Center at CABM; and Andrew Holmes for critical reading of the manuscript. The research in EA’s laboratory has been supported by NIH grants (AI 27690 and AI 36144), Janssen Research Foundation, and Hoechst General Pharma Research, a grant from NCI-ABL and the Keck Foundation, and SHH’s laboratory is sponsored in part by the National Cancer Institute, DHHS, under contract with ABL and by NIGMS.
References 1. Larder, B.A. (1993). Inhibitors of HIV reverse transcriptase as antiviral agents and drug resistance. In Reverse Transcriptase. (Skalka, A.M. & Goff, S.P., eds), pp. 205–222, Cold Spring Harbor Laboratory Press, New York. 2. Tantillo, C., et al., & Arnold, E. (1994). Locations of anti-AIDS drug binding sites and resistance mutations in the three-dimensional structure of HIV-1 reverse transcriptase: implications for mechanisms of drug inhibition and resistance. J. Mol. Biol. 243, 369–387. 3. De Clercq, E. (1995). Toward improved anti-HIV chemotherapy: Therapeutic strategies for intervention with HIV infections. J. Med. Chem. 38, 2491–2517. 4. Arnold, E., Das, K., Ding, J., Yadav, P.N.S., Hsiou, Y., Boyer, P.L. & Hughes, S.H. (1996). Targeting HIV reverse transcriptase for antiAIDS drug design. Drug Design and Discovery. 13, 29–47.
5. Kohlstaedt, L.A., Wang, J., Friedman, J.M., Rice, P.A. & Steitz, T.A. (1992). Crystal structure at 3.5 Å resolution of HIV-1 reverse transcriptase complexed with an inhibitor. Science 256, 1783–1790. 6. Smerdon, S.J., et al., & Steitz, T.A. (1994). Structure of the binding site for non-nucleoside inhibitors of the reverse transcriptase of human immunodeficiency virus type 1. Proc. Natl. Acad. Sci. USA 91, 3911–3915. 7. Ren, J., et al., & Stammers, D. (1995). High resolution structures of HIV-1 RT from four RT-inhibitor complexes. Nat. Struct. Biol. 2, 293–302. 8. Ding, J., et al., & Arnold, E. (1995). Structure of HIV-1 reverse transcriptase in a complex with the non-nucleoside inhibitor a-APA R 95845 at 2.8 Å resolution. Structure 3, 365–379. 9. Ding, J., et al., & Arnold, E. (1995). Structure of HIV-1 RT/TIBO R 86183 reveals similarity in the binding of diverse non-nucleoside inhibitors. Nat. Struct. Biol. 2, 407–415. 10. Ren, J., et al., & Stuart, D. (1995). The structure of HIV-1 reverse transcriptase complexed with 9-chloro-TIBO: lessons for inhibitor design. Structure 3, 915–926. 11. Jacobo-Molina, A., et al., & Arnold, E. (1993). Crystal structure of human immunodeficiency virus type 1 reverse transcriptase complexed with double-stranded DNA at 3.0 Å resolution shows bent DNA. Proc. Natl. Acad. Sci. USA 90, 6320–6324. 12. Rodgers, D.W., et al., & Harrison, S.C. (1995). The structure of unliganded reverse transcriptase from the human immunodeficiency virus type 1. Proc. Natl. Acad. Sci. USA 92, 1222–1226. 13. Esnouf, R., Ren, J., Ross, R., Jones, Y., Stammers, D. & Stuart, D. (1995). Mechanism of inhibition of HIV-1 reverse transcriptase by nonnucleoside inhibitors. Nat. Struct. Biol. 2, 303–308. 14. Georgiadis, M.M., Jessen, S.M., Ogata, C.M., Telesnitsky, A., Goff, S.P. & Hendrickson, W.A. (1995). Mechanistic implications from the structure of a catalytic fragment of Moloney murine leukemia virus reverse transcriptase. Structure 3, 879–892. 15. Jäger, J., Smerdon, S.J., Wang, J., Boisvert, D.C. & Steitz, T.A. (1994). Comparison of three different crystal forms shows HIV-1 reverse transcriptase displays an internal swivel motion. Structure 2, 869–876. 16. Spence, R.A., Kati, W.M., Anderson, K.S. & Johnson, K.A. (1995). Mechanism of inhibition of HIV-1 reverse transcriptase by nonnucleoside inhibitors. Science 267, 988–993. 17. Rittinger, K., Divita, G. & Goody, R. (1995). Human immunodeficiency virus reverse transcriptase substrate-induced conformational changes and the mechanism of inhibition by non-nucleoside inhibitors. Proc. Natl. Acad. Sci. USA 92, 8046–8049. 18. Debyser, Z., et al., & De Clercq, E. (1991). An antiviral target on reverse transcriptase of human immunodeficiency virus type 1 revealed by tetrahydroimidazo-[4,5,1-jk][1,4]benzodiazepinone-2(1H)one and -thione derivatives. Proc. Natl. Acad. Sci. USA 88, 1451–1455. 19. Palaniappan, C., Fay, P.J. & Bambara, R.A. (1995). Nevirapine alters the cleavage specificity of ribonuclease H of human immunodeficiency virus 1 reverse transcriptase. J. Biol. Chem. 270, 4861–4869.
860
Structure 1996, Vol 4 No 7
20. Boyer, P.L., Ferris, A.L. & Hughes, S.H. (1992). Cassette mutagenesis of the reverse transcriptase of human immunodeficiency virus type 1. J. Virol. 66, 1031–1039. 21. Boyer, P.L., Ferris, A.L. & Hughes, S.H. (1992). Mutational analysis of the fingers domain of human immunodeficiency virus type-1 reverse transcriptase. J. Virol. 66, 7533–7537. 22. Boyer, P.L., et al., & Hughes, S.H. (1994). Mutational analysis of the fingers and palm subdomains of human immunodeficiency virus type-1 (HIV-1) reverse transcriptase. J. Mol. Biol. 243, 472–483. 23. Kleim, J.-P., et al., & Riess, G. (1996). Selective pressure of a quinoxaline class non-nucleoside inhibitor of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) on HIV-1 replication results in the emergence of nucleoside RT inhibitor specific (RT L74→V/I, V75→L/I) HIV-1 mutants. Proc. Natl. Acad. Sci. USA 93, 34–38. 24. Clark, A.D., Jr., Jacobo-Molina, A., Clark, P., Hughes, S.H. & Arnold, E. (1995). Crystallization of human immunodeficiency virus type-1 reverse transcriptase with and without nucleic acid substrates, inhibitors and an antibody Fab fragment. In Methods in Enzymology. (Campbell, J.L., ed.), pp. 171–185, Academic Press, New York. 25. Teng, T.Y. (1990). Mounting of crystals for macromolecular crystallography in a free-standing thin film. J. Appl. Cryst. 24, 387–391. 26. Rodgers, D.W. (1994). Cryocrystallography. Structure 2, 1135–1140. 27. Otwinowski, Z. (1993). Oscillation Data Reduction Program. In Proceeding of the CCP4 Study Weekend: Data Collection and Processing. (Sawyer, L., Issacs N., Bailey, S., eds), pp 55–62, SERC Daresbury Laboratory, UK. 28. Miner, W. (1993). XDISPLAYF Program, Purdue University. 29. Collaborative Computational Project, No. 4. (1994). The CCP4 Suite: Programs for Protein Crystallography. Acta Cryst. D 50, 760–763. 30. Brünger, A.T. (1993). X-PLOR Manual Version 3.1: A system for Xray crystallography and NMR. Yale University Press, New Haven and London. 31. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Cryst. A 47, 110–119. 32. Carson, M. (1987). Ribbon models of macromolecules. J. Mol. Graphics 5, 103–106.