The p66 Immature Precursor of HIV-1 Reverse Transcriptase

Abstract

In contrast to the wealth of structural data available for the mature p66/p51 heterodimeric human immunodeficiency virus type 1 reverse transcriptase (RT), the structure of the homodimeric p66 precursor remains unknown. In all X-ray structures of mature RT, free or complexed, the processing site in the p66 subunit, for generating the p51 subunit, is sequestered into a β-strand within the folded ribonuclease H (RNH) domain and is not readily accessible to proteolysis, rendering it difficult to propose a simple and straightforward mechanism of the maturation step. Here, we investigated, by solution NMR, the conformation of the RT p66 homodimer. Our data demonstrate that the RNH and Thumb domains in the p66 homodimer are folded and possess conformations very similar to those in mature RT. This finding suggests that maturation models which invoke a complete or predominantly unfolded RNH domain are unlikely. The present study lays the foundation for further in-depth mechanistic investigations at the atomic level.

INTRODUCTION

Reverse transcriptase (RT) plays a central role in the replication of all retroviruses and related retrotransposons.^1–5 In the human immunodeficiency virus type 1 (HIV-1) life cycle, RT is expressed as part of the Gag-Pol polypeptide, which is processed by retroviral protease into several proteins. The mature RT enzyme is a heterodimer, composed of two subunits, p66 and p51 (Fig. 1A).^6–9 The RT p51 subunit is generated by removal of the C-terminal RNase H (RNH) domain from p66.^10–15 In vivo processing of the Gag-Pol polyprotein is complex, and the detailed mechanism of RT maturation into the heterodimer is still unclear. Based on data obtained from model systems, cleavage at the p51-RNH processing site is assumed to occur in a p66 homodimer.^15–20 However, in all known RT X-ray structures, as well as those of the isolated RNH domain, the p51-RNH cleavage site is located within the folded RNH domain, sequestered into the center of a β-sheet, and thus seemingly inaccessible to the protease (Fig. 1B).^{16, 21–25} No significant motions were observed at the p51-RNH processing site in the isolated RNH domain,^{26, 27} which may have suggested partial accessibility of the site. In addition, the lack of structural information on the “immature” p66 RT precursor renders any mechanistic explanation(s) tentative. We, therefore, embarked on studies aimed at providing the foundation for structurally elucidating RT processing.

Figure 1.

Ribbon representation of the structures of (A) p66/p51 RT heterodimer and (B) the RNH domain, indicating the p51-RNH processing site (arrow), and (C) amino acid sequence of p66. In (A)-(C), the Thumb and RNH domains in the p66 subunit are shown in green and orange, respectively; the p51 subunit in cyan. Structures in (A) and (B) were drawn using PDB:1DLO.

To evaluate protein conformation in solution, Nuclear Magnetic Resonance (NMR) spectroscopy provides powerful approaches,^28–30 since it permits the investigation of conformational equilibria and protein dynamics at the amino acid residue level.^31–33 However, NMR studies of HIV RT are challenging, given the protein’s large molecular mass (117 kDa); to date, NMR of RT has been mostly limited to observing methyl groups of side chains, such as in methionine or isoleucine.^34–36 Although methyl resonances are valuable probes for obtaining general qualitative information about a protein’s conformation in solution,^{37, 38} they report only on a limited number of positions and, therefore, cannot inform on the secondary and tertiary structural details that are mirrored by a protein’s backbone chemical shifts.³⁹

Here, we present an investigation of the p66 homodimeric RT. The p66 dimer possesses enzymatic activity^40–42 and is widely considered to function as the RT precursor.^{15–20, 43} We took advantage of the extreme sensitivity of backbone amide resonance frequencies to assess conformational similarities between different protein constructs. In the ¹H-¹⁵N heteronuclear single-quantum coherence (HSQC) spectrum of the p66 homodimer, over 240 resonances were observed. Comparison of the p66 spectrum with the spectra of the isolated domains revealed that greater than 60% of the isolated Thumb domain and more than 40% of the isolated RNH domain resonances, respectively, are in very similar positions. In contrast, only18% of the Finger-Palm domain resonances match those of the p66 homodimer. This establishes that both the Thumb and RNH domains are stably folded in the immature p66 homodimeric RT and exhibit essentially the same structures as in the isolated domains. With these findings in mind, the question arises how HIV-1 protease gains access to the p51-RNH processing site in the p66 homodimer. Our data suggest that maturation models, which invoke a complete unfolded or predominantly unfolded RNH domain^{17, 21, 22} are unlikely, and suggest that p51-RNH processing may involve selection of a minor conformation or a protease-binding induced structure, which is cleaved during maturation.

MATERIALS AND METHODS

Sample preparation

The coding sequence for the RT p66 subunit was amplified from the p6HRT-PROT vector, kindly provided by Dr. Sluis-Cremer, using 5′-acc gca cat atg ccc att agc cct att gag act gta-3′ and 5′-gca gat ctc gag tag tat ttt cct gat tcc agc act gac-3′ as forward and backward primers, respectively. The amplified product was inserted into the pET21a(+) vector (Invitrogen, Carlsbad, CA), which encodes a six histidine tag at the C-terminus of the protein construct. After initial expression and purification trials, a codon-optimized C280S/C38V double cysteine version was created for increased protein expression in E. coli (DNA 2.0 gene synthesis, Menlo Park, CA). The coding sequence for the p51 subunit with an N-terminal Strep-tag was created by amplification of the appropriate region (coding for residues 1-440) of the p66 RT codon optimized sequence, using 5′-cc gca tcc atg gat tgg agt cac ccg cag ttc gag aaa cca atc agc cca atc gaa acg gtc cc -3′ and 5′-ccg cat ctc gag tta gaa cgt ttc cgc gcc aac aat cgg ttc ttt ctc c-3′ as forward and reverse primers, respectively, and the amplified product was then inserted into a pET28a+ vector (Invitrogen). Constructs coding for residues 1-216 (Finger-Palm domain), 237-318 (Thumb domain) and 427-556 (RNH domain) were amplified from the p66 sequence using 5′-gca gct cat atg cca atc agc cca atc gaa acg gtc cc-3′ and 5′-gca gat ctc gag ggt cgt cag acc cca acg cag cag atg c-3′ as forward and backward primers, respectively, for the finger-palm domain, 5′-cgt acg cat atg gat aaa tgg aca gta cag cct ata gtg ctg cc-3′ and 5′-cgt acg ctc gag ata cac tcc atg tac tgg ttc ttt tag aat ctc-3′ as forward and backward primers, respectively, for the thumb domain, and 5′-gca gat cat atg tat caa ctg gag aaa gaa ccg att gtt ggc-3′ and 5′-gca gat ctc gag cag gat ttt gcg aat acc tgc gct cac c-3′ as the forward and backward primers, respectively, for the RNH domain. Each amplified product was inserted into a pET21a(+) vector, which resulted in a hexa-histidine tag at the C-terminus of each protein construct. The sequences of all constructs were verified by DNA sequencing, using primers for the T7 promoter and terminator (Genewiz, South Plainfield, NJ).

All proteins were produced in E. coli BL21 (DE3) gold cells (Agilent Technologies, Santa Clara, CA), using 0.5 mM IPTG for induction over 16 hours at 27°C. Uniform ¹⁵N- labeling of the isolated Finger-Palm, Thumb, and RNH domains was achieved using modified minimal media containing ¹⁵NH₄Cl as the nitrogen source. Uniform ¹⁵N- and ¹³C- labeling of the Thumb and RNH domains, for assignment purposes, was carried out, using ¹⁵NH₄Cl and ¹³C₆-glucose as sole nitrogen and carbon sources, respectively. Uniform ¹⁵N- and ²H- labeling of the p66 and the p51 protein was achieved in modified minimum medium containing ²H₂O and ¹⁵NH₄Cl. Cells were harvested by centrifugation at 5887 g, re-suspended in lysis buffer containing 25 mM sodium phosphate (pH 7.5), 25 mM imidazole and 500 mM NaCl, and lysed using a microfluidizer. Cell debris was removed by centrifugation at 34530 g and the supernatant was applied to a 5 mL HisTrap (GE healthcare Life Sciences, Piscataway, NJ) column, equilibrated in lysis buffer. Thumb, RNH, and p66 proteins were eluted using a linear gradient of 0.025 – 0.5 M imidazole. Protein-containing fractions were further purified over a 5 mL HiTrap SP column (GE healthcare), equilibrated with 25 mM sodium phosphate (pH 6.5) and eluted using a linear gradient of 0 – 0.5 M NaCl. Final purification of the p66 protein involved separation on a HiLoad 26/60 Superdex 200 gel filtration column (GE Healthcare), in phosphate-buffered saline.

For the p51 protein, which contained and N-terminal Strep-tag, the supernatant after cell lysis and centrifugation was loaded onto a 5 mL HisTrap column. The flow through was collected and applied to a 5 mL StrepTrap column equilibrated with 25 mM sodium phosphate (pH 7.5), 6 mM KCl, 280 mM NaCl, and eluted in equilibration buffer that contained 3 mM d-Desthiobiotin (Sigma-Aldrich, St. Louis, MO). Protein containing fractions were further purified over a 5 mL HiTrap SP (GE healthcare) column, equilibrated with 25 mM sodium phosphate (pH 6.5), and eluted using a linear gradient of 0–0.5 M NaCl. Protein fractions were pooled and concentrated in an Amicon Ultra concentrator (EMD Millipore, Billerica, MA) to ~10 μM. 50% v/v glycerol was added and samples were stored at −80 °C.

NMR experiments

Samples were buffer-exchanged into NMR buffer (25 mM sodium phosphate, pH 6.8, 100 mM NaCl) in an Amicon Ultra concentrator (EMD Millipore). In the final exchange step, the sample was concentrated to 350 μL and supplemented with 10% (v/v) D₂O. Final protein concentrations for NMR experiments were ~200 μM (in monomer). All NMR spectra were recorded at 303 K on Bruker 600 and 900 MHz AVANCE spectrometers, equipped with 5-mm triple-resonance, z-axis gradient cryoprobes. For p66 and p51 proteins, a TROSY version the ¹H-¹⁵N HSQC NMR experiments was used.^{44, 45} For backbone assignments of the Thumb and RNH domains 3D CBCAONH, HNCACB, and ¹H-¹⁵N NOESY-HSQC ^46–49 spectra were recoded. All data were processed with NMRPipe and analyzed with CCPN.^{50, 51} Chemical shifts of amide resonances, as defined by Eq. (1), were deemed identical if the picked peaks in two spectra resided within ± 0.03 ppm.,

$$\mathrm{\Delta }{\delta }_{\mathit{obs}}^i=\sqrt{{\left({\delta }_H^i-{\delta }_H^{\mathit{free}}\right)}^2+{\left[\left({\delta }_N^i-{\delta }_N^{\mathit{free}}\right)\times \frac{{\gamma }_N}{{\gamma }_H}\right]}^2}$$ (1)

Here, δ indicates chemical shift in ppm units, and γ indicates gyromagnetic ratios of N and H resonances, respectively. Secondary structure elements were delineated using CSI and TALOS+.^{39, 52, 53} Chemical shifts for the Thumb domain spectrum were calculated using the coordinates from the crystal structure for residues 237 to 318 of p66 in RT p66/p51 (PDB ID: 1DLO), with the program Sparta.⁵⁴

Multi-angle Light Scattering

Size-exclusion chromatography/multi-angle light scattering (SEC-MALS) data were obtained at room temperature using an analytical Superdex 200 (10 × 300 mm, GE Healthcare) column with in-line multi-angle light scattering, refractive index (Wyatt Technology, Inc., Santa Barbara, CA) and UV (Agilent Technologies) detectors. 100 μL of a protein solution (10.9 – 59.9 μM for p66 and 19.3 – 58.6 μM for p51) was loaded onto the column, pre-equilibrated and eluted with NMR buffer at a flow rate of 0.5 mL/min. The molecular masses of the eluted protein species were determined using the ASTRA V.5.3.4 program (Wyatt Technologies). The dimer association constant for p66/p66 was extracted from peak intensities at the monomer and dimer molecular masses.

RESULTS

Amide backbone resonances of homodimeric p66

To obtain insight into the conformation of the immature p66 homodimeric RT in solution, ¹H-¹⁵N TROSY HSQC spectra of perdeuterated p66 were recorded (Fig. 2A). Excluding resonances that likely arise from amino acid side chains, at least 247 resonances were resolved. Although this number is fewer than half of all possible amide resonances (~600), it is still remarkable that such a large number is observed, given the large molecular mass of the p66 homodimer and concomitant line broadening in TROSY HSQC spectra.^{38, 55} Note that the p66 sample is predominantly a homodimer (>80%) at the concentration used (~200 μM), based on multi-angle light scattering (MALS) analysis (Fig. S1) and a previously reported dissociation constant, K_D, of 4 μM.^{12, 56, 57}

Figure 2.

¹H-¹⁵N HSQC spectra of (A) p66 (dark blue), (B) p51 (cyan), (C) Palm-Finger (purple), (D) Thumb (green), and (E) RNH (orange). In (B)-(E), individual isolated domain spectra are superimposed on the p66 spectrum (dark blue). The insets depict expanded regions of the spectra as indicated in panel (A).

For comparison, a ¹H-¹⁵N TROSY HSQC spectrum was also recorded for the p51 sample (Fig. 2B), which contained less than 60% dimer, based on MALS analysis (Fig. S1) and a reported K_D of 230 μM.^{12, 56, 57} Given that both our p66 and p51 NMR samples contain monomeric and dimeric protein, but give rise to only one set of resonances, the monomer and dimer forms are likely in fast exchange on the chemical shift scale. For p51, with a K_D ≫1 μM and an estimated on-rate >10⁶ s⁻¹/M,^{58, 59} this is not surprising. For p66, with a K_D ~4 μM, the observed chemical shift positions are most likely those of the major dimer species (>80%), irrespective of the exchange régime.

Despite the different proportion of dimerization in the p66 and p51 samples, a significant number of amide resonances exhibit the same resonance frequencies in the p51 and p66 NMR spectra; this is easily seen upon superposition of the spectra (Fig. 2B and Table I). Since NMR chemical shifts are extremely sensitive to local environments, those associated with residues that experience a different environment in two states, for example when located at domain interfaces in multimers, are expected not to be the same. Thus, the observed high similarity of the spectra, irrespective of the difference in dimer population, suggests that the observed signals represent residues that reside in identical local structures of p51 and p66. Interestingly, the fact that we observe a large number of resonances in the ¹H-¹⁵N TROSY HSQC spectrum for the p66 homodimer, despite its large molecular mass, suggests that maybe some of the smaller domains exhibit a certain degree of independent motion, faster than the overall rotational molecular diffusion.

Table I.

Percentage of amide resonances in the spectra of individual isolated domains that reside at identical frequencies in the p66 or p51 spectra

Isolated domains	p51	RNH	Thumb	Finger-Palm
Total number of counted resonances in each spectrum	219	119	83	179
Percentage of signals in the individual domain spectra that coincide with those in p66	62%	40%	59%	18%
Percentage of signals in the individual domain spectra that coincide with those in p51	-	18%	55%	18%

^#Resonances were assumed to be identical if they resonated within 0.03 ppm (see the method).

The RNH and Thumb domains are independently folded domains in the p66 homodimer

The RNH domain of HIV-RT can exist as a stably folded, individual domain with a structure similar to that in the intact protein.^{21, 25, 26} It is not known, however, whether other RT domains can also exist as stable folded sub-structures. We, therefore, prepared the Finger-Palm domain (residues1 to 216), the Thumb domain (residues 237 to 318) and the RNH domain (residues 427 to 556, see Fig. 1C). The ¹H-¹⁵N HSQC spectra for all three isolated domains exhibit well-dispersed resonances, indicative of stably folded structures (Fig. 2C–2E). To qualitatively assess whether the domain structures within p66 and in isolation are identical, the isolated domain spectra were individually superimposed onto the spectrum of p66 (Fig. 2C–2E). Amide resonances that were considered identical in frequencies (0.03 ppm in the combined chemical shift in Eq. (1)) were counted: of the three domains, the isolated Thumb and RNH domains exhibited large degrees of identity with the intact p66 protein (>40%), while for the Finger-Palm domain it was low (18%, Table I). When comparing the spectra of the individual domains with their counterparts in p51, a high degree of resonance frequency identity was observed for the Thumb domain (55%). We also evaluated whether, by chance, any spectral similarity exists between the structurally diverse RNH and Thumb domains; gratifyingly no overlap was noted, as can be appreciated from the superposition of their spectra onto the p66 spectrum (compare the orange and green resonances in Fig. 3A). Likewise, p51 and RNH domain resonances do not coincide (18% identity, compare the cyan and orange resonances), whereas significant overlap occurs for p51 and p66 (compare the cyan and dark blue resonances in Fig. 3B). Counting these coinciding cross-peaks in the p51 and RNH spectra provides a rough estimate for the percentage of amide resonances that coincidentally reside at the same ¹⁵N and ¹H frequencies, amounting to < 20%.

Figure 3.

Superposition of ¹H-¹⁵N HSQC correlation (A) of the Thumb domain (green) and of the RNH domain (orange) onto the p66 spectrum (dark blue) and (B) of the p51 domain (cyan) and the RNH domain (orange) onto the p66 spectrum (dark blue). Several resonances are labeled with amino acid names and numbers in the insets.

Since no structure is available for the isolated Thumb domain and it was not clear a priori whether the isolated domain has the same conformation as the Thumb domain in the p66 homodimer, we used NMR to investigate this question. Backbone assignments were obtained for a sample of the isolated domain and qualitatively assessed using CSI and TALOS+.^{39, 52, 53} The derived secondary structure elements (α-helices) are consistent with an overall domain structure similar to the one observed in the p66/p51 RT crystal. Likewise, Cα chemical shifts that were predicted using Sparta,⁵⁴ based on the coordinates of the Thumb domain in the RT heterodimer crystal structure (PDB ID,1DLO), match the experimental ones of the isolated Thumb domain (Fig. S2). The similarity in secondary structure elements, and probably the tertiary structure, for the Thumb domain alone and that in the p66 homodimer is further inferred from the resonance overlap noted above (59%, Table I).

The combined results imply that the Thumb and RNH domains within the p66 homodimer and the Thumb domain in the p51 exhibit the same overall structures as the isolated domains. This, however, is not the case for the Finger-Palm region (Table I). For this domain, notable chemical shift differences were observed in the p66 homodimer, possibly due to local structural changes or altered domain-domain or subunit-subunit interactions.

Conservation of the p51-RNH processing site conformation in p66

The conformation of the p51-RNH processing site (F440↓Y441) in the p66 homodimer was assessed based on the comparison of the amide resonance frequencies in the ¹H-¹⁵N HSQC spectrum of the p66 homodimer and those in the isolated RNH domain.^{25, 60} Backbone amide resonances from residues in the β-1 strand, which contains the processing site in the isolated RNH domain, were essentially identical (amino acids 438 to 447) and several of these (I434, F440, and Y441) are labeled in the Fig. 3B inset. The fact that no chemical shift differences were noted for these processing site residues suggests that they are similarly located in a β-strand within the p66 homodimer (Fig. S3). Since NMR resonance frequencies are highly sensitive to local electronic and conformational influences, the close match between amide resonances for the p51-RNH processing site in the p66 homodimer and the isolated RNH domain indicates that the site is similarly structured in both. We also did not observe a significant number of random coil resonances in the p66 spectrum, making it very unlikely that one or both of the RNH domains is unfolded or disordered in the RT p66 homodimer. Since the intensities of the observed RNH resonances are similar or slightly larger than those of the observed TrpNεH resonances of the p51 domain in the p66 spectrum (Fig. 3), it is very unlikely that another RNH conformation exists in the p66 homodimer, different from the one we identified. Even with possible exchange between an unfolded and folded RNH domain in p66, any unstructured fraction would be <5% of total, given the close match between resonance frequencies for the bona-fide folded, isolated RNH domain and the RNH domain in p66. Given all of the above, it is strongly suggested that the processing mechanism has to involve the selection of a very minor conformer in the overall conformational ensemble by the protease or a local conformational change upon protease binding.

DISCUSSION

Numerous p66/p51 RT structures are available, with and without substrate, inhibitors and other ligands, allowing for a detailed elucidation of the conformational transitions that are possible in the RT heterodimer upon ligand interaction. In all of these structures, differences in domain orientations between the two subunits are observed.^{16, 22, 24, 61, 62} In contrast, structural information for the p66 homodimer, which is assumed to be the precursor of mature RT,^{15–20, 43} is not available. Therefore, any conformational changes that accompany processing of the homodimer into the heterodimer remain ambiguous.

Here, we investigated the solution conformation of the p66 homodimer, aimed at providing the basis for elucidating RT heterodimer formation. We showed that the ¹H-¹⁵N TROSY HSQC spectrum of the p66 homodimer contains a very large number of backbone amide resonances, a rather surprising finding for a protein with a molecular mass of 132 kDa. This suggests that some of the smaller domains exhibit a certain degree of independent motion, faster than the overall rotational molecular diffusion. Our study also establishes that the overall major structures of the Thumb and RNH domains within the p66 homodimer are very similar to those of the isolated domains. In addition, the high similarity in chemical shifts for the p51-RNH processing site in the isolated RNH domain and the p66 homodimeric RT indicates that the cleavage site in the p66 homodimer exhibits essentially the same conformation as the isolated RNH domain, seemingly buried in the β-structure of the domain.

The mechanistic details involved in the formation of the mature RT p66/p51 heterodimer are not well understood, but several models have been proposed. One prevailing model assumes that cleavage of the immature p66 homodimer occurs in an unstructured RNH domain (Fig. 1C),^{21, 22} with subsequent folding of the remaining p66-RNH domain into the structure of the mature heterodimeric RT. This hypothesis is based primarily on the assumption that p51-RNH processing would be very inefficient if the processing site were located within a structured domain.¹⁷ Another proposal assumes that the p66 homodimer exists in an asymmetric conformation.^{11, 18} Indeed, the p51 domain conformation in the p66 subunit differs from that in the p51 subunit of the mature RT heterodimer, and this could be true in the p66 precursor as well. A third proposal suggests cleavage of p66 monomers, followed by folding of both p66 and p51 monomers, and formation of a mature p66/p51 heterodimeric RT (a concerted model).^{12, 13, 15} While our NMR results do not allow us to propose an exclusive model, they do provide insight into the conformational characteristics in the RT homodimer, thereby limiting possible maturation scenarios. First, the observation that the p66 NMR spectrum is a composite of those of the folded sub-domains indicates that p66, whether monomer or homodimer, is folded in solution. Second, we found that the RNH domain is structured in the p66 homodimer, with a conformation essentially identical to that seen in the isolated RNH domain, not altered by domain-domain contacts. Third, none of our data suggest that there are different conformations of RNH in the two p66 subunits of the homodimer.

Our observation of a structured RNH domain in the p66 homodimer is consistent with previous biochemical results. Protease predominantly processes p66 at the p51-RNH processing site (F440-Y441), compared to other processing sites within the RNH domain in the p66 precursor (Y483-L484, N494-I495, and Y532-L533), and removal of the RNH domain does not involve multiple cleavages.^{10, 17, 63–65} Since our data show that the RNH in the p66 homodimer is folded, this explains why other protease cleavage sites are not used and only the authentic p51-RNH site is cleaved.

In the current study, NMR signals in the Finger-Palm and Connection domains were not assigned. Therefore, our data cannot rule out a model in which the p66 homodimer exists in an asymmetric conformation. However, any model in which an overall different conformation of the RNH domain is proposed for the two subunits in the p66 homodimer is very unlikely, given our data. Even if in solution a small free energy difference between the two RNH domains could exist that may give rise to two different conformations (conformational substrates) for very short times, the overall conformational equilibrium will result a single average conformation over longer time scales. Thus, both domains will adopt the same average energy state, even if two energetically distinct conformations, separated by a low energy barrier, exist. While our manuscript was in review, Zheng, et al. published a study in which they evaluated Ile methyl resonances, and suggested selective unfolding of one of the RNH domains in the p66/p66 homodimer⁶⁶. Based on our data, we believe that major disruption of the RNH structure is unlikely, although local or temporal unfolding of a small region around the processing site cannot be ruled out.

CONCLUSION

Our results support maturation models that involve folded, rather than unfolded or disordered RNH domains (Fig. 4). This is somewhat surprising since the p51-RNH cleavage site has to be processed by the protease, thus needs to be available for binding. However, as pointed out above, if a minor conformation (<5%) were present, this may have eluded detection, given the limited sensitivity and signal overlap in the spectra. Indeed, at present we suggest that such a minor conformation of the p51-RNH processing site can be selected or induced by the protease.

Figure 4.

Possible RNH conformations in the p66 homodimers as maturation precursors: (A) the RNH domains are unfolded or disordered, (B) there is conformational difference between the two RNH domains, (C) both RNH domains in the two subunits are folded. Our data support (C).