The Arginine-Rich RNA-Binding Motif of HIV-1 Rev Is Intrinsically Disordered and Folds upon RRE Binding

Abstract

Arginine-rich motifs (ARMs) capable of binding diverse RNA structures play critical roles in transcription, translation, RNA trafficking, and RNA packaging. The regulatory HIV-1 protein Rev is essential for viral replication and belongs to the ARM family of RNA-binding proteins. During the early stages of the HIV-1 life cycle, incompletely spliced and full-length viral mRNAs are very inefficiently recognized by the splicing machinery of the host cell and are subject to degradation in the cell nucleus. These transcripts harbor the Rev Response Element (RRE), which orchestrates the interaction with the Rev ARM and the successive Rev-dependent mRNA export pathway. Based on established criteria for predicting intrinsic disorder, such as hydropathy, combined with significant net charge, the very basic primary sequences of ARMs are expected to adopt coil-like structures. Thus, we initiated this study to investigate the conformational changes of the Rev ARM associated with RNA binding. We used multidimensional NMR and circular dichroism spectroscopy to monitor the observed structural transitions, and described the conformational landscapes using statistical ensemble and molecular-dynamics simulations. The combined spectroscopic and simulated results imply that the Rev ARM is intrinsically disordered not only as an isolated peptide but also when it is embedded into an oligomerization-deficient Rev mutant. RRE recognition triggers a crucial coil-to-helix transition employing an induced-fit mechanism.

Introduction

RNA-binding proteins are fundamentally important for diverse cellular and viral regulatory processes. One widespread RNA-binding domain is the arginine-rich motif (ARM), which is present in various proteins, such as ribosomal proteins, transcriptional antitermination N proteins of bacteriophages λ and P22, and the HIV-1 regulatory proteins Tat and Rev (1–5). A preponderance of arginine residues seems to be the only common feature among these apparently simple and relatively short sequences (10–20 amino acid residues in length). While maintaining the ability to interact with their cognate targets with high affinity and specificity (6,7), they have been shown to be surprisingly dynamic and can adopt different conformations, such as α-helical, β-hairpin, or extended conformations, after binding their specific RNA targets.

Rev is an ARM-containing RNA-binding protein and a critical regulator of the HIV-1 replication cycle (8–10). Rev interacts specifically with a highly structured intronic element hosted within the Env gene known as the Rev Responsive Element (RRE), which is part of all intron-containing viral messenger RNAs (mRNAs). Upon recruitment of proteins that belong to the cellular nuclear-export machinery, Crm1 and the GTPase Ran, Rev induces export of unspliced and singly spliced mRNA to the cytoplasm for subsequent translation of virion assembly proteins (11–13). To coordinate these events, Rev must engage in transient protein-RNA interactions as well as protein-protein interactions, including oligomeric interactions among Rev monomers and interactions with cellular proteins of the nuclear import and export machineries. These adaptive recognition processes require an adequate degree of conformational flexibility.

The Rev ARM comprises 17 amino acid residues, including a total of 10 arginines (Rev_34-50), and in complex with its cognate RNA, RRE StemIIB, it adopts an α-helical conformation that fits deeply into the RRE StemIIB widened major groove (14). Four arginine residues (R35, R38, R39, and R44), a threonine (T34), and an asparagine (N40) convey specific, high-affinity binding to RRE StemIIB (4,14–16). However, Rev ARM has also been shown to adopt an extended conformation upon binding to anti-Rev RNA aptamers (15,17). Moreover, interactions with selected DNA motifs do not require the T34 or N40 side chains (18), and various arginine residues play key roles in complexes with distinctly folded RNA targets, including the primary and secondary binding sites of the RRE (8). Overall, the variation observed in Rev ARM complex structures suggests that the conformation of Rev ARM is induced by and dependent upon its specific nucleic acid binding target. Such an induced-fit or adaptive-binding model has been successfully applied to describe several protein-protein and protein-nucleic acid (DNA-RNA) interactions (19–21).

In contrast, early investigations employing circular dichroism (CD) and NMR suggested that the Rev ARM is predominantly helical in solution (4,22), supporting a conformational selection (selected-fit or population-shift) model (23–27). In this scheme, RRE binding would induce a shift in the conformational ensemble, resulting in preferential selection and stabilization of a preformed, helical, RRE-binding competent conformation of the Rev ARM. However, both of these studies optimized conditions to increase helicity, either by modifying the peptide’s termini or by working at low pH.

A recently reported x-ray structure of Rev complexed with an engineered monoclonal Fab (23) and another of an oligomerization-deficient mutant that forms a homodimer (28,29) both show that Rev’s 65 N-terminal residues adopt an antiparallel helix-turn-helix architecture, and that the ARM is helical in these oligomeric complexes. Pivotal studies employing single-molecule fluorescence have unraveled the oligomerization pathway of Rev onto its cognate RNA target sequence. The initial RRE binding and successive oligomerization occur in a stepwise manner whereby a nucleating Rev monomer binds to the high-affinity binding site StemIIB followed by highly cooperative binding of additional Rev monomers to form oligomeric assemblies stabilized by protein-protein and protein-RNA interactions (30). Dimeric Rev/Rev complexes do not constitute the basic building block for oligomeric Rev/RRE complex assembly. Because Rev populates a monomeric state only at submicromolar concentrations (31), RRE-binding-competent Rev monomers continue to evade structural characterization by x-ray crystallography or spectroscopic methods.

In this study, we employed heteronuclear, multidimensional NMR and CD spectroscopy as well as molecular-dynamics (MD) calculations to investigate whether the wild-type Rev ARM peptide adopts a helical, RRE-binding competent conformation in aqueous solution under physiological conditions. Furthermore, we examined the conformation of the ARM region in the context of an oligomerization-deficient Rev variant, the double-mutant V16D/I55N, which carries mutations in the known oligomerization interfaces and permits investigation of a mostly monomeric state of Rev protein at concentrations suitable for NMR studies (30). For all of the peptide studies, we employed a previously described Rev ARM construct that features the wild-type sequence with four amino acids (G_-4AMA) appended to the N-terminus of Rev_34-50 followed by A₊₁AAAR at the C-terminus (8). To obtain data for a well-defined helical conformation of the Rev ARM, we used two preparations: 1), a complex of the Rev ARM peptide with RRE StemIIB in physiological buffer; and 2), the Rev ARM peptide in 50% 2,2,2-trifluoroethanol (TFE)/50% physiological buffer. The TFE preparation provided a helical reference sample without introducing additional signals.

Materials and Methods

Expression and purification of Rev ARM peptide and full-length Rev constructs

We cloned and overexpressed the HIV-1 Rev ARM peptide as a cleavable GB1 fusion in Escherichia coli. The HIV-1 ARM peptide sequence GAMATRQARRNRRRRWRERQRAAAAR (native amino acids are underlined (8)) was expressed in E. coli strain BL21/DE3 from a p(H)GB1-derived vector as an N-terminal fusion with a His₆ tag, a GB1 domain, and a TEV-protease cleavage site (32). Plasmids coding for (His₆)-tagged Rev in T7 expression vector (pSG003) were previously described (33). Cultures of E. coli BL21-CodonPlus (DE3)-RIPL cells expressing (His₆)-tagged Rev wild-type and V16D/I55N mutant were grown from a single colony in 500 mL ¹⁵N- or ¹⁵N- and ¹³C-enriched M9 minimal media at 37°C overnight. Cells were grown to OD₆₀₀ of 0.8 at 37°C in LB or M9 media with 100 μg/mL ampicillin. Isopropyl-β-D-thiogalactopyranoside (IPTG) was added to 1 mM to induce expression for 4 h at 37°C, and harvested cells were resuspended in denaturing lysis buffer (50 mM Na₂HPO₄/NaH₂PO₄ pH 7.4, 500 mM KCl, 10 mM imidazole, 1 mM dithiothreitol [DTT] and 8 M urea). A urea denaturation and on-column refolding protocol was used to remove nonspecific nucleic acid contamination as previously described (34), and the resulting sample was free of RNA contamination as determined by the 260/280 nm absorbance ratio (values of ∼0.6) (34). Eluted fractions were analyzed by SDS-PAGE and pooled, and in the case of the Rev ARM peptide fusion protein, TEV protease was added and incubated at room temperature overnight. The resulting peptide was purified by RP-HPLC using a Jupiter Proteo column (Phenomenex, Torrance, CA) and a linear gradient of acetonitrile in the presence of 0.1% trifluoroacetic acid (TFA). Collected fractions were pooled and lyophilized.

For uniform ¹⁵N labeling, the constructs were expressed in M9 minimal medium supplemented with ¹⁵NH₄Cl[¹⁵N, 99%] (Cambridge Isotope Laboratories, Andover, MA) as the sole nitrogen source. For uniform ¹⁵N,¹³C labeling, the peptide was expressed in M9 minimal medium supplemented with ¹⁵NH₄Cl[¹⁵N, 99%] and ¹³C-glucose[¹³C, 99%] (Cambridge Isotope Laboratories).

RNA transcription and purification

RRE StemIIB-containing plasmids were purified with the E.Z.N.A.Plasmid Giga Kit (Omega Bio-Tek, Norcross, GA), linearized with the appropriate 3′ restriction endonuclease, phenol/chloroform extracted, and ethanol precipitated. DNA template-directed in vitro transcriptions using T7 RNA polymerase were performed in transcription buffer (40 mM Tris, pH 8.0, 1 mM spermidine, 10 mM DTT, 0.01% Triton X-100) containing 0.05 μg/mL plasmid template, 30 mM MgCl₂, 1 μM T7 RNA polymerase, and 6.5 mM each of ATP, CTP, GTP, and UTP). Transcription reactions were incubated at 37°C for 4 hr.

Transcripts were purified as previously described (35). Briefly, full-length transcripts were separated from protein, unincorporated nucleotides, and abortive transcripts by anion exchange chromatography using a HiTrapQ HP column (GE Healthcare). The RNA was exchanged into buffer containing 50 mM Na₂HPO₄/NaH₂PO₄ pH 7.4, 150 mM KCl, 1 mM EDTA, 1 mM DTT, 0.02% NaN₃). Complex spectra were acquired after addition of concentrated RRE StemIIB solution to Rev ARM peptide samples (final molar ratio: 1.1 (RNA) to 1.0 (peptide)).

CD spectroscopy

CD measurements of purified Rev ARM peptide, wild-type Rev, and V16D/I55N Rev mutant protein were recorded on an AVIV 400 spectropolarimeter (AVIV Biomedical, Lakewood, NJ) at 25°C between 198 and 260 nm, with a 0.1 cm path-length cell, a wavelength increment of 1 nm, an averaging time of 5 s, and an equilibration time of 5 min. The baseline was corrected by subtracting the spectra of the respective buffers collected under identical conditions. Spectra from three scans were averaged. CD spectra of 50 μM Rev ARM peptide samples in buffer (20 mM Na₂HPO₄/NaH₂PO₄ pH 7.4, 150 mM KCl) were acquired in the presence of 0–50% (v/v) TFE, after preincubation at 25°C for 1 h. CD spectra of 10 μM wild-type Rev and V16D/I55N Rev mutant protein samples were collected in the same buffer in the absence of TFE. The helical content was estimated using the ratio of the mean residue ellipticities (θ) at 222 and 208 nm.

NMR spectroscopy: sample preparation

The lyophilized peptide samples were dissolved in a buffer containing 50 mM Na₂HPO₄/NaH₂PO₄, pH 7.4, 150 mM KCl, 1 mM EDTA, 1 mM DTT, 0.02% NaN₃ in 500 μL of 90% H₂O/10% D₂O or 50% TFE/40% buffer/10% D₂O to a final peptide concentration of 1 mM. Aliquots of wild-type Rev and V16D/I55N Rev mutant protein were buffer exchanged against Rev NMR buffer (50 mM Na₂HPO₄/NaH₂PO₄, pH 7.4, 500 mM KCl, 1 mM EDTA, 1 mM DTT, 0.02% NaN₃) in 500 μL of 90% H₂O/10% D₂O to a final protein concentration of ∼100–200 μM.

NMR data collection and processing

NMR experiments were recorded on Bruker Avance III 600 MHz and Avance II 700 and 800 MHz spectrometers (Bruker AG, Karlsruhe, Germany). The Avance 700 and 800 are equipped with a 5 mm ¹H[¹³C/¹⁵N] triple-resonance probe featuring cryogenically cooled preamplifiers and radiofrequency coils on ¹H and ¹³C channels and the z-axis gradient, and the 600 MHz instrument features a conventional 5 mm ¹H[¹³C/¹⁵N] triple-resonance probe. All NMR experiments were carried out at either 283 or 298 K. ¹H chemical shifts were externally referenced to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS), with heteronuclear¹³C and ¹⁵N chemical shifts referenced indirectly according to the ¹X/¹H ratio via DSS.

The NMR data were processed using nmrPipe (36) and analyzed with CcpNmr analysis software (37).

Residual dipolar coupling measurements

For residual dipolar coupling (RDC) measurements, 6% stretched neutral polyacrylamide gels were cast using the apparatus described by Chou et al. (38) and prepared from a stock solution of 36% w/v acrylamide and 0.94% w/v N,N-methylenebisacrylamide (molar ratio of acrylamide/bisacrylamide of 83:1), 0.08% ammonium persulfate, and 0.7% tetramethylethylenediamine.

Alkyl-PEG bicelles for RDC measurements were prepared from a 15% bicelle stock solution (150 μL C12E5, 550 μL NMR buffer, and 300 μL D₂O) according to the protocol described by Gebel and Shortle (39). A solution of n-hexanol was titrated into the C12E5 solution in microliter additions up to a molar ratio C12E5/n-hexanol of 0.96. Bicelles were added to the peptide sample in NMR buffer to make a 6% bicelle sample containing 1 mM Rev ARM peptide.

NMR assignments and structure calculation

Unique backbone and side-chain assignments were made by combining three-dimensional (3D) HNCO, HNCA, HN(CO)CA, CBCA(CO)NH, HNCACB, HBHA(CO)NH, and H(CCCO)NH (40); ¹⁵N-separated 3D nuclear Overhauser effect spectroscopy-heteronuclear single quantum coherence (NOESY-HSQC) (41); HSQC-NOESY-HSQC (42); and 3D HNHA (43,44) experiments. Analysis of a simultaneous ¹³C/¹⁵N-separated 3D NOESY-HSQC (45) spectrum together with secondary chemical-shift-derived torsion restraints (46) yielded extensive restraints for structure elucidation. Isotropic and anisotropic one-bond ¹J(H^N,N) and ¹D(H^N,N) coupling constants were obtained from gradient-enhanced IPAP-HSQC experiments employing band-selective ¹H^aliphatic decoupling pulses (47). Vicinal ³J(H^N,Hα) coupling constants were determined using quantitative HNHA experiments. To compensate for finite ¹Hα spin flips, apparent ³J(H^N,Hα) couplings were multiplied by a uniform factor of 1.05 (44). Steady-state, heteronuclear [¹H^N] ¹⁵N NOEs were measured according to standard procedures (48). Heteronuclear NOE values are reported as the ratio of peak intensities in paired interleaved spectra collected with and without the initial proton saturation period (3 s) during the 5 s recycle delay.

Distance restraints were obtained from NOEs in simultaneous, 3D ¹³C- and ¹⁵N-resolved NOESY experiments, and subsequent assignments of interresidue NOE peaks were carried out using the CANDID (49) algorithm implemented in CYANA 2.1 (L.A. Systems, Tokyo, Japan) (50). For the structure calculation of the Rev ARM peptide in 50% TFE, we evaluated a total of 236 intramolecular NOE distance constraints, including 53 medium-range NOEs (see Fig. S2 in the Supporting Material). Additionally, we derived 48 torsion angle constraints from the chemical-shift data using TALOS+ predictions, and used them in the calculations (46).

Generation of conformational ensemble

We generated explicit ensemble descriptions of the Rev ARM peptide using the flexible-meccano (FM) algorithm (51). The algorithm accounts for the heteropolymeric nature of the primary sequence and uses amino-acid specific ϕ,ψ-torsion angle sampling along with side-chain volumes to compute the conformational ensemble. The ensemble comprised 100,000 conformers, and simulated ¹D(H^N,N) and ³J(H^N,Hα) coupling constants were averaged over all ensemble members. The Karplus coefficients (³J (Θ) = A cos²Θ+BcosΘ+C) employed for back calculations were A = 7.9, B = −1.05, and C = 0.65 Hz (52).

MD simulations

MD simulations were generated using the Amber 11 suite of programs (53) and the ff99SB force field (54) with ILDN side-chain torsion modifications (55). All simulations used the particle mesh Ewald procedure (56) under periodic boundary conditions. A time step of 5 fs was used for all simulations, and the SHAKE algorithm (57) was used to constrain the length of bonds to hydrogen atoms. Langevin dynamics (58) with a collision frequency of 1 ps⁻¹ was used to maintain temperature. All-atom representations of the molecules were solvated in truncated octahedral boxes of TIP3P water (59), and charge was neutralized by the addition of 10 chloride ions.

Initial coordinates for the Rev ARM peptide in the unfolding simulation were taken from Protein Data Bank (PDB) entry 1ETF (14). The N-terminal Asp residue was replaced with the sequence Gly-Ala-Met-Ala using UCSF Chimera (60). These four N-terminal residues were added in a parallel β-strand conformation. Coordinates for the missing hydrogen atoms were calculated using UCSF Chimera.

To generate multiple unstructured initial coordinates for the folding simulations, we built an extended peptide in the antiparallel β-strand conformation using UCSF Chimera and subjected it to 50 steps of steepest-descent energy minimization, followed by 50 steps of conjugate-gradient minimization. The equilibrated extended peptide was heated to 1500 K and held at this temperature for a 500 ps in vacuo trajectory. The trajectory was clustered, and representative structures from each of the 10 clusters were solvated using different-sized truncated octahedral boxes to obtain a similar number of water molecules for each folding simulation (Table S1).

The solvated systems were energy minimized in two stages. First, the solvent was minimized with 500 steps of steepest-descent energy minimization followed by 500 steps of conjugate-gradient minimization with positional restraints holding the peptide fixed. The second stage minimized the entire system with 1000 steps of steepest descent followed by 1500 steps of conjugate-gradient minimization. Next, the systems were equilibrated, again in two stages. Initially, the peptide was held fixed while the solvent was equilibrated over 20 ps at 1 atm and 283 K. In the second stage, we equilibrated the entire system over 100 ps by removing the restraints on the peptide. Production runs of 50 ns were performed in the NPT ensemble, and coordinates were saved every 5 fs. Trajectories of each system were extended to 50 ns with a different random number seed on each restart (61).

Trajectories were analyzed using the Amber ptraj module. Root mean-square deviations (RMSDs) were calculated using the backbone C, Cα, and N atoms of all residues unless specified otherwise. Secondary structure was determined using the DSSP (Define Secondary Structure of Proteins) algorithm. Clustering used the average linkage algorithm. The dihedral angle Θ was defined by the HN, N, Cα, and Hα atoms, to be consistent with the NMR measurements, and converted to coupling constants using values of 7.90, −1.05, and 0.65 in the Karplus relationship.

Results

CD spectroscopy

CD spectroscopy has been used extensively to study Rev, and was the first method we applied. CD spectra of the Rev ARM peptide in physiological buffer exhibited a minimum around 200 nm, typical of an unfolded state (Fig. 1 A). Upon titration with TFE, minima appeared at 208 and 222 nm and the ellipticity progressively increased, suggesting an increase in helical content. The ratio [θ₂₂₂]/[θ₂₀₈], a measure of helical content, increased from 0.42 in the absence of TFE to a maximum of 0.82 in the presence of 50% (v/v) TFE (Fig. 1 B). Accordingly, we chose to use 50% TFE as a sample condition to provide a reference helical conformation of the Rev ARM peptide. A comparison of the CD spectra collected for wild-type Rev and V16D/I55N Rev mutant protein confirmed significant α-helical content for oligomeric wild-type Rev. In contrast, the oligomerization-deficient Rev double-mutant V16D/I55N adopted a mostly random coil conformation (Fig. 1 C).

Figure 1.

TFE-dependent CD of the Rev ARM peptide and CD of Rev protein variants. (A) Mean residue molar ellipticity θ as a function of wavelength (198–260 nm) for Rev ARM in increasing concentrations of TFE cosolvent. The double minima at 208 and 222 nm indicate an α-helical secondary structure in TFE, and the TFE-free spectrum exhibiting a minimum at ∼200 nm is indicative of a more random coil conformation. (B) [θ₂₂₂]/[θ₂₀₈] ratio as a function of TFE concentration. (C) Mean residue molar ellipticity θ as a function of wavelength (198–260 nm) for wild-type Rev and V16D/I55N mutant protein in buffer, pH 7.4.

Peptide ¹H^N,¹⁵N-HSQC spectra

To obtain residue-specific information regarding the Rev ARM peptide, we recorded ¹H^N,¹⁵N-HSQC spectra (Fig. 2). In aqueous solution, the peptide shows little dispersion of the resonances (Fig. 2 B), indicative of an unfolded peptide and in accordance with the CD data. In contrast, the ¹H^N,¹⁵N-HSQC spectra of the same construct in 50% TFE (Fig. 2 C) and in complex with RRE StemIIB (Fig. 2 D) show much greater dispersion. Overall, the TFE- and RRE StemIIB-induced, weighted amide chemical shift changes (Δδ(¹H^N,¹⁵N)) are similar in magnitude, averaging 0.31 and 0.34 ppm, respectively. Interestingly, the majority of the Δδ(¹H^N,¹⁵N) values induced by 50% TFE and RRE-StemIIB binding resemble each other in direction and magnitude, suggesting that the conformations in the presence of 50% TFE and RRE-StemIIB are similar. The RRE-StemIIB-induced Δδ(¹H^N,¹⁵N) values of R35 and R44 differ considerably from those induced by 50% TFE; however, this difference is likely due to the base-specific contacts these residues make with opposite sides of the RRE-StemIIB major groove (see Fig. S1) (14).

Figure 2.

Resonance dispersion of Rev ARM in ¹H,¹⁵N-HSQC spectra. (A) Primary sequence of the Rev ARM peptide construct investigated. Nonnative N-terminal residues include G_-4 through A_-1, and the C-terminal extension consists of A₊₁ through R₊₅. (B) ¹H,¹⁵N-HSQC of Rev ARM peptide in NMR buffer (50 mM Na₂HPO₄/NaH₂PO₄ pH 7.4, 150 mM KCl, 1 mM EDTA, 1 mM DTT, 0.02% NaN₃). (C) ¹H,¹⁵N-HSQC of the Rev ARM peptide in 50% TFE/40% NMR buffer/10% D₂O. (D) ¹H,¹⁵N-HSQC of the Rev ARM peptide complexed with RRE StemIIB in NMR buffer. Insets show the side-chain W45 indole and the T34 backbone amide correlations. All ¹H,¹⁵N-HSQC experiments were acquired at 800 MHz and 283 K.

Wild-type and mutant Rev ¹H^N,¹⁵N-correlation spectra

NMR investigations of the 116-residue wild-type Rev in solution are notorious for limited protein solubility and uncontrolled oligomerization. Therefore, sequential assignments outside the ARM remain unavailable. Fewer than 20% of all backbone amides, all indicative of a disordered protein region, are observable in ¹H^N,¹⁵N-transverse relaxation optimized spectroscopy (¹H^N,¹⁵N-TROSY) spectra (62) recorded at 283 K (Fig. 3 A). However, because the V16D/I55N mutations are known to block oligomerization of Rev on the RRE (30), the introduction of these two mutations into the protein oligomerization domains allows for detection of many additional, albeit poorly dispersed, resonances located in the N-terminal domain of Rev (Fig. 3 C). Employing conventional through-bond assignment strategies, we were able to sequentially assign >70% of the backbone amide resonances in the V16D/I55N Rev mutant construct. Backbone amide assignments obtained for the V16D/I55N Rev mutant could be transferred to the wild-type spectrum, where they mapped exclusively to the extreme C-terminus (Fig. 3, A and C). Although sizable portions of the oligomerization interfaces centered around helices 1 and 2 are broadened beyond detectable limits, even in the V16D/I55N Rev mutant (Fig. 3 D), unambiguous assignments include the N-terminal part of the ARM, spanning residues T34 through N40 (Fig. 3 B). The observed similarities between the backbone amide chemical shifts in the V16D/I55N Rev mutant construct and those in the Rev ARM peptide strongly suggest that the adopted confirmation is context independent and disordered (Fig. 3 B).

Figure 3.

NMR analysis of wild-type Rev and V16D/I55N mutant protein and secondary chemical shifts of Rev V16D/I55N. (A) An 800 MHz ¹H,¹⁵N-TROSY spectrum of wild-type Rev in NMR buffer containing 500 mM KCl, pH 7.4, recorded at 283 K. Assignments for observable residues are provided. (B) Inset showing an overlay of ¹H,¹⁵N-HSQC spectrum of full-length V16D/I55N Rev (single black contour) and Rev ARM peptide (gray contours) in NMR buffer recorded at 283 K. Arrows and labels indicate the position of assigned residues in the ARM region. (C) An 800 MHz ¹H,¹⁵N-HSQC spectrum of full-length V16D/I55N Rev in NMR buffer containing 500 mM KCl, pH 7.4, recorded at 283 K. (D) Combined secondary chemical shifts, Δδ(¹³Cα,β), of full-length V16D/I55N Rev in NMR buffer, pH 7.4, are plotted against the residue number. The helix-turn-helix motif and functionally important ARM/NLS and NES are shown along with the full-length Rev primary sequence. Arrows indicate the position of the two mutations in the Rev oligomerization domains (V16D and I55N). Boxed C-terminal residues are assigned in panel A.

Secondary chemical shifts

To obtain additional insights into the conformational preferences of the Rev ARM peptide and the V16D/I55N Rev mutant, we analyzed the ¹³Cα and ¹³Cβ secondary chemical shifts. The difference between ¹³Cα and ¹³Cβ secondary chemical shifts for the Rev ARM peptide in aqueous solution at pH 7.4 averages 0.8 ppm over all residues, consistent with a random coil-like conformation (Fig. 4 A). In contrast, in the presence of 50% TFE (Fig. 4 B) or RRE StemIIB (Fig. 4 C), the Rev ARM peptide exhibits differences in ¹³Cα and ¹³Cβ secondary chemical shifts averaging 2.8 and 3.7 ppm, respectively, consistent with an α-helical conformation. Most of the residues between T34 and R50 have combined ¹³Cα and ¹³Cβ secondary chemical shifts between 3 and 4 ppm, characteristic of a fully formed α-helix (63). Secondary chemical shifts of the oligomerization-deficient V16D/I55N Rev mutant confirm that the C-terminus of Rev is disordered. Furthermore, the observed deviations from random coil for ¹³Cα and ¹³Cβ chemical shifts of residues located in the ARM region (T34 and Q36–N40) within the V16D/I55N Rev mutant (Fig. 3 D) are virtually identical to the corresponding secondary chemical shifts in the ARM peptide context. In similarity to the ARM peptide scenario, this indicates that residues T34-N40 do not adopt a stably folded α-helical conformation in the context of the mostly monomeric V16D/I55N protein variant.

Figure 4.

Secondary structure of Rev ARM as indicated by ¹³C chemical shift. Combined secondary chemical shifts, Δδ(¹³Cα,β), are plotted as the difference of experimental ¹³Cα and ¹³Cβ chemical-shift deviation from random coil chemical shifts. Δδ(¹³Cα,β) is plotted against residue number for (A) Rev ARM in NMR buffer, pH 7.4; (B) Rev ARM in 50% TFE; and (C) Rev ARM in complex with StemIIB (14,65). Carbon chemical shifts for Rev ARM peptide in complex with RRE StemIIB were taken from Battiste (65). Native Rev ARM_34-50 residues are shown in bold one-letter amino acid abbreviations, and nonnative residues are indicated by an asterisk. Values above the dashed line at 2.8 ppm are indicative of an α-helical conformation.

³J(H^N,Hα) coupling constants

Fig. 5 A summarizes the ³J(H^N,Hα) coupling constants for Rev ARM peptide in aqueous solution at pH 7.4, in 50% (v/v) TFE, and in the RRE-bound state. One can interpret the vicinal coupling data by considering the mean ³J(H^N,Hα) values predicted for α-helical and β-strand conformations (64). In helical regions, the ³J(H^N,Hα) coupling constants average 5.2 Hz, in β-sheet like conformations they average 8.5 Hz, and in random coil-like regions, intermediate values ranging from 7.1 to 7.7 Hz are found. For the Rev ARM peptide in aqueous solution at pH 7.4, we observed intermediate ³J(H^N,Hα) couplings, with an average of 6.0 Hz for residues T34–R50. In comparison, the J couplings of the Rev ARM peptide in 50% TFE are smaller, with an average value of 4.7 Hz for residues T34–R50, indicative of the presence of α-helix. Similar values were reported by Battiste (65) for the α-helical conformation of the Rev ARM peptide in complex with RRE StemIIB (with an average value of 4.4 Hz for residues T34–R50).

Figure 5.

³J(H^N,Hα) coupling constants and [¹H^N]¹⁵N-heteronuclear NOE analysis. (A) Vicinal ³J(H^N,Hα) coupling constants for Rev ARM peptide in NMR buffer (black squares) and 50% TFE (gray triangles) were determined from quantitative HNHA experiments. ³J(H^N,Hα) coupling values for Rev ARM peptide in complex with RRE StemIIB (black circles) were taken from Battiste (65). Error bars indicate the estimated error in the three-bond HNHA coupling. (B) [¹H^N]¹⁵N-hetNOEs of Rev ARM peptide in NMR buffer, pH 7.4 (black squares), in 50% TFE (gray triangles), and in complex with RRE StemIIB (black circles). Error bars indicate the SD in the measured values. Native Rev ARM_34-50 residues are shown in bold one-letter amino acid abbreviations, and nonnative residues are indicated by an asterisk.

Backbone dynamics

Analysis of [¹H^N],¹⁵N heteronuclear NOEs (hetNOEs) conveniently identifies fast-timescale protein backbone dynamics on a subnanosecond timescale (66). Fig. 5 B shows that the [¹H^N],¹⁵N hetNOE values for Rev ARM in aqueous solution at pH 7.4 average 0.47 for residues T34–R50, consistent with a disordered peptide. In contrast, the hetNOEs of the Rev ARM peptide in complex with RRE StemIIB average 0.83 for residues T34–R50, and in 50% TFE average 0.77 for residues T34–R50. These higher hetNOE values indicate higher backbone rigidity and are typical of a well-ordered secondary structure.

Structure calculations

¹H, ¹⁵N, and ¹³C chemical-shift assignments of Rev ARM in the absence and presence of 50% TFE were nearly complete, and assignments have been deposited in the BioMagResBank (accession numbers 18851 and 18852, respectively).

NOE-derived distance restraints and chemical-shift-derived backbone torsion restraints were used to calculate structural ensembles of the Rev ARM peptide in aqueous solution at pH 7.4 and in 50% TFE. In the case of the physiological sample, the simulated annealing procedure converged poorly and no single conformation could be defined that satisfied our experimental chemical shift and NOE data. In contrast, the data derived from the 50% TFE sample produced a well-defined helical ensemble. None of the final ensemble of 20 energy-minimized NMR structures shown in Fig. 6 A violated NOE distances by more than 0.5 Å. No torsion angle restraint was violated by more than 5°. The vast majority of residues (99.8%) were found in the most favored regions of the Ramachandran plot. The root mean-square deviation (RMSD) of the structure ensemble compared with the mean structure was 0.87 ± 0.35 Å for backbone and 1.69 ± 0.32 Å for all heavy atoms. A summary of the distance restraints and structural statistics is given in Table S1 and Fig. S2. The Rev ARM peptide in 50% TFE features an α-helical conformation (Fig. 6 B) that closely resembles the StemIIB-bound state (14) and the conformation adopted in 20% TFE of a succinylated R₄₂A point mutant (22), characterized by backbone RMSDs of 0.9 Å and 0.7 Å, respectively (Fig. S3, A and B). The atomic coordinates and experimental NMR restraints (PDB identifier 2M1A) for the Rev ARM structure in 50% TFE have been deposited in the PDB (http://www.rcsb.org/).

Figure 6.

NMR solution structure of Rev ARM in 50% TFE. (A) Ensemble of 20 CYANA conformers representing the solution structure of the Rev ARM peptide in the presence of 50% (v/v) TFE calculated using NOE-derived distance-restraints and TALOS+ chemical-shift-derived torsion-angle constraints. (B) Ribbon representation of the mean structure of Rev ARM in 50% TFE calculated from the ensemble using MolMol (89).

Residual dipolar couplings

To investigate whether residual conformational preferences exist in aqueous solution at pH 7.4, we oriented the Rev ARM peptide in stretched acrylamide gels (67) and polyethyleneglycol (PEG)/hexanol (68) mixtures to measure ¹D(H^N,N) RDCs. Although the absolute level of alignment differed between the two anisotropic environments, very similar general distributions of ¹D(H^N,N) emerged (Fig. 7, A and C). All but two residues at the very C-terminus of Rev ARM were found to have negative RDCs, with the largest values in the central residues. RDC values generally followed a bell-shaped distribution over the sequence, with small couplings being measured at the N- and C-terminal extremities. Such distributions have been observed and rationalized as random coil RDCs and are a direct consequence of the conformational properties embedded into the primary sequence (69–71).

Figure 7.

¹D(H^N,N) RDC analysis. (A) Experimental ¹D(H^N,N) values obtained in liquid crystalline media composed of n-dodecyl-penta(ethylene glycol) (C12E5) and n-hexanol versus PALES-predicted ¹D(H^N,N) RDCs calculated assuming steric alignment of Rev ARM peptide coordinates taken from PDB entry 1ETF (14). (B) Correlation between PALES-predicted ¹D(H^N,N) RDCs for the α-helical Rev ARM peptide conformation and the experimental ¹D(H^N,N) values obtained in C12E5/n-hexanol. The dashed line is the linear regression (slope of 2.581). (C) Experimental ¹D(H^N,N) values obtained in C12E5/n-hexanol, stretched PAGE gels, and RDCs predicted from 100,000 FM-generated conformers. (D) Correlation between FM-predicted ¹D(H^N,N) RDCs and the experimental ¹D(H^N,N) values obtained in anisotropic C12E5/n-hexanol mixture. The dashed line is the linear regression (slope of 1.266).

To rule out the presence of residual α-helical conformational preferences, we used the PALES software (72) to predict RDC data based on a steric mode of molecular alignment and the α-helical coordinates of the Rev ARM peptide observed in the complex with RRE StemIIB (PDB identifier 1ETF). A comparison of experimental ¹D(H^N,N) values obtained in PEG/hexanol mixtures with those predicted for the α-helical conformation of the Rev ARM peptide showed no obvious correlation (Fig. 7 A), as characterized by a Pearson’s correlation coefficient of 0.493 (Fig. 7 B). Thus within detectable limits, the experimental ¹D(H^N,N) values are inconsistent with the presence of an α-helical conformation.

Explicitly built conformational ensembles based on knowledge-based torsion-angle statistics have proved to be useful for characterizing intrinsically disordered proteins (IDPs) (73). RDC predictions based on the statistical FM (51) approach accurately reproduced the experimental ¹D(H^N,N) values obtained in stretched acrylamide gels and PEG/hexanol mixtures (Fig. 7 C). The correlation between experimental PEG/hexanol and FM-predicted ¹D(H^N,N) values has a Pearson’s coefficient of 0.842, suggesting that in aqueous solution at pH 7.4, the Rev ARM peptide is intrinsically disordered (Fig. 7 D). It should be noted that we corrected variations in the absolute levels of alignment by applying optimal scaling factors. Collectively, our RDC analysis demonstrates that the Rev ARM peptide in aqueous solution at pH 7.4 populates an ensemble of rapidly interconverting conformations that are mostly unstructured.

MD simulations

To further probe the lack of experimental evidence for a helical conformation of the Rev ARM peptide in aqueous solution at pH 7.4, we generated 11 different 50 ns all-atom explicit-solvent MD simulations. First, an unfolding simulation was started from the α-helical coordinates of the Rev ARM peptide observed in the NMR solution structure with RRE StemIIB (PDB identifier 1ETF). The raw RMSD from the starting coordinates plotted over the course of the simulation is large and continues to increase, but calculating the RMSD after superimposing the peptide on just the native residues (T34–R50) shows that the RMSD increases in two steps in the first 3 ns of the simulation and then remains fairly constant around 2 Å (Fig. 8 A). This indicates that after equilibrating over the first 3 ns, the majority of the motion in the unfolding simulation occurs in the termini. An examination of the secondary structure over the course of the unfolding simulation shows that a structural transition occurs around 3 ns and that the native residues in the core of the peptide remain predominantly α-helical throughout the simulation. Visual inspection of the trajectory confirms the helical conformation of the Rev ARM peptide. The unfolding simulation suggests that once the core of the peptide is formed, its helical character is quite stable over a 50 ns time span (Fig. 8 B).

Figure 8.

Unfolding simulation. (A) RMSD from starting coordinates over the course of the unfolding simulation. The upper gray line was calculated after superimposing the peptide on the backbone atoms of all residues, and the lower black line was calculated after superimposing it on the native residues (T34–R50). (B) Secondary structure of the Rev ARM peptide over the course of the unfolding simulation. Secondary structure was determined with the DSSP algorithm implemented in the Amber ptraj module. Secondary structure is colored as follows: coil (gray), β-strand (blue), turn (green), α-helix (red), and 3₁₀ helix (yellow).

Since the folded Rev ARM peptide was quite stable in our unfolding simulation, we decided to test whether the peptide would fold into a stable α-helix from random coil coordinates. To minimize the influence of the starting conformation, we generated 10 50-ns-long trajectories, each starting from different random coil coordinates. To produce starting coordinates, an extended structure of the Rev ARM peptide was built, subjected to an in vacuo high-temperature simulation, and then clustered (Fig. S4). To maximize sampling of conformational space, the temperature and length of the in vacuo simulation were increased until visualization of the trajectory revealed that the peptide had contracted and expanded multiple times, and clustering had produced clusters that were visited multiple times over the course of the trajectory with no single cluster dominating. Clustering of the trajectory into 10 clusters provided representative structures that were used as the starting coordinates for folding simulations (Fig. S5). Inspection of the secondary structure over the course of the folding trajectories shows that these simulations failed to produce any significant α-helical content (Fig. 9).

Figure 9.

Folding simulations. Secondary structure of the Rev ARM peptide over the course of the 10 folding simulations. Secondary structure was determined with the DSSP algorithm implemented in the Amber ptraj module. Secondary structure is colored as follows: coil (gray), β-strand (blue), turn (green), α-helix (red), and 3₁₀ helix (yellow).

To determine whether the simulations were consistent with the experimental NMR data, we measured the dihedral angle ϕ for all residues over the course of all 10 folding simulations and converted the values to mean coupling constants using a Karplus relation (52). The ³J(H^N,Hα) values derived from the simulations averaged ∼6.5 Hz, which is typical of a disordered peptide and consistent with the experimental values obtained in aqueous solution at pH 7.4 (Fig. S6). To examine the sampling of conformational space, we combined all 10 folding simulations and grouped the combined data into 10 clusters (Fig. S7). If conformational sampling was limited, then each cluster would correspond to one trajectory. If conformational sampling was extensive, then each cluster would be sampled from each trajectory. We found that seven of 10 clusters grouped snapshots from multiple trajectories, and that during all but one simulation the peptide transitioned to conformations sampled by other folding trajectories. The folding simulations appear to have sampled the available conformational space extensively, are consistent with the ³J(H^N,Hα) coupling constants obtained from the aqueous NMR experiments in which the peptide was disordered, and fail to show any consistent secondary structure formation, suggesting that on the 50 ns timescale, the Rev ARM peptide does not transition to an α-helical conformation.

Discussion

To date, NMR solution structures of the Rev ARM peptide in complex with its high-affinity recognition site, RRE StemIIB (14), bound to an RNA aptamer (15), and of the N-terminally succinylated R₄₂A point mutant of Rev ARM at pH 4.7 in the presence of 20% TFE (22) have been determined. Crystal structures for the N-terminal helix-turn-helix motif of a Rev double mutant (28) and one of a complex with monoclonal Fab fragments (29) have been solved. Collectively, these structures demonstrate the helical of Rev ARM upon RRE binding and homodimerization, and under conditions with TFE as a cosolvent. In contrast, when it is bound to a 35-mer RNA aptamer, Rev ARM adopts an extended conformation (17). To characterize the conformational state of free Rev ARM, we first confirmed the previously observed stabilization of α-helical structure by TFE, thereby establishing a reference helical preparation for comparison studies. Using heteronuclear 3D NMR spectroscopy, we confirmed that in 50% TFE, the structure of our Rev ARM peptide was very similar to the succinylated R₄₂A variant structure determined in 20% TFE, and to the conformation found when the peptide is bound to RRE StemIIB. When we compared the NMR data obtained in 50% TFE with those obtained in aqueous solution at pH 7.4, we found no evidence supportive of an α-helical character of wild-type Rev ARM in aqueous solution at pH 7.4. The strong tendency to self-associate has thus far thwarted attempts to obtain structural information about the full-length, wild-type Rev protein on the atomic level. Mutation of Val-16 and Ile-55 to Asp-16 and Asn-55 partially alleviated uncontrolled oligomerization and permitted unambiguous assignments for >70% of the backbone residues. Consistent with previous studies demonstrating the disordered nature of the C-terminus of the Rev L60R variant in complex with a 42 nt StemIIB hairpin (8,74), our secondary chemical shifts reveal intrinsic disorder for the C-terminus of the wild-type and the entire oligomerization-deficient V16D/I55N Rev variant, including the assignable portion of the ARM in the absence of the RRE.

To probe for low populations of α-helical ARM peptide conformations, we measured RDCs in two different alignment media. In a recent study, experimental RDCs were reproduced with the use of statistical models derived from backbone conformations sampled in coil-like structures of the PDB (73). We found that the FM statistical coil model, which describes an ensemble of unstructured, interconverting structures, accurately reproduced our experimental RDC data, whereas an α-helical conformation did not. Additionally, we generated >500 ns of MD simulations. Typically, coil-to-helix transitions are thought to be highly cooperative (75), with helix nucleation times estimated to be 20–70 ns (76), yet in 10 different 50 ns folding simulations of the Rev ARM peptide, not one coil-to-helix transition was observed.

The results from our comprehensive CD, NMR, and MD studies of the Rev ARM peptide under physiological conditions demonstrate that its conformation is most accurately represented by the FM statistical coil model, which is most likely a consequence of the abundance of positively charged arginine residues. Previous CD studies suggested that the unmodified Rev ARM (T34–R50) adopts a random coil-like structure, and that the apparent helical content of ARM peptides depends strongly on terminal modifications because it was highest for a construct featuring a succinylated N-terminus, a C-terminal extension of four alanines and one arginine residue, and an amidated C terminus (4). Similarly, qualitative homonuclear NMR studies suggesting a helical conformation for the Rev ARM peptide free in solution were performed with an N-succinylated-carboxamide R₄₂A point mutant and were carried out at pH 4.7 and 4°C, conditions that were selected because they increased the structured nature of the peptide (22).

In line with previous investigations that revealed concentration-dependent secondary structure formation for wild-type Rev and assigned a molten-globular state to the Rev monomer (77), our CD and NMR investigation of the oligomerization-deficient V16D/I55N Rev construct provides evidence that the ARM region constitutes an intrinsically disordered element in the absence of binding partners that can induce its α-helical conformation. The V16D/I55N Rev variant is predominantly monomeric, even at concentrations required for detailed NMR investigations, which facilitated the confirmation of the unstructured nature of Rev monomers with site-specific resolution.

Because nucleation of Rev assembly at the high-affinity StemIIB site involves a Rev monomer (10,30,78), we suggest that this initial interaction of the Rev ARM with RRE StemIIB proceeds via an induced-fit mechanism rather than by conformational selection. Upon encountering the unstructured ARM, StemIIB provides specific interactions that neutralize the positive charge of the numerous arginines and enable formation of the α-helical conformation. A combination of mutagenesis and CD studies previously established a correlation between specific StemIIB binding affinity and α-helical content (4). The R₄₂A variant showed increased helical content compared with the wild-type peptide, which in turn lowered the specific K_d as determined by gel-shift experiments from 40 to 35 nM. Combinatorial screening of arginine-rich peptide libraries identified that a glutamine (corresponding to Asn-40 of the Rev ARM) within a stretch of polyarginine specifically recognizes StemIIB (79). Further optimization flanked the glutamine at positions −6, −2, and +2 with nonarginine residues featuring aspartic acid, glutamic acid, and alanine at high frequencies. Since positions −6, −2, and +2 are located opposite of the glutamine within a putative α-helical conformation, it is unlikely that those nonarginine residues are specifically contacting the RRE major groove. Instead, charge neutralization and stabilization of the α-helix in general appear to be responsible for the up to 50-fold higher affinity for StemIIB of these polyarginine-derived peptides compared with the wild-type Rev ARM (80).

Conclusions

The combination of low hydrophobicity and large net charge under physiological conditions (pI_T34-R50 = 12.6) predicts intrinsic disorder for the Rev ARM peptide (81,82). Indeed, our results confirm that the 17-amino-acid, arginine-rich RNA-binding motif of Rev located within helix₂ of the N-terminal helix-loop-helix motif constitutes an intrinsically disordered sequence element within the isolated peptide and also in the context of the full-length, monomeric Rev protein. Our CD, NMR, and MD investigation of the Rev ARM peptide in aqueous solution at physiological pH reveals a statistical ensemble of interconverting structures lacking detectable α-helical subpopulations, suggesting that RRE StemIIB recognition by the Rev ARM occurs via an induced-fit (21) or coupled binding-folding mechanism (83,84). Our findings outline a principle that might constitute a general strategy for the ubiquitous class of arginine-rich RNA-binding proteins to recognize their diverse and often induced RNA-bound structures (85) while performing dual functional roles as nuclear localization signals (86–88).