Programmed Ribosomal Frameshifting in SIV is Induced by a Highly Structured RNA Stem-Loop

Abstract

Simian Immunodeficiency Virus (SIV), like its human homologues (HIV-1, HIV-2), requires a −1 translational frameshift event to properly synthesize all of the proteins required for viral replication. The frameshift mechanism is dependent upon a seven nucleotide slippery sequence and a downstream RNA structure. In SIV, the downstream RNA structure has been proposed to be either a stem-loop or a pseudoknot. Here, we report the functional, structural and thermodynamic characterization of the SIV frameshift site RNA. Translational frameshift assays indicate that a stem-loop structure is sufficient to promote efficient frameshifting in vitro. NMR and thermodynamic studies of SIV RNA constructs of varying length further support the absence of any pseudoknot interaction and indicate the presence of a stable stem-loop structure. We determined the structure of the SIV frameshift-inducing RNA by NMR. The structure reveals a highly ordered 12 nucleotide loop containing a sheared G-A pair, cross-strand adenine stacking, two G-C base-pairs, and a novel CCC triloop turn. The loop structure and its high thermostability preclude pseudoknot formation. Sequence conservation and modeling studies suggest that HIV-2 RNA forms the same structure. We conclude that, like the main sub-groups of HIV-1, SIV and HIV-2 utilize stable stem-loop structures to function as a thermodynamic barrier to translation, thereby inducing ribosomal pausing and frameshifting.

Introduction

Simian immunodeficiency virus (SIV), like human immunodeficiency virus (HIV), requires a programmed −1 translational frameshift between the gag and pol coding regions of the viral mRNA.¹ The −1 frameshift allows for stop codon read-through at the end of the gag gene, and continued synthesis of the protein encoded by pol. This results in production of a Gag-Pol polyprotein fusion product, which upon proteolysis in immature virus particles will result in the protease, reverse transcriptase, and integrase enzymes, as well as the matrix, capsid and nucleocapsid structural proteins. These structural and enzymatic proteins are found in approximately a 10:1 to 20:1 molar ratio in HIV type 1 (HIV-1), depending on the assay used.^2-8 This specific stoichiometry is required for appropriate packaging and formation of virus particles.^8-10

In retroviruses, the −1 translational frameshift is programmed by two cis-acting RNA elements: a seven nucleotide slippery sequence, and a downstream RNA structure.¹¹ In SIV, as in HIV-1,² the slippery sequence is UUUUUUA, and is universally conserved among all isolates analyzed (Los Alamos National Laboratories HIV Database, http://www.hiv.lanl.gov). In this work, we focus on a subtype of SIV (SIV_SMM) found in sooty mangabys (Cercocebus torquatus atys) and macaques (genus Macaca). This subtype is closely related to HIV type 2 (HIV-2),^12-14 while SIV from chimpanzees (Pan troglodytes) is more closely related to HIV-1.^12-15

Two models have been proposed for the SIV_SMM downstream RNA secondary structure, a stem-loop,² and an H-type pseudoknot^16,17 (Figure 1). The pseudoknot model is based on the fact that seven nucleotides in the loop are complimentary to a downstream sequence; however, due to the high degree of sequence conservation, there is no observable covariance between these sequences that could be used to help validate the pseudoknot model. No structural data exists for the SIV frameshift-inducing RNA.

Figure 1.

Two secondary structure models for the SIV RNA frameshift-inducing element. The seven nucleotide slippery sequence is underlined in each model. (A) Previously proposed stem-loop model.² (B) Proposed pseudoknot model.¹⁶,¹⁷ (C-E). RNA constructs studied by NMR. (C). SIV14−70, encompassing nucleotides 14−70 downstream of the slippery site. (D). SIV17−70. The first two C-G base-pairs were flipped to G-C pairs to facilitate in vitro transcription, and are shown in outline. Note that no base pairs are indicated within the pseudoknot portions of the secondary structures in (C) and (D), since the pseudoknot interaction has not been proven and is one of the subjects of this study. (E). SIV17−50. The first two base-pairs are as in (D).

In this study, our aims were to: 1) Determine the sequence requirement for frameshifting in SIV_SMM; 2) Determine the structure of the SIV_SMM frameshift-inducing RNA; 3) Analyze the thermodynamic properties of the structure, and 4) Compare the SIV_SMM structure to HIV-1^18-20 and HIV-2 to determine if any structural conservation exists.

Results

SIV sequence-induced frameshifting in vitro

The sequence requirement for SIV_SMM programmed translational frameshifting was investigated using a dual luciferase reporter assay.^21,22 Briefly, SIV_SMM sequences of varying length were cloned between Renilla and firefly luciferase reporter genes. In vitro translation of the frameshift reporter constructs was performed in rabbit reticulocyte lysate. Frameshift efficiencies were calculated by comparing the experimental and control levels of firefly luciferase activity, after normalizing their Renilla luciferase activities as previously described.^21,22

Results for all sequences tested showed frameshifting efficiencies from 7 to 12% (Figure 2). These frameshifting frequencies are very similar to those measured for HIV-1 in vitro.^2-8 A 78-nucleotide sequence comprising all viral sequences necessary to form the potential pseudoknot (Figure 1B) induced frameshifting to a level of 9.1 ± 0.7 % (Figure 2). There is no significant difference between this value and the frameshifting frequencies observed for shorter constructs that cannot form a pseudoknot (e.g. 57 nt, 8.8 ± 0.4 %). Sequences of 69, 63, 57, and 51-nucleotides induced frameshifting to 12.2%, 8.3%, 8.8%, and 7.1%, respectively (Figure 2). These values were determined from 10 replicate measurements obtained from the same preparations of RNA and lysate. However, higher variability (+/− 3%) in frameshifting is observed when comparing different preparations of lysate and RNA (data not shown). Regardless of the relatively small differences between the various constructs, the data indicate that pseudoknot formation is not a requirement for programmed translational frameshifting by SIV, since the 63, 57 and 51-nucleotide constructs cannot form pseudoknots but produce frameshifting efficiencies comparable to the 78-nucleotide construct. The increase in frameshift efficiency for the 69-nucleotide sequence may be due to the variability of the assay system, or it may possibly be more optimally folded in the context of the reporter gene.

Figure 2.

Results of the in vitro frameshifting assay using dual-luciferase reporters.²¹ The length of the SIV sequences analyzed is indicated and compared to the frameshifting efficiency, as determined by an average of at 10 independent experiments, with the average value and standard error mean indicated above each column.

NMR Analysis of RNA folding

The stem-loop and pseudoknot models for the SIV_SMM RNA were further investigated by NMR using three RNA constructs: 1. A 57 nucleotide sequence spanning nucleotides 14−70 (SIV14−70), which includes all nucleotides downstream of the slippery site involved in the proposed pseudoknot (Figure 1C). 2. A 54 nucleotide sequence that encompasses nucleotides 17−70 (SIV17−70) (Figure 1D). 3. A 34 nucleotide sequence that spans nucleotides 17−50 (SIV17−50) that can only form a stem-loop structure (Figure 1E). For constructs SIV17−70 and SIV17−50, the tandem C-G pairs at positions 17 and 18 where inverted to G-C pairs to facilitate transcription. Inversion of these C-G pairs had no affect on the folding of the RNA (vide infra, and Figure 3).

Figure 3.

Overlay of ¹H-¹H TOCSY spectra obtained for SIV17−50 (red) and SIV17−70 (black) and SIV14−70 (green) at 750 MHz and 35°C in 10 mM sodium phosphate buffer (pH 7.0). Peaks corresponding to the loop cytidine peaks are numbered (32−35, 37).

The chemical shifts in the RNA loop and proximal stem are essentially identical for all constructs (Figure 3). In particular, the four contiguous cytidines at the apex of the loop (C32−35), which would be involved in the potential pseudoknot interaction (if present), overlay well. There are no changes in chemical shifts for C33−35 and C37 in the various constructs. Nucleotide C32 exhibits very minor chemical shift changes, but these are likely due to differences between sample preparations, since the chemical shifts for this nucleotide vary slightly even between different preparations of the same RNA sequence (data not shown). Other differences in the NMR spectra of the various constructs are solely due to the different lengths and helical sequences of the various constructs. Additionally, 2D NOESY experiments reveal the presence of the same NOEs in the loop for all constructs (data not shown). All of the imino protons and their hydrogen-bond acceptors involved in Watson-Crick base pairing have been assigned for SIV17−50 (Figure 1E), which cannot form a pseudoknot (Figure 4). The same imino proton chemical shifts are observed for the two longer constructs (Figure 1C and D), and upon addition of 10 mM magnesium, no new imino peaks indicative of new Watson-Crick base pairs are observed (data not shown). These results indicate that the RNA adopts the same stem-loop structure in all constructs regardless of ionic conditions. Therefore, the in vitro translation data (Figure 2) and the NMR data (Figure 3) suggest that the SIV_SMM RNA structure is a stem-loop, and not a pseudoknot.

Figure 4.

Direct detection of base-pairs in the SIV17−50 frameshift-inducing stem-loop. (A) 750 MHz 1D ¹H NMR spectrum of the imino protons. Assignments are indicated by numbers above the peaks. (B) 800 MHz 2D HNN-COSY spectrum for direct detection of trans-hydrogen-bond scalar couplings. Two G-C pairs in the loop are indicated by dashed lines.

Structure and thermodynamics of the SIV RNA stem-loop

We further investigated the structure of the SIV17−50 RNA by NMR. Thirteen Watson-Crick base-pairs were confirmed by direct detection of trans-hydrogen-bond scalar couplings by HNN-COSY (Figure 4). Eleven of the Watson-Crick base-pairs correspond to the stem, while two additional G-C pairs were observed to form across the 12 nucleotide loop, between nucleotides G31-C37 and C32-G36. These assignments were confirmed by thermodynamic measurements and 1D NMR analysis of constructs with mutations of these loop nucleotides. Two cooperative melting transitions are observed: a lower-temperature transition at 66°C and a higher-temperature transition at 84°C (Figure 5A). These transitions can be assigned to the loop and stem regions of the molecule, respectively, by reference to experiments conducted on sequences that contain guanosine-to-cytidine mutations in the loop (Figure 5B), and also 1D NMR measurements as a function of temperature (data not shown). These data indicate that the stem is very stable, and the loop has a surprising degree of stability for a structure which is penalized 7.1 kcal/mole by mfold secondary structure prediction software.²³ The 66°C melting transition for the loop is also observed for the longer SIV14−70 construct (Figure 1C and Supplemental data).

Figure 5.

RNA thermal denaturation profiles for SIV17−50. (A) First derivative plot of change in UV absorbance vs. temperature (red), with fitted curves showing the deconvoluted transitions as dashed lines. The T_m values are shown above each peak. (B) First derivative plot of change in UV absorbance vs. temperature of SIV17−50 (red) compared to loop sequence mutants G31C (green), G36C (blue) and G31C, G36C (magenta).

The hydrogen-bonding data (Figure 4) and thermodynamic measurements (Figure 5) provide evidence for a highly structured loop conformation. NOESY data also provide further evidence for an unusually structured loop. A strong intranucleotide H6-H1' NOE and a relatively weak intranucleotide H6-H2' NOE for C35 indicate that the base predominately adopts the syn conformation (Supplemental data).²⁴ Additionally, cross-strand internucleotide NOEs are observed between A29 and A39 (data not shown).

Structures of the SIV frameshift-inducing stem-loop were calculated using 647 NOE-derived distance restraints, backbone torsion angle restraints for the helical regions, hydrogen-bond restraints derived from scalar couplings, and 28 RDC restraints (Figure 4 and Table 1). The 20 lowest energy structures superimpose over all heavy atoms to an r.m.s.d. of 1.29 Å (Figure 6A and Table 1). The loop consists of a “neck” region (Figure 6, red nucleotides), and a “head” region (Figure 6, green nucleotides). The neck region is comprised of a sheared G-A pair (G28-A39) followed by a cross-strand adenine stack (A29 on A39), with A30 flipped out into the minor groove. The head region is comprised of two G-C pairs across the loop (G31-C37 and C32-G36) which close a compact three cytidine triloop (C33,C34,C35). The 12-nucleotide loop structure is well defined, and converges to an r.m.s.d. of 1.13 Å (Figure 6B and Table 1), though the A-form helical stem is more precisely defined as is usually the case for RNA solution structures (Figure 6C and Table 1).

Table 1.

NOE-derived distance restraints	647
Intranucleotide	253
Internucleotide	394
Dihedral restraints	219
Hydrogen-bond restraints	36
RDC restraints	28
r.m.s.d. (all heavy atoms)	1.29 Å
over stem (residues 17−27, 40−50)	0.61 Å
over loop (residues 28−39)	1.13 Å
Average NOE r.m.s.d. (Å)	0.018
Average dihedral r.m.s.d. (o)	0.40
Average RDC r.m.s.d. (Hz)	1.46

Figure 6.

NMR structures of the SIV17−50 frameshift-inducing RNA stem-loop. (A) The 20 lowest energy structures superimposed over all heavy atoms. The A-form stem is shown in blue, the purine loop nucleotides 28−30 and 38−39 are shown are in red, and the remaining loop nucleotides 31−37 are shown in green. (B) Structures superimposed over all nucleotides in the 12-nucleotide loop (G28-A39). (C) Structures superimposed over all nucleotides in the stem (nucleotides 17−27 and 40−50). (D) Stereo view of the lowest energy structure.

Discussion

In this study, we have shown that the −1 translational frameshift in SIV is programmed by a thermodynamically stable RNA stem-loop. Analysis of frameshift efficiencies of various constructs suggests that pseudoknot formation does not play a significant role in translational frameshifting, at least in vitro. Furthermore, we observe that the SIV loop is structured in a manner that precludes pseudoknot formation, since two of the nucleotides that would participate in such a pseudoknot are sequestered within stable Watson-Crick base-pairings across the loop. Hence, the 12 nucleotide loop is reduced to a compact CCC triloop conformation. To the best of our knowledge, the CCC triloop is a novel turn conformation. The triloop conformation consists of a base-stack of C33 on C32, with C34 flipped into the minor groove, and C35 adopting a syn-conformation (Figure 7A). The syn conformation of C35 is consistent with the observed intensity of the NOE between its H1' and H6 protons (Supplemental Data). The unique CCC triloop structure may be stabilized by a potential hydrogen-bond between the carboxyl oxygen of C34 and the ribose 2'OH of C37.

Figure 7.

(A) Representative triloop structure with closing G-C pairs. (B) Overlay of the tandem G-A pairs from the 23S rRNA (white) and the G-A sheared pair with cross-stacked adenosine from SIV (red). The bulged residue A30 is not shown for clarity.

The loop contains a modular structural element, the cross-strand adenine stack upon a sheared G-A pair. NOEs that help define the sheared G-A pair are shown (Supplemental Data). A search of RNA structural motifs using the SCOR 2.0²⁵ database revealed that G28, A29, A38, and A39 adopt a conformation that is nearly identical to the tandem G-A base-pairs found in D. radiodurans 23S rRNA²⁶ (PDB code 1NKW). The D. rad. structure (white) and the SIV34 structure (red) overlay with an r.m.s.d. of 1.38 Å over all carbon atoms (Figure 7B). While there is no evidence for base-pairing of A29 and A38 in the SIV34 structure, as is seen between A2331 and G2344 in the 23S rRNA, the sequence identity and structural similarities between the two cross-strand stacking motifs are clear.

Why doesn't the SIV_SMM RNA structure form a pseudoknot? Pseudoknot formation as depicted in the secondary structure model in Figure 1B is precluded by the length of the single-stranded linker region between the helical end and the loop, which is only 8 nucleotides. The SIV_SMM stem-loop structure alone is 52 Å in length, and spans more than a full helical turn. Since the secondary structure in Figure 1B includes an additional lower helix (separated by a C-C mismatch of unknown geometry), it is likely that the 8 nucleotide linker would have to span a distance of > 60 Å, which is physically impossible. The two constructs that we have studied that could potentially form pseudoknots have single stranded linkers that are 10 and 13 nucleotides long (Figure 1C and D, respectively). These linkers are long enough to span the length of the helical turn. Since a translating ribosome paused at the slippery site would be already engaged with the mRNA approximately 13 nucleotides downstream of the first base in the P site (spanning the RNA from U1 to U13 in Figure 1A),^27,28 the 4 base-pair helix at the base of the stem would be expected to be at least partially unwound prior to the frameshift event. This condition is mimicked in the SIV14−70 and SIV17−70 constructs (Figures 1C and D), yet these constructs do not form detectable pseudoknot interactions. Thus, we conclude that the loop structure precludes pseudoknot formation by sequestering its nucleotides within a stable (Tm = 66°C) structure. Interestingly, a recently developed RNA secondary structure prediction algorithm capable of predicting pseudoknots²⁹ predicts that the stem-loop structure is more energetically favorable than the pseudoknot fold, and also predicts that the linker is capable of folding into an alternative three base-pair stem-loop structure that is incompatible with pseudoknot formation (data not shown).

SIV_SMM frameshift inducing structure is likely similar to HIV-2, but differs from HIV-1

A comparison of SIV_SMM, HIV-2 and HIV-1 sequences (www.hiv.lanl.gov) revealed that SIV_SMM and HIV-2 have nearly identical sequences for the frameshifting element, but they are both substantially different from the HIV-1 sequence (Figure 8A). Additionally, when the differences between SIV_SMM and HIV-2 are mapped to the SIV secondary structure, the only significant difference is observed as a single cytidine-to-uridine transition in the loop (Figure 8B), as all other sequence changes preserve the same base-pairing shown in Figure 1A (data not shown). Modeling and energy minimization of the HIV-2 loop structure from SIV starting structures revealed that the HIV-2 loop can adopt an identical structure, and may be further stabilized by an additional hydrogen-bond between the O4 of U34 and the 2'OH group of C32, as compared to the SIV loop (Figure 8C).

Figure 8.

Sequence comparison and analysis of SIV, HIV-2 and HIV-1 frameshift-inducing sequences. (A) Primary sequence comparison of the three viruses. Differences between the SIV and HIV-2 sequences are indicated by red asterisks. Sequence logos were created using WebLogo⁵⁶ (http://weblogo.berkeley.edu/logo.cgi). (B) Secondary structures of SIV/HIV-2 and HIV-1. Differences between SIV and HIV-2 are indicated by nucleotides in red. HIV-1 structure is shown for comparison. (C) Structure of the CCC (SIV) triloop and model of the CUC (HIV-2) triloop with closing G-C pairs. Potential hydrogen-bonds are shown in yellow.

The frameshift site RNA structure is coupled to the slippery sequence

Most viral RNAs use pseudoknot structures to promote framehshifting. It has been proposed that pseudoknots promote frameshifting through their unique folding topologies, which may be poor substrates for the ribosomal helicase.^30-34 In addition, both pseudoknots and stable stem-loops are capable of pausing ribosomes at frameshift slippery sites.^35,36 A kinetic pause likely contributes to frameshifting^36-38; however, previous studies have found no simple correlation between pausing and frameshifting.³⁵ HIV-1, SIV and HIV-2 are unique in that they utilize highly stable stem-loops for frameshifting.^2,7,19,20 We hypothesize that the nature of the frameshift site RNA structure is dependent upon the slippery site to which it is coupled.

All retroviral slippery sites have an XXXYYYZ consensus sequence.³⁹ Previous data indicate that the UUUUUUA slippery sequence used by HIV-1, HIV-2 and SIV is one of the most efficient in promoting programmed −1 translational frameshifting, suggesting that it is more “slippery” than other XXXYYYZ sequences.⁴⁰ HIV-1, HIV-2 and SIV may only require a thermodynamically stable stem-loop, rather than a pseudoknot, due to the inherent “slipperiness” of the UUUUUUA sequence. For these viruses, the stable stem-loops are sufficient to pause ribosomes and induce the appropriate level of frameshifting. Other viruses, such as the mouse mammary tumor virus (MMTV)^32,41 or infectious bronchitis virus (IBV)^35,36 may require pseudoknot structures to attain the appropriate level of frameshifting because their slippery sequences (AAAAAAC for MMTV and UUUAAAC for IBV) are not as slippery.⁴² In these cases, pausing alone may not be sufficient, and pseudoknots may be required because they are stereochemically mismatched substrates for the helicase.^43,44

Materials and Methods

RNA synthesis and purification

Milligram quantities of RNA suitable for NMR and UV spectroscopy methods were transcribed in vitro using purified His₆-tagged T7 RNA polymerase, synthetic DNA oligonucleotides (Integrated DNA Technologies, Inc.), and NTPs (Sigma-Aldrich). The first two base pairs of the SIV17−50 sequence were flipped relative to wild-type to optimize transcription yields. RNA was purified by denaturing 15% (SIV17−50 RNA) or 8% (SIV17−70 and 14−70 RNA) polyacrylamide gel electrophoresis, identified by UV absorbance, and excised from the gel. RNA was recovered by diffusion into 0.3 M sodium acetate (pH 5.0), precipitated with ethanol, purified on a 6 mL DEAE anion exchange column, again precipitated with ethanol, and desalted on a 15 mL G-25 gel filtration column. The purified RNA was lyophilized, resuspended in water, and brought to pH 7.0 by the addition of 1 M NaOH or by buffer exchange into 10 mM sodium phosphate (pH 7.0). ¹³C/¹⁵N-labelled RNA samples for NMR were prepared using ¹³C/¹⁵N-labelled NTPs (Silantes GmbH, München, Germany). Fully ¹³C/¹⁵N-labelled adenosine/uridine and guanosine/cytidine samples were prepared. A shorter RNA construct of 24 nucleotides (SIV24; corresponding to G22 through C45, Figure 1) was also prepared using NTPs which were specifically deuterated at all positions except the ribose H1'and H2' and base H6/H8 atoms (Cambridge Isotope Laboratories, Inc.). Partial alignment of RNA for RDC measurements was achieved by adding 17 mg/mL of Pf1 filamentous bacteriophage (ASLA Ltd., Riga, Latvia) to the ¹³C/¹⁵N-labelled samples.

In vitro frameshifting assay

Plasmid construction

The sequences investigated here are derived from the most commonly occurring sequence of all SIV_SMM RNA isolates analyzed (www.hiv.lanl.gov) and represent the consensus sequence. For each sequence tested, complementary synthetic oligonucleotides (Integrated DNA Technologies, Inc.) with BamH I and Sac I compatible ends were cloned into the p2luc plasmid using the unique BamH I and Sac I sites between the rluc and fluc reporter genes. Oligonucleotidess SIV51 (5'-GATCCTTTTTTAGGCCTTGGTCCATGGGGAAAGAAGCCCCGCAATTTCCCATGGCGAGCT-3'), SIV57 (5'-GATCCTTTTTTAGGCCTTGGTCCATGGGGAAAGAAGCCCCGCAATTTCCCCATGGCTCAAGTGCAGCT-3'), SIV63 (5'-GATCCTTTTTTAGGCCTTGGTCCATGGGGAAAGAAGCCCCGCAATTTCCCCATGGCTCAAGTGCATCAGAGCT-3'), SIV69 (5'-GATCCTTTTTTAGGCCTTGGTCCATGGGGAAAGAAGCCCCGCAATTTCCCCATGGCTCAAGTGCATCAGGGGCTGAGCT-3'), and SIV78 (5'-GATCCTTTTTTAGGCCTTGGTCCATGGGGAAAGAAGCCCCGCAATTTCCCCATGGCTCAAGTGCATCAGGGGCTGATGCCAACGAGCT-3') and their complements were phosphorylated, annealed, and ligated into the p2luc vector to produce the experimental constructs. Positive control sequences and their complements were also cloned into p2luc and have two thymidine residues (bold) in the slippery sequence (underlined) replaced with cytidines and an additional nucleotide inserted immediately before the Sac I complementary sequence (GAGCT). Resultant products were transformed into E. coli competent cells (JM109, Promega). Plasmid DNA was purified from cell cultures (Qiagen) and the sequences of all constructs were verified (University of Wisconsin Biotechnology Center).

In vitro transcription

Purified plasmid DNA was linearized by digestion with Pml I prior to in vitro transcription. RNA (∼3 kb) was transcribed using the T7 RiboMAX Express Large Scale RNA Production System (Promega), the DNA template was removed by treatment with RQ1 RNase-free DNase, and the mixture was extracted with phenol:cholorform 5:1 (pH 4.7) and chloroform:isoamyl alcohol 24:1 (Sigma-Aldrich). The aqueous phase was run on a 5 mL P6 gel filtration column (Bio-Rad) to remove small abortive products and unincorporated NTPs. The purified RNA was lyophilized, resuspended in water, analyzed by agarose (1%) gel electrophoresis, and stored at −80 °C.

In vitro translation

RNA samples were translated in vitro using the Flexi Rabbit Reticulocyte Lysate System (Promega). Translation reactions contained 1 μg of RNA per 10 μL of reaction mixture, 1−2 mM Mg²⁺, 70 mM KCl, and 20 μM of each amino acid except methionine and leucine (10 μM each). Reactions were incubated at 30°C for 90 minutes.

Frameshift efficiency measurements

Luminescence was measured using a Veritas microplate luminometer equipped with dual-injectors (Turner Biosystems). For each sequence length investigated, 10 replicates of the experimental sequence and 5 replicates of the positive control sequence were examined using the Dual-Luciferase Reporter Assay System (Promega). Luminescence readings were taken for both reporters 10 sec after their respective substrates were injected into the reaction mixture (2 sec lag time prior to measurement). Ratios of firefly/Renilla luminescence were calculated for each of the experimental and control sequences, and replicates were averaged. The ratio of averages (experimental/control) was then determined, which corresponds to the frameshift efficiency (Figure 2). Measurement error was propagated through these calculations to yield a final standard error mean (SEM) value for each of the determined frameshift efficiencies (Figure 2).

UV Spectroscopy

Thermal stability studies were conducted on purified RNA constructs using a Cary Model 100 Bio UV-visible spectrophotometer equipped with a Peltier heating accessory and temperature probe. All samples contained 10 mM sodium phosphate buffer (pH 7.0), 200 mM KCl, and approximately 2 μM RNA. Samples were slowly heated from 20°C to 95°C at a rate of 1°C/min and absorbance data were collected at 260 nm in 1°C increments. Thermal denaturation was determined to be performed under equilibrium conditions, as evidenced by the fact that no hysteresis was observed and similar data were obtained when samples were slowly cooled from 95°C to 20°C. A minimum of three scans were taken for each sample. The data were derivatized, linear baseline corrected, smoothed over a 3°C window, normalized (with the largest value set to 1), and fit using nonlinear least-square fit to solve for T_m and ΔH as previously described.⁴⁵ Data analysis was performed using Prism 4.0 (GraphPad).

NMR Spectroscopy

All NMR spectra were obtained on Varian Inova (800 and 900 MHz) and Bruker Avance DMX (500, 600 and 750 MHz) spectrometers at the National Magnetic Resonance Facility at Madison (NMRFAM). Spectrometers were equipped with either a conventional proton, carbon, nitrogen (HCN) room temperature probe or a single z-axis gradient HCN cryogenically cooled probe. Exchangeable resonances were assigned by reference to 2D NOESY (150 msec mixing time), ¹H-¹⁵N HMQC and 2D HNN-COSY spectra of the RNA collected in 90% H₂O/10% D₂O, pH 7.0 at 283 K. Water suppression for these experiments was achieved by using a 1−1 spin echo or watergate pulse sequence. Non-exchangeable resonances were assigned by reference to 2D NOESY (50, 100, 150, 200, 250, and 300 msec mixing times), 2D ¹H-¹H TOCSY, 2D ¹H-¹³C HSQC, 3D ¹H-¹³C-¹H HCCH TOCSY, and 3D ¹H-¹³C-¹H HCCH COSY spectra of the RNA collected in 99.99% D₂O, pH 7.0 at 308 K.

Hydrogen-bonds were detected for adenosine/uridine and guanosine/cytidine ¹³C/¹⁵N-labeled samples in 90% H₂O/10% D₂O, pH 7.0 at 283 K.⁴⁶ NOESY experiments on the specifically deuterated SIV17−50 RNA were performed to aid in assignment of highly overlapped resonances.

Residual dipolar couplings (RDCs) were measured using J-modulated ¹H-¹³C CT-HSQC experiments (18 planes) in both isotropic and partially oriented ¹³C/¹⁵N-labeled RNA samples in 90% H₂O/10% D₂O at 308 K. Peak intensities from each plane were fit to yield the ¹J_C-H coupling values, and RDCs were calculated as the difference between ¹J_C-H values for isotropic and partially aligned samples.⁴⁷ PALES software (http://spin.niddk.nih.gov/bax/software/PALES)⁴⁸ was used to estimate the values for the axial (D_a) and rhombic (R) components of the alignment tensor from converged, low energy structures calculated using XPLOR-NIH⁴⁹ in the absence of RDCs. A subsequent grid search procedure⁵⁰ performed around the estimated D_a and R was used to optimize the values, using lowest overall energies as a target. Optimal D_a/R values were determined to be −57/.23 for the stem alignment tensor. RDC values for the loop could not be fit to the same alignment tensor as the stem, and were therefore not used for final structure calculations.

Thermal unfolding of the SIV17−50 RNA (0.8 mM, 90% H₂O/10% D₂O, pH 7.0) was monitored by collecting a series of 1D ¹H spectra at increasing temperature. The temperature was increased from 5 °C to 80 °C in 5 °C increments. All spectra were collected with 32 scans. All data were processed using XwinNMR (Bruker) or NMRPipe⁵¹ software, and assignments were made using Sparky (http://www.cgl.ucsf.edu/home/sparky/).

Structure Calculations

NOE distance restraints were categorized as very strong (1.8−2.5 Å), strong (2.0−4.0 Å), medium (2.5−4.5 Å), or weak (3.0−7.0 Å) based on the integrated peak volumes obtained from 2D NOESY spectra (50, 150, 200, 250, and 300 msec mixing times). Torsion angle restraints (angles: α, β, γ, δ, ε, ζ, η, ν₁, ν₂ and χ) for residues in the stem (G17-A27, U40-C50) were constrained to A-form values⁵² (±15°), which were consistent with NOESY, TOCSY, HNN-COSY, and RDC data. Additionally, A-form phosphate backbone restraints for these regions were incorporated as previously described.⁵³ Torsion angles for residues in the loop (G28-A39) were left unrestrained, with the exception of experimentally validated glycosidic (χ) and sugar pucker angles (ν₁, ν₂). Residues with strong H1'-H2' and H1'-H3' cross-peaks (C33, C34, C35) in 2D ¹H-¹H TOCSY (40 msec mixing time) spectra were restrained to the C2'-endo range (ν₁: 25 ± 15°, ν₂: −35 ± 15°), residues without H1'-H2' cross-peaks (G28,A29,A30,G31,C37,A39) were restrained to the C3'-endo range (ν₁: −25 ± 15°, ν₂: 37 ± 15°), and residues which displayed an intermediate H1'-H2' cross-peak (C32, G36, A38, C50), indicative of an averaging of sugar puckers, were left unrestrained. Analysis of the peak volumes of internucleotide and intranucleotide H1'-H6/H8 and H2'-H6/H8 NOEs in 2D NOESY (50 msec mixing time) revealed that C35 was in the syn range; therefore, χ was restrained to 30 ± 40°, as previously observed for a syn-cytidine.²⁴ All other residues in the loop were determined to be in the anti range, and were loosely restrained relative to their sugar pucker conformation (χ_C3'-endo: −160 ± 30°, χ_C2'-endo: −125 ± 30°, all other χ angles: −140 ± 35°). Hydrogen-bond distance restraints were used for all base-pairs identified by NOESY and HNN-COSY experiments, and very weak planarity restraints (1 kcal mol⁻¹ Å⁻²) were enforced during structure calculations.

An extended structure generated in CNS 1.1⁵⁴ was used to calculate 100 starting structures from random initial velocities using distance (NOE), dihedral, hydrogen bonding, and planarity restraints. Structures were subjected to 60 psec (15 fsec time steps) of restrained molecular dynamics in torsion angle space, 90 psec of slow cooling, and 30 psec (5 fsec time steps) of restrained molecular dynamics in Cartesian coordinate space. These structures were then refined with additional RDC restraints using XPLOR-NIH.⁴⁹ Structures were heated to 3000 K and cooled to 100 K (50 K steps) for 28 psec of restrained molecular dynamics in Cartesian coordinate space, followed by 500 steps of energy minimization using the Powell algorithm and simulated annealing.

Structures were accepted based on criteria of lowest overall energy and lack of NOE violations >0.5 Å and dihedral violations >5°. Structures were viewed and analyzed using MOLMOL⁵⁵ and figures were created using PyMol (http://www.pymol.org). The final structures were used to back-calculate RDC values using PALES,⁴⁸ and the r.m.s.d. for observed versus back-calculated RDCs was found to be 1.46 Hz (Table 1).

Modeling the HIV-2 Structure

Modeling of the HIV-2 loop structure was performed using Sybyl 7.3 (Tripos, Inc.). SIV starting structures were modified by replacement of the C34 base with a uracil. Charges were added by the Gasteiger-Huckel method (Tripos), followed by 500 steps of energy minimization using the Powell algorithm and the Tripos force field. Charge addition and energy minimization were also performed on SIV starting structures for comparison. Resulting structures were analyzed using Pymol (http://www.pymol.org), and potential hydrogen-bonds (within 2.0 Å) are displayed (Figure 8).

Data Bank accession codes

Coordinates for the SIV frameshift-inducing RNA have been deposited into the RCSB Protein Data Bank (accession code 2JTP). NMR chemical shift assignments and 2D NOESY time domain data have been deposited into BioMagResBank (accession code 15417).