Structural characterization of an intermolecular RNA–RNA interaction i
http://www.100md.com
《核酸研究医学期刊》
Department of Plant Pathology, Box 7616 and 1 Department of Molecular and Structural Biochemistry, Box 7622, North Carolina State University, Raleigh, NC 27695, USA
*To whom correspondence should be addressed. Tel: +1 919 515 6992; Fax: +1 919 515 7716; Email: steve_lommel@ncsu.edu
Present addresses:
Michael A. Dolan, Tripos Inc., St Louis, MO 63144, USA
Guihua Liu, Department of Chemistry, China Medical University, Shenyang, China
ABSTRACT
The 34-nucleotide trans-activator (TA) located within the RNA-2 of Red clover necrotic mosaic virus folds into a simple hairpin. The eight-nucleotide TA loop base pairs with eight complementary nucleotides in the TA binding sequence (TABS) of the capsid protein subgenomic promoter on RNA-1 and trans-activates subgenomic RNA synthesis. Short synthetic oligoribonucleotide mimics of the RNA-1 TABS and the RNA-2 TA form a weak 1:1 bimolecular complex in vitro with a Ka of 5.3 x 104 M–1. Ka determination for a series of RNA-1 and RNA-2 mimic variants indicated optimum stability is obtained with seven-base complementarity. Thermal denaturation and NMR show that the RNA-1 TABS 8mers are weakly ordered in solution while RNA-2 TA oligomers form the predicted hairpin. NMR diffusion studies confirmed RNA-1 and RNA-2 oligomer complex formation in vitro. MC-Sym generated structural models suggest that the bimolecular complex is composed of two stacked helices, one being the stem of the RNA-2 TA hairpin and the other formed by the intermolecular base pairing between RNA-1 and RNA-2. The RCNMV TA structural model is similar to those for the Simian retrovirus frameshifting element and the Human immunodeficiency virus-1 dimerization kissing hairpins, suggesting a conservation of form and function.
INTRODUCTION
RNA viruses employ a diverse array of RNA motifs for the regulation of gene expression through interactions with other nucleic acids and proteins. The genomes of positive (+) stranded RNA viruses behave as mRNAs within host cells to allow the expression of viral proteins. However, many viral genomes are polycistronic and the expression of inaccessible open reading frames (ORFs) is facilitated by multiple strategies (1,2). Subgenomic RNA (sgRNA) synthesis is a common mechanism for the generation of viral RNAs that behave as monocistronic mRNAs for internal and 3' co-terminal ORFs (3). Essentially, there are two proposed mechanisms for the generation of (+) stranded sgRNAs by the virally encoded RNA polymerase (i) internal initiation of transcription from the full length minus (–) strand copy of the genomic RNA (4) and (ii) premature termination during (–) strand synthesis followed by independent replication of this truncated RNA template (5). For both models, the presence of a distinct subgenomic promoter element on the (–) strand or a terminator on the (+) strand, is required for the initiation of transcription (1).
Recently, it has been determined that for several RNA viruses, RNA synthesis requires long-distance cis or trans RNA–RNA interactions (6). It is suggested that viruses that utilize these interactions generate sgRNA via the premature termination mechanism. The base pairing of upstream cis-acting elements to the subgenomic promoter was found to be essential for sgRNA synthesis in Potato virus X, Tomato bushy stunt virus (TBSV) and Flock house virus (7–9). The situation is taken to its extreme for the bipartite genome of Red clover necrotic mosaic virus (Fig. 1A) (RCNMV; genus Dianthovirus, family Tombusviridae) where expression of the sgRNA from RNA-1 requires the base pairing of a sequence element in trans in RNA-2 (10). The 34-nucleotide trans-activator (TA) element from RNA-2 is predicted to form a stable hairpin structure with an eight-nucleotide loop possessing complementarity to an eight-nucleotide element immediately upstream of the sgRNA transcription start site on RNA-1, referred to as the trans-activator binding site (TABS) (Fig. 1B). Mutagenesis demonstrated that complementary base pairing between the two elements is essential for sgRNA transcription but the nature of this RNA–RNA tertiary structure has not been elucidated.
Figure 1. Genome organization, expression strategy and model of RNA-1–RNA-2 interaction facilitating dianthovirus RNA-1 sgRNA synthesis. (A) The bipartite RNA genome of Red clover necrotic mosaic virus (RCNMV) is depicted with open rectangles representing ORFs. RNA-1 ORFs p27 and p88 are involved in viral replication while the capsid protein (CP) is the virion structural protein. RNA-2 encodes the movement protein (MP) required for cell-to-cell movement of the viral RNA. p88 is expressed by a –1 ribosomal frameshifting event (–1 FS) of p27. The CP is expressed from a sgRNA generated during the replication of the virus. The RNA-1 sequence near the sgRNA transcription start site (denoted with an arrow) is shown aligned with comparable regions from Sweet clover necrotic mosaic virus (SCNMV) and Carnation ringspot virus (CRSV), the two other species comprising the Dianthovirus genus. The RNA-2 sequence of the TA region is shown below RNA-2 with the corresponding aligned regions from SCNMV and CRSV. The shaded sequences represent the complementary regions between RNA-1 and RNA-2 involved in trans-activation of transcription. The nucleotide sequences for the RCNMV oligomers used in this study are depicted below their respective genomic sequences. The nucleotide positions of the various sequences are noted. (B) Model of interaction previously proposed (10).
Determination of the tertiary structures of RNAs, particularly RNA–RNA complexes is difficult. Only a few examples of high-resolution non-duplex bimolecular RNA–RNA structures such as the Human immunodeficiency virus-1 (HIV-1) kissing hairpins and dimerization signal (11,12) and interactions within the ribosome complex (13) have been elucidated and provide limited examples of structural variation. While bimolecular RNA interactions generally involve loop–loop interactions (14), it is likely that a range of structural diversity exists in bimolecular RNA–RNA complexes. Many structures may prove to be similar to unimolecular tertiary interactions, particularly those of similar function. Here we report direct biophysical evidence for the formation of a weak, yet stable bimolecular complex between the RCNMV RNA-2 TA element and the RNA-1 subgenomic promoter. We further propose structural models for this interaction that exhibit conservation in both structure and function with other viral cis and trans RNA interactions.
MATERIALS AND METHODS
RNA oligomer preparation
Synthetic oligomers used in the study were obtained from the North Carolina State University Nucleic Acids Facility and from Dharmacon, Inc. (Lafayette, CO). Isotopically labeled RNA for the NMR studies was obtained by T7 transcription of a plasmid template containing the 21-nucleotide RCNMV RNA-2 sequence (Fig. 1A) cloned into pUC18 and terminated at the XbaI site. The transcription reaction was performed at 37°C for 8 h using a T7-MEGAshortscript High Yield Transcription Kit (Ambion, Austin, TX). For the isotopic labeling, 80 μM uniformly labeled 13C and 15N GTP sodium salt (Isotec, Miamisburg, OH) was substituted for the kit’s 75 mM GTP solution. Each oligomer was purified by anion exchange HPLC using a Machery-Nagel 250/10 Nucleogen DEAE 60-7 column (Duren, Germany). The purity of each oligomer was confirmed by denaturing polyacrylamide gel electrophoresis (PAGE). The RNA-1 oligomers representing the TABS are designated R1 with a superscript denoting the length, e.g. R18 is the eight-nucleotide oligomer (nucleotides 2356–2363) used in preliminary NMR studies. Four R18 oligomers, used in PAGE and NMR experiments have an additional + or – superscript denoting the nucleotide shift from the central sequence (nucleotides 2356–2363). The RNA-2 oligomer is similarly designated R220 (nucleotides 761–780) while mutant versions are identified by a superscript denoting mutations in the U:G base pair. The RNA-2 transcript was designated R221*. The sequences of the oligomers are shown in Figure 1A and Table 1.
Table 1. The apparent complex association constants, Ka, for a series of RCNMV RNA-1 oligomers binding to R220
Thermal denaturation
Thermal denaturation of the oligomers was monitored by UV absorbance (at 260 nm) over a temperature range of 5–90°C (1°C min–1) using a Cary 3 Spectrophotometer (Varian, Palo Alto, CA). The denaturation profile of each oligomer was conducted in triplicate at three concentrations, 0.5, 5 and 50 μM in 10 mM Na phosphate, 100 mM NaCl, pH 7.0. Both 2 and 10 mm cuvettes were used in order to maintain proper optical response. Variation of Tm was less than 1°C for all concentrations. Samples were thermally denatured and then renatured to determine transition reversibility. In all cases, a difference of less than 1°C in Tm was observed between denaturation and renaturation with no hysteresis detected.
Polyacrylamide gel electrophoresis assay
Denaturing PAGE (7 M urea, 9 mM Tris–borate, 0.25 mM EDTA in 20% polyacrylamide) was conducted to assess the purity of the synthetic oligomers and to analyze the components of the bimolecular complex. Non-denaturing PAGE (9 mM Tris–borate in 20% polyacrylamide run on ice) was used to investigate and characterize parameters affecting complex formation. After electrophoresis, gels were stained with ethidium bromide (EtBr), visualized by UV transillumination and images recorded on a Bio-Rad Gel Doc 2000 imager (Bio-Rad Laboratories, Hercules, CA). The intensities of the RNA complex bands were quantified using Bio-Rad Quantity One software. Over the concentration ranges used, it was determined that EtBr intercalation reported a linear measure of RNA concentration. Apparent association constants (Ka values) were calculated based on one site binding with the aid of PRISM software (GraphPad Software, San Diego, CA).
NMR analysis
Samples were prepared by solubilizing purified RNA in 0.2 ml of 90% H2O/10% D2O phosphate buffer to give 0.5 mM RNA. Samples were equilibrated with multiple buffer exchanges in Centricon spin columns (Millipore, Milford, MA) and placed in NMR microtubes (Shigemi, Allison Park, PA, stock # BMS005V). The NMR spectra for the 1:1 complex of R18+1:R221* were also collected at 25°C in 100% D2O phosphate buffer . Bruker AVANCE 500 MHz spectrometer with an Oxford Narrow Bore Magnet (Bruker NMR, Billerica, MA) and a 5 mm 1H/BB (109Ag-31P) inverse Triple-Axis Gradient Probe (ID500–5EB, Nalorac Cryogenic Corp., Martinez, CA) was used for all NMR experiments. The 2D NOESY spectra were acquired using a Watergate-water suppression pulse sequence (15) with mixing times of 120 and 250 ms and a recycle delay of 5.2 s. Data sets with 2048 complex points in t2 and 512 complex points in t1 were acquired with 5000 Hz sweep widths in both dimensions and 128 scans per slice. All spectra were processed with a combination of exponential and sine-skewed functions and zero-filled to 2 K x 2 K data points with an exponential weighting function or shifted sine-bell function to resolve overlapped imino protons using Bruker XWINNMR software. The longitudinal eddy current (LED) pulse sequence was used in all PFGSE-NMR diffusion experiments (16).
The gradient pulse interval, and gradient duration, , were 32 and 4 ms, respectively. The settling time was 6 ms and the relaxation 1 s. Typically, 2048 scans were collected for each spectra. The pulsed-field gradient was applied in the y direction. The gradient strength in a series of 1H experiments was incremented from 28 G cm–1 to about 56 G cm–1 in six increments. Accurate diffusion coefficients were obtained for six (three samples and three standard samples) from the slopes of ln (A/A0) versus g2, determined by least-squares regression.
Computer modeling
Structures were generated using MC-Sym software (17). For this study, scripts were written to generate structural models that both satisfied experimental constraints while also retaining, as close as possible, A-form RNA. In an effort to explore the full range of structures that could be constructed, an attempt was made to model by the step-wise addition of bases comprising the R1 oligomer to the R2 hairpin loop beginning with the 5' terminus and in a separate run, from the 3' terminus. This approach was used to eliminate the potential bias of the software. The MC-Sym scripts and constraints used in the modeling are available upon request. Output structures were energy minimized by 300 steps of steepest descent gradient using B biomolecular modeling package (N.White, http://www.scripps.edu/nwhite/B) with the AMBER force field parameters. Ribbon representations were created using MOLMOL software version 2K.2 (18). Analysis of the RNA structures was made with MC-Annotate (19). Structural comparisons were based on visual alignment and then the fit of selected backbone phosphates. The root mean square deviation (RMSD) calculated for comparison of RCNMV to Simian retrovirus-1 (SRV-1) and HIV-1 RNA structures was for 11 stem and eight loop phosphates.
Biological activity of RCNMV RNA-2 mutants
Oligomers were synthesized that incorporated site-specific mutations in the RCNMV RNA-2 TA region. Complementary oligomers were annealed and cloned into the TBSV expression vector pHST2 as previously described (10) and then assayed on Nicotiana benthamiana plants for the ability to trans-activate expression of GFP. Mutations were made at the end of the stem before the loop region such that the U:G base pair was changed to either a U:A (construct TA-M22) or a G:U base pair (construct TA-M23). RNA transcripts were synthesized in vitro and co-inoculated with an RCNMV RNA-1 construct (R1SG1) containing the green fluorescent protein (GFP) as a reporter gene onto N.benthamiana plants. sGFP expression was assayed 2–3 days post-inoculation (dpi) (10).
RESULTS
R18 and R220 form a 1:1 bimolecular complex
Non-denaturing PAGE was used to evaluate the interaction of the model oligomers representing the RNA-1 TABS and the RNA-2 TA (R18 and R220, respectively). When R18 and R220 were mixed and subjected to PAGE, an additional band was observed in the gel (Fig. 2A, lanes 3–7). When this band was excised from the gel, and re-electrophoresed under denaturing conditions, it was found to contain both R18 and R220 (Fig. 2B, lane 3) in the form of a complex. The formation of this complex band did not require magnesium ions. To determine the stoichiometry of the complex, multiple titrations of both component oligomers were subjected to PAGE and the intensities of the complex bands formed were determined by densitometry. The density values were analyzed with Prism software. The data best fit a single site binding model, indicating a 1:1 stoichiometry for the complex. The apparent Ka of the complex is 5.3 ± 0.5 x 104 M–1.
Figure 2. PAGE characterization of a complex formed when R18 and R220 are mixed. (A) Non-denaturing PAGE analysis. A titration of 30 μM of R220 with increasing amounts of R18 (15–120 μM) gives rise to an additional band. Lane 1, 30 μM R220; lane 2, 30 μM R18 (stains weakly); lanes 3–7, 30 μM R220 with 15, 30, 60, 90 and 120 μM of R18, respectively. The band in lane 7 was extracted and subjected to denaturing electrophoresis. (B) Denaturing PAGE was used to determine that the additional bands formed under non-denaturing conditions are comprised of R18 and R220. Lane 1, R220; lane 2, R18; lane 3, excised band of R18 and R220 complex.
Phylogenetic comparison of the TA elements for the three species comprising the Dianthovirus genus, all three of which trans-activate RCNMV sgRNA synthesis in vivo (10), indicated that a minimum complementarity of six bases is sufficient for viral activity (Fig. 1A). A series of oligomers that shifted an eight-nucleotide frame of the RNA-1 TABS element was synthesized and PAGE assayed to determine how complementarity affected complex formation. A fixed length of eight nucleotides was chosen to maintain the dynamics of the assay. From these experiments, it was determined that a stable R1:R2 complex formed with six-base complementarity (Table 1). But the most stable complex formed with the seven complementary bases found in R18+1. The PAGE assay was also used to confirm that R221*, a labeled in vitro transcript, formed a complex with the R18+1 oligomer.
Biophysical characterization of the complex components
The RNA oligomers were characterized by thermal denaturation over a range of 0.5–50 μM. R220 exhibited a bi-phasic denaturation profile with a Tm of 59°C (data not shown). The concentration-independent Tm indicated that R220 was a unimolecular structure. This is consistent with the mfold web server prediction that the region of RNA-2, containing this conserved TA sequence, forms a simple hairpin structure composed of a six-base pair stem and an eight-nucleotide loop (10). The imino region of the R220 2D NOESY (Fig. 3A), exhibited six well separated resonances consistent with Watson–Crick base paired imino protons. A signal for a terminal base paired imino proton was not observed due to the dynamics of the terminal base pair (20). Two resonances at 12.0 and 10.9 p.p.m. possessed strong NOE cross-peaks attributed to a U:G base pair. Unlike typical Watson–Crick base pairs that have a single imino proton, a U:G base pair has two imino protons with diagnostic strong cross-peaks (Fig. 3A). The assignment of the U:G base pair was further confirmed in the 2D 1H1-15N spectra of R221*, an isotopically labeled transcript (Fig. 3B). The stem imino resonances, with the exception of the terminal base pair, were assigned based on 2D 1H-1H NOE connectivities using the U:G base pair as a starting point (21). In contrast, the thermal denaturation profiles for the R1 oligomers exhibited only a limited increase in UV absorbance with increased temperature (data not shown) and the NMR spectra for R18+1 lacked features consistent with an ordered structure (Fig. 3C). Thermal denaturation of two additional mutant R2 sequences was also performed (Table 2). Changing the base pairings of the nucleotides at the top of the stem from U:G to U:A resulted in an increase of 4°C in the Tm, consistent with mfold web server predictions of comparable stability for the two sequences. Changing the pairing from U:G to G:U also resulted in a 20°C increase in Tm. While this oligomer formed a unimolecular structure based on concentration-independent Tm, the oligomer likely formed a different secondary structure than the six-base paired stem and eight-nucleotide loop of the TA. Structure prediction with mfold web server suggests that the sequence likely forms a more stable hairpin structure with a six-nucleotide loop, a five-base pair stem and four unpaired nucleotides. NMR analysis also supports this as the preferred conformation (data not shown).
Figure 3. NMR spectra of RCNMV oligomers. (A) An expanded contour plot for the imino region of the NOESY spectrum of R220 with 120 ms mixing time at 15°C. Six imino resonances are observed. The two resonances within the dashed box are assigned to the U6:G15 base pair that closes the stem of the hairpin. (B) 1H-15N spectra of R221* at 20°C identify those resonances that arise from labeled G residues. (C) The imino region of the 1D proton spectra for R18+1 at 10°C.
Table 2. Thermal stability, Tm, apparent association constants, Ka, for RNA2 mimics and correlation of in vitro complex formation with in vivo trans-activation for selected RCNMV RNA-2 stem sequence mutations
NMR diffusion rate and exchange regime of the R1:R2 complex
Diffusion experiments were performed to confirm that the complex formed in PAGE conditions also formed under NMR conditions. The NMR studies were conducted with R18+1 since this formed a complex with the highest Ka. When R18+1 and R221* were mixed at a ratio of 1:1, the diffusion rate shifted from 1.89 x 10–6 to 1.56 x 10–6 cm2 s–1 (Table 3). The decrease in mobility was attributed to the formation of a 29-nucleotide RNA complex. Formation of this complex decreased the signal-to-noise ratio, in the NMR spectra, 4-fold for the exchangeable protons. This decrease was not observed in the non-exchangeable signals. Loss of sensitivity arose from a change from the fast exchange for the R2 hairpin to an intermediate exchange regime for the complex. Attempts to shift back to a fast exchange by altering the temperature, pH, and addition of mono and divalent counter ions, were unsuccessful. Similar conversion to intermediate exchange regimes for other bimolecular RNA complexes with comparable weak association constants have previously been reported (22).
Table 3. Values of NMR diffusion experiments for selected oligomers
NMR of the complex
The NMR proton spectra for the R18+1:R220 complex, at 20°C, were notably different from those of the uncomplexed R220 (Fig. 4A and B). The imino resonances in the complex were broadened as compared to those of uncomplexed R220. The complex’s imino region exhibited five relatively narrow, well resolved, resonances and several broad resonances. The narrow signals at 14.1, 13.4, 12.8 and 12.3 (p.p.m.), were assigned to U18, G5, G4 and G2, respectively, in the stem of the RNA-2 hairpin. This assignment was based on connectivity in the 2D NOESY for the R18+1:R220 complex and the 2D 1H1-15N spectra of the R18+1:R221* complex (Fig. 4B and C). The retention of these resonances in the 2D 1H1-15N spectra of the R18+1:R221* complex constitutes definitive evidence for the persistence of the R2 hairpin stem in the complex and excludes the possibility that the R2 hairpin is completely relaxed upon complex formation. The remaining well separated, resonance at 12.6 p.p.m. cannot be definitely assigned, but certainly does arise from one of the labeled R221* G residues. Although this 12.6 p.p.m. signal does not have connectivity to other imino signals, it likely represents a G:C base pair based on the 2D NOESY cross-peaks. Comparison of the R220 (Fig. 3A) and the R18+1:R220 complex (Fig. 4B) 2D NOESY spectra indicated that the complex lacked the clear and characteristic U:G base pair cross-peaks prominent in the R220 hairpin (boxed region in Fig. 3A). This observation is consistent with the absence of the G6 N1 proton in the 1H1-15N spectra of the R18+1:R221* complex (Fig. 4C). Collectively, these results indicate that formation of the RNA-1:RNA-2 complex results in a partial relaxation of the RNA-2 hairpin stem and the concomitant loss of the U:G base pair closing the stem. Additional noteworthy distinctions between the R220 and complex NOESY spectra were observed in other regions as well. Figure 5 compares the spectra in three regions; the imino to aromatic (Fig. 5A), aromatic and H1' (Fig. 5B), and the sugar to aromatic (Fig. 5C). While detailed interpretation of the data from these regions is not possible without full isotopic labeling, several general observations can be made. Complex formation does not affect a majority of the NOESY cross-peaks for the non-exchangeable protons. A number of new cross-peaks arise as a consequence of complex formation. The number of new cross-peaks exceeded the number of peaks no longer observed. The shift of the complex’s exchange regime had limited effect on NOESY cross-peaks associated with non-exchangeable protons.
Figure 4. NMR spectra of the RCNMV complex. (A) The imino region of the 1D proton spectra for R220 and 1:1 complex of R18+1:R220 at 10°C. (B) The NMR, 2D NOESY spectra of the imino region for a 1:1 complex of R18+1:R220 at 10°C, with 250 ms mixing time. The connectivities of the stem backbone protons of R220 hairpin are observed in the complex. (C) The imino region of the 1H-15N spectra for a 1:1 complex of R18+1:R221* at 20°C.
Figure 5. Comparison of selected regions of the water NOESY spectra for RCNMV R220 and the R220:R18+1 complex. (A) In the region of the spectra characteristic of imino to aromatic interactions, a number of resolved cross-peaks are observed for both spectra. Using a color coded overlay of the two spectra, new resonances are observed (red) while most are unchanged (green) and others are absent (uncolored). (B) In the aromatic and H1' region, again notable differences can be observed between the two spectra. Following the same coloring scheme, a number of new resonances are observed (red) while most resonances appear unaffected (green) and several are not observed (uncolored). (C) Differences are also seen in the sugar to aromatic region. While this region has significant overlap, formation of the complex results in a noticeable increase in the complexity of this region. The broadening of resonances observed for exchangeable protons is not observed in this region.
Computer modeling of the complex
It was hoped that NMR experiments would provide enough restraints for a high resolution structural determination. But the intermediate exchange regime of the complex precluded that option. Instead, an alternative approach using the available NMR data as constraints for MC-Sym software calculations was employed to generate three-dimensional RNA structures consistent with this study’s experimental results. One of the preliminary modeling results was the determination that if the R2 stem was constrained to have five base pairs, as observed in NMR experiments, the MC-Sym algorithm was unable to generate models with seven or eight intermolecular base pairs. This result suggests that the native complex must contain fewer than the eight intermolecular base pairs initially indicated by the RCNMV sequence and mutation analysis. Supporting evidence for the complex having fewer than eight base pairs is the phylogenetic sequence comparison of all Dianthoviruses. In Carnation ringspot virus, six-base complementarity is observed between the TA and TABS. To better survey the structural possibilities that the data gathered in this study reflect, the MC-Sym algorithm was used to generate models for two separate sets of constraints. The first family of models (Model A) was generated with constraints reflecting experimental results for the most stable bimolecular complex identified by PAGE and subsequently investigated in the NMR experiments. These models have five intermolecular base pairs (Fig. 6A). By constraining the complex to having only five intermolecular base pairs, MC-Sym could generate models both with and without a constrained G:U base pair. The second family of models (Model B) was generated incorporating phylogenetic and PAGE data which collectively added the additional six intermolecular base pair (Fig. 6B) constraint. MC-Sym was only able to generate multiple structures for this constraint set when the G:U base pair constraint was removed, consistent with NMR results that indicate formation of the complex results in the relaxing of the G:U base pair.
Figure 6. Models of RCNMV R1:R2 bimolecular complex generated by MC-Sym. The backbone ribbon of RNA1 is green and RNA2 is cyan. (A) View of the mean structure Model A family constrained to have five intermolecular base pairs and a secondary structure representation. (B) View of the mean structure Model B family constrained to have six intermolecular base pairs and a secondary structure representation.
Within each set of models, the maximum RMSD between any two structures, for all heavy atoms, was no greater than 0.2 ?. Therefore, the mean structure for both families was generated and used as representative structures (Fig. 6A and B). A general comparison of the two families finds they have a similar global structure. Both families have two stacked helices consisting of (i) the stem of the R2 hairpin and (ii) a short helix formed by base pairing between the R2 loop and R1. In both families, the stem region of R2 retains five base pairs and the intermolecular helix contains a stretch of G:C base pairs that provides a core of thermodynamic stability. In both models, R2 nucleotides 7 and 8 are unpaired to allow a turn of the backbone that allows formation of the intermolecular base pairing. The similarity of the models can be seen in the secondary structures drawn for both families (Fig. 6). Despite their general similarities, there are notable differences between the two families. While the Model A family maintain A form RNA helical properties for both helical regions, the presence of an additional base pair in the Model B family results in a more compact intermolecular helix. The additional base pair, which is located in the hinge region between the two helices, results in a change in the angle between the two helical regions. The change is from about 60° for the Model A family to greater than 145° for Model B.
Closing base pair of the RNA-2 stem is biologically important
It is reported that in ribosomal and other functional RNAs U:G base pairs are not randomly distributed but are frequently located at the termination of hairpins in regions of bimolecular interaction (23). To determine whether the U:G in the RCNMV RNA-2 TA is important for either structure or function, two mutant constructs were created within RCNMV. These constructs were assayed in vitro for binding by PAGE and in vivo for their ability to trans-activate sgRNA synthesis in plants from an RCNMV RNA-1 construct (R1SG1) containing the green fluorescent protein (GFP) as a reporter gene (10). The first construct contained a single G A mutation that converted the U:G to a U:A base pair (R220U:A and TA-M22). PAGE found the Ka unaffected by the G A transition and co-inoculation of R1SG1 with TA-M22 resulted in wild-type expression of GFP at 2–3 dpi (Table 2). The second construct contained a double mutation that transposed the U:G base pair to a G:U (R220G:U and TA-M23). The transposition to a G:U base pair resulted in a 2-fold reduction in the Ka for complex formation and co-inoculation of R1SG1 with TA-M23 resulted in a severe reduction in GFP expression in vivo, 2–3 dpi (Table 2).
DISCUSSION
The aim of this study was to structurally characterize the rather unique RNA interaction involved in the regulation of RCNMV sgRNA synthesis. Using a variety of experimental approaches, we developed a model system to characterize the structure and dynamics of the RNAs involved in this process. The in vitro formation of a bimolecular RNA–RNA complex in the absence of protein supports a hypothesis that the underlying mechanism is an RNA process. However, the accessory role of host or viral proteins in this RNA–RNA interaction cannot be excluded. The data collected in this study indicate that the formation of the bimolecular complex involves structural changes to the RNA-2 hairpin. The most significant structural change observed in NMR experiments is the breaking of the U:G base pair at the top of the stem of RNA-2 in order to facilitate intermolecular base pairing in the complex. A number of experimental observations suggest that the identities of the nucleotides that close the stem are important for the optimum structure and function of the TA element. Mutagenesis of these nucleotides in the viral reporter system from U:G to G:U was sufficient to nearly abolish activity. The thermal denaturation measurements, for the RNA-2 mimics, show that changes to the nucleotides closing the hairpin stem can result in a >20°C increase in Tm. Since the native U:G RNA-2 mimic had the lowest experimental Tm, it would have the lowest barrier for relaxing the closing base pair. Relaxing of the U:G base pair during complex formation appears important in the formation of the intermolecular base pairs, as observed in the NMR experiments as well as in model building. The significance of a G:U closing the stem of a loop may not be unique to RCNMV. In Escherichia coli, U:G and G:U base pairs are not randomly distributed but are most frequently found at the termination of hairpins near regions of protein and RNA interactions (24). The increased frequency of these pairings in functional regions is likely due to the unique network of hydrogen bonds in U:G and G:U base pairs that allows greater conformational flexibility compared to other canonical and wobble pairings (25). The type of structure formed by the RCNMV TA element may be an example of a broader RNA motif that utilizes a U:G base pair to facilitate transient long range, structural interactions.
The data gathered in this study indicate that while the complex is stable, it displays dynamic properties consistent with a weak complex. The association constant, determined experimentally for the RCNMV complex (Ka = 5.3 x 104 M–1), is considered weak when compared to other biological, bimolecular interactions, being nearly two orders of magnitude lower than those of the highly efficient phage transcription regulation complexes (26). Yet, it is still within a biologically relevant range as it is less than an order of magnitude different from other RNA complexes possessing functional biological activity (26). The weak association of the RNAs makes formation of the complex an infrequent event, which may be functionally advantageous. By necessity, complex formation in RCNMV cannot be too favorable, at least early in the virus infection cycle, since this would lead to the prevention of full-length, genomic RNA synthesis. This low affinity property is likely employed as a temporal regulatory mechanism, based on an equilibrium effect. That is, the interaction is not favored until one or both components are at a sufficient concentration to drive the reaction towards complex formation. This situation occurs later in the viral replication cycle when the accumulation of both RNAs is at their greatest. This would also ensure that transcription of the capsid protein sgRNA does not occur until there is a threshold amount of both progeny viral genomic RNAs to be encapsidated by capsid protein. The direct RNA–RNA interaction provides a means of communication between the two viral RNAs to ensure that both are present in the same cell for the continuation of a productive infection.
Previous computer modeling of nucleic acids has generated structures that closely fit those structures obtained by experimental techniques. For example the model generated for the 29 packaging motor closely matched the structure obtained by biochemical studies (27). A modeling approach can be particularly valuable when working with nucleic acids that are dynamic or do not lend themselves to biochemical structural determinations. The evaluation of the two models generated in this study with MC-Annotate (19) found that the Model A family possessed key structural features consistent with a pseudoknot, in contrast to the Model B family which lacked pseudoknot features. Visual comparison of the backbone structure of Model A to RNA pseudoknot structures in the Protein Data Base confirms the general similarity (28). A more thorough comparison finds the greatest similarity to those pseudoknots involved in viral –1 ribosomal frameshifting (29). The structural similarities include (i) two stacked helical regions, (ii) 5–6 loop nucleotides base paired, (iii) G:C rich pairing in the long distance/intermolecular helical region and (iv) a limited number of unpaired loop nucleotides. The RCNMV TA TABS pseudoknot model now provides for a structural basis for interpretation of the functional activity for a series of genetic mutants in the RCNMV TA element (T.L.Sit, K.A.Turner and S.A.Lommel, unpublished data). This further supports these models as reasonable structures for the TA complex.
Alignment of the selected backbone phosphates from Model A to all pseudoknot structures in the PDB database found the greatest structural similarity to the SRV-1, –1 ribosomal frameshifting pseudoknot (30) with RMSD values of 2.0 ? (Fig. 7). This structural conservation likely also reflects functional conservation. At least one of the roles for a pseudoknot structure immediately downstream from the site of ribosomal frameshifting is to promote stalling of the advancing ribosome complex over the site of frameshifting. Similarly, the role of the RCNMV bimolecular interaction is hypothesized to periodically stall and dislodge the advancing replication complex during the process of full-length (–) strand RNA synthesis leading to a premature termination product. The 3D structural comparisons also found a significant structural similarity of the RCNMV Model A to the HIV-1 ‘kissing’ hairpin complex (Fig. 7) (11). The RMSD of selected backbone atoms was 3.2 ?. Structural similarity in the dimerization complex suggests additional conservation of form and function. Unpublished experimental data from our laboratory (V.R.Basnayake, T.L.Sit and S.A.Lommel, unpublished data) suggest that the RCNMV origin of assembly signal is located solely on RNA-2 and as a consequence, the formation of the RNA-1 and RNA-2 bimolecular complex late in the infection cycle is how RNA-1 is co-packaged into virions. While preliminary, these observations suggest a function remarkably similar to the HIV-1 dimerization complex. In essence, this RCNMV stacked helical bimolecular complex may likely possess two functions critical to the virus life cycle, which have been separately observed in two different retroviruses.
Figure 7. Comparison of the RCNMV TA model structure with the SRV-1 -1 ribosomal frameshifting pseudoknot and the HIV-1 kissing hairpin structure. Ribbon representations for backbones of: (A) SRV-1, (B) RCNMV and (C) HIV-1 dimerization element. (D) SRV-1 with RCNMV and (E) HIV-1 with RCNMV. The structures were compared by superimposing 19 backbone phosphates.
ACKNOWLEDGEMENTS
We thank Drs I. T. D. Petty and C. Hemenway for reviewing this manuscript and Carol George for her assistance in the preparation of isotopically labeled transcript. This research was supported by USDA NRI competitive grant 98-02298, NSF competitive grants (MCB-0077964) to S.A.L. and T.L.S. and (MCB-9986011) to P.F.A. M.A.D. was partly supported by an NIH grant (43237) to Dr Paul Wollenzien.
REFERENCES
Bernardi,F. and Haenni,A.L. (1998) Viruses: exquisite models for cell strategies. Biochimie, 80, 1035–1041.
Gale,M.,Jr, Tan,S.L. and Katze,M.G., (2000) Translational control of viral gene expression in eukaryotes. Microbiol. Mol. Biol. Rev., 64, 239–280.
Maia,I.G., Seron,K., Haenni,A.L. and Bernardi,F. (1996) Gene expression from viral RNA genomes. Plant Mol. Biol., 32, 367–391.
Miller,W.A., Dreher,T.W. and Hall,T.C. (1985) Synthesis of brome mosaic virus subgenomic RNA in vitro by internal initiation on (–)-sense genomic RNA. Nature, 313, 68–70.
White,K.A. (2002) The premature termination model: a possible third mechanism for subgenomic mRNA transcription in (+)-strand RNA viruses. Virology, 304, 147–154.
Eckerle,L.D. and Ball,L.A. (2002) Replication of the RNA segments of a bipartite viral genome is coordinated by a transactivating subgenomic RNA. Virology, 296, 165–176.
Kim,K.H. and Hemenway,C.L. (1999) Long-distance RNA–RNA interactions and conserved sequence elements affect potato virus X plus-strand RNA accumulation. RNA, 5, 636–645.
Zhang,G., Slowinski,V. and White,K.A. (1999) Subgenomic mRNA regulation by a distal RNA element in a (+)-strand RNA virus. RNA, 5, 550–561.
Lindenbach,B.D., Sgro,J.Y. and Ahlquist,P. (2002) Long-distance base pairing in flock house virus RNA1 regulates subgenomic RNA3 synthesis and RNA2 replication. J. Virol., 76, 3905–3919.
Sit,T.L., Vaewhongs,A.A. and Lommel,S.A. (1998) RNA-mediated trans-activation of transcription from a viral RNA. Science, 281, 829–832.
Chang,K.Y. and Tinoco,I.,Jr (1997) The structure of an RNA ‘kissing’ hairpin complex of the HIV TAR hairpin loop and its complement. J. Mol. Biol., 269, 52–66.
Mujeeb,A., Clever,J.L., Billeci,T.M., James,T.L. and Parslow,T.G. (1998) Structure of the dimer initiation complex of HIV-1 genomic RNA. Nat. Struct. Biol., 6, 432–436.
Ban,N., Nissen,P., Hansen,J., Moore,P.B. and Steitz,T.A. (2000) The complete atomic structure of the large ribosomal subunit at 2.4 ? resolution. Science, 289, 905–920.
Brunel,C., Marquet,R., Romby,P. and Ehresman,C. (2002) RNA loop–loop interactions as dynamic functional motifs. Biochimie, 84, 924–944.
Piotto,M., Saudek,V. and Sklenar,V. (1992) Gradient-tailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J. Biomol. NMR, 6, 661–665.
Gibbs,S.J. and Johnson,C.S.,Jr (1991) A PFG NMR experiment for accurate diffusion and flow studies in the presence of eddy currents. J. Magn. Reson., 93, 395–402.
Major,F., Turcotte,M., Gautheret,D., Lapalme,G., Fillion,E. and Cedergren,R. (1991) The combination of symbolic and numerical computation for three-dimensional modeling of RNA. Science, 253, 1255–1260.
Koradi,R., Billeter,M. and Wüthrich,K. (1996) MOLMOL: a program for display and analysis of macromolecular structures J. Mol. Graph., 14, 51–55.
Gendron,P., Lemieuxs,S. and Major,F. (2001) Quantitative analysis of nucleic acid three-dimensional structures. J. Mol. Biol., 308, 919–936.
Varani,G. and Tinoco,I.,Jr (1991) RNA structure and NMR spectroscopy. Q. Rev. Biophys., 24, 479–532.
Kolk,M.H., van der Graaf,M., Wijmenga,S.S., Pleij,C.W., Heus,H.A. and Hilbers,C.W. (1998) NMR structure of a classical pseudoknot: interplay of single- and double-stranded RNA. Science, 280, 434–438.
Wemmer,D.E. and Williams,P.G. (1994) Use of nuclear-magnetic resonance in probing ligand-macromolecular interactions. Methods Enzymol., 239, 737–767.
Masquida,B. and Westhof,E. (2000) On the wobble GoU and related pairs. RNA, 6, 9–15.
Szymanski,M., Barciszewska,M.Z., Erdmann,V.A. and Barciszewski,J. (2000) An analysis of G-U base pair occurrence in eukaryotic 5S rRNAs. Mol. Biol. Evol., 8, 1194–1198.
Giese,M.R., Betschart,K., Dale,T., Riley,C.K., Rowan,C., Sprouse,K.J. and Serra,M.J. (1998) Stability of RNA hairpins closed by wobble base pairs. Biochemistry, 37, 1094–1100.
Wagner,E.G. and Simons,R.W. (1994) Antisense RNA control in bacteria, phage and plasmids. Annu. Rev. Microbiol., 48, 713–742.
Hoeprich,S. and Guo,P. (2002) Computer modeling of three-dimensional structure of DNA-packaging RNA (pRNA) monomer, dimer and hexamer of Phi29 DNA packaging motor. J. Biol. Chem., 277, 20794–20803.
Berman,M., Westbrook,J., Feng,Z., Gilliland,G., Bhat,T.N., Weissig,H., Shindyalov,I.N. and Bourne,P.E. (2000) The Protein Data Bank. Nucleic Acids Res., 28, 235–242.
Hilbers,C.W., Michiels,P.J. and Heus,H.A. (1999) New developments in structure determination of pseudoknots. Biopolymers, 48, 137–153.
Michiels,P.J., Versleijen,A.A., Verlaan,P.W., Pleij,C.W., Hilbers,C.W. and Heus,H.A. (2001) Solution structure of the pseudoknot of SRV-1 RNA, involved in ribosomal frameshifting. J. Mol. Biol., 310, 1109–1123.(Richard H. Guenther, Tim L. Sit, Hanna S)
*To whom correspondence should be addressed. Tel: +1 919 515 6992; Fax: +1 919 515 7716; Email: steve_lommel@ncsu.edu
Present addresses:
Michael A. Dolan, Tripos Inc., St Louis, MO 63144, USA
Guihua Liu, Department of Chemistry, China Medical University, Shenyang, China
ABSTRACT
The 34-nucleotide trans-activator (TA) located within the RNA-2 of Red clover necrotic mosaic virus folds into a simple hairpin. The eight-nucleotide TA loop base pairs with eight complementary nucleotides in the TA binding sequence (TABS) of the capsid protein subgenomic promoter on RNA-1 and trans-activates subgenomic RNA synthesis. Short synthetic oligoribonucleotide mimics of the RNA-1 TABS and the RNA-2 TA form a weak 1:1 bimolecular complex in vitro with a Ka of 5.3 x 104 M–1. Ka determination for a series of RNA-1 and RNA-2 mimic variants indicated optimum stability is obtained with seven-base complementarity. Thermal denaturation and NMR show that the RNA-1 TABS 8mers are weakly ordered in solution while RNA-2 TA oligomers form the predicted hairpin. NMR diffusion studies confirmed RNA-1 and RNA-2 oligomer complex formation in vitro. MC-Sym generated structural models suggest that the bimolecular complex is composed of two stacked helices, one being the stem of the RNA-2 TA hairpin and the other formed by the intermolecular base pairing between RNA-1 and RNA-2. The RCNMV TA structural model is similar to those for the Simian retrovirus frameshifting element and the Human immunodeficiency virus-1 dimerization kissing hairpins, suggesting a conservation of form and function.
INTRODUCTION
RNA viruses employ a diverse array of RNA motifs for the regulation of gene expression through interactions with other nucleic acids and proteins. The genomes of positive (+) stranded RNA viruses behave as mRNAs within host cells to allow the expression of viral proteins. However, many viral genomes are polycistronic and the expression of inaccessible open reading frames (ORFs) is facilitated by multiple strategies (1,2). Subgenomic RNA (sgRNA) synthesis is a common mechanism for the generation of viral RNAs that behave as monocistronic mRNAs for internal and 3' co-terminal ORFs (3). Essentially, there are two proposed mechanisms for the generation of (+) stranded sgRNAs by the virally encoded RNA polymerase (i) internal initiation of transcription from the full length minus (–) strand copy of the genomic RNA (4) and (ii) premature termination during (–) strand synthesis followed by independent replication of this truncated RNA template (5). For both models, the presence of a distinct subgenomic promoter element on the (–) strand or a terminator on the (+) strand, is required for the initiation of transcription (1).
Recently, it has been determined that for several RNA viruses, RNA synthesis requires long-distance cis or trans RNA–RNA interactions (6). It is suggested that viruses that utilize these interactions generate sgRNA via the premature termination mechanism. The base pairing of upstream cis-acting elements to the subgenomic promoter was found to be essential for sgRNA synthesis in Potato virus X, Tomato bushy stunt virus (TBSV) and Flock house virus (7–9). The situation is taken to its extreme for the bipartite genome of Red clover necrotic mosaic virus (Fig. 1A) (RCNMV; genus Dianthovirus, family Tombusviridae) where expression of the sgRNA from RNA-1 requires the base pairing of a sequence element in trans in RNA-2 (10). The 34-nucleotide trans-activator (TA) element from RNA-2 is predicted to form a stable hairpin structure with an eight-nucleotide loop possessing complementarity to an eight-nucleotide element immediately upstream of the sgRNA transcription start site on RNA-1, referred to as the trans-activator binding site (TABS) (Fig. 1B). Mutagenesis demonstrated that complementary base pairing between the two elements is essential for sgRNA transcription but the nature of this RNA–RNA tertiary structure has not been elucidated.
Figure 1. Genome organization, expression strategy and model of RNA-1–RNA-2 interaction facilitating dianthovirus RNA-1 sgRNA synthesis. (A) The bipartite RNA genome of Red clover necrotic mosaic virus (RCNMV) is depicted with open rectangles representing ORFs. RNA-1 ORFs p27 and p88 are involved in viral replication while the capsid protein (CP) is the virion structural protein. RNA-2 encodes the movement protein (MP) required for cell-to-cell movement of the viral RNA. p88 is expressed by a –1 ribosomal frameshifting event (–1 FS) of p27. The CP is expressed from a sgRNA generated during the replication of the virus. The RNA-1 sequence near the sgRNA transcription start site (denoted with an arrow) is shown aligned with comparable regions from Sweet clover necrotic mosaic virus (SCNMV) and Carnation ringspot virus (CRSV), the two other species comprising the Dianthovirus genus. The RNA-2 sequence of the TA region is shown below RNA-2 with the corresponding aligned regions from SCNMV and CRSV. The shaded sequences represent the complementary regions between RNA-1 and RNA-2 involved in trans-activation of transcription. The nucleotide sequences for the RCNMV oligomers used in this study are depicted below their respective genomic sequences. The nucleotide positions of the various sequences are noted. (B) Model of interaction previously proposed (10).
Determination of the tertiary structures of RNAs, particularly RNA–RNA complexes is difficult. Only a few examples of high-resolution non-duplex bimolecular RNA–RNA structures such as the Human immunodeficiency virus-1 (HIV-1) kissing hairpins and dimerization signal (11,12) and interactions within the ribosome complex (13) have been elucidated and provide limited examples of structural variation. While bimolecular RNA interactions generally involve loop–loop interactions (14), it is likely that a range of structural diversity exists in bimolecular RNA–RNA complexes. Many structures may prove to be similar to unimolecular tertiary interactions, particularly those of similar function. Here we report direct biophysical evidence for the formation of a weak, yet stable bimolecular complex between the RCNMV RNA-2 TA element and the RNA-1 subgenomic promoter. We further propose structural models for this interaction that exhibit conservation in both structure and function with other viral cis and trans RNA interactions.
MATERIALS AND METHODS
RNA oligomer preparation
Synthetic oligomers used in the study were obtained from the North Carolina State University Nucleic Acids Facility and from Dharmacon, Inc. (Lafayette, CO). Isotopically labeled RNA for the NMR studies was obtained by T7 transcription of a plasmid template containing the 21-nucleotide RCNMV RNA-2 sequence (Fig. 1A) cloned into pUC18 and terminated at the XbaI site. The transcription reaction was performed at 37°C for 8 h using a T7-MEGAshortscript High Yield Transcription Kit (Ambion, Austin, TX). For the isotopic labeling, 80 μM uniformly labeled 13C and 15N GTP sodium salt (Isotec, Miamisburg, OH) was substituted for the kit’s 75 mM GTP solution. Each oligomer was purified by anion exchange HPLC using a Machery-Nagel 250/10 Nucleogen DEAE 60-7 column (Duren, Germany). The purity of each oligomer was confirmed by denaturing polyacrylamide gel electrophoresis (PAGE). The RNA-1 oligomers representing the TABS are designated R1 with a superscript denoting the length, e.g. R18 is the eight-nucleotide oligomer (nucleotides 2356–2363) used in preliminary NMR studies. Four R18 oligomers, used in PAGE and NMR experiments have an additional + or – superscript denoting the nucleotide shift from the central sequence (nucleotides 2356–2363). The RNA-2 oligomer is similarly designated R220 (nucleotides 761–780) while mutant versions are identified by a superscript denoting mutations in the U:G base pair. The RNA-2 transcript was designated R221*. The sequences of the oligomers are shown in Figure 1A and Table 1.
Table 1. The apparent complex association constants, Ka, for a series of RCNMV RNA-1 oligomers binding to R220
Thermal denaturation
Thermal denaturation of the oligomers was monitored by UV absorbance (at 260 nm) over a temperature range of 5–90°C (1°C min–1) using a Cary 3 Spectrophotometer (Varian, Palo Alto, CA). The denaturation profile of each oligomer was conducted in triplicate at three concentrations, 0.5, 5 and 50 μM in 10 mM Na phosphate, 100 mM NaCl, pH 7.0. Both 2 and 10 mm cuvettes were used in order to maintain proper optical response. Variation of Tm was less than 1°C for all concentrations. Samples were thermally denatured and then renatured to determine transition reversibility. In all cases, a difference of less than 1°C in Tm was observed between denaturation and renaturation with no hysteresis detected.
Polyacrylamide gel electrophoresis assay
Denaturing PAGE (7 M urea, 9 mM Tris–borate, 0.25 mM EDTA in 20% polyacrylamide) was conducted to assess the purity of the synthetic oligomers and to analyze the components of the bimolecular complex. Non-denaturing PAGE (9 mM Tris–borate in 20% polyacrylamide run on ice) was used to investigate and characterize parameters affecting complex formation. After electrophoresis, gels were stained with ethidium bromide (EtBr), visualized by UV transillumination and images recorded on a Bio-Rad Gel Doc 2000 imager (Bio-Rad Laboratories, Hercules, CA). The intensities of the RNA complex bands were quantified using Bio-Rad Quantity One software. Over the concentration ranges used, it was determined that EtBr intercalation reported a linear measure of RNA concentration. Apparent association constants (Ka values) were calculated based on one site binding with the aid of PRISM software (GraphPad Software, San Diego, CA).
NMR analysis
Samples were prepared by solubilizing purified RNA in 0.2 ml of 90% H2O/10% D2O phosphate buffer to give 0.5 mM RNA. Samples were equilibrated with multiple buffer exchanges in Centricon spin columns (Millipore, Milford, MA) and placed in NMR microtubes (Shigemi, Allison Park, PA, stock # BMS005V). The NMR spectra for the 1:1 complex of R18+1:R221* were also collected at 25°C in 100% D2O phosphate buffer . Bruker AVANCE 500 MHz spectrometer with an Oxford Narrow Bore Magnet (Bruker NMR, Billerica, MA) and a 5 mm 1H/BB (109Ag-31P) inverse Triple-Axis Gradient Probe (ID500–5EB, Nalorac Cryogenic Corp., Martinez, CA) was used for all NMR experiments. The 2D NOESY spectra were acquired using a Watergate-water suppression pulse sequence (15) with mixing times of 120 and 250 ms and a recycle delay of 5.2 s. Data sets with 2048 complex points in t2 and 512 complex points in t1 were acquired with 5000 Hz sweep widths in both dimensions and 128 scans per slice. All spectra were processed with a combination of exponential and sine-skewed functions and zero-filled to 2 K x 2 K data points with an exponential weighting function or shifted sine-bell function to resolve overlapped imino protons using Bruker XWINNMR software. The longitudinal eddy current (LED) pulse sequence was used in all PFGSE-NMR diffusion experiments (16).
The gradient pulse interval, and gradient duration, , were 32 and 4 ms, respectively. The settling time was 6 ms and the relaxation 1 s. Typically, 2048 scans were collected for each spectra. The pulsed-field gradient was applied in the y direction. The gradient strength in a series of 1H experiments was incremented from 28 G cm–1 to about 56 G cm–1 in six increments. Accurate diffusion coefficients were obtained for six (three samples and three standard samples) from the slopes of ln (A/A0) versus g2, determined by least-squares regression.
Computer modeling
Structures were generated using MC-Sym software (17). For this study, scripts were written to generate structural models that both satisfied experimental constraints while also retaining, as close as possible, A-form RNA. In an effort to explore the full range of structures that could be constructed, an attempt was made to model by the step-wise addition of bases comprising the R1 oligomer to the R2 hairpin loop beginning with the 5' terminus and in a separate run, from the 3' terminus. This approach was used to eliminate the potential bias of the software. The MC-Sym scripts and constraints used in the modeling are available upon request. Output structures were energy minimized by 300 steps of steepest descent gradient using B biomolecular modeling package (N.White, http://www.scripps.edu/nwhite/B) with the AMBER force field parameters. Ribbon representations were created using MOLMOL software version 2K.2 (18). Analysis of the RNA structures was made with MC-Annotate (19). Structural comparisons were based on visual alignment and then the fit of selected backbone phosphates. The root mean square deviation (RMSD) calculated for comparison of RCNMV to Simian retrovirus-1 (SRV-1) and HIV-1 RNA structures was for 11 stem and eight loop phosphates.
Biological activity of RCNMV RNA-2 mutants
Oligomers were synthesized that incorporated site-specific mutations in the RCNMV RNA-2 TA region. Complementary oligomers were annealed and cloned into the TBSV expression vector pHST2 as previously described (10) and then assayed on Nicotiana benthamiana plants for the ability to trans-activate expression of GFP. Mutations were made at the end of the stem before the loop region such that the U:G base pair was changed to either a U:A (construct TA-M22) or a G:U base pair (construct TA-M23). RNA transcripts were synthesized in vitro and co-inoculated with an RCNMV RNA-1 construct (R1SG1) containing the green fluorescent protein (GFP) as a reporter gene onto N.benthamiana plants. sGFP expression was assayed 2–3 days post-inoculation (dpi) (10).
RESULTS
R18 and R220 form a 1:1 bimolecular complex
Non-denaturing PAGE was used to evaluate the interaction of the model oligomers representing the RNA-1 TABS and the RNA-2 TA (R18 and R220, respectively). When R18 and R220 were mixed and subjected to PAGE, an additional band was observed in the gel (Fig. 2A, lanes 3–7). When this band was excised from the gel, and re-electrophoresed under denaturing conditions, it was found to contain both R18 and R220 (Fig. 2B, lane 3) in the form of a complex. The formation of this complex band did not require magnesium ions. To determine the stoichiometry of the complex, multiple titrations of both component oligomers were subjected to PAGE and the intensities of the complex bands formed were determined by densitometry. The density values were analyzed with Prism software. The data best fit a single site binding model, indicating a 1:1 stoichiometry for the complex. The apparent Ka of the complex is 5.3 ± 0.5 x 104 M–1.
Figure 2. PAGE characterization of a complex formed when R18 and R220 are mixed. (A) Non-denaturing PAGE analysis. A titration of 30 μM of R220 with increasing amounts of R18 (15–120 μM) gives rise to an additional band. Lane 1, 30 μM R220; lane 2, 30 μM R18 (stains weakly); lanes 3–7, 30 μM R220 with 15, 30, 60, 90 and 120 μM of R18, respectively. The band in lane 7 was extracted and subjected to denaturing electrophoresis. (B) Denaturing PAGE was used to determine that the additional bands formed under non-denaturing conditions are comprised of R18 and R220. Lane 1, R220; lane 2, R18; lane 3, excised band of R18 and R220 complex.
Phylogenetic comparison of the TA elements for the three species comprising the Dianthovirus genus, all three of which trans-activate RCNMV sgRNA synthesis in vivo (10), indicated that a minimum complementarity of six bases is sufficient for viral activity (Fig. 1A). A series of oligomers that shifted an eight-nucleotide frame of the RNA-1 TABS element was synthesized and PAGE assayed to determine how complementarity affected complex formation. A fixed length of eight nucleotides was chosen to maintain the dynamics of the assay. From these experiments, it was determined that a stable R1:R2 complex formed with six-base complementarity (Table 1). But the most stable complex formed with the seven complementary bases found in R18+1. The PAGE assay was also used to confirm that R221*, a labeled in vitro transcript, formed a complex with the R18+1 oligomer.
Biophysical characterization of the complex components
The RNA oligomers were characterized by thermal denaturation over a range of 0.5–50 μM. R220 exhibited a bi-phasic denaturation profile with a Tm of 59°C (data not shown). The concentration-independent Tm indicated that R220 was a unimolecular structure. This is consistent with the mfold web server prediction that the region of RNA-2, containing this conserved TA sequence, forms a simple hairpin structure composed of a six-base pair stem and an eight-nucleotide loop (10). The imino region of the R220 2D NOESY (Fig. 3A), exhibited six well separated resonances consistent with Watson–Crick base paired imino protons. A signal for a terminal base paired imino proton was not observed due to the dynamics of the terminal base pair (20). Two resonances at 12.0 and 10.9 p.p.m. possessed strong NOE cross-peaks attributed to a U:G base pair. Unlike typical Watson–Crick base pairs that have a single imino proton, a U:G base pair has two imino protons with diagnostic strong cross-peaks (Fig. 3A). The assignment of the U:G base pair was further confirmed in the 2D 1H1-15N spectra of R221*, an isotopically labeled transcript (Fig. 3B). The stem imino resonances, with the exception of the terminal base pair, were assigned based on 2D 1H-1H NOE connectivities using the U:G base pair as a starting point (21). In contrast, the thermal denaturation profiles for the R1 oligomers exhibited only a limited increase in UV absorbance with increased temperature (data not shown) and the NMR spectra for R18+1 lacked features consistent with an ordered structure (Fig. 3C). Thermal denaturation of two additional mutant R2 sequences was also performed (Table 2). Changing the base pairings of the nucleotides at the top of the stem from U:G to U:A resulted in an increase of 4°C in the Tm, consistent with mfold web server predictions of comparable stability for the two sequences. Changing the pairing from U:G to G:U also resulted in a 20°C increase in Tm. While this oligomer formed a unimolecular structure based on concentration-independent Tm, the oligomer likely formed a different secondary structure than the six-base paired stem and eight-nucleotide loop of the TA. Structure prediction with mfold web server suggests that the sequence likely forms a more stable hairpin structure with a six-nucleotide loop, a five-base pair stem and four unpaired nucleotides. NMR analysis also supports this as the preferred conformation (data not shown).
Figure 3. NMR spectra of RCNMV oligomers. (A) An expanded contour plot for the imino region of the NOESY spectrum of R220 with 120 ms mixing time at 15°C. Six imino resonances are observed. The two resonances within the dashed box are assigned to the U6:G15 base pair that closes the stem of the hairpin. (B) 1H-15N spectra of R221* at 20°C identify those resonances that arise from labeled G residues. (C) The imino region of the 1D proton spectra for R18+1 at 10°C.
Table 2. Thermal stability, Tm, apparent association constants, Ka, for RNA2 mimics and correlation of in vitro complex formation with in vivo trans-activation for selected RCNMV RNA-2 stem sequence mutations
NMR diffusion rate and exchange regime of the R1:R2 complex
Diffusion experiments were performed to confirm that the complex formed in PAGE conditions also formed under NMR conditions. The NMR studies were conducted with R18+1 since this formed a complex with the highest Ka. When R18+1 and R221* were mixed at a ratio of 1:1, the diffusion rate shifted from 1.89 x 10–6 to 1.56 x 10–6 cm2 s–1 (Table 3). The decrease in mobility was attributed to the formation of a 29-nucleotide RNA complex. Formation of this complex decreased the signal-to-noise ratio, in the NMR spectra, 4-fold for the exchangeable protons. This decrease was not observed in the non-exchangeable signals. Loss of sensitivity arose from a change from the fast exchange for the R2 hairpin to an intermediate exchange regime for the complex. Attempts to shift back to a fast exchange by altering the temperature, pH, and addition of mono and divalent counter ions, were unsuccessful. Similar conversion to intermediate exchange regimes for other bimolecular RNA complexes with comparable weak association constants have previously been reported (22).
Table 3. Values of NMR diffusion experiments for selected oligomers
NMR of the complex
The NMR proton spectra for the R18+1:R220 complex, at 20°C, were notably different from those of the uncomplexed R220 (Fig. 4A and B). The imino resonances in the complex were broadened as compared to those of uncomplexed R220. The complex’s imino region exhibited five relatively narrow, well resolved, resonances and several broad resonances. The narrow signals at 14.1, 13.4, 12.8 and 12.3 (p.p.m.), were assigned to U18, G5, G4 and G2, respectively, in the stem of the RNA-2 hairpin. This assignment was based on connectivity in the 2D NOESY for the R18+1:R220 complex and the 2D 1H1-15N spectra of the R18+1:R221* complex (Fig. 4B and C). The retention of these resonances in the 2D 1H1-15N spectra of the R18+1:R221* complex constitutes definitive evidence for the persistence of the R2 hairpin stem in the complex and excludes the possibility that the R2 hairpin is completely relaxed upon complex formation. The remaining well separated, resonance at 12.6 p.p.m. cannot be definitely assigned, but certainly does arise from one of the labeled R221* G residues. Although this 12.6 p.p.m. signal does not have connectivity to other imino signals, it likely represents a G:C base pair based on the 2D NOESY cross-peaks. Comparison of the R220 (Fig. 3A) and the R18+1:R220 complex (Fig. 4B) 2D NOESY spectra indicated that the complex lacked the clear and characteristic U:G base pair cross-peaks prominent in the R220 hairpin (boxed region in Fig. 3A). This observation is consistent with the absence of the G6 N1 proton in the 1H1-15N spectra of the R18+1:R221* complex (Fig. 4C). Collectively, these results indicate that formation of the RNA-1:RNA-2 complex results in a partial relaxation of the RNA-2 hairpin stem and the concomitant loss of the U:G base pair closing the stem. Additional noteworthy distinctions between the R220 and complex NOESY spectra were observed in other regions as well. Figure 5 compares the spectra in three regions; the imino to aromatic (Fig. 5A), aromatic and H1' (Fig. 5B), and the sugar to aromatic (Fig. 5C). While detailed interpretation of the data from these regions is not possible without full isotopic labeling, several general observations can be made. Complex formation does not affect a majority of the NOESY cross-peaks for the non-exchangeable protons. A number of new cross-peaks arise as a consequence of complex formation. The number of new cross-peaks exceeded the number of peaks no longer observed. The shift of the complex’s exchange regime had limited effect on NOESY cross-peaks associated with non-exchangeable protons.
Figure 4. NMR spectra of the RCNMV complex. (A) The imino region of the 1D proton spectra for R220 and 1:1 complex of R18+1:R220 at 10°C. (B) The NMR, 2D NOESY spectra of the imino region for a 1:1 complex of R18+1:R220 at 10°C, with 250 ms mixing time. The connectivities of the stem backbone protons of R220 hairpin are observed in the complex. (C) The imino region of the 1H-15N spectra for a 1:1 complex of R18+1:R221* at 20°C.
Figure 5. Comparison of selected regions of the water NOESY spectra for RCNMV R220 and the R220:R18+1 complex. (A) In the region of the spectra characteristic of imino to aromatic interactions, a number of resolved cross-peaks are observed for both spectra. Using a color coded overlay of the two spectra, new resonances are observed (red) while most are unchanged (green) and others are absent (uncolored). (B) In the aromatic and H1' region, again notable differences can be observed between the two spectra. Following the same coloring scheme, a number of new resonances are observed (red) while most resonances appear unaffected (green) and several are not observed (uncolored). (C) Differences are also seen in the sugar to aromatic region. While this region has significant overlap, formation of the complex results in a noticeable increase in the complexity of this region. The broadening of resonances observed for exchangeable protons is not observed in this region.
Computer modeling of the complex
It was hoped that NMR experiments would provide enough restraints for a high resolution structural determination. But the intermediate exchange regime of the complex precluded that option. Instead, an alternative approach using the available NMR data as constraints for MC-Sym software calculations was employed to generate three-dimensional RNA structures consistent with this study’s experimental results. One of the preliminary modeling results was the determination that if the R2 stem was constrained to have five base pairs, as observed in NMR experiments, the MC-Sym algorithm was unable to generate models with seven or eight intermolecular base pairs. This result suggests that the native complex must contain fewer than the eight intermolecular base pairs initially indicated by the RCNMV sequence and mutation analysis. Supporting evidence for the complex having fewer than eight base pairs is the phylogenetic sequence comparison of all Dianthoviruses. In Carnation ringspot virus, six-base complementarity is observed between the TA and TABS. To better survey the structural possibilities that the data gathered in this study reflect, the MC-Sym algorithm was used to generate models for two separate sets of constraints. The first family of models (Model A) was generated with constraints reflecting experimental results for the most stable bimolecular complex identified by PAGE and subsequently investigated in the NMR experiments. These models have five intermolecular base pairs (Fig. 6A). By constraining the complex to having only five intermolecular base pairs, MC-Sym could generate models both with and without a constrained G:U base pair. The second family of models (Model B) was generated incorporating phylogenetic and PAGE data which collectively added the additional six intermolecular base pair (Fig. 6B) constraint. MC-Sym was only able to generate multiple structures for this constraint set when the G:U base pair constraint was removed, consistent with NMR results that indicate formation of the complex results in the relaxing of the G:U base pair.
Figure 6. Models of RCNMV R1:R2 bimolecular complex generated by MC-Sym. The backbone ribbon of RNA1 is green and RNA2 is cyan. (A) View of the mean structure Model A family constrained to have five intermolecular base pairs and a secondary structure representation. (B) View of the mean structure Model B family constrained to have six intermolecular base pairs and a secondary structure representation.
Within each set of models, the maximum RMSD between any two structures, for all heavy atoms, was no greater than 0.2 ?. Therefore, the mean structure for both families was generated and used as representative structures (Fig. 6A and B). A general comparison of the two families finds they have a similar global structure. Both families have two stacked helices consisting of (i) the stem of the R2 hairpin and (ii) a short helix formed by base pairing between the R2 loop and R1. In both families, the stem region of R2 retains five base pairs and the intermolecular helix contains a stretch of G:C base pairs that provides a core of thermodynamic stability. In both models, R2 nucleotides 7 and 8 are unpaired to allow a turn of the backbone that allows formation of the intermolecular base pairing. The similarity of the models can be seen in the secondary structures drawn for both families (Fig. 6). Despite their general similarities, there are notable differences between the two families. While the Model A family maintain A form RNA helical properties for both helical regions, the presence of an additional base pair in the Model B family results in a more compact intermolecular helix. The additional base pair, which is located in the hinge region between the two helices, results in a change in the angle between the two helical regions. The change is from about 60° for the Model A family to greater than 145° for Model B.
Closing base pair of the RNA-2 stem is biologically important
It is reported that in ribosomal and other functional RNAs U:G base pairs are not randomly distributed but are frequently located at the termination of hairpins in regions of bimolecular interaction (23). To determine whether the U:G in the RCNMV RNA-2 TA is important for either structure or function, two mutant constructs were created within RCNMV. These constructs were assayed in vitro for binding by PAGE and in vivo for their ability to trans-activate sgRNA synthesis in plants from an RCNMV RNA-1 construct (R1SG1) containing the green fluorescent protein (GFP) as a reporter gene (10). The first construct contained a single G A mutation that converted the U:G to a U:A base pair (R220U:A and TA-M22). PAGE found the Ka unaffected by the G A transition and co-inoculation of R1SG1 with TA-M22 resulted in wild-type expression of GFP at 2–3 dpi (Table 2). The second construct contained a double mutation that transposed the U:G base pair to a G:U (R220G:U and TA-M23). The transposition to a G:U base pair resulted in a 2-fold reduction in the Ka for complex formation and co-inoculation of R1SG1 with TA-M23 resulted in a severe reduction in GFP expression in vivo, 2–3 dpi (Table 2).
DISCUSSION
The aim of this study was to structurally characterize the rather unique RNA interaction involved in the regulation of RCNMV sgRNA synthesis. Using a variety of experimental approaches, we developed a model system to characterize the structure and dynamics of the RNAs involved in this process. The in vitro formation of a bimolecular RNA–RNA complex in the absence of protein supports a hypothesis that the underlying mechanism is an RNA process. However, the accessory role of host or viral proteins in this RNA–RNA interaction cannot be excluded. The data collected in this study indicate that the formation of the bimolecular complex involves structural changes to the RNA-2 hairpin. The most significant structural change observed in NMR experiments is the breaking of the U:G base pair at the top of the stem of RNA-2 in order to facilitate intermolecular base pairing in the complex. A number of experimental observations suggest that the identities of the nucleotides that close the stem are important for the optimum structure and function of the TA element. Mutagenesis of these nucleotides in the viral reporter system from U:G to G:U was sufficient to nearly abolish activity. The thermal denaturation measurements, for the RNA-2 mimics, show that changes to the nucleotides closing the hairpin stem can result in a >20°C increase in Tm. Since the native U:G RNA-2 mimic had the lowest experimental Tm, it would have the lowest barrier for relaxing the closing base pair. Relaxing of the U:G base pair during complex formation appears important in the formation of the intermolecular base pairs, as observed in the NMR experiments as well as in model building. The significance of a G:U closing the stem of a loop may not be unique to RCNMV. In Escherichia coli, U:G and G:U base pairs are not randomly distributed but are most frequently found at the termination of hairpins near regions of protein and RNA interactions (24). The increased frequency of these pairings in functional regions is likely due to the unique network of hydrogen bonds in U:G and G:U base pairs that allows greater conformational flexibility compared to other canonical and wobble pairings (25). The type of structure formed by the RCNMV TA element may be an example of a broader RNA motif that utilizes a U:G base pair to facilitate transient long range, structural interactions.
The data gathered in this study indicate that while the complex is stable, it displays dynamic properties consistent with a weak complex. The association constant, determined experimentally for the RCNMV complex (Ka = 5.3 x 104 M–1), is considered weak when compared to other biological, bimolecular interactions, being nearly two orders of magnitude lower than those of the highly efficient phage transcription regulation complexes (26). Yet, it is still within a biologically relevant range as it is less than an order of magnitude different from other RNA complexes possessing functional biological activity (26). The weak association of the RNAs makes formation of the complex an infrequent event, which may be functionally advantageous. By necessity, complex formation in RCNMV cannot be too favorable, at least early in the virus infection cycle, since this would lead to the prevention of full-length, genomic RNA synthesis. This low affinity property is likely employed as a temporal regulatory mechanism, based on an equilibrium effect. That is, the interaction is not favored until one or both components are at a sufficient concentration to drive the reaction towards complex formation. This situation occurs later in the viral replication cycle when the accumulation of both RNAs is at their greatest. This would also ensure that transcription of the capsid protein sgRNA does not occur until there is a threshold amount of both progeny viral genomic RNAs to be encapsidated by capsid protein. The direct RNA–RNA interaction provides a means of communication between the two viral RNAs to ensure that both are present in the same cell for the continuation of a productive infection.
Previous computer modeling of nucleic acids has generated structures that closely fit those structures obtained by experimental techniques. For example the model generated for the 29 packaging motor closely matched the structure obtained by biochemical studies (27). A modeling approach can be particularly valuable when working with nucleic acids that are dynamic or do not lend themselves to biochemical structural determinations. The evaluation of the two models generated in this study with MC-Annotate (19) found that the Model A family possessed key structural features consistent with a pseudoknot, in contrast to the Model B family which lacked pseudoknot features. Visual comparison of the backbone structure of Model A to RNA pseudoknot structures in the Protein Data Base confirms the general similarity (28). A more thorough comparison finds the greatest similarity to those pseudoknots involved in viral –1 ribosomal frameshifting (29). The structural similarities include (i) two stacked helical regions, (ii) 5–6 loop nucleotides base paired, (iii) G:C rich pairing in the long distance/intermolecular helical region and (iv) a limited number of unpaired loop nucleotides. The RCNMV TA TABS pseudoknot model now provides for a structural basis for interpretation of the functional activity for a series of genetic mutants in the RCNMV TA element (T.L.Sit, K.A.Turner and S.A.Lommel, unpublished data). This further supports these models as reasonable structures for the TA complex.
Alignment of the selected backbone phosphates from Model A to all pseudoknot structures in the PDB database found the greatest structural similarity to the SRV-1, –1 ribosomal frameshifting pseudoknot (30) with RMSD values of 2.0 ? (Fig. 7). This structural conservation likely also reflects functional conservation. At least one of the roles for a pseudoknot structure immediately downstream from the site of ribosomal frameshifting is to promote stalling of the advancing ribosome complex over the site of frameshifting. Similarly, the role of the RCNMV bimolecular interaction is hypothesized to periodically stall and dislodge the advancing replication complex during the process of full-length (–) strand RNA synthesis leading to a premature termination product. The 3D structural comparisons also found a significant structural similarity of the RCNMV Model A to the HIV-1 ‘kissing’ hairpin complex (Fig. 7) (11). The RMSD of selected backbone atoms was 3.2 ?. Structural similarity in the dimerization complex suggests additional conservation of form and function. Unpublished experimental data from our laboratory (V.R.Basnayake, T.L.Sit and S.A.Lommel, unpublished data) suggest that the RCNMV origin of assembly signal is located solely on RNA-2 and as a consequence, the formation of the RNA-1 and RNA-2 bimolecular complex late in the infection cycle is how RNA-1 is co-packaged into virions. While preliminary, these observations suggest a function remarkably similar to the HIV-1 dimerization complex. In essence, this RCNMV stacked helical bimolecular complex may likely possess two functions critical to the virus life cycle, which have been separately observed in two different retroviruses.
Figure 7. Comparison of the RCNMV TA model structure with the SRV-1 -1 ribosomal frameshifting pseudoknot and the HIV-1 kissing hairpin structure. Ribbon representations for backbones of: (A) SRV-1, (B) RCNMV and (C) HIV-1 dimerization element. (D) SRV-1 with RCNMV and (E) HIV-1 with RCNMV. The structures were compared by superimposing 19 backbone phosphates.
ACKNOWLEDGEMENTS
We thank Drs I. T. D. Petty and C. Hemenway for reviewing this manuscript and Carol George for her assistance in the preparation of isotopically labeled transcript. This research was supported by USDA NRI competitive grant 98-02298, NSF competitive grants (MCB-0077964) to S.A.L. and T.L.S. and (MCB-9986011) to P.F.A. M.A.D. was partly supported by an NIH grant (43237) to Dr Paul Wollenzien.
REFERENCES
Bernardi,F. and Haenni,A.L. (1998) Viruses: exquisite models for cell strategies. Biochimie, 80, 1035–1041.
Gale,M.,Jr, Tan,S.L. and Katze,M.G., (2000) Translational control of viral gene expression in eukaryotes. Microbiol. Mol. Biol. Rev., 64, 239–280.
Maia,I.G., Seron,K., Haenni,A.L. and Bernardi,F. (1996) Gene expression from viral RNA genomes. Plant Mol. Biol., 32, 367–391.
Miller,W.A., Dreher,T.W. and Hall,T.C. (1985) Synthesis of brome mosaic virus subgenomic RNA in vitro by internal initiation on (–)-sense genomic RNA. Nature, 313, 68–70.
White,K.A. (2002) The premature termination model: a possible third mechanism for subgenomic mRNA transcription in (+)-strand RNA viruses. Virology, 304, 147–154.
Eckerle,L.D. and Ball,L.A. (2002) Replication of the RNA segments of a bipartite viral genome is coordinated by a transactivating subgenomic RNA. Virology, 296, 165–176.
Kim,K.H. and Hemenway,C.L. (1999) Long-distance RNA–RNA interactions and conserved sequence elements affect potato virus X plus-strand RNA accumulation. RNA, 5, 636–645.
Zhang,G., Slowinski,V. and White,K.A. (1999) Subgenomic mRNA regulation by a distal RNA element in a (+)-strand RNA virus. RNA, 5, 550–561.
Lindenbach,B.D., Sgro,J.Y. and Ahlquist,P. (2002) Long-distance base pairing in flock house virus RNA1 regulates subgenomic RNA3 synthesis and RNA2 replication. J. Virol., 76, 3905–3919.
Sit,T.L., Vaewhongs,A.A. and Lommel,S.A. (1998) RNA-mediated trans-activation of transcription from a viral RNA. Science, 281, 829–832.
Chang,K.Y. and Tinoco,I.,Jr (1997) The structure of an RNA ‘kissing’ hairpin complex of the HIV TAR hairpin loop and its complement. J. Mol. Biol., 269, 52–66.
Mujeeb,A., Clever,J.L., Billeci,T.M., James,T.L. and Parslow,T.G. (1998) Structure of the dimer initiation complex of HIV-1 genomic RNA. Nat. Struct. Biol., 6, 432–436.
Ban,N., Nissen,P., Hansen,J., Moore,P.B. and Steitz,T.A. (2000) The complete atomic structure of the large ribosomal subunit at 2.4 ? resolution. Science, 289, 905–920.
Brunel,C., Marquet,R., Romby,P. and Ehresman,C. (2002) RNA loop–loop interactions as dynamic functional motifs. Biochimie, 84, 924–944.
Piotto,M., Saudek,V. and Sklenar,V. (1992) Gradient-tailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J. Biomol. NMR, 6, 661–665.
Gibbs,S.J. and Johnson,C.S.,Jr (1991) A PFG NMR experiment for accurate diffusion and flow studies in the presence of eddy currents. J. Magn. Reson., 93, 395–402.
Major,F., Turcotte,M., Gautheret,D., Lapalme,G., Fillion,E. and Cedergren,R. (1991) The combination of symbolic and numerical computation for three-dimensional modeling of RNA. Science, 253, 1255–1260.
Koradi,R., Billeter,M. and Wüthrich,K. (1996) MOLMOL: a program for display and analysis of macromolecular structures J. Mol. Graph., 14, 51–55.
Gendron,P., Lemieuxs,S. and Major,F. (2001) Quantitative analysis of nucleic acid three-dimensional structures. J. Mol. Biol., 308, 919–936.
Varani,G. and Tinoco,I.,Jr (1991) RNA structure and NMR spectroscopy. Q. Rev. Biophys., 24, 479–532.
Kolk,M.H., van der Graaf,M., Wijmenga,S.S., Pleij,C.W., Heus,H.A. and Hilbers,C.W. (1998) NMR structure of a classical pseudoknot: interplay of single- and double-stranded RNA. Science, 280, 434–438.
Wemmer,D.E. and Williams,P.G. (1994) Use of nuclear-magnetic resonance in probing ligand-macromolecular interactions. Methods Enzymol., 239, 737–767.
Masquida,B. and Westhof,E. (2000) On the wobble GoU and related pairs. RNA, 6, 9–15.
Szymanski,M., Barciszewska,M.Z., Erdmann,V.A. and Barciszewski,J. (2000) An analysis of G-U base pair occurrence in eukaryotic 5S rRNAs. Mol. Biol. Evol., 8, 1194–1198.
Giese,M.R., Betschart,K., Dale,T., Riley,C.K., Rowan,C., Sprouse,K.J. and Serra,M.J. (1998) Stability of RNA hairpins closed by wobble base pairs. Biochemistry, 37, 1094–1100.
Wagner,E.G. and Simons,R.W. (1994) Antisense RNA control in bacteria, phage and plasmids. Annu. Rev. Microbiol., 48, 713–742.
Hoeprich,S. and Guo,P. (2002) Computer modeling of three-dimensional structure of DNA-packaging RNA (pRNA) monomer, dimer and hexamer of Phi29 DNA packaging motor. J. Biol. Chem., 277, 20794–20803.
Berman,M., Westbrook,J., Feng,Z., Gilliland,G., Bhat,T.N., Weissig,H., Shindyalov,I.N. and Bourne,P.E. (2000) The Protein Data Bank. Nucleic Acids Res., 28, 235–242.
Hilbers,C.W., Michiels,P.J. and Heus,H.A. (1999) New developments in structure determination of pseudoknots. Biopolymers, 48, 137–153.
Michiels,P.J., Versleijen,A.A., Verlaan,P.W., Pleij,C.W., Hilbers,C.W. and Heus,H.A. (2001) Solution structure of the pseudoknot of SRV-1 RNA, involved in ribosomal frameshifting. J. Mol. Biol., 310, 1109–1123.(Richard H. Guenther, Tim L. Sit, Hanna S)