HEK 293E cells are a stably transfected HEK 293 cell line that constitutively expresses the Epstein-Barr virus nuclear antigen 1 (EBNA1) (24) and were obtained from Yves Durocher (Biotechnology Research Institute, Montreal). This cell line was adapted to grow in suspension in F17 medium (Invitrogen) supplemented by 2 mM l-glutamine and 0.1% Pluronic F-68 (Gibco), and transfected by using 25 kDa linear polyethylenimine (PEI, pH 7.0) (Polysciences Inc). EBNA1 promotes amplification of plasmid containing the replication origin region (OriP) of Epstein-Barr virus, leading to high expression of proteins encoded for by these plasmids.

Purification and characterization of wild-type RHA has been described previously (25). We follow the same procedure to purify 6×His tagged mutant ΔdsRBDs-RHA containing deletion of both dsRBD1 and dsRBD2 (from amino acids 2–250). Plasmid pRHA_?dsRBD1-2 encoding ΔdsRBDs-RHA was constructed (26) and transfected into HEK 293E cells. Forty-eight hours later, transfected cells were collected, washed with ice-cold phosphate-buffered saline and then lysed in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, 0.5% Triton X-100, 10% glycerol, protease inhibitor cocktail tablets (Roche), pH 7.4). Cell lysates were cleared by centrifugation and then incubated with Ni-nitrilotriacetic acid (NTA) agarose (Qiagen) at 4°C for 2 h to capture His-tagged proteins. After extensive washing, recombinant proteins were eluted by 250 mM imidazole solution (pH 7.4). Glutathione S-transferase (GST) was isolated from HEK 293E cells as described previously (26). The purified proteins were dialyzed against dialysis buffer (25 mM HEPES, pH 7.5; 200 mM NaCl; 10% glycerol; 4 mM DTT) and then stored at −80°C. The purity and the identity of purified protein were determined by Coomassie Brilliant Blue R250 staining and western blot analysis using anti-His, respectively. Bio-Rad protein assay reagent was used to determine protein concentration.

The 3'-terminus of oligonucleotides C30-G18 and C48 was labeled with fluorescein-5-thiosemicarbazide (FTSC) as described previously (27) with minor modification. Briefly, the oligonucleotides were incubated in a volume of 200 μl of 2.5 mM NaIO₄ and 100 mM NaOAc (pH 5.0) for 1 h on ice and then precipitated with one volume of 2-propanol. After washing with 70% ethanol, the RNA was incubated in 200 μl of 100 mM NaOAc (pH 5.0) and 0.5 mM FTSC on ice overnight. The FTSC-labeled RNA was precipitated in ethanol, dissolved in 25 mM Tris–HCl, pH 7.5, and further purified by using Microspin G-25 columns (GE Healthcare).

Briefly, 20 μl reaction mixtures containing 10 nM FTSC-labeled RNA and the indicated amounts of protein in binding buffer (10 mM Tris–HCl, pH 8.0, 50 mM KCl, 2 mM MgCl₂, 2 mM dithiothreitol, 2.5 mM NaH₂PO₄, 15 mM NaCl, 4% glycerol) were incubated in 96-well Black PS HE microplate (Molecular Devices) for half an hour. Fluorescence polarization (FP) assay were then carried out in quadruplicate on a Synergy 4 multi-mode microplate reader (BioTek) to measure the polarization change. The equilibrium dissociation constant (K_d) was obtained by fitting the binding curves using KaleidaGraph (28).

DNA and RNA oligonucleotides were purchased from IDT or Dharmacon, and Cy3 and Cy5 NHS ester from GE Healthcare. For a dye labeling of oligonucleotides, excess dyes were mixed together with DNA or RNA in 100 mM NaHCO₃, pH 8.5 and incubated overnight, and then residual dyes were removed by Biorad P-6 column or ethanol precipitation. Labeling efficiency was calculated by measuring UV–VIS spectrum, which was >90% for all labeling reactions. To generate substrate duplexes, a labeled or non-labeled ssDNA or ssRNA was mixed with a desired complementary one in an annealing buffer (10 mM Tris, pH8.0 and 100 mM NaCl), heated at 90°C for 2 min, and slowly cooled down to room temperature.

We monitored the unwinding reaction of RHA in real-time using single-molecule FRET by a home-made wide-field total internal reflection fluorescence (TIRF) microscopy. Fluorophore-labeled oligonucleotide, a substrate of RHA, was immobilized on a PEG-coated quartz surface for single-molecule FRET assays. The emission from single fluorophores by an excitation of a solid-state 532 nm laser (Spectra Physics) is collected through a water-immersion Olympus objective (60×, NA = 1.2). Then it passes through 555 nm long pass filter which rejects Rayleigh scattering of the laser, and a dichroic mirror (cutoff: 630 nm) which separates the emission by two, a green (I_D) and red emission (I_A). The separated emission is detected by an Electron Multiplying Charge Coupled Device (EMCCD) camera (Andor) with a 100-ms exposure time, which is analyzed using a custom-written IDL and MATLAB program. FRET is calculated by I_A/(I_D + I_A), and total intensity by I_D + I_A. FRET histograms were obtained by analyzing initial 10 frames of FRET traces using >5000 molecules, and FRET traces by analyzing each single molecule in real-time.

All the unwinding experiments were done in the unwinding buffer (20 mM Tris, pH 7.5, 25 mM NaCl, 3 mM MgCl₂, 1 mM DTT (Dithiothreitol) and 0.1 mg/ml BSA (Bovine Serum Albumin)) with various concentration of RHA (10–80 nM for a wild-type RHA and 40–320 nM for ΔdsRBDs-RHA) and ATP (10 μM–1 mM) in combination with 5–10 mM trolox and the oxygen scavenger system (0.5% (w/v) glucose, 100 mg/ml glucose oxidase (Sigma) and 8.8 kU/ml catalase (Calbiochem)), which stabilize the fluorophore during a single-molecule data acquisition.

RHA unwinds dsRNA from 3′ to 5′ (Supplementary Figure S1). To monitor the mechanistic details involved in RNA unwinding process, we designed an unwinding substrate with 3′ ssRNA tail. A Cy3 (green)-labeled strand was annealed to a Cy5 (red)-labeled complementary strand that is also biotinylated. The annealed substrate is immobilized to a PEG-coated quartz surface via a biotin-NeutrAvidin linkage for total internal reflection fluorescence imaging (29). Due to the dye to dye distance (25 bp), this RNA yields low FRET by itself, and a FRET increase can be expected only when unwinding occurs as the coiling of the unwound ssRNA brings the two dyes closer to each other (Figure 1A).

The FRET histograms built from over 5000 single molecule traces obtained with the RNA substrate shows low FRET peak expected from the separation between the two dyes (Figure 1B). The FRET remained unchanged when RHA (40 nM) alone was added, suggesting that RHA binding does not disrupt the duplex significantly (Supplementary Figure S2). In contrast, when RHA was added with ATP (1 mM), a population of higher FRET molecules appeared, signifying a strand separation in an ATP-dependent manner (Figure 1B). The single molecule traces revealed a repetitive nature of unwinding exhibited by RHA evidenced by successive bursts of FRET peaks (Figure 1C and D). The moment at which RHA was added is demarcated with a red arrow. RHA binding to RNA can be detected as an abrupt increase in total fluorescence intensity (Figure 1C, black arrow) since the protein binding near fluorescent dye results in an enhancement of the fluorescence (30). Therefore, the total intensity is a useful marker for discerning the moment of protein binding and dissociation. For instance, the trace in Figure 1C shows no decrease after RHA binding, suggesting a single RHA activity without dissociation. We took advantage of this total intensity to calculate the off rate as shown below.

There are two interesting features of RHA unwinding that are evident from the real-time single molecule traces. First, the unwinding involves a multi-step process, and second, this multi-step process occurs repetitively. Based on the distinct steps present in the unwinding traces, we have divided these substeps into binding (B), activation (A), unwinding (U), stalling (S) and reactivation (R) (Figure 1C). The binding (B) refers to the time delay between RHA addition and RNA binding, whereas the activation (A) indicates the lag time between the RNA binding and the start of unwinding. This is followed by the unwinding phase (U) where FRET increases gradually (shown in more detail in Supplementary Figure S3), stalling state (S) where high FRET remains before it rapidly decreases, and the reactivation period (R) which is the second lag time before RHA initiates the next unwinding cycle. The decrease in FRET after the unwinding is interpreted as the re-zipping or rewinding of the dsRNA as RHA moves back to its initial binding site. The FRET decrease is not due to the dissociation of RHA since the total intensity remains at the same level (Figure 1C). This multi-step unwinding cycle is repeated multiple times before the two RNA strands are completely separated from each other, and the RHA dissociates (Figure 1D and Supplementary Figure S4). To address whether this repetitive unwinding is an intrinsic property of RHA or an artifact arising from specific substrate design, we compared three alternate substrate arrangement. RHA exhibited the repetitive unwinding in all three cases (Supplementary Figure S5A and B), indicating that the repetitive unwinding is likely an intrinsic behavior of RHA rather than induced by a fluorophore position or immobilization of RNA. The number of repetitive unwinding ranges between 2 and 10 cycles, likely underestimated due to the photobleaching of dyes. The number of repetition may be reduced for shorter length or weakly structured RNA that may be unwound easily.

The transition rates were calculated by analyzing the dwell times collected from over one hundred unwinding events (Figure 1E and supplementary Figure S6), and show that the activation state (A) is likely the main rate limiting step in the RHA unwinding reaction. Based on these observations, we propose a mechanism of RHA unwinding in which the successive unwinding that involves slipping back of RHA followed by a reactivation step in preparation for new unwinding. This cycle is repeated multiple times before the two RNA strands are completely separated from each other, and the RHA dissociates (Figure 1F).

Next, we investigated how the kinetic rates of the five distinguishable steps in unwinding (B, A, U, S, R) were affected by varying the RHA concentrations, ATP concentrations, temperature, and dsRNA sequences. As expected, the binding (B) rate increased linearly with increasing RHA concentration, yielding an association constant of 2.61 × 10⁶ M⁻¹ s⁻¹ (Figure 2A). In contrast, the activation rate and the off rate (calculated from the total bound time) showed no dependence on RHA concentration (Figure 2A). This further supports the interpretation that a single RHA molecule is responsible for the repetitive unwinding rather than successive binding of multiple RHA molecules.

When ATP concentration was changed, neither the binding (B) nor activation (A) rates were altered (Figure 2B). However, a clear ATP dependence was observed for unwinding (Supplementary Figure S7A), yielding K_m of 26 μM ATP from Michaelis–Menten fitting (Figure 2B), which is consistent with a previous report (31). At 37°C, the rates of all the substeps were increased by 2–5-fold compared to the rates at 22°C, possibly due to the faster rate of conformational change, ATP hydrolysis and RHA diffusion at the higher temperature (Figure 2C and Supplementary Figure S7B). When two different dsRNA sequences, GC-rich (60% GC) or AT-rich (40% GC), were tested, the unwinding (U) and the stalling rate (S) showed the largest difference among all the substeps, likely due to the higher thermal stability of the GC-rich strand (Figure 2D). Taken together, we show that the protein concentration, ATP concentration and base composition each modulates a specific substep(s) selectively whereas the temperature has a more global effect (Figure 2E).

To characterize the activation step in the RNA unwinding, we tested whether preloading RHA would accelerate the activation step. When unwinding reaction was initiated with (bottom) or without (top) preloading RHA, the activation rate was considerably increased in the preloaded case, suggesting that the activation step involves a proper loading of RHA to dsRNA (Figure 3 A and B). This result also suggests that the proper RHA loading to dsRNA does not require ATP, as the preloading was performed in the absence of ATP.

Since N-terminal dsRBDs are thought to be essential for RHA to target RNA substrate (26), we tested the role of this domain in RNA loading and unwinding activity by preparing dsRBD truncation mutant (ΔdsRBDs-RHA) (Figure 3C and D). We found that the ΔdsRBDs-RHA exhibited lower binding rate than the wild-type RHA (Figure 3E), which is coupled with a higher off rate in the 10–320 nM protein concentration range (Figure 3F). The high off rate of ΔdsRBDs-RHA might be responsible for the lack of repetitive unwinding in this mutant (Figure 3D and Supplementary Figure S8A), indicating that dsRBDs is causally associated with the repetitive unwinding behavior of wild-type RHA. This suggests that the dsRBDs contribute to both the increased binding affinity as well as the stability of RHA binding to dsRNA (Figure 3E and F). We confirmed the affinity effect by the fluorescence polarization measurement (Supplementary Figure S8B).

The analysis of individual rates showed that while the binding rate is decreased, the activation rate was increased in the ΔdsRBDs-RHA and the unwinding rate remained unchanged (Figure 3G). This suggests an inhibitory role of dsRBDs for the activation, but not for unwinding step. When the inhibitory dsRBD is removed, the ΔdsRBDs-RHA initiates unwinding readily. Further, the activation rate measured for ΔdsRBDs-RHA is similar to the enhanced activation rate of preloaded wild-type RHA (Figure 3B and G), indicating the dsRBDs as the primary factor responsible for the slower activation step in the wild-type RHA. Taken together, the dsRBDs increase the binding affinity of RHA to dsRNA and stabilize its binding while negatively regulating its activation for unwinding.

Unwinding activity of RHA has been proposed to play a critical role in remodeling of complicated RNA secondary structures (11,18,32,33) in which both disruption and formation of double-stranded RNA occur. RHA has also been shown to promote the annealing of HIV-1 genomic RNA to tRNA^Lys3, the primer for reverse transcriptase in HIV (32). We therefore investigated whether the repetitive unwinding activity of RHA can assist annealing of ssRNA to dsRNA. We prepared Cy3-labeled ssRNA complementary to a Cy5-labeled dsRNA such that the annealing would be detected as an appearance of high FRET (Figure 4A). The ssRNA was designed to constitute a blunt-end duplex when annealed, to avoid the unwinding of the newly annealed duplex. In this platform, the appearance of high FRET molecules can be tabulated over time to estimate the rate of annealing (Figure 4B). When the Cy3-ssRNA was added to dsRNA in the absence of RHA and ATP, we observed no apparent annealing since the spontaneous melting of dsRNA is not expected in this condition. The addition of ssRNA and RHA without ATP resulted in some annealing, likely due to partial melting of duplex caused by RHA binding (Figure 4C, gray bar, binding-induced annealing). The most efficient annealing was achieved when ssRNA was added with RHA and ATP, suggesting that ATP-stimulated RHA unwinding facilitates the annealing process (Figure 4C, white bar, unwinding-induced annealing).

To test if the repetitive nature of unwinding can facilitate the annealing, we compared the annealing efficiency between the wild-type RHA and the ΔdsRBDs-RHA which typically unwinds only once (Figures 3D and 4C). To account for the difference of binding affinity between the wild-type RHA and ΔdsRBDs-RHA, we used higher protein concentration for ΔdsRBDs-RHA (320 nM) than for wild-type RHA (40 nM). The wild-type RHA with a stronger RNA binding affinity showed a higher binding-induced annealing than ΔdsRBDs-RHA even at a lower concentration (Figure 4C). In addition, we obtained an enhancement factor which is a ratio between the unwinding-induced annealing (with ATP) and the binding-induced annealing (without ATP). The comparison of these ratios showed a sufficiently larger enhancement factor than the ΔdsRBDs-RHA (Figure 4C), indicating that the repetitive unwinding of RHA facilitates annealing by making the unwound strand more accessible for pairing with the incoming RNA (Figure 4D).