Hepatitis Delta Antigen Requires a Flexible Quasi-Double-Stranded RNA Structure To Bind and Condense Hepatitis Delta Virus RNA in a Ribonucleoprotein Complex

ABSTRACT

The circular genome and antigenome RNAs of hepatitis delta virus (HDV) form characteristic unbranched, quasi-double-stranded RNA secondary structures in which short double-stranded helical segments are interspersed with internal loops and bulges. The ribonucleoprotein complexes (RNPs) formed by these RNAs with the virus-encoded protein hepatitis delta antigen (HDAg) perform essential roles in the viral life cycle, including viral replication and virion formation. Little is understood about the formation and structure of these complexes and how they function in these key processes. Here, the specific RNA features required for HDAg binding and the topology of the complexes formed were investigated. Selective 2′OH acylation analyzed by primer extension (SHAPE) applied to free and HDAg-bound HDV RNAs indicated that the characteristic secondary structure of the RNA is preserved when bound to HDAg. Notably, the analysis indicated that predicted unpaired positions in the RNA remained dynamic in the RNP. Analysis of the in vitro binding activity of RNAs in which internal loops and bulges were mutated and of synthetically designed RNAs demonstrated that the distinctive secondary structure, not the primary RNA sequence, is the major determinant of HDAg RNA binding specificity. Atomic force microscopy analysis of RNPs formed in vitro revealed complexes in which the HDV RNA is substantially condensed by bending or wrapping. Our results support a model in which the internal loops and bulges in HDV RNA contribute flexibility to the quasi-double-stranded structure that allows RNA bending and condensing by HDAg.

IMPORTANCE RNA-protein complexes (RNPs) formed by the hepatitis delta virus RNAs and protein, HDAg, perform critical roles in virus replication. Neither the structures of these RNPs nor the RNA features required to form them have been characterized. HDV RNA is unusual in that it forms an unbranched quasi-double-stranded structure in which short base-paired segments are interspersed with internal loops and bulges. We analyzed the role of the HDV RNA sequence and secondary structure in the formation of a minimal RNP and visualized the structure of this RNP using atomic force microscopy. Our results indicate that HDAg does not recognize the primary sequence of the RNA; rather, the principle contribution of unpaired bases in HDV RNA to HDAg binding is to allow flexibility in the unbranched quasi-double-stranded RNA structure. Visualization of RNPs by atomic force microscopy indicated that the RNA is significantly bent or condensed in the complex.

INTRODUCTION

Hepatitis delta virus (HDV) is a unique human pathogen that causes severe liver disease (1). Its distinctiveness derives from the replication and structure of the viral RNA and from the dependence of HDV on coinfection with hepatitis B virus (2), which provides the envelope protein for HDV (3–6) but does not play a direct role in HDV RNA replication (7). The circular HDV RNA genome is the smallest known to infect humans and is replicated by host RNA polymerase (8). Replication occurs through a double-rolling-circle mechanism that involves the circular reverse complement of the genome, the antigenome (9, 10). For both of these circular RNAs, one-half exhibits substantial sequence complementarity to the other half, such that they collapse into linear closed hairpin structures in which short (2 to 10 nucleotides [nt]) base-paired segments are interspersed with small bulges and internal loops but no branches (11, 12). This quasi-double-stranded RNA (dsRNA) structure has frequently been referred to as an unbranched rod or a rod-like structure. An indication of the importance of this structure for the virus is that less than half of the genome is devoted to encoding the sole viral protein, hepatitis delta antigen (HDAg); the bulk of the remainder of the RNA base pairs with the coding region to form the RNA secondary structure.

HDAg is encoded by the antigenome; thus, HDV is a negative-strand RNA virus. Although HDV is distinct from other negative-strand RNA viruses in that RNA replication is accomplished using host rather than viral RNA polymerase (8), it is similar in that both the genome and antigenome are associated with the viral nucleoprotein, HDAg, in cells (6, 8, 13). Typical of negative-strand RNA viruses, HDV RNA-protein complexes (RNPs) play essential roles in many aspects of the virus replication cycle, including RNA transport to the nucleus, RNA replication, control of RNA editing, and virion formation (8, 14–18). The characterization of these complexes remains an important goal for understanding how they function.

A major limitation in characterizing HDV RNPs has been the tendency of HDAg to bind nucleic acids, including DNA, nonspecifically (19, 20). Thus, the sequence and structural features of the HDV RNA required for HDAg binding are not well understood. We recently showed that a C-terminally deleted form of HDAg, HDAg-160, binds specifically to HDV RNAs that are predicted to form the characteristic unbranched quasi-dsRNA structure (20). Remarkably, binding requires a minimal RNA length of between 299 and 311 nt (20). Several regions of the HDV RNA genome and antigenome that are greater than this length and that form unbranched quasi-dsRNA structures were examined and all were found to bind, although with various affinities and efficiencies (20). Analysis of these RNAs did not reveal any specific sequences that might be required for binding. HDAg-160 does not bind fully dsRNA derived from HDV RNA segments that do bind, even at concentrations 1,000-fold above the K_d (dissociation constant) for HDV RNAs (21).

Even less is known about the topology of the HDV RNP. It does not appear to form a helical structure (6). The protein has been shown to form large multimers, possibly octamers, in the absence of RNA (19, 21–24). The crystal structure of a 48-amino-acid (aa) segment of the 195-aa protein showed a quartet of dimers that were configured such that there is sufficient space for threading of a quasi-dsRNA through a central hole (22). Alternatively, the presence of positively charged amino acids on the outer edges of the structure led to the suggestion that wrapping of the RNA around the protein could also be possible (22). Although the region crystallized appears to play a major role in RNA binding, other regions of the protein also contribute to binding specificity (25).

In this study, we analyze the contributions of HDV RNA sequence and structure features to binding HDAg and examine the structure of the complex formed by atomic force microscopy (AFM). We find that the numerous bulges and internal loops in the quasi-dsRNA structure are essential for binding. However, changing the identities of these positions had no effect on binding affinity; thus, the nucleotide sequences of the bulges and internal loops do not appear to be recognized by the protein. Rather, our results suggest that binding requires bending of the quasi-dsRNA and that these unpaired elements contribute to bending and flexibility of the RNA. Consistent with these observations, AFM analysis of an HDV RNA segment bound to HDAg indicates that the RNA is condensed, perhaps wrapped, in a manner that is reminiscent of the way DNA is condensed in nucleosomes.

MATERIALS AND METHODS

RNA preparation.

RNAs were transcribed in vitro from PCR products into which the T7 promoter sequence was incorporated. HDVmin (previously referred to as 311L [20]) includes the HDV sequence from nt 111 to 1484 (sequence numbering is according to Wang et al. [12]); HDV228 includes the 228 nt from 71 to 1523. These RNA sequences traverse the origin of the 1,679-nt circular RNA. For HDV(CA)₂₀, 40 nt of poly(CA) was added 5′ of the HDV sequences in both the forward and reverse primers used for the amplification of the HDV228 template. For HDVmut3, HDVmut5, HDVmut7, and HDVmut11 RNAs, site-directed mutations were incorporated in the region from positions 1522 to 1484 of HDVmin (termed the 1522-1484 region) via the reverse PCR primer. The template for HDVmut57 was amplified using forward and reverse primers, including mutations to the same 1522-1484 region and the region of RNA that is partially complementary, positions 111 to 72 (111-72 region). The RNA quasi-dsK12 was derived from sequences in the human herpesvirus 8 K12 gene; the sequence was manipulated to fold identically to HDVmin using the Mfold folding algorithm (26). The template for transcribing this RNA was amplified from a synthetic oligonucleotide (gBlocks; IDT, Coralville, IA). The sequence of quasi-dsK12 from 5′ to 3′ is GACACCAAGTGACCACCAGCGAAGACACCAGCTTGTTTCGTCATATGGCCAGGTGAGTGCGTGCAGGTCGCGTCTCTTGTGTTTCCACGTATCCAGGAGCGGTCCGACCCCAGGGGCGCAGCCTCCGACACCCCTGGAAACCGAACGGCTATCCCGTTCCCTGGTATTCCTGGGTGCGGAGGACCGCCTCCTCGGTCGGCCCTCCTGCGATACGCTGCTGAAGCCCAAGATGATACGCCCTCGTATTGCCCTTACATGCCTCTTGTGGTCGTCCTCAGGCGGTCGTCTTTGGGTGCACTTCGGTGTC. Templates for RNAs used for secondary structure analysis by selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE) were designed to include previously described RNA structure cassette sequences at the 5′ and 3′ ends (27) that were added using forward and reverse PCR primers that included these sequences.

RNAs were synthesized in vitro with T7 RNA polymerase as described previously (20). For RNAs used in electrophoretic mobility shift assays (EMSAs), transcription reactions included 500 μM ATP, GTP, and UTP, and 12 μM CTP and [α-³²P]CTP (PerkinElmer, Waltham, MA); RNAs were purified from 6% native polyacrylamide gels as previously described (20). For synthesis of RNAs with inosine or diaminopurine, transcription reaction mixtures included 500 μM ITP and/or diaminopurine (DAP) triphosphate (TriLink, San Diego, CA) in addition to the four canonical nucleotide triphosphates. RNAs were folded by dissolving in 50 mM NaCl, heating to 85°C for 2 min, and then immediately placing them in an ice water bath for at least 30 min. RNAs containing inosine and diaminopurine migrated identically to HDVmin in a native 6% polyacrylamide gel.

RNAs used for secondary-structure analysis and AFM were transcribed with 500 μM each ribonucleotide triphosphate. DNA templates were removed after transcription by treatment with Turbo DNase (Invitrogen, Carlsbad, CA). Transcription reactions were extracted with phenol chloroform; RNAs were then precipitated with ethanol and resuspended in RNase-free water containing 50 mM NaCl and 2 mM EDTA. RNAs were folded as described above and concentrated using Amicon Ultra 30K Ultracel filters (EMD Millipore, Billerica, MA). A fraction of the RNA was analyzed on a 1% agarose gel and detected by ethidium bromide staining. RNAs migrated as a single band on native polyacrylamide gels with mobilities expected for the unbranched quasi-dsRNA conformation.

Protein expression and purification.

Native his-tagged HDAg-160 was expressed and purified as previously described (20, 25). Protein purity and concentration were determined by Coomassie blue staining of proteins electrophoresed on sodium dodecyl sulfate-polyacrylamide gels and by UV absorbance, both in comparison to bovine serum albumin (BSA) standards as previously described (20, 25). HDAg monomer concentrations are reported for each experiment. Three different preparations of HDAg-160 were used for binding experiments. Among these preparations, binding activities on HDVmin RNA varied by less than 2-fold. Within any experiment comparing binding of different RNAs, the same HDAg-160 preparation was used.

SHAPE.

SHAPE was performed on HDVmin, HDVmut11, and HDVmut57 RNAs containing structure cassette sequences at the 5′ and 3′ ends using the SHAPE electrophile benzoyl cyanide (BzCN), which adds a 2′O-adduct to nucleotides able to access the proper conformation (27–29). The addition of the cassette sequences did not affect HDAg-160 binding of these RNAs when analyzed by electrophoretic mobility shift. Briefly, 1 μl of 800 mM BzCN in dimethyl sulfoxide (DMSO) was added to a 20-μl reaction mixture containing 3 to 6 pmol of RNA in 160 mM Tris, pH 8.0, 1 U/μl SuperaseIn RNase inhibitor (Invitrogen) and incubated for 1 min at 37°C. Control reaction mixtures included 1 μl DMSO without BzCN. For analysis of HDVmin RNA bound to HDAg, 3 pmol (0.3 μM) of HDVmin RNA was incubated with 30 pmol (3 μM) of HDAg-160 at 37°C for 5 min prior to addition of BzCN. EMSA analysis indicated that, under these conditions, more than 90% of the RNA was bound. Following incubation with BzCN, RNAs were extracted with phenol chloroform, purified using a RNA Clean & Concentrator-5 kit (Zymo Research, Irvine, CA) as directed, and resuspended in 6 μl 10 mM Tris, pH 8.0. A one-dye system similar to that described by Pang et al. (30) was used to detect BzCN adducts. RNAs were annealed with a primer to the 3′ structure cassette sequence (27) labeled with 6-carboxyfluorescein (6-FAM). Primer extension was performed using SuperScript III (Invitrogen) according to the manufacturer's recommendations with the following modifications to the incubation conditions: 5 min at 42°C, 30 min at 55°C, 25 min at 65°C, and 15 min at 75°C. Two sequencing ladders were generated using either 0.5 mM ddATP or 0.5 mM ddCTP in the primer extension reaction. Primer extension products were precipitated with ethanol, washed to remove excess salt, and resolved by capillary electrophoresis along with an LIZ size standard (Genewiz Fragment Analysis Service, South Plainfield, NJ).

Data analysis was based on a previously described process (30). Briefly, raw electropherograms were analyzed using PeakScanner (Applied Biosystems). The peaks at each position in the electropherogram were then integrated. For each RNA analyzed, y axis scaling to correct for loading error was performed so that the background for each primer extension reaction was normalized to that of a negative-control reaction performed on RNA that was not treated with BzCN. The factor for this correction was never greater than 2. A signal decay correction was applied to the data for each reaction as previously described (31). The peaks were aligned to a ladder created from two sequencing reactions. At each position, the peak area of the negative control was subtracted from the peak area in BzCN-treated samples; these values were then converted to normalized SHAPE reactivities by dividing the subtracted peak areas by the average of the highest 2% to 10% of the subtracted peak areas (32, 33). The same normalization factor was applied to the SHAPE data obtained for RNA bound to protein. Two independent replicates were performed for each condition.

Electrophoretic mobility shift assays.

Electrophoretic mobility shift assays were performed as previously described (25). Mutated and substituted RNAs were always compared to a wild-type or unmodified RNA analyzed at the same time with the same protein preparation to control for any variations in binding activity in different protein preparations. The data shown in each graph include several independent replicates. Percent binding was calculated as the signal of the bound RNA band divided by the total signal in the lane. Data in binding curves was fitted in GraphPad Prism using the one-site binding hyperbola model: percent bound = (B_max × [HDAg])/(K_d + [HDAg]) (Bmax is the maximum specific binding to the RNA). To the extent possible, binding titrations were extended to concentrations that saturated binding. Some data points are not shown, as indicated in the figure legends. R² values were 0.9 or higher for all binding curves except for HDVmut11 substituted with DAP, which was bound very poorly.

Atomic force microscopy.

497L RNA was analyzed either alone or after incubating with HDAg-160. Binding reactions were carried out with 0.1 μM 497L RNA and 0.5 μM HDAg-160 for 5 min at 37°C in 80 mM NaCl, 10 mM Tris, pH 7.5. Freshly cleaved muscovite mica was incubated in a mixture of a 1-(3-aminopropyl)silatrane (APS) solution for 30 min to prepare APS-mica as described previously (34). Sample droplets (5 μl) were deposited on APS-mica for 2 min and then washed with deionized water and dried with nitrogen gas. The mica was attached to a metal disc with double-stick tape for imaging. Images were acquired in tapping mode in the air using a MultiMode SPM nanoscope IIIa system (Bruker/Digital Instruments, Santa Barbara, CA). Silicon tapping mode probes (Hi'Res DP14; MicroMasch, Estonia) with a curvature radius on the apex of 1 nm were used. The nominal spring constant was ∼5.0 N/m, and the resonant frequency was ∼160 Hz.

RNA contour length measurements were made by tracing the image RNA backbone using the curve tool and obtaining the readout from the microscope image module (Digital Instruments, Santa Barbara, CA). The height of the complexes was measured using the section tool from the image module. Perpendicular cross-sections were made to obtain the width in two dimensions: the height of the protein was measured by the difference in the intensity of the complex compared to the background noise, and the widths were measured at half the maximal complex height.

RESULTS

HDVmin RNA forms an unbranched quasi-double-stranded secondary structure that is preserved during HDAg binding.

In order to determine the specific RNA sequence and structural determinants for binding HDV RNA by HDAg, we focused on a minimal-length 311-nt RNA, HDVmin (20). This RNA is derived from the left end of the HDV antigenome (Fig. 1A) and binds HDAg-160 with low-nanomolar affinity to form discrete complexes that can be detected by electrophoresis in native polyacrylamide gels (20). The most energetically stable secondary structure predicted for HDVmin using the program RNAstructure (35) is the unbranched quasi-dsRNA structure characteristic of HDV RNA. However, this structure has not been specifically confirmed; other HDV RNA segments have been shown to form alternative branched structures (36–38). Therefore, we determined the secondary structure of HDVmin using SHAPE (28, 39), which assesses local backbone flexibility in RNA at single-nucleotide resolution (27). The reactivity of base positions to the SHAPE electrophile is related to secondary structure: base-paired positions are weakly reactive, while unpaired positions are more highly reactive.

FIG 1.

SHAPE analysis of HDVmin RNA either alone or bound to HDAg-160. (A) Schematic showing the HDV RNA antigenome and the HDVmin RNA segment. The HDAg open reading frame is oriented left to right. (B) Box plots of SHAPE reactivities of HDVmin RNA positions predicted by the RNAstructure folding algorithm to be paired or unpaired in the secondary structure (32, 35). Boxes show the median, 25th quartile, and 75th quartile reactivities. Whiskers show 1.5 times the interquartile range in both directions. Outliers are shown as closed circles; for RNA in the presence of HDAg, the reactivities of 2 positions were outside bounds of the figure. P values were determined with a Mann-Whitney U test. (C) Secondary structures predicted for HDVmin using RNAstructure with SHAPE reactivities included as a pseudo-free-energy term, indicating the observed SHAPE reactivity of each nucleotide position (32). Unreactive nucleotides are shown in black, moderately reactive nucleotides in green and orange, and highly reactive nucleotides are in red. Positions not analyzed due to high background are shown in gray. The predicted secondary structures were identical for HDVmin alone and bound to HDAg-160.

Incorporation of SHAPE reactivities as pseudo-free-energy terms in RNAstructure (30, 32, 35) produced an unbranched quasi-dsRNA secondary structure (Fig. 1). Structures with alternative branching helices or large regions of single-stranded RNA were not supported. The SHAPE reactivities of predicted paired positions were significantly (P < 0.0001) lower than those of predicted unpaired positions (Fig. 1B). A map of SHAPE reactivities overlaid on the secondary-structure prediction (Fig. 1C) also indicates that the SHAPE reactivities are consistent with the predicted unbranched quasi-dsRNA structure. The predicted structure contains 122 Watson-Crick base pairs, 19 bulges comprising 23 nt, 10 internal loops comprising 37 nt, and a 5-nt terminal loop. Of the 23 nucleotides in predicted bulge positions, 17 were reactive with BzCN, as were 34 of the 37 bases in internal loops and all 5 bases in the terminal loop. The high SHAPE reactivities of four predicted paired positions in the A-U-rich helical stem at the loop end of HDVmin indicate that this region is readily opened. A similar conclusion can be drawn from results obtained by Beard et al. using S1 nuclease (40). Likewise, the high SHAPE reactivity in the 6-bp helix at the open end of the RNA indicate that this A-U-rich helix is less stable. Outside these two regions, only 9 predicted paired positions exhibited high SHAPE reactivity; 5 of these were at helix ends, which are less stable. Overall, our results support the conclusion that HDVmin RNA forms an extended quasi-dsRNA hairpin containing numerous internal loops and bulges.

SHAPE was also performed after incubating HDVmin RNA with HDAg-160 to determine whether protein binding altered the RNA secondary structure. We observed that the effects of protein binding on the overall RNA secondary structure were limited; indeed, the secondary structure predicted by RNAstructure was unchanged (Fig. 1C). Nevertheless, there were discernible effects of HDAg-160 binding on the SHAPE reactivity that indicate either subtle changes in conformation of individual bases and/or decreased structural heterogeneity in the RNA population. In the presence of HDAg-160, 51 positions exhibited decreased SHAPE reactivity; all but 2 of these positions were in predicted base-paired segments. On the other hand, 66 positions exhibited increased reactivity upon HDAg-160 binding; all but 4 of these were located either at helical ends or at positions predicted to be unpaired. One predicted unpaired position, U1583, exhibited a remarkably high SHAPE reactivity of 16.2, indicating extensive overlap between the conformations accessible to this nucleotide and those that are reactive with the SHAPE electrophile BzCN (29). The changes in SHAPE reactivity in the presence of the protein were also evident in median SHAPE reactivities, which, upon protein binding, decreased for predicted paired positions and increased for predicted unpaired positions (Fig. 1B). The 2.6-fold increase in median SHAPE reactivity for predicted unpaired positions was highly statistically significant (P < 0.0001). The high SHAPE reactivity of unpaired positions when bound by HDAg indicates that these positions are not constrained by direct contacts with the protein (31). Overall, the SHAPE results indicate that HDAg-160 recognizes the unbranched quasi-dsRNA secondary structure of HDVmin RNA during RNP formation.

Changes to HDV RNA secondary structure have dramatic effects on RNP formation.

To examine the contributions of secondary structures in HDVmin RNA to binding HDAg, we mutated sequences in a region of the RNA (positions 111 to 72 and 1522 to 1484; termed 111-72/1522-1484) that includes 6 single- or double-nucleotide bulges and 1 seven-nucleotide internal loop (Fig. 2A). The 79 nt in this region comprise 25% of the 311-nt HDVmin RNA. It is important to note that this region does not make uniquely important contributions to binding but is required in the context of HDVmin. Previous deletion analysis demonstrated that longer RNAs from which this segment of the structure was deleted bound HDAg-160, and RNAs containing this region but that were less than the minimal length did not bind (20).

FIG 2.

Effects of mutations that alter HDVmin RNA secondary structure on HDAg-160 binding. (A) Schematic of the HDVmin RNA secondary structure above enlargements of the region mutated showing base changes introduced and the predicted effects of these changes on the RNA secondary structure. Mutations are shown in red; insertions are denoted by caret symbols, and deletions are denoted by open triangles. Vertical lines indicate Watson-Crick base pairs; dots denote G-U wobble pairs, and nucleotides above or below text lines indicate unpaired positions. (B) Electrophoretic mobility shift assay of HDAg-160 binding 5.2 pM HDVmin and HDV(CA)₂₀ RNAs. HDAg-160 concentrations (nM) are labeled beneath each lane; the concentration range for HDV(CA)₂₀ was 8-fold higher than that for HDVmin. The mobilities of free RNAs (open circles) and bound RNAs (filled circles) are indicated. (C and D) Effects of increased (C) and decreased (D) base pairing on HDVmin binding by HDAg-160. RNAs are labeled to the right of the corresponding binding curve. Error bars shown are from at least 4 independent replicates. HDVmin, black squares and solid black line. (C) HDVmut11, black X's and dashed black line; HDV228, black circles and dash-dot black line; HDVmut3, open gray circles and dash-dot gray line; HDVmut7, open gray squares and gray dotted line. (D) HDVmut5, open black circles and dashed black line. For HDVmin, HDVmut11, and HDV228, curves were fitted using additional data points at 25.6 nM and 51.2 nM HDAg, which are not shown on the graph.

To first determine the extent to which the sequence and structure of the 111-72/1522-1484 region of the RNA structure contribute to HDVmin binding, we replaced the 111-72 (40 nt) and 1522-1484 (39 nt) segments of the linear RNA with 20 repeats of the dinucleotide CA, (CA)₂₀, in the RNA HDV(CA)₂₀. This 312-nt RNA is predicted to form a 228-nt unbranched quasi-dsRNA structure with 40-nt single-stranded unstructured (CA)₂₀ extensions at the 5′ and 3′ ends. Due to these unstructured sequences, HDV(CA)₂₀ RNA migrates with about half the mobility of HDVmin in a native polyacrylamide gel (Fig. 2B). HDV(CA)₂₀ RNA exhibited no detectable binding of HDAg-160 up to concentrations as high as 102.4 nM (Fig. 2B), approximately 100-fold above the K_d for binding to HDVmin RNA. Furthermore, removal of these 79 nt from the HDV228 RNA sharply reduced binding to HDAg-160 (Fig. 2C). This result is consistent with our previous observations that unbranched quasi-double-stranded HDV RNAs of 298 nt and shorter do not bind well to HDAg (20). Overall, these results indicate that the sequence and/or secondary structure in the 111-72/1522-1484 region make critical contributions to binding HDAg.

To determine the roles of the unpaired sequences in the 111-72/1522-1484 region of the structure, we created the RNA HDVmut11, which contains 11 mutations that remove the internal loop and all 6 bulges to produce an extended 45-bp dsRNA segment (Fig. 2A). The secondary structure for HDVmut11 was confirmed with SHAPE. Binding of HDVmut11 RNA is indistinguishable from that of the deletion mutant HDV228 (Fig. 2C). Thus, mutation of this 79-nt region such that it is fully double stranded yields an RNA for which binding is no better than when the region is deleted. This result indicates that the 45-bp double-stranded region in HDVmut11 does not significantly contribute to the binding of this RNA to HDAg-160.

In order to determine the contributions of the internal loop and bulges in the 111-72/1522-1484 region to binding, we created two additional mutated RNAs, HDVmut3 and HDVmut7, in which the predicted structure is changed such that either the regions forming 5 of the bulges or the 7-nt internal loop, respectively, became base paired (Fig. 2A). HDVmut3 was created by 3 mutations that replaced the 7-nt internal loop with 3 paired bases in the midst of 10 consecutive base pairs. HDVmut7 was created by 7 mutations that removed 5 bulges and created 35 consecutive base pairs. Both mutations decreased binding similarly, resulting in dissociation constants 3- to 5-fold higher than that of the wild-type RNA HDVmin (Fig. 2C). Thus, both the 7-nt internal loop and the bulges at this end of HDVmin contribute to HDAg-160 binding.

RNP formation is not sequence specific.

To determine the contributions of the nucleotide sequence and secondary structure of HDVmin to RNP formation, we designed the mutated RNA HDVmut57, in which 57 of the 79 nt (72%) in the 111-72/1522-1484 region were changed but the predicted secondary structure was maintained (Fig. 3). The sequence changes were introduced into both paired and unpaired positions. Of the 32 Watson-Crick base pairs in the 111-72/1522-1484 region of HDVmin, all but 6 were changed in HDVmut57; of the 15 predicted unpaired positions, all but 2 were changed to other bases, and both of these unchanged positions were in the 7-nt loop. The predicted secondary structure of HDVmut57 was verified by SHAPE. Despite the extensive sequence modifications, the affinity of HDAg-160 binding to HDVmut57 was indistinguishable from that for binding to wild-type HDVmin RNA (Fig. 3B). Thus, the primary sequence of this region of the RNA does not play a critical role in binding HDAg, except for its ability to form an unbranched quasi-dsRNA structure with intermittent bulges and internal loops.

FIG 3.

Primary nucleotide sequence is not recognized by HDAg-160. (A) Diagram of HDVmin and two RNAs predicted to form the same secondary structure. HDVmut57 was derived from HDVmin and mutated at 57 positions (indicated in red); the region mutated is enlarged. Quasi-dsK12 is a 311-nt RNA derived from sequences in the HHV8 K12 gene; the sequence was manipulated to form the same secondary structure as HDVmin. Overall, the sequence identities of HDVmin and quasi-dsK12 are less than 15%, and at unpaired positions they are only 6.5%. The predicted secondary structure of quasi-dsK12 is depicted schematically in red to indicate the sequence divergence. (B and C) Binding curves for HDAg-160 binding HDVmut57 and quasi-dsK12. RNAs are labeled to the right of the corresponding curve. HDVmin is represented by closed black squares and a solid line; HDVmut57 and quasi-dsK12 are represented by open black circles and dashed lines. Error bars indicate standard deviations obtained from at least 3 independent replicates for each curve. (C) For HDVmin and quasi-dsK12, curves were fitted using an additional data point at 51.2 nM HDAg (not shown on the graph). (D) Electrophoretic mobility shift assay of HDAg-160 binding to 5.2 pM quasi-dsK12 RNA. HDAg-160 concentrations are labeled beneath each lane. Free RNAs (open circles) and bound RNAs (filled circles) are indicated.

We extended the mutational analysis from 25% of the minimum-length RNA to the entire minimum length by designing quasi-dsK12, a 311-nt RNA of unrelated sequence that is predicted to form an unbranched quasi-dsRNA structure with bulges and internal loops in the same locations as in HDVmin but in which the primary sequence is unrelated to HDVmin. In order to reduce potential effects of the stability of the secondary structure on binding, we designed quasi-dsK12 such that the predicted free energy of the unbranched quasi-dsRNA secondary structure is identical to that of HDVmin (−160 kcal/mol). Overall, quasi-dsK12 is only 14.4% identical to HDVmin at the sequence level; the identity rate among the 62 predicted unpaired positions is even lower, 6.5%. Remarkably, we observed that this unbranched quasi-dsRNA is bound by HDAg-160 with the same affinity as HDVmin (Fig. 3). The slightly reduced electrophoretic mobility of HDAg complexes formed with quasi-dsK12 RNA, as well as the more diffuse appearance of the bound RNA, indicate that complexes formed on this RNA vary from those formed on HDVmin and are more heterogeneous as well. Nevertheless, these results indicate that HDAg recognizes the secondary structure of the RNA but not its primary sequence.

RNA flexibility contributes to RNP formation.

The observations that removal of the internal loop and six bulges in the 111-72/1522-1484 region decreased binding (Fig. 2C) but that the sequence of the bulge and loop positions could be changed without affecting binding (Fig. 3) raise the question of how these structures contribute to RNP formation. We considered that the RNA binding properties of HDAg-160 are reminiscent of nucleic acid binding proteins that bend dsDNA (i.e., histones, eukaryotic architectural proteins such as HMG-D, and prokaryotic histone-like proteins such as FIS). These proteins do not have strict requirements for primary nucleic acid sequence, but binding affinity is affected by the flexibility of the nucleic acid (41, 42). For example, binding activity of these proteins is increased by insertion of internal loops and bulges into dsDNA sequences (43, 44). It is important to note that the B-form helix formed by double-stranded DNA is more flexible (i.e., has a lower-persistence length) than the A-form helix formed by completely double-stranded RNA (45). Thus, there may be a need for unpaired regions in the mostly double-stranded HDV RNA to achieve the same degree of bending and compaction as that of DNA in, for example, nucleosomes. To address the possibility that HDAg RNA binding is sensitive to flexibility in the quasi-dsRNA secondary structure, we created an additional mutant, HDVmut5, in which the 111-72/1522-1484 region was modified by 5 sequence insertions and substitutions such that the predicted structure includes 5 internal loops and 2 single-nucleotide bulges with no more than 4 consecutive base pairs (Fig. 2A). In contrast to the effects of increasing the double-stranded character of this region on RNP formation, these mutations, which are predicted to increase RNA bending and flexibility, increased binding affinity 4-fold (Fig. 2D).

The dsDNA binding activities of histones HMG-D and FIS have also been shown to be strengthened by incorporation of inosine, which decreases the stability of base pairs and increases flexibility of dsDNA, and weakened by diaminopurine, which increases the stability of the helix and decreases dsDNA flexibility (41, 42, 46, 47). Therefore, we examined the effects of incorporating inosine or diaminopurine on the binding of HDAg-160 to HDVmin. Similar to the reported effects of these nucleotide substitutions on nucleosome formation (41, 47), we observed that incorporation of inosine increased the affinity of HDAg for HDVmin RNA 3-fold; conversely, adding diaminopurine decreased binding 3-fold (Fig. 4A). Incorporation of both inosine and diaminopurine yielded binding activity similar to that of HDVmin RNA (not shown). We also observed that binding to HDVmut11, in which 25% of the RNA is fully double stranded, was increased by incorporation of inosine, although not to wild-type levels (Fig. 4B). Thus, the inosine substitution was able to partially counteract the negative effects of the extensive dsRNA segment in this RNA. Incorporation of diaminopurine into HDVmut11 resulted in further suppression of binding. These results are consistent with the interpretation that the flexibility of HDV RNA plays a critical role in binding HDAg.

FIG 4.

HDAg-160 binding to HDVmin and HDVmut11 is affected by base substitutions that alter RNA flexibility. (A and B) Binding curves of HDAg-160 binding 5.2 pM HDVmin and HDVmut11 RNAs synthesized with either inosine (I) or diaminopurine (DAP). RNAs are labeled to the right of the corresponding curve. HDVmin and HDVmut11 RNAs, black squares and solid lines; I-substituted RNAs, open black circles with dashed lines; DAP-substituted RNAs, black triangles with dotted lines. (A) For HDVmin and HDVmin-DAP, curves were fitted using two additional data points at 0.5 nM and 51.2 nM HDAg, which are not shown on the graph.

HDV RNA is wrapped or bent in complexes formed with HDAg-160.

The binding experiments shown in Fig. 2 and 4 suggest that the HDV RNA must be flexible or bent in order for binding to occur. To more directly determine the overall structure of the HDV RNP and investigate the topological mode of RNA binding, we examined HDV RNPs by AFM. For this analysis, we used the RNA 497L, a 497-nt RNA from the HDV antigenome that is longer than the minimum length and binds a single HDAg-160 complex (21). 497L RNA is derived from the same region of the HDV RNA as HDVmin and is also predicted to form an unbranched quasi-dsRNA secondary structure. AFM analysis of 497L RNA alone revealed molecules that were straight, curved, or bent. No branched structures were observed. The median contour length of 497L RNAs was 66.0 ± 3.9 nm (n = 59), similar to the predicted length of 67.2 to 69.7 nm for an A-form helix formed by a 498-nt dsRNA (48, 49). Thus, in agreement with the SHAPE results presented in Fig. 1, the AFM analysis suggests an unbranched quasi-dsRNA secondary structure for HDV RNA.

Analysis of RNA-protein complexes formed by incubating 497L RNA with HDAg-160 revealed structures with one (n = 45) or two (n = 14) protruding RNA tails that were substantially shorter than those of 497L RNA alone. These tails are consistent with the fact that the 497L RNA is longer than the minimum RNA length for HDAg binding but not long enough to bind a second protein complex (21). Assuming a constant ratio of RNA length per nucleotide, the length of these tails (median total length, 30.9 ± 11.7 nm) indicates that 233 nt of the 497L RNA extend away from the complex and, by subtraction, that 264 ± 90 nt of the RNA are closely associated with HDAg-160. These observations are consistent with the minimum RNA length for RNP formation of between 299 and 311 nt and with our previous observation that micrococcal nuclease treatment of RNPs formed with RNAs of lengths greater than that of HDVmin increased the mobility of these complexes to that of RNPs formed with HDVmin (20).

The topology of HDV RNA bound to HDAg is not known. Based on the crystal structure of HDAg amino acids 12 to 60, two possibilities were suggested, one in which the RNA is threaded through the center of an HDAg octamer and another in which the RNA is wrapped around the protein (22). In order to gain insight into the topology of the bound portion of the RNA in RNPs, we compared the contour length of the isolated RNA with the contour length of the RNA bound to HDAg-160. Although it is not possible to clearly distinguish the RNA from the protein in the RNP complex, we reasoned that if the RNA were threaded straight through the protein, then the putative length of RNA bound would be equal to the diameter of the complex and the putative total length would be the sum of the diameter and the length of the unbound tail (Fig. 5C). Thus, assuming straight threading of RNA through the protein, we determined that the median RNA length would be 48.2 ± 12.0 nm (Fig. 5C). This distance is significantly shorter than the length of the RNA in the absence of HDAg (P < 0.0001), indicating that the RNA is not simply threaded through the protein. The 17.8-nm difference between these two measurements indicates that the RNA is wrapped or otherwise condensed through its interaction with HDAg-160. This wrapping or condensation agrees with the observed requirement for flexibility in the RNA in order for binding to occur.

FIG 5.

AFM analysis of 497L HDV RNA bound to HDAg-160 indicates that the RNA is condensed during RNP formation. (A) Representative AFM image of 497L RNA incubated with HDAg-160 for 5 min at 37°C. Several unbound RNAs are visible, as is an RNP complex with an RNA tail. (B) Representative images of complexes formed by HDAg-160 and 497L RNA. Over 75% of the 59 complexes analyzed had one RNA tail; the remainder had two tails. (A and B) Coloring in the AFM images indicates height according to the scale on the right in panel A. (C) Distribution of the total RNA length (L_T) for unbound RNA (measured; open bars) and RNA complexed with HDAg-160 (putative; black bars). The putative total lengths of the bound RNAs were determined by adding the measured length(s) of the RNA tail(s) and the measured diameter of the complex; this measurement assumes direct threading of the RNA through the complex. A P value was determined by a Mann-Whitney U test comparing the means for L_T between the complexed 497L RNA and the 497L RNA not incubated with HDAg-160. The lack of concordance indicates that the RNA is not threaded through the complex.

DISCUSSION

The RNPs formed by HDAg and the ∼1,680-nt HDV genome and antigenome RNAs play critical roles in the virus replication cycle. However, little is known about the structures of these complexes, primarily because the length of the RNA and the tendency of the protein to bind indiscriminately to nucleic acids have complicated analysis. We have recently shown that HDAg-160, a 35-aa truncation of the 195-aa genotype 1 HDAg, can be used to circumvent nonspecific nucleic acid binding and aggregation. With this protein, we have shown that HDAg binds as multimers of fixed size to unbranched quasi-double-stranded HDV RNA segments to form discrete nuclease-resistant complexes that can be readily observed on native polyacrylamide gels (20, 21). Moreover, binding exhibits a remarkable length requirement of about 311 nt (20). We have also used HDAg-160 and a 395-nt segment of HDV RNA to show that the amino-terminal region of HDAg is closely associated with the RNA and that complex formation likely involves contact with many amino acids (25). Recent work has demonstrated that these results are not artifacts due to the use of a C-terminally truncated protein or the N-terminal His₆ tag: both HDAg-160 with an N-terminal 3×-FLAG epitope tag instead of the His₆ tag and a His₆-tagged, full-length protein from another HDV genotype exhibited the same RNA length and secondary-structure binding specificities as HDAg-160 (B. Griffin and J. Casey, unpublished).

Here, we have used HDAg-160 and two HDV RNA segments that bind a single HDAg-160 multimer to analyze the RNA structural basis for HDV RNP formation and the topology of the complexes formed. We have found that the unbranched quasi-dsRNA secondary structure predicted for HDV RNA is the major determinant of HDAg binding. Remarkably, an RNA synthetically designed to form the same secondary structure as the 311-nt HDVmin RNA segment but which differed by 85% at the primary sequence level bound to HDAg-160 with affinity indistinguishable from that of HDVmin (Fig. 3). Thus, the protein does not recognize the primary RNA sequence, which contributes only indirectly to binding in that it forms the required unbranched quasi-dsRNA structure. This result is consistent with previous findings that even though HDAg can bind to several ∼400-nt quasi-double-stranded regions of the HDV genome and antigenome, no shared sequence motifs for these RNA segments were identified (20).

Many RNA-binding proteins recognize their RNA targets by interacting with the primary sequence of single-stranded elements similar to the internal loops and bulges that separate the numerous short (2- to 11-bp) dsRNA helices in HDV RNA. Because removal of these bulges and internal loops substantially reduced HDAg-160 binding but changing the primary sequence of these elements did not (Fig. 2 and 3), it is likely that the principal role of these unpaired positions is to allow for bending of the quasi-dsRNA to fit the shape of HDAg rather than to make base-specific contacts with the protein. The increased SHAPE reactivity of unpaired nucleotides (Fig. 1) is consistent with the lack of sequence-specific recognition by the protein. In contrast to the observed increase in reactivity, such specific interactions would be expected to hinder the dynamic mobility, and decrease SHAPE reactivity, of nucleotide positions that are directly involved (29). The effects of substitutions that are expected to either increase or decrease the flexibility of double-stranded regions (Fig. 4) also support the conclusion that flexibility and bending of the RNA is required for binding. Bending of the RNA upon binding is consistent with RNP complexes observed by AFM (Fig. 5); the HDV RNA is clearly and significantly condensed in these complexes. It is important to note that the bending of the RNA in the complex is not accompanied by apparent changes in the overall secondary structure of the RNA (Fig. 1).

The slight decrease in the SHAPE reactivities of predicted base-paired positions, along with the increased reactivities of predicted unpaired positions, indicate that HDAg stabilizes the unbranched quasi-double-stranded RNA structure, resulting in a more structurally homogenous population of RNAs when bound to HDAg than when unbound. The increased SHAPE reactivity of predicted unpaired positions suggests that, when bound to HDAg, many of these positions remain structurally accessible to cellular factors that could participate in replicative processes. Similarly, the decreased stability of the helix at the left end of the rod in the absence of HDAg observed in our SHAPE experiments and in S1 nuclease studies by Beard et al. (40) could leave that end of the RNA more amenable to interaction with host polymerase or other binding partners, as previously suggested (50, 51).

In light of our results, we consider that the previous use of the terms “unbranched rod” or “rod-like” to describe the structure of the HDV RNA imply too strongly that the RNA is rigid and instead have referred to the structure as an unbranched quasi-dsRNA. Indeed, previous studies using electron microscopy (52), as well as the AFM studies described here, show that HDV RNA appears extended and unbranched, but not inflexible; individual molecules exhibit various degrees of curvature. Considering the diameters measured for the RNP complexes, the length of RNA associated with the HDAg-160 multimer in the RNP observed by AFM is consistent with a simple model in which the RNA is wrapped approximately once around the outside of the protein. However, it is not possible to directly visualize the RNA in these complexes; thus, determination of the specific structure of the RNA and protein in them will require additional approaches. The functional significance of bending or condensing the RNA is obvious for packaging the genome into viral particles, which have an external diameter less than half that of HDVmin RNA, a segment less than one-fifth the length of the full genome. Whether bending of the RNA has additional functional significance beyond RNP formation remains to be determined.

The observation that HDAg does not recognize the primary sequence of HDV RNA in order to bind suggests that the protein is able to bind cellular RNAs that form similar structures. In this regard, Han et al. observed that, in the absence of HDV RNA, HDAg accumulated in the nucleolus of transfected cells, perhaps by association with rRNA precursors (53). It remains to be determined whether such interactions occur during infection and whether they contribute to virus replication or pathogenesis.

The lack of direct sequence recognition for RNA binding likely allows for HDAg to interact with different regions along the HDV RNA or even roll along it. However, it is important to note that HDAg binding preferences appear to be more complex than a simple requirement for double-stranded helical segments interspersed with bulges and internal loops. We have found that potato spindle tuber viroid RNA, which also forms an unbranched quasi-dsRNA secondary structure, is bound by HDAg-160 with 5-fold lower affinity than HDV RNA (B.G. and J.C., unpublished). Moreover, binding varies for different regions of the HDV RNA (20). Thus, just as some dsDNA binding proteins that bend DNA exhibit indirect readout of DNA sequences that affect dsDNA flexibility, perhaps HDAg is sensitive to variations in the abilities of different regions of the HDV RNA to form the proper shape for efficient binding. Such variations could lead to preferential binding to some regions over others, providing mechanisms for functional activities, including, perhaps, assembly of multiple HDAg complexes on genome-length RNAs via ordered sequential binding. Determining the structural foundations for such variations in binding activities and their roles in the functional activity of the RNP are goals for future analysis.