Recognition of small interfering RNA by a viral suppressor of RNA silencing

siRNAs, which are the processed products of Dicer⁷ (an RNase III family nuclease), are RNA sequences 21–25 nucleotides (nt) long, composed of duplex (19–23 base pairs, bp) and 3′-overhang (2-nt) segments, with 5′-phosphate and 3′-hydroxyl termini. We first sought to define the siRNA structural elements recognized by the tomato bushy stunt virus (TBSV) p19 protein (Fig. 1a). In agreement with previous research⁶, p19 binds efficiently to a 19-bp siRNA with 3′-dinucleotide overhangs at both ends, but does not recognize the component single-stranded RNAs (ssRNAs) (Fig. 1c, lanes 1–6). p19 also recognizes duplex RNAs with blunt ends, though less efficiently for 21-bp relative to 19-bp duplexes (Fig. 1c, lanes 7–10). The molecular recognition process distinguishes RNA duplexes from their DNA counterparts, because the corresponding 19-bp DNA duplex with the same sequence does not form the p19 complex (Fig. 1c, lanes 11, 12). We next screened for p19 binding to siRNAs ranging in length from 16–25 bp, and observed optimal binding at the 19–21-bp duplex level (Fig. 1d, lanes 7–12), with gradual attenuation for either shorter (16–18 bp) (Fig. 1d, lanes 1–6) or longer (22–25 bp) (Fig. 1d, lanes 13–20) duplexes. This establishes that p19 recognizes RNA duplexes of a defined length, without the requirement of 3′-dinucleotide overhangs.

We have determined the structure of homodimeric p19 (residues 27–158) in complex with a 21-nt (19-bp) siRNA (Fig. 1b) by X-ray crystallography at 1.85-Å resolution (stereo view, Fig. 2a). The concave saddle-like β-sheet surface of the p19 dimer spans all 19 bp along one face of the siRNA duplex (Fig. 2b, c), thereby burying 1,600Ų (22%) of solvent-accessible surface area of the bound RNA. Both ends of the 19-bp RNA duplex are bracketed by the projecting helices H1 and H1′ of the bound p19 homodimer (Fig. 2a). Intermolecular protein–RNA contacts span the entire length of the p19 homodimer surface (green regions in Fig. 2d), with the dimer surface electrostatically positively charged (blue regions, Fig. 2e) for those patches that contact the RNA major groove.

Each monomer within the p19 homodimer contains four α-helices (H1–H4) and four β-strands (S1–S4) (Fig. 3a). The protein dimerizes through pairing of antiparallel β-strands (S4–S4′) and antiparallel α-helices (H4–H4′), thereby burying 1,300Ų of total solvent-accessible area. The homodimer adopts basically a rectangular two-layered structure from which project two symmetry-related H1 and H1′ (reading-head) helices. One layer (concave side) is comprised of a strongly curved eight-stranded antiparallel β-sheet (Fig. 3a), while the other layer consists of six helices packed on the convex side. The symmetry-related ‘reading-head’ helices H1 and H1′ are anchored through conserved side-chain interactions to the core β-sheet scaffold of the p19 homodimer (Fig. 3b).

The non-symmetric siRNA duplex is crossed by a crystallo-graphic dyad in the crystal (see Methods), resulting in two opposite orientations of the siRNA duplex in the complex. These two orientations are shown in Fig. 3c, together with colour-coding for temperature factors: blue (low) graded to red (high). The extent of mobility increases on proceeding from the protein-contacting to the solvent-exposed segments of the bound siRNA. The RNA helix basically adopts an A-form conformation, with the helix axis bending by ~40° towards the protein (Fig. 3c). The 3′ dinucleotide overhangs of the bound siRNA are not visible in the electron density map, indicating that they are disordered and not recognized by the p19 homodimer in the complex. This is consistent with our mobility shift experiments (Fig. 2c) where blunt-end duplexes bind p19 as well as their 3′ overhang-containing counterparts. Nevertheless, the 3′ overhangs may be important for other aspects of siRNA recognition and function⁸.

The p19 homodimer monitors the length of the siRNA duplex in the structure of the complex. It achieves this unique caliper-like capability by bracketing both ends of the duplex with a pair of tryptophans, which project off the amino-terminal ends of symmetry-related ‘reading head’ H1 and H1′ helices (Fig. 2a). Stacking between W39 with the 3′-terminal base and W42 with the 5′ terminal base essentially extends the terminal base pairs at either end by an additional step (Fig. 4a). Such bracketing by pairs of stacked tryptophans at either end is further anchored by the stacking of the long side chain of R43 over W39 (Fig. 4a). Importantly, W42 is universally conserved among all available p19 homologues, while W39 can be replaced by leucine, arginine or serine residues (Fig. 1a), suggestive of variable recognition of the 3′-end of the RNA duplex. There are also direct and water-mediated intermolecular contacts between backbone and side chains of amino acids that project from the β-strands (S2) and the ordered N-terminal segments, with the phosphate backbone and sugar 2′-hydroxyls positioned towards the RNA 5′-end (Fig. 4a). The ability of p19 to accommodate duplexes ranging from 19 to 21 bp could originate in structural plasticity of the distance separating ‘reading head’ helices, each of which is connected to the structured core by a short flexible loop and several side-chain interactions (Fig. 3b).

The saddle-like β-sheet surface of the p19 homodimer spans one and a half turns of helix formed by the 21-nt (19-bp) siRNA duplex. Intermolecular contacts cover the central minor groove and two adjacent partial major grooves of the RNA duplex, with Fig. 4c showing four ribonucleotides from each RNA strand crossing four β-strands (S3′–S4′–S4–S3) in the p19 dimer, with a crossing angle of ~40° between the RNA backbone and protein β-strands.

Several direct and water-mediated intermolecular contacts involving non-bridging oxygens of the phosphodiester backbone have been observed in the structure of the complex. Three long side-chain residues (K71, R115 and Q107′) form direct contacts with phosphate groups (11′, 12′–13′ and 14′, respectively), while both backbone (CO of K67 and G118 and N of G126′) and side-chain (S120) residues form water-mediated contacts with four phosphate groups (Fig. 4b).

The RNA minor groove-facing surface of the p19 central β-sheet is rich in serine and threonine (a total of ten per p19 dimer) residues (Fig. 4c). These hydroxyl-carrying residues plus four trapped water molecules (two OH groups per water molecule) form an unusual hydroxyl-group network that hydrogen-bonds with a total of six sugar 2′-OH groups lining the RNA minor groove. Serine and threonine (S113, T122, S120, S124′ and T111′) residues form direct and water-mediated intermolecular contacts with the 2′-OH groups on one RNA strand, with symmetry-related counterparts contacting the 2′-OH groups on the partner RNA strand (Fig. 4c). An additional 2′-OH group from each RNA strand is recognized by the backbone carbonyl groups of Q107/Q107′ in the complex (Fig. 4c).

The p19 amino acids involved in direct recognition of the siRNA duplex are highlighted (red triangles) in Fig. 1a, and the majority of these are highly conserved amongst p19 homologues. The sequence conservation amongst p19 homologues suggests that other members of p19 family bind siRNA in a similar way. The siRNA sugar–phosphate backbone contacted by the p19 homodimer is highlighted (in red) in Fig. 1b, with the contacts distributed towards the centre and ends of the siRNA duplex. All electrostatic and hydrogen-bonding interactions are directed towards the sugar–phosphate backbone in the p19 dimer–siRNA complex, thereby readily explaining the sequence-independent nature of the recognition. The observed stacking of tryptophan pairs over the terminal base pairs also reflects sequence-independent contacts in the complex. Moreover, the critical contribution of 2′-OH group recognition in the complex correlates with the ability of p19 to distinguish between RNA and DNA duplexes.

The principles of protein architecture and molecular recognition observed for the p19 homodimer–siRNA complex are unique and distinct from previous structures of double-stranded RNA-binding domain (dsRBD) protein–RNA duplex complexes^9,10, where the proteins bound as monomers and loop and helical residues were involved in recognition of the sugar–phosphate backbone, and to a lesser extent of the base edges along the minor groove.

siRNAs, which are the products of dsRNA cleavage by the RNase III nuclease Dicer, play key roles in the RNA silencing pathway. siRNAs, which may form transient complexes with Dicer, are known to interact with proteins within the RNA-induced silencing complex (RISC)¹¹, in a process whereby one of the siRNA strands targets complementary mRNA. siRNAs may also serve as primers for the activity of host RNA-dependent RNA polymerases (RdRP), which have been implicated in the amplification and spreading of the silencing signal^12-16. Because the molecular mechanisms underlying these interactions are currently unknown, it remains to be established whether the siRNA recognition principles defined from our studies of the p19 viral suppressor complex could govern recognition events in these other siRNA complexes. Nevertheless, our structure elegantly demonstrates how viral proteins can recognize and sequester siRNAs of defined length.

An improved understanding of viral suppression of RNA silencing in plants could have broad implications for manipulation of the RNA silencing machinery across species. Our structure of the p19 homodimer–siRNA complex provides molecular insights into siRNA-specific recognition processes and sets the stage for the rational engineering of p19 variants capable of siRNA targeting with improved affinity, thereby enhancing available tools to manipulate RNA silencing events in gene regulation.