Research Article

Structural Basis for Double-Stranded RNA Processing by Dicer

See allHide authors and affiliations

Science  13 Jan 2006:
Vol. 311, Issue 5758, pp. 195-198
DOI: 10.1126/science.1121638


The specialized ribonuclease Dicer initiates RNA interference by cleaving double-stranded RNA (dsRNA) substrates into small fragments about 25 nucleotides in length. In the crystal structure of an intact Dicer enzyme, the PAZ domain, a module that binds the end of dsRNA, is separated from the two catalytic ribonuclease III (RNase III) domains by a flat, positively charged surface. The 65 angstrom distance between the PAZ and RNase III domains matches the length spanned by 25 base pairs of RNA. Thus, Dicer itself is a molecular ruler that recognizes dsRNA and cleaves a specified distance from the helical end.

RNA interference (RNAi) is an ancient gene-silencing process that plays a fundamental role in diverse eukaryotic functions including viral defense (1), chromatin remodeling (2), genome rearrangement (3), developmental timing (4), brain morphogenesis (5), and stem cell maintenance (6). All RNAi pathways require the multidomain ribonuclease Dicer (7). Dicer first processes input dsRNA into small fragments called short interfering RNAs (siRNAs) (8), or microRNAs (miRNA) (9), which are the hallmark of RNAi. Dicer then helps load its small RNA products into large multiprotein complexes termed RNA-induced silencing complexes (RISC) (10). RISC and RISC-like complexes use the small RNAs as guides for the sequence-specific silencing of cognate genes through mRNA degradation (11), translational inhibition (12), and heterochromatin formation (13).

Dicer products are typically 21 to 25 nucleotides long, which is the ideal size for a gene silencing guide, because it is long enough to provide the sequence complexity required to uniquely specify a single gene in a eukaryotic genome. Several models have been proposed for how Dicer generates RNA fragments of this specific size (1416), but structural information is lacking. In an effort to deepen our understanding of the initiation step of RNAi, we determined the crystal structure of an intact and fully active Dicer enzyme.

Conservation of a highly active Dicer in Giardia intestinalis. We identified an open reading frame in Giardia intestinalis that encodes the PAZ and tandem RNase III domains characteristic of Dicer (7), but lacks the N-terminal DExD/H helicase, C-terminal double-stranded RNA binding domain (dsRBD), and extended interdomain regions associated with Dicer in higher eukaryotes (Fig. 1A). A recombinant form of this protein possesses robust dicing activity in vitro (Fig. 1B). The RNA fragments produced by Giardia Dicer are 25 to 27 nucleotides long, which is similar to a class of small RNAs associated with RNAi-mediated DNA elimination in Tetrahymena (17) and RNA-directed DNA methylation in plants (18). dsRNA cleavage by Giardia Dicer is magnesium-dependent, although several other divalent cations including Mn2+, Ni2+, and Co2+ also support catalytic activity (19). The presence of discrete dicing intermediates separated by intervals of ∼25 nucleotides indicates that Giardia Dicer processes dsRNA from the helical end in a fashion similar to human Dicer (20). However, in contrast to human Dicer, Giardia Dicer has a low affinity for its small RNA product (∼1 μM) (19) and displays multiple turnover kinetics (20).

Fig. 1.

Giardia encodes an active Dicer enzyme. (A) Schematic representation of the primary sequence of human and Giardia Dicers. (B) Time course of in vitro Giardia Dicer dsRNA cleavage assay. RNA product sizes were determined by comparison with RNase T1 and alkaline hydrolysis (OH) sequencing ladders (lanes 1 and 2). Dicing requires the protein (Dcr) and Mg2+ (lanes 3 and 4).

Structural overview. We determined the crystal structure of the full-length Giardia Dicer at 3.3 Å resolution (table S1). The structure reveals an elongated molecule that, when viewed from the front, takes on a shape resembling a hatchet; the RNase III domains form the blade and the PAZ domain makes up the base of the handle (Fig. 2A). The PAZ domain is directly connected to the RNase IIIa domain by a long α helix that runs through the handle of the molecule. This “connector” helix is encircled by the N-terminal residues of the protein, which form a platform domain composed of an antiparallel β sheet and three α helices. A large helical domain bridges the two RNase III domains and forms the back end of the blade. Viewing Dicer from the side reveals a contiguous flat surface that extends along one face of the molecule.

Fig. 2.

Crystal structure of Giardia Dicer. (A) Front and side view ribbon representations of Dicer showing the N-terminal platform domain (blue), the PAZ domain (orange), the connector helix (red), the RNase IIIa domain (yellow), the RNase IIIb domain (green) and the RNase-bridging domain (gray). Disordered loops are drawn as dotted lines. (B) Close-up view of the Dicer catalytic sites; conserved acidic residues (sticks); erbium metal ions (purple); and erbium anomalous difference electron density map, contoured at 20σ (blue wire mesh). Dashed lines indicate distances described in the text.

Two–metal-ion mechanism of dsRNA cleavage. The two RNase III domains of Dicer sit adjacent to each other in the blade region and form an internal heterodimer that is similar to the homodimeric structure of bacterial RNase III (fig. S1). Although previous bacterial RNase III crystal structures revealed a single catalytic metal ion in each RNase III domain (21), subsequent studies implicated two metal ions in the hydrolysis of each strand of the dsRNA (22).

During our biochemical characterization of Giardia Dicer, we noticed that the enzyme is potently inhibited by trivalent lanthanide cations such as Er3+ (19). Lanthanides often bind more tightly to cation binding sites than divalent cations do, a property previously used to identify transient Mn2+ binding sites in proteins (23). Inspection of the anomalous difference electron density map from a crystal derivatized with ErCl3 revealed a pair of Er3+ cations in the active site of each RNase III domain of Giardia Dicer (Fig. 2B). The prominent Er3+ metal (M1) in each domain resides between four strictly conserved acidic residues, which make up the previously identified Mn2+ binding site of bacterial RNase III (21). The second Er3+ binding site (M2) lies adjacent to the first, outside of the acidic residue cluster. The distances between the two Er3+ metals in the RNase IIIa and IIIb domains are ∼4.2 Å and ∼5.5 Å, respectively. These distances are similar to those previously observed in the active site of RNase H (4.1 Å) (24), avian sarcoma virus (ASV) integrase (3.6 Å) (25), the restriction enzyme EcoRV(4.2Å)(26), and the group I intron (3.9 Å) (27), all of which are thought to use a two–metal-ion mechanism of catalysis. The 17.5 Å distance between the metal-ion pairs closely matches the width of the dsRNA major groove. We also observed Mn2+ in all M1 and some M2 sites in crystals grown in high concentrations of MnCl2. Therefore, we propose that the Er3+ metals seen in Giardia Dicer denote true catalytic metal-ion binding sites and that Giardia Dicer uses a two–metalion mechanism of catalysis. Given the high level of sequence conservation throughout the RNase III family, it is likely that all RNase III enzymes, including bacterial RNase III and Drosha, contain similar catalytic metal-ion binding sites.

Structural features of the Dicer PAZ domain. The PAZ domain is an RNA binding module found in Dicers and in the Argonaute family of proteins that are core components of RISC and other siRNA- and miRNA-containing complexes. Previous studies of PAZ domains from several Argonaute proteins revealed a degenerate oligonucleotide/oligosaccharide-binding (OB) fold that specifically recognizes dsRNA ends containing a 3′ two-base overhang (2831). Superposition of the PAZ domains of Giardia Dicer and human Argonaute1 reveals that the two domains share the same overall fold and 3′ two-nucleotide RNA binding pocket (Fig. 3A).

Fig. 3.

Structural features of the Dicer PAZ domain. (A) Superposition of the Cα atoms of PAZ domains from Giardia Dicer (orange) and human Argonaute1 (white). Amino acids forming the 3′ overhang-binding pocket are shown as sticks. (B) Electrostatic surface representation of the PAZ domains of Giardia Dicer and Argonaute1 (hAGO1). Asterisks denote 3′ overhang-binding pockets. The RNA in Argonaute1 PAZ structure is drawn as green sticks.

The Dicer PAZ domain contains a large extended loop that is conserved among Dicer sequences and absent in Argonaute (fig. S1). The Dicer-specific loop dramatically changes the electrostatic potential and molecular surface surrounding the 3′ overhang-binding pocket relative to the Argonaute PAZ domain (Fig. 3B). The presence of many basic amino acid residues in the extra loop could substantially affect the way the RNA is recognized and perhaps handed off to other complexes by each family of proteins.

A model for siRNA formation. The structure of Giardia Dicer immediately suggests how Dicer enzymes specify siRNA length. Measuring from the active site of the RNase IIIa domain to the 3′ overhang-binding pocket in the PAZ domain gives a distance of ∼65 Å (Fig. 4), which matches the length of 25 dsRNA base pairs. To produce a likely model of a Dicer-dsRNA complex, the positions of the metal-ion pairs bound in each RNase III domain were used to anchor the two scissile phosphates of an ideal A-form dsRNA helix into the RNase III active sites. This placement positions the twofold symmetry element of the dsRNA coincident with the pseudo twofold symmetry axis relating the two RNase III domains, which is analogous to how restriction enzymes typically bind dsDNA substrates (32). Bacterial RNase III has been proposed to bind dsRNA in a similar fashion (16, 33). Outside of the RNase III region, the modeled dsRNA extends along a flat surface formed by the platform domain. This surface contains a large positively charged region that could interact directly with the negatively charged phosphodiester backbone of the modeled dsRNA helix. The 3′ end of the RNA duplex falls directly into the 3′ overhang-binding pocket of the PAZ domain, and the 5′ end lies adjacent to the Dicer-specific PAZ domain loop. There are exactly 25 nucleotides between the 3′ end of the helix bound to the PAZ domain and the scissile phosphate in the RNase IIIa domain.

Fig. 4.

A model for dsRNA processing by Dicer. Front and side views of a surface representation of Giardia Dicer with modeled dsRNA. Red and blue represent acidic and basic protein surface charge, respectively. Electrostatic surface potentials do not include contributions from bound metal ions. Putative catalytic metal ions are shown as green spheres. White arrows point to scissile phosphates. Asterisk denotes PAZ domain 3′ overhang-binding pocket.

Thus, Dicer is a molecular ruler that measures and cleaves ∼25 nucleotides from the end of a dsRNA. The length of the small RNAs produced by Dicer is set by the distance between the PAZ and RNase III domains, which is largely a function of the length of the connector helix. This model of dsRNA processing is consistent with the proposed architecture of human Dicer based on biochemical studies in which the RNase IIIa and IIIb domains were shown to produce the siRNA 5′ and 3′ ends, respectively (16). Furthermore, closing the ends of a dsRNA substrate by hybridization or ligation greatly diminished dicing activity (20, 34), which may explain why circular viral dsRNA is resistant to RNAi (35).

Giardia Dicer can support RNAi in fission yeast. Given that Giardia Dicer lacks some of the domains commonly associated with Dicer enzymes, most notably the N-terminal helicase, we wondered if the structure represents an intact Dicer or merely the catalytic subunit of a larger complex required for complete Dicer function in vivo. To address this question, we introduced the Giardia Dicer gene into a strain of the fission yeast Schizosaccharomyces pombe that contains a deletion of its endogenous Dicer (dcrΔ).

Like most Dicer proteins, the S. pombe Dicer contains an N-terminal helicase domain and a C-terminal dsRBD. The S. pombe dcrΔ strain is defective in RNAi and is hypersensitive to the microtubule-destabilizing drug thiobendazole (TBZ) because of chromosome missegregation (36). Plasmid expression of S. pombe dcr1+ fully rescued TBZ sensitivity of the dcrΔ cells. A partial functional rescue of TBZ sensitivity was also achieved by episomal expression of Giardia Dicer, indicating that Giardia Dicer can suppress the chromosome segregation defect (Fig. 5A). Furthermore, Giardia Dicer restores silencing of centromeric regions that are aberrantly transcribed in the dcrΔ mutant (Fig. 5B). These results demonstrate that Giardia Dicer is sufficient to function as an intact Dicer in vivo.

Fig. 5.

Giardia Dicer supports RNAi in vivo. (A) Overexpression (OE) of Giardia Dicer rescues the TBZ sensitivity of the S. pombe Dicer delete (dcrΔ). Growth was assayed by spotting 10-fold serial dilutions of cultures indicated. (B) Overexpression of Giardia Dicer restores transcriptional silencing at centromeres (cen). Transcript levels were determined by semiquantitative reverse-transcriptase polymerase chain reaction. Actin (act) served as an internal control. bp, base pair.

A conserved architecture in Dicer enzymes. Considering the structural role played by the connector helix that links the PAZ and RNase III domains (Fig. 2), we wondered whether larger Dicer proteins found in higher eukaryotes contain an analogous helix. Sequence alignment of the region directly following the PAZ domain of several evolutionarily diverse Dicer enzymes reveals a conserved pattern of hydrophobic and hydrophilic amino acids that is predicted to form a long α helix by secondary structural analysis (fig. S2). All Dicers contain a conserved proline about 11 amino acid residues from the predicted N terminus of the helix. In the crystal structure of Giardia Dicer, this proline induces a distinct kink that aids in directing the helix toward the RNase IIIa domain.

Most Dicer proteins contain a conserved region of ∼100 amino acids termed “domain of unknown function 283” (DUF283), which lies between the helicase and PAZ domains in the primary sequence. Low but consistent sequence homology between the N-terminal domain of Giardia Dicer and DUF283 (fig. S3) suggests that in the Dicers of higher eukaryotes, DUF283 forms a platform structure similar to that of Giardia Dicer.

The conserved Dicer architecture, together with the demonstration that Giardia Dicer can substitute for S. pombe Dicer in vivo, argues that the mechanism of Dicer-catalyzed dsRNA processing is conserved. Moreover, these results indicate that all Dicers evolved from a common ancestral enzyme. Because Giardia is one of the most anciently diverged members of the eukaryotic kingdom, we may consider that the earliest eukaryotic organisms had a similar Dicer enzyme and therefore were capable of RNAi-like processes. It will be of evolutionary interest to determine the cellular function of Dicer and RNAi in Giardia.

The structure of Giardia Dicer also provides new insight into eukaryotic RNase III enzymes in general. This family of enzymes performs a range of specific cellular functions involving the cleavage of dsRNA [reviewed in (37)]. The structure of Dicer illustrates how the presence of RNA binding modules, like the PAZ and platform domains, can impart a specific function to the otherwise nonspecific double-stranded RNase activity of the RNase III dimer (38). This is likely to be the structural paradigm for all eukaryotic RNase III enzymes that have specific activities and cellular functions.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article