Proteomics Identification of Drosophila Small Interfering RNA-associated Factors*

The Drosophila melanogaster RNA-induced silencing complex (RISC) forms a large ribonucleoprotein particle on small interfering RNAs (siRNAs) and catalyzes target mRNA cleavage during RNA interference (RNAi). Dicer-2, R2D2, Loquacious, and Argonaute-2 are examples of RISC-associated factors that are involved in RNAi. Holo-RISC is an ∼80 S small interfering ribonucleoprotein, which suggests that there are many additional proteins that participate in the RNAi pathway. In this study, we used siRNA affinity capture combined with mass spectrometry to identify novel components of the Drosophila RNAi machinery. Our study identified both established RISC components and novel siRNA-associated factors, many of which contain domains that are consistent with potential roles in RNAi. Functional analysis of these novel siRNA-associated proteins suggests that these factors may play an important role in RNAi.

The Drosophila melanogaster RNA-induced silencing complex (RISC) forms a large ribonucleoprotein particle on small interfering RNAs (siRNAs) and catalyzes target mRNA cleavage during RNA interference (RNAi). Dicer-2, R2D2, Loquacious, and Argonaute-2 are examples of RISC-associated factors that are involved in RNAi. Holo-RISC is an ϳ80 S small interfering ribonucleoprotein, which suggests that there are many additional proteins that participate in the RNAi pathway. In this study, we used siRNA affinity capture combined with mass spectrometry to identify novel components of the Drosophila RNAi machinery. Our study identified both established RISC components and novel siRNAassociated factors, many of which contain domains that are consistent with potential roles in RNAi. Functional analysis of these novel siRNA-associated proteins suggests that these factors may play an important role in RNAi.

Molecular & Cellular Proteomics 9:1866 -1872, 2010.
Small RNAs can regulate gene expression through a collection of mechanisms broadly termed RNA silencing. Small RNA-mediated silencing mechanisms occur in most species (1)(2)(3)(4)(5). The ability to silence the expression of specific genes using small RNAs via RNA interference (RNAi) 1 has greatly facilitated our understanding of gene function in eukaryotes.
In addition, small RNA-mediated gene silencing has therapeutic potential and holds promise for the treatment of specific diseases (6). Understanding the mechanism of RNAi and identifying the components of the RNAi machinery are essential for harnessing its full potential in both genome-wide screens and therapeutic applications.
Studies of the RDI complex strongly suggest that it includes no other proteins besides Dicer-2 and R2D2 (22). Additional proteins such as Ago2 are present in pre-RISC and holo-RISC, but nonetheless the complete compositions of the RISC loading complex, pre-RISC, and holo-RISC are unknown. Furthermore, holo-RISC sediments at ϳ80 S during sucrose gradient centrifugation (30). These observations indicate that additional protein factors associate with siRNAs. In this study, we identified siRNA-binding proteins from Drosophila embryo extracts. Target cleavage assays and immunoblotting of our siRNA affinity-selected proteins suggest that we purified active holo-RISC components. Proteomics analysis of the affinity matrix revealed both established and novel siRNA-associated proteins. Functional analyses of a subset of these factors suggest that they play important roles in RNAi.

EXPERIMENTAL PROCEDURES
Extracts-Drosophila embryo extracts were prepared from wildtype Canton S flies as described (31). Extracts were prepared by lysis of embryos in a Dounce homogenizer in embryo lysis buffer (ELB) (30 mM HEPES, pH 7.5, 100 mM potassium acetate, 2 mM magnesium acetate, 5 mM dithiothreitol). Protein concentrations were determined with a Bradford assay (Bio-Rad).
siRNA Affinity Chromatography-All small RNAs were purchased from Dharmacon. The siRNA sequences were as follows: Pp-luc guide, 5Ј-UCGAAGUAUUCCGCGUACGUG-3Ј; Pp-luc passenger, 5Ј-CGUACGCGGAAUACUUCGAUU-3Ј; let-7 guide, 5Ј-UGAGGUAGU-AGGUUGUAUAGA-3Ј; let-7 passenger, 5Ј-UAUACAACCUACUAC-CUCAUU-3Ј. Pp-luc and let-7 guide strand siRNAs (biotinylated at their 3Ј-ends via an LC-LC linkage) were annealed to their complementary passenger strands in vitro in annealing buffer (30 mM HEPES, pH 7.5, 100 mM potassium acetate, 2 mM magnesium acetate) with each strand present at 40.9 M. Strands were heated to 95°C for 2 min and then annealed at 37°C for 60 min. After the biotinylated siRNAs were annealed, 20.45 nmol of the siRNAs were complexed with 250 l of streptavidin beads (Pierce) for 30 min at 4°C with gentle agitation. For control experiments analyzing the association of let-7 or Pp-luc single-stranded guide RNA with proteins, 20.45 nmol of biotinylated guide alone were complexed with 250 l of avidin beads as described above. Once the siRNA-streptavidin matrix was formed, the beads were washed in ELB four times with 5 bead volumes to remove free siRNAs. After generation of the affinity matrix, the beads were mixed with embryo lysates (prepared as described above) for 30 min at room temperature. After incubation with extract, the affinity matrix was washed by centrifugation in ELB eight times with 10 bead volumes to purify stably associated factors. Immunoblotting was performed with rabbit anti-Dicer-2 antibodies (19) and anti-R2D2 antibodies (Abcam).
Target Cleavage Assays-Drosophila embryo extracts were complexed with siRNA affinity columns and washed as described above. Radiolabeled Pp-luciferase RNA target was then incubated with the affinity-purified beads or negative controls (streptavidin beads alone) for 1 h at 25°C. Samples were treated with stop buffer (100 mM Tris, pH 7.4, 25 mM EDTA, 50 mM NaCl, 2.1 g of proteinase K), and total RNA from the cleavage reactions was isolated by TRIzol extraction. The radiolabeled substrate and product RNAs from the cleavage reaction were subjected to electrophoresis in 15% polyacrylamide, 8 M urea denaturing gels and visualized with a PhosphorImager.
Chromatography and Mass Spectrometry-The proteins retained on the siRNA affinity or control chromatography preparations were reduced, alkylated, and trypsinized in 100 mM ammonium bicarbonate as described previously (32). After desalting the samples with a C 18 reverse-phase salt trap (Michrom Bioresources Inc.), the peptides were subjected to reverse-phase microcapillary LC-ESI-MS/MS. A 100-m-inner diameter microcapillary column (New Objective) was packed with 10 cm of 5-m reverse-phase material (Synergi 4 Hydro RP80a, Phenomenex). The trypsinized peptides were loaded onto the reverse-phase column equilibrated in buffer A (0.1% formic Acid, 5% acetonitrile) using a packing bomb at 400 p.s.i. The loaded column was placed in line with a Thermo Finnigan LTQ linear ion trap mass spectrometer (Thermo Electron Inc.). The peptides were eluted using a 90-min linear gradient from 0 to 80% buffer B (0.1% formic acid, 80% acetonitrile) at a flow rate of 0.2 l/min. During the chromatography gradient, the eluted ions were analyzed by one full precursor ion scan (400 -2000 m/z) followed by six MS/MS scans of the six most abundant ions. Dynamic exclusion was used to reduce the detection of abundant contaminating peptides. The raw mass spectra were acquired using Xcalibur version 2.2. The raw files were converted to MZXML files using readW version 4.0.2 under default settings. The resulting MZXML files were searched with Sequest version 27 on a SAGE-N Research, Inc. Sorcerer. The converted spectra files were searched against the European Bioinformatics Institute Drosophila proteome database containing 16,496 entries as of August 12, 2007. Sequest was searched with a fragment ion mass tolerance of 1.0 Da, a parent ion tolerance of 1.2 Da, and a single trypsin miscleavage allowance. The iodoacetamide derivative of cysteine was included in the Sequest search parameters as a static modification. Proteins identified from the Sequest Sorcerer search were loaded into Scaffold version 2_04_00. Peptide identifications in Scaffold were accepted if they could be established at greater than 95% probability as specified by the Peptide Prophet algorithm (33) and contained at least three identified peptides in a single sample. Peptides that could not be ascribed to a single protein were grouped into a single protein identifica-tion to simplify our analysis. The Scaffold unique protein counts for all samples were exported as a text file, subjected to hierarchical clustering using Cluster version 3.0 (34), and viewed with Treeview (version 1.1.1). All proteins identified are included in supplemental Fig. 1 with the number of unique peptides indicated. Supplemental Fig. 1 includes all of the protein identifications and sequence coverage for each protein in each sample. For proteins with single peptide identifications, the sequence, charge state, parent ion mass, correlation value, and tandem mass spectra are included in supplemental Fig. 2.
dsRNAs were transcribed from PCR products containing the first 500 bp of each gene. Oligonucleotides used for PCR amplification contained the T7 promoter at the 5Ј-end. dsRNA was made through transcription using the Megascript kit (Ambion) with a 9-h incubation at 37°C. Resulting dsRNAs were purified by phenol-chloroform extraction, denatured at 75°C for 15 min, and reannealed by slow cooling to room temperature.
S2 Cell Transfections and Luciferase Assays-S2 cells were plated at a density of Ϸ3 ϫ 10 6 /well (6-well plates) Ϸ24 h before use. For bathing, the medium was replaced with 0.5 ml of fresh Drosophila Schneider medium (no fetal bovine serum), and dsRNA was added to a concentration of 5 ng/well. One hour later, an additional 1.5 ml of complete medium was added. After 48 h, the firefly and Renilla luciferase plasmids as well as dsRNA against firefly luciferase were added at a final concentration of 60, 10, and 1 ng/well, respectively, using the Effectine protocol (Qiagen).
For each well, the reporter activity was calculated as a ratio of firefly luciferase (Pp-luc) to Renilla luciferase. Each data point was normalized to the negative control knockdown of white with dsRNA. A statistical analysis of the difference in firefly and Renilla luciferase signal ratios between experimental knockdowns and white knockdowns was performed using a Student's paired t test with a two-tailed distribution.

RESULTS
siRNA Affinity Capture of Drosophila RISC-Previous studies have used a model siRNA directed against firefly luciferase (Pp-luc) to direct efficient cleavage of a cognate mRNA target in vitro (21,36). To capture siRNA-interacting proteins by affinity chromatography, we used a Pp-luc siRNA that was biotinylated on the 3Ј-end of the guide strand and annealed to a non-biotinylated passenger strand. The resulting biotinylated siRNAs were bound to streptavidin beads and then incubated with three different concentrations of Drosophila embryo extract. After washing, the presence of correctly assembled RNAi effector complexes was assayed through cleavage of a cognate target mRNA (Fig. 1A). Specific target cleavage activity on the affinity matrix required a protein concentration of 30 mg/ml (Fig. 1A, lane 5). Lower concentrations of extract were insufficient for the assembly of functional RISC on the beads (Fig. 1A, lanes 6 and 7). Furthermore, samples that contained avidin beads and embryo extract alone did not generate the specific cleavage product (Fig. 1A, lanes 2-4).

siRNA-associated Proteins from Drosophila
These data confirmed that our siRNA affinity approach specifically captured functional RISC.
Next, we tested our siRNA affinity preparations for the presence of the established RNAi factors Dicer-2 and R2D2. Immunoblotting the siRNA affinity preparations for Dicer-2 and R2D2 confirmed the presence of these RDI complex components (Fig. 1B, lane 2). Dicer-2 and R2D2 were not detected in a parallel sample containing only avidin beads (Fig. 1B, lane 1). Collectively, these results show that we efficiently captured known RNAi factors and functional RISC using siRNA affinity chromatography.
Proteomics Identification of siRNA-associated Factors-To identify additional factors bound to the siRNA affinity purifications, we used a proteomics approach. Proteins from 90 mg of embryo extract were incubated with either Pp-luc siRNA or let-7 siRNA (an siRNA sequence different from Ppluc) on streptavidin beads. Bona fide RNAi factors should associate with siRNAs regardless of sequence, so requiring proteins to be detected with both siRNAs allows us to enrich for strong candidates. Furthermore, some proteins could conceivably co-purify based on their association with RISC-bound mRNAs that are present in the extract, but this potential problem is minimized by the use of distinct siRNAs (including one (Pp-luc) that lacks any natural mRNA targets in Drosophila) that would likely associate with different populations of mRNAs. As negative controls for these experiments, we incubated either avidin beads or single-stranded Pp-luc or let-7 guide strand alone with embryo extracts. Therefore, our experimental design allows us to distinguish avidin-binding proteins and nonspecific RNA-binding proteins from siRNA duplex-specific factors.
After capture of the proteins on the affinity matrices, the samples and controls were washed, trypsinized, and analyzed by LC-MS/MS mass spectrometry. Importantly, both the  -luc and let-7). The observed proteins were hierarchically clustered and displayed with Treeview. The number of unique peptides corresponding to a given protein is designated by the shade of red. A square in the cluster is black if the protein was not detected in the analysis of the mass spectrometry data. The position of Dicer-2 and other siRNA-binding proteins in the cluster is expanded to the right.

siRNA-associated Proteins from Drosophila
Pp-luc and let-7 siRNA purifications showed that Dicer-2 specifically purified with the siRNA affinity matrix but not the avidin beads alone or the guide strand only samples (Fig. 2). Dicer-2 was likely the most abundant protein in the affinity purifications as it was represented by the most total spectra and the most unique peptides and exhibited high sequence coverage among the siRNA-specific binding proteins (supplemental Fig. 1) in agreement with the central role of Dicer-2 in RNAi (21,30). The dsRBD proteins Loquacious and R2D2 were detected in Pp-luc siRNA affinity purifications but were below the limit of detection in let-7 siRNA affinity purifications (supplemental Fig. 1).
Hierarchical clustering of the samples revealed a group of seven siRNA-specific binding proteins (Fig. 2) that were detected with both siRNAs but not in any of the negative control samples. RNAi factors were strongly enriched in the siRNA-specific group with a statistically significant functional ontology (Fig. 3A). Additionally, a large percentage of the siRNA-associated protein domains were consistent with those of RNAi factors including Piwi, Paz, Helicase, and dsRBD (Fig. 3B). Among these seven proteins were the RNAi factors Dicer-2 and Argonaute-1, reinforcing our isolation of both initiator and effector factors in RNA interference.

siRNA-associated Proteins from Drosophila
In addition to these known RNAi factors, we identified Maleless, which contains both Helicase and dsRBD domains. Intriguingly, we identified an RNA-binding protein (encoded by the candidate gene CG31301) that contains both G-patch and R3H domains (Figs. 2 and 3B). G-patch nucleic-acid binding proteins are found within retroviral proteins. Because a subgroup of the predicted domains listed above confers functions consistent with known steps in the RNAi pathway, these proteins represent viable candidates for novel RNAi factors.
Functional Analysis of Novel siRNA-associated Proteins-Our proteomics screen detected seven proteins bound to siRNAs in vitro. To analyze potential functional roles for these proteins in RNAi, we depleted them from Drosophila S2 cells and then tested the ability of those cells to silence a luciferase reporter (35,37). We restricted our functional analysis to proteins that contain predicted functional domains potentially relevant to RNAi such as nucleases (Argonaute and Dicer proteins), helicases (Maleless), and RNA-binding proteins (CG31301). Proteins such as Lamin, Rudimentary-like (a metabolic enzyme), and other abundant contaminating enzymes were excluded from our analysis. In the Dual-Luciferase assay, cells are first transfected with dsRNAs that target the candidate RNAi factor. After knockdown of the candidate RNAi factor, cells are co-transfected with a Pp-luciferase plasmid, a Renilla luciferase plasmid, and a dsRNA that specifically targets Pp-luciferase mRNA for RNAi. Derepression of the Pp-luciferase signal relative to the Renilla luciferase signal in the knockdown cells (in comparison with nonspecific dsRNA controls) indicates suppression of RNA interference activity, whereas a reduction in the relative Pp-luciferase signal indicates an increase in RNAi activity (35,37).
As positive controls, transfection of S2 cells with dsRNAs targeting Dicer-2 or Argonaute-2 resulted in a derepression of luciferase activity relative to the negative control dsRNAs (Fig.  4). Transfection of dsRNAs targeting a subset of the novel siRNA-associated factors identified in our proteomics screen revealed that CG31301, a predicted RNA-binding protein, is likely involved in RNAi (Fig. 3). dsRNAs targeting CG31301 derepressed Pp-luciferase levels ϳ3.2-fold relative to a negative control dsRNA directed against the Drosophila gene white (p ϭ 0.048). This derepression is significant in comparison with Dicer-2 (2.6-fold derepression (p ϭ 0.017)) and Argonaute-2 (6.57-fold derepression (p ϭ 0.048)). Additionally, knockdown of Maleless, an RNA helicase critical for sex chromosome dosage compensation, partially compromised RNAi activity (2-fold derepression (p ϭ 0.025)) (Fig. 4). Altogether, these results indicate that our proteomics screen identified both known and novel proteins with biochemical and functional properties consistent with RNAi factors.

Multiple Factors
Converge on Distinct siRNAs-Size estimates of RISC suggest that a large number of proteins asso-ciate with siRNAs in vitro (30). Consistent with this observation, our results show that Dicer-2, dsRBD proteins (Loqs, R2D2, and Maleless), RNA helicases (Dicer-2 and Maleless), Argonaute proteins, and an additional nucleic acid-binding protein (CG31301) converge on distinct siRNAs to form large siRNP complexes. These results reinforce a model where proteins of multiple apparent functions form an active RISC complex.
Identification of Potential Novel RNAi Factor: CG31301-Our results show that CG31301 associates with two siRNAs of unrelated sequence (Pp-luc and let-7). Additionally, CG31301 is important for efficient RNAi in S2 cells. Taken together, these results strongly suggest that CG31301 functions as a novel RNAi factor. CG31301 contains two predicted protein domains, an R3H domain and a G-patch domain. The R3H domain is found in single-stranded nucleic acid-binding proteins (38), whereas the G-patch domain has been found in some retroviral nucleic acid-binding proteins (39). Interestingly, recent evidence shows that retrotransposons, many of which appear to be derived from retroviruses, are among the primary targets of endogenous siRNAs (7-10, 12, 15, 40, 41). It is tempting to speculate that metazoans could have co-opted retroviral nucleic acid-binding proteins to participate in retrotransposon suppression. Future studies of CG31301 structure and function may illuminate this possibility.
Identification of Helicase Associated with siRNAs-We observed that the predicted RNA helicase and dsRBD protein Maleless specifically purified with siRNAs. In our functional assays, knockdown of Maleless modestly compromised RNAi activity. During holo-RISC formation, the guide strand of the FIG. 4. Functional analysis of siRNA-associated factors. A subset of siRNA-binding proteins was tested for functional roles in RNA interference using the luciferase assay. The luciferase assay tests for derepression of RNAi-inhibited Pp-luciferase after knockdown of a suspected RNAi factor. dsRNAs targeting Argonaute-2 and Dicer-2 were included as positive controls, whereas dsRNA targeting the white gene was included as a negative control. Bars indicate average -fold change in luciferase activities (of independent biological replicates) relative to control cells treated with dsRNA targeting white. Each dsRNA treatment shows the average -fold difference for three different experiments. Candidate gene (CG) numbers were included for genes with no previously ascribed function.
siRNA is liberated from the passenger strand to allow hybridization with an RNA target. One mechanism of guide strand liberation is the cleavage of the passenger strand by Ago2 (42,43). Nevertheless, it is possible that one or more RNA helicases help dissociate the passenger and guide strands (with or without prior passenger strand cleavage), leading to holo-RISC formation. The mammalian protein RNA Helicase A, an apparent paralog of Maleless, has been shown to complex with siRNAs and function in RNAi (44). Therefore, a role for this RNA helicase in RNAi might be evolutionarily conserved.
In summary, we conducted a proteomics screen for siRNAassociated proteins in Drosophila. Our screen identified known RNAi factors (Dicer-2, Loquacious, and R2D2) as well as several novel siRNP candidates. These latter proteins include domains with predicted activities that could play a role in RNAi. Some of these factors appear to be important for efficient RNAi in cultured cells. Future studies will elucidate the precise molecular functions of these novel siRNA-associated factors.