A genomewide screen for components of the RNAi pathway in Drosophila cultured cells

Abstract

Posttranscriptional silencing by RNAi is initiated by dsRNAs that are processed into siRNAs that ultimately target homologous mRNAs for degradation. We used luciferase reporter constructs and a cultured cell-based assay to perform a genomewide screen for components of the RNAi pathway in Drosophila melanogaster. The screen identified seven genes that affect the RNAi response, five with previously described function (AGO2, Tis11, Hsc70-3, Hsc70-4, and hdc) and two annotated genes (CG17265 and CG10883).

Posttranscriptional silencing by RNAi and related pathways regulates the expression of genes involved in diverse cellular processes in a variety of eukaryotic organisms (for review see refs. 1 and 2). The RNAi pathway is activated by long dsRNA, which is processed by the conserved RNase III enzyme Dicer into 21- to 23-nt duplexes known as siRNAs (3–8). One strand of these siRNAs is incorporated preferentially into the RNA-induced silencing complex (RISC) (9), which typically mediates cleavage of complementary RNA species in the cell to bring about gene-silencing activity (10, 11).

Genetic and biochemical analyses have identified numerous components in the RNAi-response pathway (for review see ref. 12) and functions have been assigned for several of these factors. For example, Drosophila melanogaster Dcr-2, one of two Dicer paralogues in this organism, is responsible for the initial cleavage of the dsRNAs (13). Further, Dcr-2 associates with the dsRNA-binding protein R2D2 and the siRNA duplex (14, 15) and facilitates selection of one of the two strands (9). This RNA–protein complex then recruits additional proteins, including argonaute family members, to form RISC. Other proteins such as Tudor staphylococcal-nuclease-domain-containing protein (Tudor-SN), fragile-X-related protein (dFXR), the Vasa intronic gene product (Vig), and a homolog of p68 RNA helicase (Dmp68) have been identified biochemically as RISC components in Drosophila (16–18), although their molecular roles remain unknown.

Perhaps the best characterized proteins involved in the RNAi pathway are the argonaute proteins that emerged in early biochemical studies (19). This conserved class of proteins (20) is characterized by two recognizable domains, PAZ and Piwi, that are important for recognizing the 3′ ends of RISC-loaded siRNAs (21–25) and for endonucleolytic cleavage (26–31), respectively. Recent biochemical purification of a minimal RISC and in vitro reconstitution experiments have identified AGO2 as the sole protein essential for RISC-targeted mRNA cleavage activity (29, 32). Although such biochemical experiments have done much to focus attention on the role of the argonaute proteins in RNAi-mediated silencing, given that the holo-RISC migrates at 80S in a sucrose gradient (15) and that the in vivo process is considerably more complicated than simple targeted cleavage of an mRNA, it seems likely that essential components remain to be discovered.

The relative ease of use of RNAi-mediated gene silencing has revolutionized reverse-genetic approaches in various model systems. RNAi libraries have been generated for use in genomewide screens in mammalian (33–35) and fly cell culture lines (36–38) and for the whole animal in C. elegans (39–41). These libraries are being used to exhaustively screen for novel components involved in many different pathways with the primary limitation being the development of reasonable high throughput functional readouts. Such RNAi-based approaches have even been successfully used in C. elegans to identify components that appear to be essential for the RNAi response itself (42, 43). Here we report results from a genomewide screen in D. melanogaster-cultured cells aimed at identifying previously undescribed cellular components critical for RNAi-mediated gene silencing.

Results

Dual-Luciferase Reporter System for RNAi-Mediated Response.

We first established a dual-luciferase reporter assay to monitor RNAi-mediated response in Drosophila S2R⁺ cells by using Renilla luciferase (RL) as a reporter gene and Firefly luciferase (FL) as a transfection control. Cotransfection of the reporter constructs with dsRNA (≈500 bp in length) targeting RL (dsRL) results in ≈100-fold repression of RL activity, whereas dsRNAs targeting nonspecific genes such as yellow fluorescent protein (YFP) have no discernible effect. Bathing of such S2R⁺ cells with dsRNA targeting a gene that is critical for the RNAi-mediated response should result in higher levels of RL (relative to FL) after cotransfection of reporter constructs and dsRL. As an initial test of the potential of such an assay for a genome-wide screen, a group of genes previously implicated in the RNAi and microRNA (miRNA)-regulated pathways was targeted with dsRNA before introduction of the luciferase reporter constructs and dsRL. Cells were bathed briefly with dsRNAs targeting genes of interest and then were incubated for 48 or 96 h to allow for down-regulation of the gene by RNAi and protein turnover. The two plasmid constructs (copia Renilla and AcSV-Firefly) and dsRNA-targeting RL then were cotransfected, and luminescence was measured 24 h later. Incubation of the cells after bathing with dsRNA against Argonaute2 (AGO2) for either length of time (48 or 96 h) resulted in substantial abrogation of the RNAi-mediated repression of RL activity (Fig. 1A). None of the other tested components (including well known RNAi components such as Dcr-2 and R2D2) exhibited increased levels of RL after the shorter incubation period, and only a few exhibited marginally increased RL activity after the longer incubation period (Fig. 1A). These results are broadly consistent with a number of previous studies (14, 44).

Fig. 1.

Effects of dsRNA-mediated gene knockdown on the dual-luciferase-based RNAi reporter assay. RL activity was normalized to that of Firefly (y axis) in response to dsRNA-targeting Renilla (dsRL). (A) Effects on RNAi response (and its time dependence) after knockdown of genes implicated in the miRNA or siRNA pathway. Normalized Renilla/FL activities are compared with the ratio of these activities in the absence of dsRL (YFP columns). Reporter constructs were transfected 48 h (black columns) or 96 h (gray columns) after the initial bathing with dsRNA targeting the gene of interest (indicated on x axis). (B) Luminescence readings (Renilla/Firefly) from the 96-well plate where AGO2 is targeted by dsRNA (gray column, position E07). Column 1 represents the control where no dsRL has been added.

Genomewide Screen in Drosophila S2R⁺ Cells.

Based on the robust response of the reporter system to the AGO2 knockdown, we proceeded to use this general approach to identify other potential RNAi components. We performed a high-throughput RNAi screen by using a 96-well format and a library of ≈21,300 dsRNAs targeting annotated genes in the D. melanogaster genome (38, 45). As in the preliminary analysis, the screen involved a 48-h incubation of cells after bathing with a given dsRNA from the library followed by transfection with reporter constructs and dsRL. The dsRNAs in the complete library previously shown to result in a viability phenotype were not included in this screen (38). Twenty-four hours after transfection, luminescence readings were taken and normalized values (N = RL activity/FL activity) were calculated for each well in the 96-well plate. dsRNAs were considered preliminary candidates when N was >4 SDs (4 × SD) above the average of the whole plate. In the primary screen, 138 dsRNAs were identified as significant based on N values that exceeded this stated criterion (see Fig. 1B for luminescence readings of plate containing dsRNA against AGO2). These candidate genes were rescreened by using the original dsRNA samples, and the 96-well format and 56 genes yielded a signal above the defined cutoff (4 × SD above control cells bathed with dsYFP) on the second pass. For these 56 genes, dsRNAs were resynthesized, and the reporter assay was repeated in a 24-well format. Twenty-five of the initial candidate genes reproducibly affected the RNAi response under these conditions (Table 1) and were analyzed further.

Table 1.

Fold increase of normalized RL in knockdown cells

Z value	Gene targeted	First RNA	Second RNA	Third RNA	Predicted/known function
9.5	CG7439	78 ± 6	24 ± 7		Ago2, RNAi pathway
9.5	Tis11	19 ± 3	15 ± 3	2.1 ± 0.6	RNA catabolism, RNA binding
9.5	CG7239	14 ± 4	1.4 ± 0.5
9.3	Hsc70-4	9 ± 2	30 ± 10		Heat shock protein cognate
9.1	CG15074	13 ± 3	0.9 ± 0.4		Heparin-glucosamine 3-O-sulfotransferase activity
9.1	ab	8 ± 1	0.8 ± 0.2		Transcription factor activity
9.1	CG14358	7 ± 3	0.8 ± 0.1
9.0	fru	19 ± 2	1.4 ± 0.7		Transcription factor activity
8.6	CG12969	25 ± 2	1.0 ± 0.1		Actin binding
8.4	hdc	32 ± 2	25 ± 4		Trachea development
8.3	CG6460	7 ± 2	0.8 ± 0.2
8.2	CG17265	59 ± 6	14 ± 3	2.0 ± 0.2
8.1	Lim3	12 ± 4	1.5 ± 0.4		Transcription factor activity
7.6	Eip75B	9 ± 4	0.5 ± 0.1		Transcription factor activity
7.5	m2	14 ± 1	1.1 ± 0.1		Notch signaling pathway
7.1	CG7173	9 ± 4	1.1 ± 0.4
6.8	CG2556	24 ± 5	1.1 ± 0.4
6.4	Hsc70-3	25 ± 6	35 ± 6		Heat shock protein cognate
6.4	CG5060	12 ± 3	1.2 ± 0.6
6.3	CG8154	9 ± 3	0.8 ± 0.4
6.1	CG13188	12 ± 2	1.3 ± 0.2
5.6	CG10883	2.4 ± 0.6	2.1 ± 0.3
5.6	CG5732	6 ± 2	0.9 ± 0.5		Polynucleotide adenylyltransferase activity
4.1	EG:132E8.1	3 ± 1	1.6 ± 0.3		mRNA binding
4.6	HLHm3	6 ± 1	0.9 ± 0.2		Notch signaling pathway

Z values were calculated from the primary screen. Values represent the average of a triplicate experiment. Experiments were performed multiple times. Luminometer values were normalized (Renilla/Firefly activity) and compared with the ratio of activities of cells treated with the dsYFP control RNA in the bathing step. Predicted/known functions are listed as found in Flybase (46). Seven of the original candidate genes (bold) were confirmed based on their reproducible effects.

Validation of Candidate Gene Effects on RNAi-Mediated Silencing.

Potential off-target effects are a substantial concern in assessing the validity of any RNAi-mediated process and are of particular concern when relatively long dsRNAs are used (47–49). The likelihood of off-target effects playing a role in our screen was increased further by the identification of repetitive sequences in some of the candidate genes (see below). To address these concerns, we designed a second set of dsRNAs against distinct regions of the candidate genes (generally the nonrepetitive regions) to look again at the effects of these gene knockdowns on the RNAi reporter assay. After this analysis, seven of the original candidate genes (highlighted in bold, Table 1) were confirmed based on their reproducible effects. For two of the candidate genes, Tis11 and CG17265, a third dsRNA was generated because the first two RNAs targeted glycine- and glutamine-rich coding regions of the gene (with associated repetitive nucleotide sequences). In these cases, although the effects on the luciferase reporter assay for this third dsRNA were more modest, these genes still qualify as candidates as defined by the 4 × SD cutoff above the average of control cells bathed with dsYFP.

As an initial confirmation of these candidate genes, we simply inverted the reporter assay by using RL as the transfection control and FL as the reporter. When the seven candidate dsRNAs were used to knock down target gene expression, and FL subsequently targeted with dsRNA (dsFL), we were able to confirm six of the seven candidate genes (all but the heat shock protein cognate Hsc70-4) as important for the cellular RNAi-mediated response (Table 3, which is published as supporting information on the PNAS web site). The results for Hsc70-4 were equivocal as gene knockdown in this case appeared to substantially up-regulate RL expression.

We also tested our genes of interest for their role in mediating RNAi of a nonspecific gene, SREBP (sterol regulatory element binding protein). In these experiments, we used S2 cells rather than the adherent S2R⁺ cells to more simply allow for preparation of RNA and protein in larger quantities. As before, cells (S2) were first bathed with dsRNA targeting our genes of interest and, after 48 h, were bathed with dsRNA against SREBP. After an incubation period of 24 h, the RNA and protein levels for SREBP were examined (Fig. 2). As seen with the luciferase reporter, SREBP protein levels are reduced substantially when prebathed with a nonspecific gene, dsYFP, and then subsequently treated with dsRNA-targeting SREBP (dsSREBP; see Fig. 2, lanes 1 and 2). Likewise, steady-state SREBP mRNA levels assessed by Northern blot analysis are reduced in control cells prebathed with dsYFP (Fig. 2B, lanes 1 and 2). As before, we see that knockdown of AGO2 abrogates RNAi-mediated targeting of SREBP as evaluated both at the level of RNA and protein (Fig. 2, lane 9), consistent with the robust responses that AGO2 yielded in the luciferase reporter assays. Knockdown of Hsc70-3, another candidate gene, similarly resulted in abrogation of the RNAi response at the level of the RNA and protein (Fig. 2, lane 4). Notably, knockdown of the five remaining candidate genes did not significantly diminish the RNAi response in these cells (Fig. 2 and Table 2).

Fig. 2.

Effect of dsRNA-mediated gene knockdown of seven candidate genes (and YFP control) on dsRNA-mediated silencing of the endogenous gene SREBP. S2 cells were treated first with dsRNA targeting the candidate genes (as indicated) and then, after 48 h, with dsRNA targeting SREBP (except for lane 1). (A) Western blot analysis of protein isolated from treated cells probed with anti-SREBP and anti-Tubulin antibodies. (B) Northern blot analysis of total RNA isolated from treated cells probed for SREBP and rp49 mRNAs. The protein and RNA analyses were performed twice in duplicate.

Table 2.

mRNA levels of SREBP and AGO2 in various knockdown cells

Gene targeted	Relative SREBP mRNA levels	Relative AGO2 mRNA levels
YFP (no SREBP)	9 ± 4	1.0
YFP + SREBP	1.0	0.9 ± 0.1
Hsc70-4	0.6 ± 0.1	0.8 ± 0.1
Hsc70-3	2.4 ± 0.4	1.0 ± 0.2
hdc	0.7 ± 0.2	1.0 ± 0.2
Tis11	0.5 ± 0.2	0.6 ± 0.1
CG17265	0.9 ± 0.3	0.6 ± 0.1
CG10883	0.7 ± 0.2	0.8 ± 0.1
AGO2	2.2 ± 0.4	0.2 ± 0.03

mRNA levels were calculated relative to rp49 mRNA and normalized to the value indicated in bold (n = 4).

Evaluation of AGO2 mRNA and Protein Levels After RNAi-Mediated Knockdown of Seven Candidate Genes.

The strength and consistency of the effects on the RNAi response of AGO2 caused some concern that the effects of knockdown of the other candidate genes might simply be manifested through indirect effects on overall AGO2 expression. To evaluate this possibility, we first looked specifically at AGO2 mRNA levels by Northern blot analysis after bathing with dsRNA for the seven candidate genes. As seen in Fig. 3A and Table 2, there are no discernible effects on AGO2 mRNA levels except when dsAGO2 RNA was applied. Similar analysis of Tis11 mRNA after bathing with three distinct dsRNAs showed clear reduction in Tis11 mRNA levels showing that, in general, the dsRNAs function to degrade their designated target (Fig. 4, which is published as supporting information on the PNAS web site).

Fig. 3.

Effects of dsRNA-mediated gene knockdown of seven candidate genes on AGO2 protein and mRNA levels. (A) Northern blot analysis of total RNA from cells bathed with dsRNAs (as indicated) and probed for Ago2 mRNA. (B) Western blot analysis of protein isolated from cells bathed with dsRNAs (as indicated) and probed with anti-AGO2 antibody.

It is also conceivable that targeting of the candidate genes by RNAi somehow indirectly affects AGO2 protein levels. This explanation was of particular concern because human HSP90 had been shown to interact with human Ago2 in human and bacterial cells (26, 29), possibly stabilizing the protein and, thus, affecting its expression. To investigate this possibility, we bathed S2 cells with dsRNAs targeting the candidate genes and, after 62 h, examined AGO2 protein levels. Just as shown for mRNA levels, AGO2 protein levels were diminished relative to the tubulin control only when bathed with dsAGO2 RNA but not with the remaining six candidate genes (Fig. 3B).

Discussion

We performed a high-throughput genomewide RNAi screen to identify genes involved in the RNAi pathway in Drosophila cell culture by using a dual-luciferase reporter assay that reports on cellular RNAi-mediated degradation of RL mRNA. This screen led to the identification of seven genes that, when targeted with dsRNAs to mediate knockdown of gene function, consistently prevented the down-regulation of RL expression relative to that of the control FL. We also developed an independent assay to look at the RNA and protein expression levels of an endogenous gene, SREBP, to evaluate the RNAi response in these cells and its dependence on the seven candidate genes.

The candidate genes identified in the screen include the previously described genes AGO2, Tis11, Hsc70-3, Hsc70-4, and hdc and the two annotated genes CG17265 and CG10883. Six of these candidate genes had similar effects on the cellular RNAi response when FL was instead used as the reporter and a similar assay was performed. Notably, only AGO2 and Hsc70-3 showed substantial effects in the assay that examines the response of the endogenous SREBP gene to RNAi targeting. These data suggest that the luciferase assay may be more sensitive to perturbations in the levels of a variety of RNAi components, whereas RNAi effects on SREBP are more robust and, therefore, most dependent on the roles played by components of central importance such as AGO2 and Hsc70-3. Despite the apparent overall sensitivity of our assay, our screen failed to identify a number of known components (Dcr-2 or R2D2, for example) involved in the RNAi pathway, some of which have been identified in related RNAi-based screens (43). Although initially surprising, these results are consistent with those of other groups (14, 44) and likely reflect general limitations of RNAi-based screens including: (i) the efficacy of introduced dsRNA at inducing specific RNAi, (ii) the protein half-life of the targeted mRNA, (iii) the rate of expression and turnover of the reporter gene, and (iv) the relative timing of each of these events. Taking into consideration these limitations, it is unlikely that such a screen will be comprehensive in its identification of components in the RNAi pathway, but some genes of interest nevertheless might emerge.

Of the seven candidate genes, only AGO2 previously was shown to play an important role in the siRNA pathway. The argonaute proteins have been ubiquitously identified in biochemical and genetic approaches as essential components of the RNAi-mediated gene-silencing pathway. Importantly, the RNase-H like motif in the Piwi domain of these proteins has been identified as the catalytic core for RNAi-mediated gene cleavage (26, 28, 30). The identification of AGO2 and not AGO1 in this screen is consistent with the accumulated bias from biochemical studies that AGO2 is more centrally involved in siRNA-directed events, whereas AGO1 is more centrally involved in miRNA-directed events (32, 50, 51). Mutation of AGO1 or AGO2 in Drosophila yielded different phenotypes suggesting again that these genes play distinct roles in posttranscriptional silencing (51–54). The candidate dTis11 (known as TTP in humans) was of particular interest because this gene was initially reported to be critical for the degradation of AU-rich mRNAs. Further analysis has shown that this process depends on components of the RNAi pathway including AGO2 and on interactions with an AU-region complementary miRNA (mir16) (55).

Our screen also identified two heat shock protein cognates, Hsc70-3 and Hsc70-4. Hsc70-3 is a particularly strong candidate that shows robust effects in two independent assays evaluating the RNAi-response pathway. Two important observations potentially connect heat shock proteins to the RNAi pathway: (i) human HSP90 is associated with human Ago2 by coimmunoprecipitation studies (26), and (ii) recombinant expression of human Ago2 in Escherichia coli depends on coexpression of human HSP90 (29). Although Drosophila Hsc70-3 and Hsc70-4 are not the homologues of human HSP90, it seems possible that other heat shock proteins might play analogous roles in the Drosophila system.

Another gene identified in our screen with previously described function is headcase (hdc), which is expressed specifically in imaginal cells during development and has been shown further to play an important role in trachea development (56). There are two forms of protein produced from the hdc gene, the larger product dependent on a translation termination suppression mechanism specified by a 80-nt region 3′ to the UAA internal stop codon (57). The 3′ UTR of this gene is predicted to contain a number of miRNA-binding motifs, although the significance of this observation is not clear (58, 59). The remaining two genes, CG17265 and CG10883, are entirely uncharacterized, although CG17265 has predicted orthologs in other organisms.

Although the previously ascribed functions of the seven genes provide some clues, there is much biochemistry to be done to begin to understand the potential role of these genes in the RNAi-response pathway. A striking feature of several of the genes identified in our screen is that with the exclusion of the heat shock protein cognates, Hsc70-3 and Hsc70-4, they appear to be enriched in repetitive sequences comprised primarily of glycines, glutamines, and, in some cases, serines. For example, the N-terminal region of AGO2 is composed of 11 nearly identical 23-aa-long repeats that are rich in glycine and glutamine (19). We further note that these repeat regions are remarkably conserved at the nucleotide level, varying only in positions where there are resulting changes in the amino acid sequence (i.e., there are no silent nucleotide changes; Fig. 5, which is published as supporting information on the PNAS web site). Although Arabidopsis AGO1 appears to have a related repeat sequence at its N terminus, the repeat sequences do not appear to be broadly conserved across phylogeny (e.g., they are not found in worm or human argonaute proteins). Interestingly, there are intriguingly similar repeats that can be identified in Apis mellifera in a gene that also has a Piwi domain (Fig. 5). We note that mass spectrophotometric analysis of functional RNAi–effector complexes indicated that these repeat elements were included in the expressed AGO2 protein sequence and, thus, are potentially important for gene function (32). Further annotation of other closely related species will help in defining the extent of conservation of these repeat elements. Although the functional significance of such repeats and of the glutamine-, glycine-rich nature of these proteins is not known, we note that other components previously implicated in the RNAi pathway also share this feature (e.g., Rm62, dFMRp, and vig; Fig. 6, which is published as supporting information on the PNAS web site). Computational approaches have indicated that such glutamine-rich proteins are relatively abundant in Drosophila, whereas it has been noted in yeast that such proteins are involved in a wide range of RNA-related processes (60).

A recent screen looked at the effects of knockdown of ≈2,000 Drosophila genes in S2 cells on RNAi-based targeting of an essential gene to identify novel components in the cellular RNAi pathway (61). This study identified several genes, including AGO2 and Dcr-2, as well as a previously undescribed RNA helicase (Belle), a proteosomal subunit (Pros45), and an endocytic machinery component (Chc). Our screen did not identify RNA uptake pathway components because dsRNAs were transfected directly into the cells with the reporter constructs. Our failure to identify other candidates might be accounted for by differences in the assays (targeting of a reporter gene vs. an essential gene) or in the defined cut-off values for significance.

The results of our screen also differ from another genomewide screen for RNAi components that was performed in C. elegans and which identified many of the known components implicated in RNAi and many previously undescribed gene candidates (43). Indeed, only dAGO2 overlapped with candidates identified in that screen. A number of potential explanations might account for this apparent discrepancy: (i) the cut-off criterion for our screen was relatively stringent, (ii) the secondary screen with a second dsRNA minimized the contribution from dsRNAs with off-target effects, and (iii) use of a cell culture-based rather than an organism-based assay necessitated differences in conditions. We are not able to predict which of these differences might be the most significant.

The secondary screening of the candidate genes with Firefly as a reporter and Renilla as the transfection control gives us greater confidence in the importance of six of the genes as well as the confirmation of AGO2 and Hsc70-3 in the independent RNAi assay that targets the endogenous gene SREBP. These data further suggest that AGO2 and Hsc70-3 are the most critical components in the RNAi pathway that we have identified. The results of this secondary assay are consistent with our observation that AGO2 consistently yielded the most robust and reproducible response in our RL-based reporter assay. Our further analysis of AGO2 protein levels when the other six candidate genes were targeted by RNAi indicated that these genes do not act indirectly by affecting AGO2 expression. Future experiments should focus on developing in vitro systems for assessing the role of these proteins in the various steps of RNAi-mediated gene silencing.

Materials and Methods

Cell Culture and Constructs, dsRNAs, and Antibodies.

S2R⁺ cells were cultured in Schneider's Drosophila Medium (Invitrogen, Carlsbad, CA) supplemented with 10% heat-inactivated FBS (Invitrogen) and 100 units/ml penicillin/100 units/ml streptomycin (Cambrex BioScience, Walkersville, MD). S2 cells (Invitrogen) were maintained in Drosophila SFM (Invitrogen) supplemented with 100 units/ml penicillin/100 units/ml streptomycin (Cambrex BioScience) (Pen/Strep).

FL expression was driven by the Actin5c promoter (AcSV-FL) in the vector background described in ref. 62. RL expression was driven by the Copia promoter as described (copia-Renilla) (36).

dsRNAs of ≈500 bp in length were transcribed from PCR fragments as follows. Oligonucleotides used for PCR amplification contained the T7 promoter at the 5′ end and are as described in Tables 4–7, which are published as supporting information on the PNAS web site. To confirm the identity of PCR products generated from cDNA or genomic DNA, PCR fragments were cloned by using the TOPO-TA procedure (Invitrogen) and verified by sequencing. The resulting plasmids were used as templates for subsequent PCRs. PCR fragments were gel-purified and used in T7 transcription reactions (63). dsRNAs were purified by using NAP columns (Sephadex G25; Amersham Biosciences, Piscataway, NJ), precipitated, and resuspended in ddH₂O. dsRNAs were denatured at 75°C for 15 min and reannealed by slow cooling to room temperature. Extent of annealing was assessed by agarose gel electrophoresis.

α-AGO2 monoclonal mouse antibody raised against dAGO2 was the generous gift of M. Siomi (University of Tokushima, Tokushima, Japan) (50). α-SREBP monoclonal mouse antibody SREBP-1, 2A4 raised against human SREBP-1 was obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and cross-reacted with Drosophila SREBP (64). α-β-tubulin mouse monoclonal antibody (E7) was obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA; developed by M. Klymkowsky).

Bathing of S2R⁺ Cells (Screening Conditions).

S2R⁺ cells were seeded at a density of ≈3 × 10⁵ cells per well (24-well plate) or 6 × 10⁴ cells per well (96-well plate). For bathing experiments, the media were replaced with 50 μl (96-well) or 350 μl (24-well) of serum-free media. dsRNA was added to a concentration of 15 μg/ml, incubated for 1 h, and then supplemented with an equal volume of media containing 20% FBS and 100 units/ml penicillin/100 units/ml streptomycin. After 48 h, reporter constructs were transfected by using Effectene (Qiagen, Valencia, CA) with the following amounts of DNA: for the 96-well format, 22 ng of copia-Renilla, 1 ng of AcSV-FL, and 22 ng of dsRL (dsRNA against Renilla) or dsYFP (dsRNA against YFP); for 24-well format, 67 ng of copia-Renilla, 3 ng of AcSV-FL, and 67 ng of dsRL or dsYFP. Cells were lysed 24 h after transfection, and RL and FL activities were measured by using a dual-luciferase system (Promega, Madison, WI).

Bathing of S2 Cells, RNA, and Protein Isolation.

S2 cells were plated at a density of ≈3 × 10⁶ per well (6-well plates) ≈24 h before use. For bathing, the media were replaced with 1.5 ml of fresh Drosophila SFM media, and dsRNA was added to a final concentration of 15 μg/ml. One hour later, an additional 0.5 ml of media was added. After 48 h, the media were replaced with fresh 1.5 ml of Drosophila SFM media and dsRNA against SREBP or YFP at a final concentration of 2.5 μg/ml. Again, 1 h later, an additional 0.5 ml of media was added. Cells were collected after an additional 24 h, and total RNA was isolated by using TRIzol (Invitrogen) or Tri-reagent (Ambion, Austin, TX) according to the manufacturers' protocols. Protein was isolated from the same samples from the organic phases (Ambion Tri-reagent).

Northern Blot Analysis.

Fifteen micrograms of total RNA were loaded per lane on a 1.2% denaturing agarose gel (formaldehyde). RNA was vacuum-blotted at 127 mmHg for 90 min with 10× SSC (1.5 M sodium chloride/150 mM sodium citrate) onto Hybond-XL (Amersham Biosciences) membrane followed by UV cross-linking. Membranes were hybridized in Ultra-hybridization buffer (Ambion) at 65°C by using radiolabeled RNA probes described below. For size determination, RNA Millennium Size Marker (Ambion) was run in parallel and probed with T7 transcribed Millennium Marker Probe Template (Ambion).

RNA probes were T7 transcribed from PCR templates (see Table 8, which is published as supporting information on the PNAS web site, for oligonucleotides and templates) by using Strip-EZ T7 kit (Ambion), thus allowing for easy removal of RNA probes and multiple probings of each membrane. As above, PCR fragments amplified from cDNA or genomic DNA were TOPO-TA cloned and verified by sequencing.

Western Blot Analysis.

Proteins were separated on 7.5% SDS/PAGE and electroblotted onto nitrocellulose (0.2 μm). Primary antibodies were detected by using horseradish peroxidase-coupled anti-mouse antibody (Promega) followed by the addition of chemiluminescent substrate (SuperSignal West Pico Substrate; Pierce, Rockford, IL).

Abbreviations

FL

Firefly luciferase

miRNA

microRNA

RISC

RNA-induced silencing complex

RL

Renilla luciferase.