Large scale RNAi screen identifies novel RNAi target genes
So far, the targets for dsRNA based pest control have been identified by small scale screens and on knowledge based approaches, i.e. by testing genes where previous data indicated an essential function. However, this approach will miss many genes that have not yet been linked to an essential function in one of the model species. Therefore, we screened the data produced by the large scale RNAi screen iBeetle (Bucher, Klingler, personal communication), where randomly selected genes were downregulated by injection into pupae and larvae and the resulting phenotypes were documented in the iBeetle-Base [23]. In the iBeetle screen, about 5,000 genes had been screened [24]. Of those, 100 revealed ≥90 % mortality both 9 days after pupal and eleven days after larval dsRNA injection (Additional file 1: Table S1). In order to confirm these results and to test for sensitivity, we injected different concentrations (3 ng/μl, 30 ng/μl, 100 ng/μl, 300 ng/μl and 1 μg/μl) of the same dsRNAs into 10 larvae, respectively, and scored the survival rate every second day. The most effective 40 genes caused a mortality of 50–100 % at day eight post injection using the lowest concentrated dsRNA (Additional file 1: Table S2, Additional file 2: Figure S1 A). We focused on the 11 most effective target genes, which were marked by mortality of 100 % on day eight and at least 80 % on day six post injection (Fig. 1b–m). This high degree of lethality was confirmed by repeating the experiment using non-overlapping dsRNA fragments (1 μg/μl) making off target effects improbable (Additional file 1: Table S3).
For comparison, we performed the same experiment with the orthologs of five RNAi target genes published in the seminal paper of Baum et al. [15], which caused lethality in the western corn rootworm (WCR) upon dsRNA ingestion. Indeed, the Tribolium orthologs of these genes induced a high degree of mortality, but especially with low dsRNA concentrations, the mortality did not reach the one of the 11 candidates identified in this study (Fig. 1n–p, Additional file 2: Figure S1 B).
In order to check for stage dependence of the lethal effect, we injected dsRNAs targeting these 11 genes into ten adult beetles, respectively. Mortality rate on day eight post treatment was at least 90 % with 100 ng/μl dsRNA concentration while injection of 3 ng/μl dsRNA led to lower degree of mortality indicating a concentration response curve at these concentrations (Additional file 2: Figure S2). In summary, we identified 11 novel RNAi target genes that efficiently and rapidly induce lethality at larval, pupal and adult stages even at low doses of dsRNA and are more efficient than previously used target genes at least in Tribolium.
Double RNAi led to additive but not to synergistic effects
We asked, whether the lethality of RNAi treatments can be increased synergistically by combined injection of two dsRNAs targeting different essential genes. All 55 pairwise combinations of the 11 top RNAi target genes were injected into larvae at the same end concentration as the single injections (i.e. either 0.5 ng/μl of one dsRNA or 0.25 ng/μl of two dsRNAs summing up to 0.5 ng/μl end concentration) and the survival was documented (Additional file 1: Table S4). We find no indication for synergism. Instead, the observed deviations from the baseline in some combinations are explained by additive effects: The most efficient targets become less penetrant when “diluted” with less effective dsRNAs (e.g. Fig. 2a) while less efficient targets become more potent when supplemented with stronger target genes (e.g. Fig. 2f). In conclusion, we found no indication for synergistic effects that would be able to significantly enhance the technique (Fig. 2).
Degree of sequence conservation does not strongly influence the number of off-targets
In order to protect non-target organisms it would be desirable to use dsRNA fragments that are specific to the pest species and do not contain sequences targeting genes in non-target organisms (off targets). Therefore, we asked whether protein sequence conservation of our RNAi target genes correlated with the number of potential off target sites in other species. On the protein sequence level most of our RNAi target genes showed a strong conservation between some well sequenced species covering insect diversity (Drosophila melanogaster, Aedes aegypti (Diptera), Apis mellifera (Hymenoptera), Acyrthosiphon pisum (Hemiptera). L10, L67 and L82 were the least conserved (Fig. 3a). For protein L76 no ortholog could be identified in Aedes aegypti (Fig. 3a; see Additional file 2: Figure S3 H for phylogenetic analysis).
DsRNAs are processed by the enzyme Dicer into 21–23 nt long short interfering RNAs (siRNAs). After incorporation into the RNA-induced silencing complex (RISC), they serve as template to recognize the complementary mRNA and target it for destruction [9]. However, siRNAs with an exact sequence identity of ≥17 nt can already induce off target effects [25]. Therefore, we searched the nucleotide sequences of the 11 Tribolium target genes against the well annotated NCBI transcriptome databases of the abovementioned species for ≥17 bp long stretches of identity (http://blast.ncbi.nlm.nih.gov/Blast.cgi) [26]. For visualization, we plotted these putative off target regions targeting in genes of other species against the RNAi target gene nucleotide sequence (Fig. 4).
We did not find an overt correlation between the conservation of a protein and the number of potential off target regions (compare the most diverged genes L10, L67 and L82 with more conserved genes in Fig. 4). Likewise, within a given gene we found no enrichment of potential off target regions in more conserved stretches of the sequence (e.g. conserved protein domains) compared to less conserved stretches (e.g. non-coding UTRs; Fig. 4). Importantly, the location of potential off target sites was generally different for the different species. Together, these observations indicate that the number and location of the off target sites does not strongly correlate with protein sequence conservation. Further, given the vast diversity of taxa even in small habitats, it will be difficult if not impossible to identify target sequences without potential off-target site in any other species.
Note that most of the identified potential 17 bp off-target-regions will not lead to any RNAi response and that at least half of the potentially targeted genes will not lead to lethality [24]. Further, only individuals that actively eat the protected plants will suffer. Hence, the species specificity of the RNAi based technique remains unchallenged when compared to other methods that usually target all or at least many species. Nevertheless, efforts to increase safety should focus on selected species (like bees) that need to be protected in the respective given ecological setting.
GO term cluster identifies the proteasome as prime RNAi target gene
We tried to identify common properties of our identified RNAi targets, because this information might help identifying novel RNAi targets in species less amenable to large scale screens. We first analyzed the adult body expression levels in Tribolium and compared them to their Drosophila orthologs [27]. A striking pattern was not found apart from generally high expression of the Drosophila orthologs, which was specifically true for the central nervous system (Additional file 2: Figure S3 M).
Next, we searched for GO term clusters of the top 11 and top 40 RNAi targets [28, 29] using Drosophila melanogaster GO term annotations as the background. The clustering of the 40 RNAi target genes resulted in 10 clusters (Additional file 2: Figure S4). In order to test, in how far these GO terms are predictive, we identified 1328 Drosophila genes sharing the same GO term combinations of at least one of these clusters. For those, we determined the respective Tribolium orthologs and found that 502 of them had by chance been included in the iBeetle screen. Almost all novel genes identified by GO term combinations of cluster 1 and 2 (GO terms related to proteasome function) showed a strong lethality in the iBeetle screen (≥70 % after pupal or larval injection; Fig. 3b, c). Cluster 7 (GO terms related to cytoskeleton organization) represented the third-best cluster with 37 of 66 novel genes showing lethality in the screen (Fig. 3b, c). The clustering of the top 11 RNAi target genes did not result in clusters, which were able to predict novel RNAi target genes, which could be due to the low number of input genes, which makes statistical analysis challenging (not shown). Taken together, this analysis reveals GO term combinations that are predictive for potential RNAi target genes and reveal the proteasome as prime target for RNAi based insecticides.