Minimizing off-target effects by using diced siRNAs for RNA interference

Microarray studies have shown that individual synthetic small interfering RNAs (siRNAs) can have substantial off-target effects. Pools of siRNAs, produced by incubation of dsRNAs with recombinant Dicer or RNase III, can also be used to silence genes. Here we show that diced siRNA pools are highly complex, containing hundreds of different individual siRNAs. This high complexity could either compound the problem of off-target effects, since the number of potentially problematic siRNAs is high, or it could diminish the problem, since the concentration of any individual problematic siRNA is low. We therefore compared the off-target effects of diced siRNAs to chemically synthesized siRNAs. In agreement with previous reports, we found that two chemically synthesized siRNAs targeted against p38α MAPK (MAPK14) induced off-target changes in the abundance of hundreds of mRNAs. In contrast, three diced siRNA pools against p38α MAPK had almost no off-target effects. The off-target effects of a synthetic siRNA were reduced when the siRNA was diluted 3-fold in a diced pool and completely alleviated when it was diluted 30- or 300-fold, suggesting that when problematic siRNAs are present within a diced pool, their absolute concentration is too low to result in significant off-target effects. These data rationalize the observed high specificity of RNA interference in C. elegans and D. melanogaster, where gene suppression is mediated by endogenously-generated diced siRNA pools, and provide a strategy for improving the specificity of RNA interference experiments and screens in mammalian cells.