Off-Target Effects Plague Drosophila RNAi

Two new studies suggest a major body of research may need to be reevaluated

By | September 11, 2006

Thousands of hits from dozens of genome-wide RNA interference screens in Drosophila cell cultures may need to be reevaluated, according to a pair of studies published this week in Nature and Nature Methods. The two reports demonstrate for the first time that so-called "off-target effects" (OTEs) can occur in experiments using long double-stranded RNA triggers for RNAi. "This [Nature] paper is potentially relevant to folks using libraries that can produce multiple triggers, multiple small RNAs," Greg Hannon of Cold Spring Harbor Laboratory said. Hannon, who did not participate in either study, noted that a 500-bp dsRNA could yield two dozen or more short RNAs, some of which could conceivably go astray. "So you run a primary screen, and you'll get a mix of on- and off-target effects. However, the ways that these screens are presently done solves this problem through the use of multiple independent dsRNA triggers." Off-target effects ? knockdown of genes other than the intended target ? have long been recognized as a problem for RNAi screens conducted using short RNA triggers. But genome-wide RNA libraries in Drosophila employ double-stranded RNAs hundreds of bases long, which are then processed intracellularly into the short RNAs that actually carry out RNAi. These longer dsRNAs were not thought to produce OTEs, because any non-specific effects due to short homology regions would be more than balanced by the preponderance of specific, on-target effects. The new studies now suggest that assumption is false. Using a library of more than 21,000 dsRNAs, a Wingless-responsive luciferase reporter, and Drosophila S2R+ cells, Philip Beachy, of the Johns Hopkins University School of Medicine in Baltimore, and his group identified seven promising candidates for novel Wingless pathway components. The team attempted to validate those hits by synthesizing additional dsRNAs against each target, which would be expected to reproduce the original results. But only those dsRNAs whose sequence overlapped the original dsRNAs worked as anticipated. First author Yong Ma, a postdoc in Beachy's lab, told The Scientist that he determined that all seven candidate genes contained short regions of homology (<20 bp) to armadillo, a member of the Wingless pathway, and that the dsRNAs could knock down armadillo mRNA and protein levels accordingly. That suggested that the effects he observed were really due to OTEs on armadillo. When Ma looked back at another published RNAi screen on the Wingless pathway, he found that a disproportionate number (about 60%) of putative negative regulators share a repeated trinucleotide sequence, CAN, which is found in only about 5% of the library. "We suspect that maybe the reason these dsRNAs seem like negative components is that there's an off-target effect based on the CAN repeat," he said. Norbert Perrimon, who coauthored the earlier Wingless study and created the dsRNA library Beachy used, and who runs an RNAi screening service at Harvard Medical School, said, "Initially, no one in the field had seen an off-target effect, either in Drosophila or in C. elegans, but as more screens were done, we realized these things could happen." Perrimon coauthored the Nature Methods study, which detected evidence of OTEs in a retrospective analysis of 30 RNAi screens performed at his screening center. "That basically contaminates your dataset, because you are getting what you should get, plus some junk," he said. "The question is, how do you remove this junk?" Perrimon and Ma recommend culling libraries of dsRNAs with significant homology to other targets, and validating hits with additional, non-overlapping dsRNAs to the same target. Perrimon has already updated his library, removing dsRNAs with more than 19 bases of homology to other targets, and adding additional dsRNAs per transcript. He has also retrospectively curated some published RNAi datasets to remove suspected false positives. In some cases, Perrimon said, the false positive rate is as low as 15%. But in one 2005 study, a published list of 509 hits was narrowed down to about 300. These removed entries are not necessarily false, Perrimon cautions, just suspect. "These need to be repeated with new dsRNAs, which we are doing now," he said. A detailed commentary will be published shortly on his screening service website ( to address the issues, he added. Alex Bishop, assistant professor of cellular and structural biology, University of Texas Health Science Center, San Antonio, used Perrimon's original library to identify genes that are required for viability following exposure to DNA damage in a study that remains unpublished. Of the 400 or so high-stringency hits he gets per screen, about 80% have proven to be real. "My screen is based on survival rather than luciferase expression, and we are not selecting for a pathway that is enriched with armadillo repeat containing proteins, so it's been successful, " he said. "Too successful. We have too many hits." Jeffrey M. Perkel Links within this article Y. Ma et al., "Prevalence of off-target effects in Drosophila RNA interference screens," Nature, Sept. 10, 2006, doi:10.1038/nature05179. M.M. Kulkarni et al., "Evidence of off-target effects associated with long dsRNAs in Drosophila melanogaster cell-based assays," Nature Methods, October 2006 Greg Hannon Philip Beachy Norbert Perrimon Drosophila RNAi Screening Center K. Nybakken et al., "A genome-wide RNA interference screen in Drosophila melanogaster cells for new components of the Hh signaling pathway," Nature Genetics, December 2005, doi: 10.1038/ng1682. PM ID: 16311596


Avatar of: Jeffrey Perkel

Jeffrey Perkel

Posts: 2

September 11, 2006

Norbert Perrimon's Drosophila RNAi Screening Center has now posted its commentary on RNAi off-target effects. Entitled "Matter Arising: Issues of off-targets in Drosophila RNAi screens," the commentary is available here:\n\n

Popular Now

  1. Running on Empty
    Features Running on Empty

    Regularly taking breaks from eating—for hours or days—can trigger changes both expected, such as in metabolic dynamics and inflammation, and surprising, as in immune system function and cancer progression.

  2. Athletes’ Microbiomes Differ from Nonathletes
  3. Gut Feeling
    Daily News Gut Feeling

    Sensory cells of the mouse intestine let the brain know if certain compounds are present by speaking directly to gut neurons via serotonin.

  4. Immune Cells Deliver Cancer Drugs to the Brain