By Christopher Aston
RNA interference is fast becoming an invaluable, albeit unpredictable, tool in the molecular biologist’s armory. Researchers who utilize RNAi in human cells expressed concern in 2001 when transitive RNAi was identified in C. elegans. Transitive RNAi is triggered by RNA polymerase-mediated elongation of siRNAs, using the target mRNA as a template. The elongated dsRNA itself becomes a substrate for the ribonuclease, DICER. The resulting secondary siRNAs silence the target gene further, accounting for the potent and persistent silencing observed in this model organism. More troubling, the secondary siRNAs can also be homologous to other genes, leading to “off-target” silencing.
Because microarrays reveal genome-wide expression patterns, they are valuable tools for identifying off-target silencing. Demonstrating such events in human cells has been a competitive endeavor. Two siRNA papers that appeared this summer concluded that RNAi blocks only the targeted gene, leaving other genes unfettered. A third paper offered a diametrically opposing viewpoint.
In Proceedings of the National Academy of Sciences on May 27, Jen-Tsan Chi and colleagues in Patrick Brown’s group at Stanford published a quest for off-target effects following RNAi silencing of green fluorescent protein (GFP) transfected into human embryonic kidney cells. GFP is a non-human gene, so silencing it has no downstream transcriptional effects. More importantly, no other gene expression perturbations could be detected by microarray.
A back-to-back paper by Dimitri Semizarov and colleagues from Abbott Laboratories reported similar findings. Furthermore, Chi and colleagues specifically ruled out transitive RNAi in these cells by fusing GFP to the actin gene. Even though the GFP-actin fusion was degraded, there was no subsequent degradation of a co-transfected actin-luciferase fusion, demonstrating that secondary actin siRNAs had not been generated.
In the conflicting paper in the June issue of Nature Biotechnology, Aimee Jackson and colleagues at Rosetta Inpharmatics silenced another exogenous gene, luciferase, in human HeLa cells. Notably, they also identified multiple off-target perturbations. Off-target silencing was also observed using siRNA to an endogenous gene, MAPK14. Conventional wisdom holds that siRNA has exquisite specificity, requiring near-perfect identity.
However, after analyzing the off-target mRNA sequences, they found partial identity to the MAPK14 siRNA duplex, demonstrating that RNAi is not target-specific. It is not clear whether transitive RNAi may be responsible for some other off-target effects in genes that did not share sequence identity.
These different conclusions from seemingly similar experiments can only confuse scientists concerned that RNAi effects in humans are non-specific. With RNAi being billed as a potential therapy that exclusively suppresses expression of disease alleles without affecting the normal allele, specificity is a key issue that needs to be addressed before this technology can be developed for the clinic. To what extent can transitive or non-specific RNAi be minimized through careful design of siRNA constructs? Can these effects be accounted for by predicting “collateral damage” to bystander genes? What is clear is that microarrays will continue to play a role in experiments examining RNAi effects at the genomic level rather than at the single-gene level.