The favored technique for inducing RNAi is to introduce short interfering RNA (siRNA). These are double-stranded pieces of RNA, generally 21 base-pairs long. One reason why short pieces are used is that longer double-stranded RNA (dsRNA) activates the cell’s interferon system, an innate and non-specific component of the immune system which effectively shuts down the machinery of a cell infected by viruses; double-stranded RNA is produced during viral replication and gene expression. Carol Sledz and her colleagues at the Lerner Research Institute in Cleveland, Ohio, discovered that siRNAs also stimulate the interferon system, sounding a note of caution that "siRNAs have broad and complicating effects beyond the selective silencing of target genes when introduced into cells."

Using a panel of seven synthetic siRNAs, each specific to a different gene, Sledz and her team saw increased activity in more than 50 genes that are normally activated by interferon. This effect is entirely non-specific, although the different siRNA molecules affected different components of the interferon system somewhat differently. And it depends crucially on the presence of dsRNA-dependent protein kinase (PKR), which is activated by siRNAs and which is required for interferon activation. But the absence of PKR does not block RNA interference itself.

Aimee Jackson and her group at Rosetta Inpharmatics in Kirkland, Washington, were also not impressed with previous claims of siRNAs’ "exquisite sensitivity." In their paper at #4 they point out that most previous studies of gene silencing by siRNAs looked at only "one or a few genes in addition to the targeted gene, an approach not unlike ‘looking for keys under the lamppost’." Jackson’s group felt their way out into the dark, creating 16 siRNAs for one gene and 8 for another and then looking at the overall pattern of gene expression after inserting each siRNA into a human cell line. Each siRNA inhibited its target gene right enough, but each also had a completely different impact on all the other genes in the cell. Furthermore, siRNA targeted to a gene that has no known human homolog also reliably affected several other genes.

Examining in detail the genes affected by the siRNAs, Jackson’s group was able to divide them into different groups depending on how quickly they were affected. A group of nine genes were shut down effectively at the same time as the target gene, suggesting that the mechanism was direct, rather than depending on some interaction of the siRNA with its intended target. And all of those genes contained a sequence partially identical to the siRNA sequence. In some the similarity was a core of 14-15 nucleotides in the middle of the sequence. In the others the similarity was confined to 9 nucleotides at one end of the sequence. A similar pattern showed up in the genes affected by another two siRNAs.