This is today’s 2nd post on recent papers describing the beneficial effect of things gone wrong. The first one was from the area of bioinformatics, this one is about gene silencing by RNA interference (RNAi). A lot is known about the mechanism and application of RNAi, see e.g. the Wikipedia entry. Basically, you can use RNAi to downregulate a gene of interest by designing a short oligonucleotide (siRNA) complementary to the target mRNA and getting it into the cell to do its magic. Inside the cell, the oligonucleotide pairs with the target mRNA and the resulting duplex RNA is recognized by a specific cleavage machinery, which is present in the cell anyway. Most likely, this RNAi cleavage machinery has not been put into the cell by an intelligent designer to facilitate our lab work, but rather serves as an antiviral defense mechanism: double-stranded RNA is not normally observed within the cell and is interpreted as a sign of viral infection and thus removed.
It is somewhat naive to assume a clean downregulation of just a single targeted gene. One of the things I am doing to pay my rent is the bioinformatical analysis of microarray data. In this context, I often encounter transcriptional profiles that are the result of RNAi experiments. What you typically get is a mixture of the following effects:
- Target effect. This is what you want to see – your gene of interest is downregulated.
- Off-target effects. RNAi does not require perfect matching; your siRNA oligo will probably form imperfect duplex strands with several other (unwanted) mRNAs leading to their downregulation
- Secondary effects. Downregulation of the target genes (both wanted and unwanted) might lead to expression changes in other genes. Particularly pronounced when your target is a transcription factor.
- Unspecific dsRNA response. Cleavage is not the only antiviral response to double-stranded RNA. Many dsRNAs also stimulate the innate immunity pathway, e.g. mediated by toll-like receptors (TLRs).
The paper I am talking about has just appeared on Nature’s AOP page. It is from Mark Kleinman et al, the title is “Sequence- and target-independent angiogenesis suppression by siRNA via TLR3“. It is a very instructive story and goes to show that controls are a very important thing to do. The original idea was to to design a 21bp siRNA targeted against the vascular endothelial growth factor VEGFA with the idea to prevent a neovascularization in the eye that can lead to blinding. Apparently, the RNAi experiments did show the expected effect. However – surprise surprise – it turned out that this effect was completely unrelated to VEGFA downregulation and could be mimicked by any old siRNA. It looks like the last confounding factor in my list was the dominant one: in the studied system, dsRNA lead to an innate immune response mediated by TLR3 , with the typical concomitant induction of gamma-interferon and interleukin-12.This just emphasizes the notion that getting the desired effect does not necessarily mean that your experiment works. Or, as a former labmate of mine would put it, “if your experiment gives the the expected result first time, there must be something seriously wrong”.
Note added in proof: I just saw that Eric at the Futile Cycle has also reported on the same story.