Russian roulette

Which bases should you choose for the seed region of a single siRNA?

It’s like Russian Roulette on full-automatic, where a specific seed will result in dozens or hundreds of down-regulated genes.

If you’re lucky, none of the off-targets results in a false-positive phenotype.   But odds are that you won’t be so lucky.

A recent paper suggests that taking single siRNA drugs may be closer to real Russian Roulette than anyone would hope.

The authors show that an siRNA designed to knock down human Huntington protein (HTT), and which had shown no adverse effects in Rhesus monkeys, was toxic when administered to mice.  The siRNA had been designed to produce minimal off-target effects in human, macaques, and mice, confirming the extreme difficulty in predicting off-target effects.

By modifying a single nucleotide in the siRNA seed region, the authors were able to create a version not toxic for mice.  Whether it will also be non-toxic for macaques?  That experiment must be repeated.  Single siRNAs create a lot of work.

Using miRNA target identification algorithms, the authors identify a number of potential off-target genes that may be responsible for toxicity.  For 3 of 4 genes, they show by qPCR using brain tissue that there is significant off-target silencing by the HTT siRNA (HDS1) compared to control (Ctl):

But when they try to confirm these results in an immortalised mouse neuronal striatal cell line, only one gene (Bcl2) is still being silenced (U6 is an additional vector control):

The authors comment:

In contrast, Sdf4 and Map2k6 expression was not reduced by overexpression of miHDS1 (Figure 2D and E), suggesting that these genes may not be direct off-targets in vivo, and may reflect indirect effects of Htt suppression over time or off-target suppression in non-neuronal cells, although AAV2/1 transduces primarily neurons.

Another possibility is that Sdf4 and Map2k6 are regulated by one or more genes that are off-targets of HDS1, but which are differentially expressed in live animal cells and the immortalised cell line.

The pool of accessible off-targets (and the resulting off-target footprint) differs with each transcriptional profile.

One should not assume that single siRNAs used in different cell types, or even under different experimental conditions, will behave the same way.

Using single siRNAs as normalisation controls may be especially problematic.

Why take risks with your RNAi experiments?  Use siPools for consistent, robust phenotypes.