The Importance of 3’ Overhangs in RNAi Therapeutics

[Note: some of what follows, has already been discussed in the entry ‘On the Importance of Being Tuschl’]

There has been a flurry of activity surrounding the Tuschl patents. It appears that at stake is no less than the Tuschl II patent which claims the use of 3’ overhangs in siRNAs. The uncertainty comes from some of the owners of the Tuschl patent estates prosecuting the Tuschl I patent series such that it uses scientific data underlying the Tuschl II patents. Although I cannot see that 3’ overhangs are explicitly claimed in Tuschl I, the presence of such overhang data in Tuschl I could invalidate the claims in Tuschl II as their prior use would suggest that overhangs were already practiced in the art and/or trigger double-patenting/interference issues.

It has always been a surprise to me that virtually out of nowhere, Tuschl II data, generated at the Max Planck in Gottingen after his time at the MIT, would appear in Tuschl I, which is based on work at the Whitehead/MIT and UMass. The work underlying Tuschl I identified that short siRNAs are generated from longer dsRNAs (in Drosophila extracts) and that the short siRNAs are the likely mediators of RNAi. No mention of overhangs, and no reason to anticipate that overhangs would confer an advantage. The work by Elbashir, Lendeckel, and Tuschl that underlies Tuschl II characterized the siRNAs processed in Drosophila extracts and noted that they contained 3’ overhangs. Synthetic versions of these siRNAs were then found to mediate RNAi in Drosophila extracts and, in the famous Nature paper, human cells. The 3’ overhangs were not just functionally irrelevant consequences of RNase III processing, but found to confer a distinct advantage in RNAi silencing efficacy. This advantage has been borne out in many studies since by labs throughout the world, and the reason seems to be that the overhangs allow for efficient Argonaute loading of the guide strand. For a while, 3’ overhangs appeared to have an additional advantage over blunt-end siRNAs in that they would avoid some of the innate immune responses, although this is an area in which the jury is still out. Another finding that has been borne out by numerous studies is that 19-21 bp lengths work better than smaller or larger siRNAs (not discussing here Dicer-substrates though).

Tuschl I claims double-stranded RNAs of 21 to 23 nucleotides. Note that it says nucleotides, and not base pairs, so in theory it could be dsRNAs that do not have to be blunt ended. But again, at the time of the invention, nobody in the field would have been able to anticipate the presence, let alone an advantage of the overhangs. Tuschl II claims dsRNAs with individual strands of 19-23nt in length, with at least one of the ends having a 3’ overhang.

I do not want to comment here too much on the specifics of the case, and Doug Macron from ‘RNAi News’ has done a great job in following the story- except to say that it is clear that Max Planck (and Alnylam) would never have agreed to such use of Tuschl II data, and even if there had been some procedural issues (e.g. Tuschl ‘swore’ etc, when in reality most academic scientists trust that their patent agents would act in their best interest, and therefore just sign off patent documents that come across their desk/bench) I believe that at some point fiduciary duty has been violated by those hired and paid, also by Max Planck and Alnylam, to prosecute the Tuschl patents.

While I consider it possible that some fancy workarounds of the Alnylam IP estate may give certain freedom-to-operate to some companies, scientifically the sweet spot of siRNAs are those 19-21bp in length with at least one 3’ overhang end and the drug development economics would dictate these to be used and licensed. This area is very well covered by Alnylam through Kreutzer-Limmer, Tuschl I and II, although an unexpected outcome of the Tuschl proceedings could change the picture. On the other hand, one cannot dismiss the possibility that some non-Tuschl-like structure in combination with a certain modification pattern similar to what Silence Therapeutics claims e.g. actually works and is advantageous contrary to expectation, that is non-obvious, and I'd be happy to look at data that systematically demonstrates this.