Journal Club: Unraveling the Substrate Specificity of RIG-I and what it Means for RNAi Therapeutics

A paper by the Hartmann group in Bonn, Germany, takes the mystery out of the structural features of RNAs that activate the cytosolic viral innate immune sensor RIG-I and offers simple design rules for either avoiding, or in some cases intentionally inducing such activities with RNAi triggers. It is good news for essentially all types of synthetic siRNAi as practiced today, but requires re-evaluation of some (but not all!) DNA-directed RNAi approaches, and those approaches that depend on the preparation of RNAi triggers by in vitro transcription rather than by synthetic means.

Some of the first wave of RNAi Therapeutics candidates that were rushed into the clinic were most likely based on pre-clinical efficacy results due to the induction of non-specific innate immunity. As an aside, it always ‘surprised’ me for example how companies with no track record in nucleic acid therapeutic development would suddenly claim to be the first ones to enter an RNAi Therapeutics candidate into the clinic. Innate immunity is based on the recognition of pathogen-associated molecular patterns, PAMPs, by cellular receptors that then initiate a powerful signaling cascade leading to the successful defense against viral and bacterial infections, often resulting in the death of the infected cell itself. Therefore, to avoid mis-interpretation of RNAi data and to ensure safety, the RNAi Therapeutics field needs to take into account the structural and sequence-specific signatures of nucleic acid triggers of innate immunity.

There are two classes of nucleic acid receptors relevant to this discussion. The first one are the toll-like receptors (TLRs) 3, 7, and 8 which recognize certain types of single- and double-stranded RNAs mainly in the endosomes. This is important when RNAi triggers are delivered from the outside, but not for DNA-directed RNAi. The second one is comprised of cytosolic PAMP receptors that both synthetic and DNA-directed RNAi triggers may encounter. While PKR was initially thought to be the main cytosolic receptor relevant to RNAi, it more and more emerges that the RNA helicase RIG-I is what the field needs to be mindful of, and is the subject of the present paper by Schlee and colleagues.

Before these findings, it had been thought that any RNA with a triphosphate chemical group at the 5’ end would induce RIG-I. Blunt-end double-strand RNAs of the size of siRNAs were also thought by some to have this capacity independent of a 5-triphosphate modification. The present findings, however, show that the confusion about the exact RIG-I substrate structural features arose from the origin of the RNAs used in those studies. Since 5’-triphosphate modifications are not routinely offered by synthetic RNA vendors, it has been convenient to use RNAs generated through in vitro transcription by recombinant, purified phage polymerases which leave a triphosphate group at the 5’ end. Unfortunately, these phage polymerases have the property of generating additional species of RNAs aside from the desired one. It turns out that double-stranded RNAs, still with a 5-triphospate group, are one of those and that these are the ones actually recognized by RIG-I. The authors were thus able to show that well-defined synthetic single-strand RNAs with a 5’-triphosphate alone were not sufficient to induce RIG-I.

For RNAi Therapeutics the findings mean that RIG-I should not be a concern for essentially all synthetic siRNA therapeutics, as almost all of them are administered in a 5-hydroxylated form which are then 5’-monophosphorylated. Both modifications would abolish RIG-I activation. This is also good news for blunt-end ‘Atu RNAi’-type siRNAs as practiced by Silence Therapeutics which may have been previously suspected to trigger RIG-I. It is true, however, that in the context of 5’-triphosphates, the more traditional double-stranded Tuschl-type siRNAs which contain 3’ overhangs further diminish RIG-I activity even in a 5’-triphosphate context.

The picture is somewhat more complex for DNA-directed RNAi Therapeutics approaches. Historically, perfect complementary hairpin RNAs driven by RNA polymerase III promoters (Pol III) have been used. As these hairpins are destined to be exported into the cytoplasm, the cellular location of RIG-I, the minimally or not at all modified 5’ ends of such shRNAs, i.e. 5’-triphosphates, run the risk of triggering RIG-I responses. However, a simple mismatch of the 5’-triphosphate nucleotide with the opposite strand should abrogate most RIG-I activity and some of the widely used Pol III expression cassettes fortuitously carry such mismatches which do not appear to affect gene silencing. For RNA Polymerase II-driven DNA-directed RNAi Therapeutics, RIG-I should not be a concern for the reason alone that Pol II transcripts exhibit a 5’ modification that does not induce RIG-I. And finally, for trans-kingdom RNAi Therapeutics where the bacteria expresses the RNAi trigger, one may want to design shRNAs that are similar to the Pol III strategies. It is also good news that not any 5’-triphosphate RNA induces RIG-I, since bacterial transcripts are quite rich in 5’-triphosphates.

In summary, the paper by Schlee and colleagues indicates that innate immune activation by siRNAs in the cytosol is no show-stopper by any means. By contrast, since 5-triphosphates do not necessarily abolish the RNAi activity of double-stranded RNAs, bi-functional immunostimulatory siRNAs can be designed such as for antiviral and cancer applications- as demonstrated late last year by the same group in a collaboration with Alnylam. Of course, 5’-triphosphate dsRNAs may also be used independently of RNAi for the same reasons.