Stem–loop structures can effectively substitute for an RNA pseudoknot in -1 ribosomal frameshifting

Yu, Chien-Hung; Noteborn, Mathieu H.; Pleij, Cornelis W. A.; Olsthoorn, René C. L.

In addition to the mentioned dnaX and HIV-1 gag-pol hairpins, other examples of frameshifter hairpins are found in HTLV-2 and CfMV ( Figure 6 ). HTLV-2 gag-pro features a perfect 10 bp hairpin with CUA tri-loop which induces 9% frameshifting in RRL (16) . The CfMV 2a-2b frameshifting hairpin consists of 12 bp, one cytidine bulge close to the top, and a stable UACG tetraloop and is capable of inducing 11% of frameshifting in a wheat germ cell-free system (WGE) (17) . What these hairpins have in common is their length of 10-12 bp, their relatively low number of mismatches and bulges, their small loops and their high GC content, especially in the bottom 6 bp. These features are also applicable to the good frameshifters from our dataset. Interestingly, these features do not all apply to the minimal IBV hairpin ( Figure 6 ) that is derived from the so-called minimal IBV pseudoknot. Despite its large size of 17 bp, absence of mismatches and bulges, presence of a small loop, the stability of the middle part of the hairpin, i.e. bp 5-9, is not very high. This could be the reason why its activity in RRL is 5-10 fold lower (22) than of its parent pseudoknot, whose activity is 42% (25) . Surprisingly, in our assays the frameshift-inducing efficiency of the IBV hairpin was 26% (data not shown), which is in stark contrast to the 4-8% reported by Brierley et al. (22) . This discrepancy may be due to experimental conditions: in our experiments we used non-capped transcripts, a 7-nt spacer and RRL from Promega whereas the Brierley's lab used capped transcripts, a 6-nt spacer and in-house prepared RRL. On the other hand, the 26% we obtained for the IBV hairpin would be a factor of 1.6 lower than the 42% reported for the IBV pseudoknot (25) , and is similar to the ratio of 1.4 and 1.6 we obtained for SRV in vitro and in vivo, respectively.