In 2006 Long-Cheng Li et al.¹ first demonstrated that small double-stranded RNAs could activate gene expression in a process they called RNAa (dsRNA-induced gene activation) as opposed to scientifically accepted RNAi (RNA interference) processes which down-regulate gene expression in various ways. The paper was finally published in PNAS after two years of rejections from other high impact journals.² It was initially greeted with scepticism by the scientific community even though Li et al. presented valid data on two different targets using several different cell lines. The main criticism was that the authors could not demonstrate a plausible mechanism of action beyond the fact that Ago2, which at the time was known to be part of the RNAi machinery, was involved. However, soon after publication, a second paper from an independent group, Janowski et al.³, also demonstrated RNAa and showed that there was a 24 h to 48 h delay in the onset of gene activation which then persisted for up to 10 days; kinetics distinctly different from RNAi.

Despite the fact that the RNAa mechanism remained unclear^4,5 it took only ten years between first publication and first application to the clinic (see MiNA Therapeutics⁶). How ironic then that two recent papers have finally shed some more light on how exactly RNAa works.

To clarify if RNAa directly affects active transcription, Portnoy et al.⁷, the first of these papers published, used a nuclear run-on assay in combination with RT-qPCR to measure nascent transcript levels in prostate carcinoma (PC-3) cells transfected with short activating RNA (saRNA) to P21 (saP21) or E-cadherin (saEcad) that targeted the promoters of the p21 and E-cadherin genes respectively. They demonstrated a 9- and 28-fold increase in nascent mRNA transcripts from the respective target genes compared to control saRNA or mock-transfected cells while mRNA decay kinetics remained the same in all cases.

Using ChIP (Chromatin Immunoprecipitation) to scan the region around the saP21 target site, Portnoy et al. established that Ago2 binding to genomic DNA is significantly enriched around the saP21 target sequence at -322 upstream of the transcriptional start site (TSS) in saP21 treated cells. This was accompanied by a very large increase in RNA polymerase II (RNAPII) binding at the saRNA target site and core promoter. At the 5` initiation site Ser5 phosphorylated RNAPII (paused polymerase) declined while Ser2 phosphorylated RNAPII (elongating polymerase) increased at the 5` site and throughout the transcribed region of p21. This is consistent with the saRNA triggering transcriptional initiation and productive elongation of message and can explain the increased nascent mRNA seen in the nuclear run on assay.

In an attempt to identify proteins associated with the promoter complex Portnoy et al. use biotinyated saRNA (that retained saRNA activity) and magnetic steptavidin beads combined with protein purification and mass spectrometry to identify 27 protein candidates that were found to be uniquely associated with the saP21 antisense guide strand (the active strand in this saRNA). Focusing on two of these proteins, they showed that RHA (ATP-dependent RNA helicase A) and CTR9 (RNA polymerase-associated CTR9 homolog) both co-immunoprecipitated Ago2 and RNAP II in cells transfected with saP21, but not (CTR9), or at much lower levels (RHA), in sa-control treated cells. Using RNAi to knock down either CTR9 or RHA led to attenuation of saP21, saEcad and saKLF4 translational activation.

Interestingly, members of the RISC-loading complex such as Dicer, TRBP (HIV-1TAR RNA binding protein) and TNRC6 (Trinucleotide repeat-containing gene 6A) were not found in these experiments by Portnoy although it is possible this may have been due to the sensitivity of the approach used in this study.

Thus, RNAa is associated with increased transcription initiation and elongation by RNAPII leading to nascent mRNA production.

Portnoy and colleagues also found that in a PC-3 subclonal cell line with CRISPR induced deletions of 2/3rds of the saP21 target sequence on both alleles, saP21 transcriptional activation was reduced to levels barely above those seen in cells treated with control saRNA, demonstrating target specificity.

In similar experiments Meng et al.⁸ created a HEK293A cell line using CRISPR with a 8 bp deletion on one allele and a 1 bp insertion on the other allele in the target sequence of the highly functional PR-1611 saRNA (targeting the progesterone receptor gene). This could potentially allow some residual functionality of the target sequence on the allele containing the 1 bp insertion, particularly as it is at position 14 counted from the 5` end of the antisense strand. Nevertheless, congruent with the results reported by Portnoy et al., transcriptional activation by the PR-1611 saRNA in the mutated subclonal cells was significantly reduced, while functional PR saRNAs with differing target sites remained active.

Thus, saRNA activation of transcription is specific for, and involves direct binding to, the target sequence.

Meng et al. also investigated the functionality of saRNAs with mutations in their own sequence, either as single base pair or single nucleotide changes. When the saRNAs PR-1611 and PR-11 (from Janowski et al.³) contained single base pair mutations, especially at positions 4, 6 and 8 counted from the 5` end of the antisense strand, transcriptional activation was severely reduced. A similar reduction was also seen when the saRNA PR-1611 contained single nucleotide mutations at positions 4 and 6 in the antisense strand only, suggesting that nucleotides in the seed binding sequence are critical for saRNA activity.

Meng et al. also employ biotinylated saRNAs, labelled on either sense or antisense strand, to pull down associated chromatin and then PCR amplify the genomic DNA using target gene specific primers. They report that there is a 4-fold enrichment of PR target DNA after transfection with PR-1611 saRNA biotinylated on the antisense strand compared to the sense strand. However, using the same method, Portnoy et al. demonstrate that for saP21 only the antisense strand is functional, whereas for saEcad both strands are equally active. The authors suggest that this may be due to the thermodynamic stability profile of the respective saRNA ends since in siRNA duplexes, Ago2 uses the strand with lower thermodynamic stability at its 5′ end relative to its 3′ end as the guide, and destroys the other (passenger) strand. Indeed, for both PR-1611 and saP21, the thermodynamic stability at the antisense strand 5` end is much lower than at the sense strand 5` end while both ends of saEcad are very similar.

It does not necessarily follow that equal loading of sense or antisense strand of a particular saRNA onto Ago2 will mean functionality of both strands in the RNAa process. In fact, data from Li et al.¹ had previously suggested that only the antisense strand of saEcad was functional.

Taken together, these data strongly suggest that RNAa is dependent on the antisense strand of the saRNA and that a seed sequence in the 5` end of the saRNA antisense strand is critical.

For further clarification of the RNAa mechanism, it would be interesting to see if RNAi is still functional in cells after downregulation of RHA and CTR9 when RNAa is clearly not. Also, how would single nucleotide mutations in the sense and/or antisense strand of saEcad, similar to those done in Meng et al. on PR-1611 as well as basepair mutations that increase the thermodynamic stability at the sense strand 5` end, affect RNAa activity? Could the biotinylated saRNA-chromatin pull-down method be adapted to allow detection of saRNA-Ago2 associated RNA sense transcripts, possibly in conjunction with DNase treatment?

It seems clear that there are similarities as well as differences between RNAa and nuclear RNAi. Whatever the mechanism though, by inducing transcription of target genes in a stringent sequence specific manner and requiring the addition of one small RNA component only, saRNAs have a distinct advantage over gene addition or even gene editing strategies.