Can antisense oligonucleotides targeting transcription termination windows suppress gene expression?

A target region outside the annotated gene

Most antisense oligonucleotide (ASO) designs aim at mature mRNA or at introns inside the gene body. This study asked whether an ASO can instead bind the nascent RNA that RNA polymerase II keeps making after it has passed the polyadenylation signal (PAS), the site where the pre-mRNA is normally cut and polyadenylated. That stretch of transcription is called the termination window; in human cells the authors cite a median length of 3.3 kb, with some windows extending beyond 10 kb from the annotated gene end.

The authors call these molecules downstream-of-gene ASOs, or DG-ASOs. They used 16-nucleotide gapmer ASOs: a 10-nucleotide DNA gap flanked by three or four LNA-modified nucleotides, with a full phosphorothioate backbone. The DNA gap is what allows RNase H1, an enzyme that cuts the RNA strand of a DNA:RNA hybrid, to cleave the target. In the main experiments DG-ASOs sat within 0.5 kb downstream of the most distal PAS, and for KPNB1 one was placed 2 to 2.5 kb downstream of the two main proximal PASs and still worked.

The mRNA drop depends on RNase H1 cleavage

DG-ASOs produced a pronounced decrease in CHMP1A and KPNB1 mRNA measured by RT-qPCR, normalised to GAPDH and to the negative control. Three independent lines of evidence tie that decrease to RNase H1 rather than to simple occupancy of the RNA. First, depleting RNase H1 with siRNA rescued both mRNAs. Second, overexpressing the catalytically dead D210N RNase H1 mutant, which binds DNA:RNA hybrids but cannot cleave them, suppressed the knockdown efficiency of all ASOs tested relative to wild-type RNase H1. Third, mixmer ASOs that bind the same sites but cannot recruit RNase H1 did not reduce either mRNA.

The rescue was not equally clean for both genes. The authors note that the KPNB1 effect after RNase H1 depletion was less pronounced, which they attribute to poor co-transfection efficiency of the siRNA together with both gene-body and DG-ASOs. That is a technical caveat on one arm of the evidence, not on the overall requirement for RNase H1, which the catalytically dead mutant and the mixmer experiments support independently.

Cleavage outcompetes the 3' end processing machinery

To see what happens at the 3' end, the authors built RT-PCR primers that separate mature mRNA from a 3'-extended form. One reverse primer sat upstream of the PAS and amplified both species; a second sat beyond the PAS but before the DG-ASO binding site and amplified only the extended, RNase H1-cleaved form. Because the products were 283 to 558 bp and unsuitable for qPCR, the reaction was stopped every two cycles and read on agarose gels.

In DG-ASO-treated cells the amplicon upstream of the PAS fell, matching the drop in steady-state mRNA. The product representing the uncleaved 3'-extended form rose. The ratio of signal beyond versus upstream of the PAS moved from 0.37 to 0.02 for CHMP1A and from 1 to 0.08 for KPNB1. The authors read this as RNase H1 cutting the nascent RNA downstream of the PAS before the cleavage and polyadenylation factor (CPF) complex can cut at the PAS, so the transcript never receives a correct 3' end.

Faulty transcripts are polyadenylated and sent to the exosome

An unprocessed 3' end does not simply sit in the nucleus. The authors used a dTAG system to deplete DIS3, the catalytic subunit of the exosome, and confirmed depletion by western blot and by accumulation of three PROMPTs. With DIS3 depleted, the 3'-extended CHMP1A and KPNB1 transcripts that accumulate after DG-ASO treatment increased further, while mature mRNA levels were also affected.

The paper's model is that the misprocessed pre-mRNA is still polyadenylated but is recognised as faulty and degraded by the exosome. This matters for interpreting the mRNA readout: the decrease in mature mRNA is not only a failure to produce a transcript, it is the visible end of a degradation route that starts with an incorrectly placed cut.

XRN2 gets an earlier entry point in the window

Transcription termination normally works through XRN2, a 5'-to-3' exonuclease that loads onto the free RNA end created when CPF cuts at the PAS and then degrades nascent RNA while chasing the slowing polymerase. A cut placed earlier in the termination window should give XRN2 a head start. Chromatin-associated RNA sequencing supported this: read coverage in the 5' regions of CHMP1A and KPNB1 was unaffected, so DG-ASOs did not change transcription initiation, while coverage downstream of the ASO binding site fell.

The functional test used amplicons spanning the termination window as a proxy for read-through transcription. DG-ASOs reduced nascent RNA downstream of their binding sites, and depleting XRN2 with siRNA restored those levels to control for both genes. XRN2 depletion alone did not change CHMP1A or KPNB1 mRNA levels, which the authors take to mean XRN2 is not responsible for degrading the RNase H1 cut products that feed the mRNA readout; its role here is the accelerated removal of downstream nascent RNA.

The nascent RNA measurements also set a limit on the mechanism. By chromatin-associated RNA-seq, DG-ASOs only moderately changed nascent RNA from the two loci (log2 fold change −0.8 for CHMP1A and −0.5 for KPNB1, versus −1.5 and −0.9 for the gene-body ASOs). The authors conclude that most of the degradation happened post-transcriptionally, not by blocking transcription at the start of the gene.

Termination windows are also an off-target liability

The same mechanism that makes DG-ASOs useful makes unintended binding downstream of other genes dangerous. The authors tested a promiscuous ASO (pASO) designed against HIF1A, which is also perfectly complementary to nine other mRNAs and has an imperfect but favourable match, with one terminal mismatch, to a site 236 nucleotides downstream of the CHMP1A PAS. In THP1 cells, that off-target site in the CHMP1A termination window produced knockdown comparable to the intended on-target effect.

They then asked whether a common chemistry fix could separate the two. Adding a single 2'-O-methylation at different positions of the pASO central gap changed the off-target effect on CHMP1A, showing that the placement of such a modification can shift the balance between intended and unintended knockdown. The practical point is that current off-target checks align ASO sequences against transcriptome databases containing mature mRNA, pre-mRNA and noncoding RNAs; termination window sequence is not directly annotated there, so it is not screened.

Where the window helps, and where the evidence stops

The authors frame the termination window as extra design space. Their example is ATF4, a 2.1 kb transcription factor whose termination window is about 3.5 kb in HEK293 cells, so including it would extend the targetable sequence to 5.6 kb. They also suggest DG-ASOs could suit intron-less genes, allele-specific knockdown of single-nucleotide variants, and highly structured RNAs where hybridisation inside the gene body is difficult.

The boundaries are specific. The main DG-ASOs sat within 0.5 kb of the most distal PAS, and although the KPNB1 reagent worked 2 to 2.5 kb from the proximal PASs, the authors expect the effective distance to be set gene by gene, cell type by cell type and organism by organism, most likely depending on poly(A) signal strength. The work was done in HEK293, HeLa, SH-SY5Y and THP1 cells with transfected ASOs, plus carrier-free uptake at 5 µM for 7 days, so the potency seen at 20 nM transfection is not a clinical dosing result.

The most decision-relevant split is therefore not whether DG-ASOs work but which side of the ledger a given sequence falls on. A candidate ASO that matches a termination window of a non-target gene can silence that gene through RNase H1, exactly as an on-target reagent would, and the paper's own pASO case shows this happening at a site with only one terminal mismatch. Screening design candidates against termination window sequence, and checking where 2'-O-methylation placements move the on-target versus off-target balance, is the concrete priority this study sets.

Sources used in this answer