Unstable non-coding RNAs are produced from thousands of loci in all studied eukaryotes (and also prokaryotes), but remain of largely unknown function. The present review summarizes the mechanisms of eukaryotic non-coding RNA degradation and highlights recent findings regarding function. The focus is primarily on budding yeast where the bulk of this research has been performed, but includes results from higher eukaryotes where available.
- cryptic unstable transcript (CUT)
- RNA degradation
- unstable non-coding RNA
- Xrn1-sensitive unstable transcript (XUT)
Neatly interleaved among protein-coding genes in even the most compact genomes are transcription units for ncRNA (non-protein-coding RNA), but these ncRNAs have largely escaped attention owing to their ephemeral nature. While investigating the phenotypes of RNA degradation mutants in budding yeast, several labarotories uncovered a class of ncRNA that is undetectable in wild-type cells, but abundant in cells lacking the nuclear exosome (a key RNA degradation complex) or associated co-factors [1,2]. These RNAs were christened CUTs (cryptic unstable transcripts) to reflect their enigmatic and elusive character. Such RNAs are transcribed in wild-type cells and are widespread [3,4] (Xu et al.  identified 925 CUTs compared with 5171 expressed protein-coding genes in budding yeast), but they are instantly degraded and as such their functional importance is mostly unclear. More recently, other unstable RNA species have also been characterized, but their functions remain similarly mysterious. Figure 1 shows the general pathways for unstable ncRNA degradation so far described in budding yeast.
Genome-wide analyses in multiple organisms have revealed that RNA pol II (polymerase II) promoters are largely bi-directional [2,3,5–7]. Stable RNA species may be produced in both directions; however, a competing process occurs if the nascent RNA contains a sufficient density of binding sites for the RNA pol II-associated Nrd1–Nab3–Sen1 complex; in this case Nrd1 can direct transcriptional termination before commitment to elongation, leading to co-transcriptional recruitment of the RNA degradation machinery and rapid degradation of the RNA [8,9]. Nrd1 preferentially binds the RNA pol II CTD (C-terminal domain) when it is Ser5 phosphorylated ; since Ser5 phosphorylation decreases as transcription proceeds, termination only occurs given a sufficient density of Nrd1–Nab3-binding sites at the 5′-end of the transcript. This places an effective length limit of a few hundred base-pairs on CUTs .
Although the density of Nrd1–Nab3 sites appears to be the primary degradation signal in model CUT transcripts, this is not the whole story. Such sites are not absent from normal ORFs (open reading frame), and their distribution is not noticeably different between Nrd1-dependent CUTs and Nrd1-independent SUTs (stable unannotated transcripts) . A further complication is that the Nrd1–Nab3–Sen1 complex may be recruited to RNA independent of pol II CTD binding, as it can terminate transcription of long pol II RNAs and target RNA pol III products . Furthermore, the prolyl isomerase Ess1 was recently shown to be required for CUT degradation  and it is likely that other important regulators remain to be identified. Taken together, it seems that the factors instigating RNA degradation in budding yeast are very diverse and many subtleties are yet to be appreciated.
Human and plant cells also produce unstable RNAs and, like CUTs in budding yeast, these RNAs are localized at the 5′-end of known ORFs although their distribution is different. In plants, UNTs (upstream non-coding transcripts) are initiated upstream of the main TSS (transcriptional start site) and transcribed co-directionally through the promoter into the first exon ; in humans, PROMPTs (promoter-associated pervasive transcripts) occur in a region 0.5–2.5 kb upstream of the TSS and are transcribed in both directions . PROMPTs are also observed upstream of pol I and pol III promoters in human cells, suggesting that their production is a general feature of transcription . The observed variation in the character of UNTs and PROMPTs is reflected in the lack of clear Nrd1 and Nab3 homologues in humans or plants. So although CUT-like transcripts occur across evolution, their details and recognition mechanisms are variable.
The Nrd1–Nab3 complex interacts in vivo with TRAMP (Trf4–Air–Mtr4 polyadenylation complex), a multi-subunit RNA-processing complex containing Trf4 or Trf5, Air1 or Air2 and Mtr4, which polyadenylates CUTs before degradation [2,18,19]. Although polyadenylation is generally thought to stabilize RNAs, the addition of short poly(A) tails (recent evidence suggests that 3–5 bp are the in vivo lengths ) marks RNA for degradation by the nuclear exosome, which can be directly recruited by the TRAMP complex (reviewed in ). The nuclear exosome is a multi-subunit nuclease complex, containing an endonuclease and two 3′→5′ exonuclease domains, along with a large conserved structure probably involved in substrate selection (reviewed in ). Unlike Nab3–Nrd1, both TRAMP and the exosome are highly conserved across eukaryotes, so although the precise nature and targeting mechanisms of CUTs and CUT-like RNAs can vary, the fate of unstable transcripts in the nucleus is much the same.
The exosome, TRAMP and Nrd1–Nab3 can be co-purified from yeast, suggesting a close association between these complexes . Given that Nrd1–Nab3 bind the pol II CTD, the whole degradation machinery should be connected to the pol II holoenzyme. The TRAMP complex can be detected at CUT loci in yeast , whereas in Drosophila the exosome is directly attached to pol II via dSpt6 and is detectable at transcriptionally active sites on polytene chromosomes . This close association between transcription and degradation machineries explains the extremely rapid degradation that keeps CUTs undetectable in wild-type cells.
Although the CUT degradation pathway is well defined, it has become apparent that nuclear RNA degradation in vivo is much more complicated. For example, deletion of the TRAMP component Trf4 or the exosome component Rrp6 stabilizes CUTs as expected, but simultaneous loss of both factors has an additive effect on CUT stability. This shows that the exonuclease Rrp6 must degrade CUTs not processed by Trf4, and the poly(A)polymerase Trf4 must prepare CUTs for degradation by exonucleases other than Rrp6. Trf5 is a close paralogue of Trf4 and can target CUTs, although it does not appear to do so unless Trf4 is absent [23,25], whereas mutations in the Rrp6 co-factors Rrp47 and Mpp6 also stabilize CUTs targeted by Trf4 [8,26]. In conclusion, different parts of the CUT population appear to be processed by either the TRAMP or the Rrp47–Mpp6 complex, but neither complex is sufficient to degrade the whole population.
To complicate matters further, many RNAs classified as CUTs are significantly stabilized in cells lacking cytoplasmic RNA degradation components; presumably this fraction entirely escapes the nuclear RNA surveillance machinery and is thus packaged for export as a normal mRNA . Curiously, although the nuclear exosome is a key player in nuclear RNA surveillance, the cytoplasmic exosome (which has the same core components but different co-factors) does not appear to degrade unstable RNAs that escape nuclear surveillance. Instead, multiple non-coding RNAs including CUTs are degraded by the cytoplasmic 5′→3′ exonuclease Xrn1, following decapping by the Dcp1–Dcp2–Dhh1 complex [12,27–29].
As well as removing a small fraction of CUTs, Xrn1 is the primary exonuclease responsible for the degradation of XUTs (Xrn1-sensitive unstable transcripts). This is another widespread class of unstable ncRNA (1658 were described in budding yeast), whose members are generally larger and better defined than CUTs and that tend to occur antisense to protein-coding genes . One key mechanism for initiating cytoplasmic RNA degradation is NMD (nonsense-mediated decay) (reviewed in ), a mechanism designed to detect and destroy mRNAs carrying premature stop codons. NMD clearly has a role in unstable RNA recognition as many XUTs are stabilized in NMD mutants; however, the stabilization is small, suggesting that there is a major targeting mechanism remaining to be discovered [12,27,28].
Functional unstable RNAs
We can distinguish three classes of functional unstable RNA: those produced as a by-product of a process, those whose transcription promotes a change unrelated to the RNA itself and those which become functional when they cease to be degraded.
By-products of regulation
A number of systems have been described in which a given gene produces either a CUT or a functional mRNA depending on a regulatory process. A good example of this is the NRD1 gene, which contains Nrd1 sites in the 5′-UTR; if present at sufficient concentration, Nrd1 binds these sites, instigating transcriptional termination and production of a CUT, but if Nrd1 levels are insufficient, termination does not occur and a functional mRNA is produced . At the URA2 and IMD2 genes, the TSS is selected based on nucleotide availability, leading to the production of a CUT from an upstream TSS at high nucleotide concentrations or, otherwise, an mRNA from a downstream TSS [32,33]. In all these cases, the CUT itself is apparently unimportant, but is produced as a part of a subtle gene repression system that allows very precise auto-regulation.
Transcriptional interference is a general phenomenon in which transcription of one RNA physically blocks the transcription of another at an overlapping locus. Perhaps the best-studied example in eukaryotes occurs at the SER3 gene in budding yeast, whose promoter is overlapped by the repressive non-coding RNA SRG1 . Although it may seem obvious that passage of an RNA polymerase through a gene promoter might displace transcription factors, this is not the case, at least for SER3. Rather, transcription of SRG1 increases nucleosome density over the SER3 gene in a manner dependent on Paf1, Spt2 and Spt6, blocking access of transcription factors to the SER3 promoter [35–37]. Notably, mutations in these factors derepress SER3 without down-regulating transcription of SRG1, proving that transcription through a promoter is not in itself repressive, but pol II-associated chromatin remodelling is.
Cells have developed a number of mechanisms to prevent undesired transcriptional interference. The ATP-dependent chromatin remodeller Isw2 repositions nucleosomes at the 3′-end of genes to block the expression of cryptic transcripts from these areas , and in the absence of Isw2 transcriptional repression of neighbouring genes is observed . To repress cryptic promoters within genes, the elongating pol II-associated protein Set2 trimethylates H3 K36 that in turn recruits repressive histone deacetylase complexes (reviewed in ). Products of these intragenic cryptic promoters would presumably also be unstable, although this is yet to be directly tested.
Most mysterious of the unstable transcripts are those that can act in trans to repress other genes. The first characterized of these in budding yeast was the PHO84 antisense RNA, which is normally degraded by the nuclear exosome, but is stabilized in ageing cells. Stabilization of the antisense RNA represses the PHO84 gene in an Hda1-dependent manner . Although this would be consistent with the transcriptional interference model presented above, further research demonstrated that the repression can occur in trans , suggesting that the PHO84 antisense RNA can form a bona fide ribonucleoprotein particle with transcriptional repression activity analogous to the HOTAIR ncRNA in humans.
Another group of trans-acting unstable RNAs in budding yeast occur near the 5′-end of sub-telomeric metal ion homoeostasis genes. These RNAs are degraded by NMD, but when stabilized in NMD mutants repress their neighbouring gene . Given that NMD occurs in the cytoplasm, these RNAs must only be functional after a nuclear export/re-import cycle. This is surprising but not unknown for ncRNAs, since well-characterized snRNAs (small interfering RNAs) undergo such a maturation pathway (reviewed in ). Alternative, but more controversial explanations, would be that NMD of these RNAs occurs in the nucleus, or that retrograde communication can occur across a nuclear pore complex.
A final example is the antisense RNA RTL produced from budding yeast Ty element retrotransposons. RTL is normally degraded by the cytoplasmic exonuclease Xrn1, but when stabilized in xrn1 mutants it strongly represses the Ty element RNA and Ty element retrotransposition. This RNA is proven to act in trans, and exerts repression via the H3 K4 methyltransferase enzyme Set1 . The instability of this repressive RNA is, however, surprising; it is hard to imagine that the cell encourages retrotransposon activity under normal circumstances, and raises the more interesting prospect that the retrotransposon is capable of subverting the action of cytoplasmic RNA decay to deactivate a cellular anti-retroviral response. Alternatively, since the antisense RNA is expressed from the retrotransposon locus itself, it may be that repression of transposition under some conditions where Xrn1 is inactivated may be actively beneficial to the transposon.
RNA stability as a gene expression modulator
Almost all the above examples of RNA instability have been garnered from studies of budding yeast undergoing rapid mitotic growth, and the instability of the RNAs detected under these conditions has been extrapolated to a general phenomenon. However, evidence is emerging that cells can initiate widespread changes in gene expression through changes in their RNA degradation systems.
A very neat example of this process is a feedback response that contributes to amino acid homoeostasis in mammalian cells. Many amino acid synthesis genes are targets of the NMD machinery; however, in response to amino acid stress, NMD is down-regulated, allowing a shift of gene expression towards increased amino acid synthesis . Hypoxia also causes NMD repression, and among the NMD-regulated genes are a subset that help cells survive hypoxic conditions; this latter system is a double-edged sword for higher eukaryotes as this cellular survival mechanism helps cancerous cells survive in the hypoxic tumour microenvironment .
It is in the induction of meiosis that the clearest examples of gene expression control by RNA degradation have been found. Fission yeast growing under vegetative conditions repress key meiotic mRNAs by selective recruitment of the TRAMP and exosome-mediated RNA degradation system. The key recruitment protein Mmi1 is then sequestered on entry to meiosis, stabilizing the meiotic mRNAs [47,48]. Although wasteful in terms of transcription, this type of system provides a very robust default response; a major change in growth strategy would normally require a complex and carefully orchestrated change in gene expression pattern, but in this case simply blocking the RNA degradation in response to any cellular response allows an easy transition to meiosis utilizing pre-existing mRNAs.
Curiously, although budding yeast has no apparent equivalent to the unstable meiotic mRNAs observed in fission yeast, it specifically down-regulates its nuclear RNA degradation machinery on entry to meiosis by degradation of the exosome component Rrp6 . This does not cause any major up-regulation of protein-coding genes, but does stabilize a large number of CUTs (~700), which do not appear to have any clear effect. Nonetheless, such a major change in a highly conserved complex is likely to have a purpose, hinting that there may be an important role for stabilized CUTs in meiotic cells. It should be noted that many non-coding RNAs in budding yeast can associate with ribosomes and show evidence of translation, suggesting that some RNAs classified as non-coding may also represent protein-coding genes kept unstable until required [27,50].
Ten years ago, the suggestion that more than 20% of transcriptional units in a genome produce RNA that is instantly degraded would have been met with distinct scepticism. Now it is clear that unstable RNAs are widespread within genomes and across eukaryotes, and that the machinery that degrades them is similarly conserved. A great deal of mechanistic information on the degradation pathways of these RNAs has now been gathered, but functional information still only exists in a handful of cases. Nonetheless, the characterized RNAs play important roles in a wide variety of processes, suggesting that we are only starting to understand the importance of these elusive RNAs.
J.H. was funded by the Wellcome Trust [grant number 088335].
I thank Claudia Schneider and Stephen Frenk for a critical reading of this paper.
RNA UK 2012: An Independent Meeting held at The Burnside Hotel, Bowness-on-Windermere, Cumbria, U.K., 20–22 January 2012. Organized and Edited by Raymond O'Keefe and Mark Ashe (Manchester, U.K.).
Abbreviations: CTD, C-terminal domain; CUT, cryptic unstable transcript; ncRNA, non-coding RNA; NMD, nonsense-mediated decay; ORF, open reading frame; pol, II, polymerase II; PROMPT, promoter-associated pervasive transcript; TRAMP, Trf4–Air–Mtr4 polyadenylation complex; TSS, transcriptional start site; UNT, upstream non-coding transcript; XUT, Xrn1-sensitive unstable transcript
- © The Authors Journal compilation © 2012 Biochemical Society