Although functional RNA is generally protected against degradation, defects or irregularity during RNA biogenesis lead to rapid degradation. Cellular surveillance mechanisms therefore need to distinguish aberrant, erroneous, damaged or aging transcripts from normal RNAs in order to maintain fidelity and control of gene expression. The detection of defects seems to be primarily based on functionality or aberrant rates of a given step in RNA biogenesis, allowing efficient detection of many different errors without recognition of their specific nature. We propose that the addition of non-templated nucleotides to the 3′ end of mRNAs and small non-coding RNAs, 3′ tagging, is the primary means by which malfunctioning RNAs are labelled, promoting their functional repression and degradation. However, the addition of non-templated nucleotides to transcripts can have diverse effects which vary with location, length, substrate and sequence.
Gene expression is a multi-step process that brings the genome to life. There is a need for tight regulation, determining the level and defining both the temporal and spatial expression of each gene. Additionally, the fidelity of the process has to be policed. RNA is a key element in the flow of genetic information from DNA to protein, functioning as mRNA and a wide range of non-coding transcripts including tRNA, rRNA, snoRNA (small nucleolar RNA), miRNA (microRNA) and siRNA (small interfering RNA). RNA biogenesis involves a large number of steps and a series of transient associations with different RNPs (ribonucleoproteins, i.e. RNA–protein complexes) to transform newly synthesized transcripts into mature functional RNA. Considering the complexity of RNA formation, maturation and transport and the dynamic nature of these processes, it is inevitable that a large number of different types of error will occur. Additionally, continuous adaptation to environmental changes requires regulated RNA turnover, which must be precisely directed to appropriate transcripts. Thus it is crucial for cells to identify defective or unwanted RNA in order to maintain both the fidelity and appropriate control of gene expression.
An emerging theme is the addition of a short sequence of non-templated nucleotides to the 3′ end of transcripts, 3′ RNA tagging, which often precedes degradation as a means of marking defective, damaged or aging RNA for degradation. RNA tagging is conducted by ribonucleotidyl transferases, which are members of the superfamily of DNA polymerase β-like nucleotidyl transferases [1,2]. In addition to canonical PAPs [poly(A) polymerases], these include non-canonical PAPs that add oligo(A), and PUPs [poly(U) polymerases] or TUTases (terminal uridylate transferases) that add short U-rich tails  We aim to discuss briefly what is currently known about the role of these enzymes, multiple examples of which are found in most eukaryotes.
3′ end adenylation: a multi-functional platform where size matters
The best known example of non-templated nucleotides being added to the 3′ ends of transcripts is adenylation. In the nucleus, the addition of a short oligo(A) (~A5) is associated with degradation [4,5], as it is in prokaryotes , whereas the long poly(A) tail (>A15) facilitates stabilization and export of mRNAs to the cytoplasm . In bacteria, it is well established that oligo-adenylation mediates degradation of both mRNA and ncRNAs (non-coding RNAs) [8,9]. The bacterial PAP1 marks mRNAs for degradation if a given processing step is retarded due to misfolding, misprocessing or improper RNP formation [10,11]. Initial observations of 3′ adenylation of eukaryotic non-coding nuclear RNAs (rRNA, snRNA and snoRNA) were obtained in strains lacking at least one processing enzyme [e.g. RRP (ribosomal RNA-processing protein) 6] and, owing to the transient nature of these tagged transcripts, they are rarely detected in wild-type strains [12–15]. Nuclear oligo-adenylation therefore seems to mirror the bacterial paradigm. It appears that 3′ end tagging of RNA with A-rich tracts is a conserved way to facilitate RNA processivity and in particular the decay of highly structured regions, in both prokaryotes and eukaryotes. This is consistent with the accumulation of degradation intermediates observed if RNA is incubated with the exosome in vitro .
The consensus is that the majority of newly transcribed RNA is degraded  in response to surveillance mechanisms that monitor functionality or progression through programmed steps in biogenesis [4,17,18]. In the nucleus, RNA degradation is largely dependent on TRAMP, a conserved nuclear polyadenylation complex that oligo-adenylates RNA. Unlike polyadenylation, which acts specifically on RNA polymerase II transcripts, oligo(A) tagging is not influenced by the polymerase involved. Short A-tags can promote 5′→3′ degradation by the Rat1/Xrn2 exonuclease and/or 3′→5′ degradation mediated by either the exosome or independently by one of the exosome components, Rrp6. The exosome is a conserved multi-subunit protein complex associated with two active nucleases: Rrp44/Dis3, a bifunctional 3′→5′ exoribonuclease with an additional endonuclease activity, and Rrp6, a 3′→5′ exonuclease . Once recruited, the exosome or Rrp6 can either trim the 3′ end as part of RNA maturation or degrade defective RNA [20–22]. In Saccharomyces cerevisiae, TRAMP's non-canonical PAP (Trf4 or Trf5) is activated by an RNA-binding protein, Air1 or Air2, but its processivity is strictly limited by the RNA helicase Mtr4p . Mtr4p controls the length of A-tags by inherently suppressing poly(A) extension after only three to five adenosines . The short length prevents binding by PABP [poly(A)-binding protein], thus avoiding the possibility of the tag functioning as a poly(A) tail, which could promote stabilization and export to the cytoplasm.
In S. cerevisiae, most full-length CUTs (cryptic unstable transcripts) possess poly(A) tails, predicted to be approximately 70–90 nt long, and appear to be degraded in the cytoplasm . However, some CUTs are subject to TRAMP-exosome-mediated adenylation and degradation in the nucleus, which is Rrp6 dependent . Similarly, depletion of the exosome in human cells leads to the accumulation of short, oligo-adenylated ‘promoter upstream’ transcripts, which are normally very unstable . Nuclear oligo-adenylation can also be used to differentially regulate specific transcripts. The human non-canonical PAP, dubbed Star-PAP, specifically adenylates pre-mRNAs involved in the oxidative stress response, causing oxidative-stress-dependent cleavage and activation of specific transcripts .
Defects in tRNA that disrupt the structure and function lead to rapid adenylation and degradation [28–31]. For example, tRNAiMet lacking a methyl group at A58 is adenylated and degraded by the exosome. In vitro, the natural methylated tRNAiMet is resistant to adenylation, whereas the fully unmodified synthetic tRNAiMet is adenylated by TRAMP. This suggests that a loss of tRNA tertiary structure promotes tagging [28,29]. Furthermore, a tRNAAla with point mutations that disrupt its tertiary structure is efficiently adenylated by TRAMP, unlike its wild-type counterpart . Thus defects in tRNA structure and not individual errors trigger tagging and degradation. This is consistent with the finding that tRNAs with unstable acceptor stems are subject to rapid decay mediated by 3′ end CCACCA extension, which is catalysed by a non-canonical PAP, ATP/CTP:tRNA nucleotidyl transferase . Strengthening the tRNA secondary structure increases stability of tRNAs  and almost entirely abolished CCACCA addition . It is important to note that active tRNA requires the non-templated addition of a single CCA motif, which is mediated by the same enzyme.
Although the addition of a short poly(A) tail to aberrant ncRNAs in the nucleus by TRAMP promotes 3′→5′ degradation, cytoplasmic adenylation of ncRNAs may have the opposite function. In Populus trichocarpa, the 3′ end of many miRNAs are oligo-adenylated, preventing miRNA degradation in vitro . Similarly, 3′ adenylation of mammalian miR-122 is also implicated in its stabilization . Importantly, global analysis of miRNAs associated with the Ago (Argonaute) proteins, which direct the silencing effects of miRNAs, revealed that adenylation of small RNAs inhibited the loading into the Ago complex, thus reducing silencing .
In organelles, RNA adenylation can either stabilize, as in human mitochondria [36,37], or trigger, as in chloroplasts and plant mitochondria , RNA decay. Importantly, many mitochondrial mRNAs do not possess a stop codon, but terminate with uridine. Thus the addition of oligo(A) generates the termination codon UAA , and therefore contributes to regulation at the level of translation.
3′ end uridylation of ncRNAs
As with adenylation, uridine tagging can affect ncRNAs in three known ways: stabilization, destabilization and attenuation of function. In animals and plants, both miRNA and siRNA, as well as the 5′ mRNA fragments resulting from miRNA-directed degradation, are subject to uridylation . Addition of four uridines to the 3′ end of human U6-snRNA results in its stabilization and is essential for splicing . Dicing of a conserved miRNA, pre-let-7, in mammals and Caenorhabditis elegans, is inhibited by 3′ end oligo-uridylation [42,43]. In C. elegans, some endo-siRNAs are uridylated, which represses their activity . Uridylation of mature miRNAs and siRNAs in the alga Chlamydomonas triggers their degradation by Rrp6 , and it has been proposed that uridylation and Rrp6-dependent degradation of mature ncRNAs is a part of a surveillance mechanism for aberrant, dysfunctional and unmethylated small RNAs in Arabidopsis thaliana . Consistent with this notion, in zebrafish, unmethylated piRNAs (piwi-interacting RNAs) are uridylated and/or adenylated, becoming very unstable . In Drosophila, siRNAs are protected by 2′-O-methylation from uridylation and adenylation, non-methylated siRNAs being tagged, which in turn promotes 3′→5′ degradation . The addition of an oligo(U) tail to the 3′ end of miRNAs also appears to operate as a mechanism to control function [49,50]. Uridylation of miR-26a has no effect on stability, but reduces its ability to inhibit its mRNA target .
Pyrimidine tagging of mRNA: spelling the end?
For mRNA, the correct formation of both the 5′ cap and the 3′ poly(A) tail, in addition to being critical for functionality, represent key processing checkpoints. Until a few years ago, it was widely accepted that co-transcriptional addition of the 5′-cap and 3′ poly(A) tail were the only terminal modifications of mRNA . Addition of the 5′-cap to pre-mRNA promotes elongation of transcription, splicing and 3′ end formation . The protein complex thus formed protects mRNAs against degradation, promotes export and enhances translation and is therefore essential for gene expression and regulation. Failure to add a cap to the 5′ end activates pre-mRNA degradation by Rat1/Xrn2 and this is possibly triggered by the reduced rate of transcription/processing . PABP protects the long 3′ poly(A) tail of mRNA from 3′→5′ degradation, directly repressing the major deadenylases (PARN and Ccr4-Not) . PABP is also directly involved in assembling translationally competent mRNP by bringing both the 5′ and 3′ ends together . This closed-loop mRNA structure both protects the mRNA 5′ from degradation and enhances translation. Consequently, degradation of the vast majority of mRNA is initiated by shortening the poly(A) tail [52,56]. At a threshold length, ~A15 in fungi and A15–25 in mammals [57–59], PABP dissociates, resulting in decapping and rapid mRNA degradation .
An exception to this model are the metazoan histone mRNAs, which are capped, but not adenylated; their 3′ ends are stabilized by a stem–loop structure . The synchronized degradation of hH-mRNAs (human histone mRNAs) at the end of the S-phase involves Upf1, a conserved NMD (nonsense-mediated decay) factor . It has been established that upon inhibition of DNA replication the 3′ end of hH-mRNA is tagged with a short oligo(U) tract, which triggers its degradation from both ends . Subsequently, it was discovered that polyadenylated mRNAs in both Aspergillus nidulans  and Schizosaccharomyces pombe  are also subject to pyrimidine tagging in the cytoplasm. These direct observations are supported by EST (expressed sequence tag) sequence data and general RNA analysis, suggesting that pyrimidine tagging may be a general phenomena in metazoans . Recent data from our group has also identified U/C tagging of at least one plant transcript (I.Y. Morozov and M.X. Caddick, unpublished work). However, mammalian mRNAs subject to inactivation by miRNA are not similarly tagged , which may indicate that tagging is specific to a subset of RNA degradation mechanisms. It will therefore be important to determine the extent and distribution of cytoplasmic pyrimidine tagging in eukaryotes.
In S. pombe, one of its seven non-canonical nucleotidyl transferases, Cid1, has been implicated in uridylation , and in A. nidulans, a related nucleotidyl transferase, CutA, is critical for addition of a CU-based sequence . For hH-mRNA uridylation, three different enzymes have been implicated [62,65]. In all these cases, disruption of a single nucleotidyl transferase does not completely abolish tagging. This implies that more than one nucleotidyl transferase may be involved in the modification of specific transcripts, suggesting additional levels of complexity and regulation.
The reported absence of P-bodies (processing bodies) in ΔcutA strains  is consistent with the CU modification of transcripts being associated with subcellular compartmentalization, tagging mRNAs for decapping and degradation. Although the specific role of P-bodies is not well defined, they represent dynamic RNA–protein complexes that include translationally silenced mRNA and a range of proteins involved in mRNA turnover . It is therefore possible that 3′ pyrimidine tagging promotes transport away from ribosomes as well as enhancing degradation. However, it is not known whether the P-body formation is associated with nucleotidyl transferase activity or not.
As in the case of hH-mRNA, pyrimidine tagging in A. nidulans and S. pombe occurs at the point of mRNA decapping, and disruption of tagging reduces the rate of transcript decay, although the effect on degradation rates varies widely between transcripts. In A. nidulans, 3′ end pyrimidine tagging is found predominantly in decapped transcripts with a poly(A) tail shortened to ~15 nt, the threshold length at which decapping is triggered . However, in S. pombe, the poly(A) tail length of uridylated transcripts is not as well defined. Both pyrimidine tagging and decapping occur independently of deadenylation (>A20) when either the Ccr4–Caf1–Not complex, the major deadenylation complex in eukaryotes, or the catalytic domain of Ccr4 is disrupted [52,63]. Additionally Δccr4 leads to a higher proportion of tagged transcripts in both S. pombe  and A. nidulans . One possibility is that the Ccr4–Caf1–Not complex, by associating with the poly(A) tail and progressively deadenylating the mRNA in a controlled manner, directly inhibits 3′ tagging, preventing premature decapping and degradation. In strains deleted for cutA, the distribution of poly(A) tail length in decapped and total mRNA did not vary significantly, unlike the wild-type. These data imply that loss of pyrimidine tagging leads to both a delay in decapping, which normally occurs efficiently when the transcript has been deadenylated to A15, and a reduction in the pace of the final stages of deadenylation. This would be consistent with the 3′ pyrimidine tag promoting both the recruitment of the Lsm–Pat1 complex, which enhances decapping, and the exosome, which may be responsible for the final stages of deadenylation. In particular the Lsm complex has been shown to have greater affinity for uridylated RNA [68,69], and S. pombe strains deleted for Lsm1 accumulate transcripts with short poly(A) tails .
The involvement of UPF1 in hHmRNA tagging and degradation is particularly intriguing. UPF1 is associated with NMD, which is triggered by premature termination of translation; the termination complex formed at a distance from the poly(A) tail and associated PABP results in the recruitment of UPF1 and a range of other factors that promote RNA degradation . Data from our group also found that 3′ tagging of mRNA in A. nidulans is significantly reduced in Δupf1 strains (I.Y. Morozov and M.X. Caddick, unpublished work). This opens the possibility that 3′ tagging of mRNA is being signalled by the ribosome termination complex, which would provide a surveillance mechanism for mRNA functionality.
RNA 3′ tagging is a common phenomenon that extends from prokaryotes to eukaryotes. As discussed above, the consequences are diverse, but as yet we do not have a clear overview of either the extent of tagging or its consequences. Currently, technical difficulties have prevented the use of high-throughput sequencing to address these questions for mRNA, a key problem being the presence of the long poly(A) tail. With technological advances it will hopefully be possible to assess fully the diversity of tags, their distribution across the transcriptome and their regulation. Only then will we be in a position to fully weigh up their functional significance.
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.).
cryptic unstable transcript
human histone mRNA
ribosomal RNA-processing protein
small interfering RNA
small nucleolar RNA
We thank Meriel G. Jones for critical discussions.
Work in our laboratory is supported by the Biotechnology and Biological Sciences Research Council.