Biochemical Society Transactions

Translation UK

Aberrant termination triggers nonsense-mediated mRNA decay

N. Amrani, S. Dong, F. He, R. Ganesan, S. Ghosh, S. Kervestin, C. Li, D.A. Mangus, P. Spatrick, A. Jacobson


NMD (nonsense-mediated mRNA decay) is a cellular quality-control mechanism in which an otherwise stable mRNA is destabilized by the presence of a premature termination codon. We have defined the set of endogenous NMD substrates, demonstrated that they are available for NMD at every round of translation, and showed that premature termination and normal termination are not equivalent biochemical events. Premature termination is aberrant, and its NMD-stimulating defects can be reversed by the presence of tethered poly(A)-binding protein (Pab1p) or tethered eRF3 (eukaryotic release factor 3) (Sup35p). Thus NMD appears to be triggered by a ribosome's failure to terminate adjacent to a properly configured 3′-UTR (untranslated region), an event that may promote binding of the UPF/NMD factors to stimulate mRNA decapping.

  • faux untranslated region
  • mRNA decay
  • translation termination
  • yeast

The basics: regulation of translation termination contributes to the accuracy of gene expression

Gene expression is a highly accurate process in which the polypeptide end products contain no more than one incorrect amino acid per 10000 inserted [1,2]. Several mechanisms are responsible for maintaining fidelity in the flow of genetic information. At translation termination, quality control is implemented in at least two important ways. First, the principal release factors discriminate sense codons from nonsense codons before triggering polypeptide hydrolysis [3]. Secondly, a specialized mRNA decay mechanism serves to rid the cell of mRNAs that lack complete open reading frames. Such transcripts are typically those that contain premature termination codons, and the process which ensures that these mRNAs do not accumulate as substrates for the translation apparatus has been dubbed NMD (nonsense-mediated mRNA decay) or mRNA surveillance [46]. While nonsense-containing transcripts are generally derived from genes in which a mutation has given rise to a premature nonsense codon, the fate of an mRNA would be no different if its nonsense-containing status was attributable to errors in transcription, pre-mRNA splicing or RNA editing, or caused by failure of the ribosome to maintain the normal reading frame.

Factors regulating NMD

Yeast and worm mutants that affect NMD have been isolated in screens for extragenic allosuppressors and omnipotent suppressors, regulators of frameshifting or translation, suppressors of upstream initiation codons, or two-hybrid interactors with known factors [720]. Analyses of the genes identified in these studies have demonstrated that mutations in the yeast UPF1, NMD2(UPF2), UPF3, PRT1, HRP1, MOF2, MOF5, MOF8 and DBP2 genes result in selective stabilization and increased accumulation of nonsense-containing mRNAs. In worms, regulators of NMD include SMG1-7, of which SMG2, 3 and 4 are orthologues of UPF1, NMD2 and UPF3. The other SMGs appear to be regulators of the phosphorylation status of the SMG2 protein [21]. Mammalian NMD may also involve numerous protein components of the EJC (exon–junction complex), but the potential role of these factors is complicated by the observation that they are not essential for NMD in Drosophila [2124]. Degradation of yeast nonsense-containing mRNAs also requires factors that play a role in the 5′→3′ mRNA decay pathway, namely the Xrn1p exonuclease and the Dcp1p and Dcp2p components of the decapping enzyme [25].

Endogenous substrates of the NMD pathway

mRNAs comprising endogenous substrates of the NMD pathway have been identified by virtue of their selective stabilization in strains harbouring mutations in one or more of the UPF/NMD genes. These transcripts include: (i) inefficiently spliced pre-mRNAs that enter the cytoplasm with their introns intact [26], (ii) mRNAs in which the ribosome has bypassed the initiator AUG and commenced translation further downstream [18], (iii) some mRNAs containing upstream open reading frames [7,27,28], and (iv) transcripts with extended 3′-UTRs (untranslated regions) [6,29,30]. To ensure that this analysis was comprehensive, we used high-density oligonucleotide DNA arrays to compare the expression profiles of wild-type, upf1Δ, nmd2Δ, upf3Δ, xrn1Δ and dcp1Δ strains [25]. These experiments showed that deletions of UPF1, NMD2 or UPF3 have identical effects, leading to 2–60-fold increases of 746 transcripts (approx. 10% of the transcriptome). Deletions of the XRN1 or DCP1 genes affected the expression of 16–19% of the genes in the yeast genome and the expression profile of the three upf/nmd mutant strains overlaps, yet is distinct from those of the xrn1Δ and dcp1Δ strains. The set of NMD substrates identified included previously characterized substrate mRNAs, as well as: (i) bicistronic transcripts, (ii) transcripts from pseudogenes, (iii) transcripts known to use frameshifting in their translation, and (iv) transcripts encoded by transposable elements or their long terminal repeats.

Mechanistic models: pioneer round of translation versus faux UTR

NMD is triggered by premature translation termination, but the features distinguishing that event from normal termination are unclear. Two predominant models for the mechanism of NMD have been put forth. The first [21,31], a direct descendant of the yeast ‘surveillance complex’ model [32,33], suggests that the deposition of the EJC during early processing events allows decay-inducing factors to maintain an mRNA association unless swept off by the ribosome in the first, or pioneer, round of translation. (In yeast, Hrp1p is the EJC equivalent). In the event that termination is premature, the EJC complex (or Hrp1p) is thought to remain on the mRNA and to trigger mRNA decay by subsequent interaction with the UPF1 protein. The manner in which Upf1p triggers decay is uncertain, but Upf1p–Dcp2p interactions that we and others have observed [14,34] have led to speculation that Upf1p may recruit the decapping complex [3436].

The second model posits that premature termination and normal termination are biochemically different events and that, at least in yeast, NMD is triggered by the events accompanying aberrant termination at a premature nonsense codon [3739]. In this faux UTR model [37,40,41], proper termination of translation and normal rates of mRNA decay require interactions between a terminating ribosome and a specific RNP (ribonucleoprotein) structure or set of factors localized 3′ to the stop codon. The region downstream of the premature terminator (the ‘downstream element’ or DSE) is thought to promote aberrant termination because it lacks a regulatory factor (or factors) that are normally present on a legitimate 3′-UTR. Aberrant termination, in turn, is thought to allow binding of the UPF/NMD factors and trigger mRNA decay.

Substrates of the NMD pathway are not limited to newly synthesized mRNAs, and these mRNAs cannot acquire immunity to NMD

We utilized an inducible system coupled with microarray analysis to evaluate the effect of reactivation of NMD on global mRNA accumulation in NMD-deficient cells. In this inducible NMD system, the expression of NMD2 is under the control of the GAL1 promoter [40]. Yeast cells lacking endogenous Nmd2p, but harbouring the GAL1-NMD2 allele, lack NMD activity when grown in raffinose-containing medium, but are fully functional for NMD 20 min after the addition of galactose to their growth medium. During a 60-min time course, galactose-induced expression of NMD2 caused the down-regulation (>1.8-fold; P<0.05) of >400 transcripts (F. He, P. Spatrick, C. Li, S. Dong and A. Jacobson, unpublished work). This set of transcripts overlaps with the set of mRNAs that is up-regulated when NMD is inactivated (i.e. in upf/nmd strains) and encompasses all of the previously characterized classes of NMD substrates. These data demonstrate that most transcripts that are down-regulated upon reactivation of NMD are direct targets of the NMD pathway. We also found a significant number of transcripts that are up-regulated in upf/nmd strains that did not show significant decreases in their levels upon NMD reactivation (F. He, P. Spatrick, C. Li, S. Dong and A. Jacobson, unpublished work). Such mRNAs define the class of transcripts controlled indirectly by NMD.

These experiments not only identify the bona fide endogenous substrates of NMD, but also demonstrate that these mRNAs cannot escape degradation by NMD. Substrates of the NMD pathway must be available for NMD at each round of translation and are thus not limited to newly synthesized or newly exported mRNAs, i.e. those involved in a pioneer round of translation. These observations reinforce the notion of NMD as a cytoplasmic pathway and call into question models which suggest that NMD is triggered by the failure of elongating ribosomes to dissociate decay-inducing protein complexes deposited on the mRNA during early processing events. Recent experiments in other systems also challenge the role of such complexes [22,4345].

In vitro termination at premature nonsense codons is aberrant

We established methods for toeprint analysis of translation termination codons and used these methods to determine whether premature termination yielded the same toeprint as normal termination. Wild-type extracts were incubated with synthetic nonsense-containing mRNAs, and the translation reactions were subjected to toeprinting analyses, using sensitivity to m7GpppG, a cap analogue, as a means to distinguish legitimate toeprints from background bands. Toeprints corresponding to ribosomes stalled with a stop codon in their A sites were obtained at the expected position, 12 nt downstream of premature nonsense codons [41]. Similar experiments failed to detect any toeprint signals from normal termination codons, suggesting that ribosomes terminating prematurely are released much less efficiently than those encountering normal terminators. Addition of the elongation inhibitor CHX (cycloheximide) to the translation reactions also failed to reveal toeprints from normal terminators, but did allow detection of additional toeprints in close proximity to the locations of the early stop codons. These unanticipated bands were attributable to post-termination ribosomes that failed to be released at premature terminators and were able to scan backwards and re-initiate. Regardless of the site of post-termination re-initiation, all nonsense-containing RNAs yielded toeprints corresponding to ribosomes stalled with the stop codon in their A sites when translated in sup45-2 extracts, suggesting that, before any re-initiation event, a premature stop codon must be recognized by eRF1 (eukaryotic release factor 1) (Sup45p) and peptide hydrolysis must be triggered [46,47]. Most importantly, the aberrant toeprints were linked to NMD because they failed to accumulate in extracts lacking Upf1p. The lack of aberrant toeprints in upf1Δ extracts cannot be due to inefficient translation since these extracts display premature terminator +12 nt toeprints in the absence of CHX and normal initiator AUG toeprints in the presence of CHX. These results suggest that inactivation of Upf1p may preclude or perturb events that regulate ribosome dwell time and movement at a premature stop codon.

Aberrant termination is eliminated by flanking a premature nonsense codon with a normal 3′-UTR

To determine whether aberrant termination and its consequences resulted from the relative positioning of a nonsense codon within an mRNA, four constructs were made in which the PGK1 3′-UTR [4,48] was inserted immediately downstream of early stop codons in can1 alleles, thereby creating ‘mini’ wild-type mRNA mimics. Analysis of these RNAs showed that the aberrant cap analogue-sensitive toeprints characteristic of premature termination were absent in the ‘mini’ RNAs in wild-type extracts [41]. The lack of detectable aberrant toeprints with these mimics of wild-type mRNA indicates that backwards scanning, or at least its stable end-product, is eliminated when termination codons are flanked by normal 3′-UTR sequences. These results led us to conclude that such ‘retro-re-initiation’ is a consequence of an aberrant termination event and raise the question of why a 3′-UTR created by a premature termination codon differs from that of a wild-type mRNA.

Tethered poly(A)-binding protein (Pab1p) mimics a normal 3′-UTR, stabilizing nonsense-containing mRNAs

The faux UTR model suggests that the DSE, thought to be a key cis-acting regulator of NMD [4], promotes mRNA decay because it lacks a termination regulatory factor (or factors) that is normally present on a legitimate 3′-UTR, a hypothesis that may also explain why deletions that eliminate most coding sequences downstream of premature terminators stabilize mRNAs that would otherwise be substrates for NMD [4,48]. This inadequacy could occur because translation to the normal end of a coding region remodels an mRNP (messenger RNP) [49] or because proximity to the polyadenylated tail (and bound Pab1p) has a qualitative and/or quantitative influence on the nature of proteins bound to the 3′-UTR [37]. To test whether proximity of a termination codon to Pab1p is germane to NMD, the in vivo stability of mRNAs bearing an MS2 coat protein binding-site 3′ to premature terminators in the CAN1 and PGK1 mRNAs was assessed in cells that expressed an MS2–Pab1p fusion [41]. Pab1p tethered 37–73 nt 3′ to premature UAA, UGA or UAG codons promoted 5–10-fold increases in mRNA stability and abundance. MS2–Sxl or MS2 dimer tethered at the same position, or MS2-Pab1p tethered 164 nt downstream (3′ to the DSE), had no effect on mRNA stability or abundance. The mRNA-stabilizing effects could not be attributed to selective interference with translation because the same number of ribosomes associated with the mRNA in the presence or absence of tethered Pab1p [41].

Tethered Pab1p recruits Sup35p

Partial stabilization of nonsense-containing transcripts was engendered by tethered fragments of Pab1p [41], suggesting that Pab1p-mediated mRNA stabilization might be attributable to protein–protein interactions characteristic of its respective domains [50]. To test this possibility, we utilized anti-MS2 coat protein antibodies to immunoprecipitate mRNPs and assayed for the presence of co-immunoprecipitated regulatory factors. Western blotting demonstrated that selective immunoprecipitation of these mRNPs from extracts containing MS2–Pab1p led to the recovery of MS2–Pab1p and Sup35p, but did not lead to the recovery of the initiation factors eIF4G (eukaryotic initiation factor 4G) or eIF4E [41]. Co-immunoprecipitation of Sup35p was attributable to a specific interaction with MS2–Pab1p because no comparable recovery of Sup35p was obtained from immunoprecipitates of MS2–Pab1p lacking the Sup35p-interaction domain or from immunoprecipitates of PGK1 nonsense-containing mRNPs stabilized by deletion of UPF1 in cells that express the control protein, MS2–Sxl [41]. Tethered Sup35p also stabilized PGK1 nonsense transcripts, albeit to a lesser extent than tethered Pab1p. Similar experiments, in which Sup45p was the tethered component, failed to stabilize the same mRNA significantly, indicating that regulatory aspects of termination played a key role in antagonizing NMD and that Pab1p's role in this process most likely reflected its function as a scaffold for post-transcriptional regulators [50].


Collectively, our results are consistent with the basic tenet of the faux UTR model, namely that not all termination events are equivalent, and that, at least in yeast, NMD is triggered by a ribosome's failure to terminate adjacent to a properly configured 3′-UTR. Proper termination of translation and normal rates of mRNA decay are likely to require interactions between a terminating ribosome and a specific RNP structure or set of factors localized 3′ to the stop codon. Failure to terminate in this manner appears to slow ribosome release and facilitate re-initiation. The latter event must only be a symptom of aberrant termination, not a cause of NMD, because elimination of re-initiation site AUGs does not alter the decay rate of nonsense-containing mRNAs.


Grant support to A.J. from the U.S. National Institutes of Health (GM27757 and HD48137) is gratefully acknowledged.


  • Translation UK: Focused Meeting and Satellite to BioScience2005, held at Western Infirmary, Glasgow, U.K., 21–23 July 2005. Organized and Edited by M. Bushell (Nottingham, U.K.), S. Newbury (Newcastle upon Tyne, U.K.), G. Pavitt (Manchester, U.K.) and A. Willis (Nottingham, U.K.).

Abbreviations: CHX, cycloheximide; DSE, downstream element; eIF, eukaryotic initiation factor; EJC, exon–junction complex; NMD, nonsense-mediated mRNA decay; (m)RNP, (messenger) ribonucleoprotein; UTR, untranslated region


View Abstract