Yeast cells containing a temperature‐sensitive mutation in the PRT1 gene were found to selectively stabilize mRNAs harboring early nonsense codons. The similarities between the mRNA decay phenotypes of prt1‐1 cells and those lacking the nonsense‐mediated mRNA decay (NMD) factor Upf1p led us to determine whether both types of mutations cause the accumulation of the same mRNAs. Differential display analysis and mRNA half‐life measurements demonstrated that the HHF2 mRNA increased in abundance in prt1‐1 and upf1Δ cells, but did not manifest a change in decay rate. In both mutant strains this increase was attributable to stabilization of the SPT10 transcript, an mRNA encoding a transcriptional regulator of HHF2. Analyses of chimeric mRNAs used to identify the cis‐acting basis for NMD of the SPT10 mRNA indicated that ribosomes scan beyond its initiator AUG and initiate at the next downstream AUG, resulting in premature translation termination. By searching a yeast database for transcripts with sequence features similar to those of the SPT10 mRNA, other transcripts that decay by the NMD pathway were identified. Our results demonstrate that mRNAs undergoing leaky scanning are a new class of endogenous NMD substrate, and suggest the existence of a novel cellular regulatory circuit.
Cells contain several mechanisms aimed at maintaining a high level of accuracy in the flow of genetic information, including proofreading systems that ensure the integrity of RNA synthesis and processing (Chin and Pyle, 1995; Jeon and Agarwal, 1996), tRNA aminoacylation (Freist et al., 1996) and peptide elongation (Yarus, 1992). However, these mechanisms are not flawless and cannot compensate for errors arising from mutations within an organism's genome. As a consequence, some mRNAs lack complete open reading frames (ORFs). Another surveillance process, nonsense‐mediated mRNA decay (NMD), ensures that such transcripts do not accumulate. Shown to operate in a broad spectrum of eukaryotic organisms, this pathway triggers the accelerated decay of mRNAs containing premature in‐frame termination codons (Losson and Lacroute, 1979; Leeds et al., 1991; Peltz et al., 1993, 1994; Pulak and Anderson, 1993; Whitfield et al., 1994; Maquat, 1995; Caponigro and Parker, 1996; Jacobson and Peltz, 1996; Ruiz‐Echevarria et al., 1996).
The existence of a cellular pathway capable of assessing the translational capacity of individual transcripts raises the question of whether the substrates of this pathway are restricted to ‘aberrant’ mRNAs. In addition to mRNAs with premature termination codons, substrates of this pathway include intron‐containing RNAs that have escaped from the nucleus to the cytoplasm, mRNAs with extended 3′‐UTRs, and a subset of transcripts harboring upstream open reading frames (uORFs) (Pinto et al., 1992; He et al., 1993; Pulak and Anderson, 1993; Peltz et al., 1994; Cui et al., 1995; Oliveira and McCarthy, 1995; Muhlrad and Parker, 1999; F.Sherman, personal communication). In yeast, the factors and sequences that promote NMD have been characterized extensively. Rapid degradation of nonsense‐containing yeast mRNAs requires the trans‐acting factors encoded by the UPF1, NMD2/UPF2 and UPF3 genes (Leeds et al., 1991, 1992; Peltz et al., 1993; Cui et al., 1995; He and Jacobson, 1995; Lee and Culbertson 1995; He et al., 1997), as well as a cis‐acting sequence 3′ to the premature termination codon, termed a downstream element (DSE; Peltz et al., 1993, 1994; Zhang et al., 1995; Ruiz‐Echevarria et al., 1996).
The phenomenon of NMD also illustrates how the pathways of mRNA decay and translation can be interrelated. Other examples of such interrelationships include experiments demonstrating that: (i) inhibition of translational elongation, by drugs or mutations, can block decapping and stabilize most mRNAs (Peltz et al., 1992; Beelman and Parker, 1994); (ii) cis‐acting sequences that modulate mRNA decay can be located within protein coding regions and require ongoing translation elongation for their function (Wisdom and Lee, 1991; Heaton et al., 1992; Herrick and Jacobson, 1992; Caponigro et al., 1993; Hennigan and Jacobson, 1996); (iii) shortening or removal of the poly(A) tail and 5′ cap, mRNA appendages with specific roles in translational initiation, comprise rate‐limiting steps in the decay of several mRNAs (Munroe and Jacobson, 1990; Muhlrad and Parker, 1994; Tarun and Sachs, 1995; Jacobson, 1996); and (iv) factors involved in mRNA decay are polysome‐associated (Atkin et al., 1995, 1997; Mangus and Jacobson, 1999).
The connections between mRNA decay and translation led us to identify PRT1, the gene encoding the p90 subunit of initiation factor eIF3 (Naranda et al., 1994; Danaie et al., 1995), as an additional regulator of NMD. We have utilized cells harboring an allele of this gene (prt1‐1), or an allele of UPF1 (upf1Δ), to characterize the spectrum of endogenous substrates of this pathway. Using differential display analysis, mRNAs whose abundance increased in the mutants, but not in the parental wild‐type strains, were identified. Further analysis indicated that the increased levels of these mRNAs were attributable to either their stabilization or to increases in their transcription. The HHF2 mRNA was in the latter category and the basis for its increased transcription was investigated. We found that the half‐life of the SPT10 mRNA, which encodes a transcriptional regulator of the HHF2 gene (Dollard et al., 1994), is increased in both the prt1‐1 and upf1Δ strains. Further characterization of the SPT10 transcript indicated that this mRNA becomes a substrate for the NMD pathway as a consequence of premature translational termination resulting from leaky scanning of the ribosome past the translation initiation codon and subsequent initiation at a downstream, out‐of‐frame AUG. This atypical mode of translation and turnover may well define a novel cellular regulatory circuit.
Nonsense‐containing transcripts are stabilized in prt1‐1 cells
The importance of specific translation events in the decay of individual mRNAs has been explored in experiments analyzing the consequences of their inhibition (see Introduction). To address the role of translation initiation in mRNA decay, we measured mRNA half‐lives in a set of mutant strains that block translation initiation at various steps and to varying degrees. Our initial experiments analyzed the consequences of a temperature‐sensitive mutation in PRT1. The prt1‐1 allele is attributable to a single amino acid substitution (S518F) that confers rapid inhibition of growth and translation at 37°C, as well as a defect in the interaction of the ternary complex (eIF2‐GTP‐tRNAiMet) with the 40S ribosomal subunit in vitro (Hartwell and McLaughlin, 1969; Feinberg et al., 1982; Keierleber et al., 1986; Evans et al., 1995). After cells were incubated at 37°C for 15 min, protein synthesis decreased rapidly by 95–97% in the prt1‐1 mutant, while it increased moderately in wild‐type cells over the same time course (Figure 1A). The initiation defect associated with the prt1‐1 mutation was confirmed by analyzing its effects on cellular polysome profiles. Post‐mitochondrial extracts from wild‐type cells exhibited a normal distribution of ribosomes, whereas prt1‐1 extracts manifested a drastic reduction in polysomes and an ∼2‐fold increase in monosomes (Figure 1B).
To determine whether prt1 mutations might also affect mRNA decay rates, we measured the half‐lives of two classes of cellular mRNAs after shifting cultures of mutant or wild‐type cells to non‐permissive growth conditions. Cells were incubated at 37°C for 0 or 15 min, transcription was then inhibited by addition of the drug thiolutin (Jimenez et al., 1973; Tipper, 1973; Herrick et al., 1990), and aliquots of the respective cells were removed at various times after transcriptional inhibition for RNA isolation and Northern blot analysis. The blots were probed with DNAs complementary to mRNAs that were either constitutively unstable (inherently unstable mRNAs) or unstable as a consequence of premature translational termination.
Consistent with earlier studies showing that inhibition of translation initiation in cis does not lead to mRNA stabilization (Vega Laso et al., 1993; Beelman and Parker, 1994; Sagliocco et al., 1994), there was no significant effect of the prt1‐1 mutation on the decay rates of inherently unstable mRNAs. After a 15 min shift to 37°C, half‐lives of the STE2, HIS3 and URA5 mRNAs were comparable in prt1‐1 and wild‐type cells (Figure 1C). The lack of effect applied to other mRNAs having a broad range of decay rates and was not dependent on the time of incubation at 37°C (Table I).
In contrast to the results with ‘normal’ mRNAs, the data of Figure 1D demonstrate that the prt1‐1 mutation engenders a stabilization of nonsense‐containing mRNAs. This effect was evident with two different mRNAs: the CYH2 pre‐mRNA, a naturally occurring substrate (He et al., 1993), and the PGK1‐UAG‐2 mRNA, the product of a PGK1 allele harboring a nonsense mutation at nucleotide 361 (Peltz et al., 1993). In prt1‐1 cells, the former RNA was stabilized ∼6‐fold (wild type, t1/2 = <1 min; prt1‐1, t1/2 = 6 min) and the latter RNA was stabilized ∼3‐fold (wild type, t1/2 = 3 min; prt1‐1, t1/2 = 9 min; see Figure 1D and Table I).
To test whether the stabilizing effects of the prt1‐1 mutation on nonsense‐containing mRNAs were limited to legitimate substrates, we measured the half‐life of the PGK1‐UAG‐6 mRNA, a transcript whose ‘late’ nonsense mutation at nucleotide 1449 fails to render it a substrate for NMD (Peltz et al., 1993). Experiments similar to those of Figure 1D showed that the half‐life of this mRNA was indistinguishable in the mutant and wild‐type cells (t1/2 = >35 min; Table I).
Similar analyses were conducted with cells harboring other alleles of PRT1 (prt1‐26 and prt1‐63), as well as mutations in the genes encoding the p16 subunit of eIF3 (sui1‐1), the α subunit of eIF2 (sui2‐1), the eIF2 kinase (GCN2c) and the cap‐binding protein eIF4E (cdc33–42) (Altmann et al., 1989; Castilho‐Valavicus et al., 1990; Wek et al., 1990; Evans et al., 1995). Although most of these mutations have substantial inhibitory effects on translation initiation (e.g. see Figure 1A; Hanic‐Joyce et al., 1987; Altmann et al., 1989; Linz et al., 1997; and data not shown), none except sui1‐1 had any significant effect on the decay rates of inherently unstable or nonsense‐containing mRNAs. Cells harboring the sui1‐1 mutation showed only a very modest stabilization of the CYH2 pre‐mRNA and mRNA (data not shown).
Utilization of differential display analysis to identify mRNAs that accumulate in prt1‐1 and upf1Δ strains
Since mutations in UPF1, NMD2/UPF2 or UPF3 also cause selective stabilization of nonsense‐containing mRNAs, we considered the possibility that the PRT1 gene product was another component of this set of post‐transcriptional regulators. To assess this possibility, we used mRNA differential display analysis (Liang and Pardee, 1995) to determine whether cells harboring prt1‐1 or upf1Δ mutations promoted stabilization of the same set of mRNAs. These experiments showed that the HHF2, RPL13a, LYS14 and ACO1 mRNAs, as well as the intron‐containing pre‐mRNA encoded by the ribosomal protein gene CRY2, were differentially expressed in both prt1‐1 and upf1Δ cells (Figures 2 and 3; and data not shown). In addition, upf1Δ cells accumulated the DFR1 mRNA and prt1‐1 cells accumulated the YAP1 mRNA (data not shown). The significance of the RNAs detected in this screen was underscored by the finding that the CRY2 pre‐mRNA, previously shown to be stabilized in upf1Δ cells (F.He and A.Jacobson, unpublished data; Zhen et al., 1995), was also stabilized in prt1‐1 cells (Figure 2).
The LYS14 and HHF2 mRNAs are differentially expressed in prt1‐1 and upf1Δ cells, but are not stabilized
To determine whether the increased level of the HHF2 mRNA resulted from mRNA stabilization, we isolated RNA from prt1‐1, upf1Δ and wild‐type cells after inhibiting transcription, and measured mRNA decay rates. Figure 3A and B shows that the HHF2 mRNA is more abundant in prt1‐1 and upf1Δ cells compared with their wild‐type counterparts, but its half‐life is not increased. This indicated that the change in abundance was attributable to an increase in synthesis. A similar experiment, performed to monitor the half‐life of the LYS14 mRNA in the mutant and wild‐type strains, gave comparable results (Figure 3C and data not shown). These experiments demonstrate that the increased levels of HHF2 and LYS14 mRNAs are the result of increased transcription, not an increase in mRNA half‐life.
The half‐life of the SPT10 mRNA, encoding a transcriptional regulator of HHF2, is increased in the prt1‐1 and upf1Δ strains
Since the experiments of Figure 3A and B indicated that increases in the level of HHF2 mRNA were attributable to changes in mRNA synthesis, we attempted to identify a transcriptional regulator of HHF2 that would perhaps be a substrate for the NMD pathway. HHF2 encodes a copy of histone H4, so we analyzed the mRNA decay rates of three known transcriptional regulators of histone genes, HIR1, HIR2 and SPT10 (Osley and Lycan, 1987; Natsoulis et al., 1991, 1994; Sherwood et al., 1993; Dollard et al., 1994). Figure 3D and E shows Northern blots probed for the SPT10 mRNA after transcriptional inhibition. The half‐life of the SPT10 mRNA is increased 3‐ to 4‐fold in both the prt1‐1 and upf1Δ mutant strains, although the half‐lives of the HIR1 and HIR2 transcripts were unaffected (data not shown). Thus, the identification of HHF2 in the differential display screen appears to be the result of increasing the half‐life of the mRNA encoded by its transcriptional regulator. We observed previously a similar relationship between a subset of the URA mRNAs and their transcriptional regulator, PPR1 (Leeds et al., 1991; Peltz et al., 1994).
Sequences within the 5′ segment of the SPT10 mRNA promote NMD
Since previous work has shown that the stop codon present in a uORF can be the source of the premature termination codon that triggers NMD (Pinto et al., 1992; Cui et al., 1995; Oliveira and McCarthy, 1995), we used primer extension analysis to determine whether a putative uORF identified by sequence analysis (Figure 4B) was localized within the boundaries of the 5′ end of the SPT10 mRNA (Figure 4A). As shown in Figure 4A , the SPT10 mRNA has several transcriptional start sites, two of which include the putative uORF. These transcription start sites are specific to the SPT10 transcript, since they are not present in RNA isolated from control spt10Δ cells (Figure 4A, lane 1). To determine whether translation of the uORF was responsible for NMD of the SPT10 mRNA, the only AUG codon of the uORF was mutated to CCC and the half‐life of this mutant mRNA (construct 1; Figure 4B) was measured. (We chose to mutate the AUG and not the putative termination codon of the uORF because a second, in‐frame terminator is present within the 5′‐UTR). This change had no significant effect on the decay rate of the SPT10 mRNA in wild‐type, upf1Δ or prt1‐1 cells (Figure 4C and Table II), indicating that, in spite of the low abundance of transcripts 1 and 2, recognition of a termination codon within the uORF was not the stimulus for rapid mRNA decay.
To localize the 3′ boundary of SPT10 sequences required for NMD, we prepared chimeras between SPT10 and a 3′ segment of the well characterized PGK1 gene. The fragment of PGK1 includes the last 8% of the coding region and the 3′‐UTR, and lacks a DSE (Peltz et al., 1993). Constructs 2 and 3 of this series contain the 5′‐UTR and first 73 or 94 bp of SPT10 coding sequence fused to the PGK1 segment (Figure 5A). The half‐lives of each of the encoded mRNAs were determined in wild‐type and mutant cells by Northern blot analysis after transcriptional inhibition, and the results are shown in Figure 5B and C. The blots in both panels demonstrate that the chimera including SPT10 sequences extending 94 nt into the coding region (construct 3) encodes an unstable mRNA whose half‐life is increased 4‐fold in prt1‐1 or upf1Δ mutant cells. The mRNA with 21 fewer nucleotides of SPT10 sequence (encoded by construct 2) is only slightly destabilized in wild‐type cells and, in mutant cells, has a half‐life that is either comparable to wild type (upf1Δ), or slightly higher (prt1‐1). These results indicate that sequences sufficient to promote NMD are present within the 5′‐UTR and the first 94 nt of the coding region of the SPT10 mRNA.
Aberrant translation initiation of the SPT10 mRNA triggers NMD
The experiments of Figure 5 indicated that the region sufficient to promote rapid decay of the SPT mRNA is located within the boundaries of its 5′‐UTR and first 94 nt of coding region. Since the experiments of Figure 4 showed that the uORF present in the SPT10 5′‐UTR does not stimulate NMD, we focused on the coding region. Close examination of the SPT10 coding region sequence shown in the experiments of Figure 5 to be sufficient for triggering NMD (construct 3), indicated that this segment of mRNA harbored an AUG and a downstream UAG that are both in the +1 reading frame (Figure 6A). This, coupled with the poor sequence context of the first AUG (a U in the −3 position; Baim and Sherman, 1988; Cigan et al., 1988; Yun et al., 1996), suggested that a ribosome may occasionally scan past the first AUG, initiate at the AUG in the +1 reading frame, and then encounter a premature termination signal 15 codons downstream. Consistent with the notion that the ribosome can scan and initiate downstream of the first AUG is the observation that a frameshift mutation introduced 43 nt downstream of the initiation codon shifts the reading frame of the +1 AUG into the wild‐type frame and produces a protein with partial function (Natsoulis et al., 1994). Based on these observations, we tested the leaky scanning model by preparing another construct in which the +1 AUG was mutated to CCC to prevent its use as an initiation codon (Figure 6A, construct 3‐2). (The next available AUG is in the normal SPT10 ORF, and its use as an initiation codon would simply produce a protein lacking the first 16 amino acids; see Figure 6A). Construct 3‐2 was transformed into the prt1‐1 and upf1Δ strains, and their respective wild‐type parent strains, and decay rates of the construct 3‐2 mRNA were determined (Figure 6B). Consistent with the ‘leaky scanning’ model, the half‐life of the construct 3‐2 mRNA was increased 3‐ to 4‐fold in the wild‐type strains (compare with Figure 5A and B), but not significantly affected by the prt1‐1 and upf1Δ mutations. These results indicate that the AUG present in the +1 reading frame is necessary to promote NMD.
To test the leaky scanning model further, we prepared another chimeric construct. The results shown in Figure 5 demonstrated that the construct 2 mRNA was stabilized in the wild‐type strains, suggesting that a component necessary to promote NMD was missing from this construct. Close inspection of this mRNA (Figure 6A, construct 2) revealed that, although the +1 AUG is present, the chimera lacks the premature termination codon in the +1 reading frame (termination of the +1 reading frame in this mRNA occurs further downstream, 12 nt into the PGK1 3′‐UTR). Therefore, construct 2 was mutated to introduce a premature termination codon in the +1 reading frame six codons downstream of the +1 AUG (Figure 6A, construct 2‐1). The construct 2‐1 mRNA now has a half‐life of 3 min in wild‐type strains, i.e. the point mutation caused a 4‐fold decrease in the stability of the construct 2 mRNA (Table II and data not shown). Together, the results from the experiments of Figures 4, 5 and 6 strongly suggest that leaky scanning triggers NMD of the SPT10 mRNA.
Identification of other genes whose transcripts decay via the NMD pathway
To identify other mRNAs that may be substrates for the NMD pathway due to leaky scanning, we assessed initiation codon context. Characterization of the CYC1 and CYC7 mRNAs demonstrated that an A or a G in the −3 position relative to the AUG yielded better translation initiation than a C or a U in this position (Baim and Sherman, 1988; Yun et al., 1996). For the HIS4 gene, it was shown that A or C at position −3 yielded the greatest levels of translation (Cigan et al., 1988). Interestingly, the SPT10 mRNA has a U in the −3 position (see Figure 6A). To identify genes whose transcripts are in a suboptimal translation initiation context, we searched a database that contained sequences for approximately one‐tenth of the known yeast genes (TransTerm database; Dalphin et al., 1996). Of the 550 sequences in the database, ∼80 have a suboptimal initiation context nucleotide at the −3 position. The sequence of each of these genes was retrieved and analyzed further to determine whether the encoded mRNAs could be candidates for NMD by leaky scanning. The criteria used for screening included: (i) the presence of a U or C in the −3 position relative to the initiation codon (suboptimal sequence context); (ii) the presence of a downstream AUG that is out‐of‐frame with the initiator AUG, in an optimal sequence context, and within a distance of 90 nt; and (iii) the occurrence of a premature termination codon in the same frame as the out‐of‐frame AUG. Eighteen mRNAs that satisfied these criteria were identified and are listed in Table III. Preliminary experiments demonstrated that three of these mRNAs were sufficiently abundant to be detected by routine Northern blotting (the UBP7, REV7 and STE50 mRNAs), and they were chosen for further analysis.
Figure 7A presents the sequence of the 5′ end of the coding region of the UBP7 mRNA. The initial AUG is present in a suboptimal sequence context, having U in the −3 position. The next AUG is in the +1 reading frame and, with an A in the −3 position, is present in an optimal context. This segment of the UBP7 mRNA also contains a UAA termination codon in the +1 reading frame. Consistent with the leaky scanning hypothesis, the half‐life of the UBP7 mRNA was increased at least 2‐ to 3‐fold in the prt1‐1 and upf1Δ strains (t1/2 = >> 25 min) relative to its decay rate in the respective wild‐type strains (t1/2 = 18–20 min; Figure 7C).
Figure 7A also shows the sequence of the 5′ end of the REV7 mRNA, another transcript that satisfied our criteria for NMD by leaky scanning. Like the UBP7 mRNA, the initiation codon of REV7 is present in a suboptimal context. Furthermore, there are two AUG codons downstream in the +2 reading frame, one of which is present in a suboptimal context (U at −3) and the other in an optimal context (A at −3). A termination codon is also present in the +2 reading frame. Decay rates of the REV7 mRNA in mutant and wild‐type cells were consistent with NMD by leaky scanning since the half‐life of this mRNA was also stabilized 3‐fold in the prt1‐1 and upf1Δ cells (Figure 7B). Similarly, the half‐life of the STE50 mRNA, a third candidate for NMD by leaky scanning, was increased 2‐fold in both mutant strains (data not shown). We had shown previously that the half‐life of most wild‐type mRNAs is unaffected by mutation of the UPF1 gene (Leeds et al., 1991). Inspection of those genes indicate that they are not candidates for leaky scanning based on the criteria used here, and the half‐lives measured in UPF1 and upf1Δ strains are consistent with this interpretation. Collectively, these data strongly suggest that leaky scanning can trigger NMD.
Cells harboring the prt1‐1 allele stabilize nonsense‐containing transcripts
The documented interrelationships of mRNA decay and translation (see Introduction) prompted us to analyze mRNA turnover rates in a set of yeast mutants harboring mutations that affected the activity of specific translation initiation factors. These experiments demonstrated that one mutation, prt1‐1, led to a substantial and selective stabilization of nonsense‐containing mRNAs. All other mutations analyzed (affecting the activity of eIF2, eIF3 or eIF4E) had little or no effect on the decay rates of any mRNA studied. One explanation for the discrepancies between the different mutants is that they reflect the different extent to which translation has been inhibited. At 37°C, cells harboring the prt1‐1 allele have <5% of their normal translation activity and polyribosomes (Figure 1). Cells harboring other prt1 alleles, or other initiation factor mutations, all show less drastic effects on translation activity (Figure 1 and data not shown; Hanic‐Joyce et al., 1987; Altmann et al., 1989; Yoon and Donahue, 1992; Linz et al., 1997). Moreover, prt1‐1 cells incubated at the semi‐permissive temperature of 32°C show no effects on mRNA decay rates even though translation rates are reduced by 50% (data not shown). These data suggest that, unlike reductions in translational elongation, where even modest effects promote some degree of mRNA stabilization (Peltz et al., 1992; Zuk and Jacobson, 1998), decreases in translation initiation may need to be very severe in order to affect turnover rates of nonsense‐containing mRNAs.The effects in initiation‐compromised cells also differ from those in which elongation has been inhibited because, in the latter, all classes of mRNA are stabilized (Peltz et al., 1992; Zuk and Jacobson, 1998), presumably because of occlusion of the 5′ cap by stalled ribosomes (Beelman and Parker, 1994).
A role for eIF3 in nonsense‐mediated mRNA decay?
The substrate specificity of the mRNA stabilizing effects of the prt1‐1 mutation indicates that the decay of nonsense‐containing mRNAs, but not inherently unstable mRNAs, requires either a minimal amount of ongoing translation and/or the specific function of the PRT1 protein. This conclusion is supported in part by earlier studies which demonstrated that inhibition of yeast mRNA translation in cis, using 5′‐UTR insertions of stable stem–loop structures, did not stabilize inherently unstable mRNAs (Beelman and Parker, 1994; Sagliocco et al., 1994; Oliveira and McCarthy, 1995; Linz et al., 1997). It is unlikely, however, that the stabilizing effects of the prt1‐1 mutation are attributable to a requirement for continued synthesis of an unstable protein, since translation rates were also severely reduced in prt1‐26 cells (Figure 1) but mRNA decay rates were unaffected. A role for ongoing translation in the decay of nonsense‐containing transcripts is suggested by experiments showing that their decay: (i) occurs on polysomes (He et al., 1993; Zhang et al., 1997); (ii) can be reversed by the expression of suppressor tRNAs (Losson and Lacroute, 1979); and (iii) requires the activity of a set of factors shown also to be involved in the regulation of translational termination (Czaplinski et al., 1998).
eIF3 has been shown to stabilize the 43S preinitiation complex, promote binding of the 43S preinitiation complex to the mRNA and enhance dissociation of 40S and 60S ribosomal subunits (Trachsel and Staehelin, 1979; Feinberg et al., 1982; reviewed in Merrick, 1992). Of these functions, the one that may be linked to events occurring at premature translational termination is the dissociation of ribosomal subunits, i.e. eIF3 may initiate its dissociation function by binding to a terminating ribosome. In this respect, it is of interest that recent studies have demonstrated that: (i) the UPF1 protein interacts with, and presumably regulates, the polypeptide release factors Sup35p and Sup45p (Czaplinski et al., 1998; A.Maderazo, F.He, D.Mangus and A.Jacobson, manuscript submitted); and (ii) Escherichia coli IF3 may promote ribosome dissociation at translation termination by removing deacylated tRNA from the partial P site of the 30S subunit (Karimi et al., 1999). Thus, the comparable consequences that upf1 and prt1 mutations have for nonsense‐containing mRNAs may be indicative of disruptive effects on related events at translational termination. This interpretation of PRT1 function suggests that mRNA decay phenotypes ought to be obtained with mutations in genes encoding other eIF3 subunits, such as SUI1. While we detected only a modest stabilizing effect on nonsense‐containing mRNAs in cells harboring the sui1‐1 mutation (data not shown), we note with interest that cells containing the mof2‐1 allele of SUI1 do appear to stabilize this class of mRNAs (Cui et al., 1999).
The effects of the prt1‐1 mutation on the decay of specific yeast mRNAs have also been examined previously. Barnes et al. (1993) showed that SSA1 and SSA2 heat‐shock mRNAs were destabilized in prt1‐1 cells, and Cereghino et al. (1995) showed that SDH1 and SDH2 mRNAs suffered a similar fate. These mRNAs thus appear to require translation to maintain their stability, a phenomenon analogous to that uncovered in experiments characterizing the effects of blocking translation of yeast mRNAs in cis (Muhlrad et al., 1995) or in trans (Schwartz and Parker, 1999). These results, and those of Barnes et al. (1993) and Cereghino et al. (1995), suggest that, for some mRNAs, interruption of mRNA translation initiation at a very early step prevents formation of an mRNP structure that is capable of minimizing the rate of decapping (Muhlrad et al., 1995; Schwartz and Parker, 1999).
Identification of a new class of substrates for the NMD pathway
mRNA differential display analysis has been used successfully to identify genes that are preferentially expressed in a wide variety of eukaryotic cells (reviewed in Liang and Pardee, 1995). We took advantage of this method to determine whether cells harboring prt1 or upf1 mutations stabilized comparable sets of mRNAs. As an initial test of this approach we sought to determine whether known substrates of the NMD pathway could be detected. The CRY2 pre‐mRNA, like the CYH2, RP51B and MER2 pre‐mRNAs, is a low abundance pre‐mRNA, which has been shown previously to be stabilized in cells containing upf1 mutations (He et al., 1993; Zhen et al., 1995; F.He and A.Jacobson, unpublished data). The demonstration that the CRY2 pre‐mRNA accumulated in prt1‐1 and upf1Δ cells validated the differential display procedure. The failure to detect the CYH2 pre‐mRNA in this screen, i.e. our routine indicator for the activity of the NMD pathway (He et al., 1993, 1997), was attributable to a combination of the primer sets used and the characterization of fragments <500 bp in size.
The differential display screen, and subsequent analyses presented here, demonstrate that the SPT10 mRNA is a naturally occurring substrate for the NMD pathway and suggest that this transcript is shunted into this pathway by leaky scanning of an initiating ribosome. We base this conclusion on several observations, including: (i) an SPT10–PGK1 chimeric transcript containing the 5′‐UTR and first 73 nt of SPT10 coding region (construct 2) was stable in upf1, prt1 and wild‐type strains. This construct lacked a termination codon that would be recognized as premature in the same reading frame as the +1 AUG; (ii) a related SPT10–PGK1 chimeric transcript containing 21 additional nucleotides of SPT10 coding sequence (construct 3) was stabilized in prt1 and upf1 strains, but not in wild‐type cells. The additional sequence provided a stop codon in‐frame with the +1 AUG that could prematurely terminate translation initiated at this AUG; (iii) mutation of the +1 AUG, such that it could not be used as a translation initiation site, stabilized the chimeric mRNA; and (iv) introduction of a stop codon into construct 2 that is in‐frame with the +1 AUG created a transcript that was unstable in wild‐type cells, but stabilized by the prt1 and upf1 mutations. Collectively, these data provide another example of the linkage of mRNA decay and translation, and suggest a model in which the ribosome occasionally scans past the initiator AUG, initiates at the next AUG, terminates at the out‐of‐frame nonsense codon and triggers NMD. Analyses of leaky scanning in other yeast mRNAs suggest that initiation at the second AUG can occur as frequently as initiation at the first AUG, depending on the proximity and context of the two AUG codons (Yun et al., 1996).
Our observations with the SPT10 mRNA led us to consider the possibility that other yeast mRNAs behaved comparably. Using initiation codon context and availability of downstream out‐of‐frame initiation and termination codons as criteria, we screened the TransTerm database and identified 18 candidate mRNAs (Table III). Three of these that were sufficiently abundant for subsequent analyses (the UBP7, REV7 and STE50 mRNAs) were tested for their responses to prt1 and upf1 mutations, and shown to behave comparably to the SPT10 mRNA. These results, and the fact that the TransTerm database represents only about one‐tenth of all yeast mRNAs, lead us to conclude that as many as 3% of all yeast mRNAs could be subject to leaky scanning and NMD. Obviously, the reliability of this prediction may depend on the role of other sequence elements, including the relative importance of termination codon context (Grant and Hinnebusch, 1994; Bonetti et al., 1995) and initiation context (Cigan and Donahue, 1987; Kozak, 1997).
Leaky scanning has been recognized previously to be a mechanism by which the cell can generate multiple proteins, possibly differing in function or localization, from the same mRNA (Vivier et al., 1999). The data presented here suggest that leaky scanning may also provide the cell with a novel regulatory opportunity. Any mechanism that increases the frequency of initiation at the first AUG of mRNAs destabilized by leaky scanning would enhance the amount of full‐length protein produced from the transcript, as well as increase its half‐life. Depending on the decay rate of the ‘stabilized’ form of the mRNA, such translational regulation could lead to as much as 20‐ to 30‐fold increases in expression of selected genes. It remains to be determined whether such a mechanism is actually utilized.
Combined with earlier studies, substrates of the NMD pathway now include: (i) mRNAs that arise from genes containing nonsense mutations (Leeds et al., 1991; Peltz et al., 1993); (ii) inefficiently spliced pre‐mRNAs that enter the cytoplasm with their introns intact (He et al., 1993); (iii) some mRNAs containing uORFs (Cui et al., 1995; Vilela et al., 1998); (iv) transcripts with extended 3′‐UTRs (Pulak and Anderson, 1993; Muhlrad and Parker, 1999; F.Sherman, personal communication); and (v) some mRNAs subject to leaky scanning (this study). By inference, mRNAs that contain premature nonsense codons as a consequence of transcription or splicing errors would also be substrates. The possibility that many yeast mRNAs enter the NMD pathway by leaky scanning is in contrast to our earlier studies showing that numerous wild‐type mRNAs are not substrates for this pathway (Leeds et al., 1991; Peltz et al., 1994). Analysis of the initiation codon contexts and availability of downstream initiation and termination codons in these mRNAs, including, for example, the ACT1, ADE3, URA5 and STE2 mRNAs, indicates that these mRNAs should not initiate decay by leaky scanning.
Materials and methods
Strains, plasmids and general methods
Yeast strains and plasmids are listed in Tables IV and V, respectively. Preparation of standard yeast media and methods for cell culture, transformation, DNA manipulation and PCR amplification were as described previously (He and Jacobson, 1995; He et al., 1997). PRT1 genomic DNA used for complementation of prt1 mutant strains was isolated by colony hybridization (Sambrook et al., 1989) of a Saccharomyces cerevisiae genomic DNA library of Sau3A partial fragments constructed in YCp50 (Rose et al., 1987; provided by Duane Jenness). Approximately three genomic equivalents were screened using a PRT1 fragment (pGEM1‐PRT1) as probe. The DNA sequence of the PRT1 genomic clone was confirmed.
Measurement of translation rates and polysome analysis
As described previously (Peltz et al., 1992), cells were grown in SC‐met at 24°C to an OD600 = 0.5–0.7. Cell cultures were shifted to 37°C by the addition of an equal volume of SC‐met medium preheated to 56°C and incubated thereafter at 37°C. Aliquots of the culture (1 ml) were removed to 5 μCi of [35S]Trans label (Amersham; 15 mCi/ml, 38 TBq/mmol) and incubated for 4 min at 37°C. Incorporation of labeled amino acids was monitored by trichloroacetic acid (TCA) precipitation: 0.2 ml of cold (4°C) 50% TCA was added to each aliquot, followed by heating to 70°C for 20 min, incubation on ice for 10 min and subsequent filtration through GF/C (glass fiber) filters. The filters were washed with 10 ml of 5% TCA (4°C) and 10 ml of 95% ethanol, then dried and counted in a scintillation counter. Polyribosomes were analyzed as described previously (Peltz et al., 1992). Cells were grown at 24°C to OD600 = 0.5–0.7 and then shifted to 37°C for 15 min by the addition of an equal volume of medium prewarmed to 56°C. Cells were harvested and cytoplasmic extracts were prepared and fractionated on 15–50% sucrose gradients.
RNA extraction, RNA blotting, measurement of mRNA decay rates, transcript mapping and data analysis
Total RNA was isolated from 4 ml culture aliquots that were centrifuged and immediately frozen on dry ice at the indicated times as described previously (Herrick et al., 1990). Decay rates of individual mRNAs were determined by Northern analyses using 15 μg of RNA isolated at different times after inhibition of transcription by: (i) treatment of cells with 12 μg/ml anti‐fungal agent thiolutin (for PRT1/prt1‐1 strain pairs; Jimenez et al., 1973; Tipper, 1973; generously provided by Dr Nathan Belcher, Pfizer Central Research, Pfizer, Inc., Groton, CT); or (ii) shifting rpb1‐1 temperature‐sensitive cells to the non‐permissive temperature (for UPF1/upf1 strain pairs; Herrick et al., 1990). Temperature‐shift experiments were performed as described (Peltz et al., 1993). In brief, cells were grown at 24°C to an OD600 of 0.5–0.7, centrifuged and resuspended in 20 ml of SC‐ura medium, equilibrated to 24°C and then shifted to 37°C rapidly by the addition of 20 ml of medium prewarmed to 56°C. For PRT1 and prt1‐1 strains, cells were shifted to the non‐permissive temperature for 15 min prior to the inhibition of transcription. RNA blots, standardized for rRNA loading and integrity, were hybridized with DNA or anti‐sense RNA probes specific for the genes of interest. Probes for the ACT1, CYH2, HIS3, LEU2, MFA2, PAB1, PGK1‐UAG‐2, PGK1‐UAG‐6 and STE2 mRNAs, and the CRY2 pre‐mRNA were described previously (Herrick et al., 1990; Leeds et al., 1991; Peltz et al. 1993; Zhen et al., 1995). Probes used to detect other mRNAs included: SPT10 mRNA (full‐length), a 2 kb SalI–HindIII fragment from plasmid pFW217 (kindly provided by Fred Winston; Natsoulis et al., 1991); CDC4 mRNA, a 3.5 kb SalI fragment from plasmid SJ5141000 (kindly provided by Steve Johnson); URA5 mRNA, a 2 kb EcoRI–KpnI fragment excised from plasmid FL44 (kindly provided by Francois Lacroute); GCN4 mRNA, a 3 kb EcoRI–SalI fragment excised from plasmid p164 (kindly provided by Alan Hinnebusch); SPT10–PGK1 chimeric mRNAs, a SpeI–BglII fragment containing the 5′‐UTR and the first 73 or 94 bp of SPT10 coding region excised from constructs 2 or 3; HHF2 mRNA, a PCR fragment (∼300 bp) amplified with oligonucleotides 15 and 16 (Table VI) from plasmid pCC66 (kindly provided by Fred Winston). Using genomic DNA as template, PCR was used to generate probes for the LYS14, UBP7 and REV7 mRNAs with oligonucleotides 17 and 18, 9 and 10, and 11 and 12, respectively (Table VI). Decay rates, expressed as mRNA half‐lives, were determined by direct counting of RNA blots using a Betagen Blot Analyzer, normalizing the data such that t0 was equal to 100%, and plotting the data with respect to time on semi‐log axes (Herrick et al., 1990). Individual mRNA half‐lives in specific strains varied by less than ±15%. Primer extension analysis was performed according to the procedure described by Boorstein and Craig (1989). Total RNA (20 μg) was hybridized to an oligonucleotide primer complementary to the SPT10 sense strand at positions +30 to −5 (Table VI, oligonucleotide 4). Extension reactions utilized AMV reverse transcriptase (Boehringer Mannheim). Differential display analyses utilized the RNAimage kit from GenHunter, Inc. (Nashville, TN). DNA fragments showing differential accumulation in independent RNA samples were cloned into the vector pGEM‐T (Promega).
We thank Christine Barnes, Jef Boeke, Tom Donahue, John Hershey, Alan Hinnebusch, Jim Hopper, Steve Johnson, Gerry Johnston, Francois Lacroute, Mary Ann Osley, Roy Parker, Fred Winston and John Woolford for plasmids and strains, and members of the Jacobson laboratory (past and present) for helpful discussions and editorial comments during the preparation of this manuscript. This work was supported by a grant to A.J. from the National Institutes of Health (GM27757).
- Copyright © 1999 European Molecular Biology Organization