Abstract

Nonsense-mediated mRNA decay (NMD) performs two functions in eukaryotes, one in controlling the expression level of a substantial subset of genes and the other in RNA surveillance. In the vast majority of genes, nonsense mutations render the corresponding transcripts prone to surveillance and subject to rapid degradation by NMD. To examine whether some classes of nonsense transcripts escape surveillance, we asked whether NMD acts on mRNAs that undergo subcellular localization prior to translation. In Saccharomyces cerevisiae, wild-type ASH1 mRNA is one of several dozen transcripts that are exported from the mother-cell nucleus during mitotic anaphase, transported to the bud tip on actin cables, anchored at the bud tip, and translated. Although repressed during transport, translation is a prerequisite for NMD. We found that ash1 nonsense mutations affect transport and/or anchoring independently of NMD. The nonsense transcripts respond to NMD in a manner dependent on the position of the mutation. Maximal sensitivity to NMD occurs when transport and translational repression are simultaneously impaired. Overall, our results suggest a model in which ash1 mRNAs are insensitive to NMD while translation is repressed during transport but become sensitive once repression is relieved.

IN eukaryotes, nonsense-mediated mRNA decay (NMD) plays a role in RNA surveillance by eliminating aberrant transcripts that contain a nonsense or frameshift mutation, thereby preventing the accumulation of potentially deleterious dominant-negative proteins. In addition, a subset of functional, error-free mRNAs accumulate in a manner dependent on the NMD pathway in the yeast Saccharomyces cerevisiae (Guan et al. 2006), including transcripts with a small upstream open reading frame that initiates translation in the 5′-untranslated region (UTR) (Oliveira and McCarthy 1995), transcripts in which an internal out-of-frame open reading frame (ORF) is translated due to inefficient translation initiation at the first AUG codon (Welch and Jacobson 1999), and precursors that undergo inefficient splicing in which the intron contains an in-frame stop codon (He et al. 1993).

The UPF1, UPF2, and UPF3 genes are required for NMD in S. cerevisiae (Leeds et al. 1992). The similarities of UPF gene orthologs from different classes of organisms coincide with similarities in the pathways for NMD, including a recruitment step initiated in the nucleus involving the nucleo-cytoplasmic shuttling protein Upf3p (Shirley et al. 1998, 2002; Serin et al. 2001), followed by translation initiation, premature termination, decapping, and decay in the cytoplasm. Although NMD can trigger RNA decay during any round of translation in yeast (Maderazo et al. 2003), decay is known to occur during the pioneer round of translation while RNAs are still bound to the nuclear cap-binding complex (Gao et al. 2005).

During pioneer translation, NMD appears to be temporally and spatially coupled to nuclear export. However, in S. cerevisiae, >25 transcripts have been identified where nuclear export and translation are separated by an intervening step in which the transcripts localize via translocation on actin cables. During transport, translation is repressed. Upon arrival and anchoring at the bud tip, translational repression is relieved (Long et al. 1997; Takizawa et al. 2000; Shepard et al. 2003; Andoh et al. 2006; Aronov et al. 2007). ASH1 translation appears to utilize specialized ribosomes containing a specific subset of paralogous ribosomal proteins (Komili et al. 2007; Warner 2007). These exceptional transcripts can be exploited to learn more about NMD.

ASH1 mRNA, the best-studied transcript that localizes via actin cables, codes for a transcriptional repressor of the HO gene, which produces the endonuclease that initiates homothallic switching between a- and α-mating types (Kruse et al. 2002; Gonsalvez et al. 2005; Zarnack and Feldbrügge 2007). Asymmetric localization of the ASH1 transcript prior to translation leads to asymmetric competence to switch mating type (Chartrand et al. 2002). ASH1 is transcribed in the mother-cell nucleus during anaphase (Long et al. 1997; Takizawa et al. 1997). She2p is hypothesized to bind ASH1 mRNA in the nucleus (Kruse et al. 2002). Once in the cytoplasm, the She2p-ASH1 ribonucleoprotein particle associates with Myo4p (She1p), a type V myosin motor protein, through the adaptor protein She3p (Gonsalvez et al. 2005). As a result of these associations, ASH1 mRNA is tethered to a polarized actin cytoskeleton (Long et al. 1997; Takizawa et al. 1997).

During transport, translation of ASH1 mRNA is slowed by She2p bound at three locations in the ORF and by two translational repressors, Khd1p and Puf6p, which bind the mRNA in the 5′- and 3′-UTR, respectively (Chartrand et al. 2002; Gu et al. 2004; Paquin et al. 2007; Deng et al. 2008). Another protein, Loc1p, which affects 60S rRNA processing and ribosome assembly (Harnpicharnchai et al. 2001; Urbinati et al. 2006), represses translation and is required for anchoring at the bud tip (Long et al. 2001). Upon arrival at the bud tip, ASH1 mRNA is hypothesized to be anchored and translational repression is relieved (Gonzalez et al. 1999; Gu et al. 2004; Paquin et al. 2007; Deng et al. 2008). The ASH1 transcript cofractionates with membranes, suggesting the possibility that it may be translated by membrane-associated ribosomes (Diehn et al. 2000). Locally produced Ash1p is subsequently imported into the daughter-cell nucleus to repress transcription of HO.

None of the localized mRNAs, including ASH1, are natural targets of NMD (Diehn et al. 2000), raising the possibility that localized mRNAs that contain a nonsense mutation might be also be immune to RNA surveillance. Support for this idea came from a report that a nonsense mutation at the 5′-end of the ASH1-coding region had no effect on mRNA abundance (Gonzalez et al. 1999). To further test whether or not representative asymmetrically localized transcripts are prone to RNA surveillance through NMD, we examined the behavior of ash1 nonsense mRNAs containing mutations that terminate translation prematurely at three positions in the coding region. The results show that premature termination of translation affects mRNA localization independently of NMD. The degree of sensitivity of ash1 nonsense transcripts to NMD is influenced by the position of the nonsense mutation, the transport system, and proteins that mediate translational repression. Our results are consistent with a model presented in the discussion that is based on the postulated existence of two subpopulations of transcripts: a translationally repressed, NMD-insensitive pool and a translatable, NMD-sensitive pool. The two-pool model explains many of the phenotypes of ash1 nonsense mutations that are atypical with respect to NMD.

MATERIALS AND METHODS

Strains, plasmids, and genetic methods:

Strains and plasmids used are listed in Tables 1 and 2, respectively. Yeast transformation was performed by electroporation (Grey and Brendell 1992) or the LiAc method (Gietz and Woods 2002) Growth media were described previously (Gaber and Culbertson 1982). Yeast gene deletions were constructed using the PCR-based gene disruption method (Baudin et al. 1993; Wach et al. 1994). The accumulation and decay of nonsense and missense mRNAs were analyzed in congenic strains expressing the genes from CEN plasmids and/or chromosomally integrated alleles constructed by gene replacement.

View this table:
TABLE 1

Strains

View this table:
TABLE 2

Plasmids

Allele construction:

Full-length ASH1 was PCR cloned into the centromeric (CEN) vector pRS314, including 500 nucleotides 5′ of the ASH1 ORF and all sequences between the ASH1 stop codon and the start codon of the next downstream gene, SPE1. Site-directed PCR mutagenesis was performed to generate nonsense and missense mutations. Base substitutions were introduced at three sites in the 1750-nucleotide ASH1 ORF: +308 (site A), +968 (site B), and +1511 (site C) (Figure 1). Sites were chosen to meet three criteria:

  • Sequences in zip code regions E1, E2A, and E2B were avoided. These regions are binding sites for She2p and are required for localization (Chartrand et al. 1999; Gonzalez et al. 1999).

  • At least one consensus downstream element (TGYYGATGYYYYY) thought to be required for NMD (Ruiz-Echevarria and Peltz 1996) was located within 200 nucleotides 3′ of each mutation.

  • The nonsense codons created by base substitution were followed by an A residue, which results in optimal translation termination and efficiency of NMD (Bonetti et al. 1995).

  • Mutant alleles of ASH1 were chromosomally integrated using two-step gene replacement (Orr-Weaver and Szostak 1983). The integrity of integrated alleles was confirmed by DNA sequence analysis. To construct congenic strains, the following genes were disrupted by one-step gene replacement (Rothstein 1991): UPF1, UPF2, UPF3, SHE2, SHE3, SHE4, SHE5, KHD1, PUF6, or LOC1. Strains carrying ASH1 alleles in a she1Δ background were identified among progeny from genetic crosses.

    RNA methods:

    RNA isolation and Northern blotting were described previously (Shirley et al. 1998). Rates of RNA decay were determined by temperature shift of rpb1-1 strains from 28° to 39° or by transcription inhibition using 10 μg/ml thiolutin (Parker et al. 1991). Cells harvested before temperature shift or drug addition (t0) and at subsequent intervals were frozen in dry ice/ethanol. Total RNA was extracted and relative mRNA abundance was determined by quantitative RT–PCR using 18s rRNA as a loading control or by Northern blotting using SCR1 mRNA as loading control. Half-lives were based on average values from three trials. SigmaPlot was used to evaluate decay data using the mixed exponential decay formula y = a × exp(−b × x) + c × exp(−d × x) or the simple exponential decay formula y = a × exp(−b × x). Estimations of b, designated as B, and corresponding standard errors, designated as SE(B), were used to calculate standard error (t1/2 = log(2)/B). t1/2 ± SE(t1/2) was calculated as [log(2)/(B + SE(B)), log(2)/(B − SE(B))].

    Immunoprecipitation:

    Immunoprecipitation (IP) of She2p-cmyc or HA-Upf1p was performed as in Irie et al. (2002) with modifications. Exponentially growing yeast cultures (50 ml) were harvested at OD600 = 0.6. Cells were disrupted with acid-washed glass beads in 500 μl of lysis buffer containing 25 mm HEPES–KOH (pH 7.5), 150 mm KCl, 2 mm MgCl2 200 units/ml RNasin (Promega), 0.1% NP-40, 1 mm DTT, 0.2 μg/ml heparin, proteinase inhibitor cocktail (Sigma), and 2 mm vanadyl ribonucleoside (Sigma). Bacterial tRNA (0.2 μg) (Sigma) was used to saturate protein-G–agarose beads. IP was performed by preincubation of monoclonal anti-cmyc or anti-HA antibodies (Sigma) with protein-G–agarose at 4° overnight, followed by the addition of cell lysate at 4° for 2 hr. IP complexes were washed eight times, four with 500 μl of lysis buffer and four with 500 μl of lysis buffer containing 1 m urea.

    RNA recovery from IP and RT–PCR:

    Protein–RNA complexes were eluted from protein-G–agarose by incubation at 65° for 15 min in 100 μl of elution buffer containing 50 mm Tris–HCl (pH 8.0), 100 mm NaCl, 10 mm EDTA, and 1% SDS. RNA was extracted using phenol/chloroform, and the RNA was precipitated with ethanol and 150 mm sodium acetate (pH 5.2) overnight at −20°. The RNA pellet was washed with ice-cold 70% ethanol and treated with DNase (Ambion, Turbo DNA-free kit). RNA was quantified by two-step RT–PCR. Reverse transcription reactions were performed using the Superscript III cDNA synthesis kit (Invitrogen) or the high-capacity cDNA reverse transcription kit (Applied Biosystems). Real-time PCR reactions were performed using the Taqman universal PCR kit (Applied Biosystems) on an ABI7900HT cycler. Gene-specific primers and Taqman probes were designed using PrimerExpress software. Background mRNAs present in mock experiments performed in the absence of antibodies were 2 × 10−3 less abundant relative to the same mRNAs recovered from IP experiments.

    Statistical methods:

    Two-tailed t-tests assuming equal variance were performed and P-values were calculated to determine whether the relative levels of mRNA abundance were the same or different in pairwise sets of strains. The null hypothesis (H0) was defined as the relative mutant ash1 mRNA abundance equals relative wild-type ASH1 mRNA abundance. ANOVA F-tests were performed and P-values were calculated to determine whether the relative fold changes in mRNA levels were the same or different in strains carrying upf1Δ, upf2Δ, or upf3Δ. The null hypothesis (H0) was defined as the relative fold change in mutant ash1 or wild-type ASH1 mRNA abundance is equal in strains carrying upf1Δ, upf2Δ, or upf3Δ. Pearson's χ2 and the corresponding P-value were calculated to determine whether deletions of genes coding for motor proteins and/or translational inhibitors affect the magnitude by which NMD influences the abundance of ash1 nonsense mRNAs. The null hypothesis (H0) was defined as the effect of inactivating NMD by deleting UPF1 and the effect of deleting a gene coding for a motor protein and/or a translational inhibitor on ash1 nonsense mRNA abundance are independent. For all of the statistical tests described above, a P-value of 0.05 was used as the standard cutoff. All experiments were repeated three times (n = 3). The results of the statistical analyses are described in the supplemental tables.

    Cytological methods:

    Yeast strains W303a and AAY320 were transformed individually with 2μ plasmid pC3319 or derivatives of pC3319, where nonsense or missense mutations were introduced in the ASH1 ORF. Transformants were grown to mid-log phase in synthetic liquid medium without leucine. Cells were fixed and ASH1 mRNA localization was detected by fluorescent in situ hybridization with probes hybridizing to different parts of ASH1 mRNA (Long et al. 1997). Fifty anaphase cells with a visible ASH1 signal were counted and scored for their localization phenotype. Results were based on two trials using independent transformants.

    RESULTS

    Nonsense mutations affect ASH1 mRNA localization:

    To achieve localized protein expression, the translation of ASH1 mRNA is repressed during transport, whereas the release of translational repression is required for proper anchoring at the bud tip as a prerequisite for local translation (Chartrand et al. 2002; Gu et al. 2004; Paquin et al. 2007; Deng et al. 2008). Since premature termination of translation caused by a nonsense mutation might interfere with the release of repression and result in mRNA mislocalization, we performed experiments to assess the effects of nonsense mutations on localization. We analyzed nonsense mutations at three sites in ASH1 (Figure 1A). Site A resides upstream of the E1, E2A, E2B, and E3 binding domains for She2p, whereas the other two sites, B and C, are located between the E1/E2A and E2B/E3 domains, respectively.

    Figure 1.—

    Expression of nonsense and missense alleles of ASH1. (A) The map of ASH1 shows the locations of missense and nonsense mutations. Purple boxes indicate the locations of regions (zip codes) required for She2p/RNA binding. Red lines and red letters indicate the locations and mutational changes at sites A, B, and C. Numbers refer to nucleotides in the ASH1 open reading frame. (B) Structure of a reporter used to assay for the function of ASH1 alleles. (C) The reporter was integrated in the genome as a replacement of the wild-type CAN1 gene. Wild-type or mutant ASH1 alleles were introduced into the reporter strain on a CEN plasmid. Transcriptional repression of the HO promoter by Ash1p confers canavanine resistance. Growth was assayed on plates containing synthetic defined medium plus 100 μg/ml canavanine using 1:5 serial dilutions of mid-log cultures.

    A reporter gene was used to monitor the ability of the alleles to produce functional Ash1p, a transcriptional repressor of the HO gene. The HO promoter was fused to CAN1 (HOp-CAN1) (Figure 1B) (Bobola et al. 1996; Jansen et al. 1996). In ASH1 she2Δ strains, ASH1 mRNA mislocalizes, causing repression of HOp-CAN1 in mother and daughter nuclei and leading to canavanine resistance. Wild-type and mutant alleles of ASH1 were introduced into she2Δ strains carrying HOp-CAN1 on a CEN plasmid. Growth was monitored in the presence of canavanine (Figure 1C). Strains carrying missense mutations at sites A, B, or C were resistant to canavanine, indicating that the missense alleles produced functional Ash1p. However, strains carrying nonsense mutations at the same three sites were sensitive to canavanine, indicating significantly reduced levels of functional Ash1p.

    To examine mRNA localization in strains carrying the ash1 alleles, Cy3-labeled fluorescent probes were used to detect the mRNAs in situ (materials and methods). Anaphase cells were classified into three distinct localization phenotypes: crescent (localization at the bud tip), full bud (diffuse localization in the bud), and delocalized (diffuse localization in mother and bud) (Figure 2A). In asynchronous cultures of Nmd+ strains carrying wild-type ASH1, the Cy3 signal was localized in a crescent at the bud tip in ∼60% of cells (Figure 2B). Nonsense mutations at all three sites caused a reduction to <5% of anaphase cells showing a crescent localization pattern, whereas the percentage of cells showing a delocalized signal rose significantly. Missense mutations at sites A, B, and C had no discernible effects on localization. The results indicate that shifts from the crescent to the delocalized pattern in cells carrying the nonsense mutations are most likely due to premature termination of translation.

    Figure 2.—

    Effects of nonsense and missense mutations on ASH1 mRNA localization. (A) Nonsense codons affect the intracellular distribution of ASH1 mRNA. Yeast strain W303a was individually transformed with 2μ plasmids carrying different alleles of ASH1 and analyzed by fluorescent in situ hybridization (materials and methods; Long et al. 1997). In cells classified as “delocalized,” ASH1 mRNA was symmetrically distributed between the mother cell and the bud. In cells classified as “full bud,” ASH1 mRNA was distributed throughout the daughter cell. In cells classified as “crescent,” ASH1 mRNA formed a tight crescent at the bud tip. (B) Quantification of the effects of premature termination codons on ASH1 mRNA localization. The phenotype for 50 late-anaphase cells in A expressing mutant alleles of ASH1 mRNA were counted and classified as localized, full bud, or delocalized. Two independent trials were performed.

    In addition to the increased percentage of cells showing the delocalized signal, nonsense mutations at sites A and B caused a modest but statistically significant increase in the percentage of anaphase cells showing the full-bud mislocalization pattern. Together, the results suggest that the corresponding nonsense mRNAs were transported to the bud tip, but failed to anchor. The changes in localization caused by the nonsense mutation at site C also indicate a failure to anchor, but the effects were more dramatic. Almost 90% of cells showed a delocalized pattern. In She2p/RNA-binding experiments described in a later section, we show that nonsense mutations at sites A and B cause the mRNAs to mislocalize while remaining bound to She2p. The mRNAs are therefore still tethered to actin cables. The nonsense mutation at site C causes release of the mRNA from She2p. The mRNA probably exhibits more pronounced mislocalization because it is no longer tethered to actin cables.

    The changes in localization caused by the nonsense mutations could affect the ratio of translationally repressed mRNAs engaged in transport and the translatable mRNAs that are anchored at the bud tip. Since NMD requires translation, a shift toward more translationally repressed mRNA at the expense of translatable mRNA could reduce the overall sensitivity of the nonsense mRNAs to NMD. Furthermore, NMD itself might affect localization. When the nonsense mutations were examined in Nmd strains, the patterns of mislocalization were similar to those observed for Nmd+ strains (Figure 2C). Given limits on information that can be gained from cytological analysis, additional approaches described below were pursued to assess the potential impact of NMD on translation termination, decay, and localization of ash1 nonsense mRNAs.

    Changes in mRNA levels associated with premature translation termination and NMD:

    Since the accelerated decay of nonsense mRNAs caused by NMD is a direct consequence of premature termination of translation, we anticipated that the accumulation of missense mRNAs would be unaffected by the loss of NMD. Consistent with expectation, missense mutations at sites A, B, or C (ash1-A-ms1, ash1-B-ms1, ash1-C-ms1, and ash1-C-ms2) had no effect on mRNA abundance either in Nmd+ strains or in Nmd strains carrying upf1Δ, a deletion that inactivates NMD (Table 3).

    View this table:
    TABLE 3

    Fold changes in mRNA abundance in UPF1 and upf1Δ strains

    Since NMD accelerates the decay of nonsense mRNAs, the accumulation of nonsense mRNAs are typically reduced compared to the corresponding wild-type mRNA, whereas the wild-type level is restored when NMD is inactivated. However, the levels of ash1 nonsense mRNAs could deviate from expectation in the event of a shift favoring a higher proportion of translationally repressed, NMD-insensitive mRNAs. The cytological evidence showing that the nonsense mRNAs cause mislocalization was suggestive of this possibility.

    When transcripts produced from the nonsense alleles were analyzed in Nmd+ strains, the reduced levels typical of most nonsense mRNAs were not observed (Table 3). Instead, we found that the relative levels of ash1-A-ns1 and ash1-A-ns2 mRNAs were indistinguishable from the wild-type ASH1 mRNA. In Nmd strains carrying upf1Δ, these nonsense mRNAs were detected at a 3- to 4-fold higher level compared to ASH1 mRNA. The ash1-B-ns1 and ash1-C-ns1 mRNAs behaved differently. In Nmd+ strains, they were 1.5- to 2-fold more abundant than ASH1 mRNA. In Nmd strains, the same excess accumulation was observed as in Nmd+ strains but without any further changes that could be attributed to the inactivation of NMD.

    To summarize the data, we found that nonsense mutations at all three sites cause increased mRNA accumulation, but the only mRNAs that responded to the inactivation of UPF1 were those carrying nonsense mutations at site A. Similar experiments were performed using congenic sets of strains carrying deletions of UPF1, UPF2, or UPF3. The same results were obtained regardless of which UPF gene was deleted (Figure 3, Table 4). The NMD-dependent increases in mRNA abundance observed for mutations at site A are therefore most likely due to the inactivation of NMD rather than to the loss of function of a specific UPF gene.

    Figure 3.—

    Accumulation and decay of ash1 nonsense mRNAs. The relative fold changes in mRNA levels of the nonsense mRNAs compared to ASH1 mRNA are shown. The accumulation of ASH1 mRNA was compared with the accumulation of representative ash1 nonsense mRNAs in strains carrying null alleles of UPF1, UPF2, and UPF3. The mRNAs were detected by Northern blotting using a probe that anneals to ASH1 mRNA. The RNA levels were normalized to RPS3 mRNA, which is not affected by NMD (Guan et al. 2006).

    View this table:
    TABLE 4

    Relative mRNA abundance in Nmd strains carrying different upf gene deletions

    Transcript selection and decay of ash1-A-ns1 mRNA:

    The phenotypes described above for nonsense mutations at site A deviate from what has been observed for the typical nonsense mutation in the typical gene where mRNA abundance is usually reduced due to NMD. Since nonsense mutations at site A produced transcripts that exhibit NMD-sensitive increases in accumulation, we performed additional experiments to confirm a role for NMD and to explain the underlying reasons for the deviations from expectation.

    We asked whether Upf1p preferentially binds to the ash1-A-ns1 nonsense transcript, which is a diagnostic indicator of NMD targeting (Johansson et al. 2007). Using IP/quantitative RT–PCR (materials and methods), we found that the amount of ash1-A-ns1 mRNA associated with Upf1p was more than six-fold higher compared to ASH1 mRNA (Figure 4B). The IP was performed with lysates from strains that carried chromosomal upf1Δ and an allele of UPF1 on a CEN vector that produces a functional epitope-tagged product (HA-Upf1p). The ratios of ASH1 and ash1-A-ns1 mRNAs were calculated relative to RDR1 mRNA.

    Figure 4.—

    Co-immunoprecipitation of ASH1 and ash1-A-ns1 mRNA with Upf1p. (A) Anti-HA antibodies immunoprecipiate HA-UPF1 as detected by Western blotting of a post-IP lysate. (B) The relative amounts of ASH1 and ash1-A-ns1 mRNAs recovered in the post-IP lysates were assayed quantitatively using real-time RT–PCR. Relative RNA amounts were normalized to RDR1 RNA. (C) The rates of decay of ASH1 and ash1-A-ns1 mRNAs were determined by measuring RNA levels after inhibition of transcription using the temperature-shift method in strains carrying rpb1-1 (materials and methods). Real-time RT–PCR was used to quantify the relative amounts of RNA. 18s rRNA was used as loading control. The graphs represent the relative amounts of mRNA remaining following the inhibition of transcription at t0. The error bars show standard deviations based on three trials.

    If ash1-A-ns1 mRNA is targeted by NMD, the nonsense transcript should become more stable when NMD is inactivated and less stable compared to the wild-type mRNA in Nmd+ strains. The decay rates were examined using the temperature-shift method for transcriptional shutoff followed by Northern blotting at intervals following shutoff (materials and methods). Biphasic decay was observed for both the wild-type and the nonsense mRNA irrespective of NMD (Figure 4C). The initial phase was characterized by rapid mRNA disappearance (phase I) followed by a phase of apparent stability (phase II). The phase I decay rates for ASH1 mRNA in Nmd+ and Nmd strains were indistinguishable (2.8 ± 0.8 and 2.7 ± 0.7 min, respectively), indicating that NMD had no effect on the decay of the wild-type mRNA. The nonsense mRNA was stabilized in Nmd strains. Phase I decay rates for ash1-A-ns1 mRNA in Nmd+ and Nmd strains were 3.6 ± 0.1 and 7.2 ± 2.5 min, respectively.

    An unexpected result was observed when the decay of the wild-type and nonsense mRNAs were compared. The phase I decay rates were statistically similar (2.8 ± 0.8 and 3.6 ± 0.1, respectively) in Nmd+ strains, but the nonsense mRNA should have a faster decay rate if it is targeted by NMD. By comparison, the accumulation of the nonsense mRNA was higher than wildtype in Nmd+ strains, but NMD targeting should cause it to be lower.

    One possible explanation for the deviation from expectation is suggested by the changes in localization described above that are associated with nonsense mutations. If the mutations cause a shift toward translationally repressed mRNA at the expense of translatable mRNA, a greater proportion of the nonsense mRNA pool would be insensitive to NMD as compared to the pool of wild-type mRNA. Since the estimated phase I decay rates are composite averages of the decay rates of any subpopulations that are present, an increased pool of translationally repressed, NMD-insensitive mRNA would cause the composite decay rate of the nonsense mRNA to appear artificially higher than the decay rate of the wild-type mRNA. An increase in the proportion of more stable, NMD-insensitive mRNA might also affect phase II decay rates, but phase II is more difficult to assess because of the potential contribution of low level, residual transcription caused by incomplete inhibition of transcription. Residual transcription contributes in a minor way to phase I decay, but is more significant in phase II because low-level, ongoing transcription in phase II represents a higher proportion of the mRNAs remaining after transcriptional shutoff. Residual transcription is typically not >10% of the total mRNA (Lelivelt and Culbertson 1999; Guan et al. 2006).

    Relative pool size of She2p-bound mRNA:

    To understand the underlying causes of deviations from expectation as described above, we examined the transcript pool that binds to She2p. Since these transcripts are mostly if not entirely engaged in transport via actin cables, the She2-bound mRNA pool most likely corresponds to the translationally repressed, NMD-insensitive pool. If the relative proportion of nonsense mRNA bound to She2p increases compared to wild type, this might explain the unexpectedly high levels of accumulation observed for nonsense mRNAs and the longer-than-expected half-lives.

    The relative amounts of She2p bound to wild-type ASH1 and three ash1 nonsense transcripts were determined by IP/quantitative RT–PCR. Lysates were prepared for IP from Nmd+ and Nmd strains carrying an allele of SHE2 that produces a functional epitope-tagged product (She2p-cmyc; see materials and methods). The amount of RNA recovered by RT–PCR was normalized to a control mRNA, IST2, which localizes on actin cables but is not affected by NMD (Lelivelt and Culbertson 1999; Guan et al. 2006).

    In Nmd+ strains, the relative amount of ash1-A-ns1 mRNA that copurified with She2p-cmyc was increased 2.3- ± 0.3-fold compared to the wild-type mRNA (Figure 5A) even though both transcripts were similar in total abundance when assayed by Northern blotting (Table 3, Figure 3). In Nmd strains, the relative amount of She2p-bound ash1-A-ns1 mRNA was increased 3.7- ± 0.5-fold compared to the wild-type mRNA with a corresponding 3- to 4-fold increase in total ash1-A-ns1 mRNA abundance (Table 3, Figure 3). Since both premature termination of translation at the A site and loss of NMD contribute to an increased proportion of She2p-bound mRNA, the results support a model developed further in the discussion in which a higher proportion of the nonsense transcript is protected from NMD by translational repression, which causes distortions in the expected effects of NMD.

    Figure 5.—

    Co-immunoprecipitation of ASH1 and ash1-A-ns1 mRNA with She2p. (A) Anti-cmyc antibodies immunoprecipitate She2p-cmyc as detected by Western blotting of a post-IP lysate. (B) The relative amounts of ASH1 and ash1-A-ns1 mRNAs recovered in the post-IP lysates from Nmd+ and Nmd strains were assayed quantitatively using quantitative RT–PCR. Relative RNA amounts were normalized to IST2 RNA, which is unaffected by the inactivation of UPF genes.

    In Nmd+ strains, the relative amount of She2p-bound ash1-B-ns1 nonsense mRNA was twice that of ASH1 mRNA (Figure 5B) compared with a 1.5-fold increase in total mRNA (Figure 3, Table 3). However, in Nmd strains, the relative amount of She2-bound ash1-B-ns1 was the same as wild-type mRNA, possibly because the loss of NMD has no effect on the total abundance of ash1-B-ns1 mRNA (Table 3, Figure 3). Thus, premature translation termination caused by the ash1-B-ns1 and ash1-A-ns1 mutations have similar effects on She2-bound pool size, but the two nonsense transcripts differ in their sensitivity to NMD, possibly due to the binding of She2p at domain E1 located between sites A and site B (Figure 1A). She2p bound at E1 might affect the ability of ribosomes to reach the termination codon at site B.

    The ash1-C-ns1 mutation differed dramatically from the nonsense mutations at sites A and B. Although total ash1-C-ns1 mRNA increased twofold compared to the wild-type mRNA (Figure 3, Table3), the relative amount of mRNA bound to She2p-cmyc was reduced by 90% (Figure 5B). The mislocalization of ash1-C-ns1 nonsense mRNA (Figure 2) presumably differs from the mislocalization of mRNAs containing nonsense mutations at sites A and B because the ash1-C-ns1 nonsense mRNA is no longer bound to the actin cytoskeleton, whereas nonsense mRNAs containing mutations at sites A and B show increased binding to the actin cytoskeleton.

    Behavior of ash1 nonsense transcripts in transport-defective mutants:

    To see if disruption of the transport machinery affects the abundance and decay of ash1 nonsense transcripts, the relative levels of ash1 nonsense mRNAs were examined in Nmd+ and Nmd strains carrying she2Δ (supplemental Table S1). In Nmd+ strains, the relative levels of wild-type, ash1-A-ns1, and ash1-B-ns1 nonsense mRNAs were significantly reduced compared to SHE2 strains. The ash1-C-ns1 mRNA was modestly elevated, but with marginal statistical significance (P = 0.049). In Nmd strains, the ash1-A-ns1 mRNA was equally sensitive to NMD in both SHE2 and she2Δ strains. The ash1-C-ns1 mRNA was unaffected by NMD in both SHE2 and she2Δ strains.

    Although the ash1-B-ns1 mRNA level was not affected by NMD in SHE2 strains (Table 3), a twofold increase in abundance was observed in she2Δ Nmd strains (supplemental Table S1). Decay rates were compared in SHE2 UPF1, she2Δ UPF1, SHE2 upf1Δ, and she2Δ upf1Δ strains (Figure 6). In all four strains, biphasic decay was observed, indicating the existence of at least two pools of transcripts that decay at different rates. In SHE2 strains, the ash1-B-ns1 mRNA was insensitive to NMD (Figure 6A). However, in she2Δ strains, it was NMD sensitive. The phase I half-lives were 1.7 ± 0.1 and 2.6 ± 0.6 min in the Nmd+ and Nmd strains, respectively, which corresponds to a 1.5-fold, statistically significant difference. Since two pools of differentially decaying transcripts were detected in SHE2 strains and since the She2-bound pool was eliminated in the she2Δ strains, the NMD-insensitive pool in she2Δ strains consists of transcripts that are no longer tethered to the actin cytoskeleton but that remain translationally repressed.

    Figure 6.—

    Effects of she2Δ on the rate of ash1-B-ns1 mRNA decay in Nmd+ and Nmd strains. The decay rate of ash1-B-ns1 mRNAs was determined by measuring the disappearance of mRNA detected by Northern blotting following the inhibition of transcription with 15 μg/ml thiolutin (materials and methods). (A) Decay rates in Nmd+ and Nmd strains that carry SHE2. (B) Decay rates in Nmd+ and Nmd strains that carry she2Δ. Standard error was based on three trials.

    As an alternative approach to disrupt transport, we made use of a previously reported allele called ash1-MUT, which contains multiple mutations in the zip codes that prevent mRNA binding to She2p, but without changing the amino acid sequence of the protein product. The ash1-MUT mRNA is delocalized because it cannot bind to She2p and fails to tether to actin cables (Chartrand et al. 2002). The nonsense mutations in ash1-A-ns1 and ash1-A-ns2 were combined with ash1-MUT to produce ash1-MUT-A-ns1 and ash1-MUT-A-ns2 and then integrated at the ASH1 locus by gene replacement in congenic Nmd+ and Nmd strains. When mRNA levels were assayed by Northern blotting, we found that the inactivation of NMD had the same effect on mRNA accumulation regardless of whether tethering to the transport system was disrupted by deletion of SHE2, as described above, or by disruption of She2p binding at the zip codes (supplemental Table S2).

    We wanted to know whether a more general disruption of transport affects the behavior of ash1 nonsense mRNA. The relative levels of wild-type ASH1 and ash1-A-ns1 nonsense mRNAs were determined by Northern blotting in strains carrying deletions of genes required for transport, including she1Δ (type V myosin motor protein), she3Δ (actin–myosin adaptor protein), she4Δ (regulator of She1p), and she5Δ (actin filament assembly). For each set of sheΔ strains, the inactivation of NMD caused a three- to fivefold increase in ash1-A-ns1 mRNA abundance compared to ASH1 mRNA abundance (supplemental Table S3). Chi-square tests indicated that the magnitudes of change caused by the inactivation of NMD were similar regardless of which SHE gene was deleted (supplemental Table S4). The effects of deleting these genes were similar to effects of she2Δ or the mutations in ash1-MUT that prevent tethering of the mRNA to the transport system. Overall, the results indicate that disruption of transport did not cause increased sensitivity of ash1-A-ns1 nonsense mRNA to NMD. These results might be explained if mRNAs that are not tethered to the actin cytoskeleton remain translationally repressed.

    Behavior of ash1-A-ns1 nonsense mRNA in the absence of translational repressors:

    The protein products of PUF6, KHD1, and LOC1 have been implicated in mediating translational repression of ASH1 mRNA during transport (Long et al. 2001; Irie et al. 2002; Gu et al. 2004). The effects of deleting genes coding for translational repressors were investigated in Nmd+ and Nmd strains expressing ash1-A-ns1 mRNA (supplemental Tables S4–S6) to see if relief of translational repression might cause an increased proportion of transcripts that are sensitive to NMD. We found that puf6Δ caused an overall reduction in the abundance of ash1-A-ns1 mRNA, but the effect was unrelated to NMD. The reduction could be due to either direct or indirect effects of Puf proteins on mRNA stability (Wickens et al. 2002). Single deletions of KHD1 or LOC1 had no significant effect on mRNA abundance, and no increase in sensitivity to NMD was observed as the result of deleting these translational repressors one at a time. These experiments were extended by analyzing khd1Δ and puf6Δ double deletions in combination with a she2Δ deletion to see if ash1-A-ns1 mRNAs that are not tethered to the actin cytoskeleton are more sensitive to NMD in the absence of translational repressors. Once again, no increase in sensitivity was observed (supplemental Table S6).

    One possible explanation for the results described above is that the translational repressors have redundant effects on translation. Maximal sensitivity to NMD might occur only in transport-defective mutants where the genes for translational inhibitors are simultaneously deleted. To test this, ash1-A-ns1 mRNA was analyzed in she2Δ strains carrying deletions of UPF1, KHD1, and PUF6 in multiple combinations. When mRNA levels were compared, we found, as expected from previous results, that disruption of NMD caused increased accumulation of ash1-A-ns1 mRNA (Figure 7A). In Nmd+ strains, a trend toward higher accumulation of ash1-A-ns1 mRNA was observed with the highest level found in a strain simultaneously deleted for UPF1, KHD1, PUF6, and SHE2.

    Figure 7.—

    Accumulation and decay of ash1-A-ns1 mRNA in she2Δ strains lacking inhibitors of ASH1 translation. (A) Levels of accumulation were compared by Northern blotting. (B) The decay of ash1-A-ns1 mRNA was determined in Nmd+ and Nmd strains carrying upf1Δ, khd1Δ, puf6Δ, and she2Δ as described in Figure 6 and materials and methods.

    To assess whether changes in mRNA abundance in the quadruple mutant reflect underlying changes in mRNA half-life, we measured the kinetics of decay of ash1-A-ns1 mRNA in Nmd+ and Nmd strains carrying simultaneous deletions of KHD1, PUF6, and SHE2 (Figure 7B). The nonsense mRNA decayed with biphasic kinetics in the Nmd+ strain. The phase I decay rate was extremely rapid with an estimated half-life of 0.13 ± 0.02 min. The kinetics of decay also show the presence of a more slowly decaying pool of mRNAs with a half-life of 27 ± 2.4 min. In the Nmd strain, decay was monophasic. A dramatic stabilization was observed in which the overall half-life was 15.4 ± 2.0 min. A NMD-insensitive mRNA subpopulation corresponding to phase II could still be present in the Nmd strain, but might be obscured by the predominant NMD-sensitive subpopulation. Compared with previous observations, these results suggest that the ash1-A-ns1 nonsense transcript is hypersensitive to NMD when it cannot tether to the transport system and when it can be more efficiently translated in the absence of the Khd1p and Puf6p translational repressors.

    DISCUSSION

    To our knowledge, there have been no studies that address whether particular classes of transcripts are immune to or sequestered from the effects of NMD. We studied ASH1 mRNA to assess how mRNAs that are transported by the actin cytoskeleton prior to translation are affected by blocks in translation and whether the mRNAs are sensitive or immune to NMD. According to previously proposed models (Chartrand et al. 2002), tight regulation exists between two temporally incompatible events: translation and the transport of mRNAs tethered to the transport machinery through the binding of She2p to the mRNAs. Our data support a model in which full-length translation of ASH1 mRNA is an integral part of the maturation pathway. Blocks in translation cause mislocalization. Furthermore, ash1 nonsense mRNAs are prone to NMD, but sensitivity to NMD depends on the position of the mutation.

    The ashA-ns1 nonsense mRNA accumulated and responded to loss of NMD in a manner atypical of transcripts encoded by nonsense alleles of other genes. Typically, nonsense mRNA abundance is reduced compared to the corresponding wild-type mRNA. The reduction is caused by an acceleration of the mRNA half-life due to targeting of the mRNA by NMD. When NMD is inactivated, the abundance rises back to the same level as the wild-type mRNA (Leeds et al. 1992). By contrast, the wild-type ASH1 and nonsense ash1-A-ns1 mRNAs were equally abundant in Nmd+ strains. When NMD was inactivated, the ash1-A-ns1 nonsense mRNA was two- to fourfold more abundant than wild-type ASH1 mRNA.

    An explanation for the atypical behavior comes from the finding that the relative proportion of ash1-A-ns1 mRNA bound to She2p, and therefore tethered to the actin cytoskeleton, rose twofold in Nmd+ strains and fourfold in Nmd strains. These results reflect an anchoring defect caused by the nonsense mRNA that is accentuated when NMD is inactivated. The results are summarized in Figure 8. The underlying cause of deviations in the expected behavior of ash1-A-ns1 mRNA appears to be related to the division of ASH1 mRNA into two pools. One pool, which is bound to She2p, is engaged in transport. These mRNAs are translationally repressed. Since NMD requires translation, the She2p-bound pool is insensitive to NMD. The other pool is anchored at the bud tip. These mRNAs are translationally derepressed and sensitive to NMD. This could be explained by changes in the relative sizes of the two pools caused by premature termination of translation and are further accentuated by the inactivation of NMD.

    Figure 8.—

    Dual-pool model. Nonsense mutations at the A site (Figure 1) cause a change in the relative proportion of two pools of mRNA: (1) She2p-bound mRNAs, which are tethered to the actin cytoskeleton, are engaged in transport, and are translationally repressed and NMD insensitive and (2) mRNAs anchored at the bud-tip cortex, which are translatable and sensitive to NMD. The numbers in the circles in A and B represent the relative pool sizes on the basis of experimental evidence shown in Figure 5. The same relative ratios are shown in tabular form in C. Changes in the relative pool sizes due to premature termination at site A and to the inactivation of NMD are consistent with data presented in the text.

    According to the model (Figure 8), impaired anchoring caused by premature termination at the A site causes a change in the relative sizes of the two pools in Nmd+ strains, favoring an increase in the size of the transport pool at the expense of the anchored pool. This reduces the proportion of mRNAs that are sensitive to NMD. Because of the shift toward NMD-insensitive mRNAs, the measured half-life of ash1-A-ns1 (Figure 4) appears to be similar to wild-type ASH1 mRNA, but in reality the shift masks more rapid decay of the NMD-sensitive pool. The data on accumulation, decay, and the relative binding of ash1-A-ns1 mRNA with She2p are predictable outcomes of the model. Most notably, twice as much ash1-A-ns1 mRNA is bound to She2p compared to ASH1 mRNA in Nmd+ strains, whereas four times as much is bound in Nmd strains (Figure 5). Furthermore, the total abundance of the nonsense and wild-type mRNAs are the same in Nmd+ strains, but in Nmd strains the nonsense mRNA is more abundant (Figure 3).

    In Nmd strains, mRNAs in the anchored pool are more stable, which increases the size of the pool. If anchoring of the nonsense mRNA is a rate-limiting bottleneck for entry into this pool, slower decay of the anchored pool might slow entry into the pool and cause a rise in the size of the She2p-bound pool. Alternatively, in the absence of Upf1p, less efficient termination might also slow entry into the anchored pool since Upf1p is known to promote efficient termination of translation in conjunction with the release factors eRF1 and eRF3 as a prerequisite for decapping and decay (Wang et al. 2001). In either case, our data support the idea that anchoring depends not only on translation but also on the rate at which anchored mRNAs are degraded.

    To reveal the full potential of NMD to degrade ash1-A-ns1 mRNA, we disrupted the transport system by deleting five different SHE genes and by analyzing nonsense alleles in which the ash1-A-ns1 and ash1-A-ns2 mutations were combined with multiple mutations in the zip codes to prevent the binding of She2p. Although none of the mutations had any added effect on sensitivity to NMD by themselves or in combination with single deletions of genes coding for translational repressors of ASH1, we found that ash1-A-ns1 mRNA was hypersensitive to NMD in she2Δ strains carrying simultaneous deletions of PUF6 and KHD1. The rapid decay in the triple-deletion strain demonstrates the full efficacy of NMD-mediated degradation of ash1-A-ns1 mRNA, but this rapid decay occurs only when the mRNA is in an actin-detethered, translationally derepressed pool that does not normally exist.

    The nonsense mutations at sites B and C differ from site A since the latter is located upstream of all She2p-binding domains whereas the former are located downstream of one or more of the binding domains. Like the mutations at site A, mutations at sites B and C cause increased mRNA abundance relative to wild type in Nmd+ strains, but unlike the mutations at site A, the corresponding mRNAs are insensitive to the inactivation of NMD in otherwise wild-type strains. The nonsense mutation at site B causes a mislocalization phenotype similar to nonsense mutations at site A, but the mutation at site B confers more severe mislocalization. Like ash1-A-ns1, the ash1-B-ns1 mutation causes a shift toward increased amounts of mRNA bound to She2p in Nmd+ strains, but unlike ash1-A-ns1, no further increase was observed when NMD was inactivated, consistent with its insensitivity to NMD. Reduced binding to She2p was observed for a nonsense mutation at site C in both Nmd+ and Nmd strains.

    Most importantly, the ash1-B-ns1 nonsense mRNA, which was insensitive to NMD in SHE2 strains, became sensitive in she2Δ strains as evidenced by changes in mRNA abundance and decay rate in Nmd strains. The difference in phenotypes between mutations at sites A and B might be explained by their locations relative to positions of the binding sites for She2p. The sequence of events at the time of anchoring is unknown. However, if repression of translation initiation is relieved before the mRNA is released from binding to She2p, translation could initiate and proceed unimpeded to site A where translation terminates. However, ribosomes would have to transit across the E1-binding domain for She2p to reach a termination codon at site B. If She2p bound at E1 slows translation, as proposed earlier (Chartrand et al. 2002), this could result in inefficient termination at downstream site B. It has been shown that reduced rates of translational elongation cause inefficient termination leading to readthrough (Sandbaken and Culbertson 1988). Since efficient termination is required for NMD (Bonetti et al. 1995), sensitivity to NMD might be negatively affected.

    The nonsense mutation at site C was insensitive to NMD in all strain backgrounds tested. The insensitivity to NMD could be related to the location of the mutation 356 nucleotides upstream of the normal stop codon (Figure 1). Mutations near the normal stop codon can be insensitive to NMD. For example, the HIS4 frameshift mutation his4-713, which is located 120 nucleotides upstream of the normal stop codon and which generates a premature stop codon immediately after the site of frameshifting, is NMD insensitive (Mathison and Culbertson 1985; Leeds et al. 1991). The proposed explanation for insensitivity is that A/U-rich downstream sequence elements required for NMD are not present in between the 3′-proximal premature stop codon and the normal stop codon (Ruiz-Echevarria and Peltz 1996). We identified a potential downstream element, but it might not be functional. In that case, cells might recognize the ash1-C-ns1 premature stop codon as a normal stop codon.

    When nonsense transcripts produced by alleles carrying mutations at sites A or B fail to anchor, they mislocalize as She2-bound transcripts. However, when ash1-C-ns1 nonsense mRNA fails to anchor, it dissociates from She2p and mislocalizes without being tethered to the actin cytoskeleton. By the time a ribosome reaches site C, it has passed through and presumably displaced She2p binding at three of the four binding domains, leaving the mRNA tethered only through binding in the 3′-UTR. If termination occurs at the normal stop codon, She2p releases an mRNA anchored in the cortex of the bud tip. However, if termination occurs at site C, the mRNA dissociates from She2p without anchoring. The results suggest the existence of a mechanism to ensure that anchored transcripts can be translated full length.

    Acknowledgments

    This is Laboratory of Genetics paper no. 3635. This work was supported by the University of Wisconsin College of Agricultural and Life Sciences, the University of Wisconsin School of Medicine and Public Health, National Institutes of Health grant GM65172 (M.R.C.), National Science Foundation grant DB0744017 (M.R.C.), and Kirschstein National Research Service Award Individual predoctoral fellowship F31 GM077078 (J.S.F.).

    Footnotes

    • Communicating editor: M. Hampsey

    • Received August 29, 2008.
    • Accepted September 1, 2008.

    References

    View Abstract