|Home | About | Journals | Submit | Contact Us | Français|
The nonsense-mediated mRNA decay (NMD) pathway, present in most eukaryotic cells, is a specialized pathway that leads to the recognition and rapid degradation of mRNAs with premature termination codons and, importantly, some wild-type mRNAs. Earlier studies demonstrated that aberrant mRNAs with artificially extended 3′-untranslated regions (3′-UTRs) are degraded by NMD. However, the extent to which wild-type mRNAs with long 3′-UTRs are degraded by NMD is not known. We used a global approach to identify wild-type mRNAs in Saccharomyces cerevisiae that have longer than expected 3′-UTRs, and of these mRNAs tested, 91% were degraded by NMD. We demonstrate for the first time that replacement of the natural, long 3′-UTR from wild-type PGA1 mRNA, which encodes a protein that is important for cell wall biosynthesis, with a short 3′-UTR renders it immune to NMD. The natural PGA1 3′-UTR is sufficient to target a NMD insensitive mRNA for decay by the NMD pathway. Finally, we show that nmd mutants are sensitive to Calcofluor White, which suggests that the regulation of PGA1 and other cell wall biosynthesis proteins by NMD is physiologically significant.
Nonsense-mediated decay (NMD) is a highly conserved mechanism for recognizing and rapidly degrading mRNAs with premature termination codons. It has an important protective role because it prevents the synthesis of potentially deleterious truncated proteins and is responsible for making many mutations recessive including many human disease mutations. More recently it has been recognized that NMD has a second, important role in the regulation of wild-type gene expression. Wild-type mRNAs that are regulated by the NMD pathway have been identified by global expression profiling in the yeast Saccharomyces cerevisiae, Drosophila melanogaster and humans. When NMD is inactivated 5–10% of the yeast transcriptome is affected (1–3). NMD affects a similar percentage of transcripts in the Drosophila and human transcriptome (4,6). Thus, NMD also serves an important cellular function in regulation of gene expression.
Studies in S. cerevisiae, Caenorhabditis elegans, the plant Nicotiana, and cell lines from Drosophila and humans suggest that termination codons are recognized as premature when positioned too far upstream of the poly(A) tail (7–13). mRNAs transcribed from genes containing mutations that generate aberrant extended 3′-untranslated regions (3′-UTRs) are degraded by NMD (7,11). Also synthetic mRNAs terminating translation at normal termination codons with extended 3′-UTRs are also degraded by NMD (7,11). In vitro, translation termination at stop codons followed by long 3′-UTRs is biochemically distinct from termination at stop codons followed by a normal 3′-UTR (7). The aberrant translation termination on mRNAs with long 3′-UTRs depends on Upf1p and can be rescued by tethering poly(A)-binding protein close to the stop codon. Collectively, these data are the basis for the faux 3′-UTR model, which predicts that these mRNAs are degraded by NMD because the stop codon is not in the correct context for the translating ribosomes to interact with 3′-UTR bound proteins leading to an aberrant translation termination and NMD (7). If the faux 3′-UTR model is correct, then natural cellular (wild-type) mRNAs with longer than normal 3′-UTRs should be targets for NMD by virtue of their long 3′-UTRs. However, the extent to which wild-type mRNAs with long 3′-UTRs are degraded by NMD is not known.
Consistent with the faux 3′-UTR model, most S. cerevisiae 3′-UTRs tend to be short. They typically range in size from 50 to 200 nt, with a median length of 121 nt (14). S. cerevisiae has a small but significant number of mRNAs with longer 3′-UTRs. We hypothesized that these transcripts are a novel class of wild-type mRNAs that are substrates for the NMD pathway, thus limiting their longevity in cells.
Here, we have identified a subset of wild-type S. cerevisiae mRNAs that have longer than expected 3′-UTRs. We demonstrate for the first time, to our knowledge, that wild-type S. cerevisiae mRNAs with long 3′-UTRs are strongly enriched for degradation by the NMD pathway. Many of these mRNAs are previously undetected wild-type substrates for the NMD pathway. Further we show that the natural long 3′-UTR of the PGA1 mRNA is a NMD targeting mechanism because replacement of the 3′-UTR of a NMD-insensitive mRNA with the PGA1 3′-UTR is sufficient to target this hybrid mRNA for Upf1p-dependent decay. PGA1 encodes an enzyme that is involved in glycosylphosphatidylinositol (GPI) anchor synthesis. Mutants lacking a functional NMD pathway are sensitive to the cell wall disruptor Calcofluor White suggesting that the NMD regulation of mRNAs encoding proteins involved in GPI-anchored cell wall protein biosynthesis is physiologically significant.
The S. cerevisiae strains used in this study are W303a (MATa ade2-1 ura3-11 his3-11,15 trp1-1 leu2-3,112 can1-100), AAY320 [MATa ade2-1 ura3-11 his3-11,15 trp1-1 leu2-3,112 can1-100 upf1-Δ2 (URA3)], AAY334 (MATa ura3-1 his3-11,15 trp1-1 leu2-3,112 rpb1-1) and AAY335 (MATa ura3-1 his3-11,15 trp1-1 leu2-3,112 rpb1-1 upf1-Δ2 (URA3)]. Unless otherwise stated, S. cerevisiae strains were grown and maintained using standard techniques. Susceptibility of wild-type yeast cells and upf1Δ mutant to Calcofluor White was done as described in Ram and Klis (15).
Yeast total RNA was extracted by the hot phenol method from yeast cells harvested at mid-log phase (16). Total RNA used for 3′-RACE (3′-rapid amplification of cDNA ends) was also used for quantitative northern blot analysis. 5 μg of total RNA was reverse transcribed using SuperScript™ II RT (Invitrogen Corporation, Carlsbad, CA, USA). The cDNA was PCR amplified using the Abridged Universal Amplification Primer (AUAP) provided with the 3′-RACE kit and the one primer specific to the mRNA as described in Supplementary Figure 1. The primary PCR reactions were used as templates for nested PCR reactions as described in the 3′-RACE user manual. The secondary PCR reactions were then run on 1.5% agarose gels.
mRNA steady-state levels and half-lives were measured as described by Kebaara et al. (17). Oligolabeled DNA probes were used to probe the northern blots. DNA probes were generated using primer sets for amplifying yeast open reading frames (ORFs) based on the sequences available from Saccharomyces Genome Database (SGD).
Primary mRNA sequence was predicted by combining information from genome-wide analysis of mRNA length, global identification of transcription start sites, ORF lengths and prediction of 3′-end processing sites (Figure 1). Saccharomyces cerevisiae mRNA lengths were determined genome-wide by Hurowitz and Brown (18). mRNA lengths were compared with ORF lengths as annotated in the SGD (January 2003). They found that mRNA lengths closely approximated the ORF length plus 256 nt for the combined 5′-UTR and 3′-UTR lengths. A total of 159 transcripts were categorized as transcripts greater than maximum length (18). These transcripts were predicted to have either an extended 5′-UTR and/or 3′-UTR.
We predicted the primary mRNA sequence for 116 of the 159 transcripts categorized as greater than maximum length. These 116 mRNAs correspond to the mRNAs with verified ORFs. The 5′-UTR lengths were determined from the location of the transcription start sites, when available (19,20) and the ORF lengths were obtained from the SGD (Figure 1). The mRNA 3′-processing site predictor, a tool generated by Graber et al. [http:/harlequin.jax.org/polyA/; (21)] was used to identify probable cleavage and polyadenylation sites. Since most normal S. cerevisiae mRNAs have 3′-UTRs that range in size from 50 to 200 nt, we selected 350 nt as an arbitrary cutoff. Any transcript with a predicted 3′-UTR longer than 350 nt was considered to have a long 3′-UTR. The cleavage and polyadenylation site closest to the stop codon was chosen for most mRNAs, except when there was a much stronger site predicted downstream. Of the 116 mRNAs, 56 were predicted to have long 3′-UTRs (Supplementary Table S1).
We experimentally tested these predictions for 11 mRNAs, 6 of which are shown in Table 1. We saw a strong correlation between the length of the 3′-UTR and the expected length based on the total mRNA length (18) less than the ORF and 5′-UTR lengths. The functional classes of genes encoding mRNAs with longer than expected UTRs are significantly enriched for genes involved in the regulatory control of cellular processes, especially transcription, signal transduction, cell-cycle control and metabolism (18). Among these are transcripts that encode protein components of the kinetochore, proteins involved in telomere maintenance and cell wall biogenesis. The nmd mutants have altered kinetochore function, shorter telomeres and altered telomeric silencing (1,22,23).
If the presence of a long 3′-UTR on a wild-type mRNAs is a targeting mechanism for NMD, we expect wild-type mRNAs with long 3′-UTRs to be enriched for mRNAs that are degraded by the NMD pathway. mRNAs that are degraded by the NMD pathway accumulate to higher levels and have a longer half-life in mutants lacking a functional NMD pathway (upf1Δ strains) relative to isogenic UPF1 strains [Figure 2; (17)]. To test our hypothesis, we randomly selected 11 mRNAs with 3′-UTRs >350 nt and measured their abundance in wild-type and upf1Δ mutants. Ten of these mRNAs accumulated to higher levels in an upf1Δ mutant (Table 2). Half-lives were determined for 5 of the 10 mRNAs that accumulated in an upf1Δ mutant relative to an isogenic UPF1 strain. All had longer half-lives in the upf1Δ mutant than the isogenic UPF1 strain (Figure 2 and Table 2). Thus, there is a strong correlation between long 3′-UTRs and decay by the NMD pathway.
PGA1, MPA43, DON1 and SSY5 mRNAs were previously characterized as direct substrates for the NMD pathway (1,27). PGA1 mRNA had a longer half-life in an upf1Δ strain than an UPF1 strain, and it was downregulated upon NMD reactivation (1,27). In the NMD reactivation system, a factor required for NMD (UPF2) is under the control of a regulated promoter. Yeast cells with this system have an inactive NMD pathway unless the cells are grown in inducing conditions. Upon NMD reactivation, transcripts that are NMD substrates are rapidly degraded (27). MPA43, DON1 and SSY5 mRNAs were associated with Upf1p (27). Upf1p preferentially associates with mRNAs that are direct NMD substrates (27). Our results are consistent with these results with the exception of SSY5 mRNA. In our system, SSY5 mRNA was degraded at the same rate in upf1Δ and UPF1 strains.
A role for long 3′-UTRs in targeting wild-type mRNAs for NMD predicts that the replacement of the 3′-UTR of a NMD-insensitive mRNA with the long 3′-UTR of a wild-type mRNA should make the mRNA sensitive to NMD. This prediction was tested by replacing the 3′-UTR of the NMD-insensitive miniPGK1 mRNA with the PGA1 3′-UTR [(28); Figure 3]. Wild-type PGA1 mRNA has a ~750-nt 3′-UTR and is degraded by the NMD pathway (Figure 2). The mini-PGK1 was selected because it is not degraded by the NMD pathway and it has been used as a reporter to study the cis-requirements for NMD [(28); Figure 3A]. The 3′-UTR of the hybrid miniPGK1 with the PGA1 3′-UTR (miniPGK1-PGA1 3′-UTR mRNA) was determined to be the same as the PGA1 3′-UTR using 3′-RACE (Supplementary Figure 1B). This hybrid mRNA was degraded by the NMD pathway (Figure 3B). This experiment shows that replacing the NMD-insensitive miniPGK1 3′-UTR with the wild-type PGA1 mRNA 3′-UTR is sufficient to target the hybrid transcript to NMD. These results do not exclude the possibility that there exists some undetected NMD targeting elements within the long PGA1 3′-UTR. However, we analyzed the sequences of the long 3′-UTRs for the 10 mRNAs that are degraded by the NMD pathway using the motif-based sequence analysis Multiple EM for Motif Elicitation (MEME) tool and found no common patterns.
We also expect that replacement of the long PGA1 3′-UTR with a shorter 3′-UTR from an NMD insensitive mRNA should make the PGA1 mRNA insensitive to NMD. We therefore tested whether replacing the 3′-UTR of the PGA1 mRNA with the ACT1 3′-UTR would prevent decay of the PGA1 mRNA by the NMD pathway. The ACT1 3′-UTR was selected because ACT1 mRNA is not degraded by the NMD pathway [(16); Figure 4A). The 3′-UTR of the hybrid PGA1 mRNA with the ACT1 3′-UTR (PGA1-ACT1 3′-UTR mRNA) was identical to the authentic ACT1 3′-UTR (Supplementary Figure 1A). The hybrid mRNA was not degraded by the NMD pathway (Figures 4B and 5). The stabilization of the PGA1-ACT1 3′-UTR mRNA in UPF1 cells is not due to a protective effect of the ACT1 3′-UTR sequences because PPR1-ACT1 fusion mRNA with the ACT1 3′-UTR sequences is still subject to NMD (Figure 4C). We previously showed that wild-type PPR1 mRNA is targeted for decay by the NMD pathway by an element located within the 5′-UTR and the first 92 nt of the PPR1 ORF (29).
In contrast to the NMD-dependent decay of PGA1 mRNA seen in most cells, PGA1 mRNA was NMD insensitive in cells that also expressed PGA1-ACT1 3′-UTR from a centromeric plasmid (Figure 5A), but not in cells with extra copies of PGA1 on a centromeric plasmid. This suggests coordinated regulation with PGA1-ACT1 3′-UTR mRNA. This regulation is specific to PGA1 because SPC24 mRNA and CYH2 pre-mRNA were still NMD-sensitive (Fig. 5B and 5C and Table 2). SPC24 mRNA encodes a component of the evolutionarily conserved kinetochore-associated Ndc80 complex (Ndc80p-Nuf2p-Spc24p-Spc25p), and is involved in chromosome segregation, spindle checkpoint activity and kinetochore clustering. CYH2 encodes a component of the 60S ribosomal subunit.
The stabilization of the endogenous PGA1 mRNA due to the expression of PGA1ACT1 3′-UTR mRNA is not a general effect of mRNAs producing different mRNA isoforms. The MAK31 gene encodes two mRNA isoforms, a short form with a 3′-UTR of 200 nt and a longer form with a 3′-UTR of 920 nt. The long MAK31 mRNA is degraded by NMD while the shorter mRNA is immune to the NMD pathway (Table 2). The expression of the short MAK31 mRNA does not affect the stability of the long MAK31 mRNA unlike the PGA1 mRNA and PGA1-ACT1 3′-UTR mRNA.
PGA1 encodes an essential component of GPI-mannosyltransferase II, which is involved in GPI anchor synthesis. GPI anchors are added to proteins that are to be attached to membranes (30). Pga1p, in collaboration with Gpi18p adds the second mannose residue to the GPI precursors. NMD was previously shown to regulate a set of genes coding for additional enzymes involved in the assembly of GPI-anchored cell wall proteins (1). Further, a temperature sensitive pga1-1 mutant, which is defective in GPI anchor synthesis, was sensitive to Calcofluor White (30). Calcofluor White binds to the glucan and chitin components of yeast cell walls and interferes with cell wall integrity at higher concentrations. To examine the possibility that nmd mutants may have a defect in cell wall integrity, we tested whether upf1Δ cells are hypersensitive to Calcofluor White (Figure 6A). We found upf1Δ cells are much more sensitive to Calcofluor White compared with wild-type yeast cells (Figure 6A). We also found that expression of the PGA1-ACT 3′-UTR construct in wild-type yeast cells caused the cells to be more sensitive to Calcofluor White compared with wild-type yeast cells expressing the PGA1 mRNA from a plasmid (Figure 6B, bottom panel). Expression of the PGA1-ACT 3′-UTR construct had no effect on the Calcoflour White sensitivity of NMD mutant yeast cells (Figure 6B, bottom panel).
The observation that both pga1-1 and nmd mutants are hypersensitive to Calcofluor White makes sense because they are both defective in GPI anchor synthesis. GPI-anchor synthesis is a multistep process. An increase or decrease in any of the enzymes involved in this multistep process leads to a defect in GPI anchor synthesis and sensitivity to Calcofluor White. Sensitivity to Calcofluor White by nmd mutants cannot be exclusively attributed to Pga1p because other factors required for cell wall biosynthesis have been shown to be regulated by NMD (1).
We showed that long 3′-UTRs target wild-type mRNAs for decay by the NMD pathway. Wild-type yeast mRNAs with exceptionally long 3′-UTRs are strongly enriched for degradation by the NMD pathway. Replacement of a long 3′-UTR with a shorter 3′-UTR is sufficient to prevent degradation of the wild-type PGA1 mRNA by NMD. Further replacing the 3′-UTR of a NMD-insensitive mRNA with a long 3′-UTR is sufficient to target this hybrid mRNA for decay by the NMD pathway (Figure 3). The NMD regulation of mRNAs, including PGA1 mRNA, that encode proteins involved in GPI-anchored cell wall protein biosynthesis, is physiologically significant because mutants lacking a functional NMD pathway are sensitive to Calcofluor White (Figure 6).
PGA1 may be regulated posttranscriptionally at multiple levels. Pga1p is one member of a group of enzymes that are responsible for GPI-anchor synthesis. Its levels likely needs to be carefully regulated to coordinate expression of the enzymes required for synthesis of GPI-anchored cell wall proteins. The NMD pathway may be responsible for maintaining PGA1 mRNA at a low level. Interestingly, wild-type PGA1 mRNA becomes immune to NMD in cells that also express PGA1-ACT1 3′-UTR. mRNAs are protected from NMD when they are not being translated (31). The PGA1-ACT1 3′-UTR mRNA encodes a full length, functional Pga1p, and the mRNA is present at a high level. The effect of PGA1-ACT1 3′-UTR mRNA on wild-type PGA1 mRNA suggests that wild-type PGA1 mRNA may become stored when the amount of Pga1p protein within a cell crosses a threshold. This idea is supported by the observation that expression of PGA1-ACT1 3′-UTR mRNA makes wild-type yeast cells more sensitive to Calcofluor White compared with yeast cells expressing PGA1 mRNA from a plasmid.
We predict that most wild-type yeast mRNAs with 3′-UTRs of >300 nt are degraded by the NMD pathway. MPA43 and MAK31 genes encode mRNAs that are alternatively processed. The short MPA43 mRNA has a 3′-UTR of ~300 nt and the short MAK31 mRNA has a ~200-nt 3′-UTR. The short MPA43 mRNA is degraded by NMD, while the short MAK31 is immune to NMD (Table 2). MAK31 mRNA is not actively protected from NMD because the long MAK31 mRNA (920-nt 3′-UTR) is degraded by the NMD pathway (Table 2). This suggests that 200–300 nt may be close to the minimal necessary 3′-UTR length required to target a wild-type mRNA for decay by the NMD pathway in yeast. Further, there may not be a discrete minimal 3′-UTR length required to target mRNAs for degradation by the NMD pathway. For example, the ACT1 320-nt 3′-UTR does not trigger degradation by the NMD pathway (Figure 4B). The lack of a discrete minimal 3′-UTR length requirement to target mRNAs for decay by the NMD pathway is not unexpected, because the secondary structure conformation of a 3′-UTR is also important. For example, the insertion of secondary structure into the 3′-UTR of an mRNA with an abnormally long 3′-UTR can suppress NMD (9).
The role of long 3′-UTRs in the degradation of wild-type mRNAs by NMD is likely evolutionarily conserved. Long 3′-UTRs target reporter mRNAs for decay by the NMD pathway in human cell lines, plants and C. elegans (9,10,13). Arabidopsis and Nicotiana SMG-7 mRNA, which encodes a protein required for NMD, accumulation is regulated by the NMD pathway, and the Arabidopsis SMG-7 3′-UTR is sufficient to cause NMD-dependent regulation of a GFP reporter mRNA (10). When NMD has been inhibited by hUpf1 knockdown, 75% of the wild-type mRNAs that accumulate in human HeLa cells have 3′-UTRs that are longer than 420 nt (13). Interestingly, the functional classes of mRNAs with longer than expected UTRs in yeast and humans are similar (32). These mRNAs encode proteins involved in signal transduction, transcriptional regulation, cell-cycle control and metabolism (18,32).
The 3′-UTRs of most yeast mRNAs tend to be short and homogeneous in length, averaging 121 nt (14). Others have speculated that there is an evolutionary basis for yeast mRNAs having homogenous 3′-UTR lengths (33). We propose that NMD may provide a strong selection for short 3′-UTRs by downregulating expression of genes encoding mRNAs with long 3′-UTRs. This suggests that increase in 3′-UTR lengths are to accommodate regulatory elements that may have evolved with features to evade NMD (33).
Some mRNAs with long 3′-UTRs are not degraded by the NMD pathway suggesting that these mRNAs have evolved mechanisms to evade NMD. SSY5 mRNAs from S. cerevisiae have 3′-UTRs ranging in size from 420 to 500 nt, however this mRNA is not degraded by the NMD pathway (Table 2). A significant number of human mRNAs with long 3′-UTRs are not NMD regulated (4,6). For example, the long 3′-UTRs of Cript1 and Tram1 mRNAs do not trigger NMD of a β-globin reporter RNA (13). Potential mechanisms that might have evolved for evasion of NMD by mRNA with long 3′-UTRs include stabilizer elements and secondary structures that bring the poly(A) tail into proximity of the stop codon. For example, yeast Pub1p binds to a stabilizer element that protects wild-type mRNAs with upstream ORFs from decay by the NMD pathway (34). Further, introduction of complementary sequences which can adopt a looped secondary structure into long 3′-UTRs of reporter mRNAs, confer protection from decay by the NMD pathway (9). The tertiary structure of the 3′-UTR could also be altered by binding of proteins to their target sites in 3′-UTRs (9). Thus, this altered 3′-UTR tertiary structure could be regulated by conditional expression of the corresponding 3′-UTR binding proteins.
NMD could regulate expression of genes that use alternative 3′-end processing. Several yeast genes have alternative 3′-end processing. For example MAK31 and CTR2 utilize alternative 3′-end processing sites resulting in a subpopulation of mRNAs with long 3′-UTRs [Table 2, (35)]. Thus, an expansion of our present hypothesis is that the use of alternative cleavage and polyadenylation sites resulting in a long 3′-UTR may downregulate gene expression by targeting the mRNA for rapid decay by the NMD pathway.
National Science Foundation (MCB-0444333 to A.L.A., RIG-0642154 to B.W.K.). Funding for open access charge: MCB-0444333 and RIG-0642154.
Conflict of interest statement. None declared.
We thank Rachel Urhenholt for assisting with the mRNA primary sequence prediction and Dr Khalid Sayood for the MEME analysis. We thank Dr D.P. Weeks and Melanie Langford for critical reading of the manuscript. Any opinions, findings, conclusions, or recommendations expressed in this report are ours, and do not necessarily reflect the views of the National Science Foundation.