In this study, we used a genetic analysis to investigate the mechanisms of degradation that target the different RNAs produced by the MAT
a1 gene. We found that the unspliced pre-mRNA, as well as a partially spliced version of MAT
a1 that retains intron2 are degraded by Rnt1p, Rat1p and by the nuclear exosome (A). As shown previously for the RPS22B
), Rnt1p cleavage probably competes with splicing of the pre-mRNA, which explains the increase of abundance of the MAT
a1 mRNA in strains lacking Rnt1p ( and ; Supplementary Figure S1
). We note that the terminal loop of the stem–loop structure cleaved by Rnt1p might not be optimal, since it contains the AGNN motif but is a 7-nt loop (D) instead of the canonical tetraloop motif (20
). This suboptimal structure might explain why Rat1p is the major degradative activity that targets unspliced MAT
a1 pre-mRNAs, rather than Rnt1p.
We have postulated that the increased abundance of unspliced and partially spliced species in RNA degradation mutants reflect their stabilization. Because these species do not accumulate to significant levels in wild-type cells, we were unable to compare their turnover rates and to estimate their stability in the different backgrounds. Thus, we cannot formally rule out the possibility that the splicing efficiency of MAT
a1 is influenced by the disruption of degradative activities. For example, it was shown recently that RNA degradation complexes can modulate the rate of other steps in gene expression such as transcriptional elongation (26
), and the rate of elongation might affect splicing efficiency. While we cannot formally exclude the possibility of indirect effect for exonuclease mutants, the observation that spliced MAT
a1 mRNA levels increase in the absence of Rnt1p is consistent with a direct role for Rnt1p in cleaving unspliced mRNAs and a model suggesting that Rnt1p cleavage competes with splicing of the pre-mRNA (A).
Figure 5. Nuclear RNA quality control discards incompletely or aberrantly spliced forms of MATa1 that result in aberrant a1 isoforms. (A) Model of degradation of unspliced, partially spliced and mis-spliced forms of MATa1 by Rnt1p, Rat1p and the nuclear exosome. (more ...)
As opposed to many yeast pre-mRNAs (17
), unspliced MAT
a1 transcripts are not subject to nonsense-mediated mRNA decay (Supplementary Figure S2A
), but rely mostly on nuclear degradation pathways. It is possible that these species are not exported very fast, and thus rely mostly on nuclear degradation, as described previously for other genes (24
) and for the DYN2
genes (). Regardless of the reason why MAT
a1 seem to rely on nuclear degradation, our results show that the splicing of MAT
a1 might be suboptimal. This relative inefficient splicing might be caused by the small size of its introns (~50
nt), which are smaller than most S. cerevisiae
introns. In addition, folding of the stem–loop structure that sequesters the 5′-splice site of intron2 and prevents U1 snRNP binding (A) might also explain the inefficiency of intron2 splicing and why species containing this intron are detected in ribonuclease mutants. Thus, nuclear degradation mechanisms have been selected to limit the accumulation of mRNAs that have escaped the splicing machinery, and also to prevent their expression in the cytoplasm as unfunctional proteins.
Role of the partially spliced MATa1 pre-mRNA encoding a1′
In contrast to most intron containing genes, some partially spliced species of MAT
a1 encode proteins that do not correspond to truncated forms of the normal protein (B). The partially spliced MAT
a1 mRNA species that accumulates in Rnt1p and Rat1p-deficient strains had been previously analyzed in splicing mutants (9
) and further studied by Ner and Smith (8
) via experiments on branch point mutants. In this study, it was first tested whether the protein encoded by the MAT
a1 transcript including Intron2 (termed a1′) could have a biological role and work together with mature a1. This partially spliced transcript would encode a protein that is different and longer than mature a1, especially in the C-terminal domain (). The C-terminal domain of a1 is where DNA-binding occurs through Helix 3 when in complex with alpha2. Interactions with alpha2 occur between Helices 1 and 2 and are unaffected in a1′ based on primary sequence and structural analysis (27
) (PDB code: 1F43). Specific residues on a1 that are important for contacting DNA are five residues within Helix 3 (4
). In their study, Ner and Smith suggested that the a1′ protein does not have any biological role, as it is unable to rescue a1 deficiency (8
). However this protein might potentially act as a dominant negative inhibitor, explaining the requirement to limit its accumulation through the degradation of the mRNA encoding this isoform.
Although constitutively expressed in cells of the ‘a’ mating-type as a part of the haploid MAT
a transcriptional program, the a1 protein has not yet been characterized to have a direct function in these cells (28
). In diploid cells, this protein combines with the alpha2 protein in order to repress haploid-specific gene expression. In this study, we have assessed the role of Rat1p and Rnt1p in regulation of the MAT
a1 transcript in haploid cells. Rnt1p is already known to have important roles in cell-cycle regulation and cell division (30
), and is a crucial enzyme for correct processing of numerous non-coding RNAs as well as degradation of mRNAs. Our microarray data show that one of the genes that are considerably down-regulated (−12.3-fold) in rnt1Δ
cells encodes HO, the mating-type switching endonuclease gene that is known to be repressed in diploids by the a1 and alpha2 heterodimer (31
). Considering the accumulation of mature MAT
a1 in rnt1Δ
strains, as well as partially unspliced forms, one hypothesis is that the excess of a1 could associate with the a2 protein, which has many sequence similarities to alpha2, to act in repressing HO even within haploid cells. Further experiments are required to prove this model.
A novel RNA surveillance mechanism to degrade exon-skipped species
Strikingly, we also found that incorrectly spliced mRNAs are generated from the MAT
a1 locus, in which the exon2 is skipped and exon1 is spliced to exon3. These species are preferentially degraded by Rat1p and the nuclear exosome, but they also accumulate in the absence of Rnt1p. This suggests that formation of the target stem–loop structure that bridges exon2 and intron2 favors exon2 skipping for the transcripts that have folded in this structure (A). Formation of this structure would sequester the 5′-splice site of intron2 from U1 snRNP binding, promoting a direct splicing of exon1 to exon3 (D and A). Thus the increase of exon-skipped species in the rnt1Δ
strain (A and Supplementary Figure S1A
) might not be due to the fact that Rnt1p directly contributes to degrading exon skipped species, but rather because it eliminates a fraction of the pre-mRNAs which have folded in a manner that promotes the skipping of exon 2 (A). In contrast, our data suggest that Rat1p and the nuclear exosome are directly responsible for degrading these exon-skipped species. Strikingly, we found that exon2-skipped species of other genes that contain two introns also accumulate when Rat1p or the nuclear exosome are inactivated (). This suggests that nuclear RNA degradation plays a general role in degrading species that have been incorrectly spliced. This type of event is favored by the small size of central exons in yeast, facilitating the recognition of the downstream 3′-splice site in intron2, rather than the correct 3′-splice site of intron1.
We presumed that the increased abundance of the exon2-skipped products when Rat1p or the nuclear exosome are inactivated reflects their stabilization. However, we cannot formally exclude the possibility that inactivation of Rat1p or of the nuclear exosome indirectly affects splicing, and that this change of splicing efficiency is responsible for an increased rate of exon skipping. For instance, the accumulation of this product might reflect functions for Rat1p other than direct degradation, such as promoting transcriptional termination within coding regions (23
). Inactivation of Rat1p results in a faster RNA polymerase, and it is known that a faster rate of elongation promotes exon skipping as shown in mammalian cells. However, RAI1
-deficient cells that are also deficient in termination (22
) do not exhibit an accumulation of the exon2-skipped species. This suggests that accumulation of the exon2-skipped species is likely due to its lack of degradation.
It remains to be understood why these exon-skipped species are preferentially targeted by Rat1p or the nuclear exosome in the nucleus, and not by nonsense-mediated decay. Since Rat1p associates co-transcriptionally with the RNA polymerase (22
), it is possible that it acts non-discriminately on non-capped RNAs, and that exon-skipped species are particularly prone to Rat1p-mediated degradation. Inactivation of Xrn1p, the general cytoplasmic exonuclease involved in NMD does not increase the amount of exon2 skipped species, even when Rat1p is inactivated (A). This suggests that these exon2 skipped species are not exported to the cytoplasm but are retained in the nucleus. This would explain why these exon-skipped species are also affected by the nuclear exosome. Regardless of the precise mechanism of nuclear retention, the finding that this nuclear RNA degradation pathway targets species that are incorrectly spliced underscores the importance of nuclear RNA surveillance in the quality control of gene expression and identify a novel mechanism of RNA surveillance for incorrect splicing.