|Home | About | Journals | Submit | Contact Us | Français|
The exosome is a complex involved in the maturation of rRNA and sn-snoRNA, the degradation of short lived non-coding RNAs and in the quality control of RNAs produced in mutants. It contains two catalytic subunits, Rrp6p and Dis3p, whose specific functions are not fully understood. We analyzed the transcriptome of combinations of Rrp6p and Dis3p catalytic mutants by high-resolution tiling arrays. We show that Dis3p and Rrp6p have both overlapping and specific roles in degrading distinct classes of substrates. We found that transcripts derived from more than half of intron-containing genes are degraded before splicing. Surprisingly, we also show that the exosome degrades large amounts of tRNA precursors despite the absence of processing defects. These results underscore the notion that large amounts of RNAs produced in wild type cells are discarded before entering functional pathways and suggest that kinetic competition with degradation proofreads the efficiency and accuracy of processing.
The exosome is a protein complex with nuclease activity that degrades RNA in the nucleus and the cytoplasm (Chlebowski et al., 2011; Lykke-Andersen et al., 2011; Mitchell et al., 1997; Schmid and Jensen, 2008). It contains a catalytic inactive doughnut-shaped core composed of 9 subunits to which associate two catalytic active subunits, Rrp6p and Dis3p, homologues respectively of bacterial RNAseD and RnaseII/R (Bonneau et al., 2009; Dziembowski et al., 2007; Liu et al., 2006; Midtgaard et al., 2006). Both Rrp6p and Dis3p are exonucleases, but the latter contains in addition an endonucleolytic (PIN) domain (Lebreton et al., 2008; Schaeffer et al., 2009; Schneider et al., 2009). The essential function of Dis3p in yeast depends on the function of both catalytic activities: proteins deprived of either one sustain growth while double mutants are not viable (Lebreton et al., 2008; Schaeffer et al., 2009; Schneider et al., 2009).
The exosome is present in both the nucleus and the cytoplasm, where it associates with different sets of cofactors (Schaeffer et al., 2011; Schmid and Jensen, 2008). Only the nuclear exosome contains both catalytic subunits and associates with the TRAMP complex, endowed with a poly(A) polymerase activity that stimulates degradation (LaCava et al., 2005; Vanacova et al., 2005; Wyers et al., 2005).
The nuclear exosome is involved in the processing of ribosomal RNAs and sn/snoRNAs and in the turnover of maturation by-products (Chlebowski et al., 2011; Lykke-Andersen et al., 2011; Schmid and Jensen, 2008). It also degrades cryptic unstable transcripts (CUTs), a class of short-lived, non-coding RNAs (Neil et al., 2009; Wyers et al., 2005; Xu et al., 2009). The cytoplasmic form is responsible for the 3′ to 5′ pathway of normal mRNA turnover (Schaeffer et al., 2011; Schmid and Jensen, 2008).
The enzyme is also involved in RNA quality control and degrades defective RNA molecules produced in several mutant conditions (Houseley et al., 2006; Lebreton and Seraphin, 2008). Early work has shown that pre-mRNAs that failed to be processed in splicing mutants are degraded by the nuclear exosome. These studies also suggested that degradation by the exosome in wild type cells competes with normal processing (Bousquet-Antonelli et al., 2000; Moore et al., 2006). This finding has been extended to RNAs produced in 3′-end processing mutants and several classes of mRNA export mutants (Hilleren et al., 2001; Jensen et al., 2001; Libri et al., 2002; Torchet et al., 2002). Another facet of the surveillance function of the exosome is the degradation of hypomodified tRNAs produced in mutants for tRNA methyltransferases (Kadaba et al., 2004; Schneider et al., 2007). However, to what extent exosome surveillance also targets molecules originating from processing defects occurring stochastically (i.e. not genetically fixed) in wild type cells remains unclear.
Rrp6p and Dis3p exonuclease activities have different properties (Tomecki et al., 2011): while Dis3p is a processive exonuclease, Rrp6p degrades RNA in a distributive manner. The two enzymes are also believed to function with different modalities. The substrate RNA must be threaded through the central pore of the doughnut shaped exosome core to access the exonucleolytic domain of Dis3p (Bonneau et al., 2009; Liu et al., 2006; Lorentzen et al., 2008). Substrates with double stranded regions must contain single stranded extensions whose length exceeds the length of the central channel of the exosome ring to be degraded. Polyadenylation by the TRAMP complex is thought to provide, when necessary, such an unstructured extension (LaCava et al., 2005; Vanacova et al., 2005; Wyers et al., 2005). Thus, the core ring limits the action of Dis3p, likely preventing the pervasive action of this enzyme on cellular RNAs. Rrp6p activity, in contrast, is not affected by the core exosome (Callahan and Butler, 2008; Liu et al., 2006) and can degrade single stranded substrates independently of their length.
The relative contributions of Rrp6p and Dis3p to the degradation of the many exosome substrates are still not fully elucidated. In studies with model targets, the two enzymes have been shown to contribute to the degradation of some substrates (e.g. CUTs) but also to display specificities. For instance, Rrp6p has been shown to be required for the degradation of the last 30 nt of the 5.8S rRNA precursors (Briggs et al., 1998) and for the final trimming of snoRNA precursors (Allmang et al., 1999; van Hoof et al., 2000), but also to have little effect on the degradation of the ribosomal 5′ external transcribed spacer (ETS). The latter is, conversely, specifically degraded by Dis3p after endonucleolytic cleavage by its PIN domain (Lebreton et al., 2008; Schneider et al., 2009).
These studies, however, were limited to a few exosome substrates and quantitative assessments of the different functions of the two catalytic subunits on a genomewide scale are still missing.
Here we present the high-resolution transcriptome analysis of cells containing several combinations of catalytic mutants disrupting the exo- and endonucleolitic function of the exosome and reveal important specificities in the role of Rrp6p and Dis3p. We also identify several regions of cryptic transcription that escaped previous detection. Importantly, we show that Dis3p degrades pre-mRNAs encoded by more than half of all yeast intron-containing genes. Most importantly, we also reveal that at least 50% of all transcripts synthesized from tRNA genes are degraded by the exosome in wild type cells, thereby revealing a major pathway of pre-tRNA turnover that competes with tRNA maturation. Our findings support the notion that the exosome exerts a general control on the flow of genetic information by competing and/or proofreading error-prone maturation processes in wild type cells. They underscore the notion that a major fraction of RNAs produced by transcription in yeast cells is degraded before achieving functional competence.
Our primary objective was to address the relative contributions of Rrp6p and Dis3p catalytic activities to exosome function and particularly to study RNA quality control pathways operating in wild type cells. Since the simultaneous mutation of both Dis3p catalytic domains is lethal in yeast (Lebreton et al., 2008), we expressed Dis3p mutants (or wild type as a control) in a background allowing the transient repression of a wild type DIS3 gene through a doxycycline repressible promoter (Tet-DIS3) (Lebreton et al., 2008). We used mutants containing either the D551N mutation in the exonucleolytic domain (exo−) or the D171N mutation in the endonucleolytic domain (endo−) or a combination of the two (D551N, D171N, exo-endo−). We also expressed the dis3 exo− and exo-endo− mutations in a Tet-DIS3, rrp6Δ background. The complete set of data can be viewed at http://steinmetzlab.embl.de/dis3/ and downloaded with ArrayExpress accession number E-MTAB-1246.
Extending previous studies performed on model substrates (Lebreton et al., 2008) we found that CUTs are targeted similarly by Dis3p and Rrp6p (figures 1A–B, Pearson correlation coefficient r2=0.57, p=3.2E−51). We also found that a large fraction (30–40%) of another class of non coding transcripts, stable unannotated transcripts (SUTs, David et al., 2006; Xu et al., 2009) are significantly stabilized upon mutation of Dis3p or Rrp6p (figure 1C and supplementary figures S1A,B). The endonucleolytic activity of Dis3p plays a minor but significant role in the process because in double exo-endo− mutants these targets are generally stabilized to a higher extent than in dis3 exo− cells (p=1.6E−32 for CUTs; figure 1B; supplementary figure S1B). The level of stabilization in double rrp6/dis3 mutants was also generally higher than in either single mutants (p=4.2*10−116 for CUTs), suggesting that the two catalytic subunits of the exosome act synergistically on these transcripts (figure 1B and supplementary figure S1).
Full inactivation of exosome function in triple rrp6Δ/dis3exo-endo− mutants extended the depth of hidden transcription in yeast by revealing roughly 1600 additional regions of cryptic transcription (figure 1D and supplementary figure S1D) relative to mutation of Rrp6p alone (Xu et al., 2009), a large fraction of which are independent unstable transcripts. This underscores the notion that nuclear degradation has a major impact on the yeast transcriptome.
Since the complexity of CUTs and SUTs is by far the highest among known exosome substrates, we conclude that the two catalytic subunits of the exosome have largely overlapping targets genomewide.
Sn/snoRNAs precursors are trimmed by the exosome up to the end of the mature forms, determined by the position of the co-transcriptionally formed sn/snoRNP core. In marked contrast to what was observed at CUTs and SUTs, where both Dis3p and Rrp6p contributed similarly to degradation, Rrp6p largely predominates over Dis3p for the processing of sn/snoRNAs precursors (figure 2A). Higher stabilization was observed in rrp6Δ cells relative to dis3 endo (p=1.5E−20), exo (p= 1.87E−17) and even dis3 exo-endo mutants (p=9.5E−11). Differently from CUTs and SUTs, the impact of inactivating the endonucleolytic or the exonucleolytic activities of Dis3p was similar (figure 2A), possibly indicating that the action of the core exosome on these targets is more markedly dependent on internal cleavage than for other targets. Interestingly, inactivation of both Dis3p and Rrp6p did not further stabilize pre-snoRNAs compared to inactivation of Rrp6p alone (figure 2A), suggesting that Dis3p cannot efficiently process sn/snoRNA precursors in the absence of Rrp6p. The levels of mature sn/snoRNAs were also globally sensitive to the exosome with the same specificities as precursors (i.e. Rrp6p > Dis3p, data not shown), presumably because a fraction of mature snoRNAs is degraded during maturation.
As a marked feature of dis3 relative to rrp6 mutants, we noticed the presence of strong signals downstream of several sn/snoRNA genes and CUTs, often extending several kilobases beyond the known or presumed end of the primary transcript (figure 2B and supplementary figure S2A). Northern blot analysis of the well-studied model gene SNR13 indicates that downstream signals indeed correspond to transcripts terminating within the body of the neighboring TRS31 gene (supplementary figure S2B–C and data not shown). These signals could originate either from a transcription termination defect or from the stabilization of read-through transcripts that are constitutively formed and degraded in a wild type strain. However, Pol II signals (Churchman and Weissman, 2011) were generally dramatically reduced or non detectable in these regions downstream of the primary transcript in wild type strains (supplementary figure S2D for SNR4 and SNR34) suggesting the existence of transcription termination defects in dis3 mutants. The possible, direct or indirect, involvement of Dis3p and the exosome in transcription termination is beyond the scope of the present study and was not examined further.
Standing out in our analyses is the finding that signals derived from roughly 50% of all intronic regions were conspicuously increased in dis3 exo− mutants (log2 ratio >1, 3A–B, supplementary figure S3). Opposite to what was observed for sn/snoRNAs, Rrp6p had a minor impact on intron degradation (figure 3B, 10% with a log2 ratio >1, p=1.1E−39 for the difference between the dis3exo− and rrp6Δ distributions) and only in the context of a functional Dis3p, since intronic signals were generally not further increased when RRP6 was deleted in a dis3exo− mutant background (figure 3B). This suggests that the contribution of Rrp6p to degradation of these intronic regions requires or is strongly facilitated by the catalytic activities of Dis3p. Only a handful of introns were stabilized to a similar extent in dis3 exo− and rrp6Δ cells, all containing snoRNAs (e.g. EFB1i, RPL7Bi, SFT1i, IMD4i), indicating that increased signals derive from unprocessed snoRNAs precursors (data not shown). The endonucleolytic activity of Dis3p alone plays a minor role in degradation, which is however significant in the context of the single dis3exo− as stabilizations observed in dis3 double mutants were generally higher (figure 3C and data not shown).
Northern blot analyses revealed that the increased intronic signals mainly derive from pre-mRNAs (figure 3C), indicating that Dis3p is involved in the degradation of unspliced precursors before they engage in splicing, after spliceosome assembly failure or as they leak to the cytoplasm. Increased levels of intronic RNAs are unlikely due to indirect effects of exosome mutation since pre-mRNAs targets could be efficiently detected in association with purified, dis3 exo− exosome, presumably trapped after recognition in the absence of degradation (supplementary figure S4).
Stabilization of pre-mRNAs by exosome mutants in wild type cells was previously observed in early experiments with a few model substrates (Bousquet-Antonelli et al., 2000; Moore et al., 2006). Our findings extend these observations to the genome scale and reveal the predominant role of Dis3p over Rrp6p in the phenomenon.
It has been reported that several intron-containing precursors are degraded as a consequence of their recognition by the nonsense mediated decay (NMD) pathway after export to the cytoplasm (He et al., 1993; Rutz and Seraphin, 2000; Sayani et al., 2008). Therefore we first considered the possibility that Dis3p might degrade these pre-mRNAs because of its involvement in NMD (Mitchell and Tollervey, 2003; Takahashi et al., 2003). We reasoned that, if most intronic substrates were targeted by Dis3p as a consequence of their recognition by the NMD pathway, the levels of stabilization observed in dis3 and NMD-defective mutants should be highly correlated. Contrary to expectations, however, cross-analyses of published (Sayani et al., 2008) intron stabilization values in upf1 and xrn1Δ cells (defective respectively for the recognition and degradation of NMD substrates) revealed very low, if any, correlation with values observed in dis3 mutants (r2=0.03, figure 4; supplementary figure S5A; see also figures 3A and C for examples of Upf1p-insensitive introns, RPL25, RPL32) and very low overlap between the top stabilized features (supplementary figure S5B). Similar conclusions can be drawn when comparing the dis3endo-exo− and the xrn1Δ dataset (supplementary figure S5A–B).
We then considered that Dis3p might degrade pre-mRNAs that fail to be processed because they have intrinsic suboptimal splicing efficiency. It is expected that poorly spliced introns are generally more sensitive to splicing mutants. Therefore, we adopted as an indicator of intron splicing efficiency the published sensitivity to several splicing mutants (Pleiss et al., 2007) and compared the latter to the stabilization levels observed in dis3exo-endo− cells. Because dis3 mutation mostly stabilized pre-mRNAs, we selected mutations affecting splicing before the first catalytic step for our analyses. Yeast introns belong to two different classes, short (70–90nt) and long (300–500nt), the latter being overall more sensitive to Dis3p mutation (figure 4B). We considered the possibility that length could also impact splicing efficiency and therefore analyzed the cumulative distribution of Dis3p sensitivities separately for the four possible classes of introns defined by length and splicing efficiency. Surprisingly, we found that length rather than splicing efficiency affects more markedly the distributions of Dis3p sensitivities (figure 4C for prp6-1 mutant cells and supplemental figure S5C for additional splicing mutants). For instance, the distributions of stabilization values for short introns are not significantly different whether these are poorly or efficiently spliced (figure 4C, p=0.82). The distributions for long introns are affected by the splicing efficiency but only to some extent (figure 4C, p=0.07). However, the distribution of Dis3p sensitivity values for long and short introns are markedly different (figure 4C, p=2.55E−11), indicating that long introns are preferential Dis3p degradation targets in a manner that is not significantly dependent on their splicing efficiency.
If pre-mRNAs are degraded because they entered dead end pathways or are exported to the cytoplasm they should not be able to be converted to mature products when stabilized. Therefore we asked to what extent intron-containing pre-mRNAs stabilized by dis3 mutations can enter the splicing pathway by analyzing the effect of the mutants on the levels of mature mRNAs derived from intronic features. We found that the presence of an intron has a strong impact on the levels of mature mRNAs in degradation mutants. Among the mRNAs that are strongly stabilized in dis3exo− mutants (log2 ratio >2), a highly significant fraction (45%) is derived from intron-containing genes (that represent roughly 5% of yeast genes). Importantly, in a large fraction of cases (46%) for which the pre-mRNA is stabilized, the levels of the mRNA also increased significantly (log2 ratio >1), with a good correlation between the increase of precursor and mRNA levels (figure 4D). These findings indicate that a large fraction of precursors targeted by Dis3p in wild type strains are functional and can be converted to products when they escape degradation. It also confirms that the pre-mRNA accumulation observed in dis3 mutants does not result from a secondary effect on splicing, which would result in a decrease rather than an increase of the corresponding mature mRNAs. Finally, it further supports the notion that these intron-containing RNAs are not degraded in the cytoplasm.
These findings demonstrate the existence of a widespread nuclear surveillance pathway targeting pre-mRNAs produced in splicing proficient cells and reveal the prominent role of Dis3p over Rrp6p in this process.
Analysis of total RNA preparations from wild type and mutant cells revealed that one class of abundant RNAs was markedly increased to represent the major species in dis3exo-endo− and dis3exo-endo− rrp6Δ cells (figure 5A). Deep sequencing of this fraction revealed that it contained essentially tRNAs (data not shown). tRNA increase is due to stabilization rather than to enhanced transcription because Pol III occupancy at several tRNA genes was detected to even lower values in mutant relative to wild type cells (supplemental figure S6).
Data from the tiling microarrays and northern blots confirmed and extended these findings (figure 5B–C). Steady state levels of mature and precursor tRNAs could be found to be increased (figure 5C) and both intron-containing and intronless classes of tRNAs were stabilized ( http://steinmetzlab.embl.de/dis3/). Supporting the direct role of the exosome in the degradation of these transcripts, both mature and pre-tRNAs could be efficiently detected in association with catalytically-impaired exosome (supplementary figure S7)
tRNA levels were affected to the highest levels by mutation of Dis3p relative to Rrp6p (p=2.2E−19) and the endonucleolytic activity of Dis3p contributed significantly to degradation (p=1.5E−10 for the difference between dis3exo− and dis3exo-endo− distributions) as expected for highly structured substrates (figure 5B). Also, mutation of the endonucleolytic activity of Dis3p alone has virtually the same effect as loss of either exonucleolytic activities alone, in marked contrast to what observed at other exosome substrates (figure 5B). The effects of Rrp6p and Dis3p mutants on tRNAs accumulation were synergistic, the amount of tRNAs being higher when both factors were inactivated (figures 5B–C, p=7.5E−11). Increase in tRNA steady state levels was extensive: roughly 80% of all tRNA species were stabilized more than two times, with an average of 2.9 times for the overall population in a dis3exo− rrp6Δ mutant. Note that since estimates of fold increase are relative to same amounts of total RNA and tRNAs represent a significant fraction of total cellular RNA, the overall increase is underestimated (see supplementary methods).
These findings were surprising since tRNAs are believed to be very stable molecules. Half lives estimates in other species are measured in days (Phizicky and Hopper, 2011), while tRNA accumulation was detected after depletion of the wild-type version of DIS3 mRNA for a maximum of eight hours. This could suggest either i) that the stability of mature tRNAs is lower than expected and Dis3p is involved in their turnover or ii) that a large fraction of short lived tRNAs and/or pre-tRNAs exists that is degraded by Dis3p in a wild type strain.
To probe for the possible existence of unstable tRNA species, we first adopted a relaxation-like approach. We monitored the kinetics of total tRNA levels increase during the inactivation of Dis3p function (i.e. the progressive replacement of wt Dis3p by the double dis3 exo-endo− mutant upon doxycycline treatment). We then allowed re-expression of wt Dis3p by removing doxycyclin from the culture medium and followed the kinetics with which the system recovers to original tRNA levels. We reasoned that if a short-lived tRNA fraction exists, it should be stabilized with a kinetics closely following changes in Dis3p activity. To accurately monitor Dis3p activity we assessed the levels of the unstable CUT NEL025c whose half-life (a few minutes Wyers et al., 2005) is negligible relative to the other factors influencing Dis3p function during the time course of the experiment (e.g. DIS3 mRNA and protein half-lives). Importantly, total tRNA levels mirrored closely changes in Dis3p activity, increasing during the inactivation phase and decaying during restoration of full exosome degradation activity, with a lag between the lowest level of Dis3p activity and the peak of tRNA levels of only roughly four hours (figure 5D). This suggested the existence of a short-lived fraction of tRNAs that is affected by Dis3p mutation.
To assess whether mature tRNAs or precursors are degraded by the exosome, we monitored the fate of newly synthesized tRNAs in mutant and wild type strains with pulse-chase experiments using tritiated uracil (figure 6). We observed a very rapid and significant accumulation of tRNA precursors in dis3exo-endo− relative to wild type cells that preceded the accumulation of mature tRNA forms (figure 6A, upper panel; see also supplementary figure S8 for quantification). Due to the higher levels of precursors, the accumulation of mature tRNA forms reached a maximum at a later point (and a higher level) in the mutant (roughly 2hrs versus 20min). Importantly, the decay rate of mature tRNAs during the chase was similar in the wild type and mutant cells (figure 6A, lower panel, and 6B), indicating that Dis3p is not significantly involved in the degradation of the steady state mature tRNA fraction. The kinetics of accumulation of 5S RNA was very different, in comparison, and neither a strong increase in the mature form nor a lag in its production where observed in exosome mutants relative to the wild type (supplementary figure S8). The observed kinetics of mature tRNA decay is compatible with a global tRNA half-life of less than 9 hours, which is markedly shorter than the turnover values currently reported in the literature for other species (Phizicky and Hopper, 2011). Importantly, since the decay rates of mature tRNAs are similar in dis3 and wild type cells and mature tRNA levels are higher in the mutant relatively to the wild type, the increase in precursors cannot be explained by tRNA processing defects in mutant cells. Rather, these findings indicate that a large fraction of tRNA precursors is degraded rapidly by the exosome in wild type cells. In mutant cells this stabilized fraction quantitatively contributes to the production of mature tRNAs. Strikingly, the observed two to three fold increase in the mature tRNA population (figure 5A–C) and the observed similar turnover in exosome mutants (figure 6B), implies that transcription of tRNA genes generates, on average, more than 50% of molecules that never reach the functional pool of mature tRNAs. Although several pathways of tRNA turnover in mutant or stress conditions have been described (Alexandrov et al., 2006; Chernyakov et al., 2008; Kadaba et al., 2004; Thompson et al., 2008; Whipple et al., 2011; Wilusz et al., 2011), we provide evidence that a very large fraction of tRNA precursors are degraded in wild type cells.
Based on the observation that CUTs and SUTs together represent the class of exosome targets with the largest complexity, it can be concluded that Rrp6p and Dis3p degrade largely overlapping targets genomewide. A former study by the Andrulis laboratory (Kiss and Andrulis, 2010) has examined the effect of depleting individual exosome subunits on the protein coding Drosophila transcriptome. In this study, D. melanogaster Dis3p and Rrp6p appear to have distinct roles in degrading mRNAs, Rrp6p being predominant in targeting NMD substrates (23% of the total Rrp6 targets). In yeast, the impact of exosome mutations on mRNAs is rather limited as less than 10% of features are affected (supplementary figure S1). However, contrary to the Drosophila system, Rrp6p and Dis3p appear to have largely overlapping mRNA targets in S. cerevisiae. Indeed, even excluding intron-containing genes from the analysis, 48 of the 69 upregulated Rrp6p targets are also Dis3p targets, even though the latter has a wider impact than Rrp6p on the coding transcriptome (310 upregulated targets in dis3 exo− versus 69 in rrp6Δ cells).
Specificities in the roles of Rrp6p and Dis3p become however apparent when different classes of substrates are analyzed, revealing differences in the mechanism of degradation or recruitment that are not relevant for CUTs (and SUTs) degradation.
The prominent role of Dis3p in the degradation of tRNAs is likely to be linked to its endonucleolytic activity, as could be expected for the degradation of highly structured substrates. This is in marked contrast with the poor effect that the dis3endo− mutation alone has on CUTs and SUTs relative to the other single mutants (compare the cumulative distributions in figure 1B and supplementary figure S1B). Although we could not detect defined endonucleolytic cleavage products in pre-tRNA molecules (data not shown), we suggest that internal cleavage, presumably at several positions in tRNA molecules, provides entry points that could be exploited either by the Rrp6p or Dis3p exonucleolytic activities.
A marked specificity was observed for Rrp6p in the degradation of sn/snoRNAs precursors (figure 2A), indicating that this factor, rather than Dis3p (and presumably the core exosome), is the main actor in the maturation of these RNAs. This is consistent with earlier findings showing a limited effect of core exosome mutants on sn/snoRNA precursor levels (Allmang et al., 1999; van Hoof et al., 2000), but against the notion that the core exosome is responsible for initiating degradation of pre-sn/snoRNAs while Rrp6p would only take care of the final trimming steps (Allmang et al., 1999). Note that the strong phenotype observed for dis3 mutants on other substrates argues against the possibility that the milder effect observed on pre-sn/snoRNAs is due to incomplete depletion of wt Dis3p in our experiments. The ancillary role of Dis3p in sn/snoRNAs processing is underscored by the observation that mutation of Dis3p only affects pre-sn/snoRNA levels in the presence of Rrp6p and generally not, or poorly, in an rrp6Δ background. This might suggest that, in the absence of Rrp6p, Dis3p cannot efficiently recognize pre-sn/snoRNAs or that it cannot degrade them. What distinguishes sn/snoRNAs precursors and CUTs in terms of exosome specificity is unclear, considering that they are believed to undergo similar 3′-end processing pathways. We favor the hypothesis that Rrp6p is preferentially recruited to sn/snoRNAs, perhaps via interactions with proteins of the core particle that could be mediated by Rrp47p as recently described (Costello et al., 2011). Alternatively, it is possible that processing occurs in some sub-cellular compartment (e.g. the nucleolus) where Rrp6p and pre-sn/snoRNPs might preferentially co-localize. Recruitment of Dis3p and the core exosome to pre-sn/snoRNPs might then be mediated by Rrp6p.
Contrary to sn/snoRNA precursors, Dis3p plays a major role relative to Rrp6p in the degradation of intron-containing pre-mRNAs (figure 3). Deletion of RRP6 has a small (if any) effect, alone or in combination with Dis3p mutations (figure 3 and data not shown). It would be enticing to speculate that this is because the endonucleolytic function of this protein is required to provide an entry point for intronic substrates that might be degraded co-transcriptionally before splicing takes place. However, the minor (if any) effect of the dis3endo− mutant indicates that internal cleavage cannot be epistatic to exonucleolytic degradation. The mechanistic bases for this specificity remain, therefore, to be elucidated.
One of the most remarkable findings of this study is that pre-mRNAs derived from more than half of intron-containing genes in yeast are degraded preferentially by Dis3p and the exosome in wild type cells. Surprisingly, we found that, although poorly spliced pre-mRNAs are generally more sensitive to degradation (p=0.07 for large introns in a prp6 mutant), intron length was by far the main determinant of Dis3p sensitivity (p=2.55E−11). The reason for this finding is a matter of speculation. We considered the possibility of a bias in this analysis due to the high level of expression of genes containing long introns, generally encoding ribosomal proteins. High stabilization values could be more easily detected for these features relative to poorly expressed genes, for which pre-mRNA levels would be too low for a reliable detection in a wild type strain (therefore “flattening” potentially large differences). However, we found that the sensitivity of our analysis allowed detecting strong stabilization values even for poorly transcribed features (i.e. in the 10% with the lowest Pol II occupancy, Venters et al., 2011), such as DID4, TFC3, YLR211c. Importantly, the highest sensitivity to degradation of pre-mRNAs containing long introns still holds after excluding from the analysis the 10% or 50% of features with the lowest expression level (based on Pol II occupancy, Venters et al., 2011), possible source of the bias. In both cases the difference between short and long introns was still highly significant (p=6.1E-9 and p=1.5E-12 respectively, data not shown).
We suggest that long introns might be more efficiently targeted by the exosome. For instance it has been shown that the Nrd1 complex, a known exosome adaptor (Butler and Mitchell, 2011; Vasiljeva and Buratowski, 2006), peaks genome-wide at intronic sites (Kim et al., 2011). Since this complex recognizes short sequence motifs that are relatively frequent in the genome, long introns might be preferential Dis3p targets because they are more likely to contain such degradation signals and/or because they are exposed as nascent RNAs for longer than short introns before splicing can take place. There might be less selective pressure to eliminate these motifs from introns (only transiently present on the transcript) than from coding sequences.
Importantly, since in a large number of cases (46%), pre-mRNAs that escape degradation in exosome mutants can be spliced, degradation does not generally target dead-end products but functional molecules. These might be stalled intermediates that can re-enter the splicing pathway when degradation is slow, or normal molecules that are degraded faster than they can undergo maturation, as earlier proposed based on a few model cases (Bousquet-Antonelli et al., 2000).
One important finding of the present study is the extensive degradation of tRNA precursors, pointing to the striking conclusion that roughly one out of every two molecules produced by tRNA genes is discarded before or during maturation. These results are consistent with recent findings from the Tollervey laboratory showing prominent crosslinking of exosome subunits to tRNA precursors (C. Schneider and D. Tollervey, pers. comm.). Degradation of tRNAs or tRNA-like molecules that are hypomodified or mistructured in cells defective for tRNA processing or modification has been previously described (Alexandrov et al., 2006; Chernyakov et al., 2008; Kadaba et al., 2004; Thompson et al., 2008; Whipple et al., 2011; Wilusz et al., 2011). In the present study we show that a significant fraction of tRNA precursors is degraded in a wild type strain, i.e. in the absence of systematic processing defects. Endonucleolytic cleavage of mature tRNAs under stress conditions to generate ½ molecules has been previously described in wild type cells (Thompson et al., 2008; Thompson and Parker, 2009). However, the exosome is not involved in this process since ½ tRNAs could be easily detected in cells deprived of exosome catalytic activity (D.L., unpublished data).
Our results could suggest a stringent requirement to eliminate stochastically occurring errors, which could strongly affect the overall balance of tRNA production. Alternatively, it is possible that the exosome degrades a fraction of normal molecules. Because failure to degrade precursors translates into a major increase in mature tRNAs, degradation appears to be in competition with processing. It is unclear, however, whether the excess of mature tRNAs produced in exosome mutants contains only normal molecules or malformed molecules that can be nevertheless processed in these conditions. Although we were not able to detect aberrancies in the pool of stabilized molecules (data not shown), it is possible that these are diluted by an excess of wild type molecules and are therefore very difficult to detect.
“Default” commitment to competing degradation for tRNA transcripts might be determined by the early recruitment of the exosome by the Nrd1 complex, which has been recently shown to associate with pre-tRNAs and favor their degradation (Wlotzka et al., 2011) or directly during transcription as Dis3p has been shown to co-purify with several subunits of RNA polymerase III (Krogan et al., 2006).
These findings call for the important question of why such a large amount of pre-tRNAs or pre-mRNAs are degraded in strains that are wild type for processing. Maturation of tRNAs and intron-containing mRNA precursors are both multistep processes for which stochastically occurring errors or incomplete reactions add up at each step and limit the overall efficiency of the process. Fast disposal of aberrant processing intermediates is likely crucial as they might inappropriately enter functional pathways or pollute the nuclear environment by titrating processing factors in dead-end intermediates.
The salient question is how the exosome specifically recognizes such a multitude of substrates that are unlikely to share common structural or sequence features. We suggest that kinetic competition between processing and degradation sets up a default destruction timer that diverts the RNA to the degradation path if maturation does not occur sufficiently fast. Exosome proofreading would then take place based on the kinetics with which the challenged molecules enter processing rather than on the specific recognition of aberrancies. Default tagging of molecules for degradation, as binding of the Nrd1 complex (Kim et al., 2011) or default recruitment of the exosome directly by RNA polymerase III (Krogan et al., 2006; Wlotzka et al., 2011), might set the molecular timer defining the kinetic competition with processing.
Major losses in the flow of events going from production to the final molecules might not look economical. However, the development of highly efficient processes might conceivably be too costly to be affordable by organisms that have to maintain evolutionary-prone variability. The maintenance of non-optimized reactions together with generic proofreading processes could be overall more economical in an evolutionary (and regulatory) perspective.
Yeast manipulations and molecular biology analyses were performed with standard procedures. Yeast strains are all derived from W303 and are listed in supplementary table 1 and originally described by Lebreton et al. (Lebreton and Seraphin, 2008). Oligonucleotides used in this study are listed in supplementary table 2. Pulse chase experiments and relaxation-like approach are described in supplementary methods.
To minimize secondary effects we analyzed the transcriptome of dis3exo-endo− cells by high density tiling arrays eight hours after addition of doxycycline, when growth rate started to decline. Hybridizations were performed as previously described (Xu et al., 2011). Additional experimental details and methods for data analysis can be found in the supplementary methods section.
We would like to thank J. Boulay for technical assistance, S.C. Muenster for help with tiling arrays hybridization, M. Riva for the gift of anti-Pol III antibodies and G. Chanfreau for critical reading of the manuscript. This work was supported by the Danish National Research Foundation (D.L.), the ANR (D.L. and A.J., ANR-08-Blan-0038-01), the CNRS (D.L., A.J. and B.S.), Institut Pasteur (A.J.), La Ligue Nationale Contre le Cancer Equipe Labellisée 2011 (B.S.) and IGBMC (B.S.), the National Institut of Health (L.M.S.). R.K.G. received a PhD fellowship from Région Ile de France and a “Roux” post-doctoral fellowship from Pasteur Institute. This research was carried out within the scope of the Associated European Laboratory LEA ‘Laboratory of Nuclear RNA Metabolism’.
All tiling microarray data are available with the ArrayExpress accession number E-MTAB- 1246.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errorsmaybe discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.