Comparative poly(A)+ RNA interactome capture of fission yeast RNA exosome mutants reveals insights into RNA biogenesis

The nuclear RNA exosome plays a key role in quality control and processing of multiple protein-coding and non-coding transcripts made by RNA polymerase II. A mechanistic understanding of exosome function remains a challenge given it has multiple roles in RNA regulation. Here we have analysed changes in the poly(A)+ RNA transcriptome and interactome provoked by mutations in three distinct subunits of the nuclear RNA exosome. We have identified multiple proteins whose occupancy on RNA is altered in the exosome mutants. We demonstrate that the Zinc-finger protein Mub1 regulates exosome dependent transcripts that encode stress-responsive proteins. Furthermore, we assess impact of the exosome inactivation upon RNA binding of the components of the mRNA processing machineries such as spliceosome and mRNA cleavage polyadenylation complex. We show that mutations in the exosome lead to accumulation of the components of U1 and U2 snRNPs on poly(A)+ RNA and depletion of the components of the activated spliceosome from RNA suggesting that the early stages of spliceosome assembly might provide a critical quality control step. Collectively, our data provide a global view of how RNA metabolism is affected in the exosome-deficient cells and reveal RNA-binding proteins that may act as novel exosome cofactors.


ABSTRACT
The nuclear RNA exosome plays a key role in quality control and processing of multiple proteincoding and non-coding transcripts made by RNA polymerase II. A mechanistic understanding of exosome function remains a challenge given it has multiple roles in RNA regulation. Here we have analysed changes in the poly(A)+ RNA transcriptome and interactome provoked by mutations in three distinct subunits of the nuclear RNA exosome. We have identified multiple proteins whose occupancy on RNA is altered in the exosome mutants. We demonstrate that the Zinc-finger protein Mub1 regulates exosome dependent transcripts that encode stress-responsive proteins. Furthermore, we assess impact of the exosome inactivation upon RNA binding of the components of the mRNA processing machineries such as spliceosome and mRNA cleavage polyadenylation complex. We show that mutations in the exosome lead to accumulation of the components of U1 and U2 snRNPs on poly(A)+ RNA and depletion of the components of the activated spliceosome from RNA suggesting that the early stages of spliceosome assembly might provide a critical quality control step.
Collectively, our data provide a global view of how RNA metabolism is affected in the exosomedeficient cells and reveal RNA-binding proteins that may act as novel exosome cofactors.
EXO-9 forms a double-layered barrel-like structure that comprises six ribonuclease (RNase) (PH)-like proteins (Rrp41, Rrp42, Rrp43, Rrp45, Rrp46, and Mtr3) and three S1/K homology (KH) "cap" proteins (Rrp4, Rrp40, and Csl4) (26). The two catalytic subunits occupy opposite ends of EXO-9 to constitute EXO-11 (27,28). Rrp6 is located at the top of the S1/KH cap ring near the RNA entry into the channel formed by the exosome cap and core, and Dis3 is found at the bottom of EXO-9 channel near the RNA exit pore. Both Rrp6 and Dis3 are 3′-5′ exonucleases, but the latter also has endonucleolytic activity (2). In yeast, Rrp6 is restricted to the nucleus and Dis3 is found in both nuclear and cytoplasmic compartment (29,30).
The conserved helicase Mtr4 is essential for RNA degradation by the exosome, however, the mechanism underpinning Mtr4 function in exosome regulation is not well understood (31). In the fission yeast, Schizosaccharomyces pombe (S. pombe), Mtr4 shares its function with the highly homologous Mtr4-like helicase (Mtl1) (32). Mtr4/Mtl1 interacts with the RNA-binding proteins and the exosome and was proposed to play a role in exosome recruitment to substrate RNAs (1). The exosome-bound Mtr4 facilitates the unwinding of the RNAs substrates and their threading through the channel (33,34). In addition to the exosome core, Mtr4/Mtl1 co-purifies with the RNA binding proteins involved in substrate recognition. In Saccharomyces cerevisiae (S. cerevisiae), Mtr4 is a part of the TRAMP complex (Trf4/5-Air1/2-Mtr4) which consists of two Zinc-finger proteins Air1 and Air2 (only Air1 in fission yeast), poly(A) polymerases Trf4 and Trf5 and Mtr4 (35)(36)(37). The TRAMP complex is recruited to RNA by the RNA and Pol II-binding protein Nrd1 during transcription (15,(38)(39)(40)(41).
However, in contrast to S. cerevisiae, the TRAMP complex seems more specialized in regulation of rRNA processing in fission yeast and human (31). In fission yeast, Mtl1 is proposed to interact with the Zinc-finger protein Red1 to constitute the Mtl1-Red1 complex (MTREC) (32,42).
Additionally, Mtl1 interacts with the conserved YTH-domain containing protein Mmi1 and associated proteins, Red1, Iss10 and Erh1 (12,32,(42)(43)(44). Mmi1 is needed for degradation of a selected group of mRNAs encoding for proteins involved in meiosis, cell cycle regulation, biosynthetic enzymes, RNA biology and ncRNAs by the exosome. Mmi1 is co-transcriptionally recruited to these transcripts by binding to an UUAAAC sequence motifs known as 'determinants of selective removal' (DSRs), and this is proposed to lead to their degradation (20,21,23,45). Mmi1 is also important for mRNA quality control and in particular for degradation of inefficiently spliced mRNAs and proper transcription termination of selected transcripts (16,20,46). In addition to Mmi1, Iss10, Erh1 and Red1, Mtl1 also co-purifies with other factors that have been functionally linked to the exosome regulation: the Zincfinger protein Red5, the poly(A) binding protein Pab2, RRM (RNA-Recognition-Motif) and PWI (Pro-Trp-Ile signature) domain containing protein Rmn1 and the CAP-binding proteins Cbc1, Cbc2 and Ars2 (12,32). Although complexes these proteins form have not been well defined biochemically, Mtl1 has been proposed to be a part of several distinct modules including Mtl1-Red1-Pab2-Red5-Rmn1, Mtl1-Red1, Mtl1-Red1-Mmi1-Iss10-Erh1 and Mtl1-Red1-Cbc1-Cbc2-Ars2. Nevertheless, the mechanisms by which these factors regulate substrate recognition and exosome targeting to substrate RNAs remain obscure. In addition to a group of RNAs regulated by Mmi1, levels of multiple other transcripts have been reported to increase in nuclear exosome mutants suggesting that additional Mmi1-independent mechanisms could contribute to their recognition (12,32). Here, we demonstrate that a subpopulation of these depends on the zing-finger-MYND (myeloid, Nervy, and DEAF-1) domain protein Mub1 for decay.
To gain further insight into function and regulation of the nuclear RNA exosome, we have compared the poly(A)+ transcriptomes and poly(A)+ RNA-bound proteomes from control and three different exosome mutants (mtl1-1, rrp6 and dis3-54). Our analyses support a functional connection between Rrp6 and Mtl1 in controlling levels of multiple mRNAs and ncRNAs as well as H/ACA snoRNA maturation in addition to regulation of unstable ncRNAs (CUTs) previously suggested for nuclear exosome (12). Our data suggest that nuclear exosome plays more prominent role in controlling fission yeast transcriptome than previously anticipated. Interestingly, analyses of the changes in poly(A)+ bound proteome induced by the exosome mutants have identified potential novel factors related to exosome regulation in fission yeast. We focus on the uncharacterised zf-MYND protein Mub1, which is highly enriched on poly(A)+ RNA in the exosome mutant. Mub1 physically interacts with the exosome and its deletion leads to up-regulation of a specific subset of exosome substrates supporting its role in exosome regulation. Furthermore, we have analysed the data to assess whether occupancy of the RNA binding components of RNA processing machineries is affected in the exosome mutants.
We report that the components of the NineTeen Complex (NTC) involved in activation of the spliceosome for catalysis are depleted from RNAs in contract to components of U1 and U2 snRNPs suggesting that the stage prior to the activation of the spliceosome is affected in the exosome mutants and might represents a critical quality control step during splicing. Finally, we also assess RNA-binding behaviour of factors linked to exosome function. We show that RNA recognition can be uncoupled from subsequent steps involved in exosome targeting to RNAs upon inactivation of Mtl1 by mutations in mtl1-1. While engagement with the RNA is lost for the exosome and Mtl1 in this mutant, factors proposed to be involved in substrate recognition (Mmi1, Iss10) as well as some other RNAbinding exosome factors such as Cbc1, Cbc2, Ars2 and Red1 remain on RNA. These data argue for a two-step mechanism leading to the exosome recruitment to its substrates. Altogether, our data provide insights into different aspects of RNA regulation by the exosome as well as furthers our understanding of the mechanism underpinning exosome function.

Poly(A)+ RNA interactome capture in the exosome mutants.
To gain further insights into how the exosome contributes to RNA regulation, we have applied an unbiased quantitative proteomic approach, called RNA interactome capture (RIC), and identified proteins directly interacting with polyadenylated RNA (poly(A)+) in WT cells and in three exosome mutants: rrp6, lacking exonuclease Rrp6; dis3-54, a Dis3 mutant which contains an amino acid substitution (Pro509 to Leu509) located within RNB domain that was reported to show reduced catalytic activity (47, 48) and mtl1-1, a mutant of the helicase Mtl1, which has mutations in the region surrounding the arch domain (32) ( Figure 1A). The underlying hypothesis behind this approach was that inactivation of the exosome would not only stabilise RNAs targeted by the exosome but also affect RNA occupancy of the proteins that are functionally linked to the exosome ( Figure 1B). The three mutant strains were cultured alongside a WT control in the presence of 4-thiouracil (4sU) to facilitate RNA-protein crosslinking with 365nm UV light. Following UV crosslinking, poly(A)+ RNA was enriched by oligo-d(T) selection and RNA-associated proteins were identified by mass spectrometry ( Figure 1B). The abundance of individual proteins in the whole cell extract (WCE) was also determined by proteomics and used to normalise the RIC data (see Material and Methods) as in (49, 50). The RIC/WCE ratio was used to determine the enrichment profile reflected the association of each individual protein with RNA in the mutants relative to the WT (Supplementary Table 1).
Additionally, RNA sequencing was carried out for oligo-d(T) enriched samples to assess levels of polyadenylated RNAs in mutant and WT cells ( Figure 1B).
As proof of principle, we assessed if crosslinking of the RNA-binding protein Mmi1, a known exosome regulatory RNA-binding factor involved in substrate recognition, that was found to be enriched in the WT poly(A)+ RNA interactome (49), has been affected in the exosome mutant. This analysis has demonstrated further 1.5, 2 and 4 times increase of Mmi1 association with poly(A)+ RNA (p-value=0.768, p-value=0.039 and p-value=0.001) in the exosome mutants dis3-54, mtl1-1 and rrp6 RIC respectively, compared to WT cells ( Figure 1C). The increased poly(A)+ RNA association of Mmi1 in exosome mutants correlates with the increased abundance of Mmi1 RNA substrates as revealed by the analysis of oligo-d(T)-enriched RNAs from these strains by deep sequencing (RNA-seq) ( Figure 1D). Indeed, Mmi1 mRNAs targets were most noticeably enriched in the rrp6 and mtl1-1 mutants compared to the dis-54 mutant. These results demonstrate that changed RNA occupancy of individual proteins identified through comparative poly(A)+ RNA interactome analysis can potentially be used for the identification of RNA-bound factors related to the exosome.
To further assess how mutations in different exosome subunits affect levels of poly(A)+ RNA globally, we analysed poly(A)+ RNA-seq data more closely. Consistent with the function of the exosome in degradation of ncRNAs (snRNAs, snoRNAs, antisense RNA), levels of these transcripts were increased in all three exosome mutants (Figure 2A, Supplementary Table 3). Interestingly, a high number of mRNAs is also increased in the exosome mutants more than 1.5-fold suggesting that the nuclear exosome regulates multiple mRNAs in addition to its known role in degradation of ncRNAs Although this corresponds to a smaller fraction of Mtl1 regulated transcripts (25%) this is likely due to milder effect of P509L mutation on Dis3 function in dis3-54 mutant. Our data supports a model where Mtl1 regulates both exosomes associated nucleases, with Rrp6 being perhaps slightly more dependent on Mtl1 compared to Dis3.
We next performed comparative analysis of proteins differentially enriched in all three interactomes.
For each RIC triplicate, proteins that are detected at least in two out of three biological repeats of this experiment are used in the analysis leading to a total number of 1146 proteins considered (see Material and Methods). From this analysis, 206, 247 and 138 were enriched more than 2-fold in mtl1-1, rrp6 and dis3-54 compared to WT respectively (Supplementary Table 2). On the other hand, we detected 196, 147 and 137 proteins whose abundance on poly(A)+ RNA was reduced in mtl1-1, rrp6 and dis3-54 (Supplementary Table 2). Consistently with the RNA-seq data, higher overlap was observed between proteins whose RNA binding is affected in rrp6 and mtl1-1 than between dis3-54 and mtl1-1. These results further support a functional link between Mtl1 and Rrp6 ( Figure 2C-D).
Next, we performed gene ontology (GO) analyses across the three different mutants to assess whether proteins with altered RNA-binding in the exosome mutants are part of a specific biological process. The analysis has revealed that nuclear proteins are most noticeably affected in all three  Figure  S1D). This might indicate that Dis3 plays less dominant role in regulation of RNA in the nucleus compared to Mtl1 and Rrp6. However, milder nuclear retention of poly(A)+ RNA observed in dis3-54 compared to other exosome mutants might be due to only partial inactivation of Dis3 in this mutant as discussed above. Interestingly, even though only mild accumulation of poly(A)+ RNA is observed, increased RNA-binding of nuclear proteins is more pronounced compared to cytoplasmic proteins in dis3-54 ( Figure 2E). This might reflect that exosome has more prominent impact on nuclear rather than cytoplasmic RNPs. Additionally, RNA-binding activity of proteins related to mRNA metabolic process (GO:0016071), ribosome biogenesis (GO:0042254) and cytoplasmic translation (GO:0002181) is altered in all mutants (Supplementary Table 4).
Altogether, this analysis reveals that proteins affected in each specific mutant tend to be linked to similar biological processes. Additionally, these data demonstrate significant changes in RNA-protein interactions in exosome deficient cells, highlighting the important role of the exosome in RNA metabolism.
Comparative RIC reveals novel proteins that are linked to the exosome.
We hypothesised that comparative RIC approach can be employed to identify novel RNA-bound exosome factors by assessing protein enrichment on poly(A)+ RNA in the exosome mutants. We first selected proteins that were more than 2-fold enriched on poly(A)+ RNA in the interactome of at least one mutant. Second, we focussed on proteins containing a classical RNA-binding domains (54) and annotated as uncharacterised. This resulted in a list of 10 potentials candidates ( Figure 3A).
Next, the candidate genes were deleted to test whether they recapitulate phenotypes associated with compromised exosome function. We have assessed steady-state levels of RNAs that are upregulated in the exosome mutants but are independent of known exosome factor Mmi1 (such as the RNA produced from the Tf2 retro-transposable element (SPAC9.04)). Increase in tf2-1 RNA levels is observed in SPBC31F10.10c and SPBC16G5.16 compared to WT ( Figure 3B, compare lanes 7 and 11 to lane 1). SPBC31F10.10c encodes for Mub1 protein, which contains an Armadillo-type domain, and a potential nucleic acid binding region represented by a zf-MYND domain. Mub1 shows a >6-fold increase in mtl1-1 RIC (p-value= 5,41E-08) ( Figure 3A) suggesting that it might be linked to Mtl1 function.
Next, we have assessed whether nuclear retention of poly(A)+ RNA, a typical phenotype of the exosome mutants, can be recapitulated in the deletion strains using poly(A)+ FISH ( Figure 3C).
This results in accumulation of heterogeneous RNAs produced from telomeric and centromeric regions of the S. pombe genome, although the underlying mechanism remains unclear. To test whether the candidate proteins might contribute to expression of heterochromatic transcripts, each individual deletion mutant was crossed with the strain bearing a ura4+ reporter gene inserted within transcriptionally silent telomeric region of chromosome I. We then monitored the capacity of these reporter strains to grow on -URA plates or plates containing 5-Fluoroorotic acid (5-FOA) at 25°C and 30°C. Interestingly, the deletion of either SPAC126.11c, srp40 or SPAC222.18 led to moderate growth on -URA plates and attenuated growth on 5-FOA compared to WT ( Figure 3D and Supplementary Figure S2B), suggesting that each of these candidates might play a role in heterochromatin formation/maintenance. It is interesting to note that SPAC126.11c and srp40 deletion cells also showed nuclear retention of poly(A)+ RNA ( Figure 3C). Both SPAC126.11c and Srp40 were also previously reported to co-purify with the exosome further supporting a direct link between these proteins and the exosome (42,56).
Taken together, this preliminary screen has identified several candidate proteins, including SPBC31F10.10c, SPAC126.11c, Srp40 or SPAC222.18, whose deletions phenocopy exosome mutants suggesting a potential relationship between these factors and the nuclear RNA exosome.
However, further studies are needed to test whether this is indeed the case.

Mub1 regulates exosome degradation of the stress-induced mRNAs.
To further understand how Mub1 contributes to RNA regulation we have investigated the functional consequences of Mub1 deletion. Interestingly, mub1 deleted cells showed a rounded morphology, suggesting an alteration in processes related to cell morphology (Supplementary Figure S3A Tables 5-6). Interestingly, many transcripts that are affected by Mub1 deletion were also increased in mtl1-1 ( Figure 4A-B) suggesting that Mtl1 and Mub1 act in the same pathway. In agreement with RNA-seq data, increased steady-state levels of gst2 and SPCC663.08c mRNAs in mub1 and mtl1-1 mutants is also detected by northern blot analyses ( Figure 4B). No additive effect was observed in the double mutant mtl1-1 mub1 compared to single mutants, which is consistent with Mtl1 and Mub1 acting together ( Figure 4B). To assess whether Mub1 and the exosome directly interact, we carried out co-immunoprecipitation experiments.
These experiments have revealed that Mub1 co-purifies with the Rrp6 subunit of the exosome supporting a physical link between Mub1 and the nuclear exosome ( Figure 4C).
Presence of Mub1 in the poly(A)+ RNA interactome suggests that Mub1 is an RNA-binding protein. Mub1 contains a predicted zf-MYND domain (471-528), which is likely to mediate its interaction with RNA. We therefore hypothesised that this domain may be critical for Mub1 function.
To test this, we deleted the zf-MYND domain of Mub1 ( Figure 4D). Deletion of the zf-MYND domain led to expected change in size of the protein, which was assessed by visualising FLAG tagged protein by Western blotting ( Figure 4E). Addition of triple FLAG tag to Mub1 didn't have any effect on cell morphology, cell growth and cellular protein levels (Supplementary Figure S3A-C).The protein level of Mub1471-528-3xFLAG (Mub1-Z-3xFLAG) was comparable to Mub1-3xFLAG, suggesting that the deletion did not affect the stability of the truncated protein ( Figure 4E). Cells expressing the truncated Mub1471-528-3xFLAG displayed the characteristic rounded shape, alike mub1 cells suggesting that function of Mub1 is compromised in this mutant (data not shown). To test the importance of the zinc-finger domain for Mub1 function, we assessed levels of gst2 and SPCC663.08c mRNAs in mub1471-528-3xFLAG mutants by northern blot. This experiment revealed that similar to mub1 levels of both mRNAs were increased in the mub1471-528-3xFLAG mutant ( Figure 4F). Taken together, these results suggest that Zinc-finger domain of Mub1 is essential for Mub1 function.

Mub1 plays a role in the heat shock response.
GO analysis of mRNAs increased in mub1 cells have shown a significant enrichment for GO "Core Environmental Stress Response induced" (46.25% (74/160), p-value = 1.55246e-28, AnGeLi tool (57)) implying that Mub1 may be required for cellular response to stress. Consistently, mub1 growth at 37°C is impaired ( Figure 5A). To further assess a role of Mub1 in the regulation of genes related to heat shock, we analysed expression of hsp16, encoding a heat shock protein implicated in cellular response to heat (58). Indeed, hsp16 levels are increased in mub1 and mtl1-1 ( Figure 5B, compare lanes 2-4 to 1). Interestingly, hsp16 mRNA levels are also altered during heat shock (4 hours at 37°C) in the mub1 and mtl1-1 (compare lanes 6-7 to lane 5), suggesting that Mub1 is required for exosome dependent expression of the heat shock genes. Interestingly, similar to complete Mub1 ablation, deletion of the zf-MYND domain also impaired growth at 37°C ( Figure 5C) suggesting that this domain is required to ensure proper expression of stress response genes at 37°C.

Processing of H/ACA box snoRNA depends on Rrp6 and Mtl1.
Next, we wanted to assess how RNA binding of the known RNA-binding proteins implicated in the RNA metabolism is affected in the exosome mutants. One of the function of the nuclear RNA exosome is 3'end trimming of the precursors of the stable non-coding RNAs such as snRNAs and snoRNAs (59). Precursor molecules are polyadenylated in contrast to mature sn/snoRNAs lacking poly(A) tails (4). Indeed, increased levels of snoRNAs especially noticeable for H/ACA box snoRNAs are observed in mtl1-1 and rrp6 compared to WT ( Figure 6A). Analysis of an individual gene, snR92, shows that accumulation of the polyadenylated snR92 precursor coincides with the loss of the mature form in mtl1-1 but not in rrp6 and dis3-54 mutants ( Figure 6B  suggesting that Mtl1 plays a less prominent role in C/D box snoRNAs processing. Interestingly, proteins that are known to bind H/ACA box snoRNAs, such as Nhp2, Nop10 and Cbf5, are strongly enriched in the poly(A)+ interactome in mtl1-1 and rrp6, suggesting that these proteins are associated with pre-snoRNAs ( Figure 6D-E). This is consistent with the reported co-transcriptional assembly of these proteins on H/ACA box snoRNAs and their role in snoRNA biogenesis (60). The same tendency is observed for C/D box RNA-binding proteins (Nop58, Nop56, Fib1, Pop7 and Bcd1), although the effect is less prominent compared to components of H/ACA box snoRNP ( Figure 6E).
These observations are in agreement with C/D box RNAs being less affected in the exosome mutants compared to H/ACA box snoRNAs and demonstrates that behaviour of protein components of RNPs reflects changes in the levels of the corresponding RNA. To assess whether compromised 3' end processing influences remodelling of H/ACA box pre-snoRNPs, we analysed RNA association of factors that are recruited at later stages of snoRNP assembly such as Gar1 protein which replaces the assembly factor Naf1. Recruitment of Gar1 to snoRNP has been proposed to complete the formation of functional H/ACA box snoRNPs (61,62). In agreement with the observed accumulation of pre-snoRNAs, Naf1 (SPBC30D10.15 in S. pombe) is strongly enriched on poly(A)+ RNA in the mtl1-1 and rrp6 interactomes. Surprisingly, we also noticed an increased association of Gar1 with poly(A)+ RNA in mtl1-1, where very little mature snoRNAs can be detected ( Figure 6E), suggesting that Gar1 is recruited to pre-snoRNA and its recruitment is not sufficient to dissociate Naf1. In contrast, the accumulation of poly(A)+ pre-snoRNAs did not correlate with increased association of Gar1 with poly(A)+ RNA in the rrp6 interactome. This data suggest that Mtl1 may contribute to suppress the premature exchange of Naf1 and Gar1 on unprocessed transcripts. Altogether, these data highlight the important role of the nuclear exosome in H/ACA snoRNAs processing and further support functional connection between Mtl1 and Rrp6 to process this specific class of ncRNA.

mutants.
Previous studies have demonstrated that splicing of selected pre-mRNAs is compromised in the exosome mutants, although the direct involvement of the exosome is still under debate (18,20). To study spliceosome recruitment to RNA in the exosome mutants, we next analysed RNA association of the splicing factors. One of the mRNAs whose splicing has been found to be compromised in dis3-54, mtl1-1 and rrp6 is rps2202 mRNA that encodes for the 40S ribosomal protein S15a ( Figure 7A-B).
Spliceosome assembly is initiated upon the recognition of the intronic features such as the 5' splice site (5'ss) and the branch point (BP) by U1 and U2 snRNPs through base pairing. Recruitment of U1 and U2 to pre-mRNA is followed by association of tri-snRNP U4-U6-U5 to form the pre-catalytic B complex. One of the key steps during spliceosome assembly is the recruitment of the NineTeen Complex (NTC, or Prp19 complex, (GO:0000974)) accompanied by the dissociation of U4 and U1, which leads to the activation of the B complex for catalysis of the first trans-esterification reaction ( Figure 7C) (63). Interestingly, while components of U1 and U2 snRNPs (specific proteins for U1 snRNP (GO:0005685) and U2 snRNP (GO:0005686)) are enriched, components of NTC are depleted in the mtl1-1 interactome ( Figure 7C). This data suggests that the activation of the spliceosome for catalysis is likely to be compromised upon inactivation of the exosome, potentially defining a time window for quality control.
Accumulation of longer poly(A) tails for individual RNA substrates upon mutation or depletion of the exosome subunits was reported in fission yeast and human cells (20,64,65). Transcripts produced by Pol II are polyadenylated by the Cleavage and Polyadenylation Factor (CPF). CPF consists of three functionally distinct activity-based modules: poly(A) polymerase, nuclease and phosphatase modules and cleavage factors CFIA and CFIB ( Figure 7D) (66,67). In addition, selected transcripts can also be polyadenylated by the non-canonical poly(A) polymerases Trf4 and Trf5 that are part of the TRAMP complex (68). However, RNA hyperadenylation observed in rrp6 is maintained in TRAMP mutants but lost when poly(A) polymerase of the CPF is compromised (12,65).
This suggest that CPF is responsible for addition of longer poly(A) tails rather than TRAMP. can be pulled down with the exosome and also with proteins involved in RNA recognition, such as Mmi1 (12,32). This suggests that Mtl1 could be involved in either RNA recognition or exosome targeting to RNA or both of these steps that we predict to precede degradation of the RNA by the exosome. To gain further insight into Mtl1 function, we next analysed association of factors functionally linked to the exosome with poly(A)+ RNA in the mtl1-1 interactome ( Figure 8A) (24,38,50). Strikingly, while the interaction of Mtl1 and the exosome subunits with poly(A)+ RNA is decreased in mtl1-1, association of factors involved in RNA recognition such as Mmi1, the Mmi1-associated proteins Iss10, Red1 and the CAP-binding complex Cbc1/2-Ars2 with RNA is either increased or unaffected in the absence of functional Mtl1 ( Figure 8B). In addition, Mmi1, Cbc1/2-Ars2 and Red1 are also enriched on RNA in rrp6 and dis3-54 mutants. These data suggest that exosome targeting and engagement with RNA but not RNA recognition by the RNA-binding proteins such as Mmi1 is likely to be affected in the mtl1-1 mutant. We therefore propose that Mtl1 plays a role in coupling RNA recognition to exosome recruitment ( Figure 8C). Finally, RNA association of the exosome factors proposed to interact with poly(A) tail such as Pab2 and Red5 is decreased in exosome mutants ( Figure 8D) suggesting that exosome either directly or indirectly contributes to assembly of these factors with RNA. genes in budding yeast (71). Furthermore, exosome was shown to be implicated in heat sock response in flies suggesting that function of the exosome in environmental adaptation is conserved (72). In addition to heat shock, exosome has been shown to play a role in cellular responses to nutrients, cell differentiation or DNA damage response in budding yeast and human (16,(73)(74)(75)(76). and ZFC3H1 supporting an evolutionary conserved role for these proteins in exosome regulation (80)(81)(82) 82,84,85). In addition to the Zinc-finger domain, ZCCHC8 also contains a hydrophobic region called "arch interacting motif" (AIM) that mediates interaction with MTR4 arch domain and the C-terminal region that stimulates helicase activity of MTR4 (86,87). AIM region is also present in other exosome co-factors-ZCCHC7 (Air2), rRNA processing factors NOP53 and NVL (88,89). It is possible that similar to other Zinc-finger proteins involved in exosome regulation, Mub1 could help to ensure the connection between the exosome and its target RNAs by mediating contacts with RNA and other proteins involved in exosome regulation such as Mtl1. Mub1 also has another domain in addition to MYND-type Zinc-finger domain-Armadillo-like helical domain. Future studies will clarify the specific role of Mub1 and its domains in the exosome regulation. Based on our findings, we propose a model in which Mub1 acts as an exosome co-factor involved either in substrate recognition and/or exosome targeting to a subset of RNAs that are induced in response to stress.

DISCUSSION
One of the key factors that is essential for the function of the nuclear exosome is the RNA helicase Mtr4/Mtl1. Two models of how Mtr4 assist RNA degradation by the exosome have been proposed based on the recent structural studies of reconstituted human and S. cerevisiae exosomes: 1) MTR4 recruitment to the core of human exosome mediates substrate channelling to DIS3 and is mutually exclusive with substrate channelling to RRP6 and RRP6 association with the top of the exosome core (33), 2) Mtr4 interacts with Rrp6, as has been demonstrated for the S. cerevisiae exosome (34). Our data suggest that in fission yeast, Mtl1 is required for both Rrp6 and Dis3 function in vivo. In agreement with our findings, it has recently been shown that the expression of Rrp6-GFP-Mmi1 chimeric protein is sufficient to supress the dysregulated expression of meiotic genes in a Mtl1 mutant (mtl1-cs5) and red1 mutants (90). Furthermore, we demonstrate that Rrp6 and Mtl1 mutants accumulate unprocessed precursors of H/ACA box snoRNAs suggesting that both proteins cooperate to mediate 3'-end maturation of these transcripts. Notably, accumulation of pre-snoRNAs correlates with enrichment of the components of pre-snoRNPs in poly(A)+ pull-downs observed in these mutants. Indeed, Nop10, Nhp2, Cbf5 and the snoRNP assembly factor S. pombe Naf1 (SPBC30D10.15) were strongly enriched in both interactomes compared to WT. It has been suggested by previous studies that the protein Gar1 replaces Naf1 during the maturation process of H/ACA snoRNA (60). Interestingly, while Naf1 is enriched in both the mtl1-1 and rrp6, Gar1 is only enriched in mtl1-1 interactome, suggesting that Mtl1 might mediate factor exchange during snoRNA processing.
Exosome mutants have been associated with phenotypes spanning diverse defects in RNA biogenesis including pre-mRNA splicing, 3'end processing and mRNA export from the nucleus, some of which may not be directly related to exosome function (16,20,69,70). Interestingly, we observe altered association of splicing factors with RNA which might explain splicing defects previously reported in exosome mutants for selected transcripts (20). We demonstrate that the factors required for activation of the spliceosome for catalyses (NTC components) are depleted from RNAs. At the same time, association of U1-U2 components is not affected in the exosome deficient mutants. This suggests that activation of the spliceosome for catalysis might be a rate limiting step. We propose that NTC recruitment could provide a regulatory step that contributes to the regulation of the rate of splicing and might link splicing rate and decay. Interestingly, Mtl1 was shown to interact with spliceosome components, some of which are part of the NTC complex (32,42,91). However, at this stage, we also cannot entirely rule out a possibility for a direct recruitment of NTC by the exosome.
Another reason for failed RNA processing could be titration of the RNA-binding proteins by various RNAs that accumulate in exosome mutants. In agreement with this idea, we also show, that components of the CPF machinery, responsible for addition of the poly(A) tail to the 3'end of Pol II transcribed transcripts, are enriched on poly(A)+ RNA in the exosome mutants. This suggest that CPF recycling might be compromised in the exosome mutants and could explain why RNA hyperadenylation and 3'end extension is observed in the exosome mutants. In agreement with this idea, accumulation of nuclear RNAs upon exosome depletion was proposed to be a reason behind transcriptional de-repression of the Polycomb targets in mouse ESC (92).
Although a number of factors have been linked to the exosome, their role in exosome regulation is not well understood. We demonstrate that exosome factors proposed to be involved in RNA recognition such as Mmi1 and its associated protein Iss10, as well as components of the CAP binding complex (Cbc1/2 and Ars2), remain bound to RNA in the Mtl1 mutant, whereas association of the exosome complex and Mtl1 itself is dramatically compromised. This data suggests that although Mtl1 was found to co-IP with Mmi1, it is not involved in RNA recognition and rather it might be playing a role in exosome targeting and engagement with RNA. Mtl1 was proposed to form the MTREC complex together with Red1. Interestingly, association of Red1 with poly(A)+ RNAs is increased in the Mtl1 mutant, while Mtl1 itself is not recruited to RNA. This suggests that Red1 is recruited to RNA independently of Mtl1 and reinforces the model of a two-steps recruitment of the nuclear exosome onto its RNAs substrates ( Figure 8C).
It is interesting to note that a recent study has assessed changes in the poly(A)+ RNA interactome caused by RNAi depletion of the RRP40 subunit of the exosome in human cells (85). In agreement with our observations, this study reports that loss of the exosome leads to an increased RNA association of the selected exosome regulatory factors such as MTR4, ZFC3H1 (Red1) and CBC-ARS complex. Interestingly, this study also identifies new exosome co-factors in human cells (ZC3H3, RBM26 and RBM27) that are involved in linking the exosome to the PAXT complex.
Collectively, our data suggest that an RNA interactome capture approach can provide valuable insights into how the exosome might contribute to RNA regulation. Based on our data, we propose a two-step mechanism responsible for the recognition and targeting of the RNA substrates of nuclear exosome for degradation. First, RNA substrates are recognised by the RNA-binding specificity factors and second, the exosome is recruited to RNA in Mtl1-dependent manner. We also propose a key role for the NTC in monitoring the rate of mRNA splicing. Finally, we identify Mub1 as an important protein that regulates a specific class of exosome dependent mRNAs implicated in the stress response fully supporting a key role of the nuclear exosome in cell adaptability and response to environment.

Yeast strains and manipulations
General fission yeast protocols and media are described in (93). All strains are listed in Supplementary Table 7. Experiments were carried out using YES medium at 30°C unless stated otherwise. Gene deletions and epitope tagging were performed by homologous recombination using polymerase chain reaction (PCR) products (94). All oligos are listed in Supplementary Table 8.
Protein extracts and western blotting were done as described in (95), and protein coimmunoprecipitation as described in (96).

Northern blotting
Northern blot experiments were essentially performed as described in (39). RNA was prepared as described in (20). 8 μg of RNA was resolved on a 1.2% agarose gel containing 6.7% formaldehyde in MOPS buffer. After capillary transfer in 10× SSC onto a Hybond N+ membrane (GE Healthcare), RNA was UV-crosslinked and stained with methylene blue to visualise ribosomal RNA bands. For snoRNAs analysis, 16µg of RNA was resolved in 8% Urea-PAGE. Electro-transfer was performed in TBE overnight onto a Hybond N+ membrane (GE Healthcare, RPN303B). Gene-specific probes were generated by random priming in the presence of ATP [α32P] using the Prime-It II Random Primer Labeling Kit (Agilent, 300385) using PCR generated DNA template produced from gDNA isolated from a wild type S.pombe strain (YP71) using oligonucleotides listed in Supplementary Table 8.

Statistical data analysis
Statistical analysis was performed essentially as described in (40)

Poly(A)+ RNA Fluorescence In Situ Hybridization (FISH)
Poly(A)+ RNA FISH was done as described in (100, 101 Before hybridization, 50 ng of the oligo-d(T) probe was mixed with 2μl of a 1:1 mixture between yeast transfer RNA (10 mg/ml, Life Technologies, AM7119) and salmon-sperm DNA (10 mg/ml, Life Technologies, 15632-011) and the mixture was dried in a vacuum concentrator. Hybridization buffer F (20% formamide, 10 mM NaHPO4 at pH 7.0; 50 μl per hybridization) was added, and the probe/buffer F solution was incubated for 3 min at 95 °C. Buffer H (4× SSC, 4 mg/ml BSA (acetylated) and 20 mM vanadyl ribonuclease complex; 50 μl per hybridization) was added in a 1:1 ratio to the probe/buffer F solution. Cells were resuspended in the mixture and incubated overnight at 37 °C. After three washing steps (10% formamide/2× SSC; 0.1% Triton-X100/2x SSC and 1x PBS), cells were resuspended in 1× PBS/DAPI and mounted into glass slides for imaging. Z-planes spaced by 0.2 μm were acquired on a Ultraview spinning-disc confocal. Acquisition was done with DAPI filter (405nm) and a FITC filter (488nm for alexa488 acquisition). Images were analysed using ImageJ software (102).

Bulky poly(A) tail length measurement
RNA was prepared as for Northern blotting experiments. Bulk RNAs were 3' end labelled essentially as described in (103). Briefly, 1µg of total RNA was incubated 18h at room temperature with 2µCi of

RNA sequencing
For spike-in normalisation, the S. cerevisiae cells were added to S. pombe at 1:10 ratio prior to RNA isolation. Total RNA was extracted from cultures in mid-log phase using a standard hot phenol method and treated with RNase-free DNase RQ1 (Promega, M6101) to remove any DNA contamination. For total RNA sequencing, experiments were done in duplicates, and ribodepletion was performed using ribominus transcriptome isolation kit (Invitrogen, K155003). Poly(A)+ RNA sequencing was performed by using 1/33 of the oligo-d(T) pull-down total volume, subjected to proteinase K treatment for 1h at 50°C. Poly(A)+ RNA was recovered by a standard hot phenol method. Experiments were done in triplicate. cDNA libraries were prepared using NEBNext® Ultra™ II Directional RNA Library Prep Kit for Illumina (NEB#E7760S) for 50ng of total RNA and using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB #E7420) for 100ng of WT1, mtl1-1, rrp6 and dis3-54 purified oligo-d(T) RNA. Paired-end sequencing was carried out on the Illumina HiSeq 500 platform. RNA sequencing data are available with the GEO number GSE148799 and GSE149187.

RNA-seq data analyses
Quality trimming of sequenced reads was performed using Trimmomatic (Galaxy Version 0.32.3, RRID:SCR_011848). Reads were aligned to the concatenated S. pombe (ASM294v2.19) using Bowtie 2 (TopHat) (104).For spike-in normalisation, reads derived from different S. pombe and S. cerevisiae chromosomes were separated. Reads mapped only once were obtained by SAMTools (105) and reads were mapped to the genome using genome annotation from (106). Differential expression analyses were performed using DESeq2 (107) in R and using the spike-in normalisation.
For poly(A)+ RNA sequencing total read count normalisation using DEseq2 (107) in R was used. The significance of RNAs list overlaps was calculated using a standard Fisher's exact test, For gene ontology analysis, up or downregulated protein coding gene lists were submitted to AnGeLi (http://bahlerweb.cs.ucl.ac.uk/cgi-bin/GLA/GLA_input), a web based tool (57).

Raw (fastq) and processed sequencing data (bedgraph) can be downloaded from the NCBI Gene
Expression with the GEO number GSE148799 and GSE149187. Mass spectrometry data are available via ProteomeXchange with identifier PXD016741.

ACKNOWLEDGEMENT
We thank the National Bio Resource Yeast Project, S. Grewal, K. Nasmyth, P. Bernard and JP.